Drug Therapy

A 71-year-old woman came to the hospital because of generalized weakness, fatigue, and exertional dyspnea.

She had a history of anemia, recurrent urinary tract infections, and hyperactive bladder. She had been taking nitrofurantoin for a urinary tract infection and phenazopyridine for dysuria, and she noticed that her urine was dark-colored.

She was of northern European descent. She was unaware of any family history of blood-related disorders. She had been admitted to the hospital 6 weeks earlier for symptomatic anemia after taking nitrofurantoin for a urinary tract infection. At that time, she received 2 units of packed red blood cells and then was discharged. Follow-up blood work done 2 weeks later—including a glucose-6 phosphate dehydrogenase (G6PD) assay—was normal.

On physical examination, she was pale and weak. Her hemoglobin level was 5.5 g/dL (reference range 14.0–17.5), with normal white blood cell and platelet counts and an elevated reticulocyte count. A comprehensive metabolic panel showed elevated indirect bilirubin and lactate dehydrogenase levels. A direct Coombs test for autoimmune hemolytic anemia was negative, as was a haptoglobin assay to look for intravascular hemolytic anemia. G6PD levels were normal, yet a peripheral blood smear (Figure 1) showed features of G6PD deficiency.

What was the cause of her anemia?

GLUCOSE-6-PHOSPHATE DEHYDROGENASE DEFICIENCY

G6PD deficiency is an X-linked disorder¹ that can present as hemolytic anemia. Symptoms of hemolysis can range from mild to severe on exposure to an inciting agent. Men are more commonly affected than women, and affected women are mostly heterozygous. The severity of hemolysis in heterozygous women depends on inactivation of the unaffected X chromosome in some cells.

When exposed to oxidizing agents, people with G6PD deficiency do not have enough nicotinamide adenine dinucleotide phosphate to protect red blood cells.² This leads to oxidative denaturation of hemoglobin, formation of methemoglobin, and denaturation of globulin. These products are insoluble; they collect in red blood cells and are called Heinz bodies.³ When red blood cells containing Heinz bodies pass through the liver and spleen, the insoluble masses are taken up by macrophages, causing hemolysis and the formation of “bite cells”⁴ (so named because macrophages “bite” the Heinz bodies out of the red blood cells).

Patients with G6PD deficiency have all the clinical features of hemolytic anemia. On laboratory testing, the Coombs test is negative, the G6PD level is low, and the peripheral smear shows bite cells. The G6PD level is falsely normal or elevated during acute hemolysis because red blood cells deficient in G6PD are removed from circulation and replaced by young red blood cells. The G6PD level is also elevated after blood transfusion. Thus, the G6PD level should be tested 3 months after an acute event.

Hemolysis in G6PD is usually intermittent and self-limited. No treatment is needed except for avoidance of triggers and transfusion for symptomatic anemia. Of note, triggers include some of the drugs commonly used for urinary tract infections (sulfa drugs, nitrofurantoin, phenazopyridine) and antimalarials. Fava beans are also known to cause hemolytic crisis. A complete list of things to avoid can be found at www.g6pd.org/en/G6PDDeficiency/SafeUnsafe/DaEvitare_ISS-it.

There is no commercially available genetic testing kit for G6PD deficiency. Mutation analysis and G6PD gene sequencing are possible but are neither routinely done nor widely available.

BACK TO OUR PATIENT

Our patient’s hemolytic anemia was most likely drug-induced, secondary to a relative deficiency of G6PD. She had been taking nitrofurantoin and phenazopyridine; both of these are oxidizing agents and are known to cause acute hemolytic anemia in people with G6PD deficiency. The G6PD level can be normal after a recent blood transfusion and, as in our patient, during an acute episode of hemolysis.

Because of the strong suspicion of G6PD deficiency, both drugs were stopped when the patient was discharged from the hospital. She did not take either drug for 3 months. Her G6PD level was then retested and was found to be low, confirming the diagnosis. The patient was then advised not to take those drugs again. Since then, her hemoglobin level has remained stable and she has not needed any more blood transfusions.

A 71-year-old woman came to the hospital because of generalized weakness, fatigue, and exertional dyspnea.

She had a history of anemia, recurrent urinary tract infections, and hyperactive bladder. She had been taking nitrofurantoin for a urinary tract infection and phenazopyridine for dysuria, and she noticed that her urine was dark-colored.

She was of northern European descent. She was unaware of any family history of blood-related disorders. She had been admitted to the hospital 6 weeks earlier for symptomatic anemia after taking nitrofurantoin for a urinary tract infection. At that time, she received 2 units of packed red blood cells and then was discharged. Follow-up blood work done 2 weeks later—including a glucose-6 phosphate dehydrogenase (G6PD) assay—was normal.

On physical examination, she was pale and weak. Her hemoglobin level was 5.5 g/dL (reference range 14.0–17.5), with normal white blood cell and platelet counts and an elevated reticulocyte count. A comprehensive metabolic panel showed elevated indirect bilirubin and lactate dehydrogenase levels. A direct Coombs test for autoimmune hemolytic anemia was negative, as was a haptoglobin assay to look for intravascular hemolytic anemia. G6PD levels were normal, yet a peripheral blood smear (Figure 1) showed features of G6PD deficiency.

What was the cause of her anemia?

GLUCOSE-6-PHOSPHATE DEHYDROGENASE DEFICIENCY

G6PD deficiency is an X-linked disorder¹ that can present as hemolytic anemia. Symptoms of hemolysis can range from mild to severe on exposure to an inciting agent. Men are more commonly affected than women, and affected women are mostly heterozygous. The severity of hemolysis in heterozygous women depends on inactivation of the unaffected X chromosome in some cells.

When exposed to oxidizing agents, people with G6PD deficiency do not have enough nicotinamide adenine dinucleotide phosphate to protect red blood cells.² This leads to oxidative denaturation of hemoglobin, formation of methemoglobin, and denaturation of globulin. These products are insoluble; they collect in red blood cells and are called Heinz bodies.³ When red blood cells containing Heinz bodies pass through the liver and spleen, the insoluble masses are taken up by macrophages, causing hemolysis and the formation of “bite cells”⁴ (so named because macrophages “bite” the Heinz bodies out of the red blood cells).

Patients with G6PD deficiency have all the clinical features of hemolytic anemia. On laboratory testing, the Coombs test is negative, the G6PD level is low, and the peripheral smear shows bite cells. The G6PD level is falsely normal or elevated during acute hemolysis because red blood cells deficient in G6PD are removed from circulation and replaced by young red blood cells. The G6PD level is also elevated after blood transfusion. Thus, the G6PD level should be tested 3 months after an acute event.

Hemolysis in G6PD is usually intermittent and self-limited. No treatment is needed except for avoidance of triggers and transfusion for symptomatic anemia. Of note, triggers include some of the drugs commonly used for urinary tract infections (sulfa drugs, nitrofurantoin, phenazopyridine) and antimalarials. Fava beans are also known to cause hemolytic crisis. A complete list of things to avoid can be found at www.g6pd.org/en/G6PDDeficiency/SafeUnsafe/DaEvitare_ISS-it.

There is no commercially available genetic testing kit for G6PD deficiency. Mutation analysis and G6PD gene sequencing are possible but are neither routinely done nor widely available.

BACK TO OUR PATIENT

Our patient’s hemolytic anemia was most likely drug-induced, secondary to a relative deficiency of G6PD. She had been taking nitrofurantoin and phenazopyridine; both of these are oxidizing agents and are known to cause acute hemolytic anemia in people with G6PD deficiency. The G6PD level can be normal after a recent blood transfusion and, as in our patient, during an acute episode of hemolysis.

Because of the strong suspicion of G6PD deficiency, both drugs were stopped when the patient was discharged from the hospital. She did not take either drug for 3 months. Her G6PD level was then retested and was found to be low, confirming the diagnosis. The patient was then advised not to take those drugs again. Since then, her hemoglobin level has remained stable and she has not needed any more blood transfusions.

A 71-year-old woman came to the hospital because of generalized weakness, fatigue, and exertional dyspnea.

She had a history of anemia, recurrent urinary tract infections, and hyperactive bladder. She had been taking nitrofurantoin for a urinary tract infection and phenazopyridine for dysuria, and she noticed that her urine was dark-colored.

She was of northern European descent. She was unaware of any family history of blood-related disorders. She had been admitted to the hospital 6 weeks earlier for symptomatic anemia after taking nitrofurantoin for a urinary tract infection. At that time, she received 2 units of packed red blood cells and then was discharged. Follow-up blood work done 2 weeks later—including a glucose-6 phosphate dehydrogenase (G6PD) assay—was normal.

On physical examination, she was pale and weak. Her hemoglobin level was 5.5 g/dL (reference range 14.0–17.5), with normal white blood cell and platelet counts and an elevated reticulocyte count. A comprehensive metabolic panel showed elevated indirect bilirubin and lactate dehydrogenase levels. A direct Coombs test for autoimmune hemolytic anemia was negative, as was a haptoglobin assay to look for intravascular hemolytic anemia. G6PD levels were normal, yet a peripheral blood smear (Figure 1) showed features of G6PD deficiency.

What was the cause of her anemia?

GLUCOSE-6-PHOSPHATE DEHYDROGENASE DEFICIENCY

G6PD deficiency is an X-linked disorder¹ that can present as hemolytic anemia. Symptoms of hemolysis can range from mild to severe on exposure to an inciting agent. Men are more commonly affected than women, and affected women are mostly heterozygous. The severity of hemolysis in heterozygous women depends on inactivation of the unaffected X chromosome in some cells.

When exposed to oxidizing agents, people with G6PD deficiency do not have enough nicotinamide adenine dinucleotide phosphate to protect red blood cells.² This leads to oxidative denaturation of hemoglobin, formation of methemoglobin, and denaturation of globulin. These products are insoluble; they collect in red blood cells and are called Heinz bodies.³ When red blood cells containing Heinz bodies pass through the liver and spleen, the insoluble masses are taken up by macrophages, causing hemolysis and the formation of “bite cells”⁴ (so named because macrophages “bite” the Heinz bodies out of the red blood cells).

Patients with G6PD deficiency have all the clinical features of hemolytic anemia. On laboratory testing, the Coombs test is negative, the G6PD level is low, and the peripheral smear shows bite cells. The G6PD level is falsely normal or elevated during acute hemolysis because red blood cells deficient in G6PD are removed from circulation and replaced by young red blood cells. The G6PD level is also elevated after blood transfusion. Thus, the G6PD level should be tested 3 months after an acute event.

Hemolysis in G6PD is usually intermittent and self-limited. No treatment is needed except for avoidance of triggers and transfusion for symptomatic anemia. Of note, triggers include some of the drugs commonly used for urinary tract infections (sulfa drugs, nitrofurantoin, phenazopyridine) and antimalarials. Fava beans are also known to cause hemolytic crisis. A complete list of things to avoid can be found at www.g6pd.org/en/G6PDDeficiency/SafeUnsafe/DaEvitare_ISS-it.

There is no commercially available genetic testing kit for G6PD deficiency. Mutation analysis and G6PD gene sequencing are possible but are neither routinely done nor widely available.

BACK TO OUR PATIENT

Our patient’s hemolytic anemia was most likely drug-induced, secondary to a relative deficiency of G6PD. She had been taking nitrofurantoin and phenazopyridine; both of these are oxidizing agents and are known to cause acute hemolytic anemia in people with G6PD deficiency. The G6PD level can be normal after a recent blood transfusion and, as in our patient, during an acute episode of hemolysis.

Because of the strong suspicion of G6PD deficiency, both drugs were stopped when the patient was discharged from the hospital. She did not take either drug for 3 months. Her G6PD level was then retested and was found to be low, confirming the diagnosis. The patient was then advised not to take those drugs again. Since then, her hemoglobin level has remained stable and she has not needed any more blood transfusions.

Glycosuria used to be a sign of uncontrolled diabetes and was something to be corrected, not a therapeutic mechanism. But now we have a new class of drugs that lower plasma glucose levels by increasing the renal excretion of glucose.

Here, we will review canagliflozin, the first in a new class of drugs for type 2 diabetes: how it works, who is a candidate for it, and what to watch out for.

THE NEED FOR NEW DIABETES DRUGS

Diabetes mellitus affects more than 25.8 million people in the United States—8.3% of the population—and this staggering number is rising.¹ Among US residents age 65 and older, more than 10.9 million (26.9%) have diabetes.¹ People with uncontrolled diabetes are at risk of microvascular complications such as retinopathy, nephropathy, and neuropathy, as well as cardiovascular disease. Diabetes is the leading cause of blindness, chronic kidney disease, and nontraumatic lower-limb amputation in the United States.¹

Type 2 diabetes accounts for more than 90% of cases of diabetes in the United States, Europe, and Canada.² It is characterized by insulin resistance, decreased beta-cell function, and progressive beta-cell decline.³

Current American Diabetes Association guidelines for the treatment of diabetes recommend a hemoglobin A_1c target of less than 7.0%.⁴ Initial management includes lifestyle modifications such as changes in diet and an increase in exercise, as well as consideration of metformin treatment at the same time. If glucose levels remain uncontrolled despite these efforts, other drugs should be added.

A number of oral and injectable antihyperglycemic drugs are available to help achieve this goal, though none is without risk of adverse effects. Those available up to now include metformin, sulfonylureas, meglitinides, alpha-glucosidase inhibitors, thiazolidinediones, gliptins, glucagon-like peptide-1 agonists, amylin analogues, colesevelam, dopamine agonists, and insulin.⁵ Most of the available antihyperglycemics target the liver, pancreas, gut, and muscle to improve insulin sensitivity, reduce insulin resistance, or stimulate insulin secretion.

Despite the abundance of agents, type 2 diabetes remains uncontrolled in many patients. Only 57.1% of participants with previously diagnosed diabetes in the 2003–2006 National Health and Nutrition Examination Survey were at the hemoglobin A_1c goal of less than 7.0%.⁶ Possible reasons for failure include adverse effects such as hypoglycemia, weight gain, and gastrointestinal symptoms resulting in discontinued use, nonadherence to the prescribed regimen, and failure to increase the dosage or to add additional agents, including insulin, to optimize glycemic control as beta-cell function declines over time.

HOW THE KIDNEYS HANDLE GLUCOSE

In the kidney, glucose is filtered in the glomerulus and then is reabsorbed in the proximal tubule. Normally, the filtered glucose is all reabsorbed unless the glucose load exceeds the kidney’s absorptive capacity. Membrane proteins called sodium-glucose cotransporters reabsorb glucose at the proximal tubule and return it into the peripheral circulation. Glucose enters the tubular epithelial cell with sodium by passive cotransport via the sodiumglucose cotransporters, and then exits on the other side via the glucose transporter GLUT in the basolateral membrane.

Two sodium-glucose transporters that act in the proximal tubule of the kidney have been identified: SGLT1 and SGLT2. SGLT2 reabsorbs most of the glucose in the early segment of the proximal tubule, while SLGT1 reabsorbs the remaining glucose at the distal end.⁷ SGLT2 is responsible for more than 90% of renal tubular reabsorption of glucose and is found only in the proximal tubule, whereas SGLT1 is found mainly in the gastrointestinal tract.⁸

Patients with type 2 diabetes have a higher capacity for glucose reabsorption in the proximal tubule as a result of the up-regulation of SGLT2.⁹

SGLT2 INHIBITORS AND TYPE 2 DIABETES

Drugs that inhibit SGLT2 block reabsorption of glucose in the proximal tubule, lowering the renal threshold for glucose and thereby increasing urinary glucose excretion and lowering the serum glucose level in patients with hyperglycemia. This mechanism of action is insulin-independent.

On March 29, 2013, canagliflozin became the first SGLT2 inhibitor to be approved in the United States for the treatment of type 2 diabetes.¹⁰ However, it is not the first of its class to be introduced.

Dapagliflozin was the first SGLT2 inhibitor approved in Europe and has been available there since November 2012. However, the US Food and Drug Administration withheld its approval in the United States in January 2012 because of concerns of a possible association with cancer, specifically breast and bladder cancers, as well as possible liver injury.¹⁰ Canagliflozin does not appear to share this risk.

Several other SGLT2 inhibitors may soon be available. Empagliflozin is in phase III trials, and the manufacturer has filed for approval in the United States. Ipragliflozin is awaiting approval in Japan.

CANAGLIFLOZIN: PHARMACOKINETICS AND THERAPEUTIC EFFICACY

Canagliflozin reaches its peak plasma concentration within 1 to 2 hours of oral administration.¹¹ Its half-life is 10.6 hours with a 100-mg dose and 13.1 hours with a 300-mg dose. A steady state is typically achieved in 4 to 5 days.¹¹

Canagliflozin lowers fasting plasma glucose and hemoglobin A_1c levels in a dose-dependent manner.^10,11 These effects are independent of age, sex, body mass index, and race.¹² Postprandial glucose levels are also lowered.

Other potential benefits of canagliflozin include lowering of the systolic blood pressure and, especially important in obese people with type 2 diabetes, weight loss.¹² Aside from metformin, which occasionally results in modest weight loss, other oral drugs used in treating type 2 diabetes are weight-neutral or can cause weight gain.

Trials of canagliflozin

Nine phase III trials of canagliflozin have enrolled 10,285 patients, in one of the largest clinical trial programs in type 2 diabetes to date.¹⁰ Several of these trials evaluated canagliflozin as monotherapy, whereas others assessed its effect as an add-on therapy in combination with another antihyperglycemic agent such as a sulfonylurea, metformin, pioglitazone, or insulin. There has not yet been a trial directly comparing canagliflozin with metformin.

Four of the placebo-controlled trials evaluated canagliflozin as monotherapy, canagliflozin added to metformin alone, canagliflozin added to metformin plus glimepiride, and canagliflozin added to metformin plus pioglitazone.

When canagliflozin was used as monotherapy, hemoglobin A_1c levels at 26 weeks were an absolute 0.91% lower in the canagliflozin 100 mg/day group than in the placebo group, and an absolute 1.16% lower in the canagliflozin 300 mg/day group than in the placebo group (P < .001 for both).¹² Patients lost 2.8% of their body weight with canagliflozin 100 mg and 3.3% with canagliflozin 300 mg, compared with 0.6% with placebo. Systolic blood pressure fell by a mean of 3.7 mm Hg with the 100-mg dose and by a mean of 5.4 mm Hg with the 300-mg dose compared with placebo (P < .001 for both dose groups).¹²

When canagliflozin was added to metformin, with glimepiride as the comparator drug, there was a 5.2% weight reduction with the 100-mg dose, a 5.7% reduction with 300 mg, and a 1% gain with glimepiride. Hemoglobin A_1c fell about equally in the three groups.¹¹

When canagliflozin was added to metformin and a sulfonylurea, with sitagliptin as the comparator third drug, the 300-mg canagliflozin dosage group had a 2.8% weight reduction.¹¹

WHAT ARE THE ADVERSE EFFECTS?

Overall, canagliflozin seems to be well tolerated. The most common adverse effects reported in the clinical trials were genital yeast infections, urinary tract infections, and increased urination.

Genital yeast infections were more common in women than in men, occurring in 10.4% of women who received canagliflozin 100 mg and in 11.4% of women who received 300 mg, compared with only 3.2% in the placebo group.¹¹

Urinary tract infections occurred in 5.9% of the 100-mg group and in 4.3% of the 300-mg group, compared with 4.0% of the placebo group.¹¹

Postural hypotension. Lowering of blood pressure and symptoms of postural hypotension were also reported, and these may be attributed to the drug’s mild osmotic diuretic effect. The risk of adverse effects of volume depletion was dose-dependent; in patients over age 75, they occurred in 4.9% of those taking 100 mg and in 8.7% of those taking 300 mg, compared with 2.6% of those in the placebo or active-comparator groups.¹¹ Therefore, one should exercise particular caution when starting this drug in the elderly or in patients taking diuretics or multiple antihypertensive drugs.

Hypoglycemia. When canagliflozin was used as monotherapy, the incidence of hypoglycemia over 26 weeks was similar to that with placebo, occurring in 3.6% of the 100-mg group, 3.0% of the 300-mg group, and 2.6% of the placebo group.¹² Canagliflozin was associated with fewer episodes of hypoglycemia than were sulfonylureas, and the number of episodes was similar to that in patients taking gliptins. There was a higher overall incidence of hypoglycemia when canagliflozin was used in combination with a sulfonylurea or with insulin than when it was used as monotherapy.¹¹

Hyperkalemia. Patients with moderate renal impairment or who are on potassiumsparing drugs or drugs that interfere with the renin-angiotensin-aldosterone system may be at higher risk of hyperkalemia, so close monitoring of potassium is recommended. There was also a dose-dependent increase in serum phosphate and magnesium levels, more notably in patients with moderate renal impairment within the first 3 weeks of starting the drug.¹¹

Patients on canagliflozin who are also taking digoxin, ritonavir, phenytoin, phenobarbital, or rifampin should be closely monitored because of the risk of drug-drug interactions.¹¹ Specifically, there was an increase in mean peak digoxin concentrations when used with canagliflozin 300 mg, and the use of phenytoin, phenobarbital, and ritonavir decreased the efficacy of canagliflozin.

WHAT ARE THE CARDIOVASCULAR RISKS OR LONG-TERM CONCERNS?

Dose-dependent increases in low-density lipoprotein cholesterol (LDL-C) may be seen with canagliflozin. Mean changes from baseline compared with placebo were 4.4 mg/dL (4.5%) with canagliflozin 100 mg and 8.3 mg/dL (8%) with canagliflozin 300 mg.¹¹

There was also an increase in non-high-density lipoprotein cholesterol (non-HDL-C).¹² Compared with placebo, mean non-HDL-C levels rose by 2.1 mg/dL (1.5%) with canagliflozin 100 mg and 5.1 mg/dL (3.6%) with 300 mg.¹¹

In the 26-week canagliflozin monotherapy trial, archived blood samples in a small subgroup of patients (n = 349) were measured for apolipoprotein-B, which was found to increase by 1.2% with canagliflozin 100 mg and 3.5% with canagliflozin 300 mg, compared with 0.9% in the placebo group.¹²

Although small, the increase in LDL-C seen with this drug could be a concern, as diabetic patients are already at higher risk of cardiovascular events. The mechanism of this increase is not yet known, though it may be related to metabolic changes from urinary glucose excretion.¹²

The Canagliflozin Cardiovascular Assessment Study (CANVAS) is a randomized placebo-controlled trial in more than 4,000 patients with type 2 diabetes who have a history of or are at high risk of cardiovascular events. Currently under way, it is evaluating the occurrence of major adverse cardiovascular events (the primary end point) in patients randomized to receive canagliflozin 100 mg, canagliflozin 300 mg, or placebo once daily for up to 4 years. Secondary end points will be the drug’s effects on fasting plasma insulin and glucose, progression of albuminuria, body weight, blood pressure, HDL-C, LDL-C, bone mineral density, markers of bone turnover, and body composition.¹⁰ This trial will run for 9 years, to be completed in 2018.¹³

The CANVAS investigators have already reported that within the first month of treatment, 13 patients taking canagliflozin suffered a major cardiovascular event, including stroke (one of which was fatal) compared with just one patient taking placebo. These events were not seen after the first month. The hazard ratio for major adverse cardiovascular events within the first 30 days was 6.49, but this dropped to 0.89 after the first 30 days.¹⁰

Additional issues that should be addressed in long-term postmarketing studies include possible relationships with cancers and pancreatitis and the safety of the drug in pregnancy and in children with diabetes.¹⁰

WHO IS A CANDIDATE FOR THIS DRUG?

Canagliflozin is approved for use as monotherapy in addition to lifestyle modifications. It is also approved for use with other antihyperglycemic drugs, including metformin.

Obese patients with type 2 diabetes and normal kidney function may have the greatest benefit. Because of canagliflozin’s insulinin-dependent mechanism of action, patients with both early and late type 2 diabetes may benefit from its ability to lower hemoglobin A_1c and blood glucose.¹⁴

Although it can be used in patients with moderate (but not severe) kidney disease, canagliflozin does not appear to be as effective in these patients, who had higher rates of adverse effects.¹¹ It is not indicated for patients with type 1 diabetes, type 2 diabetes with ketonuria, or end-stage renal disease (estimated glomerular filtration rate < 45 mL/min or receiving dialysis).¹¹ It also is not yet recommended for use in pregnant women or patients under age 18.

The recommended starting dose of canagliflozin is 100 mg once daily, taken with breakfast. This can be increased to 300 mg once daily if tolerated. However, patients with an estimated glomerular filtration rate of 45 to 60 mL/min should not exceed the 100- mg dose. No dose adjustment is required in patients with mild to moderate hepatic impairment. It is not recommended, however, in patients with severe hepatic impairment.¹¹

Comment. Although canagliflozin is approved as monotherapy, metformin remains my choice for first-line oral therapy. Because canagliflozin is more expensive and its long-term affects are still relatively unknown, I prefer to use it as an adjunct, and believe it will be a useful addition, especially in obese patients who are seeking to lose weight.

WHAT IS THE COST OF THIS DRUG?

The suggested price is $10.53 per tablet (AmerisourceBergen), which is comparable to that of other newer drugs for type 2 diabetes.

THE BOTTOM LINE

The availability of canagliflozin as an additional oral antihyperglycemic option may prove helpful in managing patients with type 2 diabetes who experience adverse effects with other antihyperglycemic drugs.

As with any new drug, questions remain about the long-term risks of canagliflozin. However, it seems to be well tolerated, especially in patients with normal kidney function, and poses a low risk of hypoglycemia. The slight increase in LDL-C may prompt more aggressive lipid management. Whether blood pressure-lowering and weight loss will offset this increase in LDL-C is yet to be determined. Ongoing studies will help to further elucidate whether there is an increased risk of cardiovascular events.

Finally, canagliflozin distinguishes itself from other oral diabetes drugs by its added benefit of weight loss, an appealing side effect, especially in the growing population of obese individuals with type 2 diabetes mellitus.

Glycosuria used to be a sign of uncontrolled diabetes and was something to be corrected, not a therapeutic mechanism. But now we have a new class of drugs that lower plasma glucose levels by increasing the renal excretion of glucose.

Here, we will review canagliflozin, the first in a new class of drugs for type 2 diabetes: how it works, who is a candidate for it, and what to watch out for.

THE NEED FOR NEW DIABETES DRUGS

Diabetes mellitus affects more than 25.8 million people in the United States—8.3% of the population—and this staggering number is rising.¹ Among US residents age 65 and older, more than 10.9 million (26.9%) have diabetes.¹ People with uncontrolled diabetes are at risk of microvascular complications such as retinopathy, nephropathy, and neuropathy, as well as cardiovascular disease. Diabetes is the leading cause of blindness, chronic kidney disease, and nontraumatic lower-limb amputation in the United States.¹

Type 2 diabetes accounts for more than 90% of cases of diabetes in the United States, Europe, and Canada.² It is characterized by insulin resistance, decreased beta-cell function, and progressive beta-cell decline.³

Current American Diabetes Association guidelines for the treatment of diabetes recommend a hemoglobin A_1c target of less than 7.0%.⁴ Initial management includes lifestyle modifications such as changes in diet and an increase in exercise, as well as consideration of metformin treatment at the same time. If glucose levels remain uncontrolled despite these efforts, other drugs should be added.

A number of oral and injectable antihyperglycemic drugs are available to help achieve this goal, though none is without risk of adverse effects. Those available up to now include metformin, sulfonylureas, meglitinides, alpha-glucosidase inhibitors, thiazolidinediones, gliptins, glucagon-like peptide-1 agonists, amylin analogues, colesevelam, dopamine agonists, and insulin.⁵ Most of the available antihyperglycemics target the liver, pancreas, gut, and muscle to improve insulin sensitivity, reduce insulin resistance, or stimulate insulin secretion.

Despite the abundance of agents, type 2 diabetes remains uncontrolled in many patients. Only 57.1% of participants with previously diagnosed diabetes in the 2003–2006 National Health and Nutrition Examination Survey were at the hemoglobin A_1c goal of less than 7.0%.⁶ Possible reasons for failure include adverse effects such as hypoglycemia, weight gain, and gastrointestinal symptoms resulting in discontinued use, nonadherence to the prescribed regimen, and failure to increase the dosage or to add additional agents, including insulin, to optimize glycemic control as beta-cell function declines over time.

HOW THE KIDNEYS HANDLE GLUCOSE

In the kidney, glucose is filtered in the glomerulus and then is reabsorbed in the proximal tubule. Normally, the filtered glucose is all reabsorbed unless the glucose load exceeds the kidney’s absorptive capacity. Membrane proteins called sodium-glucose cotransporters reabsorb glucose at the proximal tubule and return it into the peripheral circulation. Glucose enters the tubular epithelial cell with sodium by passive cotransport via the sodiumglucose cotransporters, and then exits on the other side via the glucose transporter GLUT in the basolateral membrane.

Two sodium-glucose transporters that act in the proximal tubule of the kidney have been identified: SGLT1 and SGLT2. SGLT2 reabsorbs most of the glucose in the early segment of the proximal tubule, while SLGT1 reabsorbs the remaining glucose at the distal end.⁷ SGLT2 is responsible for more than 90% of renal tubular reabsorption of glucose and is found only in the proximal tubule, whereas SGLT1 is found mainly in the gastrointestinal tract.⁸

Patients with type 2 diabetes have a higher capacity for glucose reabsorption in the proximal tubule as a result of the up-regulation of SGLT2.⁹

SGLT2 INHIBITORS AND TYPE 2 DIABETES

Drugs that inhibit SGLT2 block reabsorption of glucose in the proximal tubule, lowering the renal threshold for glucose and thereby increasing urinary glucose excretion and lowering the serum glucose level in patients with hyperglycemia. This mechanism of action is insulin-independent.

On March 29, 2013, canagliflozin became the first SGLT2 inhibitor to be approved in the United States for the treatment of type 2 diabetes.¹⁰ However, it is not the first of its class to be introduced.

Dapagliflozin was the first SGLT2 inhibitor approved in Europe and has been available there since November 2012. However, the US Food and Drug Administration withheld its approval in the United States in January 2012 because of concerns of a possible association with cancer, specifically breast and bladder cancers, as well as possible liver injury.¹⁰ Canagliflozin does not appear to share this risk.

Several other SGLT2 inhibitors may soon be available. Empagliflozin is in phase III trials, and the manufacturer has filed for approval in the United States. Ipragliflozin is awaiting approval in Japan.

CANAGLIFLOZIN: PHARMACOKINETICS AND THERAPEUTIC EFFICACY

Canagliflozin reaches its peak plasma concentration within 1 to 2 hours of oral administration.¹¹ Its half-life is 10.6 hours with a 100-mg dose and 13.1 hours with a 300-mg dose. A steady state is typically achieved in 4 to 5 days.¹¹

Canagliflozin lowers fasting plasma glucose and hemoglobin A_1c levels in a dose-dependent manner.^10,11 These effects are independent of age, sex, body mass index, and race.¹² Postprandial glucose levels are also lowered.

Other potential benefits of canagliflozin include lowering of the systolic blood pressure and, especially important in obese people with type 2 diabetes, weight loss.¹² Aside from metformin, which occasionally results in modest weight loss, other oral drugs used in treating type 2 diabetes are weight-neutral or can cause weight gain.

Trials of canagliflozin

Nine phase III trials of canagliflozin have enrolled 10,285 patients, in one of the largest clinical trial programs in type 2 diabetes to date.¹⁰ Several of these trials evaluated canagliflozin as monotherapy, whereas others assessed its effect as an add-on therapy in combination with another antihyperglycemic agent such as a sulfonylurea, metformin, pioglitazone, or insulin. There has not yet been a trial directly comparing canagliflozin with metformin.

Four of the placebo-controlled trials evaluated canagliflozin as monotherapy, canagliflozin added to metformin alone, canagliflozin added to metformin plus glimepiride, and canagliflozin added to metformin plus pioglitazone.

When canagliflozin was used as monotherapy, hemoglobin A_1c levels at 26 weeks were an absolute 0.91% lower in the canagliflozin 100 mg/day group than in the placebo group, and an absolute 1.16% lower in the canagliflozin 300 mg/day group than in the placebo group (P < .001 for both).¹² Patients lost 2.8% of their body weight with canagliflozin 100 mg and 3.3% with canagliflozin 300 mg, compared with 0.6% with placebo. Systolic blood pressure fell by a mean of 3.7 mm Hg with the 100-mg dose and by a mean of 5.4 mm Hg with the 300-mg dose compared with placebo (P < .001 for both dose groups).¹²

When canagliflozin was added to metformin, with glimepiride as the comparator drug, there was a 5.2% weight reduction with the 100-mg dose, a 5.7% reduction with 300 mg, and a 1% gain with glimepiride. Hemoglobin A_1c fell about equally in the three groups.¹¹

When canagliflozin was added to metformin and a sulfonylurea, with sitagliptin as the comparator third drug, the 300-mg canagliflozin dosage group had a 2.8% weight reduction.¹¹

WHAT ARE THE ADVERSE EFFECTS?

Overall, canagliflozin seems to be well tolerated. The most common adverse effects reported in the clinical trials were genital yeast infections, urinary tract infections, and increased urination.

Genital yeast infections were more common in women than in men, occurring in 10.4% of women who received canagliflozin 100 mg and in 11.4% of women who received 300 mg, compared with only 3.2% in the placebo group.¹¹

Urinary tract infections occurred in 5.9% of the 100-mg group and in 4.3% of the 300-mg group, compared with 4.0% of the placebo group.¹¹

Postural hypotension. Lowering of blood pressure and symptoms of postural hypotension were also reported, and these may be attributed to the drug’s mild osmotic diuretic effect. The risk of adverse effects of volume depletion was dose-dependent; in patients over age 75, they occurred in 4.9% of those taking 100 mg and in 8.7% of those taking 300 mg, compared with 2.6% of those in the placebo or active-comparator groups.¹¹ Therefore, one should exercise particular caution when starting this drug in the elderly or in patients taking diuretics or multiple antihypertensive drugs.

Hypoglycemia. When canagliflozin was used as monotherapy, the incidence of hypoglycemia over 26 weeks was similar to that with placebo, occurring in 3.6% of the 100-mg group, 3.0% of the 300-mg group, and 2.6% of the placebo group.¹² Canagliflozin was associated with fewer episodes of hypoglycemia than were sulfonylureas, and the number of episodes was similar to that in patients taking gliptins. There was a higher overall incidence of hypoglycemia when canagliflozin was used in combination with a sulfonylurea or with insulin than when it was used as monotherapy.¹¹

Hyperkalemia. Patients with moderate renal impairment or who are on potassiumsparing drugs or drugs that interfere with the renin-angiotensin-aldosterone system may be at higher risk of hyperkalemia, so close monitoring of potassium is recommended. There was also a dose-dependent increase in serum phosphate and magnesium levels, more notably in patients with moderate renal impairment within the first 3 weeks of starting the drug.¹¹

Patients on canagliflozin who are also taking digoxin, ritonavir, phenytoin, phenobarbital, or rifampin should be closely monitored because of the risk of drug-drug interactions.¹¹ Specifically, there was an increase in mean peak digoxin concentrations when used with canagliflozin 300 mg, and the use of phenytoin, phenobarbital, and ritonavir decreased the efficacy of canagliflozin.

WHAT ARE THE CARDIOVASCULAR RISKS OR LONG-TERM CONCERNS?

Dose-dependent increases in low-density lipoprotein cholesterol (LDL-C) may be seen with canagliflozin. Mean changes from baseline compared with placebo were 4.4 mg/dL (4.5%) with canagliflozin 100 mg and 8.3 mg/dL (8%) with canagliflozin 300 mg.¹¹

There was also an increase in non-high-density lipoprotein cholesterol (non-HDL-C).¹² Compared with placebo, mean non-HDL-C levels rose by 2.1 mg/dL (1.5%) with canagliflozin 100 mg and 5.1 mg/dL (3.6%) with 300 mg.¹¹

In the 26-week canagliflozin monotherapy trial, archived blood samples in a small subgroup of patients (n = 349) were measured for apolipoprotein-B, which was found to increase by 1.2% with canagliflozin 100 mg and 3.5% with canagliflozin 300 mg, compared with 0.9% in the placebo group.¹²

Although small, the increase in LDL-C seen with this drug could be a concern, as diabetic patients are already at higher risk of cardiovascular events. The mechanism of this increase is not yet known, though it may be related to metabolic changes from urinary glucose excretion.¹²

The Canagliflozin Cardiovascular Assessment Study (CANVAS) is a randomized placebo-controlled trial in more than 4,000 patients with type 2 diabetes who have a history of or are at high risk of cardiovascular events. Currently under way, it is evaluating the occurrence of major adverse cardiovascular events (the primary end point) in patients randomized to receive canagliflozin 100 mg, canagliflozin 300 mg, or placebo once daily for up to 4 years. Secondary end points will be the drug’s effects on fasting plasma insulin and glucose, progression of albuminuria, body weight, blood pressure, HDL-C, LDL-C, bone mineral density, markers of bone turnover, and body composition.¹⁰ This trial will run for 9 years, to be completed in 2018.¹³

The CANVAS investigators have already reported that within the first month of treatment, 13 patients taking canagliflozin suffered a major cardiovascular event, including stroke (one of which was fatal) compared with just one patient taking placebo. These events were not seen after the first month. The hazard ratio for major adverse cardiovascular events within the first 30 days was 6.49, but this dropped to 0.89 after the first 30 days.¹⁰

Additional issues that should be addressed in long-term postmarketing studies include possible relationships with cancers and pancreatitis and the safety of the drug in pregnancy and in children with diabetes.¹⁰

WHO IS A CANDIDATE FOR THIS DRUG?

Canagliflozin is approved for use as monotherapy in addition to lifestyle modifications. It is also approved for use with other antihyperglycemic drugs, including metformin.

Obese patients with type 2 diabetes and normal kidney function may have the greatest benefit. Because of canagliflozin’s insulinin-dependent mechanism of action, patients with both early and late type 2 diabetes may benefit from its ability to lower hemoglobin A_1c and blood glucose.¹⁴

Although it can be used in patients with moderate (but not severe) kidney disease, canagliflozin does not appear to be as effective in these patients, who had higher rates of adverse effects.¹¹ It is not indicated for patients with type 1 diabetes, type 2 diabetes with ketonuria, or end-stage renal disease (estimated glomerular filtration rate < 45 mL/min or receiving dialysis).¹¹ It also is not yet recommended for use in pregnant women or patients under age 18.

The recommended starting dose of canagliflozin is 100 mg once daily, taken with breakfast. This can be increased to 300 mg once daily if tolerated. However, patients with an estimated glomerular filtration rate of 45 to 60 mL/min should not exceed the 100- mg dose. No dose adjustment is required in patients with mild to moderate hepatic impairment. It is not recommended, however, in patients with severe hepatic impairment.¹¹

Comment. Although canagliflozin is approved as monotherapy, metformin remains my choice for first-line oral therapy. Because canagliflozin is more expensive and its long-term affects are still relatively unknown, I prefer to use it as an adjunct, and believe it will be a useful addition, especially in obese patients who are seeking to lose weight.

WHAT IS THE COST OF THIS DRUG?

The suggested price is $10.53 per tablet (AmerisourceBergen), which is comparable to that of other newer drugs for type 2 diabetes.

THE BOTTOM LINE

The availability of canagliflozin as an additional oral antihyperglycemic option may prove helpful in managing patients with type 2 diabetes who experience adverse effects with other antihyperglycemic drugs.

As with any new drug, questions remain about the long-term risks of canagliflozin. However, it seems to be well tolerated, especially in patients with normal kidney function, and poses a low risk of hypoglycemia. The slight increase in LDL-C may prompt more aggressive lipid management. Whether blood pressure-lowering and weight loss will offset this increase in LDL-C is yet to be determined. Ongoing studies will help to further elucidate whether there is an increased risk of cardiovascular events.

Finally, canagliflozin distinguishes itself from other oral diabetes drugs by its added benefit of weight loss, an appealing side effect, especially in the growing population of obese individuals with type 2 diabetes mellitus.

Glycosuria used to be a sign of uncontrolled diabetes and was something to be corrected, not a therapeutic mechanism. But now we have a new class of drugs that lower plasma glucose levels by increasing the renal excretion of glucose.

Here, we will review canagliflozin, the first in a new class of drugs for type 2 diabetes: how it works, who is a candidate for it, and what to watch out for.

THE NEED FOR NEW DIABETES DRUGS

Diabetes mellitus affects more than 25.8 million people in the United States—8.3% of the population—and this staggering number is rising.¹ Among US residents age 65 and older, more than 10.9 million (26.9%) have diabetes.¹ People with uncontrolled diabetes are at risk of microvascular complications such as retinopathy, nephropathy, and neuropathy, as well as cardiovascular disease. Diabetes is the leading cause of blindness, chronic kidney disease, and nontraumatic lower-limb amputation in the United States.¹

Type 2 diabetes accounts for more than 90% of cases of diabetes in the United States, Europe, and Canada.² It is characterized by insulin resistance, decreased beta-cell function, and progressive beta-cell decline.³

Current American Diabetes Association guidelines for the treatment of diabetes recommend a hemoglobin A_1c target of less than 7.0%.⁴ Initial management includes lifestyle modifications such as changes in diet and an increase in exercise, as well as consideration of metformin treatment at the same time. If glucose levels remain uncontrolled despite these efforts, other drugs should be added.

A number of oral and injectable antihyperglycemic drugs are available to help achieve this goal, though none is without risk of adverse effects. Those available up to now include metformin, sulfonylureas, meglitinides, alpha-glucosidase inhibitors, thiazolidinediones, gliptins, glucagon-like peptide-1 agonists, amylin analogues, colesevelam, dopamine agonists, and insulin.⁵ Most of the available antihyperglycemics target the liver, pancreas, gut, and muscle to improve insulin sensitivity, reduce insulin resistance, or stimulate insulin secretion.

Despite the abundance of agents, type 2 diabetes remains uncontrolled in many patients. Only 57.1% of participants with previously diagnosed diabetes in the 2003–2006 National Health and Nutrition Examination Survey were at the hemoglobin A_1c goal of less than 7.0%.⁶ Possible reasons for failure include adverse effects such as hypoglycemia, weight gain, and gastrointestinal symptoms resulting in discontinued use, nonadherence to the prescribed regimen, and failure to increase the dosage or to add additional agents, including insulin, to optimize glycemic control as beta-cell function declines over time.

HOW THE KIDNEYS HANDLE GLUCOSE

In the kidney, glucose is filtered in the glomerulus and then is reabsorbed in the proximal tubule. Normally, the filtered glucose is all reabsorbed unless the glucose load exceeds the kidney’s absorptive capacity. Membrane proteins called sodium-glucose cotransporters reabsorb glucose at the proximal tubule and return it into the peripheral circulation. Glucose enters the tubular epithelial cell with sodium by passive cotransport via the sodiumglucose cotransporters, and then exits on the other side via the glucose transporter GLUT in the basolateral membrane.

Two sodium-glucose transporters that act in the proximal tubule of the kidney have been identified: SGLT1 and SGLT2. SGLT2 reabsorbs most of the glucose in the early segment of the proximal tubule, while SLGT1 reabsorbs the remaining glucose at the distal end.⁷ SGLT2 is responsible for more than 90% of renal tubular reabsorption of glucose and is found only in the proximal tubule, whereas SGLT1 is found mainly in the gastrointestinal tract.⁸

Patients with type 2 diabetes have a higher capacity for glucose reabsorption in the proximal tubule as a result of the up-regulation of SGLT2.⁹

SGLT2 INHIBITORS AND TYPE 2 DIABETES

Drugs that inhibit SGLT2 block reabsorption of glucose in the proximal tubule, lowering the renal threshold for glucose and thereby increasing urinary glucose excretion and lowering the serum glucose level in patients with hyperglycemia. This mechanism of action is insulin-independent.

On March 29, 2013, canagliflozin became the first SGLT2 inhibitor to be approved in the United States for the treatment of type 2 diabetes.¹⁰ However, it is not the first of its class to be introduced.

Dapagliflozin was the first SGLT2 inhibitor approved in Europe and has been available there since November 2012. However, the US Food and Drug Administration withheld its approval in the United States in January 2012 because of concerns of a possible association with cancer, specifically breast and bladder cancers, as well as possible liver injury.¹⁰ Canagliflozin does not appear to share this risk.

Several other SGLT2 inhibitors may soon be available. Empagliflozin is in phase III trials, and the manufacturer has filed for approval in the United States. Ipragliflozin is awaiting approval in Japan.

CANAGLIFLOZIN: PHARMACOKINETICS AND THERAPEUTIC EFFICACY

Canagliflozin reaches its peak plasma concentration within 1 to 2 hours of oral administration.¹¹ Its half-life is 10.6 hours with a 100-mg dose and 13.1 hours with a 300-mg dose. A steady state is typically achieved in 4 to 5 days.¹¹

Canagliflozin lowers fasting plasma glucose and hemoglobin A_1c levels in a dose-dependent manner.^10,11 These effects are independent of age, sex, body mass index, and race.¹² Postprandial glucose levels are also lowered.

Other potential benefits of canagliflozin include lowering of the systolic blood pressure and, especially important in obese people with type 2 diabetes, weight loss.¹² Aside from metformin, which occasionally results in modest weight loss, other oral drugs used in treating type 2 diabetes are weight-neutral or can cause weight gain.

Trials of canagliflozin

Nine phase III trials of canagliflozin have enrolled 10,285 patients, in one of the largest clinical trial programs in type 2 diabetes to date.¹⁰ Several of these trials evaluated canagliflozin as monotherapy, whereas others assessed its effect as an add-on therapy in combination with another antihyperglycemic agent such as a sulfonylurea, metformin, pioglitazone, or insulin. There has not yet been a trial directly comparing canagliflozin with metformin.

Four of the placebo-controlled trials evaluated canagliflozin as monotherapy, canagliflozin added to metformin alone, canagliflozin added to metformin plus glimepiride, and canagliflozin added to metformin plus pioglitazone.

When canagliflozin was used as monotherapy, hemoglobin A_1c levels at 26 weeks were an absolute 0.91% lower in the canagliflozin 100 mg/day group than in the placebo group, and an absolute 1.16% lower in the canagliflozin 300 mg/day group than in the placebo group (P < .001 for both).¹² Patients lost 2.8% of their body weight with canagliflozin 100 mg and 3.3% with canagliflozin 300 mg, compared with 0.6% with placebo. Systolic blood pressure fell by a mean of 3.7 mm Hg with the 100-mg dose and by a mean of 5.4 mm Hg with the 300-mg dose compared with placebo (P < .001 for both dose groups).¹²

When canagliflozin was added to metformin, with glimepiride as the comparator drug, there was a 5.2% weight reduction with the 100-mg dose, a 5.7% reduction with 300 mg, and a 1% gain with glimepiride. Hemoglobin A_1c fell about equally in the three groups.¹¹

When canagliflozin was added to metformin and a sulfonylurea, with sitagliptin as the comparator third drug, the 300-mg canagliflozin dosage group had a 2.8% weight reduction.¹¹

WHAT ARE THE ADVERSE EFFECTS?

Overall, canagliflozin seems to be well tolerated. The most common adverse effects reported in the clinical trials were genital yeast infections, urinary tract infections, and increased urination.

Genital yeast infections were more common in women than in men, occurring in 10.4% of women who received canagliflozin 100 mg and in 11.4% of women who received 300 mg, compared with only 3.2% in the placebo group.¹¹

Urinary tract infections occurred in 5.9% of the 100-mg group and in 4.3% of the 300-mg group, compared with 4.0% of the placebo group.¹¹

Postural hypotension. Lowering of blood pressure and symptoms of postural hypotension were also reported, and these may be attributed to the drug’s mild osmotic diuretic effect. The risk of adverse effects of volume depletion was dose-dependent; in patients over age 75, they occurred in 4.9% of those taking 100 mg and in 8.7% of those taking 300 mg, compared with 2.6% of those in the placebo or active-comparator groups.¹¹ Therefore, one should exercise particular caution when starting this drug in the elderly or in patients taking diuretics or multiple antihypertensive drugs.

Hypoglycemia. When canagliflozin was used as monotherapy, the incidence of hypoglycemia over 26 weeks was similar to that with placebo, occurring in 3.6% of the 100-mg group, 3.0% of the 300-mg group, and 2.6% of the placebo group.¹² Canagliflozin was associated with fewer episodes of hypoglycemia than were sulfonylureas, and the number of episodes was similar to that in patients taking gliptins. There was a higher overall incidence of hypoglycemia when canagliflozin was used in combination with a sulfonylurea or with insulin than when it was used as monotherapy.¹¹

Hyperkalemia. Patients with moderate renal impairment or who are on potassiumsparing drugs or drugs that interfere with the renin-angiotensin-aldosterone system may be at higher risk of hyperkalemia, so close monitoring of potassium is recommended. There was also a dose-dependent increase in serum phosphate and magnesium levels, more notably in patients with moderate renal impairment within the first 3 weeks of starting the drug.¹¹

Patients on canagliflozin who are also taking digoxin, ritonavir, phenytoin, phenobarbital, or rifampin should be closely monitored because of the risk of drug-drug interactions.¹¹ Specifically, there was an increase in mean peak digoxin concentrations when used with canagliflozin 300 mg, and the use of phenytoin, phenobarbital, and ritonavir decreased the efficacy of canagliflozin.

WHAT ARE THE CARDIOVASCULAR RISKS OR LONG-TERM CONCERNS?

Dose-dependent increases in low-density lipoprotein cholesterol (LDL-C) may be seen with canagliflozin. Mean changes from baseline compared with placebo were 4.4 mg/dL (4.5%) with canagliflozin 100 mg and 8.3 mg/dL (8%) with canagliflozin 300 mg.¹¹

There was also an increase in non-high-density lipoprotein cholesterol (non-HDL-C).¹² Compared with placebo, mean non-HDL-C levels rose by 2.1 mg/dL (1.5%) with canagliflozin 100 mg and 5.1 mg/dL (3.6%) with 300 mg.¹¹

In the 26-week canagliflozin monotherapy trial, archived blood samples in a small subgroup of patients (n = 349) were measured for apolipoprotein-B, which was found to increase by 1.2% with canagliflozin 100 mg and 3.5% with canagliflozin 300 mg, compared with 0.9% in the placebo group.¹²

Although small, the increase in LDL-C seen with this drug could be a concern, as diabetic patients are already at higher risk of cardiovascular events. The mechanism of this increase is not yet known, though it may be related to metabolic changes from urinary glucose excretion.¹²

The Canagliflozin Cardiovascular Assessment Study (CANVAS) is a randomized placebo-controlled trial in more than 4,000 patients with type 2 diabetes who have a history of or are at high risk of cardiovascular events. Currently under way, it is evaluating the occurrence of major adverse cardiovascular events (the primary end point) in patients randomized to receive canagliflozin 100 mg, canagliflozin 300 mg, or placebo once daily for up to 4 years. Secondary end points will be the drug’s effects on fasting plasma insulin and glucose, progression of albuminuria, body weight, blood pressure, HDL-C, LDL-C, bone mineral density, markers of bone turnover, and body composition.¹⁰ This trial will run for 9 years, to be completed in 2018.¹³

The CANVAS investigators have already reported that within the first month of treatment, 13 patients taking canagliflozin suffered a major cardiovascular event, including stroke (one of which was fatal) compared with just one patient taking placebo. These events were not seen after the first month. The hazard ratio for major adverse cardiovascular events within the first 30 days was 6.49, but this dropped to 0.89 after the first 30 days.¹⁰

Additional issues that should be addressed in long-term postmarketing studies include possible relationships with cancers and pancreatitis and the safety of the drug in pregnancy and in children with diabetes.¹⁰

WHO IS A CANDIDATE FOR THIS DRUG?

Canagliflozin is approved for use as monotherapy in addition to lifestyle modifications. It is also approved for use with other antihyperglycemic drugs, including metformin.

Obese patients with type 2 diabetes and normal kidney function may have the greatest benefit. Because of canagliflozin’s insulinin-dependent mechanism of action, patients with both early and late type 2 diabetes may benefit from its ability to lower hemoglobin A_1c and blood glucose.¹⁴

Although it can be used in patients with moderate (but not severe) kidney disease, canagliflozin does not appear to be as effective in these patients, who had higher rates of adverse effects.¹¹ It is not indicated for patients with type 1 diabetes, type 2 diabetes with ketonuria, or end-stage renal disease (estimated glomerular filtration rate < 45 mL/min or receiving dialysis).¹¹ It also is not yet recommended for use in pregnant women or patients under age 18.

The recommended starting dose of canagliflozin is 100 mg once daily, taken with breakfast. This can be increased to 300 mg once daily if tolerated. However, patients with an estimated glomerular filtration rate of 45 to 60 mL/min should not exceed the 100- mg dose. No dose adjustment is required in patients with mild to moderate hepatic impairment. It is not recommended, however, in patients with severe hepatic impairment.¹¹

Comment. Although canagliflozin is approved as monotherapy, metformin remains my choice for first-line oral therapy. Because canagliflozin is more expensive and its long-term affects are still relatively unknown, I prefer to use it as an adjunct, and believe it will be a useful addition, especially in obese patients who are seeking to lose weight.

WHAT IS THE COST OF THIS DRUG?

The suggested price is $10.53 per tablet (AmerisourceBergen), which is comparable to that of other newer drugs for type 2 diabetes.

THE BOTTOM LINE

The availability of canagliflozin as an additional oral antihyperglycemic option may prove helpful in managing patients with type 2 diabetes who experience adverse effects with other antihyperglycemic drugs.

As with any new drug, questions remain about the long-term risks of canagliflozin. However, it seems to be well tolerated, especially in patients with normal kidney function, and poses a low risk of hypoglycemia. The slight increase in LDL-C may prompt more aggressive lipid management. Whether blood pressure-lowering and weight loss will offset this increase in LDL-C is yet to be determined. Ongoing studies will help to further elucidate whether there is an increased risk of cardiovascular events.

Finally, canagliflozin distinguishes itself from other oral diabetes drugs by its added benefit of weight loss, an appealing side effect, especially in the growing population of obese individuals with type 2 diabetes mellitus.

With the variety of drugs available for treating depression, choosing one can be daunting. Different agents have characteristics that may make them a better choice for different types of patients, but even so, treating any kind of mental illness often requires an element of trial and error.

Primary care providers are on the frontline of treating mental illness, often evaluating patients before they are seen by a psychiatrist. The purpose of this article is to provide insight into the art of prescribing antidepressants in the primary care setting. We will discuss common patient presentations, including depressed patients without other medical comorbidities as well as those with common comorbidities, with our recommendations for first-line treatment.

We hope our recommendations will help you to navigate the uncertainty more confidently, resulting in more efficient and tailored treatment for your patients.

BASELINE TESTING

When starting a patient on antidepressant drug therapy, we recommend obtaining a set of baseline laboratory tests to rule out underlying medical conditions that may be contributing to the patient’s depression or that may preclude the use of a given drug. (For example, elevation of liver enzymes may preclude the use of duloxetine.) Tests should include:

• A complete blood cell count
• A complete metabolic panel
• A thyroid-stimulating hormone level.

Electrocardiography may also be useful, as some antidepressants can prolong the QT interval or elevate the blood levels of other drugs with this effect.

GENERAL TREATMENT CONSIDERATIONS

There are several classes of antidepressants, and each class has a number of agents. Research has found little difference in efficacy among agents. So to simplify choosing which one to use, we recommend becoming comfortable with an agent from each class, ie:

• A selective serotonin reuptake inhibitor (SSRI)
• A selective serotonin-norepinephrine reuptake inhibitor (SNRI)
• A tricyclic antidepressant (TCA)
• A monoamine oxidase (MAO) inhibitor.

Each class includes generic agents, many of which are on the discount lists of retail pharmacies. Table 1 shows representative drugs from each class, with their relative costs.

Start low and go slow. In general, when starting an antidepressant, consider starting at half the normal dose, titrating upward as tolerated about every 14 days. This approach can minimize side effects. For example, if prescribing fluoxetine, start with 10 mg and titrate every 2 weeks based on tolerance and patient response. That said, each patient may respond differently, requiring perhaps a lower starting dose or a longer titration schedule.

Anticipate side effects. Most of the side effects of an antidepressant drug can be explained by its mechanism of action. Although side effects should certainly be considered when choosing an agent, patients can be reassured that most are transient and benign. A detailed discussion of side effects of antidepressant drugs is beyond the scope of this article, but a review by Khawam et al¹ was published earlier in this journal.

Reassess. If after 4 to 6 weeks the patient has had little or no response, it is reasonable to switch agents. For a patient who was on an SSRI, the change can be to another SSRI or to an SNRI. However, if two SSRIs have already failed, then choose an SNRI. Agents are commonly cross-tapered during the switch to avoid abrupt cessation of one drug or the increased risk of adverse events such as cytochrome P450 interactions, serotonin syndrome, or hypertensive crisis (when switching to an MAO inhibitor).

Beware of interactions. All SSRIs and SNRIs are metabolized through the P450 system in the liver and therefore have the potential for drug-drug interactions. Care must be taken when giving these agents together with drugs whose metabolism can be altered by P450 inhibition. For TCAs, blood levels can be checked if there is concern about toxicity; however, dosing is not strictly based on this level. Great care should be taken if a TCA is given together with an SNRI or an SSRI, as the TCA blood level can become significantly elevated. This may result in QT interval prolongation, as mentioned earlier.

Refer. Referral to a psychiatrist is appropriate for patients for whom multiple classes have failed, for patients who have another psychiatric comorbidity (such as psychosis, hypomania, or mania), or for patients who may need hospitalization. Referral is also appropriate if the physician is concerned about suicide risk.

PATIENTS WITH MAJOR DEPRESSION ONLY

For a patient presenting with depression but no other significant medical comorbidity, the first-line therapy is often an SSRI. Several generic SSRIs are available, and some are on the discount lists at retail pharmacies.

Symptoms should start to improve in about 2 weeks, and the optimal response should be achieved in 4 to 6 weeks of treatment. If this does not occur, consider either adding an augmenting agent or switching to a different antidepressant.

PATIENTS WITH CHRONIC PAIN

Chronic pain and depression often go hand in hand and can potentiate each other. When considering an antidepressant in a patient who has both conditions, the SNRIs and TCAs are typically preferred. Some SNRIs, namely duloxetine and milnacipran, are approved for certain chronic pain conditions, such as fibromyalgia. SNRIs are frequently used off-label for other chronic pain conditions such as headache and neuropathic pain.²

TCAs such as amitriptyline, nortriptyline, and doxepin are also often used in patients with chronic pain. These agents, like the SNRIs, inhibit the reuptake of serotonin and norepinephrine and are used off-label for neuropathic pain,^3,4 migraine, interstitial cystitis,⁵ and other pain conditions.^6–9

For TCAs and SNRIs, the effective dose range for chronic pain overlaps that for depression. However, TCAs are often given at lower doses to patients without depression. We recommend starting at a low dose and slowly titrating upward to an effective dose. SNRIs are often preferred over TCAs because they do not have anticholinergic side effects and because an overdose is much less likely to be lethal.

PATIENTS WITH SEXUAL DYSFUNCTION

One of the more commonly reported side effects of antidepressants is sexual dysfunction, generally in the form of delayed orgasm or decreased libido.¹⁰ Typically, these complaints are attributed to SSRIs and SNRIs; however, TCAs and MAO inhibitors have also been associated wth sexual dysfunction.

Both erectile dysfunction and priapism have been linked to certain antidepressants. In particular, trazodone is a known cause of priapism. Even if using low doses for sleep, male patients should be made aware of this adverse effect.

Switching from one agent to another in the same class is not likely to improve sexual side effects. In particular, all the SSRIs are similar in their likelihood of causing sexual dysfunction. In a patient taking an SSRI who experiences this side effect, switching to bupropion¹¹ or mirtazapine¹² can be quite useful. Bupropion acts primarily on dopamine and norepinephrine, whereas mirtazapine acts on serotonin and norepinephrine but in a different manner from SSRIs and SNRIs.

Adjunctive treatment such as a cholinergic agonist, yohimbine (contraindicated with MAO inhibitors), a serotonergic agent (eg, buspirone), or a drug that acts on nitric oxide (eg, sildenafil, tadalafil) may have some utility but is often ineffective. Dose reduction, if possible, can be of value.

PATIENTS WITH ANXIETY

Many antidepressants are also approved for anxiety disorders, and still more are used off-label for this purpose. Anxiety and depression often occur together, so being able to treat both conditions with one drug can be quite useful.¹³ In general, the antidepressant effects are seen at lower doses of SSRIs and SNRIs, whereas more of the anxiolytic effects are seen at higher doses, particularly for obsessive-compulsive disorder.¹⁴

First-line treatment would be an SSRI or SNRI. Most anxiety disorders respond to either class, but there are some more-specific recommendations. SSRIs are best studied in panic disorder, generalized anxiety disorder, social anxiety disorder, posttraumatic stress disorder, and obsessive-compulsive disorder. Fluoxetine, citalopram, escitalopram, and sertraline¹⁵ can all be effective in both major depressive disorder and generalized anxiety disorder. Panic disorder also tends to respond well to SSRIs. SNRIs have been evaluated primarily in generalized anxiety disorder but may also be useful in many of the other conditions.

Additionally, mirtazapine (used off-label)¹² and the TCAs^16–18 can help treat anxiety. Clomipramine is used to treat obsessive-compulsive disorder.¹⁹ These drugs are especially useful for nighttime anxiety, as they can aid sleep. Of note, the anxiolytic effect of mirtazapine may be greater at higher doses.

MAO inhibitors often go unused because of the dietary and medication restrictions involved. However, very refractory cases of certain anxiety disorders may respond preferentially to these agents.

Bupropion tends to be more activating than other antidepressants, so is often avoided in anxious patients. However, some research suggests this is not always necessary.²⁰ If the anxiety is secondary to depression, it will often improve significantly with this agent.

When starting or increasing the dose of an antidepressant, patients may experience increased anxiety or feel “jittery.” This feeling usually passes within the first week of treatment, and it is important to inform patients about this effect. “Start low and go slow” in patients with significant comorbid anxiety. Temporarily using a benzodiazepine such as clonazepam may make the transition more tolerable.

PATIENTS WITH CHRONIC FATIGUE SYNDROME OR FIBROMYALGIA

Increasing recognition of both chronic fatigue syndrome and fibromyalgia has led to more proactive treatment for these disorders. Depression can go hand in hand with these disorders, and certain antidepressants, namely the SNRIs, can be useful in this population.

More data exist for the treatment of fibromyalgia. Both duloxetine and milnacipran are approved by the US Food and Drug Administration (FDA) for the treatment of fibromyalgia.²¹ Venlafaxine is also used off-label for this purpose. SSRIs such as fluoxetine and citalopram have had mixed results.^21–23 TCAs have been used with some success; however, their side effects and lethal potential are often limiting.^21,24,25 A recent study in Spain also suggested there may be benefit from using MAO inhibitors for fibromyalgia, but data are quite limited.²⁶

The data for treating chronic fatigue syndrome with SSRIs, SNRIs, or MAO inhibitors are conflicting.^27–29 However, managing the co-existing depression may provide some relief in and of itself.

PATIENTS WITH FREQUENT INSOMNIA

Insomnia can be a symptom of depression, but it can also be a side effect of certain antidepressants. The SSRIs and SNRIs can disrupt sleep patterns in some patients by shortening the rapid-eye-movement (REM) stage.^30,31

In patients with severe insomnia, it may be best to first recommend taking the antidepressant in the morning if they notice worsening sleep after initiating treatment. Patients can be told with any antidepressant, “If it makes you tired, take it at night, and if it wakes you up, take it in the morning.” Of note, a recent South African study suggested that escitalopram may be able to improve sleep.³²

If that does not solve the problem, there are other options. For instance, mirtazapine, particularly in doses of 15 mg or 30 mg, aids depression and insomnia. At higher doses (45 mg), the sleep-aiding effect may be reduced. Low doses of TCAs, particularly doxepin, maprotiline (technically speaking, a tetracyclic antidepressant), amitriptyline, and nortriptyline can be effective sleep aids. These agents may be used as an adjunct to another antidepressant to enhance sleep and mood. However, the TCAs also shorten the REM stage of sleep.³³

The previously mentioned drug interactions with SSRIs and SNRIs also need to be considered. Caution should be used when discontinuing these medications, as patients may experience rebound symptoms in the form of much more vivid dreams. MAO inhibitors may worsen insomnia because they suppress REM sleep.³⁴

Trazodone is another agent that at lower doses (25–150 mg) can be an effective, nonaddicting sleep aid. When used as an antidepressant, it is generally prescribed at higher doses (300–400 mg), but its sedating effects can be quite limiting at these levels. It is important to remember the possibility of priapism in male patients.

GERIATRIC PATIENTS

Old age brings its own set of concerns when treating depression. Elderly patients are more susceptible to potential bradycardia caused by SSRIs. The TCAs have the more worrisome cardiac side effect of QTc prolongation. TCAs can slow cognitive function, whereas the SSRIs, bupropion, and the SNRIs tend not to affect cognition. Escitalopram and duloxetine have been suggested to be particularly effective in the elderly.^35,36 A study from the Netherlands linked SSRIs with increased risk of falling in geriatric patients with dementia.³⁷ Constipation, which could lead to ileus, is increased with TCAs and certain other agents (ie, paroxetine) in the geriatric population.

Mirtazapine is often very useful in elderly patients for many reasons: it treats both anxiety and depression, stimulates appetite and weight gain, can help with nausea, and is an effective sleep aid. Concerns about weight, appetite, and sleep are particularly common in the elderly, whereas younger patients can be less tolerant of drugs that make them gain weight and sleep more. Normal age-related changes to the sleep cycle contribute to decreased satisfaction with sleep as we age. In addition, depression often further impairs sleep. So, in the elderly, optimizing sleep is key. Research has also shown mirtazapine to be effective in patients with both Alzheimer dementia and depression.³⁸

DIABETIC PATIENTS

One of the more worrisome side effects of psychiatric medications in diabetic patients is weight gain. Certain antidepressants have a greater propensity for weight gain and should likely be avoided as first-line treatments in this population.¹² Typically, these agents include those that have more antihistamine action such as paroxetine and the TCAs. These agents also may lead to constipation, which could potentially worsen gastroparesis. Mirtazapine and the MAO inhibitors are also known to cause weight gain.

Bupropion and nefazodone are the most weight-neutral of all antidepressants. Nefazodone has fallen out of favor because of its potential to cause fulminant liver failure in rare cases. However, it remains a reasonable option for patients with comorbid anxiety and depression who have significant weight gain with other agents.

SSRIs and MAO inhibitors may improve or be neutral toward glucose metabolism, and some data suggest that SNRIs may impair this process.³⁹

PATIENTS WITH CARDIAC CONDITIONS

Major depression often coexists with cardiac conditions. In particular, many patients develop depression after suffering a myocardial infarction, and increasingly they are being treated for it.⁴⁰ Treatment in this situation is appropriate, since depression, if untreated, can increase the risk of recurrence of myocardial infarction.⁴¹

However, there are many concerns that accompany treating depression in cardiac patients. Therefore, a baseline electrocardiogram should be obtained before starting an antidepressant.

TCAs and tetracyclic agents have a tendency to prolong the QTc interval and potentiate ventricular arrhythmias,⁴² so it may be prudent to avoid these in patients at risk. These agents can also significantly increase the pulse rate. This tachycardia increases the risk of angina or myocardial infarction from the anticholinergic effects of these drugs.

In February 2013, the FDA issued a warning about possible arrhythmias with citalopram at doses greater than 40 mg in adult patients⁴³; however, research has suggested citalopram is effective in treating depression in cardiac patients.⁴⁴ Research has not shown an increase in efficacy at doses greater than 40 mg daily, so we recommend following the black-box warning.

TCAs and MAO inhibitors can also cause orthostatic hypotension. On the other hand, consuming large amounts of tyramine, in foods such as aged cheese, can precipitate a hypertensive crisis in patients taking MAO inhibitors.

Which antidepressants tend to be safer in cardiac patients? Sertraline has been shown to be safe in congestive heart failure and coronary artery disease,^45–47 but the SSRIs are typically safe. Fluoxetine has shown efficacy in patients who have had a myocardial infarction.⁴⁸ Mirtazapine has also been shown to be efficacious in cardiac patients.⁴⁹ Nefazodone, mirtazapine, bupropion, SSRIs, and SNRIs have little or no tendency toward orthostatic hypotension.

With the variety of drugs available for treating depression, choosing one can be daunting. Different agents have characteristics that may make them a better choice for different types of patients, but even so, treating any kind of mental illness often requires an element of trial and error.

Primary care providers are on the frontline of treating mental illness, often evaluating patients before they are seen by a psychiatrist. The purpose of this article is to provide insight into the art of prescribing antidepressants in the primary care setting. We will discuss common patient presentations, including depressed patients without other medical comorbidities as well as those with common comorbidities, with our recommendations for first-line treatment.

We hope our recommendations will help you to navigate the uncertainty more confidently, resulting in more efficient and tailored treatment for your patients.

BASELINE TESTING

When starting a patient on antidepressant drug therapy, we recommend obtaining a set of baseline laboratory tests to rule out underlying medical conditions that may be contributing to the patient’s depression or that may preclude the use of a given drug. (For example, elevation of liver enzymes may preclude the use of duloxetine.) Tests should include:

• A complete blood cell count
• A complete metabolic panel
• A thyroid-stimulating hormone level.

Electrocardiography may also be useful, as some antidepressants can prolong the QT interval or elevate the blood levels of other drugs with this effect.

GENERAL TREATMENT CONSIDERATIONS

There are several classes of antidepressants, and each class has a number of agents. Research has found little difference in efficacy among agents. So to simplify choosing which one to use, we recommend becoming comfortable with an agent from each class, ie:

• A selective serotonin reuptake inhibitor (SSRI)
• A selective serotonin-norepinephrine reuptake inhibitor (SNRI)
• A tricyclic antidepressant (TCA)
• A monoamine oxidase (MAO) inhibitor.

Each class includes generic agents, many of which are on the discount lists of retail pharmacies. Table 1 shows representative drugs from each class, with their relative costs.

Start low and go slow. In general, when starting an antidepressant, consider starting at half the normal dose, titrating upward as tolerated about every 14 days. This approach can minimize side effects. For example, if prescribing fluoxetine, start with 10 mg and titrate every 2 weeks based on tolerance and patient response. That said, each patient may respond differently, requiring perhaps a lower starting dose or a longer titration schedule.

Anticipate side effects. Most of the side effects of an antidepressant drug can be explained by its mechanism of action. Although side effects should certainly be considered when choosing an agent, patients can be reassured that most are transient and benign. A detailed discussion of side effects of antidepressant drugs is beyond the scope of this article, but a review by Khawam et al¹ was published earlier in this journal.

Reassess. If after 4 to 6 weeks the patient has had little or no response, it is reasonable to switch agents. For a patient who was on an SSRI, the change can be to another SSRI or to an SNRI. However, if two SSRIs have already failed, then choose an SNRI. Agents are commonly cross-tapered during the switch to avoid abrupt cessation of one drug or the increased risk of adverse events such as cytochrome P450 interactions, serotonin syndrome, or hypertensive crisis (when switching to an MAO inhibitor).

Beware of interactions. All SSRIs and SNRIs are metabolized through the P450 system in the liver and therefore have the potential for drug-drug interactions. Care must be taken when giving these agents together with drugs whose metabolism can be altered by P450 inhibition. For TCAs, blood levels can be checked if there is concern about toxicity; however, dosing is not strictly based on this level. Great care should be taken if a TCA is given together with an SNRI or an SSRI, as the TCA blood level can become significantly elevated. This may result in QT interval prolongation, as mentioned earlier.

Refer. Referral to a psychiatrist is appropriate for patients for whom multiple classes have failed, for patients who have another psychiatric comorbidity (such as psychosis, hypomania, or mania), or for patients who may need hospitalization. Referral is also appropriate if the physician is concerned about suicide risk.

PATIENTS WITH MAJOR DEPRESSION ONLY

For a patient presenting with depression but no other significant medical comorbidity, the first-line therapy is often an SSRI. Several generic SSRIs are available, and some are on the discount lists at retail pharmacies.

Symptoms should start to improve in about 2 weeks, and the optimal response should be achieved in 4 to 6 weeks of treatment. If this does not occur, consider either adding an augmenting agent or switching to a different antidepressant.

PATIENTS WITH CHRONIC PAIN

Chronic pain and depression often go hand in hand and can potentiate each other. When considering an antidepressant in a patient who has both conditions, the SNRIs and TCAs are typically preferred. Some SNRIs, namely duloxetine and milnacipran, are approved for certain chronic pain conditions, such as fibromyalgia. SNRIs are frequently used off-label for other chronic pain conditions such as headache and neuropathic pain.²

TCAs such as amitriptyline, nortriptyline, and doxepin are also often used in patients with chronic pain. These agents, like the SNRIs, inhibit the reuptake of serotonin and norepinephrine and are used off-label for neuropathic pain,^3,4 migraine, interstitial cystitis,⁵ and other pain conditions.^6–9

For TCAs and SNRIs, the effective dose range for chronic pain overlaps that for depression. However, TCAs are often given at lower doses to patients without depression. We recommend starting at a low dose and slowly titrating upward to an effective dose. SNRIs are often preferred over TCAs because they do not have anticholinergic side effects and because an overdose is much less likely to be lethal.

PATIENTS WITH SEXUAL DYSFUNCTION

One of the more commonly reported side effects of antidepressants is sexual dysfunction, generally in the form of delayed orgasm or decreased libido.¹⁰ Typically, these complaints are attributed to SSRIs and SNRIs; however, TCAs and MAO inhibitors have also been associated wth sexual dysfunction.

Both erectile dysfunction and priapism have been linked to certain antidepressants. In particular, trazodone is a known cause of priapism. Even if using low doses for sleep, male patients should be made aware of this adverse effect.

Switching from one agent to another in the same class is not likely to improve sexual side effects. In particular, all the SSRIs are similar in their likelihood of causing sexual dysfunction. In a patient taking an SSRI who experiences this side effect, switching to bupropion¹¹ or mirtazapine¹² can be quite useful. Bupropion acts primarily on dopamine and norepinephrine, whereas mirtazapine acts on serotonin and norepinephrine but in a different manner from SSRIs and SNRIs.

Adjunctive treatment such as a cholinergic agonist, yohimbine (contraindicated with MAO inhibitors), a serotonergic agent (eg, buspirone), or a drug that acts on nitric oxide (eg, sildenafil, tadalafil) may have some utility but is often ineffective. Dose reduction, if possible, can be of value.

PATIENTS WITH ANXIETY

Many antidepressants are also approved for anxiety disorders, and still more are used off-label for this purpose. Anxiety and depression often occur together, so being able to treat both conditions with one drug can be quite useful.¹³ In general, the antidepressant effects are seen at lower doses of SSRIs and SNRIs, whereas more of the anxiolytic effects are seen at higher doses, particularly for obsessive-compulsive disorder.¹⁴

First-line treatment would be an SSRI or SNRI. Most anxiety disorders respond to either class, but there are some more-specific recommendations. SSRIs are best studied in panic disorder, generalized anxiety disorder, social anxiety disorder, posttraumatic stress disorder, and obsessive-compulsive disorder. Fluoxetine, citalopram, escitalopram, and sertraline¹⁵ can all be effective in both major depressive disorder and generalized anxiety disorder. Panic disorder also tends to respond well to SSRIs. SNRIs have been evaluated primarily in generalized anxiety disorder but may also be useful in many of the other conditions.

Additionally, mirtazapine (used off-label)¹² and the TCAs^16–18 can help treat anxiety. Clomipramine is used to treat obsessive-compulsive disorder.¹⁹ These drugs are especially useful for nighttime anxiety, as they can aid sleep. Of note, the anxiolytic effect of mirtazapine may be greater at higher doses.

MAO inhibitors often go unused because of the dietary and medication restrictions involved. However, very refractory cases of certain anxiety disorders may respond preferentially to these agents.

Bupropion tends to be more activating than other antidepressants, so is often avoided in anxious patients. However, some research suggests this is not always necessary.²⁰ If the anxiety is secondary to depression, it will often improve significantly with this agent.

When starting or increasing the dose of an antidepressant, patients may experience increased anxiety or feel “jittery.” This feeling usually passes within the first week of treatment, and it is important to inform patients about this effect. “Start low and go slow” in patients with significant comorbid anxiety. Temporarily using a benzodiazepine such as clonazepam may make the transition more tolerable.

PATIENTS WITH CHRONIC FATIGUE SYNDROME OR FIBROMYALGIA

Increasing recognition of both chronic fatigue syndrome and fibromyalgia has led to more proactive treatment for these disorders. Depression can go hand in hand with these disorders, and certain antidepressants, namely the SNRIs, can be useful in this population.

More data exist for the treatment of fibromyalgia. Both duloxetine and milnacipran are approved by the US Food and Drug Administration (FDA) for the treatment of fibromyalgia.²¹ Venlafaxine is also used off-label for this purpose. SSRIs such as fluoxetine and citalopram have had mixed results.^21–23 TCAs have been used with some success; however, their side effects and lethal potential are often limiting.^21,24,25 A recent study in Spain also suggested there may be benefit from using MAO inhibitors for fibromyalgia, but data are quite limited.²⁶

The data for treating chronic fatigue syndrome with SSRIs, SNRIs, or MAO inhibitors are conflicting.^27–29 However, managing the co-existing depression may provide some relief in and of itself.

PATIENTS WITH FREQUENT INSOMNIA

Insomnia can be a symptom of depression, but it can also be a side effect of certain antidepressants. The SSRIs and SNRIs can disrupt sleep patterns in some patients by shortening the rapid-eye-movement (REM) stage.^30,31

In patients with severe insomnia, it may be best to first recommend taking the antidepressant in the morning if they notice worsening sleep after initiating treatment. Patients can be told with any antidepressant, “If it makes you tired, take it at night, and if it wakes you up, take it in the morning.” Of note, a recent South African study suggested that escitalopram may be able to improve sleep.³²

If that does not solve the problem, there are other options. For instance, mirtazapine, particularly in doses of 15 mg or 30 mg, aids depression and insomnia. At higher doses (45 mg), the sleep-aiding effect may be reduced. Low doses of TCAs, particularly doxepin, maprotiline (technically speaking, a tetracyclic antidepressant), amitriptyline, and nortriptyline can be effective sleep aids. These agents may be used as an adjunct to another antidepressant to enhance sleep and mood. However, the TCAs also shorten the REM stage of sleep.³³

The previously mentioned drug interactions with SSRIs and SNRIs also need to be considered. Caution should be used when discontinuing these medications, as patients may experience rebound symptoms in the form of much more vivid dreams. MAO inhibitors may worsen insomnia because they suppress REM sleep.³⁴

Trazodone is another agent that at lower doses (25–150 mg) can be an effective, nonaddicting sleep aid. When used as an antidepressant, it is generally prescribed at higher doses (300–400 mg), but its sedating effects can be quite limiting at these levels. It is important to remember the possibility of priapism in male patients.

GERIATRIC PATIENTS

Old age brings its own set of concerns when treating depression. Elderly patients are more susceptible to potential bradycardia caused by SSRIs. The TCAs have the more worrisome cardiac side effect of QTc prolongation. TCAs can slow cognitive function, whereas the SSRIs, bupropion, and the SNRIs tend not to affect cognition. Escitalopram and duloxetine have been suggested to be particularly effective in the elderly.^35,36 A study from the Netherlands linked SSRIs with increased risk of falling in geriatric patients with dementia.³⁷ Constipation, which could lead to ileus, is increased with TCAs and certain other agents (ie, paroxetine) in the geriatric population.

Mirtazapine is often very useful in elderly patients for many reasons: it treats both anxiety and depression, stimulates appetite and weight gain, can help with nausea, and is an effective sleep aid. Concerns about weight, appetite, and sleep are particularly common in the elderly, whereas younger patients can be less tolerant of drugs that make them gain weight and sleep more. Normal age-related changes to the sleep cycle contribute to decreased satisfaction with sleep as we age. In addition, depression often further impairs sleep. So, in the elderly, optimizing sleep is key. Research has also shown mirtazapine to be effective in patients with both Alzheimer dementia and depression.³⁸

DIABETIC PATIENTS

One of the more worrisome side effects of psychiatric medications in diabetic patients is weight gain. Certain antidepressants have a greater propensity for weight gain and should likely be avoided as first-line treatments in this population.¹² Typically, these agents include those that have more antihistamine action such as paroxetine and the TCAs. These agents also may lead to constipation, which could potentially worsen gastroparesis. Mirtazapine and the MAO inhibitors are also known to cause weight gain.

Bupropion and nefazodone are the most weight-neutral of all antidepressants. Nefazodone has fallen out of favor because of its potential to cause fulminant liver failure in rare cases. However, it remains a reasonable option for patients with comorbid anxiety and depression who have significant weight gain with other agents.

SSRIs and MAO inhibitors may improve or be neutral toward glucose metabolism, and some data suggest that SNRIs may impair this process.³⁹

PATIENTS WITH CARDIAC CONDITIONS

Major depression often coexists with cardiac conditions. In particular, many patients develop depression after suffering a myocardial infarction, and increasingly they are being treated for it.⁴⁰ Treatment in this situation is appropriate, since depression, if untreated, can increase the risk of recurrence of myocardial infarction.⁴¹

However, there are many concerns that accompany treating depression in cardiac patients. Therefore, a baseline electrocardiogram should be obtained before starting an antidepressant.

TCAs and tetracyclic agents have a tendency to prolong the QTc interval and potentiate ventricular arrhythmias,⁴² so it may be prudent to avoid these in patients at risk. These agents can also significantly increase the pulse rate. This tachycardia increases the risk of angina or myocardial infarction from the anticholinergic effects of these drugs.

In February 2013, the FDA issued a warning about possible arrhythmias with citalopram at doses greater than 40 mg in adult patients⁴³; however, research has suggested citalopram is effective in treating depression in cardiac patients.⁴⁴ Research has not shown an increase in efficacy at doses greater than 40 mg daily, so we recommend following the black-box warning.

TCAs and MAO inhibitors can also cause orthostatic hypotension. On the other hand, consuming large amounts of tyramine, in foods such as aged cheese, can precipitate a hypertensive crisis in patients taking MAO inhibitors.

Which antidepressants tend to be safer in cardiac patients? Sertraline has been shown to be safe in congestive heart failure and coronary artery disease,^45–47 but the SSRIs are typically safe. Fluoxetine has shown efficacy in patients who have had a myocardial infarction.⁴⁸ Mirtazapine has also been shown to be efficacious in cardiac patients.⁴⁹ Nefazodone, mirtazapine, bupropion, SSRIs, and SNRIs have little or no tendency toward orthostatic hypotension.

With the variety of drugs available for treating depression, choosing one can be daunting. Different agents have characteristics that may make them a better choice for different types of patients, but even so, treating any kind of mental illness often requires an element of trial and error.

Primary care providers are on the frontline of treating mental illness, often evaluating patients before they are seen by a psychiatrist. The purpose of this article is to provide insight into the art of prescribing antidepressants in the primary care setting. We will discuss common patient presentations, including depressed patients without other medical comorbidities as well as those with common comorbidities, with our recommendations for first-line treatment.

We hope our recommendations will help you to navigate the uncertainty more confidently, resulting in more efficient and tailored treatment for your patients.

BASELINE TESTING

When starting a patient on antidepressant drug therapy, we recommend obtaining a set of baseline laboratory tests to rule out underlying medical conditions that may be contributing to the patient’s depression or that may preclude the use of a given drug. (For example, elevation of liver enzymes may preclude the use of duloxetine.) Tests should include:

• A complete blood cell count
• A complete metabolic panel
• A thyroid-stimulating hormone level.

Electrocardiography may also be useful, as some antidepressants can prolong the QT interval or elevate the blood levels of other drugs with this effect.

GENERAL TREATMENT CONSIDERATIONS

There are several classes of antidepressants, and each class has a number of agents. Research has found little difference in efficacy among agents. So to simplify choosing which one to use, we recommend becoming comfortable with an agent from each class, ie:

• A selective serotonin reuptake inhibitor (SSRI)
• A selective serotonin-norepinephrine reuptake inhibitor (SNRI)
• A tricyclic antidepressant (TCA)
• A monoamine oxidase (MAO) inhibitor.

Each class includes generic agents, many of which are on the discount lists of retail pharmacies. Table 1 shows representative drugs from each class, with their relative costs.

Start low and go slow. In general, when starting an antidepressant, consider starting at half the normal dose, titrating upward as tolerated about every 14 days. This approach can minimize side effects. For example, if prescribing fluoxetine, start with 10 mg and titrate every 2 weeks based on tolerance and patient response. That said, each patient may respond differently, requiring perhaps a lower starting dose or a longer titration schedule.

Anticipate side effects. Most of the side effects of an antidepressant drug can be explained by its mechanism of action. Although side effects should certainly be considered when choosing an agent, patients can be reassured that most are transient and benign. A detailed discussion of side effects of antidepressant drugs is beyond the scope of this article, but a review by Khawam et al¹ was published earlier in this journal.

Reassess. If after 4 to 6 weeks the patient has had little or no response, it is reasonable to switch agents. For a patient who was on an SSRI, the change can be to another SSRI or to an SNRI. However, if two SSRIs have already failed, then choose an SNRI. Agents are commonly cross-tapered during the switch to avoid abrupt cessation of one drug or the increased risk of adverse events such as cytochrome P450 interactions, serotonin syndrome, or hypertensive crisis (when switching to an MAO inhibitor).

Beware of interactions. All SSRIs and SNRIs are metabolized through the P450 system in the liver and therefore have the potential for drug-drug interactions. Care must be taken when giving these agents together with drugs whose metabolism can be altered by P450 inhibition. For TCAs, blood levels can be checked if there is concern about toxicity; however, dosing is not strictly based on this level. Great care should be taken if a TCA is given together with an SNRI or an SSRI, as the TCA blood level can become significantly elevated. This may result in QT interval prolongation, as mentioned earlier.

Refer. Referral to a psychiatrist is appropriate for patients for whom multiple classes have failed, for patients who have another psychiatric comorbidity (such as psychosis, hypomania, or mania), or for patients who may need hospitalization. Referral is also appropriate if the physician is concerned about suicide risk.

PATIENTS WITH MAJOR DEPRESSION ONLY

For a patient presenting with depression but no other significant medical comorbidity, the first-line therapy is often an SSRI. Several generic SSRIs are available, and some are on the discount lists at retail pharmacies.

Symptoms should start to improve in about 2 weeks, and the optimal response should be achieved in 4 to 6 weeks of treatment. If this does not occur, consider either adding an augmenting agent or switching to a different antidepressant.

PATIENTS WITH CHRONIC PAIN

Chronic pain and depression often go hand in hand and can potentiate each other. When considering an antidepressant in a patient who has both conditions, the SNRIs and TCAs are typically preferred. Some SNRIs, namely duloxetine and milnacipran, are approved for certain chronic pain conditions, such as fibromyalgia. SNRIs are frequently used off-label for other chronic pain conditions such as headache and neuropathic pain.²

TCAs such as amitriptyline, nortriptyline, and doxepin are also often used in patients with chronic pain. These agents, like the SNRIs, inhibit the reuptake of serotonin and norepinephrine and are used off-label for neuropathic pain,^3,4 migraine, interstitial cystitis,⁵ and other pain conditions.^6–9

For TCAs and SNRIs, the effective dose range for chronic pain overlaps that for depression. However, TCAs are often given at lower doses to patients without depression. We recommend starting at a low dose and slowly titrating upward to an effective dose. SNRIs are often preferred over TCAs because they do not have anticholinergic side effects and because an overdose is much less likely to be lethal.

PATIENTS WITH SEXUAL DYSFUNCTION

One of the more commonly reported side effects of antidepressants is sexual dysfunction, generally in the form of delayed orgasm or decreased libido.¹⁰ Typically, these complaints are attributed to SSRIs and SNRIs; however, TCAs and MAO inhibitors have also been associated wth sexual dysfunction.

Both erectile dysfunction and priapism have been linked to certain antidepressants. In particular, trazodone is a known cause of priapism. Even if using low doses for sleep, male patients should be made aware of this adverse effect.

Switching from one agent to another in the same class is not likely to improve sexual side effects. In particular, all the SSRIs are similar in their likelihood of causing sexual dysfunction. In a patient taking an SSRI who experiences this side effect, switching to bupropion¹¹ or mirtazapine¹² can be quite useful. Bupropion acts primarily on dopamine and norepinephrine, whereas mirtazapine acts on serotonin and norepinephrine but in a different manner from SSRIs and SNRIs.

Adjunctive treatment such as a cholinergic agonist, yohimbine (contraindicated with MAO inhibitors), a serotonergic agent (eg, buspirone), or a drug that acts on nitric oxide (eg, sildenafil, tadalafil) may have some utility but is often ineffective. Dose reduction, if possible, can be of value.

PATIENTS WITH ANXIETY

Many antidepressants are also approved for anxiety disorders, and still more are used off-label for this purpose. Anxiety and depression often occur together, so being able to treat both conditions with one drug can be quite useful.¹³ In general, the antidepressant effects are seen at lower doses of SSRIs and SNRIs, whereas more of the anxiolytic effects are seen at higher doses, particularly for obsessive-compulsive disorder.¹⁴

First-line treatment would be an SSRI or SNRI. Most anxiety disorders respond to either class, but there are some more-specific recommendations. SSRIs are best studied in panic disorder, generalized anxiety disorder, social anxiety disorder, posttraumatic stress disorder, and obsessive-compulsive disorder. Fluoxetine, citalopram, escitalopram, and sertraline¹⁵ can all be effective in both major depressive disorder and generalized anxiety disorder. Panic disorder also tends to respond well to SSRIs. SNRIs have been evaluated primarily in generalized anxiety disorder but may also be useful in many of the other conditions.

Additionally, mirtazapine (used off-label)¹² and the TCAs^16–18 can help treat anxiety. Clomipramine is used to treat obsessive-compulsive disorder.¹⁹ These drugs are especially useful for nighttime anxiety, as they can aid sleep. Of note, the anxiolytic effect of mirtazapine may be greater at higher doses.

MAO inhibitors often go unused because of the dietary and medication restrictions involved. However, very refractory cases of certain anxiety disorders may respond preferentially to these agents.

Bupropion tends to be more activating than other antidepressants, so is often avoided in anxious patients. However, some research suggests this is not always necessary.²⁰ If the anxiety is secondary to depression, it will often improve significantly with this agent.

When starting or increasing the dose of an antidepressant, patients may experience increased anxiety or feel “jittery.” This feeling usually passes within the first week of treatment, and it is important to inform patients about this effect. “Start low and go slow” in patients with significant comorbid anxiety. Temporarily using a benzodiazepine such as clonazepam may make the transition more tolerable.

PATIENTS WITH CHRONIC FATIGUE SYNDROME OR FIBROMYALGIA

Increasing recognition of both chronic fatigue syndrome and fibromyalgia has led to more proactive treatment for these disorders. Depression can go hand in hand with these disorders, and certain antidepressants, namely the SNRIs, can be useful in this population.

More data exist for the treatment of fibromyalgia. Both duloxetine and milnacipran are approved by the US Food and Drug Administration (FDA) for the treatment of fibromyalgia.²¹ Venlafaxine is also used off-label for this purpose. SSRIs such as fluoxetine and citalopram have had mixed results.^21–23 TCAs have been used with some success; however, their side effects and lethal potential are often limiting.^21,24,25 A recent study in Spain also suggested there may be benefit from using MAO inhibitors for fibromyalgia, but data are quite limited.²⁶

The data for treating chronic fatigue syndrome with SSRIs, SNRIs, or MAO inhibitors are conflicting.^27–29 However, managing the co-existing depression may provide some relief in and of itself.

PATIENTS WITH FREQUENT INSOMNIA

Insomnia can be a symptom of depression, but it can also be a side effect of certain antidepressants. The SSRIs and SNRIs can disrupt sleep patterns in some patients by shortening the rapid-eye-movement (REM) stage.^30,31

In patients with severe insomnia, it may be best to first recommend taking the antidepressant in the morning if they notice worsening sleep after initiating treatment. Patients can be told with any antidepressant, “If it makes you tired, take it at night, and if it wakes you up, take it in the morning.” Of note, a recent South African study suggested that escitalopram may be able to improve sleep.³²

If that does not solve the problem, there are other options. For instance, mirtazapine, particularly in doses of 15 mg or 30 mg, aids depression and insomnia. At higher doses (45 mg), the sleep-aiding effect may be reduced. Low doses of TCAs, particularly doxepin, maprotiline (technically speaking, a tetracyclic antidepressant), amitriptyline, and nortriptyline can be effective sleep aids. These agents may be used as an adjunct to another antidepressant to enhance sleep and mood. However, the TCAs also shorten the REM stage of sleep.³³

The previously mentioned drug interactions with SSRIs and SNRIs also need to be considered. Caution should be used when discontinuing these medications, as patients may experience rebound symptoms in the form of much more vivid dreams. MAO inhibitors may worsen insomnia because they suppress REM sleep.³⁴

Trazodone is another agent that at lower doses (25–150 mg) can be an effective, nonaddicting sleep aid. When used as an antidepressant, it is generally prescribed at higher doses (300–400 mg), but its sedating effects can be quite limiting at these levels. It is important to remember the possibility of priapism in male patients.

GERIATRIC PATIENTS

Old age brings its own set of concerns when treating depression. Elderly patients are more susceptible to potential bradycardia caused by SSRIs. The TCAs have the more worrisome cardiac side effect of QTc prolongation. TCAs can slow cognitive function, whereas the SSRIs, bupropion, and the SNRIs tend not to affect cognition. Escitalopram and duloxetine have been suggested to be particularly effective in the elderly.^35,36 A study from the Netherlands linked SSRIs with increased risk of falling in geriatric patients with dementia.³⁷ Constipation, which could lead to ileus, is increased with TCAs and certain other agents (ie, paroxetine) in the geriatric population.

Mirtazapine is often very useful in elderly patients for many reasons: it treats both anxiety and depression, stimulates appetite and weight gain, can help with nausea, and is an effective sleep aid. Concerns about weight, appetite, and sleep are particularly common in the elderly, whereas younger patients can be less tolerant of drugs that make them gain weight and sleep more. Normal age-related changes to the sleep cycle contribute to decreased satisfaction with sleep as we age. In addition, depression often further impairs sleep. So, in the elderly, optimizing sleep is key. Research has also shown mirtazapine to be effective in patients with both Alzheimer dementia and depression.³⁸

DIABETIC PATIENTS

One of the more worrisome side effects of psychiatric medications in diabetic patients is weight gain. Certain antidepressants have a greater propensity for weight gain and should likely be avoided as first-line treatments in this population.¹² Typically, these agents include those that have more antihistamine action such as paroxetine and the TCAs. These agents also may lead to constipation, which could potentially worsen gastroparesis. Mirtazapine and the MAO inhibitors are also known to cause weight gain.

Bupropion and nefazodone are the most weight-neutral of all antidepressants. Nefazodone has fallen out of favor because of its potential to cause fulminant liver failure in rare cases. However, it remains a reasonable option for patients with comorbid anxiety and depression who have significant weight gain with other agents.

SSRIs and MAO inhibitors may improve or be neutral toward glucose metabolism, and some data suggest that SNRIs may impair this process.³⁹

PATIENTS WITH CARDIAC CONDITIONS

Major depression often coexists with cardiac conditions. In particular, many patients develop depression after suffering a myocardial infarction, and increasingly they are being treated for it.⁴⁰ Treatment in this situation is appropriate, since depression, if untreated, can increase the risk of recurrence of myocardial infarction.⁴¹

However, there are many concerns that accompany treating depression in cardiac patients. Therefore, a baseline electrocardiogram should be obtained before starting an antidepressant.

TCAs and tetracyclic agents have a tendency to prolong the QTc interval and potentiate ventricular arrhythmias,⁴² so it may be prudent to avoid these in patients at risk. These agents can also significantly increase the pulse rate. This tachycardia increases the risk of angina or myocardial infarction from the anticholinergic effects of these drugs.

In February 2013, the FDA issued a warning about possible arrhythmias with citalopram at doses greater than 40 mg in adult patients⁴³; however, research has suggested citalopram is effective in treating depression in cardiac patients.⁴⁴ Research has not shown an increase in efficacy at doses greater than 40 mg daily, so we recommend following the black-box warning.

TCAs and MAO inhibitors can also cause orthostatic hypotension. On the other hand, consuming large amounts of tyramine, in foods such as aged cheese, can precipitate a hypertensive crisis in patients taking MAO inhibitors.

Which antidepressants tend to be safer in cardiac patients? Sertraline has been shown to be safe in congestive heart failure and coronary artery disease,^45–47 but the SSRIs are typically safe. Fluoxetine has shown efficacy in patients who have had a myocardial infarction.⁴⁸ Mirtazapine has also been shown to be efficacious in cardiac patients.⁴⁹ Nefazodone, mirtazapine, bupropion, SSRIs, and SNRIs have little or no tendency toward orthostatic hypotension.

Early thrombus removal for lower-extremity deep venous thrombosis (DVT) is at present only modestly supported by evidence and so remains controversial. It is largely aimed at preventing postthrombotic syndrome.

The decision to pursue early thrombus removal demands weighing the patient’s risk of postthrombotic syndrome against the risks and costs associated with thrombolysis and thrombectomy, such as bleeding complications. In the final analysis, this remains a subjective decision.

With these caveats in mind, the best candidate for early thrombus removal is a young patient with iliofemoral DVT with symptoms lasting fewer than 14 days.

POSTTHROMBOTIC SYNDROME IS COMMON

Anticoagulation with heparin and warfarin is the mainstay of DVT therapy. Indeed, the safety of this therapy and its effectiveness in reducing thrombus propagation and DVT recurrence are well established. Neither heparin nor warfarin, however, actively reduces the thrombus burden. Rather, both prevent the clot from propagating while it is, hopefully, gradually reabsorbed through endogenous mechanisms.

Up to 50% of DVT patients develop postthrombotic syndrome. A variety of mechanisms are involved, including persistent obstructive thrombosis and valvular injury.¹ But much remains unknown about the etiology, and some patients develop the condition in the absence of abnormalities on objective testing.

Symptoms of postthrombotic syndrome can range from mild heaviness, edema, erythema, and cramping in the affected limb to debilitating pain with classic signs of venous hypertension (eg, venous ectasia and ulcers). It accounts for significant health care costs and has a detrimental effect on quality of life.¹ Thus, there has been interest in early thrombus removal as initial therapy for DVT.

THROMBUS REMOVAL

Venous clots can be removed with open surgery or, more typically, with percutaneous catheter-based thrombolysis and thrombectomy devices that use high-velocity saline jets, ultrasonic energy, or wire oscillation to mechanically fragment the venous clot. All of these mechanisms help with drug delivery and pose a minimal risk of pulmonary embolism.

Evidence is weak

Patients with DVT of the iliac venous system or common femoral vein are at highest risk of postthrombotic syndrome. Therefore, the Society for Vascular Surgery and the American Venous Forum have issued a grade 2C (ie, weak) recommendation in favor of early thrombus removal in patients with a first-time episode of iliofemoral DVT with fewer than 14 days of symptoms.² Moreover, patients must have a low risk of bleeding complications, be ambulatory, and have reasonable life expectancy.

The recommendation is buttressed by a Cochrane meta-analysis that included 101 patients.³ It concluded that there was a significant decrement in the development of postthrombotic syndrome with thrombolysis (but without mechanical thrombectomy) compared with standard therapy: the rate was 48% (29/61) with thrombolysis, and 65% (26/40) with standard therapy.³

More recently, the Catheter-Directed Thrombolysis Versus Standard Treatment for Acute Iliofemoral Deep Vein Thrombosis (CaVenT) study, a randomized prospective trial in 189 patients, demonstrated a lower rate of postthrombotic syndrome at 24 months and increased iliofemoral patency at 6 months with catheter-directed thrombolysis with alteplase (41.1% and 65.9%) vs anticoagulation with heparin and warfarin alone (55.6% and 47.4%).⁴

The Acute Venous Thrombosis: Thrombus Removal With Adjunctive Catheter-directed Thrombolysis (ATTRACT) trial is an ongoing prospective randomized multicenter trial of the effect of thrombolysis on postthrombotic syndrome that also hopes to clarify the relative benefits of different methods of pharmacomechanical clot removal.

While CaVenT has not been criticized extensively in the literature, other studies supporting early intervention for iliofemoral venous thrombosis generally have been noted to have a number of shortcomings, including a lack of randomization, and consequent bias, and the use of surrogate end points instead of a direct assessment of postthrombotic syndrome.

Reflecting the weakness of the evidence, the American College of Chest Physicians has issued a grade 2C recommendation against catheter-directed thrombolysis and against thrombectomy in favor of anticoagulant therapy.⁵

A subjective, case-by-case decision

The decision on standard vs interventional therapy must be made case by case. For example, thrombus removal may be more appropriate for a physically active young patient who is more likely to be impaired by postthrombotic syndrome, whereas standard warfarin therapy may be preferable for a sedentary patient. We are also more inclined to offer thrombus removal to patients who have worse symptoms.

Complicating the issue, many patients present with a mix of variables that support and oppose intervention—eg, a moderately active elderly patient with an unclear life expectancy and a history of gastrointestinal bleeding. At present, there is no way to quantitatively evaluate the risks and rewards of thrombus removal, and the final decision is essentially subjective.

Additional facts warranting consideration include the possibility that thrombolysis may require several days of therapy with daily venography for evaluation. Monitoring in the intensive care unit is normally required during the period of thrombolysis. Patients should be apprised of these elements of therapy beforehand; obviously, those who are unwilling to comply are not candidates.

Not a substitute for anticoagulation

It is important to recognize that thrombus removal is not a substitute for standard heparin-warfarin anticoagulation, which must also be prescribed.⁵ Thus, patients who cannot tolerate standard post-DVT anticoagulation should not undergo thrombus removal. Furthermore, the current evidence supports the use of standard anticoagulation over early thrombus removal of DVTs that are more distal in the lower extremity, such as those in the popliteal vein.⁵

PHLEGMASIA CERULEA DOLENS IS A SPECIAL CASE

Phlegmasia cerulea dolens—acute venous outflow obstruction associated with edema, cyanosis, and pain that in the worst cases may lead to shock, limb loss, and death—constitutes a special case. Although we lack robust supporting evidence, phlegmasia is a commonly accepted indication for early thrombus removal as a means of limb salvage.^2,6

Early thrombus removal for lower-extremity deep venous thrombosis (DVT) is at present only modestly supported by evidence and so remains controversial. It is largely aimed at preventing postthrombotic syndrome.

The decision to pursue early thrombus removal demands weighing the patient’s risk of postthrombotic syndrome against the risks and costs associated with thrombolysis and thrombectomy, such as bleeding complications. In the final analysis, this remains a subjective decision.

With these caveats in mind, the best candidate for early thrombus removal is a young patient with iliofemoral DVT with symptoms lasting fewer than 14 days.

POSTTHROMBOTIC SYNDROME IS COMMON

Anticoagulation with heparin and warfarin is the mainstay of DVT therapy. Indeed, the safety of this therapy and its effectiveness in reducing thrombus propagation and DVT recurrence are well established. Neither heparin nor warfarin, however, actively reduces the thrombus burden. Rather, both prevent the clot from propagating while it is, hopefully, gradually reabsorbed through endogenous mechanisms.

Up to 50% of DVT patients develop postthrombotic syndrome. A variety of mechanisms are involved, including persistent obstructive thrombosis and valvular injury.¹ But much remains unknown about the etiology, and some patients develop the condition in the absence of abnormalities on objective testing.

Symptoms of postthrombotic syndrome can range from mild heaviness, edema, erythema, and cramping in the affected limb to debilitating pain with classic signs of venous hypertension (eg, venous ectasia and ulcers). It accounts for significant health care costs and has a detrimental effect on quality of life.¹ Thus, there has been interest in early thrombus removal as initial therapy for DVT.

THROMBUS REMOVAL

Venous clots can be removed with open surgery or, more typically, with percutaneous catheter-based thrombolysis and thrombectomy devices that use high-velocity saline jets, ultrasonic energy, or wire oscillation to mechanically fragment the venous clot. All of these mechanisms help with drug delivery and pose a minimal risk of pulmonary embolism.

Evidence is weak

Patients with DVT of the iliac venous system or common femoral vein are at highest risk of postthrombotic syndrome. Therefore, the Society for Vascular Surgery and the American Venous Forum have issued a grade 2C (ie, weak) recommendation in favor of early thrombus removal in patients with a first-time episode of iliofemoral DVT with fewer than 14 days of symptoms.² Moreover, patients must have a low risk of bleeding complications, be ambulatory, and have reasonable life expectancy.

The recommendation is buttressed by a Cochrane meta-analysis that included 101 patients.³ It concluded that there was a significant decrement in the development of postthrombotic syndrome with thrombolysis (but without mechanical thrombectomy) compared with standard therapy: the rate was 48% (29/61) with thrombolysis, and 65% (26/40) with standard therapy.³

More recently, the Catheter-Directed Thrombolysis Versus Standard Treatment for Acute Iliofemoral Deep Vein Thrombosis (CaVenT) study, a randomized prospective trial in 189 patients, demonstrated a lower rate of postthrombotic syndrome at 24 months and increased iliofemoral patency at 6 months with catheter-directed thrombolysis with alteplase (41.1% and 65.9%) vs anticoagulation with heparin and warfarin alone (55.6% and 47.4%).⁴

The Acute Venous Thrombosis: Thrombus Removal With Adjunctive Catheter-directed Thrombolysis (ATTRACT) trial is an ongoing prospective randomized multicenter trial of the effect of thrombolysis on postthrombotic syndrome that also hopes to clarify the relative benefits of different methods of pharmacomechanical clot removal.

While CaVenT has not been criticized extensively in the literature, other studies supporting early intervention for iliofemoral venous thrombosis generally have been noted to have a number of shortcomings, including a lack of randomization, and consequent bias, and the use of surrogate end points instead of a direct assessment of postthrombotic syndrome.

Reflecting the weakness of the evidence, the American College of Chest Physicians has issued a grade 2C recommendation against catheter-directed thrombolysis and against thrombectomy in favor of anticoagulant therapy.⁵

A subjective, case-by-case decision

The decision on standard vs interventional therapy must be made case by case. For example, thrombus removal may be more appropriate for a physically active young patient who is more likely to be impaired by postthrombotic syndrome, whereas standard warfarin therapy may be preferable for a sedentary patient. We are also more inclined to offer thrombus removal to patients who have worse symptoms.

Complicating the issue, many patients present with a mix of variables that support and oppose intervention—eg, a moderately active elderly patient with an unclear life expectancy and a history of gastrointestinal bleeding. At present, there is no way to quantitatively evaluate the risks and rewards of thrombus removal, and the final decision is essentially subjective.

Additional facts warranting consideration include the possibility that thrombolysis may require several days of therapy with daily venography for evaluation. Monitoring in the intensive care unit is normally required during the period of thrombolysis. Patients should be apprised of these elements of therapy beforehand; obviously, those who are unwilling to comply are not candidates.

Not a substitute for anticoagulation

It is important to recognize that thrombus removal is not a substitute for standard heparin-warfarin anticoagulation, which must also be prescribed.⁵ Thus, patients who cannot tolerate standard post-DVT anticoagulation should not undergo thrombus removal. Furthermore, the current evidence supports the use of standard anticoagulation over early thrombus removal of DVTs that are more distal in the lower extremity, such as those in the popliteal vein.⁵

PHLEGMASIA CERULEA DOLENS IS A SPECIAL CASE

Phlegmasia cerulea dolens—acute venous outflow obstruction associated with edema, cyanosis, and pain that in the worst cases may lead to shock, limb loss, and death—constitutes a special case. Although we lack robust supporting evidence, phlegmasia is a commonly accepted indication for early thrombus removal as a means of limb salvage.^2,6

Early thrombus removal for lower-extremity deep venous thrombosis (DVT) is at present only modestly supported by evidence and so remains controversial. It is largely aimed at preventing postthrombotic syndrome.

The decision to pursue early thrombus removal demands weighing the patient’s risk of postthrombotic syndrome against the risks and costs associated with thrombolysis and thrombectomy, such as bleeding complications. In the final analysis, this remains a subjective decision.

With these caveats in mind, the best candidate for early thrombus removal is a young patient with iliofemoral DVT with symptoms lasting fewer than 14 days.

POSTTHROMBOTIC SYNDROME IS COMMON

Anticoagulation with heparin and warfarin is the mainstay of DVT therapy. Indeed, the safety of this therapy and its effectiveness in reducing thrombus propagation and DVT recurrence are well established. Neither heparin nor warfarin, however, actively reduces the thrombus burden. Rather, both prevent the clot from propagating while it is, hopefully, gradually reabsorbed through endogenous mechanisms.

Up to 50% of DVT patients develop postthrombotic syndrome. A variety of mechanisms are involved, including persistent obstructive thrombosis and valvular injury.¹ But much remains unknown about the etiology, and some patients develop the condition in the absence of abnormalities on objective testing.

Symptoms of postthrombotic syndrome can range from mild heaviness, edema, erythema, and cramping in the affected limb to debilitating pain with classic signs of venous hypertension (eg, venous ectasia and ulcers). It accounts for significant health care costs and has a detrimental effect on quality of life.¹ Thus, there has been interest in early thrombus removal as initial therapy for DVT.

THROMBUS REMOVAL

Venous clots can be removed with open surgery or, more typically, with percutaneous catheter-based thrombolysis and thrombectomy devices that use high-velocity saline jets, ultrasonic energy, or wire oscillation to mechanically fragment the venous clot. All of these mechanisms help with drug delivery and pose a minimal risk of pulmonary embolism.

Evidence is weak

Patients with DVT of the iliac venous system or common femoral vein are at highest risk of postthrombotic syndrome. Therefore, the Society for Vascular Surgery and the American Venous Forum have issued a grade 2C (ie, weak) recommendation in favor of early thrombus removal in patients with a first-time episode of iliofemoral DVT with fewer than 14 days of symptoms.² Moreover, patients must have a low risk of bleeding complications, be ambulatory, and have reasonable life expectancy.

The recommendation is buttressed by a Cochrane meta-analysis that included 101 patients.³ It concluded that there was a significant decrement in the development of postthrombotic syndrome with thrombolysis (but without mechanical thrombectomy) compared with standard therapy: the rate was 48% (29/61) with thrombolysis, and 65% (26/40) with standard therapy.³

More recently, the Catheter-Directed Thrombolysis Versus Standard Treatment for Acute Iliofemoral Deep Vein Thrombosis (CaVenT) study, a randomized prospective trial in 189 patients, demonstrated a lower rate of postthrombotic syndrome at 24 months and increased iliofemoral patency at 6 months with catheter-directed thrombolysis with alteplase (41.1% and 65.9%) vs anticoagulation with heparin and warfarin alone (55.6% and 47.4%).⁴

The Acute Venous Thrombosis: Thrombus Removal With Adjunctive Catheter-directed Thrombolysis (ATTRACT) trial is an ongoing prospective randomized multicenter trial of the effect of thrombolysis on postthrombotic syndrome that also hopes to clarify the relative benefits of different methods of pharmacomechanical clot removal.

While CaVenT has not been criticized extensively in the literature, other studies supporting early intervention for iliofemoral venous thrombosis generally have been noted to have a number of shortcomings, including a lack of randomization, and consequent bias, and the use of surrogate end points instead of a direct assessment of postthrombotic syndrome.

Reflecting the weakness of the evidence, the American College of Chest Physicians has issued a grade 2C recommendation against catheter-directed thrombolysis and against thrombectomy in favor of anticoagulant therapy.⁵

A subjective, case-by-case decision

The decision on standard vs interventional therapy must be made case by case. For example, thrombus removal may be more appropriate for a physically active young patient who is more likely to be impaired by postthrombotic syndrome, whereas standard warfarin therapy may be preferable for a sedentary patient. We are also more inclined to offer thrombus removal to patients who have worse symptoms.

Complicating the issue, many patients present with a mix of variables that support and oppose intervention—eg, a moderately active elderly patient with an unclear life expectancy and a history of gastrointestinal bleeding. At present, there is no way to quantitatively evaluate the risks and rewards of thrombus removal, and the final decision is essentially subjective.

Additional facts warranting consideration include the possibility that thrombolysis may require several days of therapy with daily venography for evaluation. Monitoring in the intensive care unit is normally required during the period of thrombolysis. Patients should be apprised of these elements of therapy beforehand; obviously, those who are unwilling to comply are not candidates.

Not a substitute for anticoagulation

It is important to recognize that thrombus removal is not a substitute for standard heparin-warfarin anticoagulation, which must also be prescribed.⁵ Thus, patients who cannot tolerate standard post-DVT anticoagulation should not undergo thrombus removal. Furthermore, the current evidence supports the use of standard anticoagulation over early thrombus removal of DVTs that are more distal in the lower extremity, such as those in the popliteal vein.⁵

PHLEGMASIA CERULEA DOLENS IS A SPECIAL CASE

Phlegmasia cerulea dolens—acute venous outflow obstruction associated with edema, cyanosis, and pain that in the worst cases may lead to shock, limb loss, and death—constitutes a special case. Although we lack robust supporting evidence, phlegmasia is a commonly accepted indication for early thrombus removal as a means of limb salvage.^2,6

A March 2013 warning by the US Food and Drug Administration that azithromycin (Zithromax, Zmax, Z-pak) may increase the risk of sudden cardiac death does not mean we must abandon using it. We should, however, try to determine if our patients have cardiovascular risk factors for this extreme side effect and take appropriate precautions.

AZITHROMYCIN: THE SAFEST OF THE MACROLIDES?

Azithromycin, a broad-spectrum macrolide antibiotic, is used to treat or prevent a range of common bacterial infections, including upper and lower respiratory tract infections and certain sexually transmitted diseases.

In terms of overall toxicity, azithromycin has been considered the safest of the macrolides, as it neither undergoes CYP3A4 metabolism nor inhibits CYP3A4 to any clinically meaningful degree, and therefore does not interfere with the array of commonly used medications that undergo CYP3A4 metabolism.

Also, in vitro, azithromycin shows only limited blockade of the potassium channel hERG. This channel is critically involved in cardiomyocyte repolarization, and if it is blocked or otherwise malfunctioning, the result can be a prolonged QT interval, ventricular arrhythmias, and even sudden cardiac death.^1–4 Therefore, lack of blockade, as reflected by a high inhibitory concentration (Table 1), boded well for the safety of azithromycin in terms of QT liability. However, we should be cautious in interpreting in vitro data.

With its broad antibiotic spectrum and perceived favorable safety profile, azithromycin has become one of the top 15 most prescribed drugs and the best-selling antibiotic in the United States, accounting for 55.4 million prescriptions in 2012, according to the IMS Institute for Healthcare Informatics.

THE FDA RECEIVES 203 REPORTS OF ADVERSE EVENTS IN 8 YEARS

However, beginning with a report of azithromycin-triggered torsades de pointes in 2001,⁵ a growing body of evidence, derived from postmarketing surveillance, has linked azithromycin to cardiac arrhythmias such as pronounced QT interval prolongation and associated torsades de pointes (which can progress to life-threatening ventricular fibrillation). Other, closely related macrolides such as clarithromycin and erythromycin are also linked to these effects.

Furthermore, in the 8-year period from 2004 to 2011, the US Food and Drug Administration (FDA) Adverse Event Reporting System (FAERS) received a total of 203 reports of azithromycin-associated QT prolongation, torsades de pointes, ventricular arrhythmia, or, in 65 cases, sudden cardiac death (Table 1).⁶

At face value, the number of FAERS reports appears to be similar between the various macrolide antibiotics. However, it is important to remember that these drugs differ substantially in the number of prescriptions written for them, with azithromycin being prescribed more often. Also, the FAERS numbers are subject to a number of well-known limitations such as confounding variables, uneven quality and completeness of reports, duplication, and underreporting. These limitations preclude the use of such adverse reporting databases in calculating and thereby comparing the true incidence of adverse events associated with the various macrolide antibiotics.^6–9

RAY ET AL FIND A HIGHER RISK OF CARDIOVASCULAR DEATH

Despite these inherent flaws, initial postmarketing surveillance reports cast enough doubt on the long-standing notion that azithromycin is the safest macrolide antibiotic to prompt Ray et al¹⁰ to assess its safety in an observational, nonrandomized study of people enrolled in the Tennessee Medicaid program.

They found that, over the typical 5 days of therapy, people taking azithromycin had a rate of cardiovascular death 2.88 times higher than in people taking no antibiotic, and 2.49 times higher than in people taking amoxicillin (Table 2).

However, the absolute excess risk compared with amoxicillin varied considerably according to baseline risk score for cardiovascular disease, with 1 excess cardiovascular death per 4,100 in the highest-risk decile compared with 1 excess cardiovascular death per 100,000 in the lowest-risk decile.^10,11

Moreover, the increase in deaths did not persist after the 5 days of therapy. This time-limited pattern directly correlated with expected peak azithromycin plasma levels during a standard 5-day course.

Ray et al used appropriate analytic methods to attempt to correct for any confounding bias intrinsic to the observational, nonrandomized study design. Nevertheless, the patients were Medicaid beneficiaries, who have a higher prevalence of comorbid conditions and higher mortality rates than the general population. Therefore, legitimate questions were raised about whether the results of the study could be generalized to populations with substantially lower baseline risk of cardiovascular disease and if differences in the baseline characteristics of the treatment groups were adequately controlled.^12,13

THE FDA REVISES AZITHROMYCIN’S WARNINGS AND PRECAUTIONS

The striking observations by Ray et al,¹⁰ coupled with the concerns raised by postmarketing surveillance reports, compelled the FDA to review the labels of azithromycin and other macrolide antibiotics.

Ultimately, the FDA opted to revise the “warning and precautions” section of the azithromycin drug label to include a warning about the potential risk of fatal arrhythmias, specifically QT interval prolongation and torsades de pointes. In a March 2013 safety announcement, it also urged health care professionals to use caution when prescribing azithromycin to patients known to have risk factors for drug-related arrhythmias, including congenital long QT syndrome, acquired QT interval prolongation, hypokalemia, hypomagnesia, bradycardia, and concurrent use of other medications known to prolong the QT interval, specifically the class IA (eg, quinidine and procainamide) and class III (eg, amiodarone, sotalol, and dofetilide) antiarrhythmics.

SVANSTRÖM ET AL FIND NO INCREASED RISK

However, just when the medical community appeared ready to accept that azithromycin may not be as safe as we thought it was, a large prospective study by Svanström et al, published in early May 2013, found no increased risk of cardiovascular death associated with azithromycin (Table 2).¹⁴

The patients were a representative population of young to middle-aged Danish adults at low baseline risk of underlying cardiovascular disease.

Interestingly, Svanström et al were careful to point out that their study was only powered to rule out a moderate-to-high (> 55%) increase in the relative risk of cardiovascular death. Furthermore, profound differences existed in the baseline risk of death and cardiovascular risk factors between their patients and the Tennessee Medicaid patients studied by Ray et al.¹⁴ Therefore, the authors suggested that their study complements rather than contradicts the study by Ray et al. They attributed the differences in the findings to treatment-effect heterogeneity, in which the risk of azithromycin-associated cardiovascular mortality is largely limited to high-risk patients, namely those with multiple preexisting cardiovascular risk factors.¹⁴

ACC/AHA RECOMMENDATION: IDENTIFY THOSE AT RISK

Collectively, the data reviewed above provide compelling evidence that azithromycin is not completely free of the QT-prolonging and torsadogenic effects that have long been associated with other macrolide antibiotics. However, the findings from both the study by Ray et al and that of Svanström et al suggest that preexisting cardiovascular risk factors play a prominent role in determining the incidence of azithromycin-associated cardiovascular death in a given population (Table 2).^10,14

These findings should prompt physicians to carefully reassess the risks and benefits of azithromycin use in their clinical practices. They also reinforce a recent call by the American Heart Association (AHA) and American College of Cardiology (ACC) to better identify, early on, patients at risk of drug-induced ventricular arrhythmias and sudden death and to subsequently improve how these patients are monitored when the use of QT-prolonging and torsadogenic drugs is medically necessary.¹⁵

AN ELECTRONIC MEDICAL RECORD FLAGS QTc ≥ 500 MS

On the heels of these AHA/ACC suggestions, our hospital has adopted an institution-wide QT alert system. Here, the electronic medical record system (Centricity EMR; GE Healthcare) uses a proprietary algorithm to detect and electronically alert ordering physicians when a patient has a prolonged QT interval, and gives information about the potential clinical significance of this electrocardiographic finding.¹⁶ Physicians also receive a warning when ordering QT-prolonging drugs in patients at risk.

This system is still in its infancy, but it has already confirmed that a prolonged QT interval (QTc ≥ 500 ms) is a powerful predictor of death from any cause and has demonstrated that mortality rates in those with prolonged QT intervals increase in a dose-dependent fashion with the patient’s number of modifiable risk factors (eg, electrolyte disturbances or QT-prolonging medications) and nonmodifiable risk factors (eg, genetic disposition, female sex, structural heart disease, diabetes mellitus).¹⁶ We have also found evidence that modifiable risk factors may have a more pronounced effect on mortality risk than non-modifiable risk factors.¹⁶

These findings suggest that information technology-based QT alert systems may one day provide physicians with an important tool to efficiently identify and possibly even modify the risk of cardiovascular death in patients at high risk, for example, by correcting electrolyte abnormalities or reducing the burden of QT-prolonging medications.

CONSIDER RISK OF QT PROLONGATION WHEN PRESCRIBING AZITHROMYCIN

For most institutions and clinical practices, such electronic QT alert systems are still years if not decades away. However, in light of the information summarized above, all physicians should begin considering risk factors for QT prolongation and torsades de pointes (summarized in Table 3) and weighing the risks and benefits of prescribing azithromycin vs alternative antibiotics with minimal QT liability. This should be relatively simple to do. Things to keep in mind:

• Although azithromycin may increase the relative risk of a cardiovascular event, for most otherwise-healthy patients, the absolute risk is miniscule.
• In a patient at risk (eg, with baseline QT prolongation or multiple risk factors for it), if azithromycin or another QT-prolonging antibiotic such as a macrolide or fluoroquinolone is medically necessary due to preferential bacterial susceptibility or patient allergies, every effort should be made to correct modifiable risk factors (eg, electrolyte abnormalities) and, if possible, to avoid polypharmacy with multiple QT-prolonging drugs.
• For patients who have multiple risk factors for QT prolongation in whom treatment with a known QT-prolonging medication is still deemed in the patient’s best interest, strong consideration should be given to inpatient administration and monitoring until the treatment has been completed.

With careful consideration of modifiable and nonmodifiable risk factors as well as a little extra caution when prescribing potential QT-prolonging medications such as azithromycin, the clinical benefit of these often-advantageous medications can be maximized and the incidence of these tragic but rare drug-induced sudden cardiac deaths can be reduced.

A March 2013 warning by the US Food and Drug Administration that azithromycin (Zithromax, Zmax, Z-pak) may increase the risk of sudden cardiac death does not mean we must abandon using it. We should, however, try to determine if our patients have cardiovascular risk factors for this extreme side effect and take appropriate precautions.

AZITHROMYCIN: THE SAFEST OF THE MACROLIDES?

Azithromycin, a broad-spectrum macrolide antibiotic, is used to treat or prevent a range of common bacterial infections, including upper and lower respiratory tract infections and certain sexually transmitted diseases.

In terms of overall toxicity, azithromycin has been considered the safest of the macrolides, as it neither undergoes CYP3A4 metabolism nor inhibits CYP3A4 to any clinically meaningful degree, and therefore does not interfere with the array of commonly used medications that undergo CYP3A4 metabolism.

Also, in vitro, azithromycin shows only limited blockade of the potassium channel hERG. This channel is critically involved in cardiomyocyte repolarization, and if it is blocked or otherwise malfunctioning, the result can be a prolonged QT interval, ventricular arrhythmias, and even sudden cardiac death.^1–4 Therefore, lack of blockade, as reflected by a high inhibitory concentration (Table 1), boded well for the safety of azithromycin in terms of QT liability. However, we should be cautious in interpreting in vitro data.

With its broad antibiotic spectrum and perceived favorable safety profile, azithromycin has become one of the top 15 most prescribed drugs and the best-selling antibiotic in the United States, accounting for 55.4 million prescriptions in 2012, according to the IMS Institute for Healthcare Informatics.

THE FDA RECEIVES 203 REPORTS OF ADVERSE EVENTS IN 8 YEARS

However, beginning with a report of azithromycin-triggered torsades de pointes in 2001,⁵ a growing body of evidence, derived from postmarketing surveillance, has linked azithromycin to cardiac arrhythmias such as pronounced QT interval prolongation and associated torsades de pointes (which can progress to life-threatening ventricular fibrillation). Other, closely related macrolides such as clarithromycin and erythromycin are also linked to these effects.

Furthermore, in the 8-year period from 2004 to 2011, the US Food and Drug Administration (FDA) Adverse Event Reporting System (FAERS) received a total of 203 reports of azithromycin-associated QT prolongation, torsades de pointes, ventricular arrhythmia, or, in 65 cases, sudden cardiac death (Table 1).⁶

At face value, the number of FAERS reports appears to be similar between the various macrolide antibiotics. However, it is important to remember that these drugs differ substantially in the number of prescriptions written for them, with azithromycin being prescribed more often. Also, the FAERS numbers are subject to a number of well-known limitations such as confounding variables, uneven quality and completeness of reports, duplication, and underreporting. These limitations preclude the use of such adverse reporting databases in calculating and thereby comparing the true incidence of adverse events associated with the various macrolide antibiotics.^6–9

RAY ET AL FIND A HIGHER RISK OF CARDIOVASCULAR DEATH

Despite these inherent flaws, initial postmarketing surveillance reports cast enough doubt on the long-standing notion that azithromycin is the safest macrolide antibiotic to prompt Ray et al¹⁰ to assess its safety in an observational, nonrandomized study of people enrolled in the Tennessee Medicaid program.

They found that, over the typical 5 days of therapy, people taking azithromycin had a rate of cardiovascular death 2.88 times higher than in people taking no antibiotic, and 2.49 times higher than in people taking amoxicillin (Table 2).

However, the absolute excess risk compared with amoxicillin varied considerably according to baseline risk score for cardiovascular disease, with 1 excess cardiovascular death per 4,100 in the highest-risk decile compared with 1 excess cardiovascular death per 100,000 in the lowest-risk decile.^10,11

Moreover, the increase in deaths did not persist after the 5 days of therapy. This time-limited pattern directly correlated with expected peak azithromycin plasma levels during a standard 5-day course.

Ray et al used appropriate analytic methods to attempt to correct for any confounding bias intrinsic to the observational, nonrandomized study design. Nevertheless, the patients were Medicaid beneficiaries, who have a higher prevalence of comorbid conditions and higher mortality rates than the general population. Therefore, legitimate questions were raised about whether the results of the study could be generalized to populations with substantially lower baseline risk of cardiovascular disease and if differences in the baseline characteristics of the treatment groups were adequately controlled.^12,13

THE FDA REVISES AZITHROMYCIN’S WARNINGS AND PRECAUTIONS

The striking observations by Ray et al,¹⁰ coupled with the concerns raised by postmarketing surveillance reports, compelled the FDA to review the labels of azithromycin and other macrolide antibiotics.

Ultimately, the FDA opted to revise the “warning and precautions” section of the azithromycin drug label to include a warning about the potential risk of fatal arrhythmias, specifically QT interval prolongation and torsades de pointes. In a March 2013 safety announcement, it also urged health care professionals to use caution when prescribing azithromycin to patients known to have risk factors for drug-related arrhythmias, including congenital long QT syndrome, acquired QT interval prolongation, hypokalemia, hypomagnesia, bradycardia, and concurrent use of other medications known to prolong the QT interval, specifically the class IA (eg, quinidine and procainamide) and class III (eg, amiodarone, sotalol, and dofetilide) antiarrhythmics.

SVANSTRÖM ET AL FIND NO INCREASED RISK

However, just when the medical community appeared ready to accept that azithromycin may not be as safe as we thought it was, a large prospective study by Svanström et al, published in early May 2013, found no increased risk of cardiovascular death associated with azithromycin (Table 2).¹⁴

The patients were a representative population of young to middle-aged Danish adults at low baseline risk of underlying cardiovascular disease.

Interestingly, Svanström et al were careful to point out that their study was only powered to rule out a moderate-to-high (> 55%) increase in the relative risk of cardiovascular death. Furthermore, profound differences existed in the baseline risk of death and cardiovascular risk factors between their patients and the Tennessee Medicaid patients studied by Ray et al.¹⁴ Therefore, the authors suggested that their study complements rather than contradicts the study by Ray et al. They attributed the differences in the findings to treatment-effect heterogeneity, in which the risk of azithromycin-associated cardiovascular mortality is largely limited to high-risk patients, namely those with multiple preexisting cardiovascular risk factors.¹⁴

ACC/AHA RECOMMENDATION: IDENTIFY THOSE AT RISK

Collectively, the data reviewed above provide compelling evidence that azithromycin is not completely free of the QT-prolonging and torsadogenic effects that have long been associated with other macrolide antibiotics. However, the findings from both the study by Ray et al and that of Svanström et al suggest that preexisting cardiovascular risk factors play a prominent role in determining the incidence of azithromycin-associated cardiovascular death in a given population (Table 2).^10,14

These findings should prompt physicians to carefully reassess the risks and benefits of azithromycin use in their clinical practices. They also reinforce a recent call by the American Heart Association (AHA) and American College of Cardiology (ACC) to better identify, early on, patients at risk of drug-induced ventricular arrhythmias and sudden death and to subsequently improve how these patients are monitored when the use of QT-prolonging and torsadogenic drugs is medically necessary.¹⁵

AN ELECTRONIC MEDICAL RECORD FLAGS QTc ≥ 500 MS

On the heels of these AHA/ACC suggestions, our hospital has adopted an institution-wide QT alert system. Here, the electronic medical record system (Centricity EMR; GE Healthcare) uses a proprietary algorithm to detect and electronically alert ordering physicians when a patient has a prolonged QT interval, and gives information about the potential clinical significance of this electrocardiographic finding.¹⁶ Physicians also receive a warning when ordering QT-prolonging drugs in patients at risk.

This system is still in its infancy, but it has already confirmed that a prolonged QT interval (QTc ≥ 500 ms) is a powerful predictor of death from any cause and has demonstrated that mortality rates in those with prolonged QT intervals increase in a dose-dependent fashion with the patient’s number of modifiable risk factors (eg, electrolyte disturbances or QT-prolonging medications) and nonmodifiable risk factors (eg, genetic disposition, female sex, structural heart disease, diabetes mellitus).¹⁶ We have also found evidence that modifiable risk factors may have a more pronounced effect on mortality risk than non-modifiable risk factors.¹⁶

These findings suggest that information technology-based QT alert systems may one day provide physicians with an important tool to efficiently identify and possibly even modify the risk of cardiovascular death in patients at high risk, for example, by correcting electrolyte abnormalities or reducing the burden of QT-prolonging medications.

CONSIDER RISK OF QT PROLONGATION WHEN PRESCRIBING AZITHROMYCIN

For most institutions and clinical practices, such electronic QT alert systems are still years if not decades away. However, in light of the information summarized above, all physicians should begin considering risk factors for QT prolongation and torsades de pointes (summarized in Table 3) and weighing the risks and benefits of prescribing azithromycin vs alternative antibiotics with minimal QT liability. This should be relatively simple to do. Things to keep in mind:

• Although azithromycin may increase the relative risk of a cardiovascular event, for most otherwise-healthy patients, the absolute risk is miniscule.
• In a patient at risk (eg, with baseline QT prolongation or multiple risk factors for it), if azithromycin or another QT-prolonging antibiotic such as a macrolide or fluoroquinolone is medically necessary due to preferential bacterial susceptibility or patient allergies, every effort should be made to correct modifiable risk factors (eg, electrolyte abnormalities) and, if possible, to avoid polypharmacy with multiple QT-prolonging drugs.
• For patients who have multiple risk factors for QT prolongation in whom treatment with a known QT-prolonging medication is still deemed in the patient’s best interest, strong consideration should be given to inpatient administration and monitoring until the treatment has been completed.

With careful consideration of modifiable and nonmodifiable risk factors as well as a little extra caution when prescribing potential QT-prolonging medications such as azithromycin, the clinical benefit of these often-advantageous medications can be maximized and the incidence of these tragic but rare drug-induced sudden cardiac deaths can be reduced.

A March 2013 warning by the US Food and Drug Administration that azithromycin (Zithromax, Zmax, Z-pak) may increase the risk of sudden cardiac death does not mean we must abandon using it. We should, however, try to determine if our patients have cardiovascular risk factors for this extreme side effect and take appropriate precautions.

AZITHROMYCIN: THE SAFEST OF THE MACROLIDES?

Azithromycin, a broad-spectrum macrolide antibiotic, is used to treat or prevent a range of common bacterial infections, including upper and lower respiratory tract infections and certain sexually transmitted diseases.

In terms of overall toxicity, azithromycin has been considered the safest of the macrolides, as it neither undergoes CYP3A4 metabolism nor inhibits CYP3A4 to any clinically meaningful degree, and therefore does not interfere with the array of commonly used medications that undergo CYP3A4 metabolism.

Also, in vitro, azithromycin shows only limited blockade of the potassium channel hERG. This channel is critically involved in cardiomyocyte repolarization, and if it is blocked or otherwise malfunctioning, the result can be a prolonged QT interval, ventricular arrhythmias, and even sudden cardiac death.^1–4 Therefore, lack of blockade, as reflected by a high inhibitory concentration (Table 1), boded well for the safety of azithromycin in terms of QT liability. However, we should be cautious in interpreting in vitro data.

With its broad antibiotic spectrum and perceived favorable safety profile, azithromycin has become one of the top 15 most prescribed drugs and the best-selling antibiotic in the United States, accounting for 55.4 million prescriptions in 2012, according to the IMS Institute for Healthcare Informatics.

THE FDA RECEIVES 203 REPORTS OF ADVERSE EVENTS IN 8 YEARS

However, beginning with a report of azithromycin-triggered torsades de pointes in 2001,⁵ a growing body of evidence, derived from postmarketing surveillance, has linked azithromycin to cardiac arrhythmias such as pronounced QT interval prolongation and associated torsades de pointes (which can progress to life-threatening ventricular fibrillation). Other, closely related macrolides such as clarithromycin and erythromycin are also linked to these effects.

Furthermore, in the 8-year period from 2004 to 2011, the US Food and Drug Administration (FDA) Adverse Event Reporting System (FAERS) received a total of 203 reports of azithromycin-associated QT prolongation, torsades de pointes, ventricular arrhythmia, or, in 65 cases, sudden cardiac death (Table 1).⁶

At face value, the number of FAERS reports appears to be similar between the various macrolide antibiotics. However, it is important to remember that these drugs differ substantially in the number of prescriptions written for them, with azithromycin being prescribed more often. Also, the FAERS numbers are subject to a number of well-known limitations such as confounding variables, uneven quality and completeness of reports, duplication, and underreporting. These limitations preclude the use of such adverse reporting databases in calculating and thereby comparing the true incidence of adverse events associated with the various macrolide antibiotics.^6–9

RAY ET AL FIND A HIGHER RISK OF CARDIOVASCULAR DEATH

Despite these inherent flaws, initial postmarketing surveillance reports cast enough doubt on the long-standing notion that azithromycin is the safest macrolide antibiotic to prompt Ray et al¹⁰ to assess its safety in an observational, nonrandomized study of people enrolled in the Tennessee Medicaid program.

They found that, over the typical 5 days of therapy, people taking azithromycin had a rate of cardiovascular death 2.88 times higher than in people taking no antibiotic, and 2.49 times higher than in people taking amoxicillin (Table 2).

However, the absolute excess risk compared with amoxicillin varied considerably according to baseline risk score for cardiovascular disease, with 1 excess cardiovascular death per 4,100 in the highest-risk decile compared with 1 excess cardiovascular death per 100,000 in the lowest-risk decile.^10,11

Moreover, the increase in deaths did not persist after the 5 days of therapy. This time-limited pattern directly correlated with expected peak azithromycin plasma levels during a standard 5-day course.

Ray et al used appropriate analytic methods to attempt to correct for any confounding bias intrinsic to the observational, nonrandomized study design. Nevertheless, the patients were Medicaid beneficiaries, who have a higher prevalence of comorbid conditions and higher mortality rates than the general population. Therefore, legitimate questions were raised about whether the results of the study could be generalized to populations with substantially lower baseline risk of cardiovascular disease and if differences in the baseline characteristics of the treatment groups were adequately controlled.^12,13

THE FDA REVISES AZITHROMYCIN’S WARNINGS AND PRECAUTIONS

The striking observations by Ray et al,¹⁰ coupled with the concerns raised by postmarketing surveillance reports, compelled the FDA to review the labels of azithromycin and other macrolide antibiotics.

Ultimately, the FDA opted to revise the “warning and precautions” section of the azithromycin drug label to include a warning about the potential risk of fatal arrhythmias, specifically QT interval prolongation and torsades de pointes. In a March 2013 safety announcement, it also urged health care professionals to use caution when prescribing azithromycin to patients known to have risk factors for drug-related arrhythmias, including congenital long QT syndrome, acquired QT interval prolongation, hypokalemia, hypomagnesia, bradycardia, and concurrent use of other medications known to prolong the QT interval, specifically the class IA (eg, quinidine and procainamide) and class III (eg, amiodarone, sotalol, and dofetilide) antiarrhythmics.

SVANSTRÖM ET AL FIND NO INCREASED RISK

However, just when the medical community appeared ready to accept that azithromycin may not be as safe as we thought it was, a large prospective study by Svanström et al, published in early May 2013, found no increased risk of cardiovascular death associated with azithromycin (Table 2).¹⁴

The patients were a representative population of young to middle-aged Danish adults at low baseline risk of underlying cardiovascular disease.

Interestingly, Svanström et al were careful to point out that their study was only powered to rule out a moderate-to-high (> 55%) increase in the relative risk of cardiovascular death. Furthermore, profound differences existed in the baseline risk of death and cardiovascular risk factors between their patients and the Tennessee Medicaid patients studied by Ray et al.¹⁴ Therefore, the authors suggested that their study complements rather than contradicts the study by Ray et al. They attributed the differences in the findings to treatment-effect heterogeneity, in which the risk of azithromycin-associated cardiovascular mortality is largely limited to high-risk patients, namely those with multiple preexisting cardiovascular risk factors.¹⁴

ACC/AHA RECOMMENDATION: IDENTIFY THOSE AT RISK

Collectively, the data reviewed above provide compelling evidence that azithromycin is not completely free of the QT-prolonging and torsadogenic effects that have long been associated with other macrolide antibiotics. However, the findings from both the study by Ray et al and that of Svanström et al suggest that preexisting cardiovascular risk factors play a prominent role in determining the incidence of azithromycin-associated cardiovascular death in a given population (Table 2).^10,14

These findings should prompt physicians to carefully reassess the risks and benefits of azithromycin use in their clinical practices. They also reinforce a recent call by the American Heart Association (AHA) and American College of Cardiology (ACC) to better identify, early on, patients at risk of drug-induced ventricular arrhythmias and sudden death and to subsequently improve how these patients are monitored when the use of QT-prolonging and torsadogenic drugs is medically necessary.¹⁵

AN ELECTRONIC MEDICAL RECORD FLAGS QTc ≥ 500 MS

On the heels of these AHA/ACC suggestions, our hospital has adopted an institution-wide QT alert system. Here, the electronic medical record system (Centricity EMR; GE Healthcare) uses a proprietary algorithm to detect and electronically alert ordering physicians when a patient has a prolonged QT interval, and gives information about the potential clinical significance of this electrocardiographic finding.¹⁶ Physicians also receive a warning when ordering QT-prolonging drugs in patients at risk.

This system is still in its infancy, but it has already confirmed that a prolonged QT interval (QTc ≥ 500 ms) is a powerful predictor of death from any cause and has demonstrated that mortality rates in those with prolonged QT intervals increase in a dose-dependent fashion with the patient’s number of modifiable risk factors (eg, electrolyte disturbances or QT-prolonging medications) and nonmodifiable risk factors (eg, genetic disposition, female sex, structural heart disease, diabetes mellitus).¹⁶ We have also found evidence that modifiable risk factors may have a more pronounced effect on mortality risk than non-modifiable risk factors.¹⁶

These findings suggest that information technology-based QT alert systems may one day provide physicians with an important tool to efficiently identify and possibly even modify the risk of cardiovascular death in patients at high risk, for example, by correcting electrolyte abnormalities or reducing the burden of QT-prolonging medications.

CONSIDER RISK OF QT PROLONGATION WHEN PRESCRIBING AZITHROMYCIN

For most institutions and clinical practices, such electronic QT alert systems are still years if not decades away. However, in light of the information summarized above, all physicians should begin considering risk factors for QT prolongation and torsades de pointes (summarized in Table 3) and weighing the risks and benefits of prescribing azithromycin vs alternative antibiotics with minimal QT liability. This should be relatively simple to do. Things to keep in mind:

• Although azithromycin may increase the relative risk of a cardiovascular event, for most otherwise-healthy patients, the absolute risk is miniscule.
• In a patient at risk (eg, with baseline QT prolongation or multiple risk factors for it), if azithromycin or another QT-prolonging antibiotic such as a macrolide or fluoroquinolone is medically necessary due to preferential bacterial susceptibility or patient allergies, every effort should be made to correct modifiable risk factors (eg, electrolyte abnormalities) and, if possible, to avoid polypharmacy with multiple QT-prolonging drugs.
• For patients who have multiple risk factors for QT prolongation in whom treatment with a known QT-prolonging medication is still deemed in the patient’s best interest, strong consideration should be given to inpatient administration and monitoring until the treatment has been completed.

With careful consideration of modifiable and nonmodifiable risk factors as well as a little extra caution when prescribing potential QT-prolonging medications such as azithromycin, the clinical benefit of these often-advantageous medications can be maximized and the incidence of these tragic but rare drug-induced sudden cardiac deaths can be reduced.

“Change is the only constant.”

—Heraclitus (c 535–475 bce)

With the cost of health care rising and money to pay for it shrinking, there has never been a greater need to reduce waste.

See related article

Ineffective treatments and adverse drug effects account for much preventable morbidity and expense. New treatments, touted as more potent, are often introduced as replacements for traditional ones that are still effective in many patients, adding to costs and the potential for harm. For the pharmaceutical industry, the search for new “blockbuster” drugs seems to have hit a wall, at least in cardiovascular medicine.¹ Advances often come at the cost of adverse effects, such as bleeding with triple antiplatelet therapy and diabetes with potent statin drugs.

The path to maximizing benefit and reducing harm now appears to lie in stratifying populations and appreciating patient individuality in response to treatment. For many decades we have known that patients vary widely in their response to drugs, owing to personal factors such as body surface area, age, environment, and genetics. And indeed, we treat our patients as individuals, for example by tailoring aminoglycoside dose to weight and renal function.

However, clinical trials typically give us an idea of the benefits only to the average patient. While subgroup analyses identify groups that may benefit more or less from treatment, the additional information they provide is not easily integrated into the clinical model of prescribing, in which one size fits all.

THE PROMISE OF PHARMACOGENETICS

The emerging field of pharmacogenetics promises to give clinicians the tools to make informed treatment decisions based on predictive genetic testing. This genetic testing aims to match treatment to an individual’s genetic profile. This often involves analyzing single-nucleotide polymorphisms in genes for enzymes that metabolize drugs, such as the cytochrome P450 enzymes, to predict efficacy or an adverse event with treatment.

Pharmacogenetics is playing an increasing role in clinical trials, particularly in the early stages of drug development, by helping to reduce the number of patients needed, prove efficacy, and identify subgroups in which alternative treatment can be targeted. At another level, a molecular understanding of disease is leading to truly targeted treatments based on genomics.

Over recent years, genetic testing has been increasingly used in clinical practice, thanks to a convergence of factors such as rapid, low-cost tests, a growing evidence base, and emerging interest among doctors and payers.

An advantage to using genetic testing as opposed to other types of laboratory testing, such as measuring the concentration of the drug in the blood during treatment, is that genetic tests can predict the response to treatment before the treatment is started. Moreover, with therapeutic drug monitoring after treatment has begun, there are sometimes no detectable measures of toxicity. For example, both carbamazepine and the antiviral drug abacavir can—fortunately only rarely—cause Stevens-Johnson syndrome. But before genetic markers were discovered, there was no method of estimating this risk apart from taking a family history.^2,3 Considering the numbers of people involved, it was not feasible until recently to suggest genetic screening for patients starting on these drugs. However, the cost of genotyping and gene sequencing has been falling at a rate inversely faster than Moore’s law (an approximate annual doubling in computer power), and population genomics is becoming a reality.⁴

The US Food and Drug Administration (FDA) recognizes the current and future value of pharmacogenetics in drug safety and development. A number of approved pharmacogenetic biomarkers are listed on the FDA website (Table 1). Black box warnings have been mandated for a number of drugs on the basis of observational evidence.

The FDA also promotes rapid approval for novel drugs with pharmacogenetic “companion diagnostics.” A recent example of this was the approval of ivacaftor for cystic fibrosis patients who have the G551D mutation.⁵ Here, a molecular understanding of the condition led to the development of a targeted treatment. Although the cost of developing this drug was high, the path is now paved for similar advances. Oncologists are familiar with these advances with the emergence of new molecularly targeted treatments, eg, BRAF inhibitors in metastatic melanoma, imatinib in chronic myeloid leukemia, and gefitinib in non-small-cell lung cancer.

PHARMACOGENETICS IN CARDIOVASCULAR MEDICINE

Cardiovascular medicine also stands to benefit from rapid advances in pharmacogenetics.

While no treatment has been developed that targets the molecular basis of cardiovascular disease, a number of genomic biomarkers have emerged that identify patients at risk of adverse reactions or treatment failure. These include genetic tests to predict the maintenance dose and risk of bleeding with warfarin,⁶ the likelihood of myopathy and myositis with simvastatin,⁷ and the risk of recurrent thrombotic events with clopidogrel.^8–10

Using pharmacogenetics in prescribing warfarin and its alternatives

The pharmacogenetics of warfarin has been extensively researched, but genotyping before prescribing this drug is not yet widely done.

In 2007, the FDA updated the labeling of warfarin to include information about the influence of two genes, VKORC1 and CYP2C9, on a patient’s response to this drug. In 2010 this was updated to add that testing for these genes could be used to predict the maintenance dose of the drug. Difficulties with algorithms used to integrate this into clinical practice have hindered adoption of this testing.

With the advent of new anticoagulants such as dabigatran, rivaroxaban, and apixaban, many have expected warfarin and its pharmacogenetics to become obsolete. However, the new agents cost considerably more. Further, they may not offer a very great advantage over warfarin: in the Randomized Evaluation of Long-term Anticoagulant Therapy (RE-LY) trial, the absolute risk reduction in intracranial hemorrhage with dabigatran vs warfarin was small.¹¹ Therefore, dabigatran is probably not cost-effective in populations at low risk of bleeding.¹² A cost-effectiveness analysis comparing warfarin with dabigatran in patients with uncomplicated atrial fibrillation has suggested that dabigatran is, however, cost-effective in patients at moderate risk.¹²

In the RE-LY trial, the international normalized ratios (INRs) of the patients in the warfarin group were in the therapeutic range only 64% of the time. The advantages of dabigatran over warfarin become less pronounced as warfarin control is tightened.¹³ Of note, pharmacogenetics and home monitoring of the INR have both been shown to lead to tighter control of the INR, with greater time within the therapeutic range.^14,15

Moreover, genetic testing can help us reduce the number of bleeding events in patients taking warfarin.¹⁶ Patients who carry the CYP2C9*2 or CYP2C9*3 polymorphism metabolize S-warfarin slower and therefore have a threefold higher risk of hemorrhage after starting warfarin.¹⁷ We could speculate that patients carrying these variants may be better served by the newer anticoagulants, though this has not been tested in any clinical trial.

It is also worth appreciating that the conditions requiring anticoagulation, such as atrial fibrillation, also have a strong genetic basis. Variants in chromosomes 4q25, 1q21, and 16q22 have all been associated with atrial fibrillation.¹⁸ The risk of atrial fibrillation is five to six times higher in carriers of multiple variants within all of these loci.¹⁹ Genetic variants at 4q25 have been associated with the response to specific antiarrhythmic drug treatments,²⁰ response to pulmonary vein isolation, ^21,22 and direct-current cardioversion.²³ One can imagine a future in which patients with palpitations, carrying multiple gene risk variants, will choose prolonged monitoring at home to confirm a diagnosis. They would then be provided with a personalized best management strategy, using their personal preferences, clinical data, and genetic profile to make a treatment decision.

Using pharmacogenetics in prescribing clopidogrel and its alternatives

The pharmacogenetics of clopidogrel is of particular interest, as it has the potential of establishing a rational basis for using newer antiplatelet drugs such as ticagrelor and prasugrel, which are considerably more expensive than generic clopidogrel.

Most of the people who do not respond to clopidogrel carry the common cytochrome P450 2C19 variants CYP2C19*2 or CYP2C19*3.⁹ These variants are present in particularly high frequency in Asians and African Americans, who often do not feature in large randomized trials.

Newer antiplatelet agents have failed to demonstrate consistent superiority to clopidogrel without a tradeoff of more bleeding. However, in the Trial to Assess Improvement in Therapeutic Outcomes by Optimizing Platelet Inhibition With Prasugrel-Thrombolysis in Myocardial Infarction 39 (TRITON TIMI-38),^24,25 patients with the *2 variant receiving prasugrel had lower cardiovascular event rates than *2 carriers receiving clopidogrel.

Similarly, the patients who benefit the most from ticagrelor are carriers of the 2C19 nonresponder variants. In a large study, clopidogrel responders who did not carry either 2C19 nonresponder genetic variants or ABCB1 variants had cardiovascular outcomes similar to those of patients receiving ticagrelor.²⁶

Clinicians have been cautious in prescribing potent antiplatelet agents to all patients because of the risk of bleeding. One could assume that by reserving newer agents for clopidogrel nonresponders, the bleeding risk could be minimized and overall benefit could be preserved with this strategy.

Cost may also be contained. The cost-effectiveness of such an approach with prasugrel has been tested with computer modeling and appears favorable.²⁷ On the other hand, a similar yet limited analysis did not find genotype-driven use of ticagrelor to be cost-effective.²⁸ This was mostly due to fewer deaths in patients receiving ticagrelor. However, the cost estimate for genotype-guided therapy was overestimated, as heterozygotes in the model were treated with ticagrelor instead of a high dose of clopidogrel.

It now appears that heterozygotes, ie, patients with one copy of the nonresponder variant, can achieve similar platelet inhibition with clopidogrel 225 mg daily as noncarriers on 75 mg daily.²⁹ Since genotype-guided antiplatelet therapy has not been tested in a randomized outcomes trial, this tailored strategy has not been widely accepted.

THE FUTURE

The barriers to adoption of pharmacogenetics are considerable. Clinicians need to be educated about it, reimbursement needs to be worked out, and the pharmaceutical industry needs to get behind it. Nevertheless, the future of pharmacogenetics is extremely promising.

Research networks are forming to support the use of pharmacogenetics in clinical practice. The Pharmgkb (www.pharmgkb.org) database serves as a hub for educating clinicians and researchers as well as curating data for reference. Vanderbilt University is piloting the BioVu project, in which DNA and genotype data on patients are being stored and matched to the electronic clinical records.³⁰ These projects not only provide clinically useful information on the current state of the art of pharmacogenetics, they also aid in disseminating new information about genotype-phenotype relationships.

Analytical software that uses “natural language processing” is being applied to clinician-generated notes to derive new observations and associations between genetic variants and clinical phenotypes. Integrating this information in real-time decision-support modules in the electronic health record provides a feedback loop for a rapid assimilation of new knowledge. Similar innovative decision-support modules are being established by Cleveland Clinic’s Center for Personalized Healthcare.³¹

The rise of ‘omic’ sciences

Pharmacogenetics and pharmacogenomics are part of a larger set of “omic” sciences. The suffix “-omics” implies a larger, more holistic view and is being applied to a number of fields—for example, the study of proteins (proteomics) and the study of metabolites (metabolomics). Profiling proteins and metabolites delivers a deluge of information on a patient that can be clustered, using pattern-recognition software, into population subgroups. Patterns of multiple proteins or metabolites are extracted from this spectral data to identify disease or response to treatment (pharmacometabolomics).

Metabolomics has been shown to predict the response to statins,³² diagnose myocardial infarction,³³ and reclassify cardiovascular risk status.³⁴ In addition, whereas traditional laboratory chemistry is reductionist, using single biomarkers for single-disease diagnosis, omic technologies hold the potential to reveal information on a number of possible health or disease states. The identification of “healthy” profiles using these technologies can potentially provide positive feedback to patients undertaking lifestyle changes and treatment.

The instrument costs for proteomic and metabolomic profiling are relatively high. However, the ongoing running costs are minimal, estimated at as low as less than $13 per test, as there are no expensive reagents.³³ High-volume testing therefore becomes very cost-effective.

Although omic science appears futuristic, proteomics and metabolomics are already used in many clinical laboratories to rapidly identify bacteria. These methods have already revolutionized the way laboratories identify microbes, since they are automated, reduce workload, and give very fast results.

The cost of genetic testing is falling

Critics of pharmacogenetics claim that the predictive value of genetic testing is poor, that evidence is lacking, and that the cost is too high. In all new technologies, the first iteration is coarse, but performance improves with use. The first major barrier is adoption. Projects like BioVu are establishing the infrastructure for a feedback loop to iteratively improve upon the status quo and provide the evidence base clinicians demand.

The cost of genetic testing is falling rapidly, with whole-genome sequencing and annotation now costing less than $5,000. The cost of a pharmacogenetic test can be as low as $100 using low-cost nanotechnology, and the test needs to be performed only once in a patient’s lifetime.²⁷

As other related molecular technologies such as proteomics and metabolomics become available and are integrated with genomics, the predictive ability of this science will improve.

AWAY FROM ONE-SIZE-FITS-ALL MEDICINE

Over the last decade there has been a trend away from “one size fits all” to customized “markets of one” in everything from consumer products to education to medicine. Mass customizing, also known as personalization, has been embraced by the internet community as a means to increase efficiency and reduce cost. This occurs by eliminating waste in redundant work or production of ineffective products.

Personalization on the Internet has been enabled through the use of informatics, mathematics, and supercomputing. The same tools that have personalized the delivery of consumer products are also being applied to the field of pharmacogenetics. Applied in an evidence-based fashion, these new technologies should profoundly improve patient care now and in the future.

“Change is the only constant.”

—Heraclitus (c 535–475 bce)

With the cost of health care rising and money to pay for it shrinking, there has never been a greater need to reduce waste.

See related article

Ineffective treatments and adverse drug effects account for much preventable morbidity and expense. New treatments, touted as more potent, are often introduced as replacements for traditional ones that are still effective in many patients, adding to costs and the potential for harm. For the pharmaceutical industry, the search for new “blockbuster” drugs seems to have hit a wall, at least in cardiovascular medicine.¹ Advances often come at the cost of adverse effects, such as bleeding with triple antiplatelet therapy and diabetes with potent statin drugs.

The path to maximizing benefit and reducing harm now appears to lie in stratifying populations and appreciating patient individuality in response to treatment. For many decades we have known that patients vary widely in their response to drugs, owing to personal factors such as body surface area, age, environment, and genetics. And indeed, we treat our patients as individuals, for example by tailoring aminoglycoside dose to weight and renal function.

However, clinical trials typically give us an idea of the benefits only to the average patient. While subgroup analyses identify groups that may benefit more or less from treatment, the additional information they provide is not easily integrated into the clinical model of prescribing, in which one size fits all.

THE PROMISE OF PHARMACOGENETICS

The emerging field of pharmacogenetics promises to give clinicians the tools to make informed treatment decisions based on predictive genetic testing. This genetic testing aims to match treatment to an individual’s genetic profile. This often involves analyzing single-nucleotide polymorphisms in genes for enzymes that metabolize drugs, such as the cytochrome P450 enzymes, to predict efficacy or an adverse event with treatment.

Pharmacogenetics is playing an increasing role in clinical trials, particularly in the early stages of drug development, by helping to reduce the number of patients needed, prove efficacy, and identify subgroups in which alternative treatment can be targeted. At another level, a molecular understanding of disease is leading to truly targeted treatments based on genomics.

Over recent years, genetic testing has been increasingly used in clinical practice, thanks to a convergence of factors such as rapid, low-cost tests, a growing evidence base, and emerging interest among doctors and payers.

An advantage to using genetic testing as opposed to other types of laboratory testing, such as measuring the concentration of the drug in the blood during treatment, is that genetic tests can predict the response to treatment before the treatment is started. Moreover, with therapeutic drug monitoring after treatment has begun, there are sometimes no detectable measures of toxicity. For example, both carbamazepine and the antiviral drug abacavir can—fortunately only rarely—cause Stevens-Johnson syndrome. But before genetic markers were discovered, there was no method of estimating this risk apart from taking a family history.^2,3 Considering the numbers of people involved, it was not feasible until recently to suggest genetic screening for patients starting on these drugs. However, the cost of genotyping and gene sequencing has been falling at a rate inversely faster than Moore’s law (an approximate annual doubling in computer power), and population genomics is becoming a reality.⁴

The US Food and Drug Administration (FDA) recognizes the current and future value of pharmacogenetics in drug safety and development. A number of approved pharmacogenetic biomarkers are listed on the FDA website (Table 1). Black box warnings have been mandated for a number of drugs on the basis of observational evidence.

The FDA also promotes rapid approval for novel drugs with pharmacogenetic “companion diagnostics.” A recent example of this was the approval of ivacaftor for cystic fibrosis patients who have the G551D mutation.⁵ Here, a molecular understanding of the condition led to the development of a targeted treatment. Although the cost of developing this drug was high, the path is now paved for similar advances. Oncologists are familiar with these advances with the emergence of new molecularly targeted treatments, eg, BRAF inhibitors in metastatic melanoma, imatinib in chronic myeloid leukemia, and gefitinib in non-small-cell lung cancer.

PHARMACOGENETICS IN CARDIOVASCULAR MEDICINE

Cardiovascular medicine also stands to benefit from rapid advances in pharmacogenetics.

While no treatment has been developed that targets the molecular basis of cardiovascular disease, a number of genomic biomarkers have emerged that identify patients at risk of adverse reactions or treatment failure. These include genetic tests to predict the maintenance dose and risk of bleeding with warfarin,⁶ the likelihood of myopathy and myositis with simvastatin,⁷ and the risk of recurrent thrombotic events with clopidogrel.^8–10

Using pharmacogenetics in prescribing warfarin and its alternatives

The pharmacogenetics of warfarin has been extensively researched, but genotyping before prescribing this drug is not yet widely done.

In 2007, the FDA updated the labeling of warfarin to include information about the influence of two genes, VKORC1 and CYP2C9, on a patient’s response to this drug. In 2010 this was updated to add that testing for these genes could be used to predict the maintenance dose of the drug. Difficulties with algorithms used to integrate this into clinical practice have hindered adoption of this testing.

With the advent of new anticoagulants such as dabigatran, rivaroxaban, and apixaban, many have expected warfarin and its pharmacogenetics to become obsolete. However, the new agents cost considerably more. Further, they may not offer a very great advantage over warfarin: in the Randomized Evaluation of Long-term Anticoagulant Therapy (RE-LY) trial, the absolute risk reduction in intracranial hemorrhage with dabigatran vs warfarin was small.¹¹ Therefore, dabigatran is probably not cost-effective in populations at low risk of bleeding.¹² A cost-effectiveness analysis comparing warfarin with dabigatran in patients with uncomplicated atrial fibrillation has suggested that dabigatran is, however, cost-effective in patients at moderate risk.¹²

In the RE-LY trial, the international normalized ratios (INRs) of the patients in the warfarin group were in the therapeutic range only 64% of the time. The advantages of dabigatran over warfarin become less pronounced as warfarin control is tightened.¹³ Of note, pharmacogenetics and home monitoring of the INR have both been shown to lead to tighter control of the INR, with greater time within the therapeutic range.^14,15

Moreover, genetic testing can help us reduce the number of bleeding events in patients taking warfarin.¹⁶ Patients who carry the CYP2C9*2 or CYP2C9*3 polymorphism metabolize S-warfarin slower and therefore have a threefold higher risk of hemorrhage after starting warfarin.¹⁷ We could speculate that patients carrying these variants may be better served by the newer anticoagulants, though this has not been tested in any clinical trial.

It is also worth appreciating that the conditions requiring anticoagulation, such as atrial fibrillation, also have a strong genetic basis. Variants in chromosomes 4q25, 1q21, and 16q22 have all been associated with atrial fibrillation.¹⁸ The risk of atrial fibrillation is five to six times higher in carriers of multiple variants within all of these loci.¹⁹ Genetic variants at 4q25 have been associated with the response to specific antiarrhythmic drug treatments,²⁰ response to pulmonary vein isolation, ^21,22 and direct-current cardioversion.²³ One can imagine a future in which patients with palpitations, carrying multiple gene risk variants, will choose prolonged monitoring at home to confirm a diagnosis. They would then be provided with a personalized best management strategy, using their personal preferences, clinical data, and genetic profile to make a treatment decision.

Using pharmacogenetics in prescribing clopidogrel and its alternatives

The pharmacogenetics of clopidogrel is of particular interest, as it has the potential of establishing a rational basis for using newer antiplatelet drugs such as ticagrelor and prasugrel, which are considerably more expensive than generic clopidogrel.

Most of the people who do not respond to clopidogrel carry the common cytochrome P450 2C19 variants CYP2C19*2 or CYP2C19*3.⁹ These variants are present in particularly high frequency in Asians and African Americans, who often do not feature in large randomized trials.

Newer antiplatelet agents have failed to demonstrate consistent superiority to clopidogrel without a tradeoff of more bleeding. However, in the Trial to Assess Improvement in Therapeutic Outcomes by Optimizing Platelet Inhibition With Prasugrel-Thrombolysis in Myocardial Infarction 39 (TRITON TIMI-38),^24,25 patients with the *2 variant receiving prasugrel had lower cardiovascular event rates than *2 carriers receiving clopidogrel.

Similarly, the patients who benefit the most from ticagrelor are carriers of the 2C19 nonresponder variants. In a large study, clopidogrel responders who did not carry either 2C19 nonresponder genetic variants or ABCB1 variants had cardiovascular outcomes similar to those of patients receiving ticagrelor.²⁶

Clinicians have been cautious in prescribing potent antiplatelet agents to all patients because of the risk of bleeding. One could assume that by reserving newer agents for clopidogrel nonresponders, the bleeding risk could be minimized and overall benefit could be preserved with this strategy.

Cost may also be contained. The cost-effectiveness of such an approach with prasugrel has been tested with computer modeling and appears favorable.²⁷ On the other hand, a similar yet limited analysis did not find genotype-driven use of ticagrelor to be cost-effective.²⁸ This was mostly due to fewer deaths in patients receiving ticagrelor. However, the cost estimate for genotype-guided therapy was overestimated, as heterozygotes in the model were treated with ticagrelor instead of a high dose of clopidogrel.

It now appears that heterozygotes, ie, patients with one copy of the nonresponder variant, can achieve similar platelet inhibition with clopidogrel 225 mg daily as noncarriers on 75 mg daily.²⁹ Since genotype-guided antiplatelet therapy has not been tested in a randomized outcomes trial, this tailored strategy has not been widely accepted.

THE FUTURE

The barriers to adoption of pharmacogenetics are considerable. Clinicians need to be educated about it, reimbursement needs to be worked out, and the pharmaceutical industry needs to get behind it. Nevertheless, the future of pharmacogenetics is extremely promising.

Research networks are forming to support the use of pharmacogenetics in clinical practice. The Pharmgkb (www.pharmgkb.org) database serves as a hub for educating clinicians and researchers as well as curating data for reference. Vanderbilt University is piloting the BioVu project, in which DNA and genotype data on patients are being stored and matched to the electronic clinical records.³⁰ These projects not only provide clinically useful information on the current state of the art of pharmacogenetics, they also aid in disseminating new information about genotype-phenotype relationships.

Analytical software that uses “natural language processing” is being applied to clinician-generated notes to derive new observations and associations between genetic variants and clinical phenotypes. Integrating this information in real-time decision-support modules in the electronic health record provides a feedback loop for a rapid assimilation of new knowledge. Similar innovative decision-support modules are being established by Cleveland Clinic’s Center for Personalized Healthcare.³¹

The rise of ‘omic’ sciences

Pharmacogenetics and pharmacogenomics are part of a larger set of “omic” sciences. The suffix “-omics” implies a larger, more holistic view and is being applied to a number of fields—for example, the study of proteins (proteomics) and the study of metabolites (metabolomics). Profiling proteins and metabolites delivers a deluge of information on a patient that can be clustered, using pattern-recognition software, into population subgroups. Patterns of multiple proteins or metabolites are extracted from this spectral data to identify disease or response to treatment (pharmacometabolomics).

Metabolomics has been shown to predict the response to statins,³² diagnose myocardial infarction,³³ and reclassify cardiovascular risk status.³⁴ In addition, whereas traditional laboratory chemistry is reductionist, using single biomarkers for single-disease diagnosis, omic technologies hold the potential to reveal information on a number of possible health or disease states. The identification of “healthy” profiles using these technologies can potentially provide positive feedback to patients undertaking lifestyle changes and treatment.

The instrument costs for proteomic and metabolomic profiling are relatively high. However, the ongoing running costs are minimal, estimated at as low as less than $13 per test, as there are no expensive reagents.³³ High-volume testing therefore becomes very cost-effective.

Although omic science appears futuristic, proteomics and metabolomics are already used in many clinical laboratories to rapidly identify bacteria. These methods have already revolutionized the way laboratories identify microbes, since they are automated, reduce workload, and give very fast results.

The cost of genetic testing is falling

Critics of pharmacogenetics claim that the predictive value of genetic testing is poor, that evidence is lacking, and that the cost is too high. In all new technologies, the first iteration is coarse, but performance improves with use. The first major barrier is adoption. Projects like BioVu are establishing the infrastructure for a feedback loop to iteratively improve upon the status quo and provide the evidence base clinicians demand.

The cost of genetic testing is falling rapidly, with whole-genome sequencing and annotation now costing less than $5,000. The cost of a pharmacogenetic test can be as low as $100 using low-cost nanotechnology, and the test needs to be performed only once in a patient’s lifetime.²⁷

As other related molecular technologies such as proteomics and metabolomics become available and are integrated with genomics, the predictive ability of this science will improve.

AWAY FROM ONE-SIZE-FITS-ALL MEDICINE

Over the last decade there has been a trend away from “one size fits all” to customized “markets of one” in everything from consumer products to education to medicine. Mass customizing, also known as personalization, has been embraced by the internet community as a means to increase efficiency and reduce cost. This occurs by eliminating waste in redundant work or production of ineffective products.

Personalization on the Internet has been enabled through the use of informatics, mathematics, and supercomputing. The same tools that have personalized the delivery of consumer products are also being applied to the field of pharmacogenetics. Applied in an evidence-based fashion, these new technologies should profoundly improve patient care now and in the future.

“Change is the only constant.”

—Heraclitus (c 535–475 bce)

With the cost of health care rising and money to pay for it shrinking, there has never been a greater need to reduce waste.

See related article

Ineffective treatments and adverse drug effects account for much preventable morbidity and expense. New treatments, touted as more potent, are often introduced as replacements for traditional ones that are still effective in many patients, adding to costs and the potential for harm. For the pharmaceutical industry, the search for new “blockbuster” drugs seems to have hit a wall, at least in cardiovascular medicine.¹ Advances often come at the cost of adverse effects, such as bleeding with triple antiplatelet therapy and diabetes with potent statin drugs.

The path to maximizing benefit and reducing harm now appears to lie in stratifying populations and appreciating patient individuality in response to treatment. For many decades we have known that patients vary widely in their response to drugs, owing to personal factors such as body surface area, age, environment, and genetics. And indeed, we treat our patients as individuals, for example by tailoring aminoglycoside dose to weight and renal function.

However, clinical trials typically give us an idea of the benefits only to the average patient. While subgroup analyses identify groups that may benefit more or less from treatment, the additional information they provide is not easily integrated into the clinical model of prescribing, in which one size fits all.

THE PROMISE OF PHARMACOGENETICS

The emerging field of pharmacogenetics promises to give clinicians the tools to make informed treatment decisions based on predictive genetic testing. This genetic testing aims to match treatment to an individual’s genetic profile. This often involves analyzing single-nucleotide polymorphisms in genes for enzymes that metabolize drugs, such as the cytochrome P450 enzymes, to predict efficacy or an adverse event with treatment.

Pharmacogenetics is playing an increasing role in clinical trials, particularly in the early stages of drug development, by helping to reduce the number of patients needed, prove efficacy, and identify subgroups in which alternative treatment can be targeted. At another level, a molecular understanding of disease is leading to truly targeted treatments based on genomics.

Over recent years, genetic testing has been increasingly used in clinical practice, thanks to a convergence of factors such as rapid, low-cost tests, a growing evidence base, and emerging interest among doctors and payers.

An advantage to using genetic testing as opposed to other types of laboratory testing, such as measuring the concentration of the drug in the blood during treatment, is that genetic tests can predict the response to treatment before the treatment is started. Moreover, with therapeutic drug monitoring after treatment has begun, there are sometimes no detectable measures of toxicity. For example, both carbamazepine and the antiviral drug abacavir can—fortunately only rarely—cause Stevens-Johnson syndrome. But before genetic markers were discovered, there was no method of estimating this risk apart from taking a family history.^2,3 Considering the numbers of people involved, it was not feasible until recently to suggest genetic screening for patients starting on these drugs. However, the cost of genotyping and gene sequencing has been falling at a rate inversely faster than Moore’s law (an approximate annual doubling in computer power), and population genomics is becoming a reality.⁴

The US Food and Drug Administration (FDA) recognizes the current and future value of pharmacogenetics in drug safety and development. A number of approved pharmacogenetic biomarkers are listed on the FDA website (Table 1). Black box warnings have been mandated for a number of drugs on the basis of observational evidence.

The FDA also promotes rapid approval for novel drugs with pharmacogenetic “companion diagnostics.” A recent example of this was the approval of ivacaftor for cystic fibrosis patients who have the G551D mutation.⁵ Here, a molecular understanding of the condition led to the development of a targeted treatment. Although the cost of developing this drug was high, the path is now paved for similar advances. Oncologists are familiar with these advances with the emergence of new molecularly targeted treatments, eg, BRAF inhibitors in metastatic melanoma, imatinib in chronic myeloid leukemia, and gefitinib in non-small-cell lung cancer.

PHARMACOGENETICS IN CARDIOVASCULAR MEDICINE

Cardiovascular medicine also stands to benefit from rapid advances in pharmacogenetics.

While no treatment has been developed that targets the molecular basis of cardiovascular disease, a number of genomic biomarkers have emerged that identify patients at risk of adverse reactions or treatment failure. These include genetic tests to predict the maintenance dose and risk of bleeding with warfarin,⁶ the likelihood of myopathy and myositis with simvastatin,⁷ and the risk of recurrent thrombotic events with clopidogrel.^8–10

Using pharmacogenetics in prescribing warfarin and its alternatives

The pharmacogenetics of warfarin has been extensively researched, but genotyping before prescribing this drug is not yet widely done.

In 2007, the FDA updated the labeling of warfarin to include information about the influence of two genes, VKORC1 and CYP2C9, on a patient’s response to this drug. In 2010 this was updated to add that testing for these genes could be used to predict the maintenance dose of the drug. Difficulties with algorithms used to integrate this into clinical practice have hindered adoption of this testing.

With the advent of new anticoagulants such as dabigatran, rivaroxaban, and apixaban, many have expected warfarin and its pharmacogenetics to become obsolete. However, the new agents cost considerably more. Further, they may not offer a very great advantage over warfarin: in the Randomized Evaluation of Long-term Anticoagulant Therapy (RE-LY) trial, the absolute risk reduction in intracranial hemorrhage with dabigatran vs warfarin was small.¹¹ Therefore, dabigatran is probably not cost-effective in populations at low risk of bleeding.¹² A cost-effectiveness analysis comparing warfarin with dabigatran in patients with uncomplicated atrial fibrillation has suggested that dabigatran is, however, cost-effective in patients at moderate risk.¹²

In the RE-LY trial, the international normalized ratios (INRs) of the patients in the warfarin group were in the therapeutic range only 64% of the time. The advantages of dabigatran over warfarin become less pronounced as warfarin control is tightened.¹³ Of note, pharmacogenetics and home monitoring of the INR have both been shown to lead to tighter control of the INR, with greater time within the therapeutic range.^14,15

Moreover, genetic testing can help us reduce the number of bleeding events in patients taking warfarin.¹⁶ Patients who carry the CYP2C9*2 or CYP2C9*3 polymorphism metabolize S-warfarin slower and therefore have a threefold higher risk of hemorrhage after starting warfarin.¹⁷ We could speculate that patients carrying these variants may be better served by the newer anticoagulants, though this has not been tested in any clinical trial.

It is also worth appreciating that the conditions requiring anticoagulation, such as atrial fibrillation, also have a strong genetic basis. Variants in chromosomes 4q25, 1q21, and 16q22 have all been associated with atrial fibrillation.¹⁸ The risk of atrial fibrillation is five to six times higher in carriers of multiple variants within all of these loci.¹⁹ Genetic variants at 4q25 have been associated with the response to specific antiarrhythmic drug treatments,²⁰ response to pulmonary vein isolation, ^21,22 and direct-current cardioversion.²³ One can imagine a future in which patients with palpitations, carrying multiple gene risk variants, will choose prolonged monitoring at home to confirm a diagnosis. They would then be provided with a personalized best management strategy, using their personal preferences, clinical data, and genetic profile to make a treatment decision.

Using pharmacogenetics in prescribing clopidogrel and its alternatives

The pharmacogenetics of clopidogrel is of particular interest, as it has the potential of establishing a rational basis for using newer antiplatelet drugs such as ticagrelor and prasugrel, which are considerably more expensive than generic clopidogrel.

Most of the people who do not respond to clopidogrel carry the common cytochrome P450 2C19 variants CYP2C19*2 or CYP2C19*3.⁹ These variants are present in particularly high frequency in Asians and African Americans, who often do not feature in large randomized trials.

Newer antiplatelet agents have failed to demonstrate consistent superiority to clopidogrel without a tradeoff of more bleeding. However, in the Trial to Assess Improvement in Therapeutic Outcomes by Optimizing Platelet Inhibition With Prasugrel-Thrombolysis in Myocardial Infarction 39 (TRITON TIMI-38),^24,25 patients with the *2 variant receiving prasugrel had lower cardiovascular event rates than *2 carriers receiving clopidogrel.

Similarly, the patients who benefit the most from ticagrelor are carriers of the 2C19 nonresponder variants. In a large study, clopidogrel responders who did not carry either 2C19 nonresponder genetic variants or ABCB1 variants had cardiovascular outcomes similar to those of patients receiving ticagrelor.²⁶

Clinicians have been cautious in prescribing potent antiplatelet agents to all patients because of the risk of bleeding. One could assume that by reserving newer agents for clopidogrel nonresponders, the bleeding risk could be minimized and overall benefit could be preserved with this strategy.

Cost may also be contained. The cost-effectiveness of such an approach with prasugrel has been tested with computer modeling and appears favorable.²⁷ On the other hand, a similar yet limited analysis did not find genotype-driven use of ticagrelor to be cost-effective.²⁸ This was mostly due to fewer deaths in patients receiving ticagrelor. However, the cost estimate for genotype-guided therapy was overestimated, as heterozygotes in the model were treated with ticagrelor instead of a high dose of clopidogrel.

It now appears that heterozygotes, ie, patients with one copy of the nonresponder variant, can achieve similar platelet inhibition with clopidogrel 225 mg daily as noncarriers on 75 mg daily.²⁹ Since genotype-guided antiplatelet therapy has not been tested in a randomized outcomes trial, this tailored strategy has not been widely accepted.

THE FUTURE

The barriers to adoption of pharmacogenetics are considerable. Clinicians need to be educated about it, reimbursement needs to be worked out, and the pharmaceutical industry needs to get behind it. Nevertheless, the future of pharmacogenetics is extremely promising.

Research networks are forming to support the use of pharmacogenetics in clinical practice. The Pharmgkb (www.pharmgkb.org) database serves as a hub for educating clinicians and researchers as well as curating data for reference. Vanderbilt University is piloting the BioVu project, in which DNA and genotype data on patients are being stored and matched to the electronic clinical records.³⁰ These projects not only provide clinically useful information on the current state of the art of pharmacogenetics, they also aid in disseminating new information about genotype-phenotype relationships.

Analytical software that uses “natural language processing” is being applied to clinician-generated notes to derive new observations and associations between genetic variants and clinical phenotypes. Integrating this information in real-time decision-support modules in the electronic health record provides a feedback loop for a rapid assimilation of new knowledge. Similar innovative decision-support modules are being established by Cleveland Clinic’s Center for Personalized Healthcare.³¹

The rise of ‘omic’ sciences

Pharmacogenetics and pharmacogenomics are part of a larger set of “omic” sciences. The suffix “-omics” implies a larger, more holistic view and is being applied to a number of fields—for example, the study of proteins (proteomics) and the study of metabolites (metabolomics). Profiling proteins and metabolites delivers a deluge of information on a patient that can be clustered, using pattern-recognition software, into population subgroups. Patterns of multiple proteins or metabolites are extracted from this spectral data to identify disease or response to treatment (pharmacometabolomics).

Metabolomics has been shown to predict the response to statins,³² diagnose myocardial infarction,³³ and reclassify cardiovascular risk status.³⁴ In addition, whereas traditional laboratory chemistry is reductionist, using single biomarkers for single-disease diagnosis, omic technologies hold the potential to reveal information on a number of possible health or disease states. The identification of “healthy” profiles using these technologies can potentially provide positive feedback to patients undertaking lifestyle changes and treatment.

The instrument costs for proteomic and metabolomic profiling are relatively high. However, the ongoing running costs are minimal, estimated at as low as less than $13 per test, as there are no expensive reagents.³³ High-volume testing therefore becomes very cost-effective.

Although omic science appears futuristic, proteomics and metabolomics are already used in many clinical laboratories to rapidly identify bacteria. These methods have already revolutionized the way laboratories identify microbes, since they are automated, reduce workload, and give very fast results.

The cost of genetic testing is falling

Critics of pharmacogenetics claim that the predictive value of genetic testing is poor, that evidence is lacking, and that the cost is too high. In all new technologies, the first iteration is coarse, but performance improves with use. The first major barrier is adoption. Projects like BioVu are establishing the infrastructure for a feedback loop to iteratively improve upon the status quo and provide the evidence base clinicians demand.

The cost of genetic testing is falling rapidly, with whole-genome sequencing and annotation now costing less than $5,000. The cost of a pharmacogenetic test can be as low as $100 using low-cost nanotechnology, and the test needs to be performed only once in a patient’s lifetime.²⁷

As other related molecular technologies such as proteomics and metabolomics become available and are integrated with genomics, the predictive ability of this science will improve.

AWAY FROM ONE-SIZE-FITS-ALL MEDICINE

Over the last decade there has been a trend away from “one size fits all” to customized “markets of one” in everything from consumer products to education to medicine. Mass customizing, also known as personalization, has been embraced by the internet community as a means to increase efficiency and reduce cost. This occurs by eliminating waste in redundant work or production of ineffective products.

Personalization on the Internet has been enabled through the use of informatics, mathematics, and supercomputing. The same tools that have personalized the delivery of consumer products are also being applied to the field of pharmacogenetics. Applied in an evidence-based fashion, these new technologies should profoundly improve patient care now and in the future.

The answer is not clear. There is evidence in favor of pharmacogenetic testing, but not yet enough to strongly recommend it. However, we do believe that physicians should consider it when starting patients on warfarin therapy.

See related commentary

WARFARIN HAS A NARROW THERAPEUTIC WINDOW

Although newer drugs are available, warfarin is still the most commonly used oral anticoagulant for preventing and treating thromboembolism.¹ It is highly effective but has a narrow therapeutic window and wide interindividual variability in dosage requirements, which poses challenges to achieving adequate anticoagulation.^1–3 Inappropriate dosing contributes to a high rate of bleeding events and emergency room visits.⁴

Warfarin is monitored using the prothrombin time. Because the prothrombin time varies depending on the assay used, the standardized value called the international normalized ratio (INR) is more commonly used.

Clinical factors such as age, body size, and drug interactions affect warfarin dosage requirements and are important to consider,⁵ even though they account for only 15% to 20% of the variability in warfarin dose.⁶

Genetic factors also affect warfarin dosage requirements. The combination of genetic and clinical factors accounts for up to 47% of the dose variability.⁷

GENES THAT AFFECT WARFARIN

Several genes are known to influence warfarin’s pharmacokinetics and pharmacodynamics. Of these, the two most clinically relevant and well studied are CYP2C9 (which codes for cytochrome P450 2C9) and VKORC1 (which codes for vitamin K epoxide reductase).⁷ These genes are polymorphic, with some variants producing less-active enzymes that allow warfarin to be more active. Therefore, patients who carry these variants need lower doses of this drug (see below).

CYP2C9 variants

The CYP2C9 gene has several variants. Of these, CYP2C9*2 and CYP2C9*3 are associated with the lowest enzyme activity.

Patients with either of these variants require significantly lower warfarin doses to reach therapeutic levels than those with the wild-type gene (ie, CYP2C9*1). CYP2C9*2 reduces warfarin clearance by 40%, and the CYP2C9*3 variant reduces it by 75%.⁷ Having a *2 or *3 allele increases the risk of bleeding during warfarin therapy and the time needed to achieve a stable dose.⁸ Other variants associated with lower warfarin dose requirements are *5, *6, and *11.

The prevalence of these variants is significantly higher in people of European ancestry (roughly one-third) than in Asian people and African Americans,⁷ although no one has recommended not testing in these low-prevalence populations. Limdi et al⁹ reported that by including the *5, *6, and *11 variants in genetic testing (in addition to *2 and *3), they could identify more African Americans (9.7%) who carried at least one of these abnormal variants than reported previously. Differences among ethnic groups need to be taken into account when interpreting pharmacogenetic studies.

VKORC1 variants

Patients also need lower doses of warfarin if they carry the VKORC1 −1639G>A variant, and they spend more time with an INR above the therapeutic range and have higher overall INR values. However, having this variant does not appear to increase the risk of bleeding.

The −1639G>A variant is the most common variant of VKORC1. Rarer ones have also been described, but most commercially available tests do not detect them.

Racial differences exist in the prevalence rates of the various VKORC1 polymorphisms, with the most sensitive (low-dose) genotype predominating in Asians and the least sensitive (high-dose) genotype predominating in African Americans. Over 50% of people of European ancestry carry the intermediate-sensitivity genotype (typical dose).⁷

CURRENT RECOMMENDATIONS FOR OR AGAINST TESTING

FDA labeling

In 2007, the US Food and Drug Administration (FDA) required that the warfarin package insert carry information about initial dosing based on CYP2C9 and VKORC1 testing. This recommendation was revised in 2010 to include a table to help clinicians select an initial warfarin dose if CYP2C9 and VKORC1 genotype information is available. However, the FDA does not require pharmacogenetic testing, leaving the decision to the discretion of the clinician.⁷

American College of Chest Physicians

The American College of Chest Physicians recommends against routine pharmacogenetic testing (grade 1B) because of a lack of evidence that it improves clinical end points or that it is cost-effective.⁵

WHAT EVIDENCE SUPORTS GENETIC TESTING TO GUIDE WARFARIN THERAPY?

To date, no large randomized, controlled trial has been published that looked at clinical outcomes with warfarin dosing based on pharmacogenetic testing. However, several smaller studies have suggested it is beneficial.

One trial found that when dosing was informed by pharmacogenetic testing, patients had significantly more time in the therapeutic range, a lower percentage of INRs greater than 4 or less than 1.5, and fewer serious adverse events (death, myocardial infarction, stroke, thromboembolism, and clinically significant bleeding events).¹⁰ Patients whose dosage was determined using pharmacogenetic algorithms as opposed to traditional clinical algorithms maintained a therapeutic INR more consistently.¹¹

In addition, compared with historical controls, patients whose physician used pharmacogenetic testing to guide warfarin dosing had a rate of hospitalization 31% lower and a rate of hospitalization specifically for bleeding or thromboembolism 28% lower during 6 months of follow-up.^12,13

Several studies have attempted to assess the cost-effectiveness and utility of pharmacogenetic testing in warfarin therapy. As yet, the results have been inconclusive.¹⁴ Larger prospective trials are under way and are estimated to be completed in late 2013.¹⁵ These include:

• COAG (Clarification of Optimal Anticoagulation Through Genetics)
• GIFT (Genetics Informatics Trial of Warfarin to Prevent Venous Thrombosis)
• EU-PACT (European Pharmacogenetics of Anticoagulant Therapy-Warfarin).

We hope these studies will provide greater clarity on the clinical utility and cost-effectiveness of pharmacogenetic testing to guide warfarin dosing.

HOW SHOULD GENETIC INFORMATION BE USED TO GUIDE OR ALTER THERAPY?

Algorithms are available for estimating initial and maintenance warfarin doses based on genetic information (CYP2C9 and VKORC1), race or ethnicity, age, sex, body mass index, smoking status, and other medications taken. In addition, models incorporating the INR on day 4 and days 6 to 11 have been developed for dose refinement.¹⁵ The algorithms explain 30% to 60% of the variability of the data, with lower values for African Americans.⁷

A well-developed dosing model that includes traditional clinical factors and patient genetic status is publicly available online at www.warfarindosing.org.⁴

CPIC: A leader in applied pharmacogenetics

In late 2009, PharmGKB joined forces with the Pharmacogenomics Research Network of the National Institutes of Health to form the Clinical Pharmacogenetics Implementation Consortium (CPIC). This organization issues guidelines that are written by expert clinicians and scientists and then are peer-reviewed, published in leading journals, and simultaneously posted to the PharmGKB website along with supplemental information and updates.

CPIC’s goal is to review the current evidence and to address barriers to the adoption of pharmacogenetic testing into clinical practice. Its guidelines do not advise when or which pharmacogenetic tests should be ordered. Rather, they provide guidance on interpreting and applying such testing, should the test results be available.⁷

CPIC has guidelines on CYP2C9 and VKORC1 genotypes and warfarin dosing.⁸ If a patient’s genetic information is available, CPIC strongly recommends the use of pharmacogenetic algorithm-based dosing. If such an algorithm is not accessible, use of a genotype dosing table is recommended.⁸

Monitoring is still needed

Many factors can affect an individual’s response to warfarin above and beyond the above-noted clinical and genetic traits. These include diet, concomitant medications (both prescription and over-the-counter and herbal), and disease state. There may also be additional genetic polymorphisms not yet identified in various racial and ethnic groups that may affect dosing requirements. And as with all medications, patient compliance and dosing errors have a large potential to affect individual response. Therefore, clinicians should still be diligent about clinical monitoring.¹⁵

Most useful for initial dose

As with most pharmacogenetic information, the greatest benefit can be achieved when this information is used to guide the initial dose, although there is also some effect noted when this information is known and acted upon into the 2nd week of treatment.⁸

Patients on long-term warfarin treatment with stable doses and those unable to achieve stable dosing because of variable adherence or dietary vitamin K intake are less likely to benefit from genetic testing.

There are no published guidelines on the utility of pharmacogenetic testing if a patient is already on a stable dose of warfarin or has a known historical stable dose. There are also no published guidelines on changing the frequency of monitoring based on known genotype.

In children, the data are sparse at this time regarding the utility of pharmacogenetically informed dosing.

HOW DOES ONE ORDER TESTING, AND WHAT IS THE COST?

The FDA has approved four warfarin pharmacogenetic test kits. To be used in clinical decision-making, these tests must be done in a laboratory certified by the Clinical Laboratory Improvement Amendments (CLIA) program.

Testing typically costs a few hundred dollars and may take days for results to be returned if not available on site.¹⁵ At Cleveland Clinic, CYP2C9 and VKORC1 testing can be run in-house at a cost of about $700. Generally, many third-party payers do not reimburse for testing without a prior-approval process.

TO TEST OR NOT TO TEST

Pharmacogenetic testing is available and may help optimize warfarin dosing early in treatment, as well as help maintain therapeutic INRs more consistently. There is preliminary evidence that using this information to guide dosing improves clinical outcomes. Several large trials are under way to address additional questions of clinical utility, with results expected in the next year. There are also readily available decision-support tools to guide therapeutic dosing, and when pharmacogenetic test results are available, utilization of a warfarin dosing algorithm is recommended.

The largest barrier remaining appears to be cost (relative to perceived benefit), and until larger trials of clinical utility and cost-effectiveness are completed and analyzed, hurdles exist to obtaining coverage for such testing.

If it is readily available (and can be paid for by insurance companies or out-of-pocket) and test results can be obtained within 24 to 48 hours or before prescribing, pharmacogenetic testing can be a valuable tool to guide and manage warfarin dosing. Particularly for patients who want to be as proactive as possible, warfarin pharmacogenetic testing offers the ability to participate in this decision-making and to potentially reduce their risk of adverse drug events. And in view of the evidence and FDA recommendations, we propose that the discussion with our patients is not whether we should consider pharmacogenetic testing, but that we have considered pharmacogenetic testing, and why we have decided for or against it.

The answer is not clear. There is evidence in favor of pharmacogenetic testing, but not yet enough to strongly recommend it. However, we do believe that physicians should consider it when starting patients on warfarin therapy.

See related commentary

WARFARIN HAS A NARROW THERAPEUTIC WINDOW

Although newer drugs are available, warfarin is still the most commonly used oral anticoagulant for preventing and treating thromboembolism.¹ It is highly effective but has a narrow therapeutic window and wide interindividual variability in dosage requirements, which poses challenges to achieving adequate anticoagulation.^1–3 Inappropriate dosing contributes to a high rate of bleeding events and emergency room visits.⁴

Warfarin is monitored using the prothrombin time. Because the prothrombin time varies depending on the assay used, the standardized value called the international normalized ratio (INR) is more commonly used.

Clinical factors such as age, body size, and drug interactions affect warfarin dosage requirements and are important to consider,⁵ even though they account for only 15% to 20% of the variability in warfarin dose.⁶

Genetic factors also affect warfarin dosage requirements. The combination of genetic and clinical factors accounts for up to 47% of the dose variability.⁷

GENES THAT AFFECT WARFARIN

Several genes are known to influence warfarin’s pharmacokinetics and pharmacodynamics. Of these, the two most clinically relevant and well studied are CYP2C9 (which codes for cytochrome P450 2C9) and VKORC1 (which codes for vitamin K epoxide reductase).⁷ These genes are polymorphic, with some variants producing less-active enzymes that allow warfarin to be more active. Therefore, patients who carry these variants need lower doses of this drug (see below).

CYP2C9 variants

The CYP2C9 gene has several variants. Of these, CYP2C9*2 and CYP2C9*3 are associated with the lowest enzyme activity.

Patients with either of these variants require significantly lower warfarin doses to reach therapeutic levels than those with the wild-type gene (ie, CYP2C9*1). CYP2C9*2 reduces warfarin clearance by 40%, and the CYP2C9*3 variant reduces it by 75%.⁷ Having a *2 or *3 allele increases the risk of bleeding during warfarin therapy and the time needed to achieve a stable dose.⁸ Other variants associated with lower warfarin dose requirements are *5, *6, and *11.

The prevalence of these variants is significantly higher in people of European ancestry (roughly one-third) than in Asian people and African Americans,⁷ although no one has recommended not testing in these low-prevalence populations. Limdi et al⁹ reported that by including the *5, *6, and *11 variants in genetic testing (in addition to *2 and *3), they could identify more African Americans (9.7%) who carried at least one of these abnormal variants than reported previously. Differences among ethnic groups need to be taken into account when interpreting pharmacogenetic studies.

VKORC1 variants

Patients also need lower doses of warfarin if they carry the VKORC1 −1639G>A variant, and they spend more time with an INR above the therapeutic range and have higher overall INR values. However, having this variant does not appear to increase the risk of bleeding.

The −1639G>A variant is the most common variant of VKORC1. Rarer ones have also been described, but most commercially available tests do not detect them.

Racial differences exist in the prevalence rates of the various VKORC1 polymorphisms, with the most sensitive (low-dose) genotype predominating in Asians and the least sensitive (high-dose) genotype predominating in African Americans. Over 50% of people of European ancestry carry the intermediate-sensitivity genotype (typical dose).⁷

CURRENT RECOMMENDATIONS FOR OR AGAINST TESTING

FDA labeling

In 2007, the US Food and Drug Administration (FDA) required that the warfarin package insert carry information about initial dosing based on CYP2C9 and VKORC1 testing. This recommendation was revised in 2010 to include a table to help clinicians select an initial warfarin dose if CYP2C9 and VKORC1 genotype information is available. However, the FDA does not require pharmacogenetic testing, leaving the decision to the discretion of the clinician.⁷

American College of Chest Physicians

The American College of Chest Physicians recommends against routine pharmacogenetic testing (grade 1B) because of a lack of evidence that it improves clinical end points or that it is cost-effective.⁵

WHAT EVIDENCE SUPORTS GENETIC TESTING TO GUIDE WARFARIN THERAPY?

To date, no large randomized, controlled trial has been published that looked at clinical outcomes with warfarin dosing based on pharmacogenetic testing. However, several smaller studies have suggested it is beneficial.

One trial found that when dosing was informed by pharmacogenetic testing, patients had significantly more time in the therapeutic range, a lower percentage of INRs greater than 4 or less than 1.5, and fewer serious adverse events (death, myocardial infarction, stroke, thromboembolism, and clinically significant bleeding events).¹⁰ Patients whose dosage was determined using pharmacogenetic algorithms as opposed to traditional clinical algorithms maintained a therapeutic INR more consistently.¹¹

In addition, compared with historical controls, patients whose physician used pharmacogenetic testing to guide warfarin dosing had a rate of hospitalization 31% lower and a rate of hospitalization specifically for bleeding or thromboembolism 28% lower during 6 months of follow-up.^12,13

Several studies have attempted to assess the cost-effectiveness and utility of pharmacogenetic testing in warfarin therapy. As yet, the results have been inconclusive.¹⁴ Larger prospective trials are under way and are estimated to be completed in late 2013.¹⁵ These include:

• COAG (Clarification of Optimal Anticoagulation Through Genetics)
• GIFT (Genetics Informatics Trial of Warfarin to Prevent Venous Thrombosis)
• EU-PACT (European Pharmacogenetics of Anticoagulant Therapy-Warfarin).

We hope these studies will provide greater clarity on the clinical utility and cost-effectiveness of pharmacogenetic testing to guide warfarin dosing.

HOW SHOULD GENETIC INFORMATION BE USED TO GUIDE OR ALTER THERAPY?

Algorithms are available for estimating initial and maintenance warfarin doses based on genetic information (CYP2C9 and VKORC1), race or ethnicity, age, sex, body mass index, smoking status, and other medications taken. In addition, models incorporating the INR on day 4 and days 6 to 11 have been developed for dose refinement.¹⁵ The algorithms explain 30% to 60% of the variability of the data, with lower values for African Americans.⁷

A well-developed dosing model that includes traditional clinical factors and patient genetic status is publicly available online at www.warfarindosing.org.⁴

CPIC: A leader in applied pharmacogenetics

In late 2009, PharmGKB joined forces with the Pharmacogenomics Research Network of the National Institutes of Health to form the Clinical Pharmacogenetics Implementation Consortium (CPIC). This organization issues guidelines that are written by expert clinicians and scientists and then are peer-reviewed, published in leading journals, and simultaneously posted to the PharmGKB website along with supplemental information and updates.

CPIC’s goal is to review the current evidence and to address barriers to the adoption of pharmacogenetic testing into clinical practice. Its guidelines do not advise when or which pharmacogenetic tests should be ordered. Rather, they provide guidance on interpreting and applying such testing, should the test results be available.⁷

CPIC has guidelines on CYP2C9 and VKORC1 genotypes and warfarin dosing.⁸ If a patient’s genetic information is available, CPIC strongly recommends the use of pharmacogenetic algorithm-based dosing. If such an algorithm is not accessible, use of a genotype dosing table is recommended.⁸

Monitoring is still needed

Many factors can affect an individual’s response to warfarin above and beyond the above-noted clinical and genetic traits. These include diet, concomitant medications (both prescription and over-the-counter and herbal), and disease state. There may also be additional genetic polymorphisms not yet identified in various racial and ethnic groups that may affect dosing requirements. And as with all medications, patient compliance and dosing errors have a large potential to affect individual response. Therefore, clinicians should still be diligent about clinical monitoring.¹⁵

Most useful for initial dose

As with most pharmacogenetic information, the greatest benefit can be achieved when this information is used to guide the initial dose, although there is also some effect noted when this information is known and acted upon into the 2nd week of treatment.⁸

Patients on long-term warfarin treatment with stable doses and those unable to achieve stable dosing because of variable adherence or dietary vitamin K intake are less likely to benefit from genetic testing.

There are no published guidelines on the utility of pharmacogenetic testing if a patient is already on a stable dose of warfarin or has a known historical stable dose. There are also no published guidelines on changing the frequency of monitoring based on known genotype.

In children, the data are sparse at this time regarding the utility of pharmacogenetically informed dosing.

HOW DOES ONE ORDER TESTING, AND WHAT IS THE COST?

The FDA has approved four warfarin pharmacogenetic test kits. To be used in clinical decision-making, these tests must be done in a laboratory certified by the Clinical Laboratory Improvement Amendments (CLIA) program.

Testing typically costs a few hundred dollars and may take days for results to be returned if not available on site.¹⁵ At Cleveland Clinic, CYP2C9 and VKORC1 testing can be run in-house at a cost of about $700. Generally, many third-party payers do not reimburse for testing without a prior-approval process.

TO TEST OR NOT TO TEST

Pharmacogenetic testing is available and may help optimize warfarin dosing early in treatment, as well as help maintain therapeutic INRs more consistently. There is preliminary evidence that using this information to guide dosing improves clinical outcomes. Several large trials are under way to address additional questions of clinical utility, with results expected in the next year. There are also readily available decision-support tools to guide therapeutic dosing, and when pharmacogenetic test results are available, utilization of a warfarin dosing algorithm is recommended.

The largest barrier remaining appears to be cost (relative to perceived benefit), and until larger trials of clinical utility and cost-effectiveness are completed and analyzed, hurdles exist to obtaining coverage for such testing.

If it is readily available (and can be paid for by insurance companies or out-of-pocket) and test results can be obtained within 24 to 48 hours or before prescribing, pharmacogenetic testing can be a valuable tool to guide and manage warfarin dosing. Particularly for patients who want to be as proactive as possible, warfarin pharmacogenetic testing offers the ability to participate in this decision-making and to potentially reduce their risk of adverse drug events. And in view of the evidence and FDA recommendations, we propose that the discussion with our patients is not whether we should consider pharmacogenetic testing, but that we have considered pharmacogenetic testing, and why we have decided for or against it.

The answer is not clear. There is evidence in favor of pharmacogenetic testing, but not yet enough to strongly recommend it. However, we do believe that physicians should consider it when starting patients on warfarin therapy.

See related commentary

WARFARIN HAS A NARROW THERAPEUTIC WINDOW

Although newer drugs are available, warfarin is still the most commonly used oral anticoagulant for preventing and treating thromboembolism.¹ It is highly effective but has a narrow therapeutic window and wide interindividual variability in dosage requirements, which poses challenges to achieving adequate anticoagulation.^1–3 Inappropriate dosing contributes to a high rate of bleeding events and emergency room visits.⁴

Warfarin is monitored using the prothrombin time. Because the prothrombin time varies depending on the assay used, the standardized value called the international normalized ratio (INR) is more commonly used.

Clinical factors such as age, body size, and drug interactions affect warfarin dosage requirements and are important to consider,⁵ even though they account for only 15% to 20% of the variability in warfarin dose.⁶

Genetic factors also affect warfarin dosage requirements. The combination of genetic and clinical factors accounts for up to 47% of the dose variability.⁷

GENES THAT AFFECT WARFARIN

Several genes are known to influence warfarin’s pharmacokinetics and pharmacodynamics. Of these, the two most clinically relevant and well studied are CYP2C9 (which codes for cytochrome P450 2C9) and VKORC1 (which codes for vitamin K epoxide reductase).⁷ These genes are polymorphic, with some variants producing less-active enzymes that allow warfarin to be more active. Therefore, patients who carry these variants need lower doses of this drug (see below).

CYP2C9 variants

The CYP2C9 gene has several variants. Of these, CYP2C9*2 and CYP2C9*3 are associated with the lowest enzyme activity.

Patients with either of these variants require significantly lower warfarin doses to reach therapeutic levels than those with the wild-type gene (ie, CYP2C9*1). CYP2C9*2 reduces warfarin clearance by 40%, and the CYP2C9*3 variant reduces it by 75%.⁷ Having a *2 or *3 allele increases the risk of bleeding during warfarin therapy and the time needed to achieve a stable dose.⁸ Other variants associated with lower warfarin dose requirements are *5, *6, and *11.

The prevalence of these variants is significantly higher in people of European ancestry (roughly one-third) than in Asian people and African Americans,⁷ although no one has recommended not testing in these low-prevalence populations. Limdi et al⁹ reported that by including the *5, *6, and *11 variants in genetic testing (in addition to *2 and *3), they could identify more African Americans (9.7%) who carried at least one of these abnormal variants than reported previously. Differences among ethnic groups need to be taken into account when interpreting pharmacogenetic studies.

VKORC1 variants

Patients also need lower doses of warfarin if they carry the VKORC1 −1639G>A variant, and they spend more time with an INR above the therapeutic range and have higher overall INR values. However, having this variant does not appear to increase the risk of bleeding.

The −1639G>A variant is the most common variant of VKORC1. Rarer ones have also been described, but most commercially available tests do not detect them.

Racial differences exist in the prevalence rates of the various VKORC1 polymorphisms, with the most sensitive (low-dose) genotype predominating in Asians and the least sensitive (high-dose) genotype predominating in African Americans. Over 50% of people of European ancestry carry the intermediate-sensitivity genotype (typical dose).⁷

CURRENT RECOMMENDATIONS FOR OR AGAINST TESTING

FDA labeling

In 2007, the US Food and Drug Administration (FDA) required that the warfarin package insert carry information about initial dosing based on CYP2C9 and VKORC1 testing. This recommendation was revised in 2010 to include a table to help clinicians select an initial warfarin dose if CYP2C9 and VKORC1 genotype information is available. However, the FDA does not require pharmacogenetic testing, leaving the decision to the discretion of the clinician.⁷

American College of Chest Physicians

The American College of Chest Physicians recommends against routine pharmacogenetic testing (grade 1B) because of a lack of evidence that it improves clinical end points or that it is cost-effective.⁵

WHAT EVIDENCE SUPORTS GENETIC TESTING TO GUIDE WARFARIN THERAPY?

To date, no large randomized, controlled trial has been published that looked at clinical outcomes with warfarin dosing based on pharmacogenetic testing. However, several smaller studies have suggested it is beneficial.

One trial found that when dosing was informed by pharmacogenetic testing, patients had significantly more time in the therapeutic range, a lower percentage of INRs greater than 4 or less than 1.5, and fewer serious adverse events (death, myocardial infarction, stroke, thromboembolism, and clinically significant bleeding events).¹⁰ Patients whose dosage was determined using pharmacogenetic algorithms as opposed to traditional clinical algorithms maintained a therapeutic INR more consistently.¹¹

In addition, compared with historical controls, patients whose physician used pharmacogenetic testing to guide warfarin dosing had a rate of hospitalization 31% lower and a rate of hospitalization specifically for bleeding or thromboembolism 28% lower during 6 months of follow-up.^12,13

Several studies have attempted to assess the cost-effectiveness and utility of pharmacogenetic testing in warfarin therapy. As yet, the results have been inconclusive.¹⁴ Larger prospective trials are under way and are estimated to be completed in late 2013.¹⁵ These include:

• COAG (Clarification of Optimal Anticoagulation Through Genetics)
• GIFT (Genetics Informatics Trial of Warfarin to Prevent Venous Thrombosis)
• EU-PACT (European Pharmacogenetics of Anticoagulant Therapy-Warfarin).

We hope these studies will provide greater clarity on the clinical utility and cost-effectiveness of pharmacogenetic testing to guide warfarin dosing.

HOW SHOULD GENETIC INFORMATION BE USED TO GUIDE OR ALTER THERAPY?

Algorithms are available for estimating initial and maintenance warfarin doses based on genetic information (CYP2C9 and VKORC1), race or ethnicity, age, sex, body mass index, smoking status, and other medications taken. In addition, models incorporating the INR on day 4 and days 6 to 11 have been developed for dose refinement.¹⁵ The algorithms explain 30% to 60% of the variability of the data, with lower values for African Americans.⁷

A well-developed dosing model that includes traditional clinical factors and patient genetic status is publicly available online at www.warfarindosing.org.⁴

CPIC: A leader in applied pharmacogenetics

In late 2009, PharmGKB joined forces with the Pharmacogenomics Research Network of the National Institutes of Health to form the Clinical Pharmacogenetics Implementation Consortium (CPIC). This organization issues guidelines that are written by expert clinicians and scientists and then are peer-reviewed, published in leading journals, and simultaneously posted to the PharmGKB website along with supplemental information and updates.

CPIC’s goal is to review the current evidence and to address barriers to the adoption of pharmacogenetic testing into clinical practice. Its guidelines do not advise when or which pharmacogenetic tests should be ordered. Rather, they provide guidance on interpreting and applying such testing, should the test results be available.⁷

CPIC has guidelines on CYP2C9 and VKORC1 genotypes and warfarin dosing.⁸ If a patient’s genetic information is available, CPIC strongly recommends the use of pharmacogenetic algorithm-based dosing. If such an algorithm is not accessible, use of a genotype dosing table is recommended.⁸

Monitoring is still needed

Many factors can affect an individual’s response to warfarin above and beyond the above-noted clinical and genetic traits. These include diet, concomitant medications (both prescription and over-the-counter and herbal), and disease state. There may also be additional genetic polymorphisms not yet identified in various racial and ethnic groups that may affect dosing requirements. And as with all medications, patient compliance and dosing errors have a large potential to affect individual response. Therefore, clinicians should still be diligent about clinical monitoring.¹⁵

Most useful for initial dose

As with most pharmacogenetic information, the greatest benefit can be achieved when this information is used to guide the initial dose, although there is also some effect noted when this information is known and acted upon into the 2nd week of treatment.⁸

Patients on long-term warfarin treatment with stable doses and those unable to achieve stable dosing because of variable adherence or dietary vitamin K intake are less likely to benefit from genetic testing.

There are no published guidelines on the utility of pharmacogenetic testing if a patient is already on a stable dose of warfarin or has a known historical stable dose. There are also no published guidelines on changing the frequency of monitoring based on known genotype.

In children, the data are sparse at this time regarding the utility of pharmacogenetically informed dosing.

HOW DOES ONE ORDER TESTING, AND WHAT IS THE COST?

The FDA has approved four warfarin pharmacogenetic test kits. To be used in clinical decision-making, these tests must be done in a laboratory certified by the Clinical Laboratory Improvement Amendments (CLIA) program.

Testing typically costs a few hundred dollars and may take days for results to be returned if not available on site.¹⁵ At Cleveland Clinic, CYP2C9 and VKORC1 testing can be run in-house at a cost of about $700. Generally, many third-party payers do not reimburse for testing without a prior-approval process.

TO TEST OR NOT TO TEST

Pharmacogenetic testing is available and may help optimize warfarin dosing early in treatment, as well as help maintain therapeutic INRs more consistently. There is preliminary evidence that using this information to guide dosing improves clinical outcomes. Several large trials are under way to address additional questions of clinical utility, with results expected in the next year. There are also readily available decision-support tools to guide therapeutic dosing, and when pharmacogenetic test results are available, utilization of a warfarin dosing algorithm is recommended.

The largest barrier remaining appears to be cost (relative to perceived benefit), and until larger trials of clinical utility and cost-effectiveness are completed and analyzed, hurdles exist to obtaining coverage for such testing.

If it is readily available (and can be paid for by insurance companies or out-of-pocket) and test results can be obtained within 24 to 48 hours or before prescribing, pharmacogenetic testing can be a valuable tool to guide and manage warfarin dosing. Particularly for patients who want to be as proactive as possible, warfarin pharmacogenetic testing offers the ability to participate in this decision-making and to potentially reduce their risk of adverse drug events. And in view of the evidence and FDA recommendations, we propose that the discussion with our patients is not whether we should consider pharmacogenetic testing, but that we have considered pharmacogenetic testing, and why we have decided for or against it.