This article aims to introduce the statistical methodology behind chi-square and Fisher’s exact tests, which are commonly used in medical research to assess associations between categorical variables. This discussion will use data from a study by Mrozek¹ in patients with acute respiratory distress syndrome (ARDS). This was a multicenter, prospective, observational study: multicenter because it included data from 10 intensive care units, prospective because the study collected the data moving forward in time, and observational because the study investigators did not have control over the group assignments but rather used the naturally occurring groups. The study objective was to characterize focal and nonfocal patterns of lung computed tomography (CT)-based imaging with plasma markers of lung injury.

The primary grouping variable was type of ARDS (focal vs nonfocal) as determined by CT scans and other lung imaging tools. In this study, there were 32 (27%) patients with focal ARDS and 87 (73%) patients with nonfocal ARDS. What will be important, however, is classifying the type of variables because this determines the type of analyses performed. Type of ARDS is a categorical variable with 2 levels.

The primary study endpoint was plasma levels of the soluble form of the receptor for advanced glycation end product. There were also a number of secondary study endpoints that can be grouped as either patient outcomes or biomarkers. Patient outcomes included the duration of mechanical ventilation and both 28- and 90-day mortality. Levels of other biomarkers included surfactant protein D, soluble intercellular adhesion molecule-1, and plasminogen activator inhibitor-1.

This article focused on the secondary outcome of 90-day mortality beginning at disease onset. Again, we are interested in classifying this variable, which is categorical with 2 levels (yes vs no). So the scenario is that we want to assess the relationship between the type of ARDS (focal vs nonfocal) and 90-day mortality (yes vs no). In its most basic form, this scenario is an investigation into the association among 2 categorical variables.

When there are 2 categorical variables, the data can be arranged in what is called a contingency table (Figure 1). Because both variables are binary (2 levels), it is called a 2 × 2 table. However, a contingency table can be generated for 2 categorical variables with any number of levels—in that case, it is called an r ×c table, where r is the number of levels for the row variable and c is the number of levels for the column variable. The actual raw counts or frequencies are recorded inside the table cells. The cell counts are often referred to as observed counts and thus the notation (O_ij) is used. The subscript i identifies the specific level of the row variable, and in this example it can equal 1 or 2 since the row variable is binary. Similarly, the subscript j identifies the specific level of the column variable and in this example it can equal 1 or 2 since the column variable is binary. Therefore, O₁₁ represents the number of patients who have the row variable = level 1 and the column variable = level 1.

In addition to the row and column variable cells, there are also the margin totals. These totals are either the row margin total (summing across the row) or the column margin total (summing down the column). For example, n₁₊ is the sum of the row where the row variable equal 1 (O₁₁ + O₁₂ = n₁₊). Finally, at the very bottom right corner is the grand total, which equals the sample size. 

The goal is to test whether or not these 2 categorical variables are associated with each other. The null hypothesis (H_o) is that there is no association between these 2 categorical variables and the alternative hypotheses (H_a) is that there is an association between these 2 categorical variables.

The next step is to translate the generic form of the hypotheses into hypotheses that are specific to the research question. In this case, the null hypothesis is that mortality is not associated with lung morphology and the alternative hypothesis is that mortality is associated with lung morphology.

The contingency table cells can be populated with the numbers found in the article. It has our outcome of focus—mortality at day 90—both the count and the percent. The results are broken down by type of ARDS (focal vs nonfocal) as follows:

• Focal ARDS = 6 patients (21.4%)
• Nonfocal ARDS = 35 patients (45.5%).

First, the row variable is lung morphology, and it has two levels (focal vs nonfocal). Next, the column variable is 90-day mortality and it has 2 levels (yes vs no). Finally, the table must be populated, but be careful not to assume that there are no missing data. Begin with the cell counts: there were 6 focal ARDS patients and 35 nonfocal ARDS patients who died within 90 days. These two numbers populate the first column and result in a column total of 41. Next, use the reported percentages to calculate the row totals. Six is 21.4% of 28, so the first row total is 28. Thirty-five is 45.5% of 77, so the second row total is 77. If there are 28 patients with focal ARDS and 77 with nonfocal ARDS, then the grand total is 28 + 77 = 105. The remaining values can be obtained by subtraction. If there are 105 total patients and 41 die within 90 days, then 105 − 41 = 64 patients who do not die within 90 days and this is the second column total. Similarly, if there are 28 focal ARDS patients and 6 die within 90 days, then 28 − 6 = 22 patients who do not die within 90 days. Lastly, if there are 77 nonfocal ARDS patients and 35 die within 90 days, then 77 − 35 = 42 patients who do not die within 90 days. Now the contingency table is complete.

Once the contingency table is built, the question becomes, “Is lung morphology associated with 90-day mortality?” To answer that question, we need to know how many patients one would expect in each table cell if the null hypothesis of no association is true. When conducting a hypothesis test, one always assumes that the null hypothesis is true and then gathers data to see how well the data aligns with that assumption.

So one must calculate how many patients to expect in each of these cells if lung morphology is not associated with 90-day mortality. One way to address this question is to ask these 2 questions:

(1) Overall, what proportion of patients die by day 90? Looking at the constructed contingency table, that answer would be 39%. This was calculated by taking the total number of patients who died by day 90 and dividing it by the total number of patients, 41/105 = 39%. This gives the overall proportion, based on the data, who would die by day 90.

(2) How many of the focal ARDS patients would be expected to die by day 90? Now it is not overall, but rather we are limiting the question to the focal ARDS group. To obtain the answer, multiply the overall proportion of patients who die by day 90 by how many focal ARDS patients are in the study. Essentially, take the answer from the previous question and multiply it by the total number of focal ARDS, which is 28. The result is (41/105) × 28 = 10.9. Thus, if there is no association among long morphology and 90-day mortality, one would expect 10.9 focal ARDS patients to die by day 90.

Now 10.9 is a very specific answer for a specific contingency table, but the answer could be written in general terms. Basically, 3 numbers were used in calculating the solution: the row margin, the column margin, and the grand total. The general formula is the following:

The notation E_ij is used to represent the expected count assuming the null hypothesis of no association among the row and column variables is true. To calculate the expected count, take the i^th row total times the j^th column total and divide by the grand total.

In the lung morphology and mortality example, what is the expected number of deaths within 90 days among the nonfocal ARDS patients? This is the second row and the first column (E₂₁). Applying the formula, one multiplies the total for the second row by the total for the first column and then divides by the grand total, (77 × 41)/105 = 30.1. This calculation is repeated for each of the 4 cells.

Because we now know the observed cell count and the expected cell count (under the null hypothesis), we can compare the observed and expected counts to see how well the data aligns with the null hypothesis. This is what the chi-square test does, and the test statistic is calculated as follows:

The sigma (Σ) means addition, so the calculation is performed on each individual cell in the contingency table and then the results are summed. A 2 × 2 table has 4 cells and thus 4 numbers will be summed. For each cell, the formula compares the observed to the expected. Basically, it computes how similar they are (that is the O minus E part). Because the differences will be positive for some cells and negative for others, the differences are squared to avoid cancellation when you add them. Finally, each squared difference is divided by the expected count to standardize the calculation.

Intuitively, if the observed counts (O_ij) are similar to the expected counts under the null hypothesis (E_ij), then these 2 numbers will be very close to each other. When taking the difference between them or subtracting them, the result is a small number. When squaring a small number, one obtains a really small number. And adding up a bunch of really small numbers results in a small number. So the test statistic is going to be small. That means that the resulting P value is going to be large. What is a P value? Think of it as an index of compatibility. How compatible is the data with the null hypothesis? Here, you get a large index of compatibility. That means that the data aligns nicely with the null hypothesis and one fails to reject the null.

Now, think about the alternative scenario. If the observed counts (O_ij) are wildly different from the expected counts under the null hypothesis (E_ij), then these 2 numbers will be quite different. When taking the difference between them or subtracting them, the result is a big number. When squaring a big number, one obtains a really big number, and adding up a bunch of really big numbers results in a large number. So the test statistic is going to be large. That means that the resulting P value is going to be small. And if you think of a P value as an index of compatibility, the data and the null hypothesis are not very compatible. That means that the data does not align nicely with the null hypothesis and one rejects the null. This is the general idea of the chi-square test. It assesses how compatible the data is with the null hypothesis that the 2 categorical variables are not associated.

To obtain the actual P value, the distribution of the test statistic (under the null hypothesis) is used to calculate the area under the curve for values equal to the test statistic or more extreme. The described test statistic has an approximate chi-square distribution with (r − 1)(c − 1) degree of freedom. Recall that r is the number of levels of the row variable and c is the number of levels of the column variable. Our example is a 2 × 2 table, so the test statistic has an approximate chi-square distribution with (2 − 1)(2 − 1) = 1 degree of freedom.

Now that the chi-square test has been fully described, the assumptions for the test must be discussed. It is important to know when you should or should not perform this test. The chi-square test assumes that observations are independent. This means that the outcome for one observation is not associated with the outcome of any other observation. This principle can be violated when multiple measurements are taken over time or when multiple measurements are taken from one patient.

Another assumption is that the chi-square large sample approximation just described is appropriate. In other words, no more than 20% of the expected counts (E_ij) are less than 5. For a 2 × 2 table, how many cells do you have? Four. So if even one of those 4 happens to have an expected count less than 5, this assumption is violated. For a 2 × 2 table, none of the expected counts can be less than 5.

Returning to the lung morphology and mortality example, were the assumptions met? The data consist of 105 unique patients. Thus, we can assume that they are independent. The minimum expected count was 10.9, which is not less than 5. Therefore, the assumptions for the chi-square test are met. Next, the test statistic is calculated using the observed and expected counts. For each cell, subtract the expected count from the observed count, square it, and divide by the expected count. Then, add the 4 resulting numbers to obtain the test statistic of 4.92.

Finally, compute the area under the chi-square distribution with 1 degree of freedom Χ²₍₁₎, at the test statistic and values more extreme. In this case, values more extreme are values greater than the test statistic. Here, the area under the curve to the right of 4.92 is .027 (Figure 3). This is the P value, which indicates that the data and the null hypothesis have very low compatibility. In this example, the area under the curve to the right of 4.92 is .027 (Figure 3). This is the P value, which indicates that the data and the null hypothesis have very low compatibility. Thus, the decision is to reject the null hypothesis. The conclusion is that lung morphology is associated with 90-day mortality (P = .027). To describe that association, one looks at the contingency table and finds a reduction in 90-day mortality with focal patterns compared to nonfocal patterns (21.4% vs 45.5%, respectively). The P value reported in the article is .026. Our hand calculation was .027, which is slightly off due to rounding. In summary, the scenario is an investigation into the association among 2 categorical variables, and, thus, a test to consider is the chi-square test, if assumptions are met.

In another example in the same study, the authors investigate whether any baseline characteristics are associated with lung morphology. For example, is neurology, specifically Parkinson disease (yes vs no), associated with lung morphology (focal vs nonfocal)? Again, the scenario is an investigation into the association between 2 categorical variables, so a chi-square test should be considered.

To start, build a contingency table arbitrarily placing lung morphology as the row variable and Parkinson disease as the column variable. Populate the contingency table based on the counts and percentages reported in the article (Figure 4). Next, check that the assumptions of the chi-square test are met. Are the observations independent? Again, because these are unique patients, we consider this assumption met. Since this is a 2 × 2 table, are all of the expected counts greater than 5? Calculations of the expected counts obtained the following: 1.1, 30.9, 2.9 and 84.1. Here, 2 of the 4 expected counts are less than 5. Therefore, methods that use large sample approximation, like the chi-squared test, may not be an appropriate choice.

Instead of using methodology that is an approximation, consider an exact test such as Fisher’s exact test. Again, refer to the contingency table where Fisher’s exact is going to calculate the exact probability (under the null hypothesis) of the observed data or results more extreme. This is the technical definition of a P value. It is, however, still quantifying how compatible the data are with the null hypothesis. The exact probability of a particular contingency table can be obtained using the hypergeometric distribution.

The symbols that resemble large parentheses are notations for a combinatorial. Because using combinatorials to calculate the probability is not user friendly, an equivalent version relies on factorials instead. Both techniques are presented above. Remember that the goal is to find the exact probability of the observed data or something more extreme.

The hypotheses are still testing whether these 2 categorical variables are associated with each other. In this particular example, we test if the proportion of patients with Parkinson disease is the same in the focal and nonfocal groups. Fisher’s exact test obtains its two-tailed P value by computing the probabilities associated with all possible tables that have the same row and column totals. Then, it identifies the alternative tables with a probability that is less than that of the observed table. Finally, it adds the probability of the observed table with the sum of the probabilities of each alternative table identified above, which results in the P value.

To explore each of those steps in detail, one must first enumerate how many tables can be built that all have the same row and column totals as the observed table. Figure 5 shows the 5 possible tables. Pick any one of the 5 2 × 2 tables; the margins are fixed. Each table has the same row totals, 32 focal and 87 nonfocal, and each table has the same column totals: 4 Parkinson and 115 non-Parkinson. Then, for each table, calculate the probability of that table. Figure 5 shows this calculation for the first 2 × 2 table, which happens to be the observed table. The probability of the table observed in the study is .2803. Such a calculation is performed on each of the other tables.

Next, one must identify the tables that have a probability smaller than the observed table. Here, we are looking for probabilities less than .2803. These are the tables deemed more extreme. Tables 3, 4, and 5 have probabilities less than .2803.

The final step is to sum the probability of the observed table and the more extreme tables (ie, those with probabilities < the observed table) (.2803 + .2337 + .0543 + .0045 = .5728). Thus, the resulting rounded P value is .57, which indicates a high level of compatibility between the data and the null hypothesis of no association. The decision is to fail to reject the null hypothesis and the conclusion is that the evidence does not support an association among lung morphology and Parkinson disease. In other words, there is insufficient evidence to claim that the proportion of Parkinson disease differs between the focal and nonfocal ARDS patients (0% vs 5%, P = .57). This matches the P value reported by Mrozek for this association.

The first objective of this article was to identify scenarios in which a chi-square or Fisher’s exact test should be considered. The general setting discussed was an investigation of the association between two categorical variables. Use of each test specifically depends on whether the assumptions have been met. Both of the examples used in our discussion happened to be binary, but that is not a restriction. Categorical variables can have more than 2 levels. All of the methods demonstrated for 2 × 2 tables can be generalized to r × c tables. 

The second objective of this article was to recognize when test assumptions have been violated. For simplicity, most researchers adhere to the following: if ≤ 20% of expected cell counts are less than 5, then use the chi-square test; if > 20% of expected cell counts are less than 5, then use Fisher’s exact test. Both methods assume that the observations are independent. Could one use the exact test when the chi-square assumptions are met? Yes, but it is more computationally expensive as it uses all possible fixed margin tables and their probabilities. If the chi-square assumptions are met, then the sample size is typically larger and these calculations become numerous. Also, it does not have to be that large of a sample for the chi-square to be a good approximation and do it very quickly.

The final objective of this article was to test claims made regarding the association of 2 independent categorical variables. We included examples from the medical literature showing step-by-step calculations of both the large sample approximation (chi-square) and exact (Fisher’s) methodologies providing insight into how these tests are conducted as well as when they are appropriate.

This article aims to introduce the statistical methodology behind chi-square and Fisher’s exact tests, which are commonly used in medical research to assess associations between categorical variables. This discussion will use data from a study by Mrozek¹ in patients with acute respiratory distress syndrome (ARDS). This was a multicenter, prospective, observational study: multicenter because it included data from 10 intensive care units, prospective because the study collected the data moving forward in time, and observational because the study investigators did not have control over the group assignments but rather used the naturally occurring groups. The study objective was to characterize focal and nonfocal patterns of lung computed tomography (CT)-based imaging with plasma markers of lung injury.

The primary grouping variable was type of ARDS (focal vs nonfocal) as determined by CT scans and other lung imaging tools. In this study, there were 32 (27%) patients with focal ARDS and 87 (73%) patients with nonfocal ARDS. What will be important, however, is classifying the type of variables because this determines the type of analyses performed. Type of ARDS is a categorical variable with 2 levels.

The primary study endpoint was plasma levels of the soluble form of the receptor for advanced glycation end product. There were also a number of secondary study endpoints that can be grouped as either patient outcomes or biomarkers. Patient outcomes included the duration of mechanical ventilation and both 28- and 90-day mortality. Levels of other biomarkers included surfactant protein D, soluble intercellular adhesion molecule-1, and plasminogen activator inhibitor-1.

This article focused on the secondary outcome of 90-day mortality beginning at disease onset. Again, we are interested in classifying this variable, which is categorical with 2 levels (yes vs no). So the scenario is that we want to assess the relationship between the type of ARDS (focal vs nonfocal) and 90-day mortality (yes vs no). In its most basic form, this scenario is an investigation into the association among 2 categorical variables.

When there are 2 categorical variables, the data can be arranged in what is called a contingency table (Figure 1). Because both variables are binary (2 levels), it is called a 2 × 2 table. However, a contingency table can be generated for 2 categorical variables with any number of levels—in that case, it is called an r ×c table, where r is the number of levels for the row variable and c is the number of levels for the column variable. The actual raw counts or frequencies are recorded inside the table cells. The cell counts are often referred to as observed counts and thus the notation (O_ij) is used. The subscript i identifies the specific level of the row variable, and in this example it can equal 1 or 2 since the row variable is binary. Similarly, the subscript j identifies the specific level of the column variable and in this example it can equal 1 or 2 since the column variable is binary. Therefore, O₁₁ represents the number of patients who have the row variable = level 1 and the column variable = level 1.

In addition to the row and column variable cells, there are also the margin totals. These totals are either the row margin total (summing across the row) or the column margin total (summing down the column). For example, n₁₊ is the sum of the row where the row variable equal 1 (O₁₁ + O₁₂ = n₁₊). Finally, at the very bottom right corner is the grand total, which equals the sample size. 

The goal is to test whether or not these 2 categorical variables are associated with each other. The null hypothesis (H_o) is that there is no association between these 2 categorical variables and the alternative hypotheses (H_a) is that there is an association between these 2 categorical variables.

The next step is to translate the generic form of the hypotheses into hypotheses that are specific to the research question. In this case, the null hypothesis is that mortality is not associated with lung morphology and the alternative hypothesis is that mortality is associated with lung morphology.

The contingency table cells can be populated with the numbers found in the article. It has our outcome of focus—mortality at day 90—both the count and the percent. The results are broken down by type of ARDS (focal vs nonfocal) as follows:

• Focal ARDS = 6 patients (21.4%)
• Nonfocal ARDS = 35 patients (45.5%).

First, the row variable is lung morphology, and it has two levels (focal vs nonfocal). Next, the column variable is 90-day mortality and it has 2 levels (yes vs no). Finally, the table must be populated, but be careful not to assume that there are no missing data. Begin with the cell counts: there were 6 focal ARDS patients and 35 nonfocal ARDS patients who died within 90 days. These two numbers populate the first column and result in a column total of 41. Next, use the reported percentages to calculate the row totals. Six is 21.4% of 28, so the first row total is 28. Thirty-five is 45.5% of 77, so the second row total is 77. If there are 28 patients with focal ARDS and 77 with nonfocal ARDS, then the grand total is 28 + 77 = 105. The remaining values can be obtained by subtraction. If there are 105 total patients and 41 die within 90 days, then 105 − 41 = 64 patients who do not die within 90 days and this is the second column total. Similarly, if there are 28 focal ARDS patients and 6 die within 90 days, then 28 − 6 = 22 patients who do not die within 90 days. Lastly, if there are 77 nonfocal ARDS patients and 35 die within 90 days, then 77 − 35 = 42 patients who do not die within 90 days. Now the contingency table is complete.

Once the contingency table is built, the question becomes, “Is lung morphology associated with 90-day mortality?” To answer that question, we need to know how many patients one would expect in each table cell if the null hypothesis of no association is true. When conducting a hypothesis test, one always assumes that the null hypothesis is true and then gathers data to see how well the data aligns with that assumption.

So one must calculate how many patients to expect in each of these cells if lung morphology is not associated with 90-day mortality. One way to address this question is to ask these 2 questions:

(1) Overall, what proportion of patients die by day 90? Looking at the constructed contingency table, that answer would be 39%. This was calculated by taking the total number of patients who died by day 90 and dividing it by the total number of patients, 41/105 = 39%. This gives the overall proportion, based on the data, who would die by day 90.

(2) How many of the focal ARDS patients would be expected to die by day 90? Now it is not overall, but rather we are limiting the question to the focal ARDS group. To obtain the answer, multiply the overall proportion of patients who die by day 90 by how many focal ARDS patients are in the study. Essentially, take the answer from the previous question and multiply it by the total number of focal ARDS, which is 28. The result is (41/105) × 28 = 10.9. Thus, if there is no association among long morphology and 90-day mortality, one would expect 10.9 focal ARDS patients to die by day 90.

Now 10.9 is a very specific answer for a specific contingency table, but the answer could be written in general terms. Basically, 3 numbers were used in calculating the solution: the row margin, the column margin, and the grand total. The general formula is the following:

The notation E_ij is used to represent the expected count assuming the null hypothesis of no association among the row and column variables is true. To calculate the expected count, take the i^th row total times the j^th column total and divide by the grand total.

In the lung morphology and mortality example, what is the expected number of deaths within 90 days among the nonfocal ARDS patients? This is the second row and the first column (E₂₁). Applying the formula, one multiplies the total for the second row by the total for the first column and then divides by the grand total, (77 × 41)/105 = 30.1. This calculation is repeated for each of the 4 cells.

Because we now know the observed cell count and the expected cell count (under the null hypothesis), we can compare the observed and expected counts to see how well the data aligns with the null hypothesis. This is what the chi-square test does, and the test statistic is calculated as follows:

The sigma (Σ) means addition, so the calculation is performed on each individual cell in the contingency table and then the results are summed. A 2 × 2 table has 4 cells and thus 4 numbers will be summed. For each cell, the formula compares the observed to the expected. Basically, it computes how similar they are (that is the O minus E part). Because the differences will be positive for some cells and negative for others, the differences are squared to avoid cancellation when you add them. Finally, each squared difference is divided by the expected count to standardize the calculation.

Intuitively, if the observed counts (O_ij) are similar to the expected counts under the null hypothesis (E_ij), then these 2 numbers will be very close to each other. When taking the difference between them or subtracting them, the result is a small number. When squaring a small number, one obtains a really small number. And adding up a bunch of really small numbers results in a small number. So the test statistic is going to be small. That means that the resulting P value is going to be large. What is a P value? Think of it as an index of compatibility. How compatible is the data with the null hypothesis? Here, you get a large index of compatibility. That means that the data aligns nicely with the null hypothesis and one fails to reject the null.

Now, think about the alternative scenario. If the observed counts (O_ij) are wildly different from the expected counts under the null hypothesis (E_ij), then these 2 numbers will be quite different. When taking the difference between them or subtracting them, the result is a big number. When squaring a big number, one obtains a really big number, and adding up a bunch of really big numbers results in a large number. So the test statistic is going to be large. That means that the resulting P value is going to be small. And if you think of a P value as an index of compatibility, the data and the null hypothesis are not very compatible. That means that the data does not align nicely with the null hypothesis and one rejects the null. This is the general idea of the chi-square test. It assesses how compatible the data is with the null hypothesis that the 2 categorical variables are not associated.

To obtain the actual P value, the distribution of the test statistic (under the null hypothesis) is used to calculate the area under the curve for values equal to the test statistic or more extreme. The described test statistic has an approximate chi-square distribution with (r − 1)(c − 1) degree of freedom. Recall that r is the number of levels of the row variable and c is the number of levels of the column variable. Our example is a 2 × 2 table, so the test statistic has an approximate chi-square distribution with (2 − 1)(2 − 1) = 1 degree of freedom.

Now that the chi-square test has been fully described, the assumptions for the test must be discussed. It is important to know when you should or should not perform this test. The chi-square test assumes that observations are independent. This means that the outcome for one observation is not associated with the outcome of any other observation. This principle can be violated when multiple measurements are taken over time or when multiple measurements are taken from one patient.

Another assumption is that the chi-square large sample approximation just described is appropriate. In other words, no more than 20% of the expected counts (E_ij) are less than 5. For a 2 × 2 table, how many cells do you have? Four. So if even one of those 4 happens to have an expected count less than 5, this assumption is violated. For a 2 × 2 table, none of the expected counts can be less than 5.

Returning to the lung morphology and mortality example, were the assumptions met? The data consist of 105 unique patients. Thus, we can assume that they are independent. The minimum expected count was 10.9, which is not less than 5. Therefore, the assumptions for the chi-square test are met. Next, the test statistic is calculated using the observed and expected counts. For each cell, subtract the expected count from the observed count, square it, and divide by the expected count. Then, add the 4 resulting numbers to obtain the test statistic of 4.92.

Finally, compute the area under the chi-square distribution with 1 degree of freedom Χ²₍₁₎, at the test statistic and values more extreme. In this case, values more extreme are values greater than the test statistic. Here, the area under the curve to the right of 4.92 is .027 (Figure 3). This is the P value, which indicates that the data and the null hypothesis have very low compatibility. In this example, the area under the curve to the right of 4.92 is .027 (Figure 3). This is the P value, which indicates that the data and the null hypothesis have very low compatibility. Thus, the decision is to reject the null hypothesis. The conclusion is that lung morphology is associated with 90-day mortality (P = .027). To describe that association, one looks at the contingency table and finds a reduction in 90-day mortality with focal patterns compared to nonfocal patterns (21.4% vs 45.5%, respectively). The P value reported in the article is .026. Our hand calculation was .027, which is slightly off due to rounding. In summary, the scenario is an investigation into the association among 2 categorical variables, and, thus, a test to consider is the chi-square test, if assumptions are met.

In another example in the same study, the authors investigate whether any baseline characteristics are associated with lung morphology. For example, is neurology, specifically Parkinson disease (yes vs no), associated with lung morphology (focal vs nonfocal)? Again, the scenario is an investigation into the association between 2 categorical variables, so a chi-square test should be considered.

To start, build a contingency table arbitrarily placing lung morphology as the row variable and Parkinson disease as the column variable. Populate the contingency table based on the counts and percentages reported in the article (Figure 4). Next, check that the assumptions of the chi-square test are met. Are the observations independent? Again, because these are unique patients, we consider this assumption met. Since this is a 2 × 2 table, are all of the expected counts greater than 5? Calculations of the expected counts obtained the following: 1.1, 30.9, 2.9 and 84.1. Here, 2 of the 4 expected counts are less than 5. Therefore, methods that use large sample approximation, like the chi-squared test, may not be an appropriate choice.

Instead of using methodology that is an approximation, consider an exact test such as Fisher’s exact test. Again, refer to the contingency table where Fisher’s exact is going to calculate the exact probability (under the null hypothesis) of the observed data or results more extreme. This is the technical definition of a P value. It is, however, still quantifying how compatible the data are with the null hypothesis. The exact probability of a particular contingency table can be obtained using the hypergeometric distribution.

The symbols that resemble large parentheses are notations for a combinatorial. Because using combinatorials to calculate the probability is not user friendly, an equivalent version relies on factorials instead. Both techniques are presented above. Remember that the goal is to find the exact probability of the observed data or something more extreme.

The hypotheses are still testing whether these 2 categorical variables are associated with each other. In this particular example, we test if the proportion of patients with Parkinson disease is the same in the focal and nonfocal groups. Fisher’s exact test obtains its two-tailed P value by computing the probabilities associated with all possible tables that have the same row and column totals. Then, it identifies the alternative tables with a probability that is less than that of the observed table. Finally, it adds the probability of the observed table with the sum of the probabilities of each alternative table identified above, which results in the P value.

To explore each of those steps in detail, one must first enumerate how many tables can be built that all have the same row and column totals as the observed table. Figure 5 shows the 5 possible tables. Pick any one of the 5 2 × 2 tables; the margins are fixed. Each table has the same row totals, 32 focal and 87 nonfocal, and each table has the same column totals: 4 Parkinson and 115 non-Parkinson. Then, for each table, calculate the probability of that table. Figure 5 shows this calculation for the first 2 × 2 table, which happens to be the observed table. The probability of the table observed in the study is .2803. Such a calculation is performed on each of the other tables.

Next, one must identify the tables that have a probability smaller than the observed table. Here, we are looking for probabilities less than .2803. These are the tables deemed more extreme. Tables 3, 4, and 5 have probabilities less than .2803.

The final step is to sum the probability of the observed table and the more extreme tables (ie, those with probabilities < the observed table) (.2803 + .2337 + .0543 + .0045 = .5728). Thus, the resulting rounded P value is .57, which indicates a high level of compatibility between the data and the null hypothesis of no association. The decision is to fail to reject the null hypothesis and the conclusion is that the evidence does not support an association among lung morphology and Parkinson disease. In other words, there is insufficient evidence to claim that the proportion of Parkinson disease differs between the focal and nonfocal ARDS patients (0% vs 5%, P = .57). This matches the P value reported by Mrozek for this association.

The first objective of this article was to identify scenarios in which a chi-square or Fisher’s exact test should be considered. The general setting discussed was an investigation of the association between two categorical variables. Use of each test specifically depends on whether the assumptions have been met. Both of the examples used in our discussion happened to be binary, but that is not a restriction. Categorical variables can have more than 2 levels. All of the methods demonstrated for 2 × 2 tables can be generalized to r × c tables. 

The second objective of this article was to recognize when test assumptions have been violated. For simplicity, most researchers adhere to the following: if ≤ 20% of expected cell counts are less than 5, then use the chi-square test; if > 20% of expected cell counts are less than 5, then use Fisher’s exact test. Both methods assume that the observations are independent. Could one use the exact test when the chi-square assumptions are met? Yes, but it is more computationally expensive as it uses all possible fixed margin tables and their probabilities. If the chi-square assumptions are met, then the sample size is typically larger and these calculations become numerous. Also, it does not have to be that large of a sample for the chi-square to be a good approximation and do it very quickly.

The final objective of this article was to test claims made regarding the association of 2 independent categorical variables. We included examples from the medical literature showing step-by-step calculations of both the large sample approximation (chi-square) and exact (Fisher’s) methodologies providing insight into how these tests are conducted as well as when they are appropriate.

This article aims to introduce the statistical methodology behind chi-square and Fisher’s exact tests, which are commonly used in medical research to assess associations between categorical variables. This discussion will use data from a study by Mrozek¹ in patients with acute respiratory distress syndrome (ARDS). This was a multicenter, prospective, observational study: multicenter because it included data from 10 intensive care units, prospective because the study collected the data moving forward in time, and observational because the study investigators did not have control over the group assignments but rather used the naturally occurring groups. The study objective was to characterize focal and nonfocal patterns of lung computed tomography (CT)-based imaging with plasma markers of lung injury.

The primary grouping variable was type of ARDS (focal vs nonfocal) as determined by CT scans and other lung imaging tools. In this study, there were 32 (27%) patients with focal ARDS and 87 (73%) patients with nonfocal ARDS. What will be important, however, is classifying the type of variables because this determines the type of analyses performed. Type of ARDS is a categorical variable with 2 levels.

The primary study endpoint was plasma levels of the soluble form of the receptor for advanced glycation end product. There were also a number of secondary study endpoints that can be grouped as either patient outcomes or biomarkers. Patient outcomes included the duration of mechanical ventilation and both 28- and 90-day mortality. Levels of other biomarkers included surfactant protein D, soluble intercellular adhesion molecule-1, and plasminogen activator inhibitor-1.

This article focused on the secondary outcome of 90-day mortality beginning at disease onset. Again, we are interested in classifying this variable, which is categorical with 2 levels (yes vs no). So the scenario is that we want to assess the relationship between the type of ARDS (focal vs nonfocal) and 90-day mortality (yes vs no). In its most basic form, this scenario is an investigation into the association among 2 categorical variables.

When there are 2 categorical variables, the data can be arranged in what is called a contingency table (Figure 1). Because both variables are binary (2 levels), it is called a 2 × 2 table. However, a contingency table can be generated for 2 categorical variables with any number of levels—in that case, it is called an r ×c table, where r is the number of levels for the row variable and c is the number of levels for the column variable. The actual raw counts or frequencies are recorded inside the table cells. The cell counts are often referred to as observed counts and thus the notation (O_ij) is used. The subscript i identifies the specific level of the row variable, and in this example it can equal 1 or 2 since the row variable is binary. Similarly, the subscript j identifies the specific level of the column variable and in this example it can equal 1 or 2 since the column variable is binary. Therefore, O₁₁ represents the number of patients who have the row variable = level 1 and the column variable = level 1.

In addition to the row and column variable cells, there are also the margin totals. These totals are either the row margin total (summing across the row) or the column margin total (summing down the column). For example, n₁₊ is the sum of the row where the row variable equal 1 (O₁₁ + O₁₂ = n₁₊). Finally, at the very bottom right corner is the grand total, which equals the sample size. 

The goal is to test whether or not these 2 categorical variables are associated with each other. The null hypothesis (H_o) is that there is no association between these 2 categorical variables and the alternative hypotheses (H_a) is that there is an association between these 2 categorical variables.

The next step is to translate the generic form of the hypotheses into hypotheses that are specific to the research question. In this case, the null hypothesis is that mortality is not associated with lung morphology and the alternative hypothesis is that mortality is associated with lung morphology.

The contingency table cells can be populated with the numbers found in the article. It has our outcome of focus—mortality at day 90—both the count and the percent. The results are broken down by type of ARDS (focal vs nonfocal) as follows:

• Focal ARDS = 6 patients (21.4%)
• Nonfocal ARDS = 35 patients (45.5%).

First, the row variable is lung morphology, and it has two levels (focal vs nonfocal). Next, the column variable is 90-day mortality and it has 2 levels (yes vs no). Finally, the table must be populated, but be careful not to assume that there are no missing data. Begin with the cell counts: there were 6 focal ARDS patients and 35 nonfocal ARDS patients who died within 90 days. These two numbers populate the first column and result in a column total of 41. Next, use the reported percentages to calculate the row totals. Six is 21.4% of 28, so the first row total is 28. Thirty-five is 45.5% of 77, so the second row total is 77. If there are 28 patients with focal ARDS and 77 with nonfocal ARDS, then the grand total is 28 + 77 = 105. The remaining values can be obtained by subtraction. If there are 105 total patients and 41 die within 90 days, then 105 − 41 = 64 patients who do not die within 90 days and this is the second column total. Similarly, if there are 28 focal ARDS patients and 6 die within 90 days, then 28 − 6 = 22 patients who do not die within 90 days. Lastly, if there are 77 nonfocal ARDS patients and 35 die within 90 days, then 77 − 35 = 42 patients who do not die within 90 days. Now the contingency table is complete.

Once the contingency table is built, the question becomes, “Is lung morphology associated with 90-day mortality?” To answer that question, we need to know how many patients one would expect in each table cell if the null hypothesis of no association is true. When conducting a hypothesis test, one always assumes that the null hypothesis is true and then gathers data to see how well the data aligns with that assumption.

So one must calculate how many patients to expect in each of these cells if lung morphology is not associated with 90-day mortality. One way to address this question is to ask these 2 questions:

(1) Overall, what proportion of patients die by day 90? Looking at the constructed contingency table, that answer would be 39%. This was calculated by taking the total number of patients who died by day 90 and dividing it by the total number of patients, 41/105 = 39%. This gives the overall proportion, based on the data, who would die by day 90.

(2) How many of the focal ARDS patients would be expected to die by day 90? Now it is not overall, but rather we are limiting the question to the focal ARDS group. To obtain the answer, multiply the overall proportion of patients who die by day 90 by how many focal ARDS patients are in the study. Essentially, take the answer from the previous question and multiply it by the total number of focal ARDS, which is 28. The result is (41/105) × 28 = 10.9. Thus, if there is no association among long morphology and 90-day mortality, one would expect 10.9 focal ARDS patients to die by day 90.

Now 10.9 is a very specific answer for a specific contingency table, but the answer could be written in general terms. Basically, 3 numbers were used in calculating the solution: the row margin, the column margin, and the grand total. The general formula is the following:

The notation E_ij is used to represent the expected count assuming the null hypothesis of no association among the row and column variables is true. To calculate the expected count, take the i^th row total times the j^th column total and divide by the grand total.

In the lung morphology and mortality example, what is the expected number of deaths within 90 days among the nonfocal ARDS patients? This is the second row and the first column (E₂₁). Applying the formula, one multiplies the total for the second row by the total for the first column and then divides by the grand total, (77 × 41)/105 = 30.1. This calculation is repeated for each of the 4 cells.

Because we now know the observed cell count and the expected cell count (under the null hypothesis), we can compare the observed and expected counts to see how well the data aligns with the null hypothesis. This is what the chi-square test does, and the test statistic is calculated as follows:

The sigma (Σ) means addition, so the calculation is performed on each individual cell in the contingency table and then the results are summed. A 2 × 2 table has 4 cells and thus 4 numbers will be summed. For each cell, the formula compares the observed to the expected. Basically, it computes how similar they are (that is the O minus E part). Because the differences will be positive for some cells and negative for others, the differences are squared to avoid cancellation when you add them. Finally, each squared difference is divided by the expected count to standardize the calculation.

Intuitively, if the observed counts (O_ij) are similar to the expected counts under the null hypothesis (E_ij), then these 2 numbers will be very close to each other. When taking the difference between them or subtracting them, the result is a small number. When squaring a small number, one obtains a really small number. And adding up a bunch of really small numbers results in a small number. So the test statistic is going to be small. That means that the resulting P value is going to be large. What is a P value? Think of it as an index of compatibility. How compatible is the data with the null hypothesis? Here, you get a large index of compatibility. That means that the data aligns nicely with the null hypothesis and one fails to reject the null.

Now, think about the alternative scenario. If the observed counts (O_ij) are wildly different from the expected counts under the null hypothesis (E_ij), then these 2 numbers will be quite different. When taking the difference between them or subtracting them, the result is a big number. When squaring a big number, one obtains a really big number, and adding up a bunch of really big numbers results in a large number. So the test statistic is going to be large. That means that the resulting P value is going to be small. And if you think of a P value as an index of compatibility, the data and the null hypothesis are not very compatible. That means that the data does not align nicely with the null hypothesis and one rejects the null. This is the general idea of the chi-square test. It assesses how compatible the data is with the null hypothesis that the 2 categorical variables are not associated.

To obtain the actual P value, the distribution of the test statistic (under the null hypothesis) is used to calculate the area under the curve for values equal to the test statistic or more extreme. The described test statistic has an approximate chi-square distribution with (r − 1)(c − 1) degree of freedom. Recall that r is the number of levels of the row variable and c is the number of levels of the column variable. Our example is a 2 × 2 table, so the test statistic has an approximate chi-square distribution with (2 − 1)(2 − 1) = 1 degree of freedom.

Now that the chi-square test has been fully described, the assumptions for the test must be discussed. It is important to know when you should or should not perform this test. The chi-square test assumes that observations are independent. This means that the outcome for one observation is not associated with the outcome of any other observation. This principle can be violated when multiple measurements are taken over time or when multiple measurements are taken from one patient.

Another assumption is that the chi-square large sample approximation just described is appropriate. In other words, no more than 20% of the expected counts (E_ij) are less than 5. For a 2 × 2 table, how many cells do you have? Four. So if even one of those 4 happens to have an expected count less than 5, this assumption is violated. For a 2 × 2 table, none of the expected counts can be less than 5.

Returning to the lung morphology and mortality example, were the assumptions met? The data consist of 105 unique patients. Thus, we can assume that they are independent. The minimum expected count was 10.9, which is not less than 5. Therefore, the assumptions for the chi-square test are met. Next, the test statistic is calculated using the observed and expected counts. For each cell, subtract the expected count from the observed count, square it, and divide by the expected count. Then, add the 4 resulting numbers to obtain the test statistic of 4.92.

Finally, compute the area under the chi-square distribution with 1 degree of freedom Χ²₍₁₎, at the test statistic and values more extreme. In this case, values more extreme are values greater than the test statistic. Here, the area under the curve to the right of 4.92 is .027 (Figure 3). This is the P value, which indicates that the data and the null hypothesis have very low compatibility. In this example, the area under the curve to the right of 4.92 is .027 (Figure 3). This is the P value, which indicates that the data and the null hypothesis have very low compatibility. Thus, the decision is to reject the null hypothesis. The conclusion is that lung morphology is associated with 90-day mortality (P = .027). To describe that association, one looks at the contingency table and finds a reduction in 90-day mortality with focal patterns compared to nonfocal patterns (21.4% vs 45.5%, respectively). The P value reported in the article is .026. Our hand calculation was .027, which is slightly off due to rounding. In summary, the scenario is an investigation into the association among 2 categorical variables, and, thus, a test to consider is the chi-square test, if assumptions are met.

In another example in the same study, the authors investigate whether any baseline characteristics are associated with lung morphology. For example, is neurology, specifically Parkinson disease (yes vs no), associated with lung morphology (focal vs nonfocal)? Again, the scenario is an investigation into the association between 2 categorical variables, so a chi-square test should be considered.

To start, build a contingency table arbitrarily placing lung morphology as the row variable and Parkinson disease as the column variable. Populate the contingency table based on the counts and percentages reported in the article (Figure 4). Next, check that the assumptions of the chi-square test are met. Are the observations independent? Again, because these are unique patients, we consider this assumption met. Since this is a 2 × 2 table, are all of the expected counts greater than 5? Calculations of the expected counts obtained the following: 1.1, 30.9, 2.9 and 84.1. Here, 2 of the 4 expected counts are less than 5. Therefore, methods that use large sample approximation, like the chi-squared test, may not be an appropriate choice.

Instead of using methodology that is an approximation, consider an exact test such as Fisher’s exact test. Again, refer to the contingency table where Fisher’s exact is going to calculate the exact probability (under the null hypothesis) of the observed data or results more extreme. This is the technical definition of a P value. It is, however, still quantifying how compatible the data are with the null hypothesis. The exact probability of a particular contingency table can be obtained using the hypergeometric distribution.

The symbols that resemble large parentheses are notations for a combinatorial. Because using combinatorials to calculate the probability is not user friendly, an equivalent version relies on factorials instead. Both techniques are presented above. Remember that the goal is to find the exact probability of the observed data or something more extreme.

The hypotheses are still testing whether these 2 categorical variables are associated with each other. In this particular example, we test if the proportion of patients with Parkinson disease is the same in the focal and nonfocal groups. Fisher’s exact test obtains its two-tailed P value by computing the probabilities associated with all possible tables that have the same row and column totals. Then, it identifies the alternative tables with a probability that is less than that of the observed table. Finally, it adds the probability of the observed table with the sum of the probabilities of each alternative table identified above, which results in the P value.

To explore each of those steps in detail, one must first enumerate how many tables can be built that all have the same row and column totals as the observed table. Figure 5 shows the 5 possible tables. Pick any one of the 5 2 × 2 tables; the margins are fixed. Each table has the same row totals, 32 focal and 87 nonfocal, and each table has the same column totals: 4 Parkinson and 115 non-Parkinson. Then, for each table, calculate the probability of that table. Figure 5 shows this calculation for the first 2 × 2 table, which happens to be the observed table. The probability of the table observed in the study is .2803. Such a calculation is performed on each of the other tables.

Next, one must identify the tables that have a probability smaller than the observed table. Here, we are looking for probabilities less than .2803. These are the tables deemed more extreme. Tables 3, 4, and 5 have probabilities less than .2803.

The final step is to sum the probability of the observed table and the more extreme tables (ie, those with probabilities < the observed table) (.2803 + .2337 + .0543 + .0045 = .5728). Thus, the resulting rounded P value is .57, which indicates a high level of compatibility between the data and the null hypothesis of no association. The decision is to fail to reject the null hypothesis and the conclusion is that the evidence does not support an association among lung morphology and Parkinson disease. In other words, there is insufficient evidence to claim that the proportion of Parkinson disease differs between the focal and nonfocal ARDS patients (0% vs 5%, P = .57). This matches the P value reported by Mrozek for this association.

The first objective of this article was to identify scenarios in which a chi-square or Fisher’s exact test should be considered. The general setting discussed was an investigation of the association between two categorical variables. Use of each test specifically depends on whether the assumptions have been met. Both of the examples used in our discussion happened to be binary, but that is not a restriction. Categorical variables can have more than 2 levels. All of the methods demonstrated for 2 × 2 tables can be generalized to r × c tables. 

The second objective of this article was to recognize when test assumptions have been violated. For simplicity, most researchers adhere to the following: if ≤ 20% of expected cell counts are less than 5, then use the chi-square test; if > 20% of expected cell counts are less than 5, then use Fisher’s exact test. Both methods assume that the observations are independent. Could one use the exact test when the chi-square assumptions are met? Yes, but it is more computationally expensive as it uses all possible fixed margin tables and their probabilities. If the chi-square assumptions are met, then the sample size is typically larger and these calculations become numerous. Also, it does not have to be that large of a sample for the chi-square to be a good approximation and do it very quickly.

The final objective of this article was to test claims made regarding the association of 2 independent categorical variables. We included examples from the medical literature showing step-by-step calculations of both the large sample approximation (chi-square) and exact (Fisher’s) methodologies providing insight into how these tests are conducted as well as when they are appropriate.

Click here to access the September 2017 Digital Edition.

Table of Contents

• Is Ketamine the New Wonder Drug for Treating Suicide?
• Development and Implementation of a Homeless Mobile Medical/Mental Veteran Intervention
• Current Approaches to Measuring Functional Status Among Older Adults in VA Primary Care Clinics
• Assessment of Free Flap Breast Reconstructions
• The Disease for Which There Is No Cure and Not Enough Conversation
• Florence A. Blanchfield: A Lifetime of Nursing Leadership

Click here to access the September 2017 Digital Edition.

Table of Contents

• Is Ketamine the New Wonder Drug for Treating Suicide?
• Development and Implementation of a Homeless Mobile Medical/Mental Veteran Intervention
• Current Approaches to Measuring Functional Status Among Older Adults in VA Primary Care Clinics
• Assessment of Free Flap Breast Reconstructions
• The Disease for Which There Is No Cure and Not Enough Conversation
• Florence A. Blanchfield: A Lifetime of Nursing Leadership

Click here to access the September 2017 Digital Edition.

Table of Contents

• Is Ketamine the New Wonder Drug for Treating Suicide?
• Development and Implementation of a Homeless Mobile Medical/Mental Veteran Intervention
• Current Approaches to Measuring Functional Status Among Older Adults in VA Primary Care Clinics
• Assessment of Free Flap Breast Reconstructions
• The Disease for Which There Is No Cure and Not Enough Conversation
• Florence A. Blanchfield: A Lifetime of Nursing Leadership

“THE PELVIC EXAM REVISITED”
ERIN HIGGINS, MD, AND
CHERYL B. IGLESIA, MD (AUGUST 2017)

Pelvic examination is essential to clinical care

I have contemplated the issue of the routine screening pelvic exam now for several years. But for the last year, I have found various problems in many “asymptomatic women.” For example: The 18-year-old who was “not sexually active” but who had Chlamydia. Or the 84-year-old who denied itching or other vulvovaginal symptoms who had either vulvar cancer or lichen sclerosis so severe her vagina was almost closed; a 30-minute review of her outside records revealed recurrent urinary tract infections requiring more than 5 courses of antibiotics in 6 months for what was actually contaminants from a urine specimen that passed through the vagina first. I think the move away from actually touching patients has completely gotten out of hand! It is appalling how many women I have seen who visited an emergency department for pelvic or abdominal pain and never had a hands-on examination. If we do not examine the part of the body that many completely ignore we may as well lose our specialty!

Christine Kneer-Aronoff, MD
Cincinnati, Ohio

Share your thoughts! Send your Letter to the Editor to [email protected]. Please include your name and the city and state in which you practice.

“THE PELVIC EXAM REVISITED”
ERIN HIGGINS, MD, AND
CHERYL B. IGLESIA, MD (AUGUST 2017)

Pelvic examination is essential to clinical care

I have contemplated the issue of the routine screening pelvic exam now for several years. But for the last year, I have found various problems in many “asymptomatic women.” For example: The 18-year-old who was “not sexually active” but who had Chlamydia. Or the 84-year-old who denied itching or other vulvovaginal symptoms who had either vulvar cancer or lichen sclerosis so severe her vagina was almost closed; a 30-minute review of her outside records revealed recurrent urinary tract infections requiring more than 5 courses of antibiotics in 6 months for what was actually contaminants from a urine specimen that passed through the vagina first. I think the move away from actually touching patients has completely gotten out of hand! It is appalling how many women I have seen who visited an emergency department for pelvic or abdominal pain and never had a hands-on examination. If we do not examine the part of the body that many completely ignore we may as well lose our specialty!

Christine Kneer-Aronoff, MD
Cincinnati, Ohio

Share your thoughts! Send your Letter to the Editor to [email protected]. Please include your name and the city and state in which you practice.

“THE PELVIC EXAM REVISITED”
ERIN HIGGINS, MD, AND
CHERYL B. IGLESIA, MD (AUGUST 2017)

Pelvic examination is essential to clinical care

I have contemplated the issue of the routine screening pelvic exam now for several years. But for the last year, I have found various problems in many “asymptomatic women.” For example: The 18-year-old who was “not sexually active” but who had Chlamydia. Or the 84-year-old who denied itching or other vulvovaginal symptoms who had either vulvar cancer or lichen sclerosis so severe her vagina was almost closed; a 30-minute review of her outside records revealed recurrent urinary tract infections requiring more than 5 courses of antibiotics in 6 months for what was actually contaminants from a urine specimen that passed through the vagina first. I think the move away from actually touching patients has completely gotten out of hand! It is appalling how many women I have seen who visited an emergency department for pelvic or abdominal pain and never had a hands-on examination. If we do not examine the part of the body that many completely ignore we may as well lose our specialty!

Christine Kneer-Aronoff, MD
Cincinnati, Ohio

Share your thoughts! Send your Letter to the Editor to [email protected]. Please include your name and the city and state in which you practice.

EXPERT COMMENTARY

The concurrent use of mechanical and pharmacologic cervical ripening is an area of active interest. Combination methods typically involve placing a Foley catheter and simultaneously administering either prostaglandins or oxytocin. Despite the long-standing belief that using 2 cervical ripening agents simultaneously has no benefit compared with using only 1 cervical ripening agent, several recent large randomized trials are challenging this paradigm by suggesting that using 2 cervical ripening agents together may in fact be superior.

Related Article:
Q: Following cesarean delivery, what is the optimal oxytocin infusion duration to prevent postpartum bleeding?

Details of the study

Schoen and colleagues conducted a randomized controlled trial that included 184 nulliparous and 139 multiparous women with an unfavorable cervix undergoing induction of labor after 24 weeks of gestation. All participants had a Foley catheter placed intracervically and then were randomly assigned to receive either concurrent oxytocin infusion within 60 minutes or no oxytocin until after Foley catheter expulsion or removal. Nulliparous and multiparous women were randomly assigned separately. Women with premature rupture of membranes and with 1 prior cesarean delivery were included in the trial, but women were excluded if they were in active labor, had suspected abruption, or had a nonreassuring fetal tracing.

The study was powered to detect a 20% increase in total delivery rate within 24 hours of Foley placement, which was the primary study outcome. Secondary induction outcomes of note included time to Foley expulsion, time to second stage, delivery within 12 hours, total time to delivery, duration of oxytocin use, and mode of delivery. Several maternal and neonatal outcomes also were examined, including tachysystole, chorioamnionitis, meconium, postpartum hemorrhage, birth weight, maternal intensive care unit (ICU) admission, and neonatal ICU admission.

Related Article:
Start offering antenatal corticosteroids to women delivering between 34 0/7 and 36 6/7 weeks of gestation to improve newborn outcomes
 

Women receiving concurrent Foley and oxytocin delivered sooner. Among nulliparous women, the overall rate of delivery within 24 hours of Foley catheter placement was 64% in the Foley with concurrent oxytocin group compared with 43% in those who received a Foley catheter alone followed by oxytocin (P = .003). The overall time to delivery was 5 hours less in nulliparous women who received combination cervical ripening compared with those who had a Foley catheter alone.

Similarly, multiparous women had an overall rate of delivery within 24 hours of 87% in the concurrent Foley and oxytocin group compared with 72% in women who received Foley catheter followed by oxytocin (P = .022).

Meanwhile, there were no statistically significant differences in mode of delivery between groups for either multiparous or nulliparous patients, and there were no differences in adverse maternal or neonatal outcomes between groups.

Related Article:
How and when umbilical cord gas analysis can justify your obstetric management

Study strengths and weaknesses

This well-designed, randomized control trial clearly demonstrated that the combination of Foley catheter and oxytocin for cervical ripening increases the rate of delivery within 24 hours compared with use of Foley catheter alone. This finding is consistent with those of 2 other large randomized trials in the past 2 years that similarly demonstrated reduced time to delivery when oxytocin infusion was used in combination with Foley catheter compared with Foley alone.^1,2

Despite these findings, important questions remain regarding concurrent use of cervical ripening agents. The study by Schoen and colleagues does not address the other option for dual cervical ripening, namely, concurrent use of Foley catheter and misoprostol. Several large randomized trials using Foley catheter with vaginal or oral misoprostol demonstrated reduced time to delivery compared with using either method alone.^1,3,4 Only 1 randomized study has compared these 2 dual cervical ripening regimens head-to-head; that study demonstrated that the misoprostol and Foley combination significantly reduced time to delivery compared with combining Foley catheter and oxytocin together.¹

Additionally, it is important to note that the study by Schoen and colleagues was not large enough to adequately evaluate potential safety risks with dual combination cervical ripening. More safety data are needed before combination cervical ripening methods can be recommended universally.

Share your thoughts! Send your Letter to the Editor to [email protected]. Please include your name and the city and state in which you practice.

EXPERT COMMENTARY

The concurrent use of mechanical and pharmacologic cervical ripening is an area of active interest. Combination methods typically involve placing a Foley catheter and simultaneously administering either prostaglandins or oxytocin. Despite the long-standing belief that using 2 cervical ripening agents simultaneously has no benefit compared with using only 1 cervical ripening agent, several recent large randomized trials are challenging this paradigm by suggesting that using 2 cervical ripening agents together may in fact be superior.

Related Article:
Q: Following cesarean delivery, what is the optimal oxytocin infusion duration to prevent postpartum bleeding?

Details of the study

Schoen and colleagues conducted a randomized controlled trial that included 184 nulliparous and 139 multiparous women with an unfavorable cervix undergoing induction of labor after 24 weeks of gestation. All participants had a Foley catheter placed intracervically and then were randomly assigned to receive either concurrent oxytocin infusion within 60 minutes or no oxytocin until after Foley catheter expulsion or removal. Nulliparous and multiparous women were randomly assigned separately. Women with premature rupture of membranes and with 1 prior cesarean delivery were included in the trial, but women were excluded if they were in active labor, had suspected abruption, or had a nonreassuring fetal tracing.

The study was powered to detect a 20% increase in total delivery rate within 24 hours of Foley placement, which was the primary study outcome. Secondary induction outcomes of note included time to Foley expulsion, time to second stage, delivery within 12 hours, total time to delivery, duration of oxytocin use, and mode of delivery. Several maternal and neonatal outcomes also were examined, including tachysystole, chorioamnionitis, meconium, postpartum hemorrhage, birth weight, maternal intensive care unit (ICU) admission, and neonatal ICU admission.

Related Article:
Start offering antenatal corticosteroids to women delivering between 34 0/7 and 36 6/7 weeks of gestation to improve newborn outcomes
 

Women receiving concurrent Foley and oxytocin delivered sooner. Among nulliparous women, the overall rate of delivery within 24 hours of Foley catheter placement was 64% in the Foley with concurrent oxytocin group compared with 43% in those who received a Foley catheter alone followed by oxytocin (P = .003). The overall time to delivery was 5 hours less in nulliparous women who received combination cervical ripening compared with those who had a Foley catheter alone.

Similarly, multiparous women had an overall rate of delivery within 24 hours of 87% in the concurrent Foley and oxytocin group compared with 72% in women who received Foley catheter followed by oxytocin (P = .022).

Meanwhile, there were no statistically significant differences in mode of delivery between groups for either multiparous or nulliparous patients, and there were no differences in adverse maternal or neonatal outcomes between groups.

Related Article:
How and when umbilical cord gas analysis can justify your obstetric management

Study strengths and weaknesses

This well-designed, randomized control trial clearly demonstrated that the combination of Foley catheter and oxytocin for cervical ripening increases the rate of delivery within 24 hours compared with use of Foley catheter alone. This finding is consistent with those of 2 other large randomized trials in the past 2 years that similarly demonstrated reduced time to delivery when oxytocin infusion was used in combination with Foley catheter compared with Foley alone.^1,2

Despite these findings, important questions remain regarding concurrent use of cervical ripening agents. The study by Schoen and colleagues does not address the other option for dual cervical ripening, namely, concurrent use of Foley catheter and misoprostol. Several large randomized trials using Foley catheter with vaginal or oral misoprostol demonstrated reduced time to delivery compared with using either method alone.^1,3,4 Only 1 randomized study has compared these 2 dual cervical ripening regimens head-to-head; that study demonstrated that the misoprostol and Foley combination significantly reduced time to delivery compared with combining Foley catheter and oxytocin together.¹

Additionally, it is important to note that the study by Schoen and colleagues was not large enough to adequately evaluate potential safety risks with dual combination cervical ripening. More safety data are needed before combination cervical ripening methods can be recommended universally.

Share your thoughts! Send your Letter to the Editor to [email protected]. Please include your name and the city and state in which you practice.

EXPERT COMMENTARY

The concurrent use of mechanical and pharmacologic cervical ripening is an area of active interest. Combination methods typically involve placing a Foley catheter and simultaneously administering either prostaglandins or oxytocin. Despite the long-standing belief that using 2 cervical ripening agents simultaneously has no benefit compared with using only 1 cervical ripening agent, several recent large randomized trials are challenging this paradigm by suggesting that using 2 cervical ripening agents together may in fact be superior.

Related Article:
Q: Following cesarean delivery, what is the optimal oxytocin infusion duration to prevent postpartum bleeding?

Details of the study

Schoen and colleagues conducted a randomized controlled trial that included 184 nulliparous and 139 multiparous women with an unfavorable cervix undergoing induction of labor after 24 weeks of gestation. All participants had a Foley catheter placed intracervically and then were randomly assigned to receive either concurrent oxytocin infusion within 60 minutes or no oxytocin until after Foley catheter expulsion or removal. Nulliparous and multiparous women were randomly assigned separately. Women with premature rupture of membranes and with 1 prior cesarean delivery were included in the trial, but women were excluded if they were in active labor, had suspected abruption, or had a nonreassuring fetal tracing.

The study was powered to detect a 20% increase in total delivery rate within 24 hours of Foley placement, which was the primary study outcome. Secondary induction outcomes of note included time to Foley expulsion, time to second stage, delivery within 12 hours, total time to delivery, duration of oxytocin use, and mode of delivery. Several maternal and neonatal outcomes also were examined, including tachysystole, chorioamnionitis, meconium, postpartum hemorrhage, birth weight, maternal intensive care unit (ICU) admission, and neonatal ICU admission.

Related Article:
Start offering antenatal corticosteroids to women delivering between 34 0/7 and 36 6/7 weeks of gestation to improve newborn outcomes
 

Women receiving concurrent Foley and oxytocin delivered sooner. Among nulliparous women, the overall rate of delivery within 24 hours of Foley catheter placement was 64% in the Foley with concurrent oxytocin group compared with 43% in those who received a Foley catheter alone followed by oxytocin (P = .003). The overall time to delivery was 5 hours less in nulliparous women who received combination cervical ripening compared with those who had a Foley catheter alone.

Similarly, multiparous women had an overall rate of delivery within 24 hours of 87% in the concurrent Foley and oxytocin group compared with 72% in women who received Foley catheter followed by oxytocin (P = .022).

Meanwhile, there were no statistically significant differences in mode of delivery between groups for either multiparous or nulliparous patients, and there were no differences in adverse maternal or neonatal outcomes between groups.

Related Article:
How and when umbilical cord gas analysis can justify your obstetric management

Study strengths and weaknesses

This well-designed, randomized control trial clearly demonstrated that the combination of Foley catheter and oxytocin for cervical ripening increases the rate of delivery within 24 hours compared with use of Foley catheter alone. This finding is consistent with those of 2 other large randomized trials in the past 2 years that similarly demonstrated reduced time to delivery when oxytocin infusion was used in combination with Foley catheter compared with Foley alone.^1,2

Despite these findings, important questions remain regarding concurrent use of cervical ripening agents. The study by Schoen and colleagues does not address the other option for dual cervical ripening, namely, concurrent use of Foley catheter and misoprostol. Several large randomized trials using Foley catheter with vaginal or oral misoprostol demonstrated reduced time to delivery compared with using either method alone.^1,3,4 Only 1 randomized study has compared these 2 dual cervical ripening regimens head-to-head; that study demonstrated that the misoprostol and Foley combination significantly reduced time to delivery compared with combining Foley catheter and oxytocin together.¹

Additionally, it is important to note that the study by Schoen and colleagues was not large enough to adequately evaluate potential safety risks with dual combination cervical ripening. More safety data are needed before combination cervical ripening methods can be recommended universally.

Share your thoughts! Send your Letter to the Editor to [email protected]. Please include your name and the city and state in which you practice.

In an age when psychiatry strives to identify the biologic causes of disease, studying endocrine-related mood disorders is particularly intriguing. DSM-5 defines premenstrual dysphoric disorder (PMDD) as a depressive disorder, with a 12-month prevalence ranging from 1.8% to 5.8% among women who menstruate.^1-3 Factors that differentiate PMDD from other affective disorders include etiology, duration, and temporal relationship with the menstrual cycle.

PMDD is a disorder of consistent yet intermittent change in mental health and functionality. Therefore, it may be underdiagnosed and consequently undertreated if a psychiatric evaluation does not coincide with symptom occurrence or if patients do not understand that intermittent symptoms are treatable.

This article summarizes what is known about the etiology of PMDD. Although there are several treatments for PMDD, many women experience adverse effects or incomplete effectiveness. Further understanding of this disorder may lead to more efficacious treatments. Additionally, understanding the pathophysiology of PMDD might shed a light on the etiology of other disorders that are temporally related to reproductive life changes, such as pregnancy-, postpartum-, or menopause-related affective dysregulation.

Making the diagnosis

The diagnosis of PMDD is made when a patient has at least 5 of 11 specific symptoms that occur during the week before onset of menses, improve within a few days after the onset of menses (shown as the “PMDD Hazard Zone” in Figure 1), and are minimal or absent post-menses.³ Symptoms should be tracked prospectively for at least 2 menstrual cycles in order to confirm the diagnosis (one must be an affective symptom and another must be a behavioral/cognitive symptom).³

The affective symptoms are:

• lability of affect (eg, sudden sadness, tearfulness, or sensitivity to rejection)
• irritability, anger, or increased inter­personal conflicts
• depressed mood, hopelessness, or self- deprecating thoughts
• anxiety or tension, feeling “keyed up” or “on edge.”

The behavioral/cognitive symptoms are:

• decreased interest in usual activities (eg, work, hobbies, friends, school)
• difficulty concentrating
• lethargy, low energy, easy fatigability
• change in appetite, overeating, food cravings
• hypersomnia or insomnia
• feeling overwhelmed or out of control
• physical symptoms (breast tenderness or swelling, headache, joint or muscle pain, bloating, weight gain).

Ruling out premenstrual exacerbation (PME). Perhaps the most common cause for misdiagnosis of PMDD is failing to rule out PME of another underlying or comorbid condition (Figure 2). In many women who have a primary mood or anxiety disorder, the late luteal phase is a vulnerable time. A patient might be coping with untreated anxiety, for example, but the symptoms become unbearable the week before menstruation begins, which is likely when she seeks help. At this stage, a diagnosis of PMDD should be provisional at best. Often, PME is treated by treating the underlying condition. Therefore, a full diagnostic psychiatric interview is important to first rule out other underlying psychiatric disorders. PMDD is diagnosed if the premenstrual symptoms persist for 2 consecutive months after treating the suspected mood or anxiety disorder. Patients can use one of many PMDD daily symptom charts available online. Alternatively, they can use a cycle-tracking mobile phone application to correlate their symptoms with their cycle and share this information with their providers.

Consider these 5 interwoven pieces

The many variables that contribute to the pathophysiology of PMDD overlap and should be considered connecting pieces in the puzzle that is the etiology of this disorder (Figure 3). In reviewing the literature, we have identified 5 topics likely to be major contributors to this disorder:

1. genetic susceptibility
2. progesterone and allopregnanolone (ALLO)
3. estrogen, serotonin, and brain-derived neurotrophic factor (BDNF)
4. putative brain structural and functional differences
5. further involvement of the hypothalamic–pituitary–adrenal (HPA) axis and hypothalamic–pituitary–gonadal (HPG) axis: trauma, resiliency, and inflammation.

Genetic susceptibility. PMDD is thought to have a heritability range between 30% to 80%.³ This is demonstrated by family and twin studies^4-7 and specific genetic studies.⁸ The involvement of genetics means an underlying neurobiologic pathophysiology is in place.

Estrogen receptor alpha (ESR1) gene. Huo et al⁸ found an associated variation in ESR1 in women with PMDD compared with controls. They speculated that because ESR1 is important for arousal, if dysfunctional, this gene could be implicated in somatic as well as affective and cognitive deficits in PMDD patients. In another study, investigators reported a relationship between PMDD and heritable personality traits, as well as a link between these traits and ESR1 polymorphic variants.¹ They suggested that personality traits (independent of affective state) might be used to distinguish patients with PMDD from controls.¹

Studies on serotonin gene polymorphism and serotonin transporter genotype. Although a study of serotonin gene polymorphism did not find an association between serotonin1A gene polymorphism and PMDD, it did show that the presence of at least 1 C allele was associated with a 2.5-fold increased risk of PMDD.⁹ Another study did not find an association between the serotonin transporter genotype 5-HTTLPR and PMDD.¹⁰ However, it showed lower frontocingulate cortex activation during the luteal phase of PMDD patients compared with controls, suggesting that PMDD is linked to impaired frontocingulate cortex activation induced by emotions during the luteal phase.¹⁰ 

Seasonal affective disorder (SAD) and PMDD have shared clinical features. A polymorphism in the serotonin transporter promoter gene 5-HTTLPR has been associated with SAD. One study found that patients with comorbid SAD and PMDD are genetically more vulnerable to comorbid affective disorders compared with patients who have SAD only.¹¹

Progesterone and ALLO. Chronic exposure to progesterone and ALLO (a main progesterone metabolite) and rapid withdrawal from ovarian hormones may play a role in the etiology of PMDD. Much like alcohol or benzodiazepines, ALLO is a potent positive allosteric modulator of GABA_A receptors and has sedative, anesthetic, and anxiolytic properties. In times of acute stress, increased ALLO is known to provide relief.^12,13 However, in women with PMDD, this typical ALLO increase might not occur.¹⁴

Patients with PMDD have been reported to have decreased levels of ALLO in the luteal phase.^15-17 In one study, women with highly symptomatic PMDD had lower levels of ALLO compared with women with less symptomatic PMDD.¹⁴ A gonadotropin-releasing hormone challenge study showed the increase in ALLO response was less in PMDD patients compared with controls.¹⁷ Luteal-phase ALLO concentrations are reported to be lower in women with premenstrual syndrome (PMS), a milder form of PMDD.^14,17

The efficacy of selective serotonin reuptake inhibitors (SSRIs) for treating PMDD could be the result of the interaction of these medications with neuroactive steroids,¹⁸ possibly because SSRIs enhance the sensitivity of GABA_A receptors or promote the formation of more ALLO (Figure 4).^19-21

Estrogen, serotonin, and BDNF. Estrogen affects multiple neurotransmitter systems that regulate mood, cognition, sleep, and eating.²² Studying estrogen in context of PMDD is important because women with PMDD can have low mood, specific food cravings, and impaired cognitive function.

Estrogen–serotonin interactions are thought to be involved in hormone-related mood disorders such as perimenopausal depression and PMDD.^23,24 However, the nature of their relationship is not yet fully understood. Ovariectomized animals have shown estrogen-induced changes related to serotonin metabolism, binding, and transmission in the regions of the brain involved in regulation of affect and cognition. Research in menopausal women also has provided some support for this interaction.²⁴

Positron emission tomography studies in humans have found increased cortical serotonin binding modulated by levels of estrogen, similar to those previously seen in rat studies.^24-27 One study showed an increased binding potential of serotonin in the cerebral cortex with estrogen treatment. This study further showed an even greater binding potential with estrogen plus progesterone, signaling a synergistic effect of the 2 hormones.²⁸

SSRIs are an effective treatment for the irritability, anxiety, and mood swings of PMDD.^29-30 Although the exact mechanism of action is unknown, the serotonergic properties are certainly of primary attention. For some PMDD patients, SSRIs work within hours to days, as opposed to days or weeks for patients with depression or anxiety, which suggests a separate or co-occurring mechanism of action is in place. In a double-blind, placebo-controlled crossover study, researchers administered the serotonin receptor antagonist metergoline to women with PMDD whose symptoms had remitted during treatment with fluoxetine and a group of healthy controls who were not receiving any medication.³¹ The women with PMDD experienced a return of symptoms 24 hours after treatment with metergoline but not with placebo; the controls experienced no mood changes.³¹

BDNF is a neurotransmitter linked to estrogen and likely related to PMDD. BDNF is critical for neurogenesis and is expressed in brain regions involved in learning and memory and also affects regulation.³² BDNF levels are increased by serotonergic antidepressants, affected by estradiol, and have cyclicity throughout the menstrual cycle.^33-35

Putative brain structural and functional differences. Imaging studies have suggested differences in brain structure in women with PMDD, with a focus on the amygdala and the prefrontal cortex. Women with PMDD have greater gray matter volume in the posterior cerebellum,³⁶ greater gray matter density of hippocampal cortex, and lower gray matter density in the parahippocampal cortex.³⁷

Some studies have shown a functional variability of the amygdala’s response to stress in women with PMDD vs healthy controls.^38,39 A proton magnetic resonance spectroscopy (1H-MRS) study of the displays the possibility of an altered GABAergic function in patients with PMDD.⁴⁰

Patients will PMDD have enhanced dorsolateral prefrontal cortex reactivity when anticipating negative stimuli (but not to the actual exposure) during the luteal phase. A positive correlation between this reactivity and progesterone levels also was observed.⁴¹ Some researchers have suggested that prefrontal cortex dysfunction may be a risk factor for PMDD.⁴²

HPA axis and HPG axis: Trauma, resiliency, inflammation. Altered cortisol levels (higher during the luteal phase⁴³ and lower during times of stress^14,44) suggest a possibly altered HPA axis in some women with PMDD. However, studies on this topic have been few and inconsistent.

Dysregulation of the HPG axis could cause vasomotor symptoms, sleep dysregulation, and mood symptoms during menopause; women with PMDD can also experience these symptoms. The influence of estrogen and progesterone on mood is also highly dependent on this axis.

Ultimately, the interplay between the HPA axis and the HPG axis is important. One study found that women with PMDD who had high serum ALLO levels (HPG-related) had blunted nocturnal cortisol levels (HPA-related) compared with healthy controls who had low ALLO levels.⁴⁵

Significant stress and trauma exposure have been associated with PMDD. A study of 3,968 women found a history of trauma and PTSD were independently associated with PMDD.⁴⁶ Another study of approximately 3,000 women found a strong correlation between abuse and PMS.⁴⁷ However, a third study found no correlations between PMDD and trauma.⁴⁸

Patients with a predisposition to PMDD may be more vulnerable to develop a posttraumatic stress-related disorder, perhaps due to decreased biologic resiliency. For example, the startle response (hyper­vigilance) has been shown to be different in women with PMDD. One study suggested that suboptimal production of premenstrual ALLO may lead to increased arousal and increased stress reactivity to psycho­social or environmental triggers.⁴⁹

The possible role of inflammation in PMDD deserves further investigation. The luteal phase entails an increase in the production of proinflammatory markers.^50,51 A 10-fold increase in progesterone is correlated with a 20% to 23% increase in C-reactive protein levels.^52,53 Women with inflammatory diseases (eg, gingivitis or irritable bowel syndrome) show worsening of symptoms prior to menstruation.^54-56 One study found increased levels of proinflammatory markers in women with PMDD compared with controls.⁵⁷

Because PMDD is heritable, it must have an underlying neurobiologic pathophysiology. Brain imaging studies show differences in structure and function in women with PMDD across the menstrual cycle. Conversion of progesterone to ALLO and the GABAergic influence of this metabolite is a topic of interest in current research. Similarly, the role of estrogen and its connection to serotonin and other neurotransmitters such as BDNF have been implicated.

The link between a history of stress, trauma, and PMDD raises the question of biologic resiliency and illness in these patients, as it connects to the HPA and HPG axis and production of inflammatory stress hormones and steroid hormones and their metabolites. PMDD can be conceptualized as variable sensitivity to hormonal response to stress,⁵⁸ thus contextualizing biochemical and psychological resiliency.

Further research is needed to clarify the possibility of a shared pathophysiology between endocrine-related mood disorders such as postpartum depression (PPD) and PMDD because current research is controversial.^59,60 In PPD, women who are exposed to high levels of progesterone and estrogen during pregnancy (just like in the mid-luteal phase) have a sudden drop in these hormones postpartum.

The ‘withdrawal theory.’ The affective symptoms of PMDD resolve almost instantaneously after the start of menstruation. Perhaps this type of immediate relief is akin to substance use disorders and symptoms of withdrawal. It could be that reinstatement of a certain amount of gonadal steroids in the follicular phase of the cycle diminishes a withdrawal-like response to these steroids.

Currently, the main leading theory is that PMDD is a result of “an abnormal response to normal hormonal changes.”⁶¹ A new study also has shown that the change in estradiol/progesterone levels (vs the steady state) was associated with PMDD symptoms.⁶² Thinking of PMDD as a disorder of withdrawal offers an alternative (yet complementary) perspective to the current theory: PMDD may be caused by the absence or diminishing of the above-named hormones and their metabolites in the late luteal phase (in the context of developed “tolerance” during the early- to mid-luteal phase).

Considering the interplay between neuro­transmitters and neurosteroids, both a “serotonin withdrawal theory” (caused by a drop in steroid hormones) and a “GABAergic withdrawal theory” (due to the decline in progesterone) could be proposed. This theory would be supported by the fact that SSRIs seem to mitigate symptoms of PMDD as well as the genetic association between PMDD and ESR1. It is more than likely that the “withdrawal” is caused by the interactions between estrogen-serotonin, progesterone-ALLO, and GABA receptors, and the complementary fashion in which progesterone and estrogen influence each other.

In an age when psychiatry strives to identify the biologic causes of disease, studying endocrine-related mood disorders is particularly intriguing. DSM-5 defines premenstrual dysphoric disorder (PMDD) as a depressive disorder, with a 12-month prevalence ranging from 1.8% to 5.8% among women who menstruate.^1-3 Factors that differentiate PMDD from other affective disorders include etiology, duration, and temporal relationship with the menstrual cycle.

PMDD is a disorder of consistent yet intermittent change in mental health and functionality. Therefore, it may be underdiagnosed and consequently undertreated if a psychiatric evaluation does not coincide with symptom occurrence or if patients do not understand that intermittent symptoms are treatable.

This article summarizes what is known about the etiology of PMDD. Although there are several treatments for PMDD, many women experience adverse effects or incomplete effectiveness. Further understanding of this disorder may lead to more efficacious treatments. Additionally, understanding the pathophysiology of PMDD might shed a light on the etiology of other disorders that are temporally related to reproductive life changes, such as pregnancy-, postpartum-, or menopause-related affective dysregulation.

Making the diagnosis

The diagnosis of PMDD is made when a patient has at least 5 of 11 specific symptoms that occur during the week before onset of menses, improve within a few days after the onset of menses (shown as the “PMDD Hazard Zone” in Figure 1), and are minimal or absent post-menses.³ Symptoms should be tracked prospectively for at least 2 menstrual cycles in order to confirm the diagnosis (one must be an affective symptom and another must be a behavioral/cognitive symptom).³

The affective symptoms are:

• lability of affect (eg, sudden sadness, tearfulness, or sensitivity to rejection)
• irritability, anger, or increased inter­personal conflicts
• depressed mood, hopelessness, or self- deprecating thoughts
• anxiety or tension, feeling “keyed up” or “on edge.”

The behavioral/cognitive symptoms are:

• decreased interest in usual activities (eg, work, hobbies, friends, school)
• difficulty concentrating
• lethargy, low energy, easy fatigability
• change in appetite, overeating, food cravings
• hypersomnia or insomnia
• feeling overwhelmed or out of control
• physical symptoms (breast tenderness or swelling, headache, joint or muscle pain, bloating, weight gain).

Ruling out premenstrual exacerbation (PME). Perhaps the most common cause for misdiagnosis of PMDD is failing to rule out PME of another underlying or comorbid condition (Figure 2). In many women who have a primary mood or anxiety disorder, the late luteal phase is a vulnerable time. A patient might be coping with untreated anxiety, for example, but the symptoms become unbearable the week before menstruation begins, which is likely when she seeks help. At this stage, a diagnosis of PMDD should be provisional at best. Often, PME is treated by treating the underlying condition. Therefore, a full diagnostic psychiatric interview is important to first rule out other underlying psychiatric disorders. PMDD is diagnosed if the premenstrual symptoms persist for 2 consecutive months after treating the suspected mood or anxiety disorder. Patients can use one of many PMDD daily symptom charts available online. Alternatively, they can use a cycle-tracking mobile phone application to correlate their symptoms with their cycle and share this information with their providers.

Consider these 5 interwoven pieces

The many variables that contribute to the pathophysiology of PMDD overlap and should be considered connecting pieces in the puzzle that is the etiology of this disorder (Figure 3). In reviewing the literature, we have identified 5 topics likely to be major contributors to this disorder:

1. genetic susceptibility
2. progesterone and allopregnanolone (ALLO)
3. estrogen, serotonin, and brain-derived neurotrophic factor (BDNF)
4. putative brain structural and functional differences
5. further involvement of the hypothalamic–pituitary–adrenal (HPA) axis and hypothalamic–pituitary–gonadal (HPG) axis: trauma, resiliency, and inflammation.

Genetic susceptibility. PMDD is thought to have a heritability range between 30% to 80%.³ This is demonstrated by family and twin studies^4-7 and specific genetic studies.⁸ The involvement of genetics means an underlying neurobiologic pathophysiology is in place.

Estrogen receptor alpha (ESR1) gene. Huo et al⁸ found an associated variation in ESR1 in women with PMDD compared with controls. They speculated that because ESR1 is important for arousal, if dysfunctional, this gene could be implicated in somatic as well as affective and cognitive deficits in PMDD patients. In another study, investigators reported a relationship between PMDD and heritable personality traits, as well as a link between these traits and ESR1 polymorphic variants.¹ They suggested that personality traits (independent of affective state) might be used to distinguish patients with PMDD from controls.¹

Studies on serotonin gene polymorphism and serotonin transporter genotype. Although a study of serotonin gene polymorphism did not find an association between serotonin1A gene polymorphism and PMDD, it did show that the presence of at least 1 C allele was associated with a 2.5-fold increased risk of PMDD.⁹ Another study did not find an association between the serotonin transporter genotype 5-HTTLPR and PMDD.¹⁰ However, it showed lower frontocingulate cortex activation during the luteal phase of PMDD patients compared with controls, suggesting that PMDD is linked to impaired frontocingulate cortex activation induced by emotions during the luteal phase.¹⁰ 

Seasonal affective disorder (SAD) and PMDD have shared clinical features. A polymorphism in the serotonin transporter promoter gene 5-HTTLPR has been associated with SAD. One study found that patients with comorbid SAD and PMDD are genetically more vulnerable to comorbid affective disorders compared with patients who have SAD only.¹¹

Progesterone and ALLO. Chronic exposure to progesterone and ALLO (a main progesterone metabolite) and rapid withdrawal from ovarian hormones may play a role in the etiology of PMDD. Much like alcohol or benzodiazepines, ALLO is a potent positive allosteric modulator of GABA_A receptors and has sedative, anesthetic, and anxiolytic properties. In times of acute stress, increased ALLO is known to provide relief.^12,13 However, in women with PMDD, this typical ALLO increase might not occur.¹⁴

Patients with PMDD have been reported to have decreased levels of ALLO in the luteal phase.^15-17 In one study, women with highly symptomatic PMDD had lower levels of ALLO compared with women with less symptomatic PMDD.¹⁴ A gonadotropin-releasing hormone challenge study showed the increase in ALLO response was less in PMDD patients compared with controls.¹⁷ Luteal-phase ALLO concentrations are reported to be lower in women with premenstrual syndrome (PMS), a milder form of PMDD.^14,17

The efficacy of selective serotonin reuptake inhibitors (SSRIs) for treating PMDD could be the result of the interaction of these medications with neuroactive steroids,¹⁸ possibly because SSRIs enhance the sensitivity of GABA_A receptors or promote the formation of more ALLO (Figure 4).^19-21

Estrogen, serotonin, and BDNF. Estrogen affects multiple neurotransmitter systems that regulate mood, cognition, sleep, and eating.²² Studying estrogen in context of PMDD is important because women with PMDD can have low mood, specific food cravings, and impaired cognitive function.

Estrogen–serotonin interactions are thought to be involved in hormone-related mood disorders such as perimenopausal depression and PMDD.^23,24 However, the nature of their relationship is not yet fully understood. Ovariectomized animals have shown estrogen-induced changes related to serotonin metabolism, binding, and transmission in the regions of the brain involved in regulation of affect and cognition. Research in menopausal women also has provided some support for this interaction.²⁴

Positron emission tomography studies in humans have found increased cortical serotonin binding modulated by levels of estrogen, similar to those previously seen in rat studies.^24-27 One study showed an increased binding potential of serotonin in the cerebral cortex with estrogen treatment. This study further showed an even greater binding potential with estrogen plus progesterone, signaling a synergistic effect of the 2 hormones.²⁸

SSRIs are an effective treatment for the irritability, anxiety, and mood swings of PMDD.^29-30 Although the exact mechanism of action is unknown, the serotonergic properties are certainly of primary attention. For some PMDD patients, SSRIs work within hours to days, as opposed to days or weeks for patients with depression or anxiety, which suggests a separate or co-occurring mechanism of action is in place. In a double-blind, placebo-controlled crossover study, researchers administered the serotonin receptor antagonist metergoline to women with PMDD whose symptoms had remitted during treatment with fluoxetine and a group of healthy controls who were not receiving any medication.³¹ The women with PMDD experienced a return of symptoms 24 hours after treatment with metergoline but not with placebo; the controls experienced no mood changes.³¹

BDNF is a neurotransmitter linked to estrogen and likely related to PMDD. BDNF is critical for neurogenesis and is expressed in brain regions involved in learning and memory and also affects regulation.³² BDNF levels are increased by serotonergic antidepressants, affected by estradiol, and have cyclicity throughout the menstrual cycle.^33-35

Putative brain structural and functional differences. Imaging studies have suggested differences in brain structure in women with PMDD, with a focus on the amygdala and the prefrontal cortex. Women with PMDD have greater gray matter volume in the posterior cerebellum,³⁶ greater gray matter density of hippocampal cortex, and lower gray matter density in the parahippocampal cortex.³⁷

Some studies have shown a functional variability of the amygdala’s response to stress in women with PMDD vs healthy controls.^38,39 A proton magnetic resonance spectroscopy (1H-MRS) study of the displays the possibility of an altered GABAergic function in patients with PMDD.⁴⁰

Patients will PMDD have enhanced dorsolateral prefrontal cortex reactivity when anticipating negative stimuli (but not to the actual exposure) during the luteal phase. A positive correlation between this reactivity and progesterone levels also was observed.⁴¹ Some researchers have suggested that prefrontal cortex dysfunction may be a risk factor for PMDD.⁴²

HPA axis and HPG axis: Trauma, resiliency, inflammation. Altered cortisol levels (higher during the luteal phase⁴³ and lower during times of stress^14,44) suggest a possibly altered HPA axis in some women with PMDD. However, studies on this topic have been few and inconsistent.

Dysregulation of the HPG axis could cause vasomotor symptoms, sleep dysregulation, and mood symptoms during menopause; women with PMDD can also experience these symptoms. The influence of estrogen and progesterone on mood is also highly dependent on this axis.

Ultimately, the interplay between the HPA axis and the HPG axis is important. One study found that women with PMDD who had high serum ALLO levels (HPG-related) had blunted nocturnal cortisol levels (HPA-related) compared with healthy controls who had low ALLO levels.⁴⁵

Significant stress and trauma exposure have been associated with PMDD. A study of 3,968 women found a history of trauma and PTSD were independently associated with PMDD.⁴⁶ Another study of approximately 3,000 women found a strong correlation between abuse and PMS.⁴⁷ However, a third study found no correlations between PMDD and trauma.⁴⁸

Patients with a predisposition to PMDD may be more vulnerable to develop a posttraumatic stress-related disorder, perhaps due to decreased biologic resiliency. For example, the startle response (hyper­vigilance) has been shown to be different in women with PMDD. One study suggested that suboptimal production of premenstrual ALLO may lead to increased arousal and increased stress reactivity to psycho­social or environmental triggers.⁴⁹

The possible role of inflammation in PMDD deserves further investigation. The luteal phase entails an increase in the production of proinflammatory markers.^50,51 A 10-fold increase in progesterone is correlated with a 20% to 23% increase in C-reactive protein levels.^52,53 Women with inflammatory diseases (eg, gingivitis or irritable bowel syndrome) show worsening of symptoms prior to menstruation.^54-56 One study found increased levels of proinflammatory markers in women with PMDD compared with controls.⁵⁷

Because PMDD is heritable, it must have an underlying neurobiologic pathophysiology. Brain imaging studies show differences in structure and function in women with PMDD across the menstrual cycle. Conversion of progesterone to ALLO and the GABAergic influence of this metabolite is a topic of interest in current research. Similarly, the role of estrogen and its connection to serotonin and other neurotransmitters such as BDNF have been implicated.

The link between a history of stress, trauma, and PMDD raises the question of biologic resiliency and illness in these patients, as it connects to the HPA and HPG axis and production of inflammatory stress hormones and steroid hormones and their metabolites. PMDD can be conceptualized as variable sensitivity to hormonal response to stress,⁵⁸ thus contextualizing biochemical and psychological resiliency.

Further research is needed to clarify the possibility of a shared pathophysiology between endocrine-related mood disorders such as postpartum depression (PPD) and PMDD because current research is controversial.^59,60 In PPD, women who are exposed to high levels of progesterone and estrogen during pregnancy (just like in the mid-luteal phase) have a sudden drop in these hormones postpartum.

The ‘withdrawal theory.’ The affective symptoms of PMDD resolve almost instantaneously after the start of menstruation. Perhaps this type of immediate relief is akin to substance use disorders and symptoms of withdrawal. It could be that reinstatement of a certain amount of gonadal steroids in the follicular phase of the cycle diminishes a withdrawal-like response to these steroids.

Currently, the main leading theory is that PMDD is a result of “an abnormal response to normal hormonal changes.”⁶¹ A new study also has shown that the change in estradiol/progesterone levels (vs the steady state) was associated with PMDD symptoms.⁶² Thinking of PMDD as a disorder of withdrawal offers an alternative (yet complementary) perspective to the current theory: PMDD may be caused by the absence or diminishing of the above-named hormones and their metabolites in the late luteal phase (in the context of developed “tolerance” during the early- to mid-luteal phase).

Considering the interplay between neuro­transmitters and neurosteroids, both a “serotonin withdrawal theory” (caused by a drop in steroid hormones) and a “GABAergic withdrawal theory” (due to the decline in progesterone) could be proposed. This theory would be supported by the fact that SSRIs seem to mitigate symptoms of PMDD as well as the genetic association between PMDD and ESR1. It is more than likely that the “withdrawal” is caused by the interactions between estrogen-serotonin, progesterone-ALLO, and GABA receptors, and the complementary fashion in which progesterone and estrogen influence each other.

In an age when psychiatry strives to identify the biologic causes of disease, studying endocrine-related mood disorders is particularly intriguing. DSM-5 defines premenstrual dysphoric disorder (PMDD) as a depressive disorder, with a 12-month prevalence ranging from 1.8% to 5.8% among women who menstruate.^1-3 Factors that differentiate PMDD from other affective disorders include etiology, duration, and temporal relationship with the menstrual cycle.

PMDD is a disorder of consistent yet intermittent change in mental health and functionality. Therefore, it may be underdiagnosed and consequently undertreated if a psychiatric evaluation does not coincide with symptom occurrence or if patients do not understand that intermittent symptoms are treatable.

This article summarizes what is known about the etiology of PMDD. Although there are several treatments for PMDD, many women experience adverse effects or incomplete effectiveness. Further understanding of this disorder may lead to more efficacious treatments. Additionally, understanding the pathophysiology of PMDD might shed a light on the etiology of other disorders that are temporally related to reproductive life changes, such as pregnancy-, postpartum-, or menopause-related affective dysregulation.

Making the diagnosis

The diagnosis of PMDD is made when a patient has at least 5 of 11 specific symptoms that occur during the week before onset of menses, improve within a few days after the onset of menses (shown as the “PMDD Hazard Zone” in Figure 1), and are minimal or absent post-menses.³ Symptoms should be tracked prospectively for at least 2 menstrual cycles in order to confirm the diagnosis (one must be an affective symptom and another must be a behavioral/cognitive symptom).³

The affective symptoms are:

• lability of affect (eg, sudden sadness, tearfulness, or sensitivity to rejection)
• irritability, anger, or increased inter­personal conflicts
• depressed mood, hopelessness, or self- deprecating thoughts
• anxiety or tension, feeling “keyed up” or “on edge.”

The behavioral/cognitive symptoms are:

• decreased interest in usual activities (eg, work, hobbies, friends, school)
• difficulty concentrating
• lethargy, low energy, easy fatigability
• change in appetite, overeating, food cravings
• hypersomnia or insomnia
• feeling overwhelmed or out of control
• physical symptoms (breast tenderness or swelling, headache, joint or muscle pain, bloating, weight gain).

Ruling out premenstrual exacerbation (PME). Perhaps the most common cause for misdiagnosis of PMDD is failing to rule out PME of another underlying or comorbid condition (Figure 2). In many women who have a primary mood or anxiety disorder, the late luteal phase is a vulnerable time. A patient might be coping with untreated anxiety, for example, but the symptoms become unbearable the week before menstruation begins, which is likely when she seeks help. At this stage, a diagnosis of PMDD should be provisional at best. Often, PME is treated by treating the underlying condition. Therefore, a full diagnostic psychiatric interview is important to first rule out other underlying psychiatric disorders. PMDD is diagnosed if the premenstrual symptoms persist for 2 consecutive months after treating the suspected mood or anxiety disorder. Patients can use one of many PMDD daily symptom charts available online. Alternatively, they can use a cycle-tracking mobile phone application to correlate their symptoms with their cycle and share this information with their providers.

Consider these 5 interwoven pieces

The many variables that contribute to the pathophysiology of PMDD overlap and should be considered connecting pieces in the puzzle that is the etiology of this disorder (Figure 3). In reviewing the literature, we have identified 5 topics likely to be major contributors to this disorder:

1. genetic susceptibility
2. progesterone and allopregnanolone (ALLO)
3. estrogen, serotonin, and brain-derived neurotrophic factor (BDNF)
4. putative brain structural and functional differences
5. further involvement of the hypothalamic–pituitary–adrenal (HPA) axis and hypothalamic–pituitary–gonadal (HPG) axis: trauma, resiliency, and inflammation.

Genetic susceptibility. PMDD is thought to have a heritability range between 30% to 80%.³ This is demonstrated by family and twin studies^4-7 and specific genetic studies.⁸ The involvement of genetics means an underlying neurobiologic pathophysiology is in place.

Estrogen receptor alpha (ESR1) gene. Huo et al⁸ found an associated variation in ESR1 in women with PMDD compared with controls. They speculated that because ESR1 is important for arousal, if dysfunctional, this gene could be implicated in somatic as well as affective and cognitive deficits in PMDD patients. In another study, investigators reported a relationship between PMDD and heritable personality traits, as well as a link between these traits and ESR1 polymorphic variants.¹ They suggested that personality traits (independent of affective state) might be used to distinguish patients with PMDD from controls.¹

Studies on serotonin gene polymorphism and serotonin transporter genotype. Although a study of serotonin gene polymorphism did not find an association between serotonin1A gene polymorphism and PMDD, it did show that the presence of at least 1 C allele was associated with a 2.5-fold increased risk of PMDD.⁹ Another study did not find an association between the serotonin transporter genotype 5-HTTLPR and PMDD.¹⁰ However, it showed lower frontocingulate cortex activation during the luteal phase of PMDD patients compared with controls, suggesting that PMDD is linked to impaired frontocingulate cortex activation induced by emotions during the luteal phase.¹⁰ 

Seasonal affective disorder (SAD) and PMDD have shared clinical features. A polymorphism in the serotonin transporter promoter gene 5-HTTLPR has been associated with SAD. One study found that patients with comorbid SAD and PMDD are genetically more vulnerable to comorbid affective disorders compared with patients who have SAD only.¹¹

Progesterone and ALLO. Chronic exposure to progesterone and ALLO (a main progesterone metabolite) and rapid withdrawal from ovarian hormones may play a role in the etiology of PMDD. Much like alcohol or benzodiazepines, ALLO is a potent positive allosteric modulator of GABA_A receptors and has sedative, anesthetic, and anxiolytic properties. In times of acute stress, increased ALLO is known to provide relief.^12,13 However, in women with PMDD, this typical ALLO increase might not occur.¹⁴

Patients with PMDD have been reported to have decreased levels of ALLO in the luteal phase.^15-17 In one study, women with highly symptomatic PMDD had lower levels of ALLO compared with women with less symptomatic PMDD.¹⁴ A gonadotropin-releasing hormone challenge study showed the increase in ALLO response was less in PMDD patients compared with controls.¹⁷ Luteal-phase ALLO concentrations are reported to be lower in women with premenstrual syndrome (PMS), a milder form of PMDD.^14,17

The efficacy of selective serotonin reuptake inhibitors (SSRIs) for treating PMDD could be the result of the interaction of these medications with neuroactive steroids,¹⁸ possibly because SSRIs enhance the sensitivity of GABA_A receptors or promote the formation of more ALLO (Figure 4).^19-21

Estrogen, serotonin, and BDNF. Estrogen affects multiple neurotransmitter systems that regulate mood, cognition, sleep, and eating.²² Studying estrogen in context of PMDD is important because women with PMDD can have low mood, specific food cravings, and impaired cognitive function.

Estrogen–serotonin interactions are thought to be involved in hormone-related mood disorders such as perimenopausal depression and PMDD.^23,24 However, the nature of their relationship is not yet fully understood. Ovariectomized animals have shown estrogen-induced changes related to serotonin metabolism, binding, and transmission in the regions of the brain involved in regulation of affect and cognition. Research in menopausal women also has provided some support for this interaction.²⁴

Positron emission tomography studies in humans have found increased cortical serotonin binding modulated by levels of estrogen, similar to those previously seen in rat studies.^24-27 One study showed an increased binding potential of serotonin in the cerebral cortex with estrogen treatment. This study further showed an even greater binding potential with estrogen plus progesterone, signaling a synergistic effect of the 2 hormones.²⁸

SSRIs are an effective treatment for the irritability, anxiety, and mood swings of PMDD.^29-30 Although the exact mechanism of action is unknown, the serotonergic properties are certainly of primary attention. For some PMDD patients, SSRIs work within hours to days, as opposed to days or weeks for patients with depression or anxiety, which suggests a separate or co-occurring mechanism of action is in place. In a double-blind, placebo-controlled crossover study, researchers administered the serotonin receptor antagonist metergoline to women with PMDD whose symptoms had remitted during treatment with fluoxetine and a group of healthy controls who were not receiving any medication.³¹ The women with PMDD experienced a return of symptoms 24 hours after treatment with metergoline but not with placebo; the controls experienced no mood changes.³¹

BDNF is a neurotransmitter linked to estrogen and likely related to PMDD. BDNF is critical for neurogenesis and is expressed in brain regions involved in learning and memory and also affects regulation.³² BDNF levels are increased by serotonergic antidepressants, affected by estradiol, and have cyclicity throughout the menstrual cycle.^33-35

Putative brain structural and functional differences. Imaging studies have suggested differences in brain structure in women with PMDD, with a focus on the amygdala and the prefrontal cortex. Women with PMDD have greater gray matter volume in the posterior cerebellum,³⁶ greater gray matter density of hippocampal cortex, and lower gray matter density in the parahippocampal cortex.³⁷

Some studies have shown a functional variability of the amygdala’s response to stress in women with PMDD vs healthy controls.^38,39 A proton magnetic resonance spectroscopy (1H-MRS) study of the displays the possibility of an altered GABAergic function in patients with PMDD.⁴⁰

Patients will PMDD have enhanced dorsolateral prefrontal cortex reactivity when anticipating negative stimuli (but not to the actual exposure) during the luteal phase. A positive correlation between this reactivity and progesterone levels also was observed.⁴¹ Some researchers have suggested that prefrontal cortex dysfunction may be a risk factor for PMDD.⁴²

HPA axis and HPG axis: Trauma, resiliency, inflammation. Altered cortisol levels (higher during the luteal phase⁴³ and lower during times of stress^14,44) suggest a possibly altered HPA axis in some women with PMDD. However, studies on this topic have been few and inconsistent.

Dysregulation of the HPG axis could cause vasomotor symptoms, sleep dysregulation, and mood symptoms during menopause; women with PMDD can also experience these symptoms. The influence of estrogen and progesterone on mood is also highly dependent on this axis.

Ultimately, the interplay between the HPA axis and the HPG axis is important. One study found that women with PMDD who had high serum ALLO levels (HPG-related) had blunted nocturnal cortisol levels (HPA-related) compared with healthy controls who had low ALLO levels.⁴⁵

Significant stress and trauma exposure have been associated with PMDD. A study of 3,968 women found a history of trauma and PTSD were independently associated with PMDD.⁴⁶ Another study of approximately 3,000 women found a strong correlation between abuse and PMS.⁴⁷ However, a third study found no correlations between PMDD and trauma.⁴⁸

Patients with a predisposition to PMDD may be more vulnerable to develop a posttraumatic stress-related disorder, perhaps due to decreased biologic resiliency. For example, the startle response (hyper­vigilance) has been shown to be different in women with PMDD. One study suggested that suboptimal production of premenstrual ALLO may lead to increased arousal and increased stress reactivity to psycho­social or environmental triggers.⁴⁹

The possible role of inflammation in PMDD deserves further investigation. The luteal phase entails an increase in the production of proinflammatory markers.^50,51 A 10-fold increase in progesterone is correlated with a 20% to 23% increase in C-reactive protein levels.^52,53 Women with inflammatory diseases (eg, gingivitis or irritable bowel syndrome) show worsening of symptoms prior to menstruation.^54-56 One study found increased levels of proinflammatory markers in women with PMDD compared with controls.⁵⁷

Because PMDD is heritable, it must have an underlying neurobiologic pathophysiology. Brain imaging studies show differences in structure and function in women with PMDD across the menstrual cycle. Conversion of progesterone to ALLO and the GABAergic influence of this metabolite is a topic of interest in current research. Similarly, the role of estrogen and its connection to serotonin and other neurotransmitters such as BDNF have been implicated.

The link between a history of stress, trauma, and PMDD raises the question of biologic resiliency and illness in these patients, as it connects to the HPA and HPG axis and production of inflammatory stress hormones and steroid hormones and their metabolites. PMDD can be conceptualized as variable sensitivity to hormonal response to stress,⁵⁸ thus contextualizing biochemical and psychological resiliency.

Further research is needed to clarify the possibility of a shared pathophysiology between endocrine-related mood disorders such as postpartum depression (PPD) and PMDD because current research is controversial.^59,60 In PPD, women who are exposed to high levels of progesterone and estrogen during pregnancy (just like in the mid-luteal phase) have a sudden drop in these hormones postpartum.

The ‘withdrawal theory.’ The affective symptoms of PMDD resolve almost instantaneously after the start of menstruation. Perhaps this type of immediate relief is akin to substance use disorders and symptoms of withdrawal. It could be that reinstatement of a certain amount of gonadal steroids in the follicular phase of the cycle diminishes a withdrawal-like response to these steroids.

Currently, the main leading theory is that PMDD is a result of “an abnormal response to normal hormonal changes.”⁶¹ A new study also has shown that the change in estradiol/progesterone levels (vs the steady state) was associated with PMDD symptoms.⁶² Thinking of PMDD as a disorder of withdrawal offers an alternative (yet complementary) perspective to the current theory: PMDD may be caused by the absence or diminishing of the above-named hormones and their metabolites in the late luteal phase (in the context of developed “tolerance” during the early- to mid-luteal phase).

Considering the interplay between neuro­transmitters and neurosteroids, both a “serotonin withdrawal theory” (caused by a drop in steroid hormones) and a “GABAergic withdrawal theory” (due to the decline in progesterone) could be proposed. This theory would be supported by the fact that SSRIs seem to mitigate symptoms of PMDD as well as the genetic association between PMDD and ESR1. It is more than likely that the “withdrawal” is caused by the interactions between estrogen-serotonin, progesterone-ALLO, and GABA receptors, and the complementary fashion in which progesterone and estrogen influence each other.

Like all physicians, psychiatrists practice in an increasingly complex health care environment, with escalating demands for productivity, rising threats of malpractice, expanding clinical oversight, and growing concerns about income. Additionally, psychiatric practice presents its own challenges, including limited resources and concerns about patient violence and suicide. These concerns can make it difficult to establish a healthy work–life balance.

Physicians, including psychiatrists, are at risk for alcohol or substance abuse/dependency, burnout, and suicide. As psychiatrists, we need to attend to our own personal and professional health so that we can best help our patients. This review focuses on the challenges psychiatrists face that can adversely affect their well-being and offers strategies to reduce the risk of burnout and enhance wellness.

The challenges of medicine and their impact on psychiatrists

The practice of medicine is inherently challenging. It requires hard work, discipline, dedication, and faithfulness to high ethical standards. Additional challenges include declining autonomy and opportunities for social support, increasing accountability, and a growing interest in reducing the cost of care by employing more non-physician health professionals—which in psychiatry typically include psychologists, nurse practitioners, and social workers. The uncertainty of the Affordable Care Act, declining income, and concerns about the nature of future medical practice are also stressors.^1,2

Factors that contribute to psychiatrists’ stress include:

• limited resources
• concerns about patient violence and suicide
• crowded inpatient units
• changing culture in mental health services
• high work demands
• poorly defined roles of consultants
• declining authority
• frustration with the inability to impact systemic change
• conflict between responsibility toward employers vs the patient
• isolation.³

Concern about patient suicide is a significant stressor.^4,5 Some evidence suggests that the impact of a patient’s suicide on a physician is more severe when it occurs during training than after graduation and is inversely correlated with the clinician’s perceived social integration into their professional network.⁵

Impediments to a physician’s well-being

Alcohol abuse/dependence. Approximately 13% of male physicians and 21% of female physicians meet Alcohol Use Disorders Identification Test Version C criteria for alcohol abuse or dependence, according to a study of approximately 7,300 U.S. physicians from all specialties.⁶ (In this study, prescription drug abuse and use of illicit drugs were rare.) Age, hours worked, male sex, being married or partnered, having children, and being in a specialty other than internal medicine were independently associated with alcohol abuse or dependence.

Fortunately, psychiatrists were among the specialties with below average likelihood to meet diagnostic criteria for alcohol abuse/dependency.⁶ However, alcohol abuse or dependency was associated with burnout, depression, suicidal ideation, lower quality of life, lower career satisfaction, and medical errors.

Burnout is a long-term stress reaction consisting of:

• physical and emotional exhaustion (feeling depleted)
• depersonalization (cynicism, lack of engagement with or negative attitudes toward patients)
• reduced sense of personal accomplishment (lack of a sense of purpose).⁷

In a 2017 survey of >14,000 U.S. physicians from 27 specialties, 42% of psychiatrists reported burnout.⁸ In another survey of approximately 300 resident physicians across all specialties in a tertiary academic hospital, 69% met criteria for burnout.⁹ This condition affects resident physicians as well as those in practice. Residents and program directors cited a lack of work–life balance and feeling unappreciated as factors contributing to burnout.

Among physicians, factors that contribute to burnout include loss of autonomy, diminished status as physicians, and increased work pressures. Burnout has a negative impact on both patients and health care systems. It is associated with an increased risk of depression and can contribute to:

• broken relationships
• alcohol abuse
• physician suicide
• decreased quality of care, including patient safety and satisfaction
• increased risk of malpractice suits
• reduced patient adherence to medical recommendations.^5,10-12

Physicians who embrace medicine as a calling (ie, committing one’s life to personally meaningful work that serves a prosocial purpose) experience less burnout. According to a survey of approximately 900 primary care physicians and 300 psychiatrists, 42% of psychiatrists strongly agreed that medicine is a calling.¹³ Overall, physicians with a high sense of calling reported less burnout than those with a lower sense of calling (17% vs 31%, respectively).¹³

Depression and suicide. Gold et al¹² analyzed a database that included information on approximately 31,600 adult suicide victims, and 203 of these victims were physicians. Compared with others, physicians were more likely to have a diagnosed mental illness or an occupation-related problem that contributed to suicide. Toxicology results also showed that physician suicide victims were significantly more likely than non-physician victims to test positive for benzodiazepines and barbiturates, but not antidepressants, which suggests that physicians with depression may not have been receiving adequate treatment.¹²

Although occupation-related stress and inadequate mental health treatment may be modifiable risk factors to reduce suicide deaths among physicians, stigma and fear of medical staff and licensure issues may deter physicians from seeking treatment.¹⁴

Steps to avoid burnout

Evidence-based interventions. There is limited evidence-based data regarding specific interventions for preventing burnout and reducing stress among physicians, particularly among psychiatrists.⁴

A randomized controlled trial of 74 practicing physicians at the Mayo Clinic in Rochester, Minnesota, evaluated the effectiveness of 19 biweekly physician-facilitated discussion groups.¹⁵ The groups covered topics such as elements of mindfulness, reflection, shared experience, and small-group learning. The institution provided 1 hour of paid time every other week for physicians to participate in this program. Physicians in the control group could schedule and use this time as they chose. Researchers also collected data on 350 non-trial participants.

The proportion of participants who strongly agreed that their work was meaningful increased 6.3% in the intervention group but decreased 6.3% in the control group and 13.4% among non-trial participants (P = .04).¹⁵ Rates of depersonalization, emotional exhaustion, and overall burnout decreased substantially in the intervention group, decreased slightly in the control group, and increased in the non-trial cohort. Results were sustained at 12 months after the study. There were no statistically significant differences in stress, symptoms of depression, overall quality of life, or job satisfaction.¹⁵

Preliminary evidence suggests that residents and fellows would find a wellness or suicide prevention program helpful. One study found that the use of one such program, which provided individual counseling, psychiatric evaluation, and wellness workshops for residents, fellows, and faculty in an academic health center, increased from 5% to 25% of eligible participants, and participants reported high levels of satisfaction with the program.¹⁶ Such programs would require institutional support for space and clinical staff.¹⁵

Empathy. As psychiatrists, we are taught to be empathetic. Yet, with the numerous challenges we face, it is not always easy. Stressors such as an increased workload or burnout can adversely affect a psychiatrist’s ability to provide empathetic care.¹⁷ However, empathetic treatment has clear benefits for both physicians and patients. Empathic skills can lead to more professional satisfaction and outcomes, which are important components of accountability, and can:

• promote patient satisfaction
• establish trust
• reduce anxiety
• increase adherence to treatment regimens
• improve health outcomes
• decrease the likelihood of malpractice suits.¹⁷

Mindfulness is a “flexible state of mind in which we are actively engaged in the present, noticing new things and sensitive to context.”^18,19 It may sound mundane to cling to phrases such as “living in the present,” but mindfulness can be a valuable tool for psychiatrists who struggle to maintain well-being in medicine’s challenging milieu. The process of mindfulness—actively drawing distinctions and noticing new things, “seeing the familiar in the novel and the novel in the familiar”—can ensure that we have active minds, that we are involved, and that we are capturing the joy of living in the stimulating present.¹⁸

Focus on issues you can control

Many of the factors that negatively influence professional satisfaction and well-being, such as loss of autonomy, demand for increased patient care volume, and increasing scrutiny on the quality of care, are beyond a psychiatrist’s control. Medical administrators can help reduce some of these issues by increasing physician autonomy, offering physicians the opportunity to work part-time, offering medical staff workshops to enhance positive communication, or addressing leader­ship problems. However, psychiatrists may benefit most by identifying modifiable issues under their own control, such as prioritizing a work–life balance, applying the fundamentals of a health prevention strategy to their own lives (Box^20,21), approaching medicine as a calling, embracing an empathetic approach to patient care, and bringing mindfulness to medical practice.

Like all physicians, psychiatrists practice in an increasingly complex health care environment, with escalating demands for productivity, rising threats of malpractice, expanding clinical oversight, and growing concerns about income. Additionally, psychiatric practice presents its own challenges, including limited resources and concerns about patient violence and suicide. These concerns can make it difficult to establish a healthy work–life balance.

Physicians, including psychiatrists, are at risk for alcohol or substance abuse/dependency, burnout, and suicide. As psychiatrists, we need to attend to our own personal and professional health so that we can best help our patients. This review focuses on the challenges psychiatrists face that can adversely affect their well-being and offers strategies to reduce the risk of burnout and enhance wellness.

The challenges of medicine and their impact on psychiatrists

The practice of medicine is inherently challenging. It requires hard work, discipline, dedication, and faithfulness to high ethical standards. Additional challenges include declining autonomy and opportunities for social support, increasing accountability, and a growing interest in reducing the cost of care by employing more non-physician health professionals—which in psychiatry typically include psychologists, nurse practitioners, and social workers. The uncertainty of the Affordable Care Act, declining income, and concerns about the nature of future medical practice are also stressors.^1,2

Factors that contribute to psychiatrists’ stress include:

• limited resources
• concerns about patient violence and suicide
• crowded inpatient units
• changing culture in mental health services
• high work demands
• poorly defined roles of consultants
• declining authority
• frustration with the inability to impact systemic change
• conflict between responsibility toward employers vs the patient
• isolation.³

Concern about patient suicide is a significant stressor.^4,5 Some evidence suggests that the impact of a patient’s suicide on a physician is more severe when it occurs during training than after graduation and is inversely correlated with the clinician’s perceived social integration into their professional network.⁵

Impediments to a physician’s well-being

Alcohol abuse/dependence. Approximately 13% of male physicians and 21% of female physicians meet Alcohol Use Disorders Identification Test Version C criteria for alcohol abuse or dependence, according to a study of approximately 7,300 U.S. physicians from all specialties.⁶ (In this study, prescription drug abuse and use of illicit drugs were rare.) Age, hours worked, male sex, being married or partnered, having children, and being in a specialty other than internal medicine were independently associated with alcohol abuse or dependence.

Fortunately, psychiatrists were among the specialties with below average likelihood to meet diagnostic criteria for alcohol abuse/dependency.⁶ However, alcohol abuse or dependency was associated with burnout, depression, suicidal ideation, lower quality of life, lower career satisfaction, and medical errors.

Burnout is a long-term stress reaction consisting of:

• physical and emotional exhaustion (feeling depleted)
• depersonalization (cynicism, lack of engagement with or negative attitudes toward patients)
• reduced sense of personal accomplishment (lack of a sense of purpose).⁷

In a 2017 survey of >14,000 U.S. physicians from 27 specialties, 42% of psychiatrists reported burnout.⁸ In another survey of approximately 300 resident physicians across all specialties in a tertiary academic hospital, 69% met criteria for burnout.⁹ This condition affects resident physicians as well as those in practice. Residents and program directors cited a lack of work–life balance and feeling unappreciated as factors contributing to burnout.

Among physicians, factors that contribute to burnout include loss of autonomy, diminished status as physicians, and increased work pressures. Burnout has a negative impact on both patients and health care systems. It is associated with an increased risk of depression and can contribute to:

• broken relationships
• alcohol abuse
• physician suicide
• decreased quality of care, including patient safety and satisfaction
• increased risk of malpractice suits
• reduced patient adherence to medical recommendations.^5,10-12

Physicians who embrace medicine as a calling (ie, committing one’s life to personally meaningful work that serves a prosocial purpose) experience less burnout. According to a survey of approximately 900 primary care physicians and 300 psychiatrists, 42% of psychiatrists strongly agreed that medicine is a calling.¹³ Overall, physicians with a high sense of calling reported less burnout than those with a lower sense of calling (17% vs 31%, respectively).¹³

Depression and suicide. Gold et al¹² analyzed a database that included information on approximately 31,600 adult suicide victims, and 203 of these victims were physicians. Compared with others, physicians were more likely to have a diagnosed mental illness or an occupation-related problem that contributed to suicide. Toxicology results also showed that physician suicide victims were significantly more likely than non-physician victims to test positive for benzodiazepines and barbiturates, but not antidepressants, which suggests that physicians with depression may not have been receiving adequate treatment.¹²

Although occupation-related stress and inadequate mental health treatment may be modifiable risk factors to reduce suicide deaths among physicians, stigma and fear of medical staff and licensure issues may deter physicians from seeking treatment.¹⁴

Steps to avoid burnout

Evidence-based interventions. There is limited evidence-based data regarding specific interventions for preventing burnout and reducing stress among physicians, particularly among psychiatrists.⁴

A randomized controlled trial of 74 practicing physicians at the Mayo Clinic in Rochester, Minnesota, evaluated the effectiveness of 19 biweekly physician-facilitated discussion groups.¹⁵ The groups covered topics such as elements of mindfulness, reflection, shared experience, and small-group learning. The institution provided 1 hour of paid time every other week for physicians to participate in this program. Physicians in the control group could schedule and use this time as they chose. Researchers also collected data on 350 non-trial participants.

The proportion of participants who strongly agreed that their work was meaningful increased 6.3% in the intervention group but decreased 6.3% in the control group and 13.4% among non-trial participants (P = .04).¹⁵ Rates of depersonalization, emotional exhaustion, and overall burnout decreased substantially in the intervention group, decreased slightly in the control group, and increased in the non-trial cohort. Results were sustained at 12 months after the study. There were no statistically significant differences in stress, symptoms of depression, overall quality of life, or job satisfaction.¹⁵

Preliminary evidence suggests that residents and fellows would find a wellness or suicide prevention program helpful. One study found that the use of one such program, which provided individual counseling, psychiatric evaluation, and wellness workshops for residents, fellows, and faculty in an academic health center, increased from 5% to 25% of eligible participants, and participants reported high levels of satisfaction with the program.¹⁶ Such programs would require institutional support for space and clinical staff.¹⁵

Empathy. As psychiatrists, we are taught to be empathetic. Yet, with the numerous challenges we face, it is not always easy. Stressors such as an increased workload or burnout can adversely affect a psychiatrist’s ability to provide empathetic care.¹⁷ However, empathetic treatment has clear benefits for both physicians and patients. Empathic skills can lead to more professional satisfaction and outcomes, which are important components of accountability, and can:

• promote patient satisfaction
• establish trust
• reduce anxiety
• increase adherence to treatment regimens
• improve health outcomes
• decrease the likelihood of malpractice suits.¹⁷

Mindfulness is a “flexible state of mind in which we are actively engaged in the present, noticing new things and sensitive to context.”^18,19 It may sound mundane to cling to phrases such as “living in the present,” but mindfulness can be a valuable tool for psychiatrists who struggle to maintain well-being in medicine’s challenging milieu. The process of mindfulness—actively drawing distinctions and noticing new things, “seeing the familiar in the novel and the novel in the familiar”—can ensure that we have active minds, that we are involved, and that we are capturing the joy of living in the stimulating present.¹⁸

Focus on issues you can control

Many of the factors that negatively influence professional satisfaction and well-being, such as loss of autonomy, demand for increased patient care volume, and increasing scrutiny on the quality of care, are beyond a psychiatrist’s control. Medical administrators can help reduce some of these issues by increasing physician autonomy, offering physicians the opportunity to work part-time, offering medical staff workshops to enhance positive communication, or addressing leader­ship problems. However, psychiatrists may benefit most by identifying modifiable issues under their own control, such as prioritizing a work–life balance, applying the fundamentals of a health prevention strategy to their own lives (Box^20,21), approaching medicine as a calling, embracing an empathetic approach to patient care, and bringing mindfulness to medical practice.

Like all physicians, psychiatrists practice in an increasingly complex health care environment, with escalating demands for productivity, rising threats of malpractice, expanding clinical oversight, and growing concerns about income. Additionally, psychiatric practice presents its own challenges, including limited resources and concerns about patient violence and suicide. These concerns can make it difficult to establish a healthy work–life balance.

Physicians, including psychiatrists, are at risk for alcohol or substance abuse/dependency, burnout, and suicide. As psychiatrists, we need to attend to our own personal and professional health so that we can best help our patients. This review focuses on the challenges psychiatrists face that can adversely affect their well-being and offers strategies to reduce the risk of burnout and enhance wellness.

The challenges of medicine and their impact on psychiatrists

The practice of medicine is inherently challenging. It requires hard work, discipline, dedication, and faithfulness to high ethical standards. Additional challenges include declining autonomy and opportunities for social support, increasing accountability, and a growing interest in reducing the cost of care by employing more non-physician health professionals—which in psychiatry typically include psychologists, nurse practitioners, and social workers. The uncertainty of the Affordable Care Act, declining income, and concerns about the nature of future medical practice are also stressors.^1,2

Factors that contribute to psychiatrists’ stress include:

• limited resources
• concerns about patient violence and suicide
• crowded inpatient units
• changing culture in mental health services
• high work demands
• poorly defined roles of consultants
• declining authority
• frustration with the inability to impact systemic change
• conflict between responsibility toward employers vs the patient
• isolation.³

Concern about patient suicide is a significant stressor.^4,5 Some evidence suggests that the impact of a patient’s suicide on a physician is more severe when it occurs during training than after graduation and is inversely correlated with the clinician’s perceived social integration into their professional network.⁵

Impediments to a physician’s well-being

Alcohol abuse/dependence. Approximately 13% of male physicians and 21% of female physicians meet Alcohol Use Disorders Identification Test Version C criteria for alcohol abuse or dependence, according to a study of approximately 7,300 U.S. physicians from all specialties.⁶ (In this study, prescription drug abuse and use of illicit drugs were rare.) Age, hours worked, male sex, being married or partnered, having children, and being in a specialty other than internal medicine were independently associated with alcohol abuse or dependence.

Fortunately, psychiatrists were among the specialties with below average likelihood to meet diagnostic criteria for alcohol abuse/dependency.⁶ However, alcohol abuse or dependency was associated with burnout, depression, suicidal ideation, lower quality of life, lower career satisfaction, and medical errors.

Burnout is a long-term stress reaction consisting of:

• physical and emotional exhaustion (feeling depleted)
• depersonalization (cynicism, lack of engagement with or negative attitudes toward patients)
• reduced sense of personal accomplishment (lack of a sense of purpose).⁷

In a 2017 survey of >14,000 U.S. physicians from 27 specialties, 42% of psychiatrists reported burnout.⁸ In another survey of approximately 300 resident physicians across all specialties in a tertiary academic hospital, 69% met criteria for burnout.⁹ This condition affects resident physicians as well as those in practice. Residents and program directors cited a lack of work–life balance and feeling unappreciated as factors contributing to burnout.

Among physicians, factors that contribute to burnout include loss of autonomy, diminished status as physicians, and increased work pressures. Burnout has a negative impact on both patients and health care systems. It is associated with an increased risk of depression and can contribute to:

• broken relationships
• alcohol abuse
• physician suicide
• decreased quality of care, including patient safety and satisfaction
• increased risk of malpractice suits
• reduced patient adherence to medical recommendations.^5,10-12

Physicians who embrace medicine as a calling (ie, committing one’s life to personally meaningful work that serves a prosocial purpose) experience less burnout. According to a survey of approximately 900 primary care physicians and 300 psychiatrists, 42% of psychiatrists strongly agreed that medicine is a calling.¹³ Overall, physicians with a high sense of calling reported less burnout than those with a lower sense of calling (17% vs 31%, respectively).¹³

Depression and suicide. Gold et al¹² analyzed a database that included information on approximately 31,600 adult suicide victims, and 203 of these victims were physicians. Compared with others, physicians were more likely to have a diagnosed mental illness or an occupation-related problem that contributed to suicide. Toxicology results also showed that physician suicide victims were significantly more likely than non-physician victims to test positive for benzodiazepines and barbiturates, but not antidepressants, which suggests that physicians with depression may not have been receiving adequate treatment.¹²

Although occupation-related stress and inadequate mental health treatment may be modifiable risk factors to reduce suicide deaths among physicians, stigma and fear of medical staff and licensure issues may deter physicians from seeking treatment.¹⁴

Steps to avoid burnout

Evidence-based interventions. There is limited evidence-based data regarding specific interventions for preventing burnout and reducing stress among physicians, particularly among psychiatrists.⁴

A randomized controlled trial of 74 practicing physicians at the Mayo Clinic in Rochester, Minnesota, evaluated the effectiveness of 19 biweekly physician-facilitated discussion groups.¹⁵ The groups covered topics such as elements of mindfulness, reflection, shared experience, and small-group learning. The institution provided 1 hour of paid time every other week for physicians to participate in this program. Physicians in the control group could schedule and use this time as they chose. Researchers also collected data on 350 non-trial participants.

The proportion of participants who strongly agreed that their work was meaningful increased 6.3% in the intervention group but decreased 6.3% in the control group and 13.4% among non-trial participants (P = .04).¹⁵ Rates of depersonalization, emotional exhaustion, and overall burnout decreased substantially in the intervention group, decreased slightly in the control group, and increased in the non-trial cohort. Results were sustained at 12 months after the study. There were no statistically significant differences in stress, symptoms of depression, overall quality of life, or job satisfaction.¹⁵

Preliminary evidence suggests that residents and fellows would find a wellness or suicide prevention program helpful. One study found that the use of one such program, which provided individual counseling, psychiatric evaluation, and wellness workshops for residents, fellows, and faculty in an academic health center, increased from 5% to 25% of eligible participants, and participants reported high levels of satisfaction with the program.¹⁶ Such programs would require institutional support for space and clinical staff.¹⁵

Empathy. As psychiatrists, we are taught to be empathetic. Yet, with the numerous challenges we face, it is not always easy. Stressors such as an increased workload or burnout can adversely affect a psychiatrist’s ability to provide empathetic care.¹⁷ However, empathetic treatment has clear benefits for both physicians and patients. Empathic skills can lead to more professional satisfaction and outcomes, which are important components of accountability, and can:

• promote patient satisfaction
• establish trust
• reduce anxiety
• increase adherence to treatment regimens
• improve health outcomes
• decrease the likelihood of malpractice suits.¹⁷

Mindfulness is a “flexible state of mind in which we are actively engaged in the present, noticing new things and sensitive to context.”^18,19 It may sound mundane to cling to phrases such as “living in the present,” but mindfulness can be a valuable tool for psychiatrists who struggle to maintain well-being in medicine’s challenging milieu. The process of mindfulness—actively drawing distinctions and noticing new things, “seeing the familiar in the novel and the novel in the familiar”—can ensure that we have active minds, that we are involved, and that we are capturing the joy of living in the stimulating present.¹⁸

Focus on issues you can control

Many of the factors that negatively influence professional satisfaction and well-being, such as loss of autonomy, demand for increased patient care volume, and increasing scrutiny on the quality of care, are beyond a psychiatrist’s control. Medical administrators can help reduce some of these issues by increasing physician autonomy, offering physicians the opportunity to work part-time, offering medical staff workshops to enhance positive communication, or addressing leader­ship problems. However, psychiatrists may benefit most by identifying modifiable issues under their own control, such as prioritizing a work–life balance, applying the fundamentals of a health prevention strategy to their own lives (Box^20,21), approaching medicine as a calling, embracing an empathetic approach to patient care, and bringing mindfulness to medical practice.

An abundance of recent neuroscience advances is directly related to psychiatric disorders, because the primary mission of the brain is to generate a mind, and every new discovery provides another piece of the psychiatric disorders puzzle. The time also is ripe to incorporate clinical neuroscience concepts and language in our clinical practice and terminology. The neuroscientification of clinical psychiatry must start with clinical neuro­science literacy.

Although the traditional training of psychiatrists has evolved, it continues to perpetuate the old-fashioned model of care exemplified by the mental status examination, which documents the patient’s appearance, speech, mood, affect, thoughts, perceptions, behavior, cognition, insight, and judgement. Evaluations and progress notes have been constrained by this decades-old formula of observing, interviewing, and documenting signs and symptoms, and arriving at a working diagnosis, followed by a treatment plan comprised of a cluster of drug names, psychotherapeutic modalities, and social or rehabilitation interventions. This widely accepted procedure is important because it focuses on the mind. But where are the details about the brain, whose structural and functional aberrations generate the anomalies of the mind and are the scientific foundations of psychiatric care?

All psychiatrists are fully aware that brain pathology is the source of every psychiatric disorder they evaluate, diagnose, and treat. But it is time to formulate every patient’s care using neuroscience data and include neural mechanisms of the psychiatric disorder in the chart. Our clinical language must be integrated with the rapidly growing neuroscience of abnormalities in brain–behavior links.

Psychiatry is lagging behind neuro­logy, its sister brain specialty, where neural pathways and processes are front and center in describing symptoms. According to Eisenberg,¹ psychiatry training in the 1980s was, for the most part, “brainless.” But it should not remain so, because neuroscience advances have skyrocketed since he made that provocative statement 3 decades ago. Yet, the psychiatric residency training curriculum in many programs is lagging behind the rapid evolution of psychiatry as a clinical neuroscience.²

To its credit, the Accreditation Council for Graduate Medical Education, which oversees and accredits residency training programs in all specialties, including psychiatry, recently announced that psychiatric residency training must emphasize neuroscience competence side-by-side with clinical competence. Psychiatric residents must increasingly incorporate neurobiology in their formulation of clinical care and determine how the selected pharmacologic therapy addresses the dysregulated neural circuitry underlying the clinical manifestation. A good example of this method is a recently published case of posttraumatic stress disorder (PTSD),³ which discussed the clinical components and treatment of this brain disorder through the prism of clinical neuroscience research data. PTSD “trauma” is not only psychological, but also neurobiological, and both must be incorporated in formulating a clinical case.

Another important step has emerged to focus on infusing neuroscience facts and concepts within the clinical training of psychiatric residents. The National Neuroscience Curriculum Initiative (www.nncionline.org) is a timely and welcome initiative that will aggressively promulgate a clinical neuro­scientification of psychiatric training, triggering a roadmap for modern, cutting-edge psychiatric practice.⁴ This will help consolidate psychiatry’s rightful place as a clinical neuroscience, without relinquishing its biopsycho­social roots.

As research continues to elucidate the neural mechanisms of key psychiatric symptoms, such as anxiety, depression, mania, impulsiveness, compulsions, delusions, or hallucinations, the trans­formation of psychiatry into an authentic clinical neuroscience is inevitable. But contemporary psychiatric practitioners must retool and start their journey toward neuroscience literacy by attending relevant continuing medical education presentations and regularly reading journals that focus on clinical psychiatric neuro­science, such as Molecular Psychiatry, JAMA Psychiatry, Biological Psychiatry, Neuropsychopharmacology, and Progress in Neuro-psychopharmacology and Biological Psychiatry. Current Psychiatry will do its part by establishing a section on psychiatric neuroscience for our readers. Dr. Stephen Stahl’s recent guest editorial in Current Psychiatry⁵ about Neuroscience-based Nomenclature is another important step toward the neuroscientification of psychiatric medications’ names, using the neural mechanism of action nomenclature instead of a single clinical indication (eg, antidepressant or antipsychotic) when there may be several approved uses for the medication.

It is my sincere hope that my fellow clinical psychiatrists will steadily grow their clinical neuroscience literacy and apply it to daily patient care. By formulating psychiatric signs and symptoms in evidence-based, neurobiological frameworks provided by clinical neuro­science advances, the myths, shibboleths, and misperceptions of what causes mental illness will dissipate and eventually vanish. And so will the stigma that shrouds our patients who suffer from brain disorders that manifest as disruptions of 1 or more complex functions of the human mind.

An abundance of recent neuroscience advances is directly related to psychiatric disorders, because the primary mission of the brain is to generate a mind, and every new discovery provides another piece of the psychiatric disorders puzzle. The time also is ripe to incorporate clinical neuroscience concepts and language in our clinical practice and terminology. The neuroscientification of clinical psychiatry must start with clinical neuro­science literacy.

Although the traditional training of psychiatrists has evolved, it continues to perpetuate the old-fashioned model of care exemplified by the mental status examination, which documents the patient’s appearance, speech, mood, affect, thoughts, perceptions, behavior, cognition, insight, and judgement. Evaluations and progress notes have been constrained by this decades-old formula of observing, interviewing, and documenting signs and symptoms, and arriving at a working diagnosis, followed by a treatment plan comprised of a cluster of drug names, psychotherapeutic modalities, and social or rehabilitation interventions. This widely accepted procedure is important because it focuses on the mind. But where are the details about the brain, whose structural and functional aberrations generate the anomalies of the mind and are the scientific foundations of psychiatric care?

All psychiatrists are fully aware that brain pathology is the source of every psychiatric disorder they evaluate, diagnose, and treat. But it is time to formulate every patient’s care using neuroscience data and include neural mechanisms of the psychiatric disorder in the chart. Our clinical language must be integrated with the rapidly growing neuroscience of abnormalities in brain–behavior links.

Psychiatry is lagging behind neuro­logy, its sister brain specialty, where neural pathways and processes are front and center in describing symptoms. According to Eisenberg,¹ psychiatry training in the 1980s was, for the most part, “brainless.” But it should not remain so, because neuroscience advances have skyrocketed since he made that provocative statement 3 decades ago. Yet, the psychiatric residency training curriculum in many programs is lagging behind the rapid evolution of psychiatry as a clinical neuroscience.²

To its credit, the Accreditation Council for Graduate Medical Education, which oversees and accredits residency training programs in all specialties, including psychiatry, recently announced that psychiatric residency training must emphasize neuroscience competence side-by-side with clinical competence. Psychiatric residents must increasingly incorporate neurobiology in their formulation of clinical care and determine how the selected pharmacologic therapy addresses the dysregulated neural circuitry underlying the clinical manifestation. A good example of this method is a recently published case of posttraumatic stress disorder (PTSD),³ which discussed the clinical components and treatment of this brain disorder through the prism of clinical neuroscience research data. PTSD “trauma” is not only psychological, but also neurobiological, and both must be incorporated in formulating a clinical case.

Another important step has emerged to focus on infusing neuroscience facts and concepts within the clinical training of psychiatric residents. The National Neuroscience Curriculum Initiative (www.nncionline.org) is a timely and welcome initiative that will aggressively promulgate a clinical neuro­scientification of psychiatric training, triggering a roadmap for modern, cutting-edge psychiatric practice.⁴ This will help consolidate psychiatry’s rightful place as a clinical neuroscience, without relinquishing its biopsycho­social roots.

As research continues to elucidate the neural mechanisms of key psychiatric symptoms, such as anxiety, depression, mania, impulsiveness, compulsions, delusions, or hallucinations, the trans­formation of psychiatry into an authentic clinical neuroscience is inevitable. But contemporary psychiatric practitioners must retool and start their journey toward neuroscience literacy by attending relevant continuing medical education presentations and regularly reading journals that focus on clinical psychiatric neuro­science, such as Molecular Psychiatry, JAMA Psychiatry, Biological Psychiatry, Neuropsychopharmacology, and Progress in Neuro-psychopharmacology and Biological Psychiatry. Current Psychiatry will do its part by establishing a section on psychiatric neuroscience for our readers. Dr. Stephen Stahl’s recent guest editorial in Current Psychiatry⁵ about Neuroscience-based Nomenclature is another important step toward the neuroscientification of psychiatric medications’ names, using the neural mechanism of action nomenclature instead of a single clinical indication (eg, antidepressant or antipsychotic) when there may be several approved uses for the medication.

It is my sincere hope that my fellow clinical psychiatrists will steadily grow their clinical neuroscience literacy and apply it to daily patient care. By formulating psychiatric signs and symptoms in evidence-based, neurobiological frameworks provided by clinical neuro­science advances, the myths, shibboleths, and misperceptions of what causes mental illness will dissipate and eventually vanish. And so will the stigma that shrouds our patients who suffer from brain disorders that manifest as disruptions of 1 or more complex functions of the human mind.

An abundance of recent neuroscience advances is directly related to psychiatric disorders, because the primary mission of the brain is to generate a mind, and every new discovery provides another piece of the psychiatric disorders puzzle. The time also is ripe to incorporate clinical neuroscience concepts and language in our clinical practice and terminology. The neuroscientification of clinical psychiatry must start with clinical neuro­science literacy.

Although the traditional training of psychiatrists has evolved, it continues to perpetuate the old-fashioned model of care exemplified by the mental status examination, which documents the patient’s appearance, speech, mood, affect, thoughts, perceptions, behavior, cognition, insight, and judgement. Evaluations and progress notes have been constrained by this decades-old formula of observing, interviewing, and documenting signs and symptoms, and arriving at a working diagnosis, followed by a treatment plan comprised of a cluster of drug names, psychotherapeutic modalities, and social or rehabilitation interventions. This widely accepted procedure is important because it focuses on the mind. But where are the details about the brain, whose structural and functional aberrations generate the anomalies of the mind and are the scientific foundations of psychiatric care?

All psychiatrists are fully aware that brain pathology is the source of every psychiatric disorder they evaluate, diagnose, and treat. But it is time to formulate every patient’s care using neuroscience data and include neural mechanisms of the psychiatric disorder in the chart. Our clinical language must be integrated with the rapidly growing neuroscience of abnormalities in brain–behavior links.

Psychiatry is lagging behind neuro­logy, its sister brain specialty, where neural pathways and processes are front and center in describing symptoms. According to Eisenberg,¹ psychiatry training in the 1980s was, for the most part, “brainless.” But it should not remain so, because neuroscience advances have skyrocketed since he made that provocative statement 3 decades ago. Yet, the psychiatric residency training curriculum in many programs is lagging behind the rapid evolution of psychiatry as a clinical neuroscience.²

To its credit, the Accreditation Council for Graduate Medical Education, which oversees and accredits residency training programs in all specialties, including psychiatry, recently announced that psychiatric residency training must emphasize neuroscience competence side-by-side with clinical competence. Psychiatric residents must increasingly incorporate neurobiology in their formulation of clinical care and determine how the selected pharmacologic therapy addresses the dysregulated neural circuitry underlying the clinical manifestation. A good example of this method is a recently published case of posttraumatic stress disorder (PTSD),³ which discussed the clinical components and treatment of this brain disorder through the prism of clinical neuroscience research data. PTSD “trauma” is not only psychological, but also neurobiological, and both must be incorporated in formulating a clinical case.

Another important step has emerged to focus on infusing neuroscience facts and concepts within the clinical training of psychiatric residents. The National Neuroscience Curriculum Initiative (www.nncionline.org) is a timely and welcome initiative that will aggressively promulgate a clinical neuro­scientification of psychiatric training, triggering a roadmap for modern, cutting-edge psychiatric practice.⁴ This will help consolidate psychiatry’s rightful place as a clinical neuroscience, without relinquishing its biopsycho­social roots.

As research continues to elucidate the neural mechanisms of key psychiatric symptoms, such as anxiety, depression, mania, impulsiveness, compulsions, delusions, or hallucinations, the trans­formation of psychiatry into an authentic clinical neuroscience is inevitable. But contemporary psychiatric practitioners must retool and start their journey toward neuroscience literacy by attending relevant continuing medical education presentations and regularly reading journals that focus on clinical psychiatric neuro­science, such as Molecular Psychiatry, JAMA Psychiatry, Biological Psychiatry, Neuropsychopharmacology, and Progress in Neuro-psychopharmacology and Biological Psychiatry. Current Psychiatry will do its part by establishing a section on psychiatric neuroscience for our readers. Dr. Stephen Stahl’s recent guest editorial in Current Psychiatry⁵ about Neuroscience-based Nomenclature is another important step toward the neuroscientification of psychiatric medications’ names, using the neural mechanism of action nomenclature instead of a single clinical indication (eg, antidepressant or antipsychotic) when there may be several approved uses for the medication.

It is my sincere hope that my fellow clinical psychiatrists will steadily grow their clinical neuroscience literacy and apply it to daily patient care. By formulating psychiatric signs and symptoms in evidence-based, neurobiological frameworks provided by clinical neuro­science advances, the myths, shibboleths, and misperceptions of what causes mental illness will dissipate and eventually vanish. And so will the stigma that shrouds our patients who suffer from brain disorders that manifest as disruptions of 1 or more complex functions of the human mind.

Dear Dr. Mossman,
I am retired, but an attorney friend of mine has asked me to help out by performing forensic evaluations. I’m tempted to try it because the work sounds meaningful and interesting. I won’t have a doctor–patient relationship with the attorney’s clients, and I expect the work will take <10 hours a week. Do I need malpractice coverage? Should I consider any other medicolegal issues before I start?

Submitted by “Dr. B”

One of the great things about being a psychiatrist is the variety of available practice options. Like Dr. B, many psychiatrists contemplate using their clinical know-how to perform forensic evaluations. For some psychiatrists, part-time work as an expert witness may provide an appealing change of pace from their other clinical duties¹ and a way to supplement their income.²

But as would be true for other kinds of medical practice, Dr. B is wise to consider the possible risks before jumping into forensic work. To help Dr. B decide about getting insurance coverage, we will:

• explain briefly the subspecialty of forensic psychiatry
• review the theory of malpractice and negligence torts
• discuss whether forensic evaluations can create doctor–patient relationships
• explore the availability and limitations of immunity for forensic work
• describe other types of liability with forensic work
• summarize steps to avoid liability.

Introduction to forensic psychiatry

Some psychiatrists—and many people who are not psychiatrists—have a vague or incorrect understanding of forensic psychiatry. Put succinctly, “Forensic Psychiatry is a subspecialty of psychiatry in which scientific and clinical expertise is applied in legal contexts….”³ To practice forensic psychiatry well, a psychiatrist must have some understanding of the law and how to apply and translate clinical concepts to fit legal criteria.⁴ Psychiatrists who offer to serve as expert witnesses should be familiar with how the courtroom functions, the nuances of how expert testimony is used, and possible sources of bias.^4,5

Forensic work can create role conflicts. For most types of forensic assessments, psychiatrists should not provide forensic opinions or testimony about their own patients.³ Even psychiatrists who only work as expert witnesses must balance duties of assisting the trier of fact, fulfilling the consultation role to the retaining party, upholding the standards and ethics of the profession, and striving to provide truthful, objective testimony.²

Special training usually is required

The most important qualification for being a good psychiatric expert witness is being a good psychiatrist, and courts do not require psychiatrists to have specialty training in forensic psychiatry to perform forensic psychiatric evaluations. Yet, the field of forensic psychiatry has developed over the past 50 years to the point that psychiatrists need special training to properly perform many, if not most, types of forensic evaluations.⁶ Much of forensic psychiatry involves writing specialized reports for lawyers and the court,⁷ and experts are supposed to meet professional standards, regardless of their training.^8-10 Psychiatrists who perform forensic work are obligated to claim expertise only in areas where their knowledge, skills, training, and experience justify such claims. These considerations explain why, since 1999, the American Board of Psychiatry and Neurology has limited eligibility for board certification in forensic psychiatry to psychiatrists who have completed accredited forensic fellowships.¹¹

Malpractice: A short review

To address Dr. B’s question about malpractice coverage, we first review what malpractice is.

“Tort” is a legal term for injury, and tort claims arise when one party harms another and the harmed party seeks money as compensation.⁹ In a tort claim alleging negligence, the plaintiff (ie, the person bringing the suit) asserts that the defendant had a legally recognized duty, that the defendant breached that duty, and that breach of duty harmed the plaintiff.⁸

Physicians have a legal duty to “possess the requisite knowledge and skill such as is possessed by the average member of the medical profession; … exercise ordinary and reasonable care in the application of such knowledge and skill; and … use best judgment in such application.”¹⁰ A medical malpractice lawsuit asserts that a doctor breached this duty and caused injury in the course of the medical practice.

Malpractice in forensic cases

Practicing medicine typically occurs within the context of treatment relationships. One might think, as Dr. B did, that because forensic evaluations do not involve treating patients, they do not create the kind of doctor–patient relationship that could lead to malpractice liability. This is incorrect, however, for several reasons.

Certain well-intended actions during a forensic evaluation, such as explaining the implications of a diagnosis, giving specific advice about a medication, or making a recommendation about where or how to obtain treatment, may create a doctor–patient relationship.^12,13 Many states’ laws on what constitutes the practice of medicine include performing examinations, diagnosing, or referring to oneself as “Dr.” or as a medical practitioner.^14-17 State courts have interpreted these laws to further define what constitutes medical practice and the creation of a doctor–patient relationship during a forensic examination.^18,19 Some legal scholars²⁰ and the American Medical Association (AMA)⁹ regard provision of expert testimony as practicing medicine because such testimony requires the application of medical science and rendering of diagnoses.

Immunity and shifts away from it

For many years, courts granted civil immunity to expert witnesses for several policy reasons.^8,9,13,20-22 Courts recognized that losing parties might want to blame whomever they could, and immunity could provide legal protection for expert witnesses. Without such protection, witnesses might feel more pressured to give testimony favorable to their side at the loss of objectivity,^23,24 or experts might be discouraged from testifying at all. This would be true especially for academic psychiatrists who testify infrequently or for retired doctors, such as Dr. B, who might not want to carry insurance for just one case.²¹ According to this argument, rather than using the threat of litigation to keep out improper testimony, courts should rely on both admissibility standards^25,26 and the adversarial nature of proceedings.²¹

Those who oppose granting immunity to experts argue that admissibility rules and cross-examination do too little to prevent bad testimony; the threat of liability, however, motivates experts to be more cautious and scientifically rigorous in their approach.²¹ Opponents also have argued that the threat of liability might reduce improper testimony, which they believe was partly responsible for rising malpractice premiums.²⁰

Courts vary in how they consider granting immunity and to what extent. For example:

• Some courts will not grant immunity to so-called “friendly experts,” while others have limited immunity for adversarial experts.^20-22
• Some courts have applied immunity to general fact witnesses but not to professional experts.^21,24,27
• When immunity is considered, it is usually regarding actual testimony. Yet, some courts have included pretrial services.^21,28-30
• Some courts have considered the testimonial issue at hand when deciding whether to extend immunity. For example, immunity may not apply if the issue is loss of profits^21,31 or if an experiment is conducted to demonstrate the extent of a physical injury.^21,32

If you plan to serve as an expert witness, find out what, if any, immunity is available in the jurisdiction where you expect to testify. If you do not have immunity, you may be subject to various malpractice claims, including alleged physical or emotional harm resulting from the evaluation¹ (perhaps caused by misuse of empathic statements³³), an accusation of negligent misdiagnosis of an evaluee,⁸ or failing to act upon a duty to warn or protect that arises during an assessment.³⁴

Other liability

Dr. B also asked about medicolegal issues other than malpractice. Although negligence is the claim that forensic psychiatrists most commonly encounter,¹⁰ other types of claims arise in practice-related legal actions. Potential causes of action include failure to obtain or attempt to obtain informed consent, breach of confidentiality, or not responding to a psychiatric emergency during evaluation. The plaintiff usually must show that the expert’s conduct was the cause-in-fact of injury.⁸

Besides civil lawsuits, forensic work may generate complaints to state medical boards.¹⁰ Occasionally, state medical boards have revoked psychiatrists’ licenses for improper testimony.²⁰ Aggrieved parties may allege violations of the Health Insurance Portability and Accountability Act of 1996, such as mishandling protected health information. Psychiatrists also may face sanction by professional societies—for example, censure by the American Psychiatric Association^9,10 or the AMA¹³ for ethics violations—if their improper testimony is considered unprofessional conduct. The theory behind this is that judges and jurors cannot be technical experts in every field, so the field must have a mechanism to police itself.^20,35,36 Finally, forensic experts can face criminal charges for perjury if they lie under oath.⁸

How to protect yourself

Even when legal claims against psychiatrists turn out to be baseless, legal costs of defending oneself can mount quickly. Knowing this, Dr. B may conclude that obtaining malpractice insurance would be wise. But a malpractice policy alone may not meet all Dr. B’s needs, because some policies do not cover ordinary negligence or other potential causes of legal action against a psychiatrist.¹³ Some companies offer these extra types of coverage for work as an expert witness at no additional cost, and some offer access to risk management services with specialized knowledge about forensic psychiatric practice.

Dear Dr. Mossman,
I am retired, but an attorney friend of mine has asked me to help out by performing forensic evaluations. I’m tempted to try it because the work sounds meaningful and interesting. I won’t have a doctor–patient relationship with the attorney’s clients, and I expect the work will take <10 hours a week. Do I need malpractice coverage? Should I consider any other medicolegal issues before I start?

Submitted by “Dr. B”

One of the great things about being a psychiatrist is the variety of available practice options. Like Dr. B, many psychiatrists contemplate using their clinical know-how to perform forensic evaluations. For some psychiatrists, part-time work as an expert witness may provide an appealing change of pace from their other clinical duties¹ and a way to supplement their income.²

But as would be true for other kinds of medical practice, Dr. B is wise to consider the possible risks before jumping into forensic work. To help Dr. B decide about getting insurance coverage, we will:

• explain briefly the subspecialty of forensic psychiatry
• review the theory of malpractice and negligence torts
• discuss whether forensic evaluations can create doctor–patient relationships
• explore the availability and limitations of immunity for forensic work
• describe other types of liability with forensic work
• summarize steps to avoid liability.

Introduction to forensic psychiatry

Some psychiatrists—and many people who are not psychiatrists—have a vague or incorrect understanding of forensic psychiatry. Put succinctly, “Forensic Psychiatry is a subspecialty of psychiatry in which scientific and clinical expertise is applied in legal contexts….”³ To practice forensic psychiatry well, a psychiatrist must have some understanding of the law and how to apply and translate clinical concepts to fit legal criteria.⁴ Psychiatrists who offer to serve as expert witnesses should be familiar with how the courtroom functions, the nuances of how expert testimony is used, and possible sources of bias.^4,5

Forensic work can create role conflicts. For most types of forensic assessments, psychiatrists should not provide forensic opinions or testimony about their own patients.³ Even psychiatrists who only work as expert witnesses must balance duties of assisting the trier of fact, fulfilling the consultation role to the retaining party, upholding the standards and ethics of the profession, and striving to provide truthful, objective testimony.²

Special training usually is required

The most important qualification for being a good psychiatric expert witness is being a good psychiatrist, and courts do not require psychiatrists to have specialty training in forensic psychiatry to perform forensic psychiatric evaluations. Yet, the field of forensic psychiatry has developed over the past 50 years to the point that psychiatrists need special training to properly perform many, if not most, types of forensic evaluations.⁶ Much of forensic psychiatry involves writing specialized reports for lawyers and the court,⁷ and experts are supposed to meet professional standards, regardless of their training.^8-10 Psychiatrists who perform forensic work are obligated to claim expertise only in areas where their knowledge, skills, training, and experience justify such claims. These considerations explain why, since 1999, the American Board of Psychiatry and Neurology has limited eligibility for board certification in forensic psychiatry to psychiatrists who have completed accredited forensic fellowships.¹¹

Malpractice: A short review

To address Dr. B’s question about malpractice coverage, we first review what malpractice is.

“Tort” is a legal term for injury, and tort claims arise when one party harms another and the harmed party seeks money as compensation.⁹ In a tort claim alleging negligence, the plaintiff (ie, the person bringing the suit) asserts that the defendant had a legally recognized duty, that the defendant breached that duty, and that breach of duty harmed the plaintiff.⁸

Physicians have a legal duty to “possess the requisite knowledge and skill such as is possessed by the average member of the medical profession; … exercise ordinary and reasonable care in the application of such knowledge and skill; and … use best judgment in such application.”¹⁰ A medical malpractice lawsuit asserts that a doctor breached this duty and caused injury in the course of the medical practice.

Malpractice in forensic cases

Practicing medicine typically occurs within the context of treatment relationships. One might think, as Dr. B did, that because forensic evaluations do not involve treating patients, they do not create the kind of doctor–patient relationship that could lead to malpractice liability. This is incorrect, however, for several reasons.

Certain well-intended actions during a forensic evaluation, such as explaining the implications of a diagnosis, giving specific advice about a medication, or making a recommendation about where or how to obtain treatment, may create a doctor–patient relationship.^12,13 Many states’ laws on what constitutes the practice of medicine include performing examinations, diagnosing, or referring to oneself as “Dr.” or as a medical practitioner.^14-17 State courts have interpreted these laws to further define what constitutes medical practice and the creation of a doctor–patient relationship during a forensic examination.^18,19 Some legal scholars²⁰ and the American Medical Association (AMA)⁹ regard provision of expert testimony as practicing medicine because such testimony requires the application of medical science and rendering of diagnoses.

Immunity and shifts away from it

For many years, courts granted civil immunity to expert witnesses for several policy reasons.^8,9,13,20-22 Courts recognized that losing parties might want to blame whomever they could, and immunity could provide legal protection for expert witnesses. Without such protection, witnesses might feel more pressured to give testimony favorable to their side at the loss of objectivity,^23,24 or experts might be discouraged from testifying at all. This would be true especially for academic psychiatrists who testify infrequently or for retired doctors, such as Dr. B, who might not want to carry insurance for just one case.²¹ According to this argument, rather than using the threat of litigation to keep out improper testimony, courts should rely on both admissibility standards^25,26 and the adversarial nature of proceedings.²¹

Those who oppose granting immunity to experts argue that admissibility rules and cross-examination do too little to prevent bad testimony; the threat of liability, however, motivates experts to be more cautious and scientifically rigorous in their approach.²¹ Opponents also have argued that the threat of liability might reduce improper testimony, which they believe was partly responsible for rising malpractice premiums.²⁰

Courts vary in how they consider granting immunity and to what extent. For example:

• Some courts will not grant immunity to so-called “friendly experts,” while others have limited immunity for adversarial experts.^20-22
• Some courts have applied immunity to general fact witnesses but not to professional experts.^21,24,27
• When immunity is considered, it is usually regarding actual testimony. Yet, some courts have included pretrial services.^21,28-30
• Some courts have considered the testimonial issue at hand when deciding whether to extend immunity. For example, immunity may not apply if the issue is loss of profits^21,31 or if an experiment is conducted to demonstrate the extent of a physical injury.^21,32

If you plan to serve as an expert witness, find out what, if any, immunity is available in the jurisdiction where you expect to testify. If you do not have immunity, you may be subject to various malpractice claims, including alleged physical or emotional harm resulting from the evaluation¹ (perhaps caused by misuse of empathic statements³³), an accusation of negligent misdiagnosis of an evaluee,⁸ or failing to act upon a duty to warn or protect that arises during an assessment.³⁴

Other liability

Dr. B also asked about medicolegal issues other than malpractice. Although negligence is the claim that forensic psychiatrists most commonly encounter,¹⁰ other types of claims arise in practice-related legal actions. Potential causes of action include failure to obtain or attempt to obtain informed consent, breach of confidentiality, or not responding to a psychiatric emergency during evaluation. The plaintiff usually must show that the expert’s conduct was the cause-in-fact of injury.⁸

Besides civil lawsuits, forensic work may generate complaints to state medical boards.¹⁰ Occasionally, state medical boards have revoked psychiatrists’ licenses for improper testimony.²⁰ Aggrieved parties may allege violations of the Health Insurance Portability and Accountability Act of 1996, such as mishandling protected health information. Psychiatrists also may face sanction by professional societies—for example, censure by the American Psychiatric Association^9,10 or the AMA¹³ for ethics violations—if their improper testimony is considered unprofessional conduct. The theory behind this is that judges and jurors cannot be technical experts in every field, so the field must have a mechanism to police itself.^20,35,36 Finally, forensic experts can face criminal charges for perjury if they lie under oath.⁸

How to protect yourself

Even when legal claims against psychiatrists turn out to be baseless, legal costs of defending oneself can mount quickly. Knowing this, Dr. B may conclude that obtaining malpractice insurance would be wise. But a malpractice policy alone may not meet all Dr. B’s needs, because some policies do not cover ordinary negligence or other potential causes of legal action against a psychiatrist.¹³ Some companies offer these extra types of coverage for work as an expert witness at no additional cost, and some offer access to risk management services with specialized knowledge about forensic psychiatric practice.

Dear Dr. Mossman,
I am retired, but an attorney friend of mine has asked me to help out by performing forensic evaluations. I’m tempted to try it because the work sounds meaningful and interesting. I won’t have a doctor–patient relationship with the attorney’s clients, and I expect the work will take <10 hours a week. Do I need malpractice coverage? Should I consider any other medicolegal issues before I start?

Submitted by “Dr. B”

One of the great things about being a psychiatrist is the variety of available practice options. Like Dr. B, many psychiatrists contemplate using their clinical know-how to perform forensic evaluations. For some psychiatrists, part-time work as an expert witness may provide an appealing change of pace from their other clinical duties¹ and a way to supplement their income.²

But as would be true for other kinds of medical practice, Dr. B is wise to consider the possible risks before jumping into forensic work. To help Dr. B decide about getting insurance coverage, we will:

• explain briefly the subspecialty of forensic psychiatry
• review the theory of malpractice and negligence torts
• discuss whether forensic evaluations can create doctor–patient relationships
• explore the availability and limitations of immunity for forensic work
• describe other types of liability with forensic work
• summarize steps to avoid liability.

Introduction to forensic psychiatry

Some psychiatrists—and many people who are not psychiatrists—have a vague or incorrect understanding of forensic psychiatry. Put succinctly, “Forensic Psychiatry is a subspecialty of psychiatry in which scientific and clinical expertise is applied in legal contexts….”³ To practice forensic psychiatry well, a psychiatrist must have some understanding of the law and how to apply and translate clinical concepts to fit legal criteria.⁴ Psychiatrists who offer to serve as expert witnesses should be familiar with how the courtroom functions, the nuances of how expert testimony is used, and possible sources of bias.^4,5

Forensic work can create role conflicts. For most types of forensic assessments, psychiatrists should not provide forensic opinions or testimony about their own patients.³ Even psychiatrists who only work as expert witnesses must balance duties of assisting the trier of fact, fulfilling the consultation role to the retaining party, upholding the standards and ethics of the profession, and striving to provide truthful, objective testimony.²

Special training usually is required

The most important qualification for being a good psychiatric expert witness is being a good psychiatrist, and courts do not require psychiatrists to have specialty training in forensic psychiatry to perform forensic psychiatric evaluations. Yet, the field of forensic psychiatry has developed over the past 50 years to the point that psychiatrists need special training to properly perform many, if not most, types of forensic evaluations.⁶ Much of forensic psychiatry involves writing specialized reports for lawyers and the court,⁷ and experts are supposed to meet professional standards, regardless of their training.^8-10 Psychiatrists who perform forensic work are obligated to claim expertise only in areas where their knowledge, skills, training, and experience justify such claims. These considerations explain why, since 1999, the American Board of Psychiatry and Neurology has limited eligibility for board certification in forensic psychiatry to psychiatrists who have completed accredited forensic fellowships.¹¹

Malpractice: A short review

To address Dr. B’s question about malpractice coverage, we first review what malpractice is.

“Tort” is a legal term for injury, and tort claims arise when one party harms another and the harmed party seeks money as compensation.⁹ In a tort claim alleging negligence, the plaintiff (ie, the person bringing the suit) asserts that the defendant had a legally recognized duty, that the defendant breached that duty, and that breach of duty harmed the plaintiff.⁸

Physicians have a legal duty to “possess the requisite knowledge and skill such as is possessed by the average member of the medical profession; … exercise ordinary and reasonable care in the application of such knowledge and skill; and … use best judgment in such application.”¹⁰ A medical malpractice lawsuit asserts that a doctor breached this duty and caused injury in the course of the medical practice.

Malpractice in forensic cases

Practicing medicine typically occurs within the context of treatment relationships. One might think, as Dr. B did, that because forensic evaluations do not involve treating patients, they do not create the kind of doctor–patient relationship that could lead to malpractice liability. This is incorrect, however, for several reasons.

Certain well-intended actions during a forensic evaluation, such as explaining the implications of a diagnosis, giving specific advice about a medication, or making a recommendation about where or how to obtain treatment, may create a doctor–patient relationship.^12,13 Many states’ laws on what constitutes the practice of medicine include performing examinations, diagnosing, or referring to oneself as “Dr.” or as a medical practitioner.^14-17 State courts have interpreted these laws to further define what constitutes medical practice and the creation of a doctor–patient relationship during a forensic examination.^18,19 Some legal scholars²⁰ and the American Medical Association (AMA)⁹ regard provision of expert testimony as practicing medicine because such testimony requires the application of medical science and rendering of diagnoses.

Immunity and shifts away from it

For many years, courts granted civil immunity to expert witnesses for several policy reasons.^8,9,13,20-22 Courts recognized that losing parties might want to blame whomever they could, and immunity could provide legal protection for expert witnesses. Without such protection, witnesses might feel more pressured to give testimony favorable to their side at the loss of objectivity,^23,24 or experts might be discouraged from testifying at all. This would be true especially for academic psychiatrists who testify infrequently or for retired doctors, such as Dr. B, who might not want to carry insurance for just one case.²¹ According to this argument, rather than using the threat of litigation to keep out improper testimony, courts should rely on both admissibility standards^25,26 and the adversarial nature of proceedings.²¹

Those who oppose granting immunity to experts argue that admissibility rules and cross-examination do too little to prevent bad testimony; the threat of liability, however, motivates experts to be more cautious and scientifically rigorous in their approach.²¹ Opponents also have argued that the threat of liability might reduce improper testimony, which they believe was partly responsible for rising malpractice premiums.²⁰

Courts vary in how they consider granting immunity and to what extent. For example:

• Some courts will not grant immunity to so-called “friendly experts,” while others have limited immunity for adversarial experts.^20-22
• Some courts have applied immunity to general fact witnesses but not to professional experts.^21,24,27
• When immunity is considered, it is usually regarding actual testimony. Yet, some courts have included pretrial services.^21,28-30
• Some courts have considered the testimonial issue at hand when deciding whether to extend immunity. For example, immunity may not apply if the issue is loss of profits^21,31 or if an experiment is conducted to demonstrate the extent of a physical injury.^21,32

If you plan to serve as an expert witness, find out what, if any, immunity is available in the jurisdiction where you expect to testify. If you do not have immunity, you may be subject to various malpractice claims, including alleged physical or emotional harm resulting from the evaluation¹ (perhaps caused by misuse of empathic statements³³), an accusation of negligent misdiagnosis of an evaluee,⁸ or failing to act upon a duty to warn or protect that arises during an assessment.³⁴

Other liability

Dr. B also asked about medicolegal issues other than malpractice. Although negligence is the claim that forensic psychiatrists most commonly encounter,¹⁰ other types of claims arise in practice-related legal actions. Potential causes of action include failure to obtain or attempt to obtain informed consent, breach of confidentiality, or not responding to a psychiatric emergency during evaluation. The plaintiff usually must show that the expert’s conduct was the cause-in-fact of injury.⁸

Besides civil lawsuits, forensic work may generate complaints to state medical boards.¹⁰ Occasionally, state medical boards have revoked psychiatrists’ licenses for improper testimony.²⁰ Aggrieved parties may allege violations of the Health Insurance Portability and Accountability Act of 1996, such as mishandling protected health information. Psychiatrists also may face sanction by professional societies—for example, censure by the American Psychiatric Association^9,10 or the AMA¹³ for ethics violations—if their improper testimony is considered unprofessional conduct. The theory behind this is that judges and jurors cannot be technical experts in every field, so the field must have a mechanism to police itself.^20,35,36 Finally, forensic experts can face criminal charges for perjury if they lie under oath.⁸

How to protect yourself

Even when legal claims against psychiatrists turn out to be baseless, legal costs of defending oneself can mount quickly. Knowing this, Dr. B may conclude that obtaining malpractice insurance would be wise. But a malpractice policy alone may not meet all Dr. B’s needs, because some policies do not cover ordinary negligence or other potential causes of legal action against a psychiatrist.¹³ Some companies offer these extra types of coverage for work as an expert witness at no additional cost, and some offer access to risk management services with specialized knowledge about forensic psychiatric practice.