Posttraumatic Stress Disorder and Risk for CHD
Posttraumatic Stress Disorder and Risk for CHD
A comprehensive search resulted in 1,194 unique articles, of which 23 qualified for full-text review (Figure 1). Six articles met the final inclusion criteria. There was one disagreement between the reviewers at the full-text stage, which was resolved by discussion and consensus. A second disagreement occurred at the data extraction stage because one study published a tightly controlled estimate of the association of PTSD with incident CHD in a sample of twins from a larger registry, which could yield an estimate of the association of PTSD with cardiovascular mortality in a larger sample. We decided to include the more precise estimate from the smaller sample, with the recognition that including the estimate from the larger sample may have increased our overall meta-analytic estimate but would have represented a less rigorous test of our primary study aim.
(Enlarge Image)
Figure 1.
Search strategy flowchart.
Table gives detailed information about study characteristics, which included 402,274 participants cumulatively. Participants were enrolled between 1982 and 2000 and were followed up for a mean of 2.9 to 15 years. Most participants were military veterans, but many participants were also drawn from the World Trade Center registry of those exposed to the attacks on 9/11 and from a representative sample of women from East Baltimore exposed to various traumas. Accordingly, there were diverse types of traumatic experiences represented in the meta-analysis.
Posttraumatic stress disorder was measured by self-report questionnaire, diagnostic interview, or International Classification of Diseases, Ninth Revision (ICD-9) diagnostic codes. Articles considered a broad range of covariates, from known cardiac risk factors (eg, systolic blood pressure, hyperlipidemia, and smoking) to intelligence scores. Five of the 6 studies included depression as a covariate. Incident CHD and cardiac-specific mortality were assessed by self-report of physician diagnosis, medical records, and/or death certificate searches.
There was an indication of heterogeneity in estimates (Q5 = 10.17; P = .07; I = 50.83), with HRs ranging from 1.39 to 3.28. Figure 2 shows the HRs for incident CHD and cardiac-specific mortality associated with PTSD for each of the 6 studies (N = 402,274) with no adjustment for depression. The random-effects model yielded an aggregated HR of 1.55 (95% CI 1.34–1.80).
(Enlarge Image)
Figure 2.
Forest plot of the association of PTSD with incident CHD. Note: the area of each square is proportional to the study's weight in the meta-analysis, and each line represents the CI around the estimate. The diamond represents the aggregate estimate, and its lateral points indicate CIs for this estimate.
Although difficult to assess with only 5 studies, there was visual but no statistical evidence of publication bias (Figure 3) because the fail-safe N (number of studies with null findings that would be needed to conclude that the true HR was 1) was 210. Orwin's fail-safe N suggested, however, that 48 studies with null findings would yield a trivial HR of less than 1.04.
(Enlarge Image)
Figure 3.
Funnel plot to assess publication bias across studies.
There was also a significant heterogeneity in the depression-adjusted estimates (Q5 = 19.09, P = .002; I = 73.81), thereby supporting the use of a random-effects model to pool data. The aggregate HR for the 5 depression-adjusted estimates (N = 362,950) in a random-effects model was 1.27 (95% CI 1.08–1.49). All studies that adjusted for depression did so after adjustment for clinical and behavioral CHD risk factors. Depression-adjusted estimates reported in each study and the method used for assessing depression in each are given in Table.
Possible Moderators. Although there were not enough studies to test potential moderators of the association of PTSD with incident CHD, we conducted a number of descriptive analyses for likely moderators.
Sex. There appear to be sex differences in risk for PTSD, and there may be a stronger association of PTSD with incident CHD in women. Without adjustment for depression (only 1 study of women was adjusted for depression), the aggregate HR for the association of PTSD with incident CHD in men was 1.62 (95% CI 1.39–1.89), whereas the HR for women was 1.94 (95% CI 1.13–3.33). Although underpowered to detect a significant difference, a mixed-effects analysis suggested that there may be a sex difference in the association of PTSD with incident CHD (Q2 = 4.74; P = .09).
Race/Ethnicity. Some studies suggest that minority race/ethnicity status is associated with increased risk for PTSD, and there are established racial/ethnic differences in CHD risk. We conducted an exploratory meta-regression analysis to determine the association between the racial composition of each study sample and its estimate of the association between PTSD and CHD by coding each study for the proportion of its participants who identified as non-Hispanic white. Studies ranged from 52% to 100% non-Hispanic white, but racial composition explained less than 1% of the variability in effect size estimates (P = .89).
Type of PTSD Assessment. Meta-analyses of the rate of PTSD caused by some types of trauma demonstrate that the method for determining a PTSD "case" (ie, positive screen on a self-report questionnaire vs diagnostic interview) influences the estimate of PTSD prevalence. In the included studies, the HR in those that assessed PTSD by clinical interview was 1.81 (95% CI 1.13–2.91), whereas the estimate for those that used screening questionnaires was 1.62 (95% CI 1.42–1.84). Although underpowered to detect a significant difference, a mixed-effects analysis suggested no difference in the association of PTSD with incident CHD by type of PTSD assessment (Q1 = 0.21; P = .65).
Date of PTSD Assessment. Because definitions of both PTSD and CHD have changed over the last 30 years, we tested whether the date of PTSD assessment was associated with studies' effect size estimates using meta-regression. Date of PTSD assessment accounted for only 1% of the variability in HR estimates (P = .50).
Results
Literature Search
A comprehensive search resulted in 1,194 unique articles, of which 23 qualified for full-text review (Figure 1). Six articles met the final inclusion criteria. There was one disagreement between the reviewers at the full-text stage, which was resolved by discussion and consensus. A second disagreement occurred at the data extraction stage because one study published a tightly controlled estimate of the association of PTSD with incident CHD in a sample of twins from a larger registry, which could yield an estimate of the association of PTSD with cardiovascular mortality in a larger sample. We decided to include the more precise estimate from the smaller sample, with the recognition that including the estimate from the larger sample may have increased our overall meta-analytic estimate but would have represented a less rigorous test of our primary study aim.
(Enlarge Image)
Figure 1.
Search strategy flowchart.
Study Characteristics
Table gives detailed information about study characteristics, which included 402,274 participants cumulatively. Participants were enrolled between 1982 and 2000 and were followed up for a mean of 2.9 to 15 years. Most participants were military veterans, but many participants were also drawn from the World Trade Center registry of those exposed to the attacks on 9/11 and from a representative sample of women from East Baltimore exposed to various traumas. Accordingly, there were diverse types of traumatic experiences represented in the meta-analysis.
Posttraumatic stress disorder was measured by self-report questionnaire, diagnostic interview, or International Classification of Diseases, Ninth Revision (ICD-9) diagnostic codes. Articles considered a broad range of covariates, from known cardiac risk factors (eg, systolic blood pressure, hyperlipidemia, and smoking) to intelligence scores. Five of the 6 studies included depression as a covariate. Incident CHD and cardiac-specific mortality were assessed by self-report of physician diagnosis, medical records, and/or death certificate searches.
Association of PTSD With Incident CHD and Mortality
There was an indication of heterogeneity in estimates (Q5 = 10.17; P = .07; I = 50.83), with HRs ranging from 1.39 to 3.28. Figure 2 shows the HRs for incident CHD and cardiac-specific mortality associated with PTSD for each of the 6 studies (N = 402,274) with no adjustment for depression. The random-effects model yielded an aggregated HR of 1.55 (95% CI 1.34–1.80).
(Enlarge Image)
Figure 2.
Forest plot of the association of PTSD with incident CHD. Note: the area of each square is proportional to the study's weight in the meta-analysis, and each line represents the CI around the estimate. The diamond represents the aggregate estimate, and its lateral points indicate CIs for this estimate.
Although difficult to assess with only 5 studies, there was visual but no statistical evidence of publication bias (Figure 3) because the fail-safe N (number of studies with null findings that would be needed to conclude that the true HR was 1) was 210. Orwin's fail-safe N suggested, however, that 48 studies with null findings would yield a trivial HR of less than 1.04.
(Enlarge Image)
Figure 3.
Funnel plot to assess publication bias across studies.
Adjustment for Depression
There was also a significant heterogeneity in the depression-adjusted estimates (Q5 = 19.09, P = .002; I = 73.81), thereby supporting the use of a random-effects model to pool data. The aggregate HR for the 5 depression-adjusted estimates (N = 362,950) in a random-effects model was 1.27 (95% CI 1.08–1.49). All studies that adjusted for depression did so after adjustment for clinical and behavioral CHD risk factors. Depression-adjusted estimates reported in each study and the method used for assessing depression in each are given in Table.
Descriptive Information
Possible Moderators. Although there were not enough studies to test potential moderators of the association of PTSD with incident CHD, we conducted a number of descriptive analyses for likely moderators.
Sex. There appear to be sex differences in risk for PTSD, and there may be a stronger association of PTSD with incident CHD in women. Without adjustment for depression (only 1 study of women was adjusted for depression), the aggregate HR for the association of PTSD with incident CHD in men was 1.62 (95% CI 1.39–1.89), whereas the HR for women was 1.94 (95% CI 1.13–3.33). Although underpowered to detect a significant difference, a mixed-effects analysis suggested that there may be a sex difference in the association of PTSD with incident CHD (Q2 = 4.74; P = .09).
Race/Ethnicity. Some studies suggest that minority race/ethnicity status is associated with increased risk for PTSD, and there are established racial/ethnic differences in CHD risk. We conducted an exploratory meta-regression analysis to determine the association between the racial composition of each study sample and its estimate of the association between PTSD and CHD by coding each study for the proportion of its participants who identified as non-Hispanic white. Studies ranged from 52% to 100% non-Hispanic white, but racial composition explained less than 1% of the variability in effect size estimates (P = .89).
Type of PTSD Assessment. Meta-analyses of the rate of PTSD caused by some types of trauma demonstrate that the method for determining a PTSD "case" (ie, positive screen on a self-report questionnaire vs diagnostic interview) influences the estimate of PTSD prevalence. In the included studies, the HR in those that assessed PTSD by clinical interview was 1.81 (95% CI 1.13–2.91), whereas the estimate for those that used screening questionnaires was 1.62 (95% CI 1.42–1.84). Although underpowered to detect a significant difference, a mixed-effects analysis suggested no difference in the association of PTSD with incident CHD by type of PTSD assessment (Q1 = 0.21; P = .65).
Date of PTSD Assessment. Because definitions of both PTSD and CHD have changed over the last 30 years, we tested whether the date of PTSD assessment was associated with studies' effect size estimates using meta-regression. Date of PTSD assessment accounted for only 1% of the variability in HR estimates (P = .50).
Source...