FY indicates fiscal year; PCL, PTSD Checklist; PTSD, posttraumatic stress disorder.
eTable 1. Variable Definitions
eTable 2. Type 2 Diabetes Outcome, Cumulative Incidence %, and Incidence Rate Per 1000 Person-Years (N = 1598)
eFigure 1. Retrospective Cohort Design
eFigure 2. Unweighted and Weighted SMD% for Covariates for Patients With <20 Point PCL Decrease vs ≥20 Point PCL Decrease
Customize your JAMA Network experience by selecting one or more topics from the list below.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Scherrer JF, Salas J, Norman SB, et al. Association Between Clinically Meaningful Posttraumatic Stress Disorder Improvement and Risk of Type 2 Diabetes. JAMA Psychiatry. Published online August 21, 2019. doi:10.1001/jamapsychiatry.2019.2096
Is clinically meaningful posttraumatic stress disorder symptom decrease (≥20-point decrease on the Posttraumatic Stress Disorder Checklist score) associated with a lower risk of incident type 2 diabetes compared with less than a clinically meaningful or no improvement?
In this cohort study of medical records from 1598 patients, clinically meaningful posttraumatic stress disorder improvement compared with less than clinically meaningful or no improvement was associated with a 49% lower risk of incident type 2 diabetes.
Long-term chronic health conditions associated with posttraumatic stress disorder may be less likely to occur among patients who experience clinically meaningful symptom reduction through treatment or spontaneous improvement.
Posttraumatic stress disorder (PTSD) is associated with increased risk of type 2 diabetes (T2D). Improvement in PTSD has been associated with improved self-reported physical health and hypertension; however, there is no literature, to our knowledge, on whether PTSD improvement is associated with T2D risk.
To examine whether clinically meaningful PTSD symptom reduction is associated with lower risk of T2D.
Design, Setting, and Participants
This retrospective cohort study examined Veterans Health Affairs medical record data from 5916 patients who received PTSD specialty care between fiscal years 2008 and 2012 and were followed up through fiscal year 2015. Eligible patients had 1 or more PTSD Checklist (PCL) scores of 50 or higher between fiscal years 2008 and 2012 and a second PCL score within the following 12 months and at least 8 weeks after the first PCL score of 50 or higher. The index date was 12 months after the first PCL score. Patients were free of T2D diagnosis or an antidiabetic medication use for 12 months before the index date and had at least 1 visit after the index date. Data analyses were completed during January 2019.
Reduction in PCL scores during a 12-month period was used to define patients as those with a clinically meaningful improvement (≥20-point PCL score decrease) and patients with less or no improvement (<20-point PCL score decrease).
Main Outcomes and Measures
Incident T2D diagnosed during a 2- to 6-year follow-up.
Medical records from a total of 1598 patients (mean [SD] age, 42.1 [13.4] years; 1347 [84.3%] male; 1060 [66.3%] white) were studied. The age-adjusted cumulative incidence of T2D was 2.6% among patients with a clinically meaningful PCL score decrease and 5.9% among patients without a clinically meaningful PCL score decrease (P = .003). After control for confounding, patients with a clinically meaningful PCL score decrease were significantly less likely to develop T2DM compared with those without a clinically meaningful decrease (hazard ratio, 0.51; 95% CI, 0.26-0.98).
Conclusions and Relevance
The findings suggest that clinically meaningful reductions in PTSD symptoms are associated with a lower risk of T2D. A decrease in PCL score, whether through treatment or spontaneous improvement, may help mitigate the greater risk of T2D in patients with PTSD.
Posttraumatic stress disorder (PTSD) is a chronic condition that affects up to 12% of civilians and up to 30% of the veteran population and is associated with increased risks for multicomorbidity.1-4 Completion of evidence-based psychotherapy is associated with clinically meaningful reductions in PTSD symptoms5,6 and can result in improvement in psychiatric comorbidities and perceived health.7-9 Improvement in PTSD is associated with parallel improvements in depression and general emotional well-being,7,8,10,11 sleep,12,13 blood pressure,9,14 general physical concerns (eg, back pain, headache, and cough), and perceived health.15
Posttraumatic stress disorder is associated with increased risk of type 2 diabetes (T2D),3,15-20 and this association may be partly explained by the high prevalence of obesity, glucose dysregulation, inflammation, the metabolic syndrome, depression, and other T2D risk factors among those with vs without PTSD.20-22 To our knowledge, no studies have examined whether a clinically meaningful reduction in PTSD symptoms is associated with lower risk of incident T2D.
We hypothesized that clinically meaningful PTSD symptom reduction would be associated with a decreased risk of incident T2D during a 2- to 6-year follow-up period. In addition, to explore potential mechanisms for the association between clinically meaningful PTSD symptom reduction and risk of T2D, we evaluated change in body mass index (BMI) (calculated as weight in kilograms divided by height in meters squared), hemoglobin A1c values, and depression symptoms in patients with and without clinically meaningful PTSD symptom reduction.
This retrospective cohort study used medical record data from Veterans Health Affairs (VHA) patients who used a PTSD specialty clinic at 1 of 5 VHA medical centers across the United States between fiscal years (FYs) 2008 and 2012. Follow-up continued until FY2015. Medical record data included diagnostic codes, type of clinic encounter (eg, primary care, PTSD psychotherapy, and physical therapy), medications, laboratory results, vital signs, and demographic measures. The study was approved by the institutional review boards of Saint Louis University and the Harry S. Truman Veterans Administration Medical Center with a waiver of informed consent because data were deidentified.
From patients 18 to 70 years of age who had 2 or more visits to PTSD specialty care, we randomly selected 5916 patients with PTSD. Eligible patients had a PTSD Checklist (PCL) score of 50 or higher (ie, above the threshold for probable PTSD)23,24 in FY2008 to FY2012. We required patients to have at least 1 PCL score within the following 12 months of their first PCL score of 50 or higher. The last PCL score in this 12-month period was required to be at least 8 weeks after the first PCL score of 50 or higher. The index date was 12 months after the first PCL score of 50 or higher. Index dates could occur from FY2009 to FY2013. Therefore, follow-up time ranged from 2 to 6 years. All patients had to be free of a T2D diagnosis or not receiving an antidiabetic medication for 12 months before the index date and have had at least 1 visit after the index date. After eligibility criteria were applied, 1598 patients with PTSD and free of diabetes at the index date were available for analysis. The sample selection is shown in the Figure, and the retrospective cohort design is shown in eFigure 1 in the Supplement.
Posttraumatic stress disorder was defined as 2 or more outpatient visits within a 12-month period or 1 inpatient stay with an International Classification of Diseases, Ninth Revision (ICD-9) code of 309.81. Requiring 2 or more PTSD diagnoses has good positive predictive value (82%) compared with a criterion standard PCL score of 50 or higher25 and has good agreement (79.4%) with lifetime diagnosis according to the Structured Clinical Interview for DSM-IV.26
Incident T2D was defined as the first occurrence of ICD-9 codes 250.x0, 250.x2, 357.2, 362.0x, or 366.41 or a fill for an antidiabetic medication. Follow-up time was measured as days since the index date to T2D or censor date, which was the last clinical encounter in FY2008 to FY2015.
The PCL scores were obtained from medical record abstraction and administrative data. Because VHA administrative medical record data do not capture all PCL scores (eg, those stored in physician notes), medical record abstraction was conducted by trained medical record abstractors from Abt Associates (https://www.abtassociates.com/). During 6 months, Abt Associates abstracted 22 287 valid PCL scores from eligible patients with encounters between FY2008 and FY2015. After administrative data were added and duplicate scores removed, 26 631 valid PCL values for 4441 patients with PTSD were available from encounters between FY2008 and FY2015.
The difference between the last PCL score in the exposure year and first PCL score of 50 or higher (PCL score range, 17-85) was used to classify patients as having a clinically meaningful decrease in PTSD symptoms, defined as a 20-point or greater decrease vs less than or no PTSD symptom decrease defined as a less than 20-point decrease. Research suggests that 10 points reflect clinically meaningful improvement.24 To increase the ability to detect an association with PTSD improvement, we used 20 points to define meaningful PCL score reduction.
Detailed variable definitions are given in eTable 1 in the Supplement. Sociodemographic variables were measured at the start of the observation period. Some patients had missing data for race/ethnicity; therefore, we created a missing category to retain all cases in analyses. Other covariates were measured up to the index date. These covariates included depression, anxiety disorders, alcohol and drug abuse or dependence, smoking, sleep disorder, adequate acute-phase antidepressant treatment, receipt of any atypical antipsychotic medication, hypertension, hyperlipidemia, and obesity. We selected covariates that are correlated with PCL score change (eg, depression) and medications and comorbid metabolic conditions that are correlated with PTSD and T2D. Receipt of minimally adequate PTSD psychotherapy (≥9 visits in 15 weeks) was measured from the first PCL score of 50 or higher to the index date. We controlled for PTSD psychotherapy to reduce the risk of confounding treatment seeking, adherence, and treatment response with health prevention behaviors that would reduce the risk of T2D. Our goal was to measure the total association between PCL score decrease (whether attributable to treatment or spontaneous remission) and risk of T2D and not to specifically assess the effect of treatment response.
We used propensity scores and inverse probability of exposure weighting (IPEW) to balance the distribution of potential confounders between patients who did and did not have a clinically meaningful PCL score decrease. The propensity score is a binary logistic regression model that was used to measure the probability of a clinically meaningful PCL decrease vs less than a clinically meaningful decrease given covariates.27 The propensity score was used to compute a stabilized weight,28,29 which is the marginal probability of experiencing a clinically meaningful PCL score decrease divided by the propensity score for having a clinically meaningful PCL decrease for those with a decrease of 20 points or higher and for those with a less than 20-point decrease. Well-behaved weights have a mean close to 1 and a maximum value less than 10; therefore, weights of 10 or higher were trimmed.29-31 A pseudo-population was created after applying IPEW, and confounding was controlled when variables balanced between patients with and without a clinically meaningful PCL decrease. Balance is evidenced by a standardized mean difference less than 10%.29
Using a subsample of patients with available data, we computed exploratory analyses to evaluate whether decreased T2D risk after clinically meaningful PTSD improvement was associated with concurrent decreases in BMI (n = 1405), Patient Health Questionnaire 9 (PHQ-9) scores (n = 324), or hemoglobin A1c values (n = 585). We assessed changes in hemoglobin A1c and BMI values from the first available value in the exposure year to the first available value that occurred at least 12 months later. For PHQ-9, we compared changes from the first PHQ-9 score in the exposure year with the last in the exposure year that occurred at least 8 weeks after the first PHQ-9 score. A repeated-measures difference in differences analysis for each measure was conducted using mixed-effects regression models. The difference in differences analysis provided an estimation of the mean change (ie, slope) in each measure in the PCL groups and assesses whether the 2 slopes are significantly different.
To determine whether an unmeasured confounder could completely explain our results, we computed the E-value.32 The E-value is the minimum magnitude of association that an unmeasured confounder would have with both the exposure and the outcome to completely explain the association between clinically meaningful PTSD improvement and incident T2D.32
Descriptive analysis used χ2 tests for categorical variables and 2-tailed, independent-sample t tests for continuous variables to estimate the association between potential confounders and PCL score decrease. For unweighted data, we computed Poisson regression models to calculate T2D incidence rates per 1000 person-years. Bivariate Cox proportional hazards regression models were used to measure the association of each covariate with incident T2D.
For unweighted and weighted data, we used Cox proportional hazards regression models to test the association between clinically meaningful PCL score decrease and incident T2D. Weighted Cox proportional hazards regression models controlled for all measured confounders (Table 1 and eFigure 2 in the Supplement). An expanded Cox proportional hazards regression model included adjustment for hypertension, obesity, and hyperlipidemia that could be present after the index date. Results were expressed by hazard ratios (HRs) and 95% CIs. Robust, sandwich-type variance estimators were used to calculate 95% CIs and P values for weighted data. For all models, the proportional hazards assumption was tested and met. SAS statistical software, version 9.4 (SAS Institute Inc) was used for all analyses, and α = .05 indicated statistical significance. Data analyses were completed during January 2019.
Medical records from a total of 1598 patients (mean [SD] age, 42.1 [13.4] years; 1347 [84.3%] male; 1060 [66.3%] white) were studied (Table 1). Older age was associated with a clinically meaningful PCL score decrease (mean [SD], 43.6 [13.5] years in the group with PCL score decrease ≥20 vs 41.7 [13.4] years in the group with PCL score decrease <20, P = .02). Other demographic variables were not associated with PCL score decrease. Although statistically significant, the mean (SD) of the first PCL score was similar in both groups (66.1 [8.6] in the group with PCL score decrease ≥20 vs 64.2 [9.1] in the group with PCL score decrease <20; P < .001). The distribution of psychiatric disorders, substance use disorders, sleep disorders, smoking, hypertension, hyperlipidemia, and obesity did not significantly differ between groups. Minimally adequate duration of PTSD psychotherapy was significantly more prevalent among those who did (200 [59.0%]) vs did not (537 [42.6%]) experience a clinically meaningful PCL score decrease (P < .001). Receipt of acute-phase antidepressant medication (959 [76.2%] vs 232 [68.4%], P = .004) and receipt of any antipsychotic fill (388 [30.8%] vs 80 [23.6%], P < .001) were significantly more prevalent among patients with less than a clinically meaningful PCL score decrease.
Bivariate associations between potential confounders and incident T2D are given in Table 2. Older age (HR, 1.05; 95% CI, 1.04-1.07; P < .001) and black race (HR, 1.86; 95% CI, 1.23-2.83; P = .004) were significantly associated with incident T2D. High primary care utilization (HR, 1.55; 95% CI, 1.04-2.31; P = .03) and minimally adequate PTSD psychotherapy duration (HR, 1.73; 95% CI, 1.17-2.55; P = .006) were associated with incident T2D. Hypertension (HR, 3.46; 95% CI, 2.33-5.16), hyperlipidemia (HR, 2.82; 95% CI, 1.91-4.16), and obesity (HR, 3.32; 95% CI, 2.12-5.21) were significantly associated with incident T2D (P < .001).
The overall T2D incidence rate per 1000 person-years is given in eTable 2 in the Supplement. For the unweighted data, the T2D incidence rate was 18.0 per 1000 person-years. The age-adjusted T2D incidence rate was significantly lower among patients with a clinically meaningful PCL score decrease (7.3 per 1000 person-years) compared with those with less than a clinically meaningful PCL score decrease (16.0 per 1000 person-years) (P = .005).
The IPEW results are shown in eFigure 2 in the Supplement. The IPEW balanced all confounders (standardized mean difference <10%) between those who did and did not experience a clinically meaningful PCL score decrease. Stabilized weights ranged from 0.39 to 3.30 (mean [SD], 1.00 [0.25]), and there were no extreme weights of 10 or greater (ie, no weights were trimmed).
Results of the Cox proportional hazards regression models revealed that, before weighting in model 1, a clinically meaningful PCL score decrease vs less than a clinically meaningful PCL score decrease was associated with a significantly lower risk of T2D (HR, 0.50; 95% CI, 0.27-0.90; P = .02). This estimate remained largely unchanged after adjusting for age (HR, 0.45; 95% CI, 0.25-0.82; P = .01) and after computing the model in weighted data (HR, 0.51; 95% CI, 0.26-0.98; P = .04). Adding postindex hypertension, obesity, and hyperlipidemia to the model produced similar results (HR, 0.49; 95% CI, 0.26-0.95; P = .03).
Results of exploratory analyses, given in Table 3, revealed similar BMI values before and after a clinically meaningful PCL score decrease and before and after less than a clinically meaningful decrease. Among patients with a clinically meaningful PCL score decrease, the first mean (SD) BMI was 28.7 (5.4), and 12 months later, the mean (SD) BMI was 29.3 (5.4). Similar results were obtained among those with less than a clinically meaningful decrease. At measurements of the first and 12-month PCL scores, mean hemoglobin A1c values ranged from 5.4% to 5.5% (to convert to proportion of total hemoglobin, multiply by 0.01) in both groups. Among 324 patients with PHQ-9 scores, at the time of the first PCL score measurement, we observed a mean (SD) PHQ-9 score of 16.5 (5.4) among patients with less than a clinically meaningful PCL score decrease compared with a mean (SD) PHQ-9 score of 14.8 (6.5) among those with a clinically meaningful decrease. Twelve months later, patients who had less than a clinically meaningful PCL score decrease experienced a mean PHQ-9 decrease of 1.05 (95% CI, −1.74 to −0.37), which was significantly less than the decrease of 6.12 (95 CI, −7.71 to −4.52) among patients who had a clinically meaningful PCL decrease (P < .001).
To assess whether our results simply reflect improvement in comorbid depression or a clinically meaningful PCL score decrease is associated with lower T2D risk in patients without depression, we conducted a post hoc analysis limited to patients with only PTSD (n = 433). Among patients without depression, the age-adjusted T2D incidence rate was 2.0 per 1000 person-years among patients who experienced a clinically meaningful PCL score reduction and 12.4 per 1000 person-years among those who did not experience a clinically meaningful PCL score reduction.
The E-value was 3.33 for the association between a clinically meaningful PCL score decrease and incident T2D. An unmeasured confounder would require an association of 3.33 with a clinically meaningful PCL score decrease and incident T2D to explain our results.
In this large cohort of VHA patients with PTSD, a clinically meaningful PCL score decrease (defined as a decrease in PCL score ≥20 points) compared with less than a clinically meaningful decrease was significantly associated with a lower risk of incident T2D. This result was independent of numerous demographics and psychiatric and physical comorbidities. The association was also independent of the number of PTSD psychotherapy sessions used, suggesting that a healthy adherer effect, or a general orientation to improve health, is unlikely to explain our observations.
With use of a subsample of patients with available data, exploratory analyses suggest that depression symptoms, not BMI and hemoglobin A1c values, decreased in patients who experienced clinically meaningful PTSD improvement. These results combined with findings from our expanded Cox proportional hazards regression models suggest that clinically meaningful PTSD improvement and lower T2D risk is not associated with comorbid hypertension, hyperlipidemia, obesity, and hemoglobin A1c levels. However, modeling change in depression, BMI, and hemoglobin A1c did not include multivariate adjustment, and the role of these factors in the association between meaningful PTSD improvement and risk of T2D should be considered preliminary.
Depression remission may contribute to lower T2D incidence. Decades of research have shown a bidirectional association between depression and T2D.33-37 In patients with PTSD alone, reduction in PCL scores were associated with lower risk of T2D. On the basis of our main analyses, among patients with PTSD who had comorbid depression, a reduction in PTSD symptoms and depression was associated with lower risk for T2D. These findings suggest that improved depression does not account for our results but may be a necessary component for an association with lower T2D risk among patients with PTSD and comorbid depression.
Depression and PTSD may be linked to incident T2D through hypothalamic-pituitary-adrenal axis and cortisol dysregulation.38 Evidence for lower cortisol levels after improvement in PTSD and depression is not consistent.39 However, 1 study40 found that cortisol levels decreased in patients who experienced decreased PTSD and depression symptoms after PTSD psychotherapy. On the basis of the existing literature, we cautiously speculate that normalization of hypothalamic-pituitary-adrenal axis and cortisol levels could be one mechanism behind our results. PTSD is associated with inflammation, which may in turn be associated with increased risk for T2D.21 Thus, another possible explanation for our results is through reduction in PTSD-related inflammation through PTSD improvement or in combination with selective serotonin reuptake inhibitor use.41
Anxiety and depression have long been known to contribute to the development of cardiometabolic disease.42,43 Numerous studies, but not all,44 suggest that decreased depression is associated with improvement in insulin resistance and glycemic control45-49 and adherence to selective serotonin reuptake inhibitor therapy may be associated with reduced proinflammatory markers.50 The current study expands this field to PTSD with and without comorbid depression and T2D risk.
Prospective intervention studies are warranted to measure these factors, which are not routinely available in administrative medical records. Such a study could determine whether large decreases in PCL scores are associated with improved insulin resistance and reduced inflammation. These measures may be more sensitive to improvements in PTSD and depression and complement hemoglobin A1c values.
Unmeasured confounding could influence our results. For instance, we did not measure use of antihypertensives, some of which have potential beneficial effects on mood.51 Variables not in the medical record, such as social support, could help patients make lifestyle changes.32 However, unmeasured confounding is not likely to completely explain our findings because the E-value was 3.33. Such a value is unlikely given that the largest observed magnitude of association between measured confounders and incident T2D was hypertension. Our follow-up time is insufficient to conclude that PTSD improvement is associated with reduced T2D during the lifetime. We did not have data to link the qualifying traumatic event to PTSD diagnosis, and we were unable to draw inferences about the potential association of unique types of trauma with our results. The duration of PTSD and treatments used before the start of the data collection period could influence our results in unknown ways, but this confounding may be controlled because we balanced PTSD severity between groups. Patients with less than 2 PCL scores and not eligible for analyses had fewer PTSD psychotherapy sessions; therefore, results may not apply to younger patients because young age appears to be the only robust correlate of PTSD psychotherapy dropout.52 Also, results may not be generalizable to non-VHA patients; however, results have previously been replicated from analysis of VHA patient data in private sector cohorts on a range of topics including the association between opioids and depression53 and the association between metformin vs sulfonylurea and reduced risk for incident dementia.54
The findings suggest that clinically meaningful PTSD improvement is associated with a decreased risk of developing T2D. Patient education regarding potential health benefits of PTSD treatment may incentive psychotherapy use.
Accepted for Publication: July 26, 2019.
Corresponding Author: Jeffrey F. Scherrer, PhD, Department of Family and Community Medicine, Saint Louis University School of Medicine, 1402 N Grand Blvd, St Louis, MO 63104 (firstname.lastname@example.org).
Published Online: August 21, 2019. doi:10.1001/jamapsychiatry.2019.2096
Author Contributions: Dr Scherrer and Ms Salas had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Scherrer, Salas, Norman, Schnurr, Chard, Tuerk, Schneider, Friedman, Lustman.
Acquisition, analysis, or interpretation of data: Scherrer, Salas, Tuerk, Schneider, van den Berk-Clark, Cohen, Friedman, Lustman.
Drafting of the manuscript: Scherrer, Salas, Schneider, Lustman.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Scherrer, Salas.
Obtained funding: Scherrer, Salas, Lustman.
Administrative, technical, or material support: Scherrer, Chard, Tuerk, van den Berk-Clark.
Conflict of Interest Disclosures: Dr Scherrer reported receiving grants from the National Heart, Lung, and Blood Institute during the conduct of the study. Ms Salas reported receiving grants from the National Heart, Lung, and Blood Institute during the conduct of the study. Dr Norman reported receiving grants from the National Institutes of Health during the conduct of the study. Dr Schnurr reported receiving other support from Noblis Therapeutics outside the submitted work. Dr Tuerk reported receiving other support from Saint Louis University during the conduct of the study. Dr Lustman reported receiving grants from the National Heart, Lung, and Blood Institute during the conduct of the study. No other disclosures were reported.
Funding/Support: This study was supported by grant R01HL125424 from the National Heart, Lung, and Blood Institute (Dr Scherrer). This material is the result of work supported with resources and the use of facilities at the Harry S. Truman Memorial Veterans’ Hospital.
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: The views expressed in this report do not necessarily reflect those of the Veterans Health Administration.
Meeting Presentation: This paper was presented at the Annual Meeting of the North American Primary Care Research Group; November 16, 2019; Toronto, Ontario, Canada.
Additional Contributions: The following people provided uncompensated critiques of the original submission: Gregory Simon, MD, MPH (Kaiser Permanente Washington Health Research Institute), Mark Sullivan, MD (Department of Psychiatry and Behavioral Health, University of Washington), and Kenneth Freedland, PhD (Department of Psychiatry, Washington University School of Medicine).
Create a personal account or sign in to: