Post-Traumatic Stress Disorder and Risk of Parkinson’s Disease in a Veteran Cohort

METHODS

This was a retrospective cohort study of two groups of Veterans living at one of two Marine Corps base camps between 1975 and 1985. We obtained data from the Agency for Toxic Substances and Disorders Registry (ATSDR) for 340,489 US Veterans who were based at Camp Lejeune, North Carolina or Camp Pendleton, California, and identified those who utilized any VHA health care services (through 2/17/2021) and/or had Medicare claims (through 12/31/2018). 158,122 Veterans (46% of initial cohort) had any healthcare utilization. We reviewed the VHA electronic medical records (EMR) of all Veterans with ≥1 diagnostic codes for PD (ICD9 332.0 or ICD10 G20) to validate this diagnosis based on documentation in the EMR including symptomatology and PD medication use and to determine incident date of diagnosis. After validation, we identified 430 Veterans with PD; as previously reported in Goldman et al. a significantly higher proportion of these Veterans had resided at Camp Lejeune.⁷

We defined a Veteran as having had PTSD if s/he had ≥ two encounters, at least 3 months apart, with a documented diagnosis of PTSD based on ICD9 (309.81) or ICD10 (F43.10, F43.11, F43.12) codes in the VHA EMR or in Medicare claims files. Using a nested case-control design, we tested associations of PTSD and PD by matching 10 controls to each Veteran with PD based on age at PD diagnosis (“index date”), sex, race, and rank (i.e., enlisted, officer). We adjusted for Camp (Lejeune, Pendleton) in conditional logistic regression models, and conducted sensitivity analyses restricted to cases and controls with any VHA usage prior to index date, in sex and race-specific strata (White, non-White), and in each Camp separately. We also tested models including cigarette smoking (ever/never), though this variable was missing for a substantial proportion of cases. Camp and smoking status were included to remove them as potential confounders in our analyses.

This study was reviewed and approved by the institutional review boards of the University of California San Francisco, the San Francisco VA and Hines VA Hospital.

RESULTS

PTSD was identified in 15.1% (n = 65) of Veterans with PD and 12.6% (n = 544) of matched controls (Table 1). Veterans with PD were less likely than controls to have ever smoked (38.1% versus 47%, p < 0.0005).

In regression models adjusted for camp, the risk of PD was significantly higher in Veterans with PTSD (OR 1.23, 95% CI 1.09–1.40) (Table 2). Inclusion of smoking had minimal impact on the association with PTSD. Risk was similar in models restricted to males or females-only, though statistical power was limited for the latter given the small number of females. Risk was also similar in White and non-white Veterans (data not shown).

In camp-specific models, the association of PD and PTSD was stronger in Veterans from Camp Pendleton than in those from Camp Lejeune (OR 1.66 versus 1.17), but a multiplicative interaction term was not statistically significant (p = 0.34).

When we only included cases (n = 259) and matched controls (n = 2,515) with any VHA health care use prior to being diagnosed with PD, the association with PTSD was strengthened (OR = 1.49; CI: 1.29–1.72; p < 0.0001). Limiting analyses to cases in which a PTSD diagnosis occurred before the first PD diagnosis, resulted in a 1.53 times greater odds of developing PD than not having a prior PTSD diagnosis (p < 0.0001; Table 2). Finally, restricting our sample to only cases where PTSD was documented at least one year prior to the PD diagnosis resulted in a non-significant association for PTSD and PD.

DISCUSSION

We found a statistically significant association of PTSD and PD using rigorous case-finding methods, replicating several prior reports.^4,5 Importantly, accurate PD diagnosis requires a high degree of clinician expertise, and it is often misdiagnosed in administrative datasets.^8,9 We identified PD cases using a combination of ICD codes, and/or use of PD medications and validated diagnoses by chart review. From the list of 1,095 cases with a PD ICD code, chart reviews validated a diagnosis of PD in only 430 cases (39%). This raises concerns for studies that only use diagnostic codes to identify persons with a diagnosis of PD, as many of these individuals may in fact, have other neurologic-related conditions such as secondary parkinsonism from antipsychotic medications.

The association of PTSD and PD is biologically plausible. Stress enhances neuronal monoamine turnover, increases central and peripheral markers of inflammation, activates microglia, and enhances pathologic severity in a variety of animal models of parkinsonism.^10,11

The validation effort was a strength of this project as we continue to see that PD is both difficult to reliably diagnose and is often not managed by movement disorders neurologists who have specialized training to diagnose PD.^8,9,12 We were also able to draw from both VA and Medicare data to improve the sensitivity of case ascertainment. Of note, however, our study cohort was young (mean age < 63 years), which limited the number likely to utilize Medicare; this may result in reduced power to assess the association between PTSD and PD. While we did not validate the PTSD diagnosis, we required at least two diagnoses in the EMR at least 3 months apart to increase accuracy.

Although PTSD was significantly associated with PD in both Camps, the association was stronger at Camp Pendleton. There are many possible reasons for this including differential rates of combat deployment and/or seeking care within the VA. It may explain why requiring at least a year between PTSD and PD diagnoses resulted in a non-significant association, as more Pendleton cases were dropped, affecting our sample size and statistical power.

The primary limitation to these analyses is the difficulty in determining the temporal association between PTSD and PD. Documentation of PTSD in the EMR likely reflects when the provider learned about the PTSD rather than when it actually occurred. PTSD diagnosis in Veterans is often associated with their military service and would likely be documented closer to the occurrence of symptoms. The event(s) underlying the PTSD may have occurred much earlier than when it was documented in the VA records, as our data only go back to October 1999. Our cohort of Veterans were active military 15–25 years prior to when we had any diagnostic information for them. Even with this limitation, we did find increased odds of being diagnosed with PD following a documented diagnosis of PTSD in our cohort. Additionally, though we adjusted for potential confounders and explored sensitivity models stratified by sex and race, we were unable to consider other potential modifiers of the relationship between PTSD and PD. For example, a recent paper using a very large cohort of military Veterans found that a prior documented diagnosis of PTSD or traumatic brain injury increased the odds of developing PD and the combination of these conditions increased the odds beyond either diagnosis alone.¹³

PTSD is associated with increased risk of PD in this well-validated cohort of US Marine Corps Veterans. Using a rigorous case-finding approach and a nested case-control design, these results strengthen the evidence of the association between PTSD and PD. Stress mitigation and treatment of PTSD may potentially help to delay or prevent PD.