Post-Traumatic Stress Disorder Symptoms Are Associated With Incident Chronic Back Pain

A Longitudinal Twin Study of Older Male Veterans

Pradeep Suri, MD, MS; Edward J. Boyko, MD, MPH; Nicholas L. Smith, PhD; Jeffrey G. Jarvik, MD, MPH; Gail P. Jarvik, MD, PhD; Frances M.K. Williams, PhD; Rhonda Williams, PhD; Jodie Haselkorn, MD, MPH; Jack Goldberg, PhD


Spine. 2019;44(17):1220-1227. 

In This Article

Materials and Methods

A challenge in conducting observational research is how best to isolate the effects of potential risk factors on an outcome by controlling for relevant confounders. Even with the most rigorous methods for multivariate statistical adjustment, there remains the potential for residual confounding due to factors which are unknown and unmeasured. An alternative approach is to use study designs that account for some of the unknown or unmeasured confounding factors. The co-twin control design using monozygotic (MZ) twins is one such approach. This design capitalizes on the fact that MZ ("identical") twins within a pair are perfectly matched for genetics and are also matched for "early" (childhood to adolescence) family environmental factors, as well as a host of other factors which are not directly measured, but which twins share in common. This offers a degree of control for confounding that cannot be attained in conventional studies of unrelated individuals.

Study Sample

This was a longitudinal study involving members of the Vietnam Era Twin (VET) Registry. The VET Registry was assembled from military discharge records, and includes more than 7000 male twin pairs, both of whom served on active duty during the Vietnam era (1965–1975).[24] It is the only active national twin registry in the US, and members live in all 50 states. VET Registry participants are comparable to older adult males from the US general population with respect to sociodemographic factors including educational attainment and income.[25] Further details of the VET Registry construction and zygosity ascertainment are described elsewhere.[26–29]

Between 2010 and 2012, living VET Registry members were invited to participate in an observational study of PTSD among Veterans [Cooperative Studies Program (CSP) #569]. This included a survey obtaining information regarding a broad range of physical and mental health conditions (Figure 1). Informed consent was obtained from all participating VET Registry members as part of CSP #569, and this protocol was approved by the Veterans' Affairs (VA) Puget Sound Institutional Review Board. During CSP #569, study participants reported whether they had ever had "chronic back pain" in the past, without specifying the precise location within the back, duration, or frequency of pain. We identified all 171 MZ twin pairs (n = 342 individuals) participating in CSP #569 in which both twins within a pair ("co-twins") did not report having had CBP in the past, but only one cotwin in the pair met criteria for having current PTSD symptoms. MZ twin pairs were the focus of the current study to permit within-MZ-pair analyses, which by design are controlled for underlying genetic or early family environmental factors that might predispose certain individuals to having both PTSD symptoms and CBP. These participants were contacted by mail or telephone-administered surveys between 2015 and 2017, 5 years after the time of CSP #569, to complete new data collection specifically for the purposes of the current study (Figure 1). Participants were not informed of the current study's research questions examining associations between PTSD and CBP.

Figure 1.

Flowchart of study participation. *After 5-year follow-up data collection was completed, nine respondents were subsequently found to have had missing data for chronic back pain at baseline (during CSP #569). CSP #569 indicates Cooperative Studies Program #569; PCL= PTSD, (post-traumatic stress disorder) checklist.

Assessment of PTSD Symptoms at Baseline

PTSD symptoms were assessed using the PTSD Checklist Civilian version (PCL).[30] The PCL has high test-retest reliability (r = 0.96[31]), internal consistency (α = 0.96[32]), and concurrent validity in Veterans when compared with a formal diagnosis of PTSD.[33–35] It consists of 17 Likert items that correspond to PTSD symptoms from the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV). For each item, respondents report the degree to which they were bothered by symptoms in the past month, ranging from 1 (not at all) to 5 (extremely). Individual PCL items are summed to produce a summary severity score. We defined participants with PCL scores <30 as not having current PTSD symptoms, and those with PCL scores ≥30 as having current PTSD symptoms. The PCL cut point of 30 has high sensitivity (87%) and specificity (79%) compared with a reference standard of the World Health Organization Composite Diagnostic Interview diagnosis of current PTSD,[35] and corresponds to the lowest number recommended for use in general population samples (30–35).[36]

Assessment of Covariates

We considered as confounders a selected list of baseline variables with a conceptual rationale for potentially being associated with both PTSD and back pain. Data on age and race were previously abstracted from military service records during the 1980s. Other data collected by participant self-report during CSP #569 included educational attainment, annual family income, having ever used VA health care, having ever received or applied for VA disability compensation for any health condition, current cigarette smoking, and height and weight [used to calculate body mass index (weight in kilograms/height in meters)[2]]; these variables are further described elsewhere.[25] Current depression symptoms were measured using the Patient Health Questionnaire (PHQ-9), a measure of depression that grades the severity of nine depression symptoms over the past 2 weeks (range 0–27, with higher scores representing greater depression symptoms).[37] A cut point of 5 or higher was used to define depression.[37]

Assessment of Back Pain Outcomes at 5-year Follow-up

At the 5-year follow-up, participants were asked the question "In the past 4 weeks, have you had low back pain?" anchored to a pain diagram indicating the lumbar region. Subsequent questions inquired about low back pain duration and intensity on a 0 to 10 numerical rating scale (NRS), where "0" reflects no pain and 10 reflects the worst pain possible, using question items from the minimal dataset of the US National Institutes of Health (NIH) task force on research standards for chronic low back pain.[38] These questions were then asked in regards to mid/upper back pain in the past 4 weeks, also defined as "thoracic" pain and accompanied by an illustration depicting the thoracic region. Participants completed the Oswestry Disability Index (ODI), accounting for back symptoms in the low back and mid/upper back over the past month. The ODI is a widely used and validated index of back pain-related functional limitations; scores range from 0 to 100, where 0 represents no limitations and 100 represents the most severe limitations possible.[39] The primary study outcome was incident CBP (duration ≥3 months) in the low back and/or mid/upper back at 5-year follow-up. Secondary outcomes by specific region of the back included low back pain of ≥3 months duration ("incident chronic low back pain") and mid/upper back pain of ≥3 months duration ("incident chronic mid/upper back pain").

Statistical Analysis

We assessed normality of continuous variables using graphical methods and found none in violation of this assumption. We compared baseline characteristics by incident CBP at 5-year follow-up. We conducted within-MZ-pair analyses using generalized linear models to examine the association of baseline PTSD symptoms with incident CBP at 5-year follow-up. This was done first by examining bivariable associations from models including baseline PSTD symptoms and incident CBP, and next in multivariable analyses adjusting for potential confounders. We identified possible confounders of the PTSD-CBP relationship by examining for ≥10% changes in parameter estimates between bivariable models and multivariable models adjusting for the potential confounding variable.[40] We calculated risk ratios (RRs) and 95% confidence intervals (95% CIs), setting the threshold for statistical significance at P ≤ 0.05. The same methods were applied to the secondary back pain outcomes. Sample size was determined a priori based on assumptions of a 34% incidence of CBP at 5-year follow-up using annual incidence data from a prior study of Veterans,[41] two-sided α of 0.05, 80% power to detect a relative risk of CBP ≥2.0 for those with PTSD,[42] and a 60% response rate from the targeted group of 342 VET Registry members who were sent surveys. Analyses were conducted using STATA v.15.1 (STATA Corp., College Station, TX).