Which Chronic Low Back Pain Patients Respond Favorably to Yoga, Physical Therapy, and a Self-care Book? Responder Analyses from a Randomized Controlled Trial

Abstract

Purpose: To identify baseline characteristics of adults with chronic low back pain (cLBP) that predict response (i.e., a clinically important improvement) and/or modify treatment effect across three nonpharmacologic interventions.

Design: Secondary analysis of a randomized controlled trial.

Setting: Academic safety net hospital and seven federally qualified community health centers.

Subjects: Adults with cLBP (N = 299).

Methods: We report patient characteristics that were predictors of response and/or modified treatment effect across three 12-week treatments: yoga, physical therapy [PT], and a self-care book. Using preselected characteristics, we used logistic regression to identify predictors of "response," defined as a ≥30% improvement in the Roland Morris Disability Questionnaire. Then, using "response" as our outcome, we identified baseline characteristics that were treatment effect modifiers by testing for statistical interaction (P < 0.05) across two comparisons: 1) yoga-or-PT vs self-care and 2) yoga vs PT.

Results: Overall, 39% (116/299) of participants were responders, with more responders in the yoga-or-PT group (42%) than the self-care (23%) group. There was no difference in proportion responding to yoga (48%) vs PT (37%, odds ratio [OR] = 1.5, 95% confidence interval = 0.88 - 2.6). Predictors of response included having more than a high school education, a higher income, employment, few depressive symptoms, lower perceived stress, few work-related fear avoidance beliefs, high pain self-efficacy, and being a nonsmoker. Effect modifiers included use of pain medication and fear avoidance beliefs related to physical activity (both P = 0.02 for interaction). When comparing yoga or PT with self-care, a greater proportion were responders among those using pain meds (OR = 5.3), which differed from those not taking pain meds (OR = 0.94) at baseline. We also found greater treatment response among those with lower (OR = 7.0), but not high (OR = 1.3), fear avoidance beliefs around physical activity.

Conclusions: Our findings revealed important subgroups for whom referral to yoga or PT may improve cLBP outcomes.

Trial registration: ClinicalTrials.gov NCT01343927.

Keywords: Chronic Low Back Pain; Effect Modification; Health Disparities; Physical Therapy; Predictor of Response; Primary Care; Yoga.

© The Author(s) 2020. Published by Oxford University Press on behalf of the American Academy of Pain Medicine. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.

Figures

Figure 1

Flow diagram of postintervention response comparisons for yoga, physical therapy, and self-care.

Figure 2

Characteristics that are predictors of response and/or treatment effect modifiers, or neither, when comparing yoga or PT with self-care. For each panel, the baseline characteristic is shown as a potential predictor of response on the left and a potential treatment effect modifier on the right. The corresponding P values are from logistic regression models adjusted for age and baseline back-specific physical function on the Roland Morris Disability Questionnaire. In panel a, sex was neither a predictor of response (P = 0.46) nor a treatment effect modifier (P for interaction = 0.69). In panel b, the number of chronic conditions was a predictor of response (P < 0.001) but not a treatment effect modifier (P for interaction = 0.21). In panel c, higher fear avoidance beliefs around physical activity was not a predictor of response (P = 0.60) but was a treatment effect modifier (P for interaction = 0.02). In panel d, baseline pain medicine use was both a predictor of response (P < 0.001) and a treatment effect modifier (P for interaction = 0.02).

Figure 3

Forest plot of effect modification on the multiplicative scale. Odds ratios indicate likelihood of being a responder. For example, participants using pain medications at baseline were more likely to be responders if they were in the yoga or physical therapy (PT) group compared with self-care group (odds ratio [OR] = 5.3, 95% confidence interval [CI] = 2.0 − 14.2). In contrast, among those not taking pain medications at baseline, the proportion of responders was not statistically different in yoga or PT compared with self-care (OR = 0.94, 95% CI = 0.31 − 2.9).

Appendix Figure 1

Clinical decision-making tree.