Identifying Treatment Effect Moderators in the TARGET Trial: A Secondary Analysis

Abstract

The purpose of this secondary analysis was to identify treatment effect moderators from a large, pragmatic, cluster randomized trial for prevention of chronic low back pain (LBP). The TARGET trial was conducted at 76 primary care clinics in 4 health systems across the United States between May 2016 and June 2018. Practices were randomly assigned (1:1) to stratified care (intervention) or usual care (control). The intervention included identifying high risk patients with acute LBP and triggering an electronic best practice alert for referral for psychologically informed physical therapy (PIPT). We estimated and compared average Oswestry Disability Questionnaire (ODQ) scores at 6 months using linear mixed models controlling for baseline ODQ scores with site (fixed effects) and clinic (random effects) to account for cluster randomization. Baseline factors were considered treatment effect moderators if the type III test for the treatment x factor interaction resulted in a p-value ≤0.15. These analyses included 1,250 patients with completed baseline data and 6-month ODQ scores. Baseline characteristics were (Intervention, Control respectively): under 45 years of age (44%, 43%), female (61%, 62%), White (81%, 74%), non-Hispanic (91%, 92%), obese (50%, 48%), and non-smoking (65%, 68%). Potential effect modifiers were pain medication and smoking (p's ≤ 0.15). The intervention group had larger 6 month ODQ improvement for patients taking ≥3 pain medications, (7.0 points, P = 0.05) and current smokers (5.7 points, P = 0.02). In the TARGET trial smoking and taking ≥3 pain medications were patient level factors related to larger ODQ improvements from PIPT. The magnitude of improvements did not exceed clinically meaningful thresholds, however these factors still could be used to tailor receipt of PIPT treatment. These findings add to the existing treatment moderation literature for LBP but need to confirmed in additional trials. Patient-Centered Outcomes Research Institute (PCORI) contract # PCS-1402-10867. The purpose of this secondary analysis was to identify treatment effect moderators from a large, pragmatic, cluster randomized trial for prevention of chronic low back pain (LBP). The TARGET trial was conducted at 76 primary care clinics in 4 health systems across the United States between May 2016 and June 2018. Practices were randomly assigned (1:1) to stratified care (intervention) or usual care (control). The intervention included identifying high risk patients with acute LBP and triggering an electronic best practice alert for referral for psychologically informed physical therapy (PIPT). We estimated and compared average Oswestry Disability Questionnaire (ODQ) scores at 6 months using linear mixed models controlling for baseline ODQ scores with site (fixed effects) and clinic (random effects) to account for cluster randomization. Baseline factors were considered treatment effect moderators if the type III test for the treatment x factor interaction resulted in a p-value ≤0.15. These analyses included 1,250 patients with completed baseline data and 6-month ODQ scores. Baseline characteristics were (Intervention, Control respectively): under 45 years of age (44%, 43%), female (61%, 62%), White (81%, 74%), non-Hispanic (91%, 92%), obese (50%, 48%), and non-smoking (65%, 68%). Potential effect modifiers were pain medication and smoking (p's ≤ 0.15). The intervention group had larger 6 month ODQ improvement for patients taking ≥3 pain medications, (7.0 points, P = 0.05) and current smokers (5.7 points, P = 0.02). In the TARGET trial smoking and taking ≥3 pain medications were patient level factors related to larger ODQ improvements from PIPT. The magnitude of improvements did not exceed clinically meaningful thresholds, however these factors still could be used to tailor receipt of PIPT treatment. These findings add to the existing treatment moderation literature for LBP but need to confirmed in additional trials. Patient-Centered Outcomes Research Institute (PCORI) contract # PCS-1402-10867.