Application of Heterogeneity of Treatment-Effects Methods: Exploratory Analyses of a Trial of Exercise-Based Interventions for Knee Osteoarthritis

Abstract

Objective: To evaluate heterogeneity of treatment effects in a trial of exercise-based interventions for knee osteoarthritis (OA).

Methods: Participants (n = 350) were randomized to standard physical therapy (PT; n = 140), internet-based exercise training (IBET; n = 142), or wait list (WL; n = 68) control. We applied qualitative interaction trees (QUINT), a sequential partitioning method, and generalized unbiased interaction detection and estimation (GUIDE), a regression tree approach, to identify subgroups with greater improvements in Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) score over 4 months. Predictors included 24 demographic, clinical, and psychosocial characteristics. We conducted internal validation to estimate optimism (bias) in the range of mean outcome differences among arms.

Results: Both QUINT and GUIDE indicated that for participants with lower body mass index (BMI), IBET was better than PT (improvements of WOMAC ranged from 6.3 to 9.1 points lower), and for those with higher BMI and a longer duration of knee OA, PT was better than IBET (WOMAC improvement was 6.3 points). In GUIDE analyses comparing PT or IBET to WL, participants not employed had improvements in WOMAC ranging from 1.8 to 6.8 points lower with PT or IBT versus WL. From internal validation, there were large corrections to the mean outcome differences among arms; however, after correction, some differences remained in the clinically meaningful range.

Conclusion: Results suggest there may be subgroups who experience greater improvement in symptoms from PT or IBET, and this finding could guide referrals and future trials. However, uncertainty persists for specific treatment-effects size estimates and how they apply beyond this study sample.

Trial registration: ClinicalTrials.gov NCT02312713.

Figure 1.

QUINT subgroups with 2 active arms Minimum sample size for subgroup is n=40 total participants (i.e., at least n=20 in each arm); The subgroups are as follows: subgroup 1: BMI ≤ 24.31 kg/m² (n=44; mean difference =−7.2 points); subgroup 2: BMI > 24.31 kg/m², OA Symptom duration ≤ 9.5 years, and score on Social Support for Exercise ≤ 56.5 (n=57; mean difference = −6.3 points); subgroup 3: BMI > 24.31 kg/m², OA Symptom duration ≤ 9.5 years, and score on Social Support for Exercise > 56.5 (n=50; mean difference = 3.1 points); and subgroup 4: BMI > 24.31 kg/m², OA Symptom duration > 9.5 years, (n=127; mean difference = 6.3 points).

Figure 2.

QUINT subgroups with two active arms Minimum sample size per subgroup is n=30 total participants (i.e., at least n=15 per arm); The subgroups are as follows: subgroup 1: BMI ≤ 23.94 kg/m² (n=38; mean difference =−9.1 points); subgroup 2: BMI > 23.94 kg/m², Age ≤ 55.54 years(n=33; mean difference = −8.0 points); subgroup 3: BMI > 23.94 kg/m², Age >55.54 years, and number of chair stands ≤ 8.5 (n=86; mean difference = 9.6 points); subgroup 4: BMI > 23.94 kg/m², Age >55.54 and ≤72.03 years, and number of chair stands > 8.5, (n=79; mean difference = 4.7 points), and subgroup 5: BMI > 23.94 kg/m², Age >72.03 years, and number of chair stands > 8.5, (n=42; mean difference = −4.9 points).

Figure 3.

GUIDE subgroups with two active arms IBET and PT, n=278 with minimum sample size per subgroup of n=40 total participants (i.e., at least n=20 in each arm); The subgroups are as follows: subgroup 1: OA Symptom duration ≤ 9.5 years and BMI ≤ 29.45 kg/m² (n=59; unadjusted mean difference =−6.9 points, adjusted mean difference =−5.0 points); subgroup 2=:OA Symptom duration ≤ 9.5 years and BMI > 29.45 kg/m² (n=72; unadjusted mean difference =0.1 points, adjusted mean difference =−0.5 points); subgroup 3: OA Symptom duration > 9.5 years and OA Symptom duration ≤ 18.5 years (n=80; unadjusted mean difference = 3.5 points, adjusted mean difference=3.1 points); and subgroup 4: OA Symptom duration > 18.5 years (n=67; unadjusted mean difference = 5.7 points, adjusted mean difference=7.5 points).

Figure 4.

GUIDE subgroups with all three arms IBET, PT and WL. n=345 with minimum sample size per subgroup of n=60 total participants (i.e., at least n=20 in each arm); mean differences for subgroups (nodes) are IBET-Usual care followed by PT-WL with negative values indicating greater improvement in treatment arm (IBET or PT) compared to WL. The subgroups are as follows: subgroup 1: Not employed (n=205; unadjusted IBET- WL mean difference =−3.8 points, adjusted mean difference=−5.4; unadjusted PT-WL mean difference=−6.4, adjusted mean difference=−7.4); subgroup 2: Employed (n=140; unadjusted IBET- WL mean difference =0.7 points, adjusted mean difference=1.2; unadjusted PT-WL mean difference=1.6, adjusted mean difference=2.1).

Figure 5.

GUIDE subgroups with all three arms IBET, PT and WL n=345 with minimum sample size per subgroup of n=45 total participants (i.e., at least n=15 in each arm); Subgroups are as follows: subgroup 1: Not employed and OA symptom duration ≤ 10.5 years (n=115; unadjusted IBET- WL mean difference =−5.7 points, adjusted mean difference=−5.7; unadjusted PT-WL mean difference=−6.8, adjusted mean difference=−6.5); subgroup 2: Not employed and OA symptom duration > 10.5 years (n=90; unadjusted IBET-WL mean difference =−1.8 points, adjusted mean difference=−5.8; unadjusted PT-WL mean difference=−6.2, adjusted mean difference=−11.4); subgroup 3: Employed and number of chair stands ≤ 9.5 (n=64; unadjusted IBET-WL mean difference =−1.8 points, adjusted mean difference=−1.4; unadjusted PT-WL mean difference=0.7, adjusted mean difference=0.7); subgroup 4: Employed and number of chair stands > 9.5 (n=76; unadjusted IBET- WL mean difference =2.6 points, adjusted mean difference=3.7; unadjusted PT-WL mean difference=2.1, adjusted mean difference=2.3)