Supplementary Table 7. Threshold-gradient sensitivity analysis excluding BMI.

Top-two-day threshold	Weekend warrior, HR (95% CI); P	Regularly active, HR (95% CI); P
≥ 50%	0.779 (0.741, 0.818); < 0.001	0.751 (0.707, 0.798); < 0.001
≥ 55%	0.782 (0.741, 0.826); < 0.001	0.757 (0.718, 0.799); < 0.001
≥ 60%	0.770 (0.725, 0.816); < 0.001	0.769 (0.731, 0.809); < 0.001
≥ 65%	0.770 (0.720, 0.824); < 0.001	0.769 (0.733, 0.807); < 0.001
≥ 70%	0.794 (0.735, 0.857); < 0.001	0.764 (0.729, 0.801); < 0.001
≥ 75%	0.809 (0.739, 0.885); < 0.001	0.764 (0.729, 0.801); < 0.001
≥ 80%	0.781 (0.699, 0.874); < 0.001	0.768 (0.734, 0.805); < 0.001
≥ 85%	0.736 (0.633, 0.856); < 0.001	0.771 (0.736, 0.807); < 0.001
≥ 90%	0.674 (0.554, 0.821); < 0.001	0.772 (0.737, 0.808); < 0.001
≥ 95%	0.792 (0.602, 1.042); 0.096	0.769 (0.735, 0.805); < 0.001

HRs were estimated using Cox proportional hazards models as a sensitivity analysis excluding BMI from the covariate set. The exposure was the physical activity pattern defined separately at each top-two-day threshold: inactive (<150 min/week of MVPA), WW (>=150 min/week and at least the stated percentage of total weekly MVPA on the two most active days), or regularly active (>=150 min/week and below the stated percentage). Inactive participants were the reference group. The outcome was incident overall OA, defined as the first recorded ICD-10 M15-M19 diagnosis during follow-up among participants without OA at baseline. Models adjusted for age at accelerometer assessment, sex, ethnicity, education, Townsend deprivation index, smoking status, and alcohol intake frequency. Abbreviations: BMI, body mass index; CI, confidence interval; HR, hazard ratio; ICD-10, International Classification of Diseases, 10th Revision; MVPA, moderate-to-vigorous physical activity; OA, osteoarthritis; WW, weekend warrior.