Table 2.

The Cox model-estimates associated with isometric log-ratios (ilrs) expressing the work (3 ilrs) and leisure (4 ilrs) time compositions of physical behaviors. Cox proportional hazards regression models were performed to investigate the association between work and leisure time composition of physical behaviors and risk of long-term sickness absence

Variable	With LBP (n = 406)			Without LBP (n = 519)
	HR	95% CI	p	HR	95% CI	p
Work
Ilr1[ln(MVPA:sedentary, stand,LIPA)]	3.12	1.73–5.62	< 0.001	0.72	0.38–1.37	0.31
Ilr2[ln(LIPA:stand, sedentary)]	0.63	0.37–1.09	0.10	1.15	0.68–1.93	0.61
Ilr3[ln(stand:sedentary)]	1.07	0.72–1.59	0.75	0.90	0.61–1.31	0.57
Leisure
Ilr1[ln(MVPA:sedentary, stand,LIPA, time in bed)]	0.52	0.29–0.93	0.03	0.77	0.43–1.39	0.39
Ilr2[ln(sedentary:stand, LIPA, time in bed)]	1.31	0.64–2.68	0.47	1.11	0.52–2.38	0.78
Ilr3[ln(stand:LIPA, time in bed)]	2.17	0.85–5.52	0.11	0.78	0.32–1.86	0.57
lr4[ln(LIPA: time in bed)]	0.78	0.37–1.65	0.51	1.52	0.67–3.47	0.32

These estimates are from the Cox model where all ilrs from both work and leisure were included in the same model. Only estimates for ilr₁ are interpretable as it contains information for the whole composition. ilr₁ (MVPA) in the table represents the log ratio between MVPA as the numerator and the geometric mean of the remaining behaviors as the denominator. The estimates should be interpreted as one unit difference in LTSA risk corresponding to one unit change in each ilr adjusted for other ilrs and confounders

LIPA light physical activity, MVPA moderate-to-vigorous physical activity, ilr isometric log ratio, HR hazard ratio, p level of statistical significance