Table 1

Adult life course models testing the way in which exposure to high BMI (top quartile) in adulthood might affect later life knee OA

	BMI trajectory		BMI (kg/m²)				Life course models (test of fit against the saturated model^*)
	1=high		Median
	26 Years	43 Years	26 Years	43 Years	N (%)	Knee OA (%)	Accumulation (duration)^†	Accumulation (duration and age)^‡	Period^§ (early adulthood)	Period (mid-adulthood)
Men	0	0	21.9	24.2	829 (66.1)	6.2	χ² (2)=1.65; p=0.438	χ² (1)=1.19; p=0.276	χ² (2)=3.79; p=0.150	χ² (2)=1.37; p=0.504
	1	0	26.0	26.0	114 (9.1)	8.8
	0	1	23.7	28.8	118 (9.4)	11.0
	1	1	27.1	30.5	194 (15.5)	9.3
Women	0	0	21.0	22.9	870 (65.6)	8.5	χ² (2)=3.31; p=0.191	χ² (1)=0.55; p=0.458	χ² (2)=26.16; p<0.001	χ² (2)=5.09; p=0.078
	1	0	24.8	24.9	120 (9.1)	14.2	χ² (2)=3.31; p=0.191	χ² (1)=0.55; p=0.458	χ² (2)=26.16; p<0.001	χ² (2)=5.09; p=0.078
	0	1	22.6	29.1	130 (9.8)	21.5
	1	1	26.2	31.1	206 (15.5)	26.7

↵* The saturated model allows each BMI trajectory to affect the likelihood of knee OA. The model that differs the least from the saturated model in terms of explaining knee OA is the one with most support. Hence, larger p values indicate a better model (highlighted in boldface).
↵† Accumulation (duration): this model assumes that the effect of having a high BMI is the same at each age, that is, it is only the amount of time exposed that is important (see online supplementary text S2).
↵‡ Accumulation (duration and age): this model is the same as the duration model but also allows the effect of having a high BMI to differ at each age, that is, exposure time and age of exposure are allowed to influence the likelihood of knee OA (see online supplementary text S2).
↵§ Period: These models allow exposure to affect knee OA at one age only.
BMI, body mass index; OA, osteoarthritis.