Table 3.

Multivariable-adjusted risks for incident type 2 diabetes associated with modeled circadian rhythm variables.

	Model 1		Model 2		Model 3
	HR [95% CI]	p-value	HR [95% CI]	p-value	HR [95% CI]	p-value
Mean amplitude	0.95 [0.91–1.00]a	0.031	0.96 [0.92–1.00]a	0.044	0.99 [0.95–1.04]	0.66
Min. amplitude	0.92 [0.89–0.96]a	<0.0001	0.92 [0.89–0.96]a	<0.0001	0.93 [0.90–0.97]a	0.00061
Max. amplitude	0.99 [0.94–1.03]	0.53	1.00 [0.95–1.04]	0.96	1.05 [1.00–1.10]a	0.039
Phase variability	1.04 [1.02–1.07]a	0.0016	1.04 [1.01-1.06]a	0.011	1.03 [1.01-1.06]a	0.017
Mean phase (percentile group)
0–20%	1.39 [1.20–1.60]a	<0.0001	1.25 [1.08–1.45]a	0.0034	1.18 [1.01-1.37]a	0.032
20–40%	1.30 [1.12–1.50]a	0.00059	1.27 [1.09–1.47]a	0.0021	1.26 [1.08–1.47]a	0.0032
40–60% (ref.)	–	–	–	–	–	–
60–80%	1.07 [0.91–1.25]	0.41	1.10 [0.94–1.29]	0.22	1.06 [0.90–1.24]	0.49
80–100%	1.29 [1.11–1.49]a	0.00073	1.30 [1.12–1.50]a	0.00069	1.17 [1.01-1.36]a	0.042

Data are hazard ratios (95% CIs) and represent differences in risk of incident diabetes associated with a one standard deviation increase in each circadian rhythm variable: mean, min., and max. amplitude, and phase variability. For mean phase, hazard ratios represent the diabetes risk associated with each percentile group relative to the 40–60% reference group (−0.40 to 0.40 h relative to the sample circular mean phase), centered at the sample circular mean phase (03:50 h). Hourly mean phase ranges for the other percentile groups were: −12 to −1.16 h (0–20%), −1.16 to −0.40 h (20–40%), 0.40 to 1.16 h (60–80%), and 1.16 to 12 h (80–100%). Model covariates: model 1: age, sex, and ethnicity; model 2: model 1 covariates plus income, material deprivation, education, and employment status; and model 3: model 2 covariates plus smoking status, alcohol consumption, healthy diet, physical activity, and urbanicity.

^ap < 0.05.