Table 5.

The relationship between RFM and incident diabetes is analyzed using a two-piecewise Cox proportional hazards regression model in different sensitivity analyses.

Incident DM	Female (HR, 95%CI, P)	Male (HR, 95%CI, P)	All participants (HR, 95%CI, P)
Model 4
Fitting model by standard linear regression	1.15 (1.05, 1.27) 0.0026	1.06 (0.99, 1.13) 0.1181	1.08 (1.03, 1.14) 0.0033
Fitting model by two-piecewise Cox proportional hazards regression
The inflection point of RFM	38.99	21.54	21.08
≤ Inflection point	1.12 (1.02, 1.23) 0.0141	0.95 (0.87, 1.04) 0.2724	0.96 (0.88, 1.05) 0.3448
> Inflection point	1.40 (1.18, 1.67) 0.0002	1.15 (1.05, 1.25) 0.0018	1.12 (1.06, 1.19) 0.0001
P for log-likelihood ratio test	0.011	0.002	0.003
Model 5
Fitting model by standard linear regression	1.13 (1.03, 1.24) 0.0104	1.04 (0.97, 1.12) 0.2628	1.07 (1.01, 1.14) 0.0143
Fitting model by two-piecewise Cox proportional hazards regression
The inflection point of RFM	38.66	23.29	22.29
≤ Inflection point	1.10 (1.00, 1.20) 0.0525	0.97 (0.89, 1.05) 0.4154	0.97 (0.89, 1.05) 0.3934
> Inflection point	1.42 (1.19, 1.68) < 0.0001	1.17 (1.06, 1.29) 0.0025	1.12 (1.05, 1.19) 0.0005
P for log-likelihood ratio test	0.003	0.001	0.001

Model 4 was a sensitivity analysis in participants with age < 65 years.

Model 5 was a sensitivity analysis conducted on participants with SBP < 140mmHg and DBP < 90mmHg.

Note 1: In all participants, we adjusted gender, age, BMI, alcoholic intake, smoking status, exercise habits, SBP, DBP, ALT, AST, GGT, HDL-C, TC, TG, HbA1c, and FPG.

Note 2: For female and male subgroups, we adjusted for age, BMI, alcoholic intake, smoking status, exercise habits, SBP, DBP, ALT, AST, GGT, HDL-C, TC, TG, HbA1c, and FPG.

HR: hazard ratios; CI: confidence; DM: diabetes mellitus; RFM: relative fat mass