2.1 Study population

This prospective cohort study was derived from the China Health and Retirement Longitudinal Study of 10 257 households and 17 708 middle-aged and elderly participants from 150 counties or districts, 450 villages or urban communities in 28 provinces.^{25, 26} The respondents were followed up four times every 2 years through a face-to-face personal interview from June 2011 and March 2012 to July 2018 and March 2019. Moreover, the blood sample was collected in 2011 and 2015.

Individuals with complete data on serum creatinine, WC, and cystatin C and who were not diabetic at baseline were included (n = 6318). We excluded those with self-reported cancer (n = 52) and end-stage renal disease²⁷ (estimated GFR [eGFR] <15 mL/min/1.73m²) (n = 3). Participants who did not complete at least two visits were excluded (n = 212). We also excluded those with outliers of WC (<means minus 3 SD or > means plus 3 SD) (n = 113). Finally, a total of 5938 participants were included in the study (Figure 1).

The ethics review committee approved this study of Peking University. All participants signed written informed consent.

2.2 Measurement

At baseline, information on demographic characteristics, lifestyle, and medical history were collected by structured questionnaires. Smoking and drinking status were classified as never, former, and current. As for anthropometric measurement, body weight and height were measured to the nearest 0.1 kg and 0.1 cm, respectively, by light clothes and without shoes. BMI was calculated as weight (kg) divided by height (m) squared. WC was measured midway between the lower rib and the upper iliac crest to the nearest 0.1 cm. Systolic (SBP) and diastolic blood pressures were measured three times in the seated position after 10 min of rest by using a sphygmomanometer at intervals of 45 s, and the average was used for analysis. Professional nurses collected all venous blood samples after fasting for at least 12 h at night.²⁸ Blood glucose, total cholesterol (TC), triglycerides (TG), low-density lipoprotein (LDL-C), and high-density lipoprotein (HDL-C) were measured by enzymatic colormetric test. The glycosylated hemoglobin (HbA1c) was measured by boronate affinity high-performance liquid chromatography. eGFR was calculated from the creatinine level by using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equations.²⁹ eGFR_CKD-EPI was calculated as follows: for females with a plasma creatinine ≤0.7, (plasma creatinine/0.7)^−0.329 × (0.993)^age (×144 if white or other); for females with a plasma creatinine >0.7, (plasma creatinine/0.7)^−1.209 × (0.993)^age (×144 if white or other); for males with a plasma creatinine ≤0.9; (plasma creatinine/0.9)^−0.411 × (0.993)^age (×141 if white or other); for males with a plasma creatinine >0.9, (plasma creatinine/0.9)^−1.209 × (0.993)^age (×144 if white or other). The serum creatinine was tested by rate-blanked and compensated Jaffe creatinine method. Cystatin C was measured by particle-enhanced turbimetric assay. CCR was calculated as [creatinine (mg/dL)/cystatin C (mg/L)] × 100. CCR/WC ratios were calculated as creatinine to cystatin C to waist circumference.

2.3 Definition

Based on the American Diabetes Association criteria,³⁰ T2DM was defined as fasting blood glucose ≥7.0 mmol/L, and/or random blood glucose ≥11.1 mmol/L, and/or HbA1c ≥6.5%, and/or self-report diabetes, and/or taking antidiabetes treatment.

2.4 Statistical analysis

Baseline characteristics were described with median (interquartile range) for nonnormally distributed continuous variables and frequency (percentage) for categorical variables. The participants were classified by three CCR/WC tertiaries. Differences in baseline variables among the three categories were detected using analysis of variance for continuous variables and chi-squared tests for categorical variables. Cox proportional hazards regression models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for incident T2DM by tertiaries of CCR/WC, with the lowest tertiary as the reference. Model 1 was adjusted for age and sex. Model 2 was adjusted for model 1 plus marital status, educational level, smoking status, and drinking status. Model 3 was adjusted for model 2 plus BMI, SBP, TC, TG, LDL-C, HDL-C, and uric acid. P values for trends across CCR/WC tertiaries were calculated by a median value as a continuous variable. We used restricted cubic spline regression to explore possible linear or nonlinear correlation between incident T2DM and CCR/WC. Subgroup analyses were conducted by sex, age, and BMI. In the sensitivity analysis, first, we removed individuals who had T2DM within the first 2 years to control the potential contribution of reverse causation. Second, we used multiple imputations for missing data. Two-sided p < .05 was considered as statistically significant. The statistical analyses were performed using SPSS software version 26 (SPSS Inc) and R statistical software version 3.6.1 (R Foundation).

3.1 Baseline characteristics of study participants

In total, 5938 participants (men = 2758, women = 3180) were involved in our analysis. The median age at baseline was 59.00 years and 53.6% were women. Table 1 shows the characteristics of the participants. According to CCR/WC tertiaries, participants were divided into three subgroups: tertiary 1 (Q1), CCR/WC ≤0.83; tertiary 2 (Q2), 0.84 ≤ CCR/WC ≤1.02; tertiary 3 (Q3), CCR/WC ≥1.03. Participants in higher tertiaries were more likely to be younger, male, married, high education level, alcohol drinkers, smokers, more likely to have lower hypertension, hyperlipidemia, and cardiovascular disease; were more likely to have higher levels of height, TC, HDL-C, creatinine, uric acid; and were more likely to have lower weight, BMI, SBP, TG, LDL-C, WC, and cystatin C. However, no significant difference in blood glucose, TG, and LDL-C was found.

3.2 Risk of incident T2DM by tertiaries of CCR/WC

After 37679.00 person-years follow-up (median 7.00 years), 766 participants developed T2DM. Thus, the T2DM incidence was 20.33 per 1000 person-years. Table 2 indicates the relationship between tertiaries of CCR/WC and T2DM incidence. With the CCR/WC increased, the incidence of T2DM declined in the crude and adjusted model. After adjusting for potential confounders, including age, sex, marital status, educational level, smoking status, drinking status, BMI, SBP, TC, TG, LDL-C, HDL-C, and uric acid. HRs (95% CIs) for T2DM across tertiaries of CCR/WC were 1.000 (reference), 0.772(0.647, 0.921), and 0.724 (0.596, 0.880; p for trend = .001).

A nonlinear dose–response relationship was observed between CCR/WC and the risk of T2DM using restricted cubic spline regression (Figure 2). The HRs for the association between CCR/WC and incident T2DM decreased with increasing CCR/WC levels. The risk of T2DM decreased rapidly until around 0.91 of ratio and then started relatively flat afterwards (p for nonlinearity = .03). As CCR/WC elevated, the negative association between CCR/WC and the risk of T2DM became flatter when CCR/WC exceeded 0.85 for females. Further subgroup analyses of the ratio and incident T2DM were performed according to gender, age, and BMI (Figure 3). CCR/WC ratio was strongly associated with the incidence of T2DM in women but attenuated in men. We found at age < 65 years the HR (95% CI) for the association between CCR/WC ratio and T2DM incidence was 0.665 (0.524–0.844) and 0.804 (0.645–1.003) comparing Q3 and Q2 with Q1. In age ≥ 65 years, compared with Q1, a higher ratio was associated with reduced risk of T2DM in Q2 (HR, 0.713 [95% CI: 0.527–0.965]) and in Q3 (HR, 0.881 [95% CI: 0.630–1.232]). The relationship between the ratio and T2DM incidence is independent of BMI group.

The Kaplan–Meier curve indicated a significant difference in the incidence of T2DM among the CCR/WC tertiary (log-rank test, p < .0001). The cumulative risk of T2DM increased over time by this ratio. Moreover, it also found statistically significant differences in males and females respectively (Figure 4).

3.3 Sensitivity analyses

We performed a sensitivity analysis to analyze potential confounders, shown in Tables S1 and S2. Excluding participants with impaired fasting glucose and T2DM in the first 2 years did not materially change the associations.

3.4 Multiple Imputation for Missing Data

The study contained 60 missing values of SBP, 30 of height, 29 of weight, 2 of education, 4 of smoking status, 4 of drinking status, 2 of HDL-C, and 2 of LDL-C. We observed the same consequence in tertiaries of CCR/WC and incident T2DM after multiple imputations for missing data, shown in Table S3. In model 3, HRs (95% CIs) for T2DM across tertiaries of CCR/WC were 1.000 (reference), 0.760 (0.638, 0.905), and 0.717 (0.591, 0.869; p for trend = .001).