Study design and data source
This retrospective cohort study was based on a computerized database established by the Rich Healthcare Group in China, namely, the ‘DATADRYAD’ database (www.Datadryad.org). We downloaded the raw data for free from the site, provided by Chen et al. [15] from: association of body mass index and age with incident diabetes in Chinese adults: a population-based cohort study. (Dryad Digital Repository. 10.1136/bmjopen-2018-021768). The original study enrolled a total of 685,277 Chinese persons ≥ 20 years old who attended at least two visits from 2010 to 2016 across 32 sites and 11 cities in China (Shanghai, Beijing, Nanjing, Suzhou, Shenzhen, Changzhou, Chengdu, Guangzhou, Hefei, Wuhan, Nantong). The time of cohort entry was defined according to the date of the initial visit. At each visit to the health check center, participants completed a detailed questionnaire assessing demographic, lifestyle and family history of chronic disease. The trained staff administered the clinical measurements, including measurements of body weight, height and blood pressure. Fasting venous blood samples were collected after a fast of at least a 10 h at each visit. Plasma glucose levels were measured by the glucose oxidase method on an autoanalyzer (Beckman 5800). BMI was equal to the weight divided by the square of height. The data were collected under standardized conditions in accordance with uniform procedures. Laboratory methods were also carefully standardized through stringent internal and external quality controls.
The authors of the original study waived all copyright and related ownership of the raw data. Therefore, we could use these data for secondary analysis without infringing on the authors’ rights. Furthermore, the original study was approved by the Rich Healthcare Group Review Board, and the information was retrieved retrospectively. The original study was conducted in accordance with the Declaration of Helsinki, as was this secondary research. The data were anonymous, and the requirement for informed consent was waived by the Rich Healthcare Group Review Board due to the observational nature of the study.
Study sample
Consistent with the original study, participants aged 20–99 years who attended at least two visits between 2010 and 2016 were eligible for inclusion in our research. Participants were excluded at baseline in the original study if they met any of the following criteria: (1) no available information on weight, height or sex; (2) extreme BMI values (< 15 kg/m2 or > 55 kg/m2); (3) visit intervals < 2 years; (4) no available fasting plasma glucose values; and (5) diagnosis of diabetes at baseline or undefined diabetes status at follow-up. A total of 211,833 participants remained after applying the exclusion criteria in the original study [15]. In the present study, we further excluded participants with incomplete blood pressure and extreme PP values (mean ± 3 standard deviations, n = 2198). Figure 1 depicts the participant selection process. Finally, our study included 209,635 participants in the secondary analysis.
Exposure and outcome measures
The outcome of interest was incident diabetes. Diabetes mellitus was defined as fasting plasma glucose ≥ 7.00 mmol/L and/or self-reported diabetes during the follow-up period [15]. Patients were censored at the time of diagnosis of diabetes or the last visit, whichever came first.
The exposure of interest was PP which was defined as the difference between SBP and DBP. Blood pressure values were obtained by trained staff using standard mercury sphygmomanometers through office blood pressure measurements. Covariates of interest included age, sex, BMI, fasting plasma glucose (FPG), smoking status, alcohol consumption status, and family history of diabetes.
Statistical analyses
Continuous variables are expressed as the means ± standard deviations (normal distribution) or medians (quartiles) (skewed distribution), and categorical variables are expressed as frequency or percentages. Missing values for each categorical covariate (smoking and alcohol consumption status) are considered as a group.
A multiple Cox regression model was used to explore the association between PP at baseline and diabetes risk, expressed as HRs with 95% CI, which were calculated both for each 10 mmHg increase in PP and across the quartiles of PP. Covariates in the multivariable models included age, BMI, baseline FPG, smoking status, alcohol consumption status and family history of diabetes. All BP measures were not included simultaneously in regression analysis to avoid any collinearity that these independent variables may have. Sensitivity analysis was carried out in different models after excluding participants with missing values. The E-value was calculated to evaluate unmeasured confounding, which is defined as the minimum strength of association on the risk ratio scale, that an unmeasured confounder would need to have with both the treatment and the outcome to fully explain a specific treatment–outcome association, conditional on the measured covariates [16]. Multivariate adjusted smooth curve fitting was used to explore sex differences in the association between PP and diabetes risk (expressed as log RR for incident diabetes). A P value ≤ 0.05 was considered statistically significant. All statistical analyses were performed with SPSS 25.0 (IBM SPSS Inc, Chicago, IL) and R version 3.5.3.