OBJECTIVETo investigate the association between dietary patterns and the risk of developing hyperglycemia in Nanjing.

METHODSUsing multi-stage stratified random cluster sampling, the baseline survey was conducted on local residents older than 30 years in 7 communities from 2 urban districts from June to September 2007 in Nanjing. The total eligible subjects were 3376. Excluding the 476 previously diagnosed hyperglycemia patients, 2900 non-hyperglycemia subjects were used as the baseline sample for the follow-up survey from June to September 2010. Using specially designed food frequency questionnaire (FFQ), factor analysis was applied to identify food patterns. Multivariable linear and Cox regression models were used to analyze the association between different dietary patterns and risk of hyperglycemia.

RESULTSThe follow-up rate was 72.2%, with 2093 subjects participated the follow-up survey in 3 years. Three-year cumulative incidence of hyperglycemia was 7.5% (158/2093). The incidence rate was 7.1% (62/873) for males and 7.9% (96/1220) for females, but the differences were not statistically significant (χ(2) = 0.43, P = 0.513). Five dietary patterns were identified by factor analysis: condiment, animal and plant protein, traditional healthy, sweet food and alcohol drinking. By multivariable linear regression, on average, an increase in traditional healthy pattern and sweet food pattern of 1 unit was associated with a -0.054, 0.050 mmol/L increase in fasting blood glucose, respectively, and the differences were both statistically significant (t = -2.38, 2.27, respectively, P values were both less than 0.05). By multivariable Cox regression, the pattern sweet food was positively significantly associated with hyperglycemia risk in men. The incidence of hyperglycemia was 4.7% (14/295) for the lowest tertile of the factor score (T1), and 9.7% (26/269) for the highest tertile of the factor score (T3) (T3:T1: RR = 1.88, 95%CI: 1.04 - 3.54). The pattern traditional healthy was inversely associated with hyperglycemia risk in women. The incidence of hyperglycemia was 10.7% (45/421) for T1 and 6.3% (21/335) for T3 (T3:T1: RR = 0.59, 95% CI: 0.35 - 0.99). Conversely, a statistically significant positively association was observed for the pattern alcohol drinking in women. The incidence of hyperglycemia was 8.1% (38/472) for T1 and 11.1% (33/297) for T3 (T3:T1: RR = 1.35, 95%CI: 0.84 - 2.16).

CONCLUSIONDietary patterns are associated with hyperglycemia. The sweet food pattern is a risk factor for hyperglycemia in men. In women, healthy dietary pattern is healthy and the alcohol drinking pattern is a risk factor for hyperglycemia.