BackgroundThe relation of the variety and quantity of different sources of dietary proteins intake and diabetes remains uncertain. We aimed to investigate the associations between the variety and quantity of proteins intake from eight major food sources and new-onset diabetes, using data from the China Health and Nutrition Survey (CHNS).Methods16,260 participants without diabetes at baseline from CHNS were included. Dietary intake was measured by three consecutive 24-h dietary recalls combined with a household food inventory. The variety score of protein sources was defined as the number of protein sources consumed at the appropriate level, accounting for both types and quantity of proteins. New-onset diabetes was defined as self-reported physician-diagnosed diabetes or fasting glucose ≥7.0mmol/L or glycated hemoglobin ≥6.5% during the follow-up.ResultsDuring a median follow-up of 9.0 years, 1100 (6.8%) subjects developed diabetes. Overall, there were U-shaped associations of percentages energy from total protein, whole grain-derived and poultry-derived proteins with new-onset diabetes; J-shaped associations of unprocessed or processed red meat-derived proteins with new-onset diabetes; a reverse J-shaped association of the fish-derived protein with new-onset diabetes; L-shaped associations of egg-derived and legume-derived proteins with new-onset diabetes; and a reverse L-shaped association of the refined grain-derived protein with new-onset diabetes (all P values for nonlinearity<0.001). Moreover, a significantly lower risk of new-onset diabetes was found in those with a higher variety score of protein sources (per score increment; HR, 0.69; 95%CI, 0.65–0.72).ConclusionsThere was an inverse association between the variety of proteins with appropriate quantity from different food sources and new-onset diabetes.