Abstract
Background
Despite the well-recognised health benefits of fresh fruit consumption, substantial uncertainties remain about its potential effects on incident diabetes and, among those with diabetes, on risks of death and major vascular complications.
Methods and findings
Between June 2004 and July 2008, the nationwide China Kadoorie Biobank study recruited 0.5 million adults aged 30–79 (mean 51) y from ten diverse localities across China. During ~7 y of follow-up, 9,504 new diabetes cases were recorded among 482,591 participants without prevalent (previously diagnosed or screen-detected) diabetes at baseline, with an overall incidence rate of 2.8 per 1,000 person-years. Among 30,300 (5.9%) participants who had diabetes at baseline, 3,389 deaths occurred (overall mortality rate 16.5 per 1,000), along with 9,746 cases of macrovascular disease and 1,345 cases of microvascular disease. Cox regression yielded adjusted hazard ratios (HRs) associating each disease outcome with self-reported fresh fruit consumption, adjusting for potential confounders such as age, sex, region, socio-economic status, other lifestyle factors, body mass index, and family history of diabetes. Overall, 18.8% of participants reported consuming fresh fruit daily, and 6.4% never/rarely (non-consumers), with the proportion of non-consumers about three times higher in individuals with previously diagnosed diabetes (18.9%) than in those with screen-detected diabetes (6.7%) or no diabetes (6.0%). Among those without diabetes at baseline, higher fruit consumption was associated with significantly lower risk of developing diabetes (adjusted HR = 0.88 [95% CI 0.83–0.93] for daily versus non-consumers, p < 0.001, corresponding to a 0.2% difference in 5-y absolute risk), with a clear dose–response relationship. Among those with baseline diabetes, higher fruit consumption was associated with lower risks of all-cause mortality (adjusted HR = 0.83 [95% CI 0.74–0.93] per 100 g/d) and microvascular (0.72 [0.61–0.87]) and macrovascular (0.87 [0.82–0.93]) complications (p < 0.001), with similar HRs in individuals with previously diagnosed and screen-detected diabetes; estimated differences in 5-y absolute risk between daily and non-consumers were 1.9%, 1.1%, and 5.4%, respectively. The main limitation of this study was that, owing to its observational nature, we could not fully exclude the effects of residual confounding.
