In this rural elderly Chinese cohort followed for seven years, we found that both lower cognitive function and lower BMI are independently associated with increased mortality risk compared to individuals with average cognitive function and normal weight. In addition, we found that higher cognitive function is associated with lower mortality risk. At the same time, we found no significant difference in mortality risk between overweight/obese participants and those with normal weight.
Previous studies have reported increased mortality risk of individuals with poor cognitive function and have attributed the increased risk to cognitive impairment and dementia [3, 20–23]. When we excluded participants who died within the first year of baseline, poor cognition was no longer significantly associated with increased mortality, suggesting that terminal illnesses may be a cause for poor cognition . However, few studies have separately examined mortality risk in individuals with higher than average cognitive scores. Our result of lower mortality risk in individuals from the top cognitive quintile group compared to those with average cognitive scores suggests a potential protective mechanism of high cognitive function in the absence of disease pathology. This protective effect could result from more years of formal education, a more intellectually stimulating job, life-long pursuit of leisure activities with cognitive benefit, or healthier lifestyles . In our study, individuals with the highest cognitive function also had the highest proportion of schooling, although education itself was not an independent predictor of mortality in the final multivariate model. It is also interesting to observe that the highest cognitive functioning individuals in our cohort had similar levels of medical comorbidities as participants in the other cognitive groups suggesting that the lower mortality risk in the high cognition group is unlikely due to better physical health in this group.
In terms of BMI, our results support previous findings of increased mortality risk for underweight participants as previously reported in a large Chinese study that also included younger study subjects (mean age 52 years) , in white adults , and in non-Hispanic black adults . The increased mortality risk in underweight individuals was thought to be due to increased comorbidity burdens in these individuals. However, our final multivariate models included diabetes, stroke and cognitive function and still found significantly increased mortality hazard for the underweight participants, hence these comorbid conditions are unlikely to account for the increased mortality in this group. The underweight individuals were also suspected to be suffering from terminal diseases that resulted in weight loss. However, when we tested this hypothesis by excluding individuals who died within one year of baseline evaluation, we still observed an increased mortality risk in the underweight individuals. In fact, the survival plot in Figure 2 seems to suggest a steady rate of death in the underweight group across the seven years of follow-up.
While our analyses lend support to a protective effect of cognitive reserve, our results are inconclusive on the effect of a physiologic reserve. Unlike previous studies [6, 25, 26], we did not find an increased mortality risk for those overweight or obese. We also did not find lower mortality among those overweight and obese participants as reported in other studies . Since this cohort has few participants with BMI in the obese range, we anticipated insufficient power to examine the risk associated with obesity. Nevertheless, the survival plot in Figure 2 and the hazard ratio estimate for the overweight group suggest that there are little differences in mortality risk between the overweight and normal weight participants. Thus for physiologic reserves defined by BMI, our results support the maintenance of normal weight in the elderly, but offer no evidence of additional protection from BMI higher than normal weight on mortality risk.
Other predictors that have been consistently shown to be risk factors for mortality include older age, male gender, history of diabetes and cardiovascular diseases . Interestingly, in our cohort education was not associated with mortality after adjusting for cognitive function, although the majority of our cohort has little or no education (62%). The low variation in education levels among our participants may explain the lack of association between education and mortality. Alternatively, the effect of education may be attenuated in the presence of cognitive function in the survival models.
Results on the relationship between APOE ϵ4 genotype and mortality risk have not been consistent as some studies found increased mortality risk in APOEϵ4 carriers [28, 29] while others reported no relationship [30, 31]. The discrepancy in results was sometimes attributed to whether cognitive status was adjusted in the models or to low ϵ4 allele frequencies in the population studied . In our cohort, APOEϵ4 was not a significant predictor of mortality even without adjusting for cognitive function. However, this cohort does have low APOEϵ4 frequency (8.8%)  which may account for the non-significant relationship between APOEϵ4 and mortality.
Our study has a number of strengths. The first is our large cohort size with relatively long follow-up. A comprehensive set of cognitive measures were used in our study ensuring robust results in grouping participants according to their composite cognitive scores. The cohort is also unique as few studies with cognitive and body composition measures have been conducted in rural elderly Chinese.
This study also has important limitations. Our analysis focused on all causes mortality as many of our participants died at home and accurate information on causes of death was not available. It is also possible that other predictors of mortality, especially socio-economic information other than education, were not included in our models.