To the best of our knowledge, this is the first study using the MR method to investigate the potential causal association between domain-specific sedentary behaviors and the risk of type 2 diabetes. Our study clarified the directionality of the association between television watching and type 2 diabetes, supporting that prolonged television watching time increases the risk of type 2 diabetes. However, no evidence was observed for the association of computer use and driving behavior with the risk of type 2 diabetes. Furthermore, our study suggested that the association between television watching and the risk of type 2 diabetes may be mainly mediated by BMI or WHR and partially explained by educational attainment.
Our findings are in line with the results of previous studies, which also suggested that prolonged time spent on television watching was a risk factor of type 2 diabetes. For instance, an American prospective cohort study with 68,497 women, 6 years of fellow-up and adjustment for other covariates indicated that each 2 hours per day increment in television watching was associated with a 14% increased risk for type 2 diabetes. In addition, another European study with 23,855 individuals and 7.8 years of follow-up found that the amount of time spent on television watching was positively associated with incident diabetes, but this association was largely attenuated after adjusting for anthropometric measures. Furthermore, one cross-sectional study conducted among older people suggested that excessive television watching was associated with higher risk of type 2 diabetes (OR = 1.56, 95% CI = 1.10–2.21), but this was not the case for other domain-specific sedentary behaviors including computer use and transport.
Actually, television watching is the most commonly used proxy for sedentary behavior in observational studies, as it is easier to recall and shows considerable accuracy.[37, 38] In addition, television is almost solely performed at home, which is more modifiable by intervention.[39, 40] Based on our findings and previous literature, several hypotheses have been put forward as to why prolonged television watching seems to increase the risk of type 2 diabetes. A commonly acknowledged mechanism is that excessive television watching increases obesity, a major risk factor for type 2 diabetes. Specifically, long-term leisure television watching is related to a lower expenditure of energy, as well as a higher energy intake and a relatively unhealthy dietary pattern, leading to obesity and weight gain, which in turn increases the risk of type 2 diabetes.[41–43] Our study found that the association between television watching and the risk of type 2 diabetes attenuated larger when adjusting WHR than BMI, which suggested that abdominal obesity might be a more proper mechanistic element for this association. Second, since physical activity was an established protective factor for type 2 diabetes, several studies suggested that the association between television watching and type 2 diabetes may be explained by lack of physical activity.[44, 45] However, some other studies indicated poor correlation between television watching and physical activity, supporting that television watching was associated with type 2 diabetes independently of physical activity.[46, 47] Consistent with this finding, the association was little attenuated after adjustment for physical activity in our study. Third, the attenuation of the association between television watching and type 2 diabetes after adjustment for education attainment may represent another explanatory mechanism. In line with this, previous study indicated that higher educational attainment was negatively associated with incident diabetes.
Previous studies have reported null association between computer use and the risk of type 2 diabetes.[36, 49] Energy expenditure for computer use might be greater than that for television watching. Interestingly, negative association between computer use and type 2 diabetes was observed in our study though not significant. This potential association may be possibly explained by the confounding bias of television watching and educational attainment. First, in the original UK Biobank study, from which we obtained summary dataset, television watching was inversely correlated with computer use. Second, significant association between computer use and educational attainment was observed in our study, and education was reported negatively related to incident diabetes in previous study. Further studies are warranted to explore the nature association between computer use and type 2 diabetes.
The major strength of present study was the MR design, which could reduce the bias of confounding and reverse causation. In addition, bidirectional MR analysis and multivariable MR analysis was used to further minimize the bias of reverse causation and confounding, respectively. Mediation analysis was performed to investigate the proportion mediated via specific potential mediator. Another strength was the large sample sizes for the both data sources used for genetic associations with sedentary behaviors and the type 2 diabetes. A further strength was the large number of SNPs identified as IVs for television watching. This two together resulted in high precision of the results in our study. However, the estimates calculated by MR-Egger analysis were imprecise and must be interpreted with caution.
There were potential limitations to current study. First, the pleiotropic effect could not be completely ruled out, which was an established limitation to the MR analysis. However, symmetrical graphics were observed in the funnel plot, and robust results were observed in the analyses using multiple methods. In addition, multivariable MR analysis was used to adjust for other traits. Second, dietary energy and diet quality were not adjusted in our multivariable MR analysis, since the genetic association with dietary factors was not available. Dietary factor was a potential mediator for the association between television watching and type 2 diabetes, thus further studies are needed to determine the effect of dietary factors on this association. Third, the number of SNPs identified as IVs for computer use and driving behavior were relevantly small, which may lead to low power of the analyses. Thus, the corresponding results might be imprecise and must be interpreted with caution. Forth, because our study was limited to individuals of European ancestry, it was unclear whether our results could be generalized to other ancestral populations. Fifth, individual-level data were not available, which precluded us from the sex-specific analyses. Finally, we only clarified the association between sedentary behaviors with the risk of type 2 diabetes from a genetic perspective, further investigations are warranted.