Nevertheless,our study also has limitations.First,thepossibility that the observed association was due to unmeasured or
residual confounding should be considered,and inadequate control for confounders might bias the results in either direction,
toward exaggeration or underestimation of risk estimates