Nevertheless,our study also has limitations.First,the possibility 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.Second,
several studies with few death numbers of liver cancer were
included, which might increase the sampling error