How the risk of liver cancer changes after alcohol cessation: A review and meta-analysis of the current literature

Background: It is well established that drinking alcohol raises the risk of liver cancer (hepatocellular carcinoma). However, it has not been sufficiently established as to whether or not drinking cessation subsequently reduces the risk of liver cancer and if it does reduce the risk how long it takes for this heightened risk to fall to that of never drinkers. This question is important for effective policy design and evaluation, to establish causality and for motivational treatments.Methods: A systematic review and meta-analysis using the current available evidence and a specific form of Generalised Least Squares is performed to assess how the risk of liver cancer changes with time for former drinkers.Results: Four studies are found to have quantified the effect of drinking cessation on the risk of liver cancer. The meta-analysis suggests that the risk of liver cancer does indeed fall after cessation by 6-7% a year, but there remains a large uncertainty around this estimate both statistically and in its interpretation. As an illustration it is estimated that a time period of 23 years is required after drinking cessation, with a correspondingly large 95% confidence interval of 14 to 70 years, for the risk of liver cancer to be equal to that of never drinkers.Conclusion: This is a relatively under researched area and this is reflected in the uncertainty of the findings. It is our view that it is not possible to extrapolate the results found here to the general population. Too few studies have addressed this question and of the studies that have, all have significant limitations. The key issue amongst the relevant studies is that it appears that current drinkers, abstainers and former drinkers are not composed of, or effectively adjusted to be, similar populations making inferences about risk changes impossible. This is a very difficult area to study effectively, but it is an important topic. More work is required to reduce both statistical uncertainty and tackle the various study limitations this paper highlights and until this is done, the current result should be considered preliminary.