The heterogeneity in the results of observational studies that investigated the association between alcohol consumption and risk of liver cirrhosis was analysed by means of a meta-analysis that included 15 articles published from 1978 to 1997. Relative risks associated with low levels of alcohol intake (25 g/day) ranged from 1.5 [95% confidence interval (CI): 1.4-1.5] for a linear model fitting the results of the six studies performed in Mediterranean areas, to 3.6 (95% CI 3.1-4.3) for a quadratic model fitting the results of the nine studies performed in other areas. A strong indication of heterogeneity was observed when combining all studies. Quadratic term of alcohol intake, quality of the study and area in which the study was performed explained most of this heterogeneity. Efforts should be made to explain the strong heterogeneity in the trend estimates. Reproducible methods to collect relevant and valid information on alcohol intake should be developed and the role of drinking patterns and viral and nutritional factors in modifying the effect of alcohol on the risk of liver cirrhosis should be investigated.