How can results across different studies or even within studies (e.g., between men and women) be compared? Risk-ratio estimates cannot simply be pooled because of differences in the estimated rates of alcoholism. Estimates of the prevalence of alcoholism are highly variable, depending on how alcoholism is defined. Thus, Kendler and colleagues (1992) reported a lifetime prevalence of alcohol dependence defined by DSM–III–R of 8 to 10 percent in women, whereas Hrubec and Omenn (1981) reported a prevalence in men of alcoholism treated through the VA system of only 2.6 to 3.1 percent. Estimates of risk ratios for relatives of alcoholics, which express risk to relatives as a ratio of the risk in the general population, are similarly variable.
