The current study has several limitations. (i) The sample consisted of only males, and the findings may not be generalizable to females. In general, findings from prediction-based studies will be more useful when they are generalizable to different strata of the population from which the sample is drawn. Therefore, we suggest that future predictive studies aim for samples from both genders. (ii) The outcome groups (AUD and CTL) were not matched for age, as the age difference was statistically significant. Although age was not significantly associated with the key features, either in each group or in the total sample, the results would have been more credible if the groups were matched for age. Therefore, future studies should also aim to apply predictive models on age-matched groups, to avoid confounding the results. (iii) Finally, the sample size for a predictive model was rather small, although the random forest algorithm is known to handle such situations more effectively than other machine learning models. Therefore, similar studies with larger sample sizes are needed to confirm the findings of the present study. A larger
