Abstract: Our study was designed in 1994 when no CONSORT guidelines existed for cluster randomised trials.Thus we had no opportunity to refer to CONSORT at that time; however, the study fulfilled almost all requirements.The randomisation scheme depends to a large extent on local circumstances.Randomisation was done by a statistician independent of our study and the allocation was done randomly using restricted block randomisation, ensuring almost equal distribution of clusters into the study groups.The number of clusters must again reflect the local circumstances.A small number of clusters was used in our study to minimise contamination in view of the high population density in Kerala; a greater number of smaller clusters would have resulted in closer geographic proximity.Two previous interim analyses ignored the cluster design so the differences in risk factor prevalence were significant. 1,2The final analysis took into account the cluster design, and the study groups were well balanced for age, sex, religion, socioeconomic status, pan-tobacco chewing, and alcohol use, but the proportion of trials using more rigorous methods and perhaps aided by sophisticated adjunctive techniques.