The sample size was estimated so as to measure 20% MPD prevalence with a 2.0 pp margin of error and associations with a prevalence ratio greater than 1.8, a 95% confidence level, 80% statistical power and a 14:1 unexposed: exposed ratio with regard to pesticide poisoning. A further 10% was added to account for missing data and refusals as well as 15% for confounding factors. The sample size required was about 2600 workers. Based on an estimated of 2.5 workers per farm, 1040 farms needed to be studied.