Estimation of sample size using different computer programmes for biomedical investigations

Abstract

Estimation of sample size is important in biomedical research studies. The use of statistical software requires a thorough study of potential drawbacks of these tools, especially when different program-mes may estimate different sample sizes when given the same basic data. Objective: To present sample size estimations with six different softwares in the study of means, proportions, cohorts and case-control designs. Results: Significant differences in sample size estimates were observed using the same basic data. When studying means EpiDat suggests half the sample size estimations performed by other softwares. In proportions Medcalc output is 58% of the larger estimation. In cohort or case-control studies with groups of the same size, estimations are not all equal though similar depending on hypothesis test applied. In a case-control study with 4 controls per case there is a striking difference between EpiInfo and EpiDat softwares, but Power provides an explanation establishing the need to consider the hypothesis test to be used. Conclusions: Valuable support is to be expected when estimating sample size in different scenarios by using appropriate softwares. However, the investigator should be aware of the biostatistical concepts underlying these estimations in order to take better decisions in computer assisted estimations. It is recommended to become familiar with one specific software.