ירחון החברה הישראלית לפדיאטריה אמבולטורית (חיפ"א) גיליון 2018-1

9 Statistics The pitfalls of large international studies - a paradigm common to Israel In the New England Journal of Medicine, Yusuf and Wittes 1 call into question how we interpret results from large multinational studies. Such studies are needed in order to include a larger sample for an uncommon disease, pool resources for multinational regulation or other statistical reasons. We often presume that the multinational nature of such studies increase their reliability due to their population diversity and reproducibility in different locations. But is everything as it seems? Can non –measured variations in different societies, such as nutrition, reporting differences and their overall differences in culture of disease and medicine alter reporting and therefore, subsequent study results? In at least one trial, they describe how variations of results in different countries lead to FDA changes. Alternatively, in regard to less common diseases, small populations in individual countries may represent statistical flukes rather than true disease reflection. The perfect clinical trial is hard to come by. After taking into account population variances, social differences, as well as more subtle changes through viral mutations, epigenetics and environmental changes, truly reproducible and objective results are hard to come by. In Israel, we often need to rely on studies performed elsewhere and extrapolate that data for our patient population. In addition, our patients come from a myriad of genetic backgrounds with mini-populations of various ethnicities, each comprising of only a microcosm of our patient pool. We can not throw up our hands in a fury of inconclusive studies, yet, when we read the medical literature, we should always be aware of the limits of our true understanding of the statistics we so often rely on. Bottom line: Even large multinational studies have inherent bias. 1. Yusuf S and Wittes J. Interpreting Geographic Variations in Results of Randomized, Controlled Trials. N Engl J Med 2016; 375:2263-2271 December 8, 2016 DOI:10.1056/NEJMra1510065