OBJECTIVES: In advance of a new medication being introduced to market, it is important to understand contemporary patterns of care and associated clinical outcomes, in order to document current treatment gaps the new asset will address. Retrospective chart review is a powerful methodology for addressing these questions. A limitation of chart review studies is that the process of extracting comprehensive data from charts can impact sample size feasibility.  In contrast to the hypothesis-testing framework, for a burden of illness study, the role of sample size is to improve precision around estimates of descriptive outcomes, and ensure sufficient representation is achieved amongst subgroups of clinical interest. Determining how many patient charts to include in such a study is not straightforward: common sample size calculations developed for a hypothesis-testing framework have limited applicability in this context.   The objective in this abstract is to develop and present rigorous approaches for sample size calculation for such studies.   METHODS: Proposed methods are described and an illustrative case study is presented of a retrospective chart review of advanced melanoma, and the precision obtained for a sample size of 655 patients across three countries, including ability to consider patient subgroups and compare costs by country. For cost outcomes, the ratio of standard deviation to mean ranged from 0.3-4.5 across countries, with a median of 0.7 and mean of 1.0.   RESULTS:  Using the proposed methods and assuming a ratio of 1.0, to achieve precision of 20% around mean cost would require a sample size of 96; parameters can be adjusted as needed to reflect alternative requirements.  CONCLUSIONS: This methodological study addresses an important knowledge gap, as sample sizes are frequently determined using ad-hoc approaches and/or based only on feasibility considerations. The approach presented here is methodologically rigorous and designed for practical application in real-world retrospective chart review studies.