It is well recognized in the medical community that a medical intervention (e.g. a drug or a device) may have different effects on different people, e.g. some patients may benefit from an intervention whereas others do not or even are harmed by it. Application of a medical intervention to patients who do not benefit from it not only puts unnecessary financial burden on individual patients and the society, but also results in suboptimal health outcomes for these patients. Patients may miss the best treatment window because of the use of an intervention that has a null effect for them. On the other hand, as drug development has become increasingly expensive with high failure rates, there is an urgent need to identify patients who benefit from a new drug even though others may not. A major gap exists in the methodology to extract knowledge on who will or will not benefit from a medical intervention. Moreover, there is no software specifically designed to serve this purpose on a routine basis by analyzing medical data. Our long-term objective is to produce generic software specifically designed for the analysis of treatment effect in different subpopulations, which can be applied on a routine basis using data from randomized or observational studies.
The specific aims of this proposal are to: 1) develop a set of statistical methods to examine whether or not there exists subpopulations who benefit from or are harmed by a medical intervention, and if yes how to identify the subpopulations using patient characteristics and estimate the treatment effect in these subpopulations; 2) apply the developed methods to data from both randomized trials and a real-world electronic medical records (EMRs) database to understand the effect of a medical device called implantable cardioverter defibrillator (ICD) in reducing mortality; and 3) implement the methods in a popular statistical software that is publicly available. The project has impact. First, the methodological research in this proposal addresses key AHQR priority areas of focus on accelerating implementation of Patient-Centered Outcomes Research and make health care safer. As a general statistical methodology, the methods to be developed can be applied to essentially any diseases/conditions to study heterogeneity in treatment effect (HTE) including treatment benefit and harm for patient-centered outcomes. Second, knowledge obtained by our method generates more informative and relevant information to facilitate decision-making for patients/clinicians and guideline-making for policymakers. Third, this project will provide critical evidence towards improvement of ICD guidelines that account for the heterogeneity in response to ICDs.
A major gap exists in the methodology to extract reliable, informative and easy-to-understand knowledge from medical data on who will or will not benefit from a medical intervention to facilitate decision- and policy-making. The objective of this projec to develop a suite of new statistical methods to study treatment effect in different groups of people, apply it to a well-recognized medical problem in using ICD to reduce sudden cardiac death, and produce and disseminate software to implement the developed statistical methods. The successful completion of this project will generate analysis tools that can be applied on a routine basis and evidence towards the goal of guideline adjustment to optimize the use of ICDs.
|Kundi, Harun; Valsdottir, Linda R; Popma, Jeffrey J et al. (2018) Impact of a Claims-Based Frailty Indicator on the Prediction of Long-Term Mortality After Transcatheter Aortic Valve Replacement in Medicare Beneficiaries. Circ Cardiovasc Qual Outcomes 11:e005048|