An interview with Miguel Hernán, Harvard University Kolokotrones Professor of Biostatistics and Epidemiology
On March 15, 2023, the United States Center for Medicare and Medicaid Services (CMS) issued draft guidance on the implementation of the Drug Price Negotiation Program, established under the Inflation Reduction Act (IRA) of 2022. This program includes a definition of maximum fair price, based on key elements including comparative effectiveness data and information about a drug’s impact on specific Medicare populations. To inform those topics, the CMS intends to review existing literature and real-world evidence, conduct internal analytics, and consult subject matter and clinical experts.Rigorous methodologies employed on fit-for-purpose real-world data, such as target trial emulation (TTE), produce comprehensive and robust evidence packages to support the drug price negotiation process. The TTE framework is used to estimate comparative treatment effects by emulating a randomized controlled trial, using available observational data and appropriate methodology to avoid common sources of bias and study design flaws.
Kolokotrones Professor of Biostatistics and Epidemiology Miguel Hernán of Harvard University describes why TTE is best suited to meet the new standards of CMS, and how manufacturers can prepare. He also describes the role of TTE in value assessments during later life cycle stages.
Given the IRA’s requirements for comparative effectiveness evidence from drug manufacturers for their drug and therapeutic alternatives, broadly speaking, do you see target trial emulation (TTE) as a part of the solution to fill evidence gaps in the US?
Certainly. The findings from randomized trials can be used to answer many questions about comparative effectiveness, but it is just not possible to conduct enough randomized target trials to answer every single question about comparative effectiveness. There will be questions about multiple drug indications, treatment strategies, patient characteristics, clinical outcomes, and periods of follow-up. We cannot reasonably expect that randomized trials can be used to assess the effectiveness of every treatment strategy in every subgroup of clinical interest for every outcome over several years of use. Rather, we will need to use observational real-world data to explicitly emulate the target trials that decision makers need.
How can TTE support the IRA’s requirement for drug manufacturers to generate evidence on the comparative effectiveness of their drug and its therapeutic alternatives in specific populations, such as individuals with disabilities, the elderly, the terminally ill, children, and other patient populations who may not have been available in pivotal trial evidence?
This is a perfect example of how evidence from both randomized trials and their observational emulations can complement each other. Randomized trials often impose strict eligibility criteria that limit the transportability of their results to other populations, but they yield reliable effect estimates in the studied populations. These effect estimates can be used as benchmarks for target trial emulations using real-world data. The idea is that we first emulate a target trial in exactly the same patient population that was included in the original or pivotal trial. Then, after confirming that the real-world data are sufficiently rich to replicate those results, we extend the inferences to patient populations that were not well represented in the pivotal trial. Benchmarking, combined with the typically much larger sample size of real-world observational databases, may increase our confidence in the extension of causal inferences to understudied populations.
The IRA will require manufacturers to develop strategies that span the product life cycle, so called “Life Cycle Value Assessment.” This may be similar to the Swiss setting where the value of approved drugs is re-evaluated every 3 years. Value and evidence requirements have historically been applied at product launch when therapeutic effectiveness may remain uncertain. What role does TTE have in value assessment during later life-cycle stages, how might this be incorporated into manufacturers’ evidence-generation programs to support the entire product life cycle? Are “living trials” something we should be thinking about?
The “Life Cycle Value Assessment” will better align drug evaluation in the United States with that in other countries. Not only may the effects of a drug not be fully known at the time of FDA approval, but approval also does not guarantee that the drug is, and will continue to be, effective in comparison with other (existing and future) drugs and under changing practice patterns. Therefore, it only makes sense that payers are interested in periodic evaluations of a drug’s comparative value. It would be, however, unrealistic to expect that such evaluations can rely exclusively on classical randomized trials, which cannot provide information in real time. Rather, in an ideal world, many drugs would be administered as part of systemwide randomized trials to evaluate comparative effectiveness and safety. If equipoise exists, health systems would randomize drug assignment at the point of care until it can be concluded that a particular treatment is superior to the others. In practice, however, point-of-care randomized trials prove challenging for many health systems and regulatory environments. They are also not feasible when physician’s preferences or consensus favor one particular drug. Therefore, explicit emulation of target trials will continue to play an important role for comparative effectiveness during a drug’s life cycle.
What is your view on the current state of real-world data in the US? Are there sufficient quality data that are fit-for-purpose to support TTE applications?
Yes, as many examples show, but there are also some challenges. Prominent among them is the fragmentation of the data. Take commercial databases of administrative insurance claims, which can be purchased from several sources. A person changing jobs can transition across databases every couple of years and data users may not be able to track the person. This setting constraints the time horizon of the clinical questions that can be asked and threatens the completeness of the data on medical history. On the bright side, Medicare data can be used to track individuals after age 65 and many states are building all-payer claims databases. Another challenge that is being gradually overcome is the traditional exclusive reliance on insurance claims databases. Because claims often do not contain sufficiently rich clinical information on confounders, these databases are not a good fit for many causal inference applications. Fortunately, an increasing number of integrated health systems are making their databases available for research.
What kinds of investments in the development of real-world evidence will drug manufacturers need to consider with respect to meeting the IRA price negotiation evidence requirements?
Drug manufacturers will need to invest in both data expertise and causal inference expertise. Regardless of whether their research ends up being conducted in-house or contracted out, manufacturers need teams of scientists who know how to ask causal questions in terms of target trials, who can evaluate the adequacy of the methodology used to emulate target trials, and who can identify real-world data sources that are fit for purpose when a new emulation is required. I imagine these causal inference experts at industry, together with their counterparts at government agencies and medical journals, will greatly elevate the quality of observational research by conducting and publishing sound emulations of target trials.
Want to learn more? Contact us!
Miguel Hernán uses health data and causal inference methods to learn what works. As Director of the CAUSALab at Harvard, he and his collaborators repurpose real world data into scientific evidence for the prevention and treatment of infectious diseases, cancer, cardiovascular disease, and mental illness. This work has shaped health policy and research methodology worldwide. As the Kolokotrones Professor of Biostatistics and Epidemiology, he teaches at the Harvard T.H. Chan School of Public Health, where he has mentored dozens of trainees and students, and at the Harvard-MIT Division of Health Sciences and Technology. His free online course “Causal Diagrams” and book “Causal Inference: What If,” co-authored with James Robins, are widely used for the training of researchers.
Evie Merinopoulou is a Health Economist and Real-World Data Scientist working on applications of Real-World Evidence in support of regulatory and HTA decision making. Ms. Merinopoulou has worked in the healthcare consulting industry for over 10 years. She currently serves as a Director and Research Principal at Cytel, based in London, UK. She leads the design and execution of observational research projects using global real-world data. Ms. Merinopoulou particularly focuses on projects involving real-world synthetic control arms, quantitative bias analysis, head-to-head comparisons using target trial emulation, and transportability analysis.
Read more from Perspectives on Enquiry & Evidence:
Sorry no results please clear the filters and try again