The founder and president of the Institute for Clinical and Economic Review responds to the commentary on heterogeneity in value assessment.
Am J Manag Care. 2019;25(11):542-543To assert that value assessment is at fault for ignoring heterogeneity in relative effectiveness, or for minimizing the importance of subgroups (the relationship between the heterogeneity and subgroups being critical but often obscured), is a bit like finding a man building a house out of the wood he can find or borrow from neighbors and criticizing him for not using bricks that no one will sell him. He has a need for shelter; he does the best he can with the resources he can get; and he would love to have bricks, but powers beyond his control make that impossible.
Let’s start with the goal of value assessment. What are we trying to build? Is the aim to provide a tool to help inform the clinical care of individual patients? Value assessment can indeed be oriented to serve the interests of enhanced shared decision making for individual patients.1 Evidence-based tools can help frame the many different elements of clinical decisions that are important to patients and provide summaries of evidence from population averages or, ideally, from results for patients with similar clinical characteristics. What most distinguishes this form of patient-targeted value assessment is that it keeps all the various elements of risks, benefits, and other elements of value disaggregated so patients can place their own unique “weights” on them and add them up or otherwise consider them in some quasi or formally quantitative process.
This is value assessment in service of what I would call individual heterogeneity—the variation among individual patients that a skilled clinician can illuminate and apply to tailor the care for a patient in their best interest. Individual patients will have unique clinical, emotional, social, and other characteristics that providers should always consider to help select the “best” drug or other treatment option, a critical goal of good medical care.
But there is a second kind of heterogeneity that can be called population heterogeneity, and this is the home territory of the value assessment performed by health technology assessment (HTA) agencies and research groups around the world. The goal of HTA here is not to inform individual clinical decisions but to inform the decisions that are taken at the population level: coverage and pricing. The heterogeneity that matters most in these decisions reflects variation in outcomes at a higher level than the individual and has two forms: one that is knowable in advance of treatment and one that is unknowable to the patient and clinician because its causes are unknown to all. In the latter case, evidence may show that patient outcomes appear something like a bell curve, with some patients receiving “average” benefits and harms, whereas others experience better or worse outcomes. The key feature of unknowable population heterogeneity is that there are no signposts, biomarkers, or key clinical indicators that can helpfully predict whether a specific patient will have average outcomes. Although it is still helpful to understand the range of outcomes for different patients, unknowable population heterogeneity leaves patients, clinicians, and policy makers largely reliant on population averages.
However, sometimes evidence can provide a guide to help identify when patients can be expected to experience relatively better or worse outcomes. And here lies the connection to subgroup analysis. Formal subgroup analysis is the most powerful way for HTA to identify how the risks and benefits of treatment may vary systematically within a larger population.2 I would argue that it is misleading to claim that HTA has been slow or recalcitrant in recognizing, seeking, and applying subgroup information to create precise value assessments at the population level. Seeking subgroups for which a drug might be most effective and cost-effective is a vigorous part of HTA. As one example, at the National Institute for Health and Care Excellence in the United Kingdom, this effort leads the agency to designate positive funding decisions for subgroups for many drugs that would otherwise fail a general test of cost-effectiveness across the entire labeled population.3
For my HTA agency, the Institute for Clinical and Economic Review (ICER), the hunt for subgroups has led us to create stratified cost-effectiveness findings for different patient subgroups in reviews for treatments such as proprotein convertase subtilisin/kexin type 9 drugs for hypercholesterolemia, Spinraza and Zolgensma treatments for spinal muscular atrophy, and preventive treatments for migraine.4 In many cases, we have found strikingly different cost-effectiveness data across subgroups. We are eager for more data and omnivorous in our appetite for evidence from various sources. In the US healthcare system, where there is a single price for a drug regardless of its use, we pursue the use of subgroups to the point of calculating separate value-based price benchmarks for each subpopulation, even though we must also recognize the reality of the US system and do a weighted calculation across all subgroups to calculate a single population value-based price.
Why have we not been able to identify and model subpopulations within every review we have done? The simple answer is that data are often not available on both the treatment of interest and its main comparator that can be used to assess subgroup effects, but there is a deeper and more complex issue in play. Drug makers face very conflicting incentives in helping to identify subpopulations who might have more—or less—benefit from their drug. On one hand, a special niche within a broader label in which patients can be shown to have superior outcomes might help the drug maker compete for market share against other drugs. Conversely, slicing the data from the overall labeled population into subpopulations might show diminished benefit in large swaths of the patient population, helping clinicians and payers limit use of the drug to the narrower subgroups that benefit most within the broader label. For that reason, and perhaps others, even though we at ICER continue to make routine requests to drug makers at the initiation of every review for stratified or patient-level data, we continue to routinely have this request unfulfilled. Whether we, like the man building his house who would love to be able to do so with bricks, should be criticized for lacking bricks and doing the best we can with the wood we can cobble together, is dubious.
Ultimately, when value assessment seeks to build an analysis to inform population-level policies, it will never be able to consider the granular details that make each patient unique. But the middle ground of subgroup analysis is ripe for common efforts and an area in which we can hope to see progress in our ability to generate, analyze, and apply data in service of more sophisticated coverage and pricing policies.Author Affiliation: Institute for Clinical and Economic Review, Boston, MA.REFERENCES
1. Patient perspective value framework. Milken Institute FasterCures website. fastercures.org/programs/patients-count/patient-perspective-value-framework. Accessed October 14, 2019.
2. Espinoza MA, Manca A, Claxton K, Sculpher MJ. The value of heterogeneity for cost-effectiveness subgroup analysis: conceptual framework and application. Med Decis Making. 2014;34(8):951-964. doi: 10.1177/0272989X14538705.
3. Barham L. NICE numbers at 20. PharmaPhorum website. pharmaphorum.com/views-analysis-market-access/nice-numbers-20. Published July 18, 2019. Accessed October 14, 2019.
4. Topics. Institute for Clinical and Economic Review website. icer-review.org/topics. Accessed October 14, 2019.