In my introductory blog post, I noted that genomics might present new challenges for health economics and called for more discussion about appropriate methods in this context. I didn’t anticipate a particularly rapid response, but just a few days after posting I became aware of a new article published in PharmacoEconomics that engaged with many of the issues raised in my introductory post. Titled “Concepts of ‘personalization’ in personalised medicine: Implications for economic evaluations”, this paper reports the results of a workshop which considered where extensions to standard methods might be required in genomics and is a welcome addition to the limited existing literature on this subject. We covered some similar ground in a related paper published in Pharmacogenomics last year, and it is heartening to see that this new paper has reached some similar conclusions and developed a number of these issues further.
Cutting to the chase, the paper reports two key findings from the workshop. The first, that complex model-based economic evaluations are required in personalized medicine (PM) to synthesise a wide variety of data from difference sources, and of varying quality, is a point worth making, but is not necessarily breaking new ground.
Where the paper makes a real contribution to the literature is the second key finding: that two distinct but linked interpretations of the concept of PM exist, each presenting specific challenges to the use of standard economic evaluation methods. These two concepts are termed ‘physiology-based personalization’ and ‘preference-based personalization’, and I believe that making this distinction is beneficial going forward, particularly in terms of planning economic evaluations in PM.
On the physiological side, challenges include the representation of complex care pathways, accounting for spillover effects, data requirements for populating models, and analysis of uncertainty. I would encourage the interested reader to go to the paper because each challenge is discussed in detail. I will, however, highlight one specific issue which has been mostly neglected in PM economic evaluations to date: the impact of test thresholds on cost-effectiveness results. The interpretation of genomic tests is often a challenge because binary outcomes are uncommon. A test may be designed to identify the presence of 15 single-nucleotide polymorphisms (SNPs) that are relevant for a particular disease, but it is as unlikely that all 15 will be present as it is that none of the 15 will be present. Where do you draw the line in this situation, in terms of defining a test-positive result? Once you’ve defined the line, what do you do when the rapidly evolving field of genomics research identifies another 4 relevant SNPs? Does the line move? Remember that you’re probably making this decision whilst relying on a poor quality evidence base (RCTs in genomics are thin on the ground…). Studies which consider potential solutions to this issue would be very welcome.
On the preferences side, challenges include a lack of data on both revealed and stated preferences, issues surrounding risk attitudes and perceptions, and uncertainty concerning the most appropriate evaluative framework. The focus here isn’t solely on patient preferences; clinician preferences are also important, and could have a major impact on test uptake. We’d know this for sure if we had good data on uptake and adherence, but we don’t. Rogowski et al. suggest a number of solutions, and these are well worth considering going forward.
Issues related to risk attitudes and perceptions in PM offer such a wide variety of interesting research angles for health economists that it always surprises me that these topics do not crop up more often at conferences and in journal articles. Perhaps the greater need for cross-discipline collaboration precludes health economists from getting involved, which is a shame. The main issues surround the considerable heterogeneity in both attitudes and perceptions and how to incorporate this into economic evaluations. Returning to our previous example, risk-averse Patient A might consider the presence of 5 SNPs to be an appropriate test cut-off, whereas risk-seeking Patient B might be willing to tolerate the presence of 10 SNPs before accepting some sort of intervention. How do we manage these preferences in economic evaluations of genomic testing?
The authors end by focusing on what I consider to be the key issue in health economics and genomics: what is the appropriate evaluative framework in PM, welfarism or extra-welfarism? I will return to this in subsequent posts, so I won’t linger here (and the authors provide a neat summary anyway). The authors note that the choice of framework is intimately linked to the acceptance (or not, depending on your point of view) that patient preferences are more important in PM than in other areas. I can appreciate that there are opposing arguments here, and there are wider issues too regarding the value judgements that are being made (implicitly) and the appropriate remit of the decision-making bodies using information generated by health economists to make resource allocation decisions. I would encourage all health economists working in this area to engage with this issue in their work to further this debate.
So, to summarise, this paper presents a clear summary of some of the key challenges in this context. I’d encourage you to give it a read as it is very accessible, and a good starting point for health economists looking to delve further into this area. Your thoughts on any of the issues raised in the paper are very welcome in the comments section below.