Nermina Ferizovic | Health Economics Expert

For the next blog post in the series where I review papers published as part of the “Economics of Gene and Cell Therapies” in PharmacoEconomics, I will review: “Surrogacy and the Valuation of ATMPs: Taking Our Place in the Evidence Generation/Assessment Continuum” from Gladwell et al., 2024 found at the following link: https://link.springer.com/article/10.1007/s40273-023-01334-y.

The premise of the study is that ‘commonly used value assessment frameworks struggle to characterize, quantify, and address uncertainties that result from advanced therapy medicinal products (ATMPs)’ typical evidence base’ with the authors focussing on how surrogate endpoints are addressed in ATMP health technology assessment (HTA) submissions. The authors note that frameworks and methods used to assess purported surrogate relationships were largely created in the 1980s-90s before development of ATMPs and are not suitable for ATMPs. The authors carefully describe the contextual basis of their argument and describe current methods used in HTA when robust multiple clinical trial data are not available:

• ‘Empiric’ methods: ignoring the intermediate response endpoint entirely
• ‘Naïve’ methods: ignoring the consistent warning that the available relationship between surrogate and final endpoint is likely technology specific

Managed access agreements are referenced as used alongside the current ‘empiric’ and ‘naïve’ methods to help reduce uncertainty. 

The authors propose what they term a ‘mechanistic in silico model’ (essentially ordinary differential equations) to assess the relationship between surrogate endpoints and final endpoints in HTA of ATMPs. This proposal consists of the following ideas:

• At the point of submission to HTA bodies, there is already a wealth of data available for manufacturers and HTA bodies, particularly early phase data and data used to inform regulatory submissions
• Accounting for biological, pathophysiological, and pharmacological processes using longitudinal data to assess the potential impact of the therapy
• As longer-term data and non-trial data (including data captured through managed access agreements) become available, these can be used to assess the ‘mechanistic in silico models’ generated from earlier phase trials and regulatory data
  • The authors term this the ‘evidence generation/assessment continuum’ as data from early phases to managed access agreements could be used to help develop and refine these ‘mechanistic in silico models’ 
• Managed access agreements would be instrumental for these ‘mechanistic in silico models’ to function appropriately

In summary, the authors propose the consistent use of mathematical models, utilising data from early phases to post launch while accounting for biological mechanisms, with iterative refinement to validate surrogate relationships in HTA of ATMPs. 

It is important to note that this paper is an opinion piece rather than a methodological paper and despite proposing an interesting solution based on pre-existing precedence, this proposal needs further development. 

Key areas for further research and development include understanding best practise on how these models should be specified and whether recommendations for model specification would differ for different ATMP types. Were this a suitable approach for assessing surrogacy, recommendations would need to be made to manufacturers for different types of ATMPs to account for their heterogeneity. The opinion piece focusses on CAR-T therapies to make the argument, with these therapies being the most common ATMP type reaching the market. Would this approach, which is heavily reliant on data availability on the biological, pathophysiological, and pharmacological processes of therapies, also be appropriate for less common ATMPs and disease areas? 

Further, there may still be limitations with the development of these models as a method of assessing surrogacy in an area synonymous with limited data even with the combined use of early phase data through to data collected as part of managed access agreements. In a situation where data captured as part of managed access agreements is not captured accurately (such as in rare diseases) how well would these ‘mechanistic in silico models’ perform? In this instance, there is a risk that this method could increase the workload for manufacturers and HTA bodies alike but with limited gain. 

Overall, this is an interesting opinion piece which lays the foundational contextual argument for an alternative approach to assessing surrogacy endpoints in HTA of ATMPs. However, it requires further consideration before a guiding framework which takes advantage of the evidence generation/assessment continuum could be developed.