There is no single formula: Making early HTA useful under uncertainty

If you have been asked what a health technology is worth before it has a final design, a trial result or a price, you are in luck: early health technology assessment (early HTA) has been developed to address this question. Early HTA has been defined as HTA conducted to inform decisions on the further development, research or investment of a concept or emerging technology [1]. Knowing what it can and cannot tell you is the first step towards using it well.

Traditional HTA mainly informs reimbursement once evidence has been accumulated. Early HTA is used while technology and investment decisions can still change, and its main users are innovators and funders [1]. Its methods (headroom analysis, early decision-analytic modelling and value-of-information (VOI) analysis) resemble those of traditional HTA but serve research and development (R&D) prioritization rather than coverage decisions [1].

Specifications, design and market strategy keep changing during development, so early HTA models are updated as evidence emerges. Results therefore need careful reading. Wiwatpanit et al. caution that early value-for-money conclusions are preliminary, highly uncertain and may be biased towards favourable outcomes [1]. They attribute this to limited evidence, such as preliminary data and assumptions or estimates supplied by innovators, and to value that may change as the technology develops [1]. Early HTA is not intended to inform coverage programmes, and economic modelling cannot substitute for clinical studies [1].

What this means for innovators developing technologies for public use

For innovators and funders, early HTA can clarify potential value, define target product profiles (TPPs), identify evidence gaps and examine plausible development scenarios [1].

A Thai early HTA of tongue swab (TS) testing for pulmonary tuberculosis illustrates several of these [2]. Clinical trials were still under way, so TS performance had to be assumed. Mou et al. plan to update the model once trial data become available, and no update has yet been made [2]. The base case favoured TS with real-time polymerase chain reaction (TS RTPCR), with an incremental cost-effectiveness ratio (ICER) of 10,401.33 Baht per quality-adjusted life year (QALY) against the standard practice of using acid-fast bacillus smear microscopy and sputum Xpert (AFB&Xpert) in people aged five or older, below the 160,000 Baht threshold. The study also reported three analyses aimed at development decisions:

• TPP. When one parameter is varied at a time, TS with loop-mediated isothermal amplification (TS LAMP) would need its price to fall from 860 to 177.08 Baht to surpass sputum RTPCR. 
• Scenario analysis. Two scenarios changed the results for sputum-based tests: that 10% of patients cannot produce usable sputum, and that smear microscopy requires three samples [2].
• VOI. In people aged five or older, population expected value of perfect information (EVPI) was 236,713.39 Baht at the Thai threshold. The four parameters with the highest expected value of partial perfect information (EVPPI) were the specificity of sputum RTPCR, the sensitivity and specificity of TS RTPCR, and the mortality hazard ratio for active pulmonary tuberculosis without early treatment compared with early treatment [2].

Implications for LMICs

A recent LMIC-focused example is an early HTA model of a point-of-care test for neonatal sepsis, built to the WHO target product profile [3]. Because practice and costs differ across countries, Fedorov et al. built one global decision model. They drew parameters from multi-country studies where possible, filled gaps through expert elicitation and tested uncertainty probabilistically. They also report the minimum test sensitivity and specificity needed to avoid net harm and provide an Excel version that countries can populate with local data [3]. The paper does not report such country adaptations, so practical questions remain. Which components are feasible with local data? How should assumptions be documented when evidence is thin? Can identifying the uncertainties that matter most help direct scarce research resources?

Practical considerations when conducting early HTA

• If performance data are not yet available, report target values alongside ICERs. 
• If several strategies compete, use VOI to identify which parameters justify further research. 
• Watch out for reading early results as reimbursement advice [1].
• Start by building comparators from the local care pathway, ideally before the first clinical trial, and plan to update the model as evidence accrues [1].

When you read the full papers, pay attention to Table 1 in the perspective [1], full text: https://doi.org/10.1038/s44401-026-00102-2 and Table 3 in the TS study [2], full text: https://doi.org/10.1016/j.lansea.2025.100533. 

References

1. Wiwatpanit T, Teerawattananon Y, Wang Y. Introduction of early health technology assessment for innovators, clinicians, and funders. npj Health Syst. 2026;3:50. doi:10.1038/s44401-026-00102-2
2. Mou L, Wiwatpanit T, Piriyapol A, et al. Early health technology assessment of tongue swab for non-sputum based pulmonary tuberculosis diagnosis in Thailand. Lancet Reg Health Southeast Asia. 2025;33:100533. doi:10.1016/j.lansea.2025.100533
3. Fedorov MS, Bansal A, Bradley G, et al. Cost-effectiveness of a target point-of-care triage test for neonatal sepsis in low- and middle-income countries. CGD Working Paper 744. Washington, DC: Center for Global Development; 2026. https://www.cgdev.org/publication/cost-effectiveness-target-point-care-triage-test-neonatal-sepsis-low-and-middle-income