The four decisions with an expiry date: why HEOR belongs in the IEP before the protocol locks

In 2021, when NICE weighed mogamulizumab for cutaneous T-cell lymphoma, it said something no bridging study could ever settle: "Collecting further data is unlikely to address the clinical uncertainty because of the limitations in the trial design." [1] Read that again. Not "your data is thin," not "come back with longer follow-up." The problem was welded into the trial, and no study run afterwards could reach back in and fix it.

The pivotal trial, MAVORIC, had compared mogamulizumab against vorinostat, a drug "not licensed for use in the UK" that "did not represent NHS standard care." [1] So the committee could not see how the drug performed against the treatment British patients actually receive. The sponsor tried to patch the hole with a post-hoc unanchored indirect treatment comparison against Hospital Episode Statistics; it was rejected as insufficient. On that reasoning NICE refused the drug the Cancer Drugs Fund: no volume of data collection would fix a comparator the trial had never run. [1]

Here is the part a resource-lean team should sit with, because it is worse news than a flat rejection would be. Mogamulizumab did reach NHS patients, recommended for routine use in Sézary syndrome after at least one systemic treatment. [1] It got there through a revised commercial arrangement: a price concession deep enough to bring the cost-effectiveness estimate to £28,233 per QALY, inside the range NICE will accept. [1] The evidence gap itself stayed exactly as open as the committee had described. The company could not make the evidence better, so it made the medicine cheaper. The one input it could not retrofit, the comparator, it bought its way past in a different currency. A company without the margin to hand over that discount has no such door to walk through.

That whole sequence ran on a belief most of the industry shares: that HEOR and access evidence are workstreams you start once approval is in sight, and whatever you missed at trial design you generate later, with a bit of RWE, a bridging study, a managed-access scheme. It is a comforting story. For a specific and important set of decisions, it is also false.

Four decisions in your Phase 2/3 design have an expiry date: which comparator you run against; which patient-reported outcome and utility instrument you field; how long you follow patients; whether you capture resource-use data. All four are yours to change up to the day the protocol locks, and none can be added afterwards. Miss one and you have not made a scheduling slip you can fix by booking the HEOR conversation sooner.

You have poured concrete.

A protocol is wet concrete

Most teams treat integrating HEOR into clinical trial design as a calendar question: get the market-access people in the room a few months earlier. That instinct is understandable, but it mistakes the mechanism. A protocol is wet concrete. While you pour it you can shape it however you like, swapping a comparator, adding an instrument, extending follow-up. Once it cures, the shape is fixed. You are not up against a deadline that can slip a quarter; you are up against a phase change, and it arrives the day the protocol locks.

So "involve HEOR earlier" is true but useless on its own. These four decisions are the ones that expire, so they are the four a lean team must get right on the first pour, because there is no second one. That is what an IEP exists to do: surface those gaps while the design is still soft, before someone external does it for you. We have argued the case for a lean, gap-hunting IEP in the most underutilised tool in drug development and why an IEP is critical when resources are tight.

Take the four in turn.

The comparator you didn't run

The most expensive of the four, and the least retrievable. If your pivotal control arm is not the standard of care in the market where you mean to be reimbursed, no later analysis reliably rebuilds the comparison, because an HTA body cannot judge relative value against a treatment your trial never contained.

Mogamulizumab is the subtlest version: a comparator that sat inside the trial but not inside the clinic, vorinostat set against a UK standard of care it did not resemble, a mismatch no post-hoc analysis could rebuild. [1] Pembrolizumab shows another failure mode; in its appraisal the missing comparator was best supportive care, never in the trial to begin with, so there was nothing to weigh the drug against for the population the committee cared about. [2] Pralsetinib went further still: its pivotal ARROW trial was single-arm, with no comparator at all. [3]

The field number shows how routine this is. Across 543 oncology HTA evaluation reports, 22% needed an indirect treatment comparison to patch a missing comparator, and only 30% of those ITCs were accepted overall. In France, the acceptance rate was zero. [4] The ITC is the standard retrofit for a comparator you didn't run, and most of the time it does not clear the bar. So at design, pick the comparator that reflects the standard of care in the market you intend to be reimbursed in. A single-arm trial only postpones this bill, and the deferred version is the one that usually gets rejected. That said, "standard of care" is itself market-specific: NICE and HAS disagreed on the reimbursement recommendation for the same oncology drugs 62% of the time between 2015 and 2021. [5]

The utility values you never measured

Cost-per-QALY models need preference-based utility data, in practice most often EQ-5D. If your pivotal trial did not field the instrument, your health economists are left mapping across from a clinical PRO, and that mapping is a weak substitute HTA bodies routinely reject.

Pralsetinib again. ARROW collected no EQ-5D, so the company mapped utilities from EORTC QLQ-C30 instead. The verdict: "utilities derived in this way were not considered robust enough to inform decision making." [3]

Not recommended.

The registry shows how avoidable this is, inside a single sponsor's own portfolio. Kite/Gilead's axi-cel pivotal, ZUMA-1, registered no PRO or utility instrument in its pivotal cohorts; the follow-on randomised trial for the same drug, ZUMA-7, registered EQ-5D-5L and EORTC QLQ-C30 from the design stage onward. [6] Same molecule, same company, one trial built to support an HTA submission and one not. Novartis's tisagenlecleucel pivotal, JULIET, registered no comparator and no utility instrument anywhere. [6]

You cannot photograph a scene after the shutter has closed. Nusinersen makes the point outside oncology: its open-label extension, SHINE, added a paediatric quality-of-life measure and a caregiver-burden instrument years after the pivotal trial, ENDEAR, had locked. [7] Neither appears in ENDEAR, and neither is a preference-based utility index of the kind a QALY model needs. Retrofitting costs you twice: late, and often the wrong thing entirely. So field EQ-5D, or whichever utility instrument your target market accepts, in the pivotal protocol rather than the extension study. It is a short questionnaire and it cannot be back-filled. The IMI PREFER programme's central lesson points the same way: preference data has to be planned into the evaluation, not appended to it. [8]

The follow-up you cut short

Cost-effectiveness modelling needs an outcome curve long enough to project beyond the last data point. A pivotal study that reads out early hands the modeller a short curve and a long extrapolation, and committees discount what they cannot yet see.

I will not claim those extrapolations are systematically wrong; whether immature survival data actually misleads is genuinely contested in the methods literature. What is not contested is what HTA bodies do when the data is immature: they hedge. Across NICE cancer recommendations from 2018 to 2022, 56% rested on immature survival data; of those, 39% were routed into the Cancer Drugs Fund and 7% rejected outright. [9] Short follow-up rarely kills an appraisal on the spot. It very often costs you years parked in managed access.

Tisagenlecleucel is the worked example. Its single-arm ELIANA data had immature follow-up, and the technology went into the Cancer Drugs Fund. The retrofit was extraordinary: the team recruited seven external oncologists and ran a formal expert-elicitation protocol just to narrow a five-year survival estimate the trial could not yet provide. [10] A structured survey of seven clinicians, standing in for a number a longer follow-up window would simply have measured. Follow-up duration is a protocol clause; specifying a longer minimum, or pre-planning an extension that carries the same outcome measures forward, costs a fraction of that.

The costs nobody counted

Economic models also need healthcare-resource-use and cost data: admissions, procedures, the real pattern of care. Efficacy trials routinely don't collect it, and unlike the first three gaps this one carries no single dramatic rejection you can point to. What it carries is a slow, expensive reconstruction problem.

Neither JULIET nor ZUMA-1 collected resource use or cost as an outcome. The data arrived years later, out of retrospective claims: NCT06271369 mines US Medicare claims to compare tisagenlecleucel and axi-cel, and NCT05349201 does the same through the French PMSI database for Kymriah and Yescarta. [11] Two reconstructions, in two markets, long after launch, doing work a resource-use case report form could have done alongside each pivotal trial. Collecting economic data inside the trial, running it "piggyback" on the clinical protocol, has been standard methodology since at least 2005 [12], so this is hardly an exotic ask. A resource-use CRF costs pennies against a claims-linkage study commissioned three years post-approval, once for every market you enter.

What "collect it later" actually buys you

Let's be blunt about the escape hatch, because every reader has reached for it. The answer to all four gaps is supposed to be "collect it later," so look at what later delivers.

Pembrolizumab spent three years, 2018 to 2021, in the Cancer Drugs Fund specifically to gather the missing real-world and comparator data. NICE's verdict on that effort: "The additional clinical data collected by Public Health England as part of the Systemic Anti-Cancer Therapy dataset while pembrolizumab was in the Cancer Drugs Fund did not contribute to this review." [2] Three years of it, and the appraisal did not move: the missing comparator was never in the trial design, and no volume of downstream data could conjure one. We have unpicked this drug's wider development story in the pembrolizumab case study.

Tisagenlecleucel's seven-oncologist panel is the other face of "later": not free, not fast, approximating something the trial could have measured directly. You pay for the retrofit in time, in years parked in managed access; in cash, in bespoke studies and elicitation panels; and often you still miss the bar. "Generate it later" is a worse deal than the design clause it replaces, and for three of the four that clause costs almost nothing at the point of pouring.

The case for not gilding your pivotal trial

Now the honest counterargument, a resource-lean biotech should design its pivotal trial to get approved, full stop. Every arm you add, every instrument you field, every extra month of follow-up piles cost, operational complexity and recruitment risk onto the single study that decides whether the company lives, and it piles them on for markets it may never reach. Managed-access schemes exist precisely so missing data can be gathered after approval; one reading of the Cancer Drugs Fund experience is that most cancer technologies that enter managed access are eventually recommended once they exit it a figure of around 84%, roughly 16 of 19, has been cited, though I have not verified it and would not lean on it. Regulator first; access is tomorrow's problem.

I take the first half of that argument seriously, but I reject its conclusion, for two reasons. First, no one is asking you to gold-plate the trial for every market on earth. The ask is narrower: make the handful of choices that are cheap to include and impossible to retrofit. EQ-5D is a short questionnaire, a resource-use CRF is close to free, a comparator that reflects standard of care in your lead reimbursement market often costs no more than the placebo arm you would run anyway, and follow-up is a clause. This is minimum-viable-evidence discipline, the vital few rather than boiling the ocean, and it is the same 90/10 logic we apply to every other evidence decision, laid out in the 90/10 minimum viable evidence framework.

Second, "collect it later" is exactly the retrofit this post has watched fail. Managed access recovers a choice you deferred; it cannot create one you never made. Pembrolizumab's three CDF years did not contribute; mogamulizumab was refused the Fund because the comparator gap was structural. Where the missing input is a comparator that was never in the trial, no scheme reaches back and inserts it. And the optimistic recovery statistics, even if they hold, describe drugs that reached managed access in the first place, itself a delay measured in years, not the tidy "approve now, access later" the counterargument imagines.

Where the HEOR voice needs a seat

We said earlier that "involve HEOR earlier" is true but useless on its own. Cashed out, it means three specific insertion points, each before the concrete sets.

• The target product profile. The TPP is where you name the market you are building the asset to be reimbursed in, before a single endpoint is chosen; set the access target there and the comparator, instrument and follow-up decisions inherit from it. This is our recommended way of using the TPP rather than a documented industry standard, but the logic holds: an access target that is never written down at the TPP stage will not drive the design. We have shown what a rigorous TPP looks like in creating target product profiles without the guesswork.
• The IEP gap analysis. The four design-locked decisions should sit as explicit line items in your IEP's gap review, scored for impact and feasibility, before the protocol is finalised, not discovered during an appraisal.
• Formal HTA scientific advice, during design. This is the mechanism, and the HTA bodies say so themselves. NICE's stated optimum is that "the optimum time to seek advice on the clinical development and evidence generation plan... is during the design period and before the initiation of the main studies." [13] And it works: in EUnetHTA's Joint Action 3 parallel-consultation data, scientific advice changed the development plan in 12 of 21 cases. [13]

One precision worth holding, because the two get conflated. Europe now runs two separate mechanisms at two points on the timeline. The Joint Scientific Consultation sits before trial design and can shape the study; the Joint Clinical Assessment sits after the marketing-authorisation application and reads the study you already built. One can still work the concrete, the other can only inspect it once set. And know which market you are building for. Because NICE and HAS diverge so often, 62% of the time on the same oncology drugs, and reject on different grounds (NICE mainly on cost-effectiveness, HAS mainly on insufficient clinical evidence) [5], the design-stage question stops being "will this satisfy an HTA body" in the abstract and becomes "which market am I building this trial to be reimbursed in, and what does that body need to see." That target belongs in the TPP, set before the protocol locks. It's the kind of gap we built InovaCS to hunt for, precisely because it only shows up while the concrete is still wet.

The one deadline

The whole argument collapses to a single deadline. Four decisions are available to you only while the protocol is still wet, and permanently unavailable once it sets: comparator, utility instrument, follow-up duration, resource-use capture. Almost everything else in your access strategy can genuinely be built later. These four cannot. The HEOR conversation has exactly one home: before the main studies start, the last moment the concrete will take an impression. Not launch planning. Not a vague gesture at "getting people in the room earlier."

And the stakes are not really about dossiers. An approved medicine that no health system will fund is, for the patient waiting on it, not far from a medicine that was never approved at all. Those four choices are what separate the one from the other. Pour carefully.

References

[1] National Institute for Health and Care Excellence. "Mogamulizumab for treating relapsed or refractory mycosis fungoides or Sézary syndrome (TA754)." Final guidance, 15 December 2021; recommends mogamulizumab for routine NHS use via a revised commercial arrangement and declines the Cancer Drugs Fund, superseding the February 2021 final appraisal document. https://www.nice.org.uk/guidance/ta754

[2] National Institute for Health and Care Excellence. "Pembrolizumab for treating locally advanced or metastatic urothelial carcinoma after platinum-containing chemotherapy (TA692)." Guidance, 28 April 2021; updates and replaces TA519 (2018), which was available through the Cancer Drugs Fund. https://www.nice.org.uk/guidance/ta692

[3] Al Khayat MNMT, et al. (2023). "Pralsetinib for RET Fusion-Positive Advanced Non-small-Cell Lung Cancer: An Evidence Review Group Perspective of a NICE Single Technology Appraisal." PharmacoEconomics. PMID: 36757608. https://pubmed.ncbi.nlm.nih.gov/36757608/

[4] Macabeo B, et al. (2024). "The Acceptance of Indirect Treatment Comparison Methods in Oncology by Health Technology Assessment Agencies in England, France, Germany, Italy, and Spain." PharmacoEconomics Open. PMID: 38097828. https://pubmed.ncbi.nlm.nih.gov/38097828/

[5] Trouiller JB, Laramée P. (2023). "Comparative Assessment of Reimbursement Recommendations by NICE and HAS for Oncology New Medicines Indicated for the Treatment of Solid Tumors from 2015 to 2021." Medical Decision Making. PMID: 37480275. https://pubmed.ncbi.nlm.nih.gov/37480275/

[6] Kite Pharma/Gilead. "ZUMA-1." ClinicalTrials.gov: NCT02348216. https://clinicaltrials.gov/study/NCT02348216 ; Kite Pharma/Gilead. "ZUMA-7." ClinicalTrials.gov: NCT03391466. https://clinicaltrials.gov/study/NCT03391466 ; Novartis. "JULIET." ClinicalTrials.gov: NCT02445248. https://clinicaltrials.gov/study/NCT02445248

[7] Biogen. "SHINE." ClinicalTrials.gov: NCT02594124. https://clinicaltrials.gov/study/NCT02594124 ; Biogen. "ENDEAR." ClinicalTrials.gov: NCT02193074. https://clinicaltrials.gov/study/NCT02193074

[8] Janssens R, et al. (2023). "How can patient preferences be used and communicated in the regulatory evaluation of medicinal products? Findings and recommendations from IMI PREFER and call to action." Frontiers in Pharmacology. PMID: 37663265. https://pubmed.ncbi.nlm.nih.gov/37663265/

[9] Gibbons CL, Latimer NR. (2025). "Prevalence of Immature Survival Data for Anticancer Drugs Presented to the National Institute for Health and Care Excellence Between 2018 and 2022." Value in Health. PMID: 39725010. https://pubmed.ncbi.nlm.nih.gov/39725010/

[10] Walton M, et al. (2019). "Tisagenlecleucel for the Treatment of Relapsed or Refractory B-cell Acute Lymphoblastic Leukaemia in People Aged up to 25 Years: An Evidence Review Group Perspective of a NICE Single Technology Appraisal." PharmacoEconomics. PMID: 30982165. https://pubmed.ncbi.nlm.nih.gov/30982165/ ; Cope S, et al. (2019). "Integrating expert opinion with clinical trial data to extrapolate long-term survival: a case study of CAR-T therapy for children and young adults with relapsed or refractory acute lymphoblastic leukemia." BMC Medical Research Methodology. PMID: 31477025. https://pubmed.ncbi.nlm.nih.gov/31477025/

[11] "Real-world comparison of tisagenlecleucel and axicabtagene ciloleucel (US Medicare claims)." ClinicalTrials.gov: NCT06271369. https://clinicaltrials.gov/study/NCT06271369 ; "Real-world comparison of Kymriah and Yescarta (French PMSI claims)." ClinicalTrials.gov: NCT05349201. https://clinicaltrials.gov/study/NCT05349201

[12] O'Sullivan AK, et al. (2005). "Collection of health-economic data alongside clinical trials: is there a future for piggyback evaluations?" Value in Health. PMID: 15841896. https://pubmed.ncbi.nlm.nih.gov/15841896/

[13] Ibargoyen-Roteta N, et al. (2022). "A systematic review of the early dialogue frameworks used within health technology assessment and their actual adoption from HTA agencies." Frontiers in Public Health. PMID: 36276363. https://pubmed.ncbi.nlm.nih.gov/36276363/