Base rates: what are the odds?
Computed from our full archive of 23,433 decisions back to 2000. Denominators are decided outcomes only (withdrawals, non-submissions and advice-only items excluded).
Yes here, no there
Of the 1,264 molecules that more than one agency has given a verdict on, 741 were recommended or restricted somewhere and not recommended somewhere else. The same evidence, the same molecule, opposite answers.
That is 59% overall, but the number on its own is misleading, because the more agencies look at a medicine the more chances there are for one of them to say no. Read it by exposure instead.
| Agencies with a verdict | Molecules | Disagreed | Rate |
|---|---|---|---|
| 2 | 445 | 141 | 32% |
| 3 | 178 | 85 | 48% |
| 4 | 137 | 91 | 66% |
| 5 | 95 | 59 | 62% |
| six or more | 409 | 367 | 90% |
A disagreement is one agency recording “recommended” or “restricted” and another recording “not recommended”, with at least two agencies having issued a verdict; terminated and non-verdict appraisals are excluded. It is a conservative count: Germany and Italy record “no additional benefit proven”, a negative finding in those systems, which is not treated here as a no. Counting it would give 798.
What the base rates do not tell you
The rates above are each computed over that agency’s own archive. Those archives open in different years and cover different shares of what each system assesses, so the column is not a ranking of how generous the agencies are. Our own file shows the size of that: the Scottish acceptance rate is 80% across the whole archive and 88% on recent submissions — 8 points from period alone.
The same caution applies to coverage. This archive holds far more French dossiers than English appraisals, because HAS publishes an opinion on almost everything and NICE appraises a selection.
| Agency | Decisions | Molecules | Span |
|---|---|---|---|
| HAS · France | 6,500 | 2,262 | 2003–2026 |
| PBAC · Australia | 2,361 | 872 | 2014–2026 |
| SMC · Scotland | 2,059 | 1,259 | 2002–2026 |
| PHARMAC · New Zealand | 1,946 | 1,193 | 1995–2026 |
| AWMSG · Wales | 1,789 | 1,034 | 2002–2026 |
| TLV · Sweden | 1,512 | 1,036 | 2002–2026 |
| CIPM · Spain | 1,082 | 534 | 2019–2026 |
| G-BA · Germany | 1,000 | 468 | 2011–2026 |
| NICE · England | 930 | 537 | 2000–2026 |
| CONITEC · Brazil | 893 | 546 | 2012–2026 |
| NYE METODER · Norway | 871 | 487 | 2014–2026 |
| AOTMIT · Poland | 738 | 418 | 2016–2026 |
| ZIN · Netherlands | 442 | 305 | 2006–2026 |
| AIFA · Italy | 431 | 210 | 2017–2026 |
| DMC · Denmark | 425 | 254 | 2017–2026 |
| HIRA · South Korea | 197 | 161 | 2019–2026 |
| CHUIKYO · Japan | 148 | 128 | 2019–2026 |
| EU JCA · EU joint assessment | 23 | 23 | 2025–2026 |
Questions this data answers
How often do HTA agencies disagree about the same medicine?
Across this archive, 741 molecules have been recommended or restricted by one agency and not recommended by another. That is 59% of the 1,264 molecules that more than one agency has actually given a verdict on — but the figure rises with the number of agencies involved, from 32% at two to 90% at six or more.
What counts as a disagreement?
One agency recording "recommended" or "restricted" and another recording "not recommended", on the same molecule, with at least two agencies having issued a verdict. Appraisals that were terminated or that record no verdict are excluded. The count is deliberately conservative: Germany and Italy record "no additional benefit proven", which is a negative finding in those systems but is not counted here as a "no". Counting it would give 798 rather than 741.
Can I compare the base rates between agencies?
Not directly. Each rate is computed over that agency's own archive, and those archives start in different years and cover different shares of what each system assesses. Our own file shows the size of that: the Scottish acceptance rate is 80% across the whole archive and 88% on recent submissions — 8 points from period alone.
How much of this is free?
Every molecule in the archive is searchable here — all 6,583 of them — with how many decisions it has, how many agencies have ruled on it, whether they disagreed, and the year of the most recent one. Decision-level detail, meaning the agency, date, outcome and a link to that agency's own record, is open for Pembrolizumab and Semaglutide as worked examples. The rest of the decision-level archive, the divergence list and the API are in the subscription.
How current is the HTA decision archive?
The most recent decision in the archive is dated 2026-09-16. The HTA archive refreshes twice a month.
Citing this
Altima Partners, “HTA outcomes: 23,083 decisions across 18 agencies”, altimapartners.co.uk/hta-outcomes, decisions to 2026-09-16. If you quote the divergence figure, quote its definition with it.
Decision records by molecule
534 molecules have a full cross-agency record on this site: every decision we hold, what each agency concluded, and where they disagree.
A–Z: abaloparatide–cholecalciferol ciclosporin–fidaxomicin filgotinib–maribavir marstacimab–ranibizumab ranolazine–trastuzumab emtansine tremelimumab–zonisamide
By agency: AIFA AOTMIT AWMSG CHUIKYO CIPM CONITEC DMC EU JCA G-BA HAS HIRA NICE NYE METODER PBAC PHARMAC SMC TLV ZIN
By therapy class: Antidiarrheals / intestinal anti-inflammatories Diabetes Other alimentary tract and metabolism Antithrombotics Antihemorrhagics Lipid modifying agents Sex hormones and modulators Pituitary and hypothalamic hormones Antivirals Antineoplastics Endocrine therapy Immunosuppressants Analgesics Antiepileptics Psycholeptics Psychoanaleptics Other nervous system Obstructive airway diseases Ophthalmologicals
What does this mean for your submission?
We read HTA precedent the way committees do: which comparators, which endpoints, which price signals swung past decisions in your indication. We build it into payer-ready models and value dossiers. Fixed fees, senior-built.
Talk to us about your asset