A1SLIDES · THE DOSE
The Denominator Is the Argument
Issue 003 · August 2026
IN THIS ISSUE
In August, three separate decisions turned on the same question: which population are we actually talking about? A drug was approved on a lower response rate than the one presented. One indication generated seven separate comparator questions. A "flat" quarter concealed two products growing over 60%. The number was never the argument. The denominator was.
Executive Summary
Three things happened in August that share a structure.
A regulator approved a drug on different numbers than the ones presented. On 6 August, FDA granted accelerated approval to Replimune's Tudriqev (vusolimogene oderparepvec-wtpg) with nivolumab in unresectable advanced cutaneous melanoma after progression on anti-PD-1. The approval rests on an objective response rate of 24.2% and median duration of response of 14.1 months. The IGNYTE figures in wide circulation before approval were an ORR of 33.6% and median DOR of 24.8 months. Both are accurate. They describe different populations. IGNYTE enrolled 140 patients; 91 had at least one non-injected lesion and formed the efficacy-evaluable population.
Europe made comparator definition the central access problem. The second and third JCA reports, covering two extensive-stage small cell lung cancer therapies, required one PICO and seven PICOs respectively — a sevenfold difference in scope driven by breadth of claimed indication.
Aggregate numbers concealed the story in both major earnings updates. Bayer's pharma sales were flat at roughly €4.46bn while Nubeqa grew 63.9% and Kerendia 82.9%. BioNTech's revenue fell 59% to €105.6m against a prior-year quarter that included a one-time Pfizer payment.
The common thread: the headline number was true and useless. The unit of analysis was the argument.
The Dose · Issue 003
Get the next issue before your next pitch.
Funding signals, regulatory moves, and the slide that wins the room — straight to your inbox, once a month.
One email a month. Unsubscribe anytime.
01 · The Signal
One structural shift. Explained clearly.
The question that decided every major August event was not "what is the number?" It was "what is the number a number of?"
Take the two events furthest apart in kind - a US accelerated approval and a European HTA assessment - and the same structure appears.
IGNYTE enrolled 140 patients. Ninety-one had at least one non-injected lesion, and that subgroup formed the efficacy-evaluable population. This mattered because RP1 is injected directly into tumours, and the central regulatory question was whether observed responses reflected systemic antitumour activity attributable to RP1 rather than to nivolumab. The approval rests on ORR 24.2% and median DOR 14.1 months. The figures circulating for the week before — 33.6% and 24.8 months — describe a different denominator.
Nobody was wrong. But a Medical Affairs team that built pre-approval slides on 33.6% now holds a deck that does not match the label, and an MSL who quotes it in the field is outside it. If your response-rate slide does not state the evaluable population and why it differs from enrolment, it needs rebuilding the moment a label lands.
Figure 1 · IGNYTE · Replimune Tudriqev
The same trial, two response rates
The HTA Coordination Group published its second and third Joint Clinical Assessments in July, both covering extensive-stage SCLC. For PharmaMar's lurbinectedin, Member States defined a single PICO, reflecting a narrowly specified population and treatment setting. For tarlatamab, the same process required seven PICOs across multiple subpopulations, because the claimed indication was broader. In both cases randomised controlled trials formed the evidence base, and all Member State PICOs were addressed through a combination of direct and indirect methods.
Seven PICOs is seven distinct population–comparator–outcome questions, each needing its own evidence and, frequently, its own indirect treatment comparison. Published simulation work suggests the realistic range runs higher: an ISPOR simulation for two anticancer investigational products predicted 8–29 and 12–31 PICOs, generating between roughly 230 and 1,570 analysis requests. A review of 35 published PICO scoping exercises across 20 indications found an average of eight consolidated PICOs with seven countries participating. A first-line mNSCLC scoping simulation across 25 countries generated 67, driven largely by biomarker and histology subpopulations.
Figure 2 · EU Joint Clinical Assessment
One disease area, two assessment scopes
Lurbinectedin
Narrowly specified population and treatment setting
Tarlatamab
Broader claimed indication, multiple subpopulations
Each block is a distinct population-comparator-outcome question requiring its own evidence.
The single-population value story is finished in Europe. A JCA-era value deck is not one narrative with regional annexes; it is a scaffold that resolves to eight or thirty different population–comparator pairs without being rebuilt each time. The window is unforgiving: 90 days between PICO release and dossier submission, running in parallel with the EMA process rather than after it, which means the dossier is due before the CHMP opinion.
The presentation consequence is concrete. Every slide carrying evidence needs a visible unit of analysis: which population, which comparator, which outcome, and where that combination came from. Teams that treat this as a labelling detail will spend the 90 days rebuilding decks. Teams that build it into the template will spend the 90 days on the analyses.
02 · Capital & Deal Pulse
Notable market movements this August
AstraZeneca–Bristol Myers Squibb — reported, then denied
Approx. $400bn combined | AZ shares fell as much as 7%
The Financial Times reported on 2 August that the two companies had discussed a merger. AstraZeneca's shares dropped as much as 7% the following day. By 6 August, a senior source was quoted denying the talks. Neither company has confirmed anything at any stage.
Worth sitting with: the $400bn figure was always simple addition — AstraZeneca at roughly $264bn and Bristol Myers Squibb at roughly $133bn on the Friday before the report. No premium was ever disclosed, because no deal was ever confirmed. Analysts moved quickly to "unlikely," citing antitrust exposure and commercial overlap. BMO pointed specifically to non-small cell lung cancer, where BMS's Opdivo brought in $10.05bn worldwide in 2025 against AstraZeneca's Imfinzi at $6.06bn.
The lesson is narrower than "investors want a value-creation bridge." It is that a market will price a thesis it has not been given, and it will price it conservatively. AstraZeneca went in with a specific vulnerability: a late-stage cardiovascular trial had failed earlier in the month after management communicated confidence, denting credibility ahead of a story that needed a lot of it.
Replimune / Tudriqev — approved
FDA accelerated approval, 6 August | Third review cycle, after two complete response letters
The Cellular, Tissue, and Gene Therapies Advisory Committee voted 10–3 on 30 July that IGNYTE's efficacy results were evaluable and clinically meaningful. FDA granted accelerated approval four days after the 2 August PDUFA date. Continued approval is contingent on the confirmatory Phase 3 IGNYTE-3 trial, reading out in 2030.
Worth sitting with: the three dissenting panellists did not argue the drug lacked value. One stated plainly that the vote was driven by evidentiary standards rather than clinical meaningfulness. That distinction matters for anyone drawing lessons from this adcomm — see Section 03.
BioNTech Q2 2026
Revenue €105.6m, down 59% | FY guidance cut to €1.6–1.9bn from €2.0–2.3bn
Worth sitting with: the decline is real, but the comparator flatters it. The prior-year quarter included a one-time compensation payment from Pfizer. Second-half guidance includes a €613m collaboration payment from Bristol Myers Squibb. Cash and security investments stand at €16.6bn against 14 ongoing pivotal trials, with more than 17 late-stage or pivotal readouts targeted through 2030. Adjusted SG&A rose sharply as commercial infrastructure is built ahead of first product approvals. Incoming CEO Guido Oelkers takes office by 1 February 2027; Ugur Sahin remains involved.
Bayer Q2 2026
Pharma sales approx. €4.46bn, flat | Group sales €10.872bn, up 2.2% (Fx & portfolio adjusted)
Worth sitting with: "flat" is the least informative word available. Nubeqa grew 63.9% to €880m and Kerendia 82.9% to €329m — together €1.2bn, up roughly 69% — while Eylea fell 27% under biosimilar pressure and Xarelto continued its expected decline. Pharmaceuticals EBITDA margin compressed to 23.7% from 24.5%. Bayer nonetheless confirmed its currency-adjusted Group outlook and lowered its net financial debt expectation. Free cash flow was negative €2.7bn in H1, driven by €2.5bn in litigation payouts.
Figure 3 · Bayer Pharmaceuticals · Q2 2026
What "flat" concealed
The headline
Flat
Inside the flat · year-on-year change
0%
None of it visible in the headline number. Source: Bayer Q2 2026 results, 4 Aug 2026. Growth rates currency and portfolio adjusted. Xarelto bar shows direction only; no comparable quarterly growth rate disclosed. Percentage changes are not weighted by revenue contribution.
03 · Pipeline Watch
Regulatory and evidence signals
Approved
Replimune, Tudriqev (vusolimogene oderparepvec-wtpg) + nivolumab · Unresectable advanced cutaneous melanoma after anti-PD-1
The reading circulating all week is that framing changed the outcome. That is not quite what happened, and the difference is worth getting right.
An advisory committee and a complete response letter answer different questions. The CTGTAC was asked two narrow ones: whether the single-arm IGNYTE study as designed allowed reliable determination of response rate and durability in the proposed population, and whether the observed response rate and duration were clinically meaningful. It voted 10–3 yes. Dissenters were explicit that they were voting on evidentiary standards, not on whether the drug helps patients. FDA reviewers' underlying objection — that without a comparator arm RP1's contribution cannot be isolated from nivolumab, and that an intratumourally injected agent raises questions about systemic effect — was never resolved. It was contextualised against unmet need and then handled through the accelerated approval pathway, with the confirmatory trial carrying the burden.
the honest lesson is not that presentation changed the answer. It is that presentation changed which question was in front of the room. Contested evidence became decision-ready when the evidence hierarchy, its limitations, the clinical context and the available alternatives sat in one frame — with Amtagvi (lifileucel) as the only approved comparator in this setting and the confirmatory readout as the stated resolution path. Uncertainty was not removed. It was located, bounded, and given a date.
label figures (ORR 24.2%, DOR 14.1 months) differ from the widely circulated IGNYTE figures (33.6%, 24.8 months) because the evaluable population differs. Any field or medical material built on pre-approval numbers needs auditing now.
JCA Watch
Second and third Joint Clinical Assessments published
Both covered extensive-stage SCLC. Lurbinectedin required one PICO; tarlatamab required seven. Both products received EMA marketing authorisation at the end of May 2026, so the JCA outputs now feed national HTA procedures.
Key signal: breadth of claimed indication is a direct multiplier on evidence workload. A wider label is not a free option; it is seven evidence packages instead of one.
Recurring evidence weaknesses flagged in the first year of JCAs: risk of bias in single-blinded trials; concerns where trials were conducted outside Europe, since patient characteristics, treatment pathways and standards of care may differ; and inadequate handling of missing data. Transparent justification where evidence is limited or absent has been highlighted as important by the JCA Subgroup.
Commercial Watch
Bayer, growth products versus legacy erosion
Nubeqa and Kerendia against Eylea and Xarelto is the clearest current example of why aggregate divisional reporting fails as a narrative device. Bayer has upgraded its pharma outlook on the strength of two products invisible in the headline number.
On The Calendar
- Late October 2026AstraZeneca and Bristol Myers Squibb Q3 earnings, within days of each other. First formal opportunity for either to address the merger reporting on the record.
- Late 2026 / early 2027Bayer's asundexian launches expected in the US and China; priority reviews ongoing in multiple markets.
- PendingBMS pivotal readouts for milvexian and Cobenfy.
- Through 2026JCA volume scales from the first year's ten assessments; 15 medical device categories identified as potentially eligible, the first year high-risk devices meaningfully enter EU-level assessment.
- 1 February 2027Guido Oelkers assumes BioNTech CEO role.
- 2030IGNYTE-3 confirmatory readout, on which Tudriqev's continued approval depends.
04 · The Deck
How clinical data actually gets read
Third instalment in our running series drawn from The Life Sciences Presentation Report 2026–27. Standard 01 covered chart-type accuracy. Standard 02 covered survival curve misreading. Standard 03 is the row almost every deck deletes.
Standard 03: The number-at-risk row is not decoration. It is the denominator.
Every Kaplan-Meier curve is a series of conditional probabilities computed on a shrinking population. At month 36, the curve may rest on a handful of patients. The number-at-risk row is the only element on the chart that tells the reader this. Delete it — or set it in six-point type that vanishes at projection scale — and the flat tail reads as a durable plateau rather than an artefact of a population thinned to almost nothing.
This is the Replimune denominator gap drawn as a picture. In both cases a true number becomes misleading because the population it was computed on is not visible.
Foundational evidence: Cleveland and McGill's 1984 work in the Journal of the American Statistical Association established that position judgments on a common scale are read substantially more accurately than angle judgments — which is what a pie chart asks for, and why patient-segment pies and market-size bubbles sit at the bottom of the accuracy ranking. More recent work on clinician interpretation of survival plots reports low baseline accuracy on Kaplan-Meier curves specifically, with two errors recurring: reading a flattening tail as declining risk when it reflects a shrinking at-risk population, and reading two separated curves as a statistical comparison when no test has been run.
The standard, in four rules:
The number-at-risk row is legible at projection scale, or the slide is not finished. Minimum 10pt at 16:9; test on the room's screen, not a laptop.
Censoring marks stay visible. Their absence implies complete follow-up that does not exist.
The comparative statistic lives with the curve, not on the source slide it was lifted from. If no test was run, say so on the chart.
State the analysis population and its size on the chart, not only in the footnote — and where it differs from enrolment, say why.
Rule 4 is new to this standard and is a direct consequence of the Replimune approval.
05 · Editor's View
Perspective from the desk
From my experience across Medical Affairs, market research and life-sciences strategy, the 24.2% versus 33.6% example is generated through an evidence governance test. This is not a reporting error. Both figures in fact are technically correct, speak a truth of their own. However, once the once the result is separated from its attributable population definition, it becomes vulnerable to misinterpretation across medical, regulatory, market access and commercial channels and teams.
A comparable situation can be seen with Leqembi (lecanemab) in early Alzheimer's disease. In the US, the FDA converted Leqembi from accelerated to traditional approval after the CLARITY AD confirmatory study verified clinical benefit. Per the US label, treatment with drug is is permissible in patients with mild cognitive impairment or mild dementia due to Alzheimer's disease, while managing the higher risk of amyloid-related imaging abnormalities among ApoE ε4 homozygotes through genetic testing, warnings and MRI monitoring and not by formally excluding that population.
The EU assessed the same underlying evidence through a different benefit–risk lens. Its April 2025 authorization restricted Leqembi to patients with confirmed amyloid pathology who are ApoE ε4 non-carriers or heterozygotes, effectively excluding homozygotes from the authorized population.
The evidence base did not fundamentally change; the eligible denominator did.
In the US, risk was primarily managed through labelling, testing and monitoring. In the EU, risk management also changed the population for whom the treatment could be prescribed.
For global healthcare teams, this reinforces an important point: one global evidence deck cannot simply be replicated across markets.
The evidence architecture should anticipate jurisdiction-specific populations, comparators, safety thresholds and decision criteria. Medical, regulatory, HEOR and market-access teams therefore require a common evidence backbone, supported by modular and traceable claims that can be adapted without losing their original context.
Evidence communication must be governed with the same discipline as evidence generation. The denominator is not a footnote it is part of the claim and a very essential variable.
06 · Off The Deck
What's new at A1 Slides
Strengthening the Framework Behind Every Delivery
As our work with global organisations continues to grow, we are strengthening the governance and protection framework behind our delivery model.
A1 Slides has initiated its ISO 27001 certification process to further formalise information-security practices across our operations. Alongside this, we maintain comprehensive Professional Indemnity Insurance coverage, adding another layer of assurance for the work entrusted to us.
Together with our NDA-led workflows and controlled project practices, these measures reflect our continued investment in security, accountability and enterprise-ready delivery.
A1 Slides is strengthening data protection through ISO 27001-aligned security practices—learn how we safeguard client information.
Adding a New Dimension to Life Sciences Communication
A1 Slides is expanding its life sciences capabilities through a new collaboration with a South Korean specialist in 3D scientific visualization and animation.
The collaboration adds a new layer to how we can support complex scientific communication — transforming concepts such as DNA, drug mechanisms, molecular structures and complex scientific data into engaging 3D visuals and videos that are easier for audiences to understand.
Combined with our presentation and data-storytelling capabilities, this enables us to support life sciences teams beyond the slide — from complex information to visual storytelling and immersive scientific communication.
07 · People Also Ask
Because the label reports on the efficacy-evaluable population, frequently a subgroup of those enrolled, defined by the specific regulatory question at issue. IGNYTE enrolled 140 patients; 91 had at least one non-injected lesion — the population relevant to whether RP1 produced a systemic effect. Circulating figures were ORR 33.6% and median DOR 24.8 months; the approval rests on 24.2% and 14.1 months. Both describe the same trial. Field and medical material built on pre-approval figures should be audited against the label immediately after approval.
Two narrow questions: whether the single-arm IGNYTE study as designed allowed reliable determination of response rate and durability in the proposed population, and whether the observed response rate and duration were clinically meaningful. The 10–3 vote was on those questions, not on approval, and committee recommendations are non-binding. Dissenting members stated they were voting on standards of evidence rather than clinical value.
It depends almost entirely on breadth of claimed indication. The two JCAs published in July required one and seven respectively for two extensive-stage SCLC therapies. A review of 35 published PICO scoping exercises across 20 indications found an average of eight consolidated PICOs with seven countries participating. Simulations for individual oncology products have predicted 8–29 and 12–31, and a 25-country first-line mNSCLC exercise generated 67, driven by biomarker and histology subpopulations.
There is a 90-day window between release of the PICO framework and dossier submission, and the JCA runs in parallel with the EMA regulatory process rather than after it — the dossier is due before the CHMP opinion. Advance preparation ahead of formal EMA submission is the consistent recommendation from teams who have been through it.
Risk of bias in single-blinded trials; concerns where trials were conducted outside Europe, because patient characteristics, treatment pathways and standards of care may differ; and inadequate handling of missing data. Transparent justification where evidence is limited or absent has been highlighted as important by the JCA Subgroup.
It changes what the evidence supports and adds an explicit condition. Tudriqev's approval is based on objective response rate and duration of response, with continued approval contingent on verification of clinical benefit in a confirmatory trial — IGNYTE-3, reading out in 2030. That contingency is part of the claim and belongs on the slide, not in a footnote.
Analysts questioned strategic rationale rather than scale, citing antitrust exposure and commercial overlap, particularly in non-small cell lung cancer, where BMS's Opdivo recorded $10.05bn in 2025 worldwide sales against AstraZeneca's Imfinzi at $6.06bn. The $400bn figure was the sum of two market capitalisations with no disclosed premium. A senior source subsequently denied the talks.
It is the row beneath a Kaplan-Meier curve showing how many patients remain in the analysis at each time point — the denominator for every point on the curve. Without it, a flattening tail computed on a very small remaining population reads as a durable plateau. It should be legible at projection scale, a minimum of 10pt at 16:9.
Only with an explicit statement that no comparison was performed. Reading two separated curves as a statistical comparison when no test has been run is one of the most consistently documented misinterpretations of survival plots.
As a scaffold rather than a narrative. The core evidence architecture stays fixed while population, comparator and outcome resolve differently per PICO. Each slide carries its unit of analysis visibly, so switching PICO does not require rebuilding the deck.
Sources & Evidence
- Replimune — FDA accelerated approval of TUDRIQEV (vusolimogene oderparepvec-wtpg), 6 August 2026. ir.replimune.com
- Replimune — Form 8-K, CTGTAC advisory committee outcome, 30 July 2026. sec.gov
- BioPharma Dive — Replimune rebounds to win FDA approval of melanoma drug, 6 August 2026
- STAT News — FDA approves Replimune melanoma drug previously rejected twice, 6 August 2026
- CancerNetwork — FDA Advisory Committee Backs RP1/Nivolumab for Advanced Melanoma
- FDA Law Blog — The FDA AdComm Reboot: Part 2, Lessons from the Meeting on Replimune's RP1, August 2026
- Becaris Publishing — Second and third Joint Clinical Assessments published, addressing all Member State PICOs with comparative evidence, July 2026
- Avalere Health Advisory — Europe's First Joint Clinical Assessment Report is Out, 17 June 2026
- ICON plc — Lessons learned in EU HTA Regulation's first year, 13 March 2026
- ISPOR — Predicting Evidence Requirement Implications of the EU JCA Through a PICO Simulation for Two Anticancer IMPs
- European Joint Clinical Assessment PICO Scoping Process: Analysis of Current Approaches and Recommendations, PMC12821599
- Financial Times — AstraZeneca holds talks with Bristol Myers Squibb over $400bn tie-up, 2 August 2026
- CNBC — AstraZeneca slides after reports of Bristol Myers merger talks leave analysts 'perplexed', 3 August 2026
- PharmExec — Source Denies AstraZeneca–Bristol Myers Merger Talks, 6 August 2026
- BioSpace — 'Unlikely' AstraZeneca–BMS mega-merger would be largest pharma deal ever (BMO commentary), August 2026
- BioNTech — Second Quarter 2026 Financial Results and Corporate Update, 4 August 2026
- Bayer — Q2 2026 Results, 4 August 2026
