Fewer than 10% of drug candidates entering clinical trials reach regulatory approval, and most of the failures that define this statistic were not unforeseeable. They were scenarios that a structured pre-mortem analysis could have identified before the budget was committed, before the Phase III protocol was locked, before the hundreds of millions of dollars that distinguish a pivotal programme from a calculated experiment were deployed. [4] The pre-mortem technique, popularised by psychologist Gary Klein and now used by some pharmaceutical portfolio teams, asks decision-makers to imagine that a programme has already failed and to work backward from that failure to identify what went wrong. This prospective failure analysis often surfaces risks that conventional forward-looking risk assessment misses, because it shifts the cognitive frame from defending a decision to explaining its failure.[1]
At Pienomial, we built KnolPersona and KnolAI specifically to provide the evidence intelligence that makes pharma scenario planning rigorous rather than anecdotal. The pre-mortem is only as good as the evidence it draws on. A team that imagines failure scenarios without access to competitive precedent, HTA body evidence requirements, regulatory history, and payer positioning intelligence is running a thought experiment. A team that grounds each failure scenario in sourced, current intelligence is running a scenario planning in healthcare exercise that can genuinely change a go/no-go decision. This post explains how to run a pre-mortem on a clinical trial budget decision with the discipline and evidence grounding that the financial stakes demand.[9]
The go/no-go decision at the Phase II to Phase III transition is the most consequential resource allocation decision in drug development. Phase III trials require between $20 million and $100 million or more in direct costs, with average per-patient costs of $113,030. Each protocol amendment costs an estimated $500,000 or more in direct costs and introduces three or more months of delay. Screen failure rates above 50% can double the per-enrolled-patient cost of the entire trial. [5] Committing to Phase III without systematically exploring the most likely failure modes of the evidence architecture is not a risk accepted knowingly. It is a risk that has not been seriously confronted.
A scoping review published in Clinical Pharmacology and Therapeutics in 2025 examined decision-making criteria and methods at the Phase II to Phase III transition, finding that while the probability of success concept plays a central role in informing go/no-go decisions, most frameworks focus on statistical significance in efficacy rather than the full multi-stakeholder picture that actually determines commercial success: regulatory approval, market access, financial viability, and competitive performance simultaneously. [3] A pre-mortem that addresses only clinical endpoint risk is a partial pre-mortem. A rigorous pre-mortem addresses all five failure modes: clinical, regulatory, HTA, competitive, and commercial.
Leading biopharma companies routinely discontinue 21 to 22% of their pipeline programs annually as part of disciplined portfolio management, recognising that the no-go decision at the right stage is not failure but precisely calibrated capital allocation. [8] The pre-mortem is the process that makes those decisions earlier and with more confidence rather than later and with more sunk cost.
2. What a Pre-Mortem Is and How It Differs From a Risk Register
The pre-mortem and the conventional risk register are both risk identification tools, but they approach the same problem from opposite cognitive directions. A risk register asks: what could go wrong with this programme? It is a forward-looking exercise performed by a team that has already developed an attachment to the programme they are assessing, which is why risk registers consistently underestimate the most consequential risks: the ones that require the most uncomfortable conclusions.[1]
The pre-mortem asks a fundamentally different question: it is five years from now, this programme has failed completely, what happened? The instruction to imagine failure as already having occurred, rather than as a possibility to be assessed, produces a qualitatively different set of identified risks because it removes the psychological friction of arguing against a proposal one's team has spent months developing. In the pre-mortem frame, identifying the most devastating failure scenario is the task, not a criticism of the programme's proponents.[1]
The output of a pre-mortem is a structured set of identified failure scenarios, each grounded in the specific evidence that makes that scenario plausible, with a prioritised set of evidence actions that, if taken before budget commitment, would either close the failure risk or allow the team to make an informed decision to accept it. The output of a risk register is a list of risks with probability and impact scores that often get reviewed once and filed without driving specific evidence actions.[9]
3. The Five Failure Mode Categories for a Clinical Trial Pre-Mortem
A rigorous clinical trial pre-mortem must address five failure mode categories, each requiring different evidence to assess and each capable of producing the programme failure the team is asked to imagine.[3]
Failure Mode 1, Clinical failure: The trial does not demonstrate adequate efficacy or safety. This is the most commonly anticipated failure mode and the one most aggressively assessed in standard probability of success calculations. The pre-mortem question is not whether the trial might fail to achieve statistical significance, but specifically why it would fail: an insufficient effect size relative to the chosen comparator, an endpoint that does not capture the benefit the mechanism produces, a patient population that dilutes the responder effect, or an active comparator arm that outperforms the expected standard of care baseline.
Failure Mode 2, Regulatory failure: The trial succeeds clinically but the regulatory filing fails or receives a CRL. Pre-mortem questions for regulatory failure include: has the chosen comparator been accepted in this indication before? Has the endpoint package satisfied the relevant agency in analogous products? Are there safety signals in the mechanism class that have triggered RMP requirements the protocol does not anticipate? Has the development programme been discussed with the FDA at an End-of-Phase-2 meeting?[4]
Failure Mode 3, HTA failure: The product receives regulatory approval but fails to achieve the reimbursement required for commercial success. This is the failure mode most consistently underweighted in pre-mortems, yet it is now the most commercially consequential for products entering the EU HTA landscape under the Joint Clinical Assessment framework. Pre-mortem questions include: does the trial comparator match what NICE, G-BA, and HAS will use for their assessment? Does the endpoint package satisfy G-BA's patient-relevant endpoint requirements? Is the subgroup pre-specification adequate for the PICO scope the JCA is likely to require?
Failure Mode 4, Competitive failure: The product reaches the market but the competitive landscape has shifted sufficiently that the positioning assumptions embedded in the clinical strategy no longer hold. Pre-mortem questions include: which competitor programmes are most likely to read out before this trial completes, and what would their results need to show to make this programme commercially untenable? Which competitor HTA outcomes would shift the standard-of-care comparator away from what this trial is testing against?[1]
Failure Mode 5, Operational and budget failure: The programme cannot be executed as designed within the committed budget and timeline. Pre-mortem questions include: what is the screen failure rate risk given the eligibility criteria? What is the protocol complexity risk given the endpoint burden? What is the site activation risk in the target geographies? Each month a trial runs over schedule carries significant cost in staff time, site fees, and opportunity cost, and the average Phase III takes 30.5 months even without delays.[5]
4. Step 1: Define the Decision Being Pre-Mortemed with Precision
A pre-mortem produces its most useful output when the decision being assessed is stated with the specificity required to surface decision-relevant failure scenarios rather than generic programme risks. A vague question produces vague scenarios. A precise question produces actionable ones.[9]
The pre-mortem question for a Phase II to Phase III go decision should specify: the exact indication and patient population that Phase III will target, the primary endpoint and the clinically meaningful difference the trial is powered to detect, the comparator and the evidence quality justification for that comparator choice, the primary HTA bodies the programme will target and the approximate submission timeline, and the competitive programmes that will be operating in the same indication over the same timeline.
This level of specificity allows the pre-mortem team to generate failure scenarios that are plausibly specific to this programme in this landscape rather than generic pharmaceutical development risks. Generic risks, inadequate effect size, safety concerns, regulatory rejection, are already in every risk register. The pre-mortem's value is in the specific, programme-and-landscape-specific failure scenarios that do not appear in generic frameworks.[8]
5. Step 2: Ground Each Failure Scenario in Current Evidence Intelligence
The single most important quality-determining factor in a clinical trial pre-mortem is the currency and depth of the evidence intelligence that the failure scenarios are grounded in. A pre-mortem conducted from institutional memory and general industry experience will identify the same failure scenarios that every team in the indication has already identified. A pre-mortem grounded in current, comprehensive intelligence will identify the specific scenarios that this programme, in this competitive landscape, at this regulatory and HTA moment, is most vulnerable to.[9]
KnolAI supports this evidence grounding by generating a structured intelligence brief covering all five failure mode categories before the pre-mortem session begins. The clinical precedent brief covers the endpoint packages that have achieved regulatory approval and HTA reimbursement in the indication and the ones that have not, sourced to the specific trial and assessment documents that document those outcomes. The competitive intelligence brief covers the current pipeline for the indication including trial designs, endpoints, comparators, and expected readout timelines, sourced to clinical trial registry updates. The HTA precedent brief covers what NICE, G-BA, and HAS have required for analogous products in the indication and what the emerging JCA PICO scope is likely to specify.[9]
This intelligence brief transforms the pre-mortem from a creative brainstorming exercise into a structured evidence synthesis exercise. The team is not imagining plausible failure scenarios. They are systematically working through documented failure scenarios from comparable programmes and assessing which ones apply to this programme's specific design.[4]
6. Step 3: Run the KnolPersona Assessor Simulation Before the Pre-Mortem
KnolPersona, Pienomial's AI expert intelligence module within Knolens, provides a specific capability that significantly strengthens the HTA failure mode analysis in a clinical trial pre-mortem: it simulates the review perspective of NICE technical teams, G-BA scientific advisors, and JCA assessment bodies against the current evidence architecture, identifying the specific challenges those reviewers are most likely to raise based on their documented assessment behaviour for analogous products.[9]
Running a KnolPersona assessor simulation before the pre-mortem session means the team arrives at the HTA failure mode discussion with a sourced, precedent-grounded list of the specific weaknesses a NICE technical reviewer would identify in the current evidence package. This is qualitatively different from the team imagining what NICE might challenge: it is what NICE has actually challenged in comparable submissions, applied systematically to the current programme's evidence architecture.
The KnolPersona output for a pre-mortem context typically reveals three to five high-priority HTA failure scenarios: a comparator gap where the trial comparator differs from what NICE or G-BA has historically defined, an endpoint patient-relevance gap where the primary endpoint does not satisfy G-BA's patient-relevant endpoint standard, a subgroup pre-specification gap where the statistical analysis plan does not cover subgroups the JCA PICO scope is likely to require, and a survival extrapolation challenge where the planned trial follow-up is insufficient for the endpoint NICE requires for reimbursement recommendation. Each of these scenarios is addressable before Phase III protocol lock. None of them is addressable after.[9]
7. Step 4: Generate the Failure Scenario List and Prioritise by Impact and Addressability
The output of the pre-mortem evidence intelligence brief and the KnolPersona simulation is a structured failure scenario list. The pre-mortem session's task is to review, extend, and prioritise this list by two dimensions: the impact of the scenario on the programme's commercial outcome if it materialises, and the addressability of the scenario through evidence actions that can be taken before the budget is committed.[1]
The prioritisation matrix produced by this exercise divides failure scenarios into four categories. High-impact, high-addressability scenarios are the ones the pre-mortem was designed to surface: programme-threatening failure modes that can be closed by specific protocol design changes, subgroup pre-specification additions, or RWE commissioning decisions before Phase III begins. These are the scenarios that justify delaying the go decision until the evidence actions are taken. High-impact, low-addressability scenarios are the ones that define the residual risk the team is accepting by proceeding: failure modes that cannot be closed with the evidence available, where the decision is to proceed knowing the risk exists. Low-impact scenarios, whether addressable or not, are documented and monitored but do not affect the go/no-go decision.[8]
8. Step 5: Build the Evidence Action Plan and the Decision Threshold
The evidence action plan that emerges from the pre-mortem is the document that separates a pre-mortem that changes decisions from one that produces a retrospective comfort. The action plan specifies, for each high-impact, high-addressability failure scenario: the specific evidence action required to close the risk, the team responsible for executing it, the timeline for completion, and the decision threshold that the evidence action must meet to satisfy the go/no-go criterion.[3]
A concrete example: if the KnolPersona simulation identifies that NICE has consistently challenged PFS as the primary endpoint for analogous products in the indication and requested OS data or a validated surrogate correlation before accepting PFS as the basis for a positive recommendation, the evidence action is to identify whether a validated surrogate correlation exists in the published literature and, if so, to ensure it is pre-specified in the Phase III protocol. The decision threshold is that the surrogate validation evidence must be sufficient to satisfy the NICE DSU's technical evidence standard for surrogate acceptance in this indication before the Phase III protocol is finalised.[9]
The discipline of stating specific decision thresholds before the budget is committed is what prevents the common failure mode of pre-mortems: identifying failure scenarios clearly, taking no specific action in response, and committing the budget anyway because no one ever defined what evidence would be sufficient to close each risk.[4]
9. How Fast Can Your Team Run an Evidence-Grounded Pre-Mortem with KnolAI and KnolPersona?
A manual pre-mortem for a Phase III go decision, conducted using the team's existing knowledge and available literature, typically requires one to two weeks of preparation to assemble the clinical precedent, competitive landscape, HTA history, and regulatory precedent that the failure scenario generation needs. The pre-mortem session itself requires a half-day to a full day. KnolAI and KnolPersona compress the preparation to two to three days, allowing the full pre-mortem process to be completed within a single week.[9]
Day 1 to 2, Intelligence brief generated by KnolAI: KnolAI generates a structured evidence intelligence brief covering all five failure mode categories from the Knolens knowledge layer. Clinical precedent, competitive landscape, HTA precedent, regulatory history, and operational risk intelligence are available as a sourced, structured document by end of day two. The brief includes specific precedent cases including analogous programmes that failed at each stage and the specific evidence gap that produced the failure.
Day 3, KnolPersona assessor simulation completed: KnolPersona runs the assessor challenge simulation against the current evidence architecture for NICE, G-BA, and JCA simultaneously. The output is a structured challenge report identifying the HTA failure scenarios with the highest precedent support, grounded in documented assessor behaviour from the Knolens knowledge graph.[9]
Day 4 to 5, Pre-mortem session and evidence action plan: The clinical development, HEOR, regulatory, and market access team conducts the pre-mortem session with the KnolAI intelligence brief and the KnolPersona challenge report as the evidence foundation. The failure scenario prioritisation matrix is completed. The evidence action plan and decision thresholds are documented. The go/no-go recommendation is made on the basis of whether the high-impact, high-addressability scenarios have been closed to the required threshold or whether specific actions are required before budget commitment.
Conclusion
The pre-mortem is the most underused structured decision-making tool in pharmaceutical clinical development, given how clearly the evidence supports its value and how consequential the decisions it is designed to improve actually are. With Phase III costs ranging from $20 million to $100 million or more, Phase I to approval success rates at 6.7%, and nearly three-quarters of sponsors acknowledging their trial designs are more complex than necessary, the case for structured prospective failure analysis before committing Phase III budget is not theoretical. It is the difference between programmes that avoid failure modes that were visible in advance and programmes that encounter them after the point where they can no longer be addressed.
At Pienomial, we built KnolAI and KnolPersona to make evidence-grounded pharma scenario planning achievable at the speed and depth that clinical trial budget decisions require. The combination of KnolAI's multi-domain intelligence brief and KnolPersona's assessor simulation produces a pre-mortem evidence foundation in days that a manual process would require weeks to assemble, and grounded in the specific documented precedents that matter rather than in general industry intuition. [9]
CTA: See how KnolAI and KnolPersona power evidence-grounded pre-mortem analysis for clinical trial decisions. Book a demo with the Pienomial team today.
