An asset you already own is the cheapest shot on goal you will ever get. The toxicology package is done and the manufacturing line is standing. The safety database grows with every patient dosed, and the KOLs are already on the phone. Set that against the bill for starting over: DiMasi and colleagues put the out-of-pocket cost of a de novo programme at roughly $1,395 million per approved compound, and about $2,558 million once you capitalise the cost of failure and time (2013 dollars) [1]. So when a lead indication disappoints, reaching for a second disease, a re-indication, is not desperation. For a lean biotech betting its future on one or two assets, it can be the smart move.
Here is what the arithmetic hides. Every advantage you inherit, the tox, the CMC, the safety file, the relationships, answers a question nobody is actually asking. None of it tells you whether the drug will work in the next disease. And the reflex in most re-indication searches is to start at the wrong end: pick a promising new indication first, then reverse-engineer a mechanism story to justify it. That order is backwards.
Before you go looking for a good new disease, answer a harder question: why did it actually fail the first time?
Because a failure has a location. Some failures are local to the original disease, its population, its endpoint, its comparator, and relocating the molecule clears them completely. Other failures live in the molecule itself, and those follow it into every disease you try next. Run the wrong search and you inherit the exact failure you were trying to escape.
Think of it as a differential diagnosis you run on your own asset. Is the pathology local to one organ, so that moving the patient resolves it, or is it systemic and present wherever the patient goes? You would not transfer a patient between wards without answering that. Teams re-indicate assets without answering it all the time.
Consider fingolimod. Novartis first developed it as an immunosuppressant to prevent kidney-transplant rejection, and it lost, cleanly, to mycophenolate mofetil in that setting (NCT00099736, 696 patients, completed 2005) [2]. The reason recorded in the literature is that it offered no advantage over a strong incumbent: comparable efficacy but a less favourable tolerability profile, not a fundamental safety flaw in the drug [2]. That is a local failure. It says something about transplant medicine and about the comparator; it says almost nothing about the molecule's worth elsewhere.
Elsewhere turned out to be multiple sclerosis. The same sphingosine-1-phosphate-receptor modulation that underwhelmed against mycophenolate became the first oral disease-modifying therapy for relapsing-remitting MS, through TRANSFORMS (NCT00340834, 1,292 patients, against interferon β-1a) and FREEDOMS (NCT00289978, 1,272 patients, against placebo) [2]. The transplant trials wound down in 2005 and 2006; the MS pivotals began in 2006. Same asset, same sponsor, a near-immediate pivot. The failure stayed behind in the ward it belonged to.
Now tanezumab. In 2010 the FDA placed a partial clinical hold on the entire anti-NGF antibody class, after reports of rapidly progressive osteoarthritis and joint destruction [3]. That hold stopped Pfizer's osteoarthritis programme. In the same year, it also stopped the separately run chronic low back pain programme (NCT00924664, 849 enrolled, terminated) [3]. Two different diseases, one hold, because blocking nerve growth factor, which helps protect joints under mechanical load, had nothing to do with either specific disease and everything to do with the molecule. Pfizer later resumed both indications at lower doses (NCT02528188 in osteoarthritis, 3,021 patients; NCT02528253 in chronic low back pain, 1,832 patients) [3]. Tanezumab still holds no FDA approval in any indication.
There is the distinction the whole decision turns on. Fingolimod carried a local failure and walked away from it. Tanezumab carried a systemic one, and it followed the molecule into a disease that had nothing in common with the first except the drug. The failure travels.
Free download
The Rescue Readiness Checklist
A six-point diagnostic for a programme after an ambiguous readout: what's salvageable, and what to pull together before any rescue conversation.
Get the checklist →The pull toward the convenient indication is strong and mostly unspoken. You have sunk years and cash into this molecule, and sunk cost quietly reframes itself as conviction. Your KOLs already know the asset, so their suggestions arrive pre-vetted and flattering. The manufacturing is set up for this compound; the safety database is largest here. And there is almost always a mechanism story elegant enough to make the next disease sound obvious. The literature on portfolio decisions has names for some of this, champion bias and the "not-invented-here" reflex among them (Bieske et al. 2023) [4]. None of those forces is evidence. Each is a reason to start the search from the disease rather than the diagnosis.
The base rate does not help you resist; it reassures you instead. Krishnamurthy and colleagues identified 11,814 records and reviewed 115 papers on why drugs get abandoned; the commonest reason by a distance was lack of efficacy or superiority to other therapies (59 papers), followed by strategic and business reasons (35), safety (28), and trial-design decisions (12) [5]. Read that quickly and it sounds like permission: most abandonments are not fundamental asset-level failures, so the prize is real. That said, read it properly and it says something sharper. The base rate tells you the average failure is probably relocatable. It tells you nothing about whether yours is.
The clearest cautionary tale is intepirdine. GSK ran it through four failed trials; Roivant's Axovant licensed it for $5 million and bet on combining it with donepezil, without a fresh diagnosis of why the prior failures would not simply recur. The 1,300-patient MINDSET Phase 3 failed, and Axovant's stock fell roughly 70% on the readout [6]. The pivot appears as convenience, and the failure it ignored duly repeated.
These teams are rarely careless, and usually the opposite. The convenience pull is strongest precisely when the team is competent, invested, and under time pressure, which is to say, always.
Here is the process, and it costs you a week not a financing round. Run it before you screen a single new indication.
Step 1. Locate the break. Where in the causal chain did the first programme actually fail? An efficacy miss, a safety signal, or an exposure-response and PK problem? "The trial failed" is not a diagnosis. "We never reached the exposure the target needs, and at the exposure we did reach there was no separation from placebo" is.
Step 2. Ask whether the break is local or systemic. Would the same mechanism cause the same problem in the candidate disease? Tanezumab is the systemic case: nerve growth factor protects joints under load in any patient with joints, so the mechanism carries its liability into every musculoskeletal indication you try. Fingolimod is the local case: losing to mycophenolate in transplant told you about transplant, not about the molecule's behaviour in autoimmune neurology.
Two cautions belong inside this step, because "it failed" and "it doesn't work" are not the same statement.
The first is the diluted signal. Nelivaptan was described as showing statistically significant antidepressant efficacy in only one of two earlier Phase 2 trials, and was set aside. A 2025 reanalysis found the response in the active arm was bimodal: a high-responder subgroup with a mean change of roughly −17.14 and a low-responder subgroup at about −3.85 (Zu Eulenburg et al. 2025) [7]. That is not a drug that does not work. That is a real effect drowned in an unenriched population. The compound is back in Phase 2 with a genetic companion diagnostic. The failure was the population and the design, not the molecule.
The second is the endpoint. In the RUTH trial, raloxifene had no significant effect on its primary coronary endpoint (hazard ratio 0.95, 95% CI 0.84 to 1.07) and, in the same trial, significantly reduced invasive breast cancer (hazard ratio 0.56, 95% CI 0.38 to 0.83) (Barrett-Connor et al. 2006) [8]. One drug, one trial, two endpoints, opposite verdicts. Whether that molecule "failed" depends entirely on which line of the results table you read. Your diagnosis has to name the endpoint, not just the outcome.
Step 3. Separate the molecule from the mechanism. This is where re-indication decisions quietly go wrong. Torcetrapib is the cautionary illustration, though not itself a repositioning case: it was abandoned outright, and its failure traced to molecule-specific off-target effects rippling across roughly 200 plasma proteins, not to any flaw in CETP inhibition as a mechanism (Williams et al. 2018) [9]. The mechanism was fine; the molecule was not. Later, cleaner CETP inhibitors revived the approach, obicetrapib among them (Prajapathi et al. 2026) [10]. The practical test: if the failure was molecule-specific and you are still carrying that molecule, it is a systemic failure by another name, and it travels with you. A mechanism can be redeemed by a better molecule. Your molecule cannot be redeemed by a better mechanism story.
Step 4. Only now, screen candidate indications, at in-licensing rigour. Once the diagnosis clears the molecule, the new indication is a fresh asset decision, and it deserves the same discipline you would apply to a drug you were licensing in from a stranger: would you pay for it, and could you get paid for it? The lower bar teams apply to their own molecule, because it is already in the building, is exactly the bias to resist. Run the indication-prioritisation screen as if the asset were external. What an unstructured search produces at scale is not encouraging: Heijman and colleagues catalogued 52 authorised drugs repurposed for osteoarthritis across 220 studies and concluded that success across that literature "remains largely elusive", blaming underpowered, non-standardised designs and disease heterogeneity [11]. That is the landscape you are competing against when you screen casually, so screen it properly.
And do not expect a regulator to make you do any of this, the discipline is entirely self-imposed. Which is exactly why so few teams impose it.
Diagnose first, screen second. Every molecule carries its own history into the next disease you point it at, and the only question worth answering before you commit to that re-indication is which parts of that history are local and which are heritable.
Nobody will enforce this on you. There is no guidance to cite, no reviewer who will reject a search for skipping the step. The teams that run the differential before the readout, rather than reconstructing a rationale after it, are simply the ones who tend not to end up needing a programme rescue. That is usually where this diagnosis gets run in anger: under pressure, on a struggling asset, with the convenient pivot already halfway to the board. It is a great deal cheaper to run it a week before the KOL call sets the agenda.
Get the monthly digest
The 5 things evidence leads need to know each month: regulatory moves, RWE developments and what they mean in practice. No pitch, one email a month.
[1] DiMasi JA, Grabowski HG, Hansen RW. (2016). "Innovation in the pharmaceutical industry: New estimates of R&D costs." Journal of Health Economics;47:20-33. PMID: 26928437. https://pubmed.ncbi.nlm.nih.gov/26928437/
[2] Novartis. "Fingolimod (FTY720) in renal transplantation." ClinicalTrials.gov: NCT00099736. https://clinicaltrials.gov/study/NCT00099736 · Tedesco-Silva H, et al. (2006). "Randomized controlled trial of FTY720 versus MMF in de novo renal transplantation." Transplantation;82(12). PMID: 17198261. https://pubmed.ncbi.nlm.nih.gov/17198261/ (FTY720 2.5 mg non-inferior to MMF on the primary efficacy endpoint, but provided no benefit over standard care and carried a less favourable tolerability profile.) · Novartis. "TRANSFORMS — fingolimod vs interferon β-1a in relapsing-remitting MS." ClinicalTrials.gov: NCT00340834. https://clinicaltrials.gov/study/NCT00340834 · Novartis. "FREEDOMS — fingolimod vs placebo in relapsing-remitting MS." ClinicalTrials.gov: NCT00289978. https://clinicaltrials.gov/study/NCT00289978
[3] Pfizer. "Tanezumab in chronic low back pain." ClinicalTrials.gov: NCT00924664. https://clinicaltrials.gov/study/NCT00924664 · Pfizer. "Tanezumab in osteoarthritis (resumed programme)." ClinicalTrials.gov: NCT02528188. https://clinicaltrials.gov/study/NCT02528188 · Pfizer. "Tanezumab in chronic low back pain (resumed programme)." ClinicalTrials.gov: NCT02528253. https://clinicaltrials.gov/study/NCT02528253 (Anti-NGF osteoarthritis and chronic-low-back-pain trials were placed under an FDA partial clinical hold in 2010 over rapidly progressive osteoarthritis and joint destruction; the NCT00924664 record notes the hold halted dosing/enrolment on 19 July 2010. Pfizer and Eli Lilly discontinued the tanezumab programme in October 2021 after a 19–1 FDA advisory-committee vote against approval. https://www.fiercebiotech.com/biotech/lilly-pfizer-stop-development-osteoarthritis-drug-after-fda-rejection-extending-bleak-run)
[4] Bieske L, Zinner M, Dahlhausen F, Truebel H. (2023). "Trends, challenges, and success factors in pharmaceutical portfolio management: Cognitive biases in decision-making and their mitigating measures." Drug Discovery Today. PMID: 37572999. https://pubmed.ncbi.nlm.nih.gov/37572999/
[5] Krishnamurthy N, Grimshaw AA, Axson SA, Choe SH, Miller JE. (2022). "Drug repurposing: a systematic review on root causes, barriers and facilitators." BMC Health Services Research. PMID: 35906687. https://pubmed.ncbi.nlm.nih.gov/35906687/
[6] Axovant Sciences / intepirdine (RVT-101). MINDSET Phase 3: "A Phase 3, Double-blind, Randomized Study of RVT-101 Versus Placebo When Added to Existing Stable Donepezil Treatment in Subjects With Mild to Moderate Alzheimer's Disease." ClinicalTrials.gov: NCT02585934 (Axovant Sciences Ltd., 1,315 enrolled, added to donepezil). https://clinicaltrials.gov/study/NCT02585934 (GSK developed the compound as SB-742457 across multiple Phase 2 trials; licensing terms and the ~70% stock reaction are from trade-press reconstruction, presented as an industry case.)
[7] Zu Eulenburg C, et al. (2025). "Toward precision psychiatry: theoretical implications of bimodal response patterns to vasopressin V1b receptor inhibition in depression." Frontiers in Psychiatry. PMID: 41179820. https://pubmed.ncbi.nlm.nih.gov/41179820/
[8] Barrett-Connor E, et al. (2006). "Effects of raloxifene on cardiovascular events and breast cancer in postmenopausal women (RUTH)." New England Journal of Medicine. PMID: 16837676. https://pubmed.ncbi.nlm.nih.gov/16837676/
[9] Williams SA, et al. (2018). "Improving Assessment of Drug Safety Through Proteomics: Early Detection and Mechanistic Characterization of the Unforeseen Harmful Effects of Torcetrapib." Circulation. PMID: 28974520. https://pubmed.ncbi.nlm.nih.gov/28974520/
[10] Prajapathi D, et al. (2026). "From Failure to Promise: Obicetrapib and the Renaissance of Cholesteryl Ester Transfer Protein Inhibition in Atherosclerotic Cardiovascular Disease." Journal of the American Heart Association. PMID: 42017323. https://pubmed.ncbi.nlm.nih.gov/42017323/
[11] Heijman MWJ, et al. (2025). "The Potential of Authorised Drugs to be Repurposed for the Treatment of Osteoarthritis: A Scoping Review of Clinical Studies." Drugs. PMID: 41619152. https://pubmed.ncbi.nlm.nih.gov/41619152/
[12] FDA. "Applications Covered by Section 505(b)(2)." DRAFT guidance, 1999 (Federal Register notice of availability, 8 December 1999); not finalised as of this writing. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/applications-covered-section-505b2
[13] FDA. "Determining Whether to Submit an ANDA or a 505(b)(2) Application." FINAL guidance, May 2019. Docket FDA-2017-D-5974. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/determining-whether-submit-anda-or-505b2-application
[14] European Commission (STAMP) / EMA-HMA. "EU medicines repurposing pilot." Supports not-for-profit organisations and academia ("champions") in repurposing off-patent, already-authorised medicines for a new indication. https://www.ema.europa.eu/en/news/repurposing-authorised-medicines-pilot-support-not-profit-organisations-and-academia
[16] Ridker PM, et al. (2017). "Antiinflammatory therapy with canakinumab for atherosclerotic disease (CANTOS)." New England Journal of Medicine. PMID: 28845751. https://pubmed.ncbi.nlm.nih.gov/28845751/