The Eight-Unit Year: A Reconstruction of the Cordavex Launch

A hundred-unit forecast does not mean a hundred patients walked through the door and eight were chosen. It means the commercial team identified, with their fourteen primary transplant-centre relationships, roughly that many patients eligible for the therapy across the launch year. The actual count of patient cases opened in the company's coordination system between April 12, 2023, and the Chapter 11 filing was 120. The number that reached infusion was 8.

What happened to the other one hundred and twelve is the post-mortem. The company's case-management records, examined after the filing, place each loss in one of six categories. The largest is the simplest:

The funnel does not show a single point of failure. It shows four. The twenty-eight cases denied by a payer with no appeal launched are the prior-authorisation failure. The thirty-two cases abandoned after physician identification but before any order was placed are the case-coordination failure. The three units manufactured, shipped, and destroyed are the manufacturing-conditioning timing failure. And the entire population on this chart — every case, every centre, every payer — is the centre-activation failure: a launch that was supposed to be at sixty active transplant programmes by year-end and ended at fifteen.

What follows is each, in the order it happened.

Every Cordavex unit required individual prior authorisation. The product was a medical-benefit cell therapy: it did not flow through a pharmacy formulary, it did not run on PBM auto-adjudication. Each case opened a paper trail of roughly fifteen pages: a cover letter establishing medical necessity, a clinical summary mapping the patient's history to the payer's coverage criteria, citations of the relevant Coverage Policy Bulletin — Aetna’s CPB-1000 for cord blood transplantation, Cigna’s PHC-1183, UnitedHealthcare’s MCP-1066 — and an appeal template, pre-drafted, for the first denial.

The packages were not hard to assemble. They were repetitive. Eighty per cent of each document was boilerplate: the FDA approval basis, the Phase 3 data, the cost-benefit framing against extended neutropenic hospitalisation, the REMS attestation. The remaining twenty per cent was patient-specific (diagnosis codes, prior therapy lines, performance status) and payer-specific (the policy citation, the appeal window). The company employed two case managers. Each package took, by the team’s own time-tracking, between seven and eleven hours.

By the end of July, thirty-five cases were open. Six had received a payer decision. The other twenty-nine were either waiting or had been denied without yet having an appeal drafted. Mr. Reyes — the launch lead at the time, since departed — wrote in the weekly tracker, “PA throughput is our binding constraint. Asking finance for two more case-manager FTEs.”

The two FTEs were not approved before September. Of the twenty-eight cases that would, by the year’s end, be classified as “denied with no appeal launched,” nineteen entered that state in the Q2 window. Each represented a thirty-day appeal clock that quietly expired while a case manager was working on a different case.

The first manufactured-and-destroyed event was case CDV-2023-0010, an adult acute myeloid leukemia patient at a mid-tier academic centre. The case had passed prior authorisation on a fast track in the first week of September. Cord blood unit selection completed on Day 4. Manufacturing initiated on Day 6. The expected arrival at the centre was Day 27 — the standard twenty-one-day manufacturing window plus a one-day shipping buffer. The centre’s BMT scheduler, who was new to the role, booked the patient into conditioning on Day 22. Conditioning began on schedule. The manufactured unit landed at the centre on Day 27, five days past the seventy-two-hour viability window after conditioning. The unit was destroyed on arrival.

The seventy-two-hour viability window was a known constraint. It was on the label. It was in the FDA-approved REMS materials. It was in the centre training packet Helika’s clinical liaison had delivered on activation. What was not — anywhere — was a single calendar that connected manufacturing-completion to centre-conditioning. Manufacturing was tracked inside the QC system at Helika. Conditioning was tracked inside the centre’s BMT scheduling tool. Shipping was tracked by the logistics vendor. The three systems did not talk to each other except through email. Three hundred and fifteen thousand dollars of cost was destroyed because nobody owned the calendar.

Two further manufactured-and-destroyed events followed, one in late October and one in mid-November. Each was a different centre, a different proximate cause — a centre infectious-disease consultation flagged late, a patient-side dose intolerance during conditioning that the manufacturing team did not learn about until ten days after — and the same root cause: no end-to-end calendar owned by anyone. The total cost of destroyed units across the launch year, by the company’s post-bankruptcy reconciliation, was nine hundred and forty-five thousand dollars.

The activation plan, drafted in March 2023, named sixty target centres. The criterion was “FACT-accredited adult or combined-programme transplant centres with annual allogeneic volume above one hundred.” The CIBMTR directory listed seventy-three such centres in the United States and Canada. The plan was to activate, on average, five per month through year one. By October 1, fifteen were activated. By December 31, the figure had moved to sixteen.

The four-person account team had visited or called all sixty target centres. None of those visits had been built on a per-centre intelligence brief. The Centre for International Blood and Marrow Transplant Research publishes annual activity reports listing each centre’s transplant volume, donor-source mix, and disease-area focus. ClinicalTrials.gov lists each centre’s active trial portfolio and the principal investigators running them. PubMed lists each transplant director’s recent publications. All three sources are public, free, and queryable. The Helika commercial team did not have, by the bankruptcy date, a workflow for synthesising them.

What this looked like, in practice: an account manager would arrive at a transplant centre programme office for a thirty-minute meeting with the director, having read the centre’s website and the FDA-approved label. The director, in those thirty minutes, would ask three questions. What is your neutrophil-engraftment data versus my current cord-blood unit source? What is the time-to-clinical-decision on your prior authorisation? Have you seen any data in severe aplastic anemia? The answers existed — in the Phase 3 publication, in the case-management records, in the SAA pivotal trial in progress at four centres including, in two cases, the very centre the account manager was sitting in. The account manager did not have those answers at the meeting. By the time the follow-up brief landed in the director’s inbox, three weeks later, the meeting’s receptivity had cooled. A small handful of centres followed up regardless. Most did not.

For all the operational failure, eight cases moved through the entire coordination process — identification, prior authorisation, cord blood unit selection, manufacturing, shipping, conditioning, infusion, engraftment — and resulted in a successful transplant. They are recorded, anonymised, in the company’s case-management system:

Each of these eight cases represented the version of the launch that the plan had assumed would happen ninety-two more times. Each one was managed, in the company’s after-action accounting, by one of two case managers who happened to have bandwidth on the day the case opened. The variable that distinguished a case that infused from one that did not was not, in most instances, clinical eligibility or payer rigour. It was workload.

The diagnosis, on the evidence of a year, is not that any single system failed. Each system did what it was built to do. The CIBMTR file was published, in March 2023 and again in March 2024, on schedule. The ClinicalTrials.gov registry accepted every recruiting protocol. PubMed indexed every publication. The payers’ Coverage Policy Bulletins were available on every payer’s public-facing website, with their CPT codes and ICD-10 mappings and appeal windows. The doctors identified patients. The patients arrived.

What failed was the synthesis. Four operational systems — case management, manufacturing logistics, centre intelligence, prior-authorisation drafting — each operated by a different person on a different cadence, held four fragments of one launch. The two case managers had the queue, read daily, no way to flag which appeals were on a closing clock. The manufacturing team had the QC system, no view of conditioning schedules at the receiving end. The four-person account team had cold-call schedules, no per-centre intelligence brief. The medical-affairs team had the payer policies, no workflow for delivering them into case-specific PA packages. Each of the four people knew everything they were supposed to know. None of them knew enough to act in time.

The cost, in the industry, is well-documented. Cell-therapy launches disproportionately fail their first-year volume targets — Pharmaceutical Commerce reported in 2024 that fewer than thirty per cent of new cell-therapy products hit their year-one plan, citing per-case coordination overhead and inadequate centre activation strategy as the two largest contributors. ^[†]Pharmaceutical Commerce, ‘Why Cell and Gene Therapy Launches Fail Their First Year,’ 2024. Neither cause is, strictly, a failure of clinical knowledge or strategic judgement. They are failures of attention, distributed across a workforce that was, in every individual instance, paying perfectly good attention to what was in front of them.

Helika filed for Chapter 11 protection on February 7, 2024. The case was assigned to the United States Bankruptcy Court for the District of Delaware. Of the eighty employees on the payroll at filing, fifty-two were retained through the proceeding; the remaining twenty-eight were severed by mid-March. The two case managers, both of whom had asked for additional headcount in the third quarter of 2023, were among the retained.

The company emerged from bankruptcy in June 2024 under new ownership. A private-credit firm provided the post-petition financing and the equity recap that followed. The board hired a new chief executive — a turnaround specialist with a prior small-biotech launch-fix in his record — and a new chief commercial officer with explicit responsibility for “repairing launch operations.” The new commercial plan, dated July 2024, names each of the four operational failures described above and assigns to each a single accountable executive. The PA drafting work was scoped, in the same plan, as a templating-and-tooling project rather than a case-manager-FTE problem. The centre activation work was scoped as a research-and-brief workflow rather than a cold-call cadence. The manufacturing-conditioning calendar was assigned to a single supply-chain director, with read access to both the centre BMT schedulers and the in-house QC system. By the end of 2025, Helika reported, in a press release accompanying its second indication approval (refractory severe aplastic anemia), that the year-two-of-restart unit volume was running at three times the year-one rate.

The eight units delivered in 2023 — to the eight patients listed in the table above — represent, by any honest accounting, the eight successes of a failed launch. They also represent the proof, if proof were needed, that the science worked. The launch did not. None of the people involved had been bad at their jobs. They had been operating four systems that did not connect.

For the record, the timeline below tracks what was known, in which system, by which person, on which day across the ten months of Cordavex’s first year on the market — from FDA approval to Chapter 11 filing. The events are ones that, with hindsight, could have been assembled into a single picture by any observer with simultaneous access to all four files. No one had that access. By the time someone did, on the morning of February 7, 2024, the company was already in court.