A man with a rare blood disorder became well enough for a transplant

AI system

System: Every Cure’s early drug-repurposing system [3] [6] [7] [8] [9] [10]
Model: Every Cure’s early algorithm for scoring possible uses of drugs against diseases. The accounts do not identify which of its algorithms produced these scores. [3] [6] [7] [8] [9] [10]
Release: The January 2024 configuration has no published release identifier. [3] [6] [7] [8] [9] [10]
Application developer: Every Cure [3] [6] [10]
Role: Not assessed
Configuration: The system scored candidate drugs for POEMS syndrome in January 2024. Physicians reviewed the scores and chose the treatment. [3] [6] [7] [8] [9] [10]
Inputs and outputs: Not assessed
Autonomy: Not assessed
Workflow: The consulting physician recognized similarities between POEMS syndrome and multiple myeloma and reviewed the AI scores for drugs that were used for myeloma. He discussed treatment with the physician responsible for Coates, who assessed the risks and administered the drugs. [1] [2] [3] [4] [5] [6] [7] [8] [9]
Human oversight: Fajgenbaum recommended the regimen; Wayne Gao weighed the risks and administered it.
Relation to AI: Fajgenbaum said that he had considered three drugs used for multiple myeloma and that Every Cure’s AI algorithm also scored them highly for POEMS syndrome. In an earlier AAMC interview, he described reviewing the scores and recommending the drugs to Coates’s treating physician. Wayne Gao weighed the risks and administered the treatment. The AI therefore supported the physicians’ assessment of candidate drugs. The sources do not show that it independently devised the regimen or that the benefit would have been impossible without it. The Lancet Haematology commentary contains a brief vignette of a matching unnamed POEMS patient, explicitly labeled unpublished data. No detailed clinical case report, underlying medical records, or case-specific model output was located. The commentary names KGML-xDTD as one of several algorithms in Every Cure’s platform, but it does not establish which algorithm or release was used for Coates. [1] [2] [3] [4] [5] [6] [7] [8] [9]
AI contribution: A consulting physician considered three myeloma drugs and reviewed the high scores that Every Cure’s algorithm gave them for POEMS syndrome. The treating physician assessed the risks and administered them. The available accounts do not isolate the AI’s contribution to recovery. [1] [2] [3] [4] [5] [6] [7] [8] [9]

Benefits

Outcome identifiers: Not assessed
Benefit: Coates reported that he responded to treatment within one or two weeks and later became well enough for a stem cell transplant. [1] [2] [3] [4] [5] [6] [7] [8] [9]
Category: Improved Physical Health
Status: Not assessed
Beneficiary groups: People. [1] [2] [3] [4] [5] [6] [7] [8] [9]
Beneficiaries: Not assessed
Reach: Not assessed
Distribution and access: Not assessed
Duration: Not assessed
Expected?: Yes [1] [2] [3] [4] [5] [6] [7] [8] [9]
Costs and harms: Not assessed

Measurement

Studies: Not assessed
Study design: Not assessed
Measurement: Not assessed
Setting and sample: Not assessed
Measures: he responded within one or two weeks
Comparison: the treatments his doctors had tried, with hospice as the remaining option
Results: Reference: Coates was now entering a terminal stage: his heart and kidneys were failing, and fluid had to be drained from his belly several times a week. He was too sick to undergo a stem-cell transplant, which could save his life. Result: within days Coates started to improve. A few months later, he'd recovered sufficiently to receive a stem-cell transplant. He is now in remission.
Additional value from AI: Not assessed
Limitations: Not assessed
Denominators: Not assessed
Materials for replication: Not assessed

Money

Kind of value: Not assessed
Amount: Not assessed
Currency and prices: Not assessed
Period and unit: Not assessed
Realization: Not assessed
Recipients and payers: Not assessed
Gross or net: Not assessed
Included costs: Not assessed
Comparison: Not assessed
Method: Not assessed
Return and payback: Not assessed

People and places

Participants: Developer: Every Cure (co-founded by David Fajgenbaum, University of Pennsylvania). Operator: David Fajgenbaum (Every Cure / University of Pennsylvania); regimen administered by Wayne Gao, Coates's treating hematologist and oncologist.
Jurisdictions: Benefit: Renton, United States. Developer: United States.
Ownership and control: Not assessed
Dependencies: Not assessed
Cross-border connections: Not assessed
Geopolitical implications: Not assessed

Follow-up

Conditions for replication: Clinicians need to assess the patient’s condition, the drugs, and their risks. A high AI score does not establish suitability. In this instance, recovery also depended on a response sufficient to permit a later transplant. [1] [2] [3] [4] [5] [6] [7] [8] [9]
Replication attempts: Not assessed
Later findings: Every Cure’s 2025 annual report states that Coates remains in remission. In Penn Medicine’s November 2025 documentary, he says that full remission followed five months after his transplant. [1] [2] [3] [4] [5] [6] [7] [8] [9]
Availability: Not assessed
Institutional decision: Not assessed
Grounds: Not assessed
Action and result: Not assessed
Success criterion: Not assessed

Evidence

Source numbers

AIBD-0127-source-0001–0010 [1] [2] [3] [4] [5] [6] [7] [8] [9] [10]

Citations

Numbered references below. [1] [2] [3] [4] [5] [6] [7] [8] [9] [10]

Language

Not assessed

Source relationships

Independent, Developer, Relationship unreported, Operator [1] [2] [3] [4] [5] [6] [7] [8] [9] [10]

Original evidence

Original interviews provide testimony from Coates and his physicians. Every Cure, Penn Medicine, and ARPA-H later describe the same treatment. The accounts share witnesses and investigators rather than providing independent clinical confirmation. [1] [2] [3] [4] [5] [6] [7] [8] [9] [10]

Access and versions

Patient and physician interviews describe treatment and recovery. Detailed clinical records, drug doses, and the patient’s AI scores are unavailable. The accounts differ on the timing of treatment and recovery. [1] [2] [3] [4] [5] [6] [7] [8] [9] [10]

Editor’s note. The unnamed patients in the Lancet commentary, WHYY, and ARPA-H are matched to Coates by their disease, treatment, and circumstances (references 5, 8–9). ARPA-H gives a relative date that differs from Every Cure’s January 2024 chronology. The Lancet commentary labels its vignette unpublished data.

Supported findings

Not assessed

Evidence links

Not assessed

Evidence strength

Not assessed

Missing evidence

Not assessed

Grades

Assessed benefits: People. [1] [2] [3] [4] [5] [6] [7] [8] [9]
Depth: People: 5. [1] [2] [3] [4] [5] [6] [7] [8] [9]
Reasons: Coates was critically ill and too sick for a transplant. He reported that he responded to the drugs within one or two weeks, later received a transplant, and reached remission five months afterward. The benefit was life-preserving, with the contribution from AI inseparable from the rest of his care. [1] [2] [3] [4] [5] [6] [7] [8] [9]
Observed or intended: Not assessed
Rubric: Not assessed
Assessor and date: Not assessed
Alternative assessments: Not assessed
Fill: Independently reported [1] [2] [3] [4] [5] [6] [7] [8] [9] [10]

Entry details

Kind: Not assessed
Related entries: Not assessed
Field: Healthcare - rare disease / drug repurposing. Individual off-label treatment of a terminal POEMS syndrome patient.
Purpose: Every Cure ranks existing drugs for potential use against other diseases. Its early system helped a physician assess drugs that he had considered for Coates. [1] [2] [3] [4] [5] [6] [7] [8] [9]
Dates: Source published: 20 March 2025. Updated: 24 October 2026.
Stage: Not assessed
Version: SHA-256: fe0be1b2a614…

References

Other identifiers: oecd_aim: 2025-03-21-aa9f.
Category mappings: Not assessed
Brief: Not assessed
Revisions: 19 revisions. Latest: 24 October 2026.

Sources

News report Independent author

Morgan, K. (2025). Doctors Told Him He Was Going to Die. Then A.I. Saved His Life. The New York Times, March 20, 2025.

Original interviews with Coates, Fajgenbaum, and treating physician Wayne Gao support the reported condition, treatment decision, response within a week, later transplant, and remission. Opening patient narrative and final section, including the interview with treating physician Wayne Gao. Original English article read through the archived copy. Access copy

Original journalism with firsthand patient, treating physician, and researcher testimony. No medical records or clinical case report for Coates are supplied.
News report Independent author

Khullar, D. (2025). Can A.I. Find Cures for Untreatable Diseases—Using Drugs We Already Have? The New Yorker, Open Questions, 15 July 2025.

Separate interview with Fajgenbaum identifies the three drugs and clinical decision. Paragraph beginning “Last year, Fajgenbaum received an e-mail”.

Independent reporter, same underlying case and researcher; not a new medical assessment.
Press release Developer

Every Cure. Repurposing Portfolio, “Carfilzomib-cyclophosphamide-dexamethasone for POEMS syndrome.”

Developer specifies January 2024 and high algorithm scores alongside biological similarity. Carfilzomib-cyclophosphamide-dexamethasone for POEMS syndrome.

Developer account; its only case link returns to the NYT article.
Report Developer

Every Cure. 2025 Annual Report, “Joseph,” p. 15.

Later institutional account of remission. Printed p. 15, Joseph.

Every Cure account; patient testimony and developer report, not an independent clinical follow-up.
Report Authorship unassessed

Advanced Research Projects Agency for Health. Fiscal Year 2027 Congressional Justification for Appropriations Committees, 17, “Repurposing for rare disease.”

Names MATRIX and attributes a matching POEMS treatment to that toolkit. Printed p. 17, Repurposing for rare disease.

Funding agency’s retrospective account. The patient is unnamed and the relative date conflicts with the developer’s January 2024 chronology.
Other Developer

Penn Medicine. “Finding Hidden Cures | Dr. David Fajgenbaum & Joseph’s Story.” Video, 14:19, November 17, 2025, 8:26–9:04 and 9:52–11:25. Publisher-supplied English captions.

Original interviews and patient videos explain how human judgment and AI scores contributed, give Coates’s response timeline, and place full remission after transplantation. 6:44–8:16, failed prior treatment and hospice discussion; 8:26–9:04, knowledge graph, physician’s candidates, and AI scores; 9:52–10:00, response and transfer from CCU; 10:43–11:25, kidney recovery, transplantation, and later remission.

Produced by Penn Medicine, Fajgenbaum’s institution. Coates, Tara Theobald, and Fajgenbaum give firsthand testimony. Patient videos add contemporaneous testimony, but the film is not an independent clinical evaluation.
News report Deployer

Stacy Weiner. “This doctor saved his own life. Now he’s on a mission to save thousands more.” AAMCNews, September 24, 2024, response to “Has Every Cure had any successes yet?”

An interview published in September 2024 names Joseph, dates the appeal to January, and describes reviewing AI scores before recommending three myeloma drugs to his doctor. Response to “Has Every Cure had any successes yet?” and preceding discussion of Every Cure’s platform.

Original interview conducted by Stacy Weiner for AAMCNews. The relevant answer is Fajgenbaum’s firsthand account as the recommending physician and platform cofounder, not independent clinical verification.
Other Developer

David C. Fajgenbaum, Sally Nijim, Grant Mitchell, Matej Macak, Chris Bizon, Alexander Tropsha, and David Koslicki. “Pioneering a new field of computational pharmacophenomics to unlock the life-saving potential of existing medicines.” The Lancet Haematology 12, no. 2 (February 2025): e94–e96. Commentary, paragraph beginning “We have already begun to pilot these processes.” https://doi.org/10.1016/S2352-3026(24)00278-3.

The investigators describe reviewing platform scores for a refractory POEMS patient facing hospice and report nine months of remission at submission. They also describe the platform and identify KGML-xDTD as one of several algorithms. Paragraph beginning “We have already begun to pilot these processes”; platform description beginning “The Every Cure platform is built to integrate multiple knowledge sources”; reference 10. Access copy

Commentary by the platform’s investigators, with a brief clinical vignette explicitly labeled unpublished data. The patient is unnamed, so identification as Coates is inferred from the disease, hospice setting, drug class, and related firsthand accounts. It is not an independent study or a detailed clinical case report.
News report Independent author

Liz Tung. “UPenn professor receives $48.3 million to find new treatments in old drugs.” WHYY, March 10, 2024, opening POEMS account and “Harnessing the power of AI.”

An early original interview reports that a Washington POEMS patient had left intensive care and the hospital after the three-drug recommendation. It also names MATRIX and describes the platform as being in beta in March 2024. Opening six paragraphs and “Harnessing the power of AI,” especially the paragraph beginning “Currently, Every Cure’s matchmaking program is in beta mode.”

Original reporting by Liz Tung. The clinical claims come from Fajgenbaum, and the patient is unnamed. Identification as Coates is inferred from Washington state, POEMS, the timing, the hospice plan, and the myeloma-drug recommendation. This is additional early testimony from the same physician, not independent clinical replication.
News report Developer

Ann-Marie Roche. “Elsevier and Every Cure: treating rare diseases with generic drugs — at scale.” Elsevier Connect, January 30, 2024.

Contemporaneous interview establishes that Every Cure was using several algorithms, including customized Elsevier algorithms. Opening description of the algorithms and responses about Elsevier’s contribution.

Technical partner’s interview with Fajgenbaum. It does not identify which algorithm supplied Coates’s scores or corroborate his clinical outcome.