#121 – Azra Raza, M.D.: Why we're losing the war on cancer

Raza trained on this regimen in 1977 and uses it unchanged today. The improvement in outcomes since 1977 is entirely attributable to better infection management, blood product availability, and supportive care protocols — not to the drugs themselves. She uses 7+3 as her canonical example of 'the local rather than the express stream, a few more stops' — a phrase from a 1961 Peter De Vries novel that described leukemia treatment at the time and remains equally accurate today.

Raza mentions the FLT3 mutation as one example of a truly targeted therapy — patients with FLT3 mutations may respond better to FLT3 inhibitors, especially when treated early rather than at stage four. This is consistent with her central thesis: targeted therapies may work much better against earlier disease when fewer co-mutations are present.

Raza credits Gleevec as a 'fantastic success' but deliberately contextualizes its population impact: CML is rare (~5,000 new cases/year in the US), it is essentially a pre-malignancy in chronic phase, and the same drug fails once the disease evolves additional mutations. The celebration of Gleevec as proof that targeted therapy can conquer all cancers is, in her view, a category error — it works because chronic CML has one target; common solid tumors have dozens of co-drivers and enormous intratumoral heterogeneity.

Raza credits this as one of oncology's genuine successes but pointedly notes that Western scientists did not derive it through rational drug design — Chinese clinicians were curing APL with 'vitamins and arsenic' for years before the West investigated. The molecular mechanism (RAR-alpha lesion) was worked out in retrospect after the clinical cure was already established. She uses this to challenge the narrative of bench-to-bedside translational medicine: 'scientists do not take credit in front of me for acute promyelocytic leukemia — you did not work it out, you only worked it out in retrospect.'

She references Toshiba's announcement of a 13-cancer detection chip using microfluidics on a single drop of blood in 4 hours for $180. Grail (acquired by Illumina) is in phase-2 validation of a cell-free DNA methylation-based multi-cancer early detection test. The M-chip she holds up is FDA-approved for PSA monitoring — she proposes building a 'barcode' of 50+ cancer-specific biomarkers for which each cancer gets its own lane, giving daily or weekly tracking of a panel that would detect each major cancer type at the earliest possible molecular signal. She also calls for research using every biological compartment: blood, sweat, tears, saliva, urine, microbiome — all are underexplored.

Raza's bank is now intersecting with three converging technologies: whole-genome and whole-exome sequencing (to find every somatic mutation in every sample), single-cell RNA sequencing (to characterize the heterogeneity of each sample at cellular resolution), and high-dimensional proteomics/metabolomics. Together, these can trace the clonal evolution of a single patient's disease from MDS through AML transformation over 10+ years. The scientific question she is hunting: some MDS patients transform to AML in months; others live 15 years without transformation. What separates them? The answer is in the serial samples.

Raza describes Umar — her best friend's son, 38 years old, Oxford and Columbia graduate, newly engaged, who presented with osteogenic sarcoma with microscopic vascular invasion. From the moment that pathology came back, she knew: 'his chances of survival beyond a couple of years at most are 0.00.' The choice was: die of cancer (painful) or attempt treatment knowing the treatment itself may kill him first. She uses this to make two points: (1) for patients who are at this stage, the framing of '3.7 months median improvement in survival' as a triumph is a moral obscenity, and (2) the only real path out is to find Umar's cancer before it reaches that stage. She cites Emily Dickinson: 'tell all the truth but tell it slant / success in circuit lies.'

Attia corroborates: his reading of the literature is that lifestyle modifications (exercise, diet, weight management, no tobacco) could reduce cancer risk by 30–50%, making it a high-leverage intervention — especially when combined with the early detection framework. Raza's personal practice of daily rebounding is based on her understanding that lymphatic circulation, unlike cardiovascular circulation, has no dedicated pump and depends entirely on physical movement. This detail is offered as personal practice, not as a clinical recommendation with evidence-grade backing.

Raza's analysis: the incidence spike from the 1940s–1970s was driven by tobacco. The 'decline' reverses that spike. The moment you control for smoking, the underlying survival benefit from 50 years of drug development is essentially flat for the solid-tumor metastatic diseases — lung, pancreas, colorectal, ovarian — that account for the majority of cancer deaths. She specifically distinguishes this from real progress in hematologic malignancies (Hodgkin's, CML via Gleevec, APL via ATRA), which she credits as genuine victories but notes together account for less than 10% of total cancer deaths.

The recently FDA-approved liposomal encapsulation of 7+3 (CPX-351/Vyxeos) is Raza's primary illustration of the economics problem: it is the identical drug combination inside a fatty capsule, approved on a median overall survival improvement of 3.7 months in cherry-picked trial patients, at a cost of $45,000 per cycle versus $5,000 for the standard formulation. The eligibility criteria for the clinical trial require normal kidney, liver, and lung function — selecting the best-prognosis subset — and even in that group only ~30% experienced the 3.7-month benefit (which is a median, meaning half did not even reach it). Physicians who do not prescribe the approved version risk malpractice liability.

Raza says she listens to CAR-T YouTube lectures every morning while rebounding on her mini-trampoline and has never heard a single serious researcher in a public forum acknowledge the organ-destruction problem. The exception is B-cell cancers: CD19-targeted CAR-T kills all CD19-expressing B cells, but B-cell aplasia is manageable with IVIG. Attempt the same approach on pancreatic cancer (e.g., mesothelin-targeted CAR-T) and you destroy the whole pancreas, a non-replaceable organ. The case-study reports of dramatic single-patient responses in cholangiocarcinoma, pancreatic, and colon cancer represent individually engineered neoantigens — important as proof-of-concept but not scalable to population-level oncology with the current technology.

Raza describes a PhD student who wrote to her after reading 'The First Cell': she was testing breast cancer drugs against 2D and 3D in-vitro cell systems, then planning to move to the mouse model, then to human trials. Raza's response: your cell lines are no longer genetically representative of breast cancer; your mouse model uses immunocompromised animals that don't spontaneously develop the tumor; you are testing drug efficacy in a system that is twice removed from the disease you want to treat. The student asked 'should I spend the next 10 years of my life doing this?' — which captures the human cost of the flawed paradigm.

Raza's argument is that the predictive value of animal safety studies for human toxicity is very low — 'what you learn in animals by using drugs is only correct for animals.' She would not use animals for safety either, though she distinguishes this from basic biology research where animal models remain valuable. The practical implication is a dramatic compression of the IND-to-first-human timeline and a shift of resources away from the pre-clinical phase that has consumed huge fractions of the $2.5 billion average drug development cost.

The scientific question the repository is designed to answer: what separates a pre-leukemia (MDS) that transforms to AML within months from one that takes 15 years? If you can find the earliest molecular signatures of the about-to-transform cell — the 'first cell' — you can identify the disease at the one-million-cell stage rather than the one-billion-cell stage. The repository spans the entire arc from MDS diagnosis through remission, relapse, and transformation, with matched germline controls on every patient. Raza calls it 'a national treasure' that has been impossible to fully analyze until recently because the technology did not exist to interrogate it at scale.