I thought this was interesting because of the general discussion about iron levels. chatGPT’s analysis is almost longer than the original paper.
chatGPT(6AstraMaxPaid):
The underlying trial provides an interesting signal that iron reduction might reduce cancer risk, but this opinion article overstates what it establishes. Its novelty is limited, and its headline estimates contain arithmetic errors and an invalid combination of overlapping outcomes.
The paper is Paul Pisarik’s 2025 article in Frontiers in Oncology: fonc-15-1695261.pdffonc-15-1695261.pdf.
The paper argues that periodically removing blood to reduce iron stores could substantially reduce cancer incidence and mortality. It presents no new patient data. Instead, it revisits the cancer findings of the FeAST randomised trial, originally published in 2008.
FeAST included 1,277 patients with peripheral arterial disease, averaging approximately 67 years old and almost entirely men. Participants were assigned to calibrated phlebotomy or a control group. Cancer outcomes were recorded prospectively, with average follow-up of 4.5 years. This randomised design is a substantial strength compared with simply observing that blood donors have better health. pubmed.ncbi.nlm.nih.gov
The opinion article reports that mean ferritin in the intervention group fell from approximately 122 to 80 ng/mL. Although the protocol targeted a trough ferritin of 25 ng/mL, that was not the average concentration achieved.
The following calculations use the event counts reported in the uploaded paper:
| Outcome | Iron reduction | Control | Absolute difference | Approximate number needed to treat |
|---|---|---|---|---|
| New visceral cancer diagnoses | 38/636, or 6.0% | 60/641, or 9.4% | 3.39 percentage points | 30 |
| Cancer deaths | 14/636, or 2.2% | 36/641, or 5.6% | 3.41 percentage points | 30 |
The number needed to treat, or NNT, estimates how many people would need the intervention to prevent one additional event. These are crude estimates over the observed follow-up, rather than precise estimates at a uniform follow-up time.
The reported hazard ratio for cancer incidence was 0.65, with a 95% confidence interval of 0.43 to 0.97 and p = 0.036. The article also cites reductions in cancer-specific mortality and all-cause mortality among participants who developed cancer.
The author extrapolates these findings to ten years, recommends blood donation for men with peripheral arterial disease, and suggests that the accepted ferritin reference range may need narrowing if the findings are confirmed.
The novelty is principally a numerical reframing of previously published evidence. The article adds NNT calculations, ten-year projections and a comparison with prostate cancer screening. It does not provide an independent replication, a newly demonstrated mechanism, a comparison of different ferritin targets, or evidence establishing benefits in the general population.
Its useful contribution is to draw attention to a potentially important intervention signal occurring at ferritin concentrations commonly considered normal. However, the cancer findings themselves were already published in 2008, and the article’s strongest new numerical claims are its least defensible.
My main criticisms concern the calculations, interpretation and clinical recommendations.
-
The headline figures are internally inconsistent.
The paper states that an NNT of 13 means 49 people benefit per 1,000 treated. In fact, an NNT of 13 corresponds to approximately 77 per 1,000. A benefit of 49 per 1,000 corresponds to an NNT of approximately 20.
It also describes 98 beneficiaries per 1,000 as an NNT of 7. That corresponds to an NNT of approximately 10.2, while an NNT of 7 corresponds to approximately 143 per 1,000.
These are substantive arithmetic errors in the quantities used to communicate the intervention’s effectiveness.
-
Cancer prevention and cancer-death prevention cannot simply be added as separate beneficiaries.
The paper combines the reduction in cancer diagnoses with the reduction in cancer deaths to derive its especially favourable overall NNT.
These outcomes overlap. Preventing a cancer that would have proved fatal can contribute to both reductions. For deaths arising from newly diagnosed cancers, a combined endpoint of “cancer diagnosis or cancer death” must count each person only once.
The article does not provide an analysis that resolves this overlap. Its combined NNT of 7 is therefore unsupported, independently of the arithmetic inconsistencies.
-
The ten-year projections are assumptions, not trial findings.
The author explicitly assumes that the absolute benefit continues accumulating linearly beyond the observed follow-up.
Even accepting that assumption, the raw trial proportions project approximately 75 to 76 fewer events per 1,000 at ten years for each outcome considered separately, rather than 49.
More fundamentally, the trial does not establish that effects continue linearly. Adherence, cancer risk and deaths from other causes can change over time. This matters particularly in an older population with vascular disease. A credible ten-year estimate would require longer follow-up or a justified survival model with uncertainty estimates.
-
The parent trial did not establish an overall survival benefit.
In the original FeAST report, all-cause mortality was 125/636 in the iron-reduction group versus 148/641 in controls. The hazard ratio was 0.85, with a 95% confidence interval of 0.67 to 1.08 and p = 0.17: not statistically significant. The principal cardiovascular composite outcome was also not significantly reduced. JAMA Network
This does not disprove a cancer-specific benefit. It does mean that lower mortality among participants who developed cancer must not be presented as proof that the intervention extended survival across the whole trial population.
The cancer-incidence finding also rests on just 98 diagnoses, with a confidence interval extending close to no effect. It warrants independent confirmation.
-
An achieved mean ferritin of 80 ng/mL does not establish an optimal target.
Participants were randomised to an intervention, not to maintaining ferritin at 25, 50, 80 or 120 ng/mL.
Consequently, the study cannot determine whether 80 ng/mL is optimal, whether lower concentrations improve outcomes further, or which starting concentrations predict benefit.
Similarly, a proposal to redefine the normal range as 20 to 100 ng/mL requires considerably more evidence. Laboratory reference intervals and treatment targets answer different questions.
-
The mechanistic language treats ferritin too simplistically.
The article refers to a “toxic effect of ferritin”. This conflates serum ferritin, stored iron and the chemically reactive labile iron pool.
Ferritin can protect cells by sequestering iron. Experimental manipulation of ferritin has demonstrated reductions in labile iron and resistance to oxidative injury. Serum ferritin also requires interpretation in relation to inflammation, something the original FeAST eligibility criteria explicitly recognised. pubmed.ncbi.nlm.nih.gov
Therefore, a benefit from iron removal would not by itself establish that ferritin protein is toxic. Nor does lower ferritin among older survivors prove that higher ferritin caused earlier deaths. The article supplies no direct measurements establishing the proposed mechanism.
-
The comparison with prostate cancer screening is not valid.
The claim that phlebotomy is approximately 140 times more effective compares an unsupported combined outcome with prevention of death from one particular cancer.
It also compares different populations, interventions and evidence bases. Such a ratio cannot establish comparative effectiveness, even though the author acknowledges that the comparison is imperfect.
-
The recommendation to donate blood goes beyond the evidence presented.
The author appropriately acknowledges that extension to the general population remains theoretical. Nevertheless, the recommendation for men with peripheral arterial disease is stronger than this additional analysis supports.
FeAST used a calibrated protocol in selected participants. A prevention recommendation also needs a balanced assessment of adverse effects. The large INTERVAL randomised trial found that more frequent donation increased iron deficiency and donation-related symptoms, although it studied substantially more frequent donation schedules than those described here. PubMed
The paper’s emphasis on benefits gives insufficient attention to this balance.
A useful next study would independently test cancer prevention as a prespecified primary objective, measure both benefits and harms, distinguish cancer incidence from survival after diagnosis, and compare clearly defined iron-reduction strategies. That would address the promising question raised by FeAST without relying on the unsupported ten-year claims in this article.