onc brain

The longer read

The value of this report is not that active surveillance is safe, which fifteen-year data already established, but that it is the first dataset long enough to test whether the safety holds once the lead time of screen-detected disease has fully run out. That distinction matters because the standard reassurance offered in clinic rests on follow-up shorter than the natural history of the tumours being watched. Median lead time for screen-detected prostate cancer is variously put at five to twelve years, and small low-grade tumours plausibly longer, so a cohort followed fifteen years is describing a period in which many of these cancers had not yet had the opportunity to declare themselves. Extending to a median of eighteen years, with survival curves out to twenty-five, is the first look past that horizon.

What the extension shows is a divergence between the two endpoints that clinicians tend to conflate. Prostate cancer-specific survival barely moves: 97% at fifteen years, 94% at twenty-five. Failure-free survival, by contrast, falls from 81% to 68% over the same window and shows no inflection where it might safely be truncated. Nearly half of all failures occurred more than ten years after diagnosis. The practical implication is that the reassurance which holds for mortality does not transfer to curability, and the interval over which a man remains at risk of losing his window for definitive treatment is longer than most surveillance programmes are designed, funded, or staffed to cover.

The comparison to ProtecT is the one that should move confidence, and it moves it in a specific direction. ProtecT randomised, found a higher metastasis rate in the monitoring arm, and thereby established that surveillance carries a real cost in curability even when mortality is unaffected. This cohort is not randomised and so cannot quantify that cost, but it observes the same phenomenon over a longer horizon and in a more favourable population: eighteen men reached PSA relapse after radical prostatectomy or radiotherapy without yet starting hormones, and forty-four went onto hormonal therapy, mostly for symptoms. Those men were not overdiagnosed in any meaningful sense. The mortality comparison with PIVOT and SPCG-4 runs the other way and is largely an artefact of case mix rather than management: screen-detected disease is a different entity from clinically diagnosed disease, and the low mortality here should not be read as evidence that surveillance outperformed those trials' treatment arms.

Two features should temper how far the numbers travel. Sextant biopsy until 2009 and prediagnostic MRI in twenty-one of 488 men means baseline risk assignment was systematically softer than it would be today, which cuts both ways: some men labelled very low risk harboured undersampled higher-grade disease, inflating the apparent failure rate of favourable disease, while genuinely low-risk men would today be identified more confidently and surveilled with more assurance. The associate editor's argument that contemporary outcomes should be better is reasonable but untested. Second, the composite failure endpoint pools a man who died of prostate cancer with a man carrying a low untreated PSA relapse, and the 68% figure is doing work that its most severe components do not support on their own. For a radiation oncologist the operative question is narrower than the headline: whether the men arriving from surveillance years later are still candidates for definitive rather than salvage-intent treatment, and this cohort suggests that for a meaningful minority the answer is no.