915-Megahertz Microwave Hyperthermia

My dear Weldon,

The dossier arrived on the 14th of May, and I have sat with it through two evenings at University College before writing to you. The finding: 915-megahertz microwave hyperthermia, applied adjunct to therapy in sarcoma, yields a 42 per cent survival benefit. Hazard ratio 0.58 for local progression-free survival in the Issels series, n=341. I did not want this result. You know my disposition toward the clinicians who wave physical agents at tumours and call it science. I expected the contingency table to collapse into noise. It did not.

State the null as one must: hyperthermia adds nothing to standard therapy. Test it. The chi-squared rejects at conventional thresholds, and the meta-analytic odds ratio of 2.67 for complete pathological response in cervical carcinoma points the same direction. Effect size alongside the p-value, always - and here the effect size is not trivial. A hazard ratio of 0.58 in n=341 is not a precise measurement of sample size. It is a finding.

Here is what troubles me, and what I write to you plainly. My resistance was itself a hypothesis, and I had never stated it, never tested it, never assigned it a threshold. I was prepared to commit the very sin I have spent thirty years prosecuting: to treat my prior conviction as the verdict rather than as the thing to be weighed. The sampling distribution is the court’s standard of evidence, and I had appointed myself judge before the evidence was called. The verdict came back against me. Not proven, I told myself at first - small samples, heterogeneous protocols, the usual caveats now attached to the dossier. But caveats are not acquittal. Failure to reject my scepticism would require power I never had, because I never designed the test.

So a correction, stated as a closing decision. When I began this examination I believed that a result unwelcome to the investigator carries a heavier presumption against itself. I now believe the opposite: that the presumption runs the other way, and that the investigator who dislikes a result owes it a stricter, not a laxer, arithmetic. The heating of tissue by microwaves is not my sort of science. The table does not care what sort of science I prefer. Catalogue it as confirmed, with caveats, and let my discomfort stand recorded as the error it was.

Yours ever, K.P.