[blithedale]

ARGH. ARGH ARGH ARGH.

Almost no one reads these papers when stuff like this gets posted on HN.

And there is shameful statistical fuckery afoot in this paper. People are only repeating the hype statistics at the top of the paper.

Their "pathological response" rates touted at the top of paper only come from the 1/3 of women able to actually carry out the diet. This is the "per-protocol" language in the abstract. An honest evaluation here would be "intention to treat" and analyze chemo responses in every assigned to each arm.

And they do that evaulation, but bury it later in their paper. Turns out when you account for the 2/3 of women who can't do that fasting, there's NO DIFFERENCE in response rate:

   "The overall pCR rate was 11.7% and did not differ between the two groups (10.8% in FMD group versus 12.7% in control group; OR 0.830, 95% CI 0.282–2.442, P = 0.735."

Argh...

[blithedale]

A lot of people might think "Ah, well, fasting is clearly effective for the 1/3 that are able to carry it out, just gotta make sure you have the willpower to be the 1/3."

But hold on. Slow down. There's a good chance the "women who fasted had better chemo responses" story completely REVERSES cause and effect. Here's how:

    WOMEN 1 - Has indolent biology, slower growing tumor than your average breast cancer at this stage.
 - Symptoms: Less pain, neuropathy, less swelling, tumors not affecting distant organs as much

    WOMEN 2 - Has aggressive biology, faster growing tumor than your average breast cancer at this stage.
 - Symptoms: More pain, neuropathy, more swelling, tumors begin to affect distant organs

So ask yourself: Which women is more likely to be able to follow a strictly regimented diet?

Clearly WOMEN 1.

[blithedale]

This is like making people with cancer do an Ironman triathalon.

People about to enter hospice and pass on in a few weeks are gonna be way less able to complete it than someone who has months left to live.

But all you've done is stratify your patients.

The triathalon doesn't make anyone live longer.

[markus92]

That's absolutely right and it's a major problem of this study. There is one caveat though, in the most common ER+ breast cancer, radiological response is more predictive of survival than pCR. And in this study, they do observe a better radiological response in the fasting group, even when looking at ITT (though just barely significant).