Is it true that a low-fat diet protects against breast cancer?
A low-fat diet offers no clearly proven protection against breast cancer; the type of fat appears to matter far more than the total amount, so it is better to focus on eating less saturated fat and more healthy fats such as olive oil.
The idea that a low-fat diet protects against breast cancer is a popular assumption, but the evidence for it is weak and partly contradicted. Some reviews find a modest protective effect of eating less fat, but the underlying studies are inconsistent and of varying quality. Total fat intake on its own was not found to have a significant link with breast cancer mortality in more detailed research.
One specific type of fat does stand out, however: women with the highest intake of saturated fat had a 51% greater chance of dying from breast cancer than women with the lowest intake. That is a meaningful difference, although it is an observational association rather than a proven causal relationship. Eating less saturated fat therefore seems more worthwhile than simply eating 'less fat' in general.
It is precisely here that the PREDIMED study, a randomised trial, offers a surprising counterpoint. Participants who followed a Mediterranean diet with extra olive oil, in other words a high-fat diet, developed invasive breast cancer 68% less often than the group eating a low-fat diet. This is a secondary analysis based on only 35 cases, and the result needs confirmation in larger research. It does show, however, that eating less fat is not necessarily the best approach: the type of fat matters.
After a breast cancer diagnosis, a low-fat diet does appear to be beneficial. Several studies combined show that women who ate a low-fat diet after their diagnosis had a 23% lower chance of relapse, and overall mortality was possibly 17% lower. Those figures are encouraging, but here too the research is limited and firm conclusions are premature.
Based on systematic reviews and meta-analyses (PMID 22948801, 12737717, 39710959, 25692500, 24606431) and the PREDIMED randomised trial (PMID 26365989). All associations are observational or derived from secondary analyses, except for PREDIMED. The numbers in the PREDIMED sub-analysis are small (n=35 cases).