Organic food is one of the few categories where the marketing claim and the research question are almost identical: people who choose it are different from people who don’t. That is the whole problem, and one of the two studies below is about it directly.
The headline, stated exactly
Baudry and colleagues’ NutriNet-Santé cohort, in JAMA Internal Medicine 2018, is the study the internet cites. Sixty-eight thousand nine hundred and forty-six French adults, 78.0 per cent female, mean age at baseline 44.2 years with a standard deviation of 14.5. For sixteen products, participants reported how often they ate labelled organic food — never, occasionally or most of the time — and those answers were compiled into an organic food score running from 0 to 32. Follow-up ran from 10 May 2009 to 30 November 2016.
One thousand three hundred and forty first incident cancers, of which 459 breast, 180 prostate, 135 skin, 99 colorectal, 47 non-Hodgkin lymphomas.
The result: high organic food scores were inversely associated with overall cancer risk, hazard ratio 0.75 for quartile 4 versus quartile 1 (95% CI 0.63 to 0.88, P for trend = 0.001), absolute risk reduction 0.6 per cent, hazard ratio 0.92 per five-point increase (95% CI 0.88 to 0.96).
Now read those two numbers the way the authors wrote them. A relative risk of 0.75 sounds dramatic. The absolute risk reduction is six tenths of one per cent. And the conclusion does not claim an effect on cancer risk at all: it says a higher frequency of organic food consumption was associated with a reduced risk of cancer, and that if these findings are confirmed, further research is necessary to determine the underlying factors involved.
One disclosure belongs in the reading too: one author reported serving since 2018 as a scientific expert, with no honoraria or personal funding, in two nonprofit foundations recently funded in France. That does not invalidate anything. It is part of the paper.
The study that explains the problem
Andersen and colleagues’ analysis of the Danish Diet, Cancer and Health cohort, published in Public Health Nutrition 2022, is not about cancer. It is about who buys organic food, and it is more useful for this question than any cancer study.
Forty-three thousand two hundred and nine men and women aged 54 to 73 completed detailed questionnaires between 1999 and 2002. Fifteen per cent reported never consuming organic food, 39 per cent low consumption, 37 per cent medium and 10 per cent high.
The relative likelihood of consuming organic food rather than never was highest among women, among people with a BMI below 25 kg/m2, among those with low alcohol intake, among those participating in sports, among non-smokers or former smokers, and among those who adhered to the Danish national dietary guidelines.
So: women, leaner, drinking less, exercising, not smoking, following the national dietary guidelines. Every one of those is independently associated with lower cancer risk. A cohort in which the organic buyers are systematically all of the above will produce an organic food score that predicts cancer risk even if the food itself did nothing at all.
The authors’ conclusion is a sentence every nutrition researcher should have stapled to their own paper: these findings have to be considered in the adjustment strategy for future studies linking organic food consumption with health outcomes.
What is left to test, and whether anyone tested it
Jadotte, Koos and Lane’s 2021 protocol in JBI Evidence Synthesis sets out what a proper synthesis would need: studies comparing organic with conventional food consumption and measuring cancer incidence in adults. Registered as PROSPERO CRD42019126224.
The inclusion criteria include one detail that explains the entire state of the field. Studies are included from 2009 or later, because 2009 was the date of a previous systematic review on organic food consumption and health outcomes — and that earlier review found no studies with cancer-related outcomes.
Read that again. Before 2009 there was nothing to review. This is not a case of one good cohort being ignored; it is a case of an evidence base that did not exist.
The comparison nobody has run
There is no randomised trial here, and there cannot easily be one — you cannot assign a shopping basket to sixty thousand households for a decade. What would move the question is not a bigger observational cohort with the same confounding. It is a study that holds the lifestyle constant and varies the food: a feeding trial with matched produce, or a comparison within a single population that adjusts for the Danish-style covariates that Andersen and colleagues listed.
Until that exists, the NutriNet-Santé result describes people, not apples.
What this evidence does not show
It does not show that organic food does not reduce cancer risk — the association is real in that cohort and the authors call for confirmation, not dismissal. It does not show that the 0.75 is wrong, only that a consumer group that differs on seven health behaviours at once cannot isolate the food. And the Danish study is cross-sectional: it describes who bought organic food between 1999 and 2002, not what happened to them afterwards.
This article summarises published cohort studies and a systematic review protocol. It is not medical advice and it is not a recommendation for any individual.
