INDEPENDENT THINKING. WHOLE-FOOD LIVING.NO ANIMALS / NO OIL
THENANODIET
← Oncology / Health Desk

Inside a chemotherapy trial, diet was still observational

A stage III colon cancer analysis illustrates why the randomized label must stay with the assigned treatment.

Whole-food ingredients prepared without added oil

Illustrative food image, not a photograph of the study intervention.

Randomized treatment, reported diet

Meyerhardt and colleagues studied 1,009 patients with stage III colon cancer enrolled in the CALGB 89803 adjuvant chemotherapy trial. Although the parent study randomized chemotherapy, the dietary analysis was prospective and observational. Patients reported intake during chemotherapy and again six months afterward using a food-frequency questionnaire. Researchers used factor analysis to identify two dietary patterns, labeled prudent and Western. No group was randomly assigned to either eating pattern. Calling this a randomized diet trial would transfer the parent study's design to an exposure that was never randomized, changing what the result can establish.

Meyerhardt JA, Niedzwiecki D, Hollis D, Saltz LB et al.. JAMA, 2007. PMID 17699009 · DOI

What the pattern labels contained

The prudent pattern featured higher intakes of fruits, vegetables, poultry, and fish. The Western pattern featured meat, fat, refined grains, and desserts. These were statistical groupings of reported foods, not controlled menus or a simple vegan-versus-nonvegan comparison. In particular, prudent should not be relabeled as exclusively plant-based. The analysis asked how differing pattern scores related to outcomes after diagnosis. It did not identify a single harmful ingredient by experimentally removing it, and the pattern labels cannot assign responsibility for the association to one food while holding everything else constant.

Meyerhardt JA, Niedzwiecki D, Hollis D, Saltz LB et al.. JAMA, 2007. PMID 17699009 · DOI

The composite outcome needs its name

During a median 5.3 years of follow-up, 324 patients experienced recurrence; 223 died with recurrence, and 28 died without documented recurrence. Higher Western-pattern scores were associated with worse disease-free survival. Comparing the highest with the lowest fifth, the adjusted hazard ratio was 3.25, with a 95% confidence interval of 2.04 to 5.19. Here disease-free survival concerned recurrence or death, so the estimate should not be described simply as a death hazard or as recurrence alone. These event categories overlap and should not be added together as though they were mutually exclusive groups.

Meyerhardt JA, Niedzwiecki D, Hollis D, Saltz LB et al.. JAMA, 2007. PMID 17699009 · DOI

A prognosis association is not a treatment effect

The association concerns patients with stage III disease treated with surgery and adjuvant chemotherapy, rather than primary prevention in healthy adults. Adjustment cannot fully exclude alternative explanations for differences between self-selected diets. The paper therefore cannot show that changing an individual's pattern will reverse the reported hazard ratio, or that a particular home meal improves survival. Its useful contribution is a prognosis hypothesis with a defined population and outcome. Keep the dietary exposure observational, keep the composite endpoint explicit, and resist describing this analysis as proof of a randomized nutritional treatment for colorectal cancer recurrence.

Meyerhardt JA, Niedzwiecki D, Hollis D, Saltz LB et al.. JAMA, 2007. PMID 17699009 · DOI