Why Do Nutrition Meta‑Analyses Often Misinterpret Effects?

When scientists study what we eat, they often need to account for total energy intake. Imagine comparing two recipes: one high in saturated fat and another rich in fish. Without adjusting for calories, you might attribute an effect to a food that actually stems from the overall energy change.\n\nThis is the core of a recent meta‑scientific review: many meta‑analyses linking saturated fat or fish to heart disease ignore whether the original studies adjusted for energy intake. Without this adjustment, results reflect "substitution effects"—what happens when you replace a portion of energy from one food with another—but most meta‑analyses treat them as direct effects.\n\nThe authors examined eight meta‑analyses, 82 primary studies, and 144 statistical models. Only one meta‑analysis explicitly checked the energy‑adjustment strategy; the others mixed incompatible estimates, producing vague conclusions.\n\nWhy does this matter? Nutritional guidelines rely on these syntheses. If the estimates are muddled, dietary advice can be misleading. Experts therefore recommend: (1) verify energy adjustment in original studies, (2) clearly distinguish substitution effects, and (3) report specific subgroup analyses. Only then will future meta‑analyses provide reliable, understandable recommendations for everyone.

Energy adjustment is a statistical practice that controls for total caloric intake variability among participants. In regression models, total energy intake can be included as a covariate (residual method) or the exposure expressed as a percentage of total energy (density method). Both methods transform the estimated effect into a "substitution effect": for example, a 5% increase in saturated fat at the expense of carbohydrates.\n\nThe issue arises when primary studies use different strategies or fail to report them, yet meta‑analyses pool coefficients without distinguishing the effect type. This results in an "apples‑and‑oranges" meta‑analysis, summing incomparable estimates.\n\nTo solve this, it is recommended to build a detailed database of adjustment strategies, classifying each model as (i) direct (no adjustment), (ii) residual, or (iii) density. Then, meta‑analyses should conduct separate subgroup analyses for each category, or convert all effects to a common substitution scale using transformation formulas based on energy balances. Additionally, confidence intervals and heterogeneity I² should be reported for each subgroup to assess result consistency.\n\nFinally, systematic review guidelines (e.g., PRISMA‑NUT) should include an explicit item on energy adjustment, requiring authors to describe the adopted strategy and justify the choice. With such transparency, the quality of nutritional evidence can improve, leading to more robust dietary recommendations.