The use of the GRADE dose-response gradient domain in nutrition evidence syntheses varies considerably

This study aimed to identify and describe the use of the Grading of Recommendations Assessment, Development and Evaluation (GRADE) dose-response gradient domain to upgrade the certainty of evidence (CoE) in nutrition systematic reviews (SRs).

We searched for SRs of observational studies of nutrition topics that used GRADE and upgraded the CoE of at least one outcome for a dose-response gradient or reported reasons for not upgrading.

Within eligible SRs (21/281), 123 of 371 outcomes were upgraded for a dose-response gradient. For 118 outcomes, the authors conducted linear dose-response analyses, and for 106 outcomes, the authors conducted nonlinear dose-response analyses. From these, 107 outcomes showed a statistically significant (P < 0.05) association in the linear dose-response model, and for 28 outcomes, the test for nonlinearity was statistically significant. The CoE for 0.8% of all outcomes was rated as high, 47.2% as moderate, 43.9% as low, and 8.1% as very low. Fifty-five percent of outcomes that were upgraded for a dose-response gradient were already downgraded for at least one domain. This is contrary to GRADE guidance.

The approach for rating up the CoE for dose-response relationship is not consistent in nutrition reviews, likely because of a lack of clear guidance for when and how to do it. Therefore, more comprehensive GRADE guidance is necessary to enhance the correct use and comparability of dose-response upgrading.