Table 1

Comparison of the rating criteria between GRADE and NutriGRADE for assessing evidence from systematic reviews and meta-analyses1

GRADENutriGRADE
RCTsObservational studiesRCTsObservational studies
Assessment criteria
 Risk of bias1) Random sequence generation; 2) Allocation concealment; 3) Blinding of participants and researchers; 4) Blinding of outcome assessment; 5) Incomplete outcome data; 6) Selective reporting; 7) Other biases1) Exposed and nonexposed cohorts from same population; 2) Confidence in exposure assessment; 3) Outcome of interest not present at start of study; 4) Match or adjusted for confounding variables; 5) Confidence in assessment of confounding factors; 6) Confidence in assessment of outcome; 7) Follow-up adequateSame as GRADE, except blinding of participants and researchers cannot readily be achieved1) Ascertainment of exposure (criteria specific to type of dietary assessment, i.e., validated food-frequency questionnaire, diet-associated biomarkers); 2) Basic and outcome-relevant statistical adjustments; 3) Assessment of outcome; 4) Adequacy of follow-up
 PrecisionNumber of cases (events), sample size, and inspection of the 95% CIs around the best estimate of the absolute effectSame as for RCTsSimilar to GRADE, additional point awarded for 1) 400–2,000 participants, but 95% CI excludes the null value; 2) >2000 participantsSimilar to GRADE, additional point awarded for 1) ≥500 events and the 95% CI excludes the null value; 2) ≥500 events, but 95% CI overlaps the null value (i.e., CI includes RR of 1.0), and 95% CI excludes important benefit (RR of <0.8) or harm (RR of >1.2)
 ConsistencySimilarity of point estimates and the overlap of their confidence intervals, as well as statistical criteria for heterogeneity (e.g., I2 and Cochrane Q test)Same as for RCTsSame as GRADESame as GRADE
 DirectnessNo important differences in the population or intervention; hard clinical outcome (vs. surrogate marker)Same as for RCTsSame as GRADESame as GRADE
 Publication  bias1) Number of studies; 2) Evidence for publication bias detected using funnel plots or statistical tests (e.g., Egger, Begg-Mazumdar)Same as for RCTsSame as GRADESame as GRADE
 Funding biasN/AN/AHigher points awarded for academic or research institutions, without industry funding or significant conflicts of interestSame as for RCTs
Criteria for upgradingN/A1) Large magnitude of effect (+1 for RR >2 or <0.5 or +2 for RR >5 or <0.2, in situations with an effect); 2) Dose-response gradient; 3) Adjustment for any potential residual confounders would only strengthen the observed associationThese criteria are imbedded in the NutriGRADE scoring system, including for moderate (+1 for RR or HR <0.80–0.50 and >1.20–2 comparing extreme quantiles) or large effect size (+2 for RR or HR <0.5 or >2 comparing extreme quantiles) and dose-response analysisSame as for RCTs
Criteria for downgrading1) Risk of bias; 2) Imprecision; 3) Inconsistency; 4) Indirectness; 5) Publication biasSame as for RCTsThese criteria are imbedded in the NutriGRADE scoring system, including for moderate (+1 for RR or HR <0.80–0.50 and >1.20–2 comparing extreme quantiles) or large effect size (+2 for RR or HR <0.5 or >2 comparing extreme quantiles) and dose-response analysisSame as for RCTs
Overall assessmentInitially rated as high, downgrade per above criteriaInitially rated as low, upgrade or downgrade per above criteriaInitial rating does not depend explicitly on study design, only on the overall score: 1) ≥8 points: high; 2) 6–7.99 points: moderate; 3) 4–5.99 points: low; 4) 0–3.99 points: very lowSame as for RCTs
  • HR, hazard ratio; N/A, not applicable; RR, relative risk.

  • 1 Original reference article for GRADE: (26); and for NutriGRADE: (33).