Table 3.
GRADE assessment for the certainty of evidence.
| Study design | Outcome | No. of participants (studies) | Certainty of evidence (GRADE) | Relative effect RR (95% CI) | Risk with placebo/reference |
|---|---|---|---|---|---|
| Prospective cohort studies | Atopic diseases in offspring | 123,445 (8 studies) | Very lowᵃ | 0.88 (0.82–0.95) | 124 per 1,000 |
| Asthma | 33,205 (6 studies) | Low | 0.81 (0.68–0.96) | 137 per 1,000 | |
| Eczema | 33,334 (5 studies) | Very lowᵇ | 0.93 (0.82–1.07) | 131 per 1,000 | |
| Randomized controlled trials | Atopic diseases in offspring | 4,271 (8 RCTs) | Lowᵃ,ᵇ | 0.87 (0.72–1.06) | 241 per 1,000 |
| Asthma | 2,241 (7 RCTs) | High | 0.80 (0.61–1.05) | 170 per 1,000 | |
| Eczema | 1,929 (7 RCTs) | Lowᶜ,ᵈ | 1.21 (0.87–1.67) | 197 per 1,000 | |
| Food allergy | 444 (4 RCTs) | Moderateᵉ | 0.75 (0.40–1.41) | 142 per 1,000 | |
| Rhinoconjunctivitis | 1,863 (4 RCTs) | High | 0.83 (0.66–1.03) | 147 per 1,000 |
CI, confidence interval; GRADE, Grading of Recommendations Assessment, Development and Evaluation; LC-PUFAs, long-chain polyunsaturated fatty acids; RR, risk ratio; RCT, randomized controlled trial. The absolute risk difference was calculated from the assumed risk in the placebo/reference group and the relative effect estimate. Negative values indicate fewer events with n–3 LC-PUFAs, whereas positive values indicate more events. ᵃStudies with a relatively large sample size were considered to have a higher risk of bias. ᵇThe 95% CI crosses the null value (RR = 1), and the confidence interval was considered insufficiently precise for clinical interpretation. ᶜModerate heterogeneity was observed (I2 = 53%, p = 0.04), with widely varying point estimates (RR 0.39–3.77) and some non-overlapping confidence intervals. ᵈThe confidence interval crosses the null value and is wide (0.87–1.67). ᵉThe funnel plot was clearly asymmetric. GRADE certainty was assessed separately for prospective cohort studies and RCTs because of differences in study design and risk of bias.