TABLE 1.
Comparison of the evidence for disease outcomes examined by 3 types of studies1
| Meta-analysis of observational studies | Meta-analysis of RCTs | MR study | ||||||
|---|---|---|---|---|---|---|---|---|
| Outcomes | Risk estimates (95% CI) | Evidence class2 | Risk of bias level3 | Risk estimates (95% CI) | Evidence | Risk of bias level3 | Risk estimate (95% CI) | Evidence |
| All-cause mortality | 0.61 (0.52, 0.76) | Ⅲ | High | 0.93 (0.88, 0.98) | P < 0.05, 95%PI excluded the null | High | 0.77 (0.62, 0.95) | n = 95,766 |
| Breast cancer | 0.99 (0.98, 1.00) | NS | High | 0.97 (0.95, 1.00) | P > 0.05, 95%PI excluded the null | High | 1.02 (0.97, 1.08) | n = 228,95; P = 0.47; power = 0.13 |
| CVD mortality | 0.88 (0.80, 0.96) | Ⅳ | High | 0.96 (0.93, 1.00) | P > 0.05, 95%PI excluded the null | High | 0.77 (0.55, 1.08) | n = 95,766 |
| Colorectal cancer | 0.87 (0.77, 0.99) | Ⅳ | High | 0.94 (0.91, 0.97) | NA | High | 0.92 (0.76, 1.10) | n = 95,906; P = 0.36; power = 0.29 |
| Asthma | 0.92 (0.77, 1.10) | NS | High | 0.73 (0.58, 0.92) | P < 0.05, 95%PI excluded the null | High | 1.03 (0.90, 1.19) | n = 146,761; P = 0.63; power = 0.25 |
| All-cancer incidence | 0.89 (0.81, 0.97) | Ⅳ | High | 0.98 (0.93, 1.03) | P > 0.05, 95%PI excluded the null | High | 0.97 (0.91, 1.04) | n = 438,870 |
| All-cancer mortality | 0.80 (0.70, 0.91) | Ⅲ | High | 0.87 (0.79, 0.96) | P < 0.05, 95%PI excluded the null | High | 0.97 (0.84, 1.11) | n = 438,870 |
Data are consistent results of meta-analyses of observational studies and MR studies with the results of meta-analyses of RCTs. CVD, cardiovascular disease; MR, Mendelian randomization; NA, not available; NS, not significant (P ≥ 0.05); PI, prediction interval; RCT, randomized controlled trial; ROBIS, Risk of Bias in Systematic Reviews.
Evidence class criteria: class I (convincing), a statistical significance level of P < 10–6, >1000 cases, a significant result at P < 0.05 reported by the study with the largest sample size in the meta-analysis, a 95% PI that excluded the null, a low heterogeneity (I2 < 50%), no evidence of small-study effects (P > 0.10) and of excess significance bias (P > 0.10); class II (highly suggestive), a statistical significance level of P < 10–6, >1000 cases, and a significant result at P < 0.05 reported by the study with the largest sample size in the meta-analysis; class III (suggestive), a statistical significance level of P < 10–3 and >1000 cases; class IV (weak), a statistical significance level of P < 0.05.
The risk of bias was assessed by ROBIS: low risk of bias (Low), the findings of the meta-analysis are likely to be reliable, domain 1–4 (study eligibility criteria, identification and selection of studies, data collection and study appraisal, and synthesis and findings) did not raise any concerns with the review process or concerns were appropriately considered in the review conclusions, the conclusions were supported by the evidence and included consideration of the relevance of included studies; high risk of bias (High), ≥1 of the concerns raised during the domain 1–4 assessment was not addressed in the review conclusions, the review conclusions were not supported by the evidence, or the conclusions did not consider the relevance of the included studies to the review question; unclear risk of bias (Unclear), there is insufficient information reported to make a judgement on risk of bias