Table 1. Summary of evidence grading for meta-analyses of randomized controlled trials on menopausal hormone therapy and incidence of diseases and other health outcomes.

Evidence	Criteria	Meta-analyses of randomized controlled trialsa
		Outcomes with beneficial effects	Outcomes with harmful effects
Consistent	95% CI of the mean effect excludes null value with no heterogeneity, or predictive distributionb contains an extreme proportion (>90%) of true effects in the direction of the mean effect	All fracture, vertebral fracture, nonvertebral fracture, colorectal cancer (EPT), cardiovascular mortality (ET)	Cardiovascular disease, cerebrovascular disease, stroke, nonfatal stroke, venous thromboembolism, deep vein thrombosis, gallbladder disease requiring surgery, endometrial hyperplasia (ET), lung cancer mortality (EPT)
Highly suggestive	95% CI of the mean effect excludes null value, with heterogeneity present but predictive distribution not estimablec, or predictive distribution contains a substantial proportion (70%–90%) of true effects in the direction of the mean effect	Vasomotor symptomsd, vasomotor symptom severityd, vaginal atrophy (intravaginal ET)d, sexual function, urinary incontinenced, diabetes mellitus	Cardiovascular disease recurrence
Suggestive	95% CI of the mean effect includes null value, predictive distribution not estimablec, and 95% CI of the most precise studye excludes null value	Breast cancer (ET), endometrial cancer (EPT), hip fracture (EPT), sleep quality (EPT), skeletal muscle strength (EPT), recurrent urinary tract infection (ET)d	Breast cancer (EPT), dementia (EPT), pulmonary embolism (EPT), irregular vaginal bleeding (ET)
Controversial	Predictive distribution contains a non-negligible proportion (>30%) of true effects in both the same and the opposite direction of the mean effect	None	None
Insufficient	Insufficient evidence to draw conclusions	All-cause mortality, all cancer incidence and mortality, lung cancer incidence and mortality (ET), breast cancer recurrence and mortality, ovarian cancer incidence and overall survival, endometrial cancer (ET), colorectal cancer incidence (ET) and mortality, cardiovascular mortality (EPT), cerebrovascular disease recurrence, stroke recurrence, fatal stroke incidence and recurrence, nonfatal stroke recurrence, transient ischemic attack incidence and recurrence, coronary heart disease incidence and recurrence and mortality, myocardial infarction incidence and recurrence, fatal and nonfatal myocardial infarction incidence and recurrence, angina pectoris incidence and recurrence, cardiac death, coronary revascularization, venous thromboembolism recurrence, deep vein thrombosis recurrence, pulmonary embolism incidence (ET) and recurrence, dementia (ET), Alzheimer disease, cognitive function (in healthy women and in women with dementia), hip fracture (ET), endometrial hyperplasia (EPT), irregular vaginal bleeding (EPT), sleep quality (ET), occurrence and recurrence of depressive symptoms

Small-study effects existed for all fracture, sexual function, urinary incontinence, and deep vein thrombosis; the Egger regression test was used to examine whether smaller studies tended to show more pronounced effects than larger studies; it was applied only in meta-analyses of ≥10 studies; more information is available in S3 Text. Meta-analysis results were robust to severe or extreme publication bias for all fracture, vasomotor symptoms, stroke, nonfatal stroke, venous thromboembolism, gallbladder disease requiring surgery, and endometrial hyperplasia (ET); “robust to severe or extreme publication bias” means that the meta-analysis results cannot be explained away by hypothetical publication bias that greatly exceeds empirical estimates of publication bias severity in medicine (i.e., hypothetical publication bias in which statistically significant positive effects are 4-fold more likely to be published and meta-analyzed than nonsignificant or negative effects); more information is available in S3 Text. CI, confidence interval; EPT, estrogen plus progestin therapy; ET, estrogen-alone therapy.

^aThe average effects of any menopausal hormone therapy (ET or EPT) in perimenopausal or postmenopausal women. When subgroup analysis by type of hormone therapy indicated a qualitative difference or statistically significant results were found for only 1 type of hormone therapy, results are presented separately for them. The effects refer to outcome incidence, unless otherwise indicated.

^bThe predictive distribution describes how the true effect sizes across studies are distributed around the summary average effect. Predictive distribution was estimated only in meta-analyses of ≥10 studies. More information is available in S3 Text.

^cDue to a small number of studies (<10) being included in the meta-analysis.

^dIn women who already have the outcome of interest.

^eThe study with the smallest standard error in each meta-analysis.