Table 3.

GRADE assessment per meta-analytical model.

Type of analysis	Number of studies, sample size	Random effect model (crude Peto OR, [95%CI])		Heterogeneity test (I², p for Q statistics) b	GRADE assessment a
					Risk of bias	Inconsistency	Indirectness	Imprecision	Publication bias	Level of Evidence
E2-based COC vs EE-based COC	k=4 n E2-based= 72,210 n oCOC= 487,942	0.67 [0.51;0.87]		I² 0%, p=0.46	Not serious	Not serious	Not serious	Not serious	Not serious	Moderate
E2-based vs oCOC	k=4 n E2-based= 67,008 n oCOC= 97,082	0.60 [0.45;0.81]		I² 0%, p=0.68	Not serious	Not serious	Not serious	Not serious	Not serious	Moderate
E2-based vs EE/LNG	k=4 n E2-based= 72,081 n oCOC= 390,860	0.80 ([0.54;1.17]		I² 6%, p=0.36	Not serious	Not serious	Not serious	Not serious	Not serious	Moderate
EE/LNG vs oCOC	k=3 n EE/LNG= 85,689 n oCOC=96,572	1.32 [0.79;2.21]		I² 76%, p=0.02	Not serious	Serious	Not serious	Not serious	Not serious	Low
Additional analyses in studies providing adjusted HR
Type of analysis	Number of studies, sample size	Random effect model (crude Peto OR, [95%CI])	Random effect model (Adjusted HR, [95%CI]) c	Heterogeneity test (I², p for Q statistics) b	GRADE assessment a
					Risk of bias	Inconsistency	Indirectness	Imprecision	Publication bias	Level of Evidence
E2-based vs EE/LNG in studies with adjustment for confounding factors	k=3 n E2-based= 66,811 (E2/DNG) 66,834 (E2/NOMAC) n EE/LNG= 85,689	E2/DNG 0.64 [0.41;1.00] E2/NOMAC 0.74 [0.48;1.15]	E2/DNG 0.49 [0.29;0.85] E2/NOMAC 0.68 [0.37;1.25]	E2/DNG: I²=0%, p=0.79 E2/NOMAC: I²=37%, p=0.20	Not serious	Not serious	Not serious	Not serious	Not serious	Moderate
Studies from ZEG E2-based vs EE/LNG	k=2 n EE/LNG= 66,778 n oCOC=83,177	0.64 [0.40;1.02]	0.51 [0.29;0.90)	I²=0%, p=0.66	Not serious	Not serious	Not serious	Serious	Not serious	Low

^aGRADE assessment: 1) Study quality was considered as serious if low study quality was reported for more than 75% of the included studies, 2) Inconsistency was considered as serious in case of unexplained substantial significant heterogeneity, 3) Indirectness was considered serious if presence of factors that limit the generalizability of the results, 4) Imprecision was considered serious if k<3 or large 95% Confidence Intervals, 5) Publication bias: due to the low number of included studies (k=5 in the global meta-analytical model), a proper evaluation of publication bias could not be performed. However, because of the quality of the manual search performed for this systematic review and meta-analysis, any missing of evidence was considered as very unlikely and no serious publication bias was considered in the GRADE assessment.

^bI² is reported for adjusted HR when available, otherwise it is reported for crude Peto OR.

^cPotential confounders considered in analyses: Reed et al.: age, body mass index, family history of VTE and current duration of HC use; Schink et al.: age at cohort entry, cardiovascular diseases, coagulation disorders and other blood diseases, diabetes or use of antidiabetics or insulin; migraine with aura, Varicose veins of lower extremitie, obesity, Paresis, hospitalization, surgery, fractures or trauma, Current use of ASA, antiplatelets, antithrombotics or DOACs, Current use of NSAIDs, current use of glucocorticoids or other corticoids, Current use of antidepressants or antipsychotics; Bauerfeind et al.: age, body mass index (BMI), duration of current hormonal contraceptive use and family history of VTE; for ATE, it included age, BMI, smoking, treated hypertension and a family history of fatal ATE.