Table 2. Incidences of cervical cancer by metformin exposure and hazard ratios comparing exposed to unexposed in the original cohort and the matched cohort, respectively.

Metformin use	Case number	Incident cervical cancer	%	Person-years	Incidence rate (per 100,000 person-years)	Hazard ratio (95% confidence interval)	P
I. Original cohort
Never users	6940	38	0.55	31307.79	121.38	1.000
Ever users	132971	438	0.33	641413.41	68.29	0.558 (0.401-0.778)	0.0006
Tertiles of cumulative duration of metformin therapy (months)
Never users	6940	38	0.55	31307.79	121.38	1.000
<23.0	43778	254	0.58	161462.49	157.31	1.272 (0.904-1.790)	0.1679
23.0-47.9	44026	146	0.33	221949.93	65.78	0.523 (0.366-0.747)	0.0004
>47.9	45167	38	0.08	258001.00	14.73	0.109 (0.070-0.172)	<0.0001
II. Matched cohort
Never users	6940	38	0.55	31303.68	121.39	1.000
Ever users	6940	21	0.30	32891.65	63.85	0.522 (0.306-0.889)	0.0168
Tertiles of cumulative duration of metformin therapy (months)
Never users	6940	38	0.55	31303.68	121.39	1.000
<25.1	2287	12	0.52	8017.01	149.68	1.227 (0.639-2.355)	0.5383
25.1-50.4	2294	8	0.35	11443.50	69.91	0.562 (0.262-1.205)	0.1388
>50.4	2358	1	0.04	13431.15	7.45	0.061 (0.008-0.447)	0.0059

Cox regression models were created by incorporation with the inverse probability of treatment weighting using propensity score created from variables in Table 1 plus the entry date of the patients.