Table 2.

Patient selection from source population to matched cohorts in each dataset

Cohort	Dataset
	Finland	Sweden	Netherlands hospital	Netherlands GP	UK GP	UK GP-hospital	Total
Diabetic patients who initiated or switched treatment
Source population	438 229	295 001	208 962	63 284	317 632	306 301	1 629 409
Eligible patients after applying all exclusion criteria
Exposed to pioglitazone	18 794	3712	9081	2829	12 662	14 504	61 587
Not exposed to pioglitazone	332 255	224 742	131 028	42 174	70 471	78 037	878 707
Cohort sizes after matching*
Exposed to pioglitazone	18 794	3 712	7491 (8965)	2823	11 408	12 109	56 337
Not exposed to pioglitazone
Nearest matched cohort	18 794	3 712	7491 (8965)	2823	11 408	12 109	56 337
Multiple matched cohort	157 263	31 040	42 048 (50 493)	15 775	35 799	35 184	317 109

*Numbers in parentheses are the cohort size before overlap between Netherland cohorts was removed.