Klip 2004.

Methods	Retrospective cohort.
Participants	The study cohort consisted of 26,428 women diagnosed with subfertility problems between 1980 and 1995 in all 12 IVF hospitals in the Netherlands, with known date of birth, and > 18 years at the time of first visit to the fertility clinic. Women that emigrated, were lost to follow‐up, refused to participate, were diagnosed with cancer before entering the cohort, or had unknown date of first IVF treatment or privacy issues were excluded, leaving 24,692 in the analytic cohort, 18,310 and 6382 in the exposed and unexposed group, respectively. Attempt was made to frequency match the control group according to the distribution of the subfertility diagnoses in the IVF group. All cohort members received a risk factor questionnaire, and subfertility data were collected from their medical records. Median follow‐up, 5.2 years in exposed; 8.0 years in unexposed group.
Interventions	At least one IVF treatment cycle with ovarian stimulation. Information on date, type of IVF treatment, dosages and type of fertility drugs used in each phase of the menstrual cycle (human menopausal gonadotropin, follicle‐stimulating hormone, clomiphene citrate, human chorionic gonadotropin, progesterone and luteal phase support) was available.
Outcomes	Endometrial cancer identified through record linkage to the population‐based Netherlands Cancer Registry for the period 1989 to 1997. Incident cases in total cohort, 12; in exposed cohort, 6.
Notes	Subfertile women were unable to achieve conception after 1 or more years of frequent unprotected intercourse. The cause of subfertility was medically verified and classified as tubal factor, male factor, ovarian disorder (including ovulation disorder, polycystic ovary syndrome, premature menopause), cervical factor, uterine abnormality, endometriosis, or unexplained.
Risk of bias
Bias	Authors' judgement	Support for judgement
Selection bias (comparability)	High risk	Quote: "The following women were excluded from all analyses: 80 women diagnosed with cancer before entering the cohort (including two women with breast cancer and six women with ovarian cancer)". Comment: Outcome not likely to be present at start. "The unexposed group consisted of 6,588 women whose subfertility was diagnosed in the four participating clinics that had a computerised registry of all subfertile women evaluated after 1980. We extensively checked whether women in the unexposed group received IVF in other hospitals". Comment: Non‐exposed drawn from the same population as the exposed cohort. Comment: No comparability on cause of subfertility, diabetes mellitus, polycystic ovary syndrome (PCOS), and obesity was ensured.
Selection bias (confounding)	High risk	"Standardised Incidence Ratio (SIR), defined as the ratio of the observed (O) and expected (E) number of cancers in the study population. Expected numbers were calculated by applying the person‐year distribution in the cohort to sex‐, age‐ and calendar period‐specific reference data from the NCR, Hazard Ratios adjusted for age at the end of follow‐up". Comment: Analyses adjusted only for age.
Performance bias	High risk	Quote: "Research assistants specifically trained for data collection in this study abstracted detailed information from the medical records of 13,216 women in the cohort". Comment: Exposure to ovary‐stimulation drugs was ascertained by a secure source, namely, medical records. Blindness regarding the allocated interventions not guaranteed.
Detection bias	Low risk	Quote: "The study outcome, cancer incidence, was assessed through record linkage with the NCR for the period 1989‐1997 and through the health questionnaire survey (including medical verification of self‐reported tumours) for the periods before 1989 and after 1997". Comment: Ascertainment of outcome by record linkage.
Attrition bias	Low risk	Quote from Table 1: "Loss to follow‐up: 1.9%" Comment: Follow‐up was expected to be rather complete.
Selective reporting (reporting bias)	Low risk	Comment: All of the study’s prespecified (primary and secondary) outcomes that were of interest in the review have been reported in the prespecified way.
Other bias	High risk	Quote from Table 2: "a median follow‐up duration of 16.9 years". Comment: Inadequate length of follow‐up (< 10 years). Comment: Non‐RCT study