Lily Eghdami - UBC Department of Obstetrics & Gynaecology

Fertility and Sterility®International Statistical Classification of Diseases and RelatedHealth Problems (ICD)-9 codes 390–459 (diseases of the circulatorysystem) or 745–747 (anomalies of the cardiovascularsystem), or ICD-10 codes I (diseases of the circulatory system)and Q20–28 (anomalies of the cardiovascular system). Mortalitydue to cancer was based on cause of death attributedto ICD-9 codes 140–208 (neoplasms) and ICD-10 code C (neoplasms).Analyses of cause-specific mortality excluded 419participants with unknown cause of death.Statistical AnalysisBaseline characteristics by type of menopause (surgical due tobilateral oophorectomy vs. natural) were compared using thec 2 test. Multivariable Cox proportional hazards regressionmodels, with age (in days) as the time scale, provided estimatesof the relative risks (RR) and the 95% confidence intervals(CI) for all-cause mortality, cardiovascular mortality, andcancer mortality. Participants were followed from cohort entryuntil the first of the following events: death, move out ofthe United States, or end of follow-up period, which was December31, 2007. The models for all-cause mortality were adjustedfor baseline, self-reported measures of BMI (16–24.9kg/m 2 ,25–29.9 kg/m 2 ,30–54.9 kg/m 2 , or unknown), race(white, African American, or other), smoking history (neversmoker, former smoker, or current smoker), prior history ofdiabetes (yes or no), prior history of high blood pressure(yes or no), prior history of heart attack/myocardial infarction(yes or no), and prior history of at least one of the followingcancers: lung, leukemia, Hodgkin’s lymphoma, other lymphoma,colorectal, thyroid, and melanoma (yes or no). Themodel with all participants was additionally adjusted forage at exposure (natural menopause or bilateral oophorectomy).The models for cardiovascular mortality were configuredin the same manner as described above for all-causemortality, except they were not adjusted for prior history ofheart attack/myocardial infarction. The models for cancermortality were configured in the same manner as describedabove for all-cause mortality, except that they were not adjustedfor prior personal history of at least one of the followingcancers: lung, leukemia, Hodgkin’s lymphoma, otherlymphoma, colorectal, thyroid, and melanoma (yes or no);and they were additionally adjusted for family history of atleast one of the following cancers: breast, endometrium,ovary, cervix, lung, leukemia, Hodgkin’s lymphoma, otherlymphoma, colorectal, prostate, thyroid, or melanoma (yesor no). All Cox models were fit using the PHREG procedurein SAS version 9.2 (SAS Institute). In light of the longfollow-up period under consideration the STRATA¼VARI-ABLE statement in the PHREG procedure was used to adjustfor any cohort effect; the strata variable was set to age (inyears) at cohort entry, a proxy for birth year. The use of ageat cohort entry in years vs. use of a categoric variable forage at cohort entry (i.e., 5-year intervals) in the STRATAstatement ensures that any effect of a single important publicationor event would be adjusted for and is made possible bythe large sample size of the CTS.Because many women who undergo surgical menopausedue to bilateral oophorectomy begin HT use shortly thereafter,we also performed stratified analyses based on HT use (evervs. never) at baseline. Ever-users were further stratified byformulation (ET only vs. estrogen plus progestin therapy[EPT] only) in an analysis that excluded participants whoused more than one formulation. Models were also stratifiedby age at natural/surgical menopause (

116 VOL. 97 NO. 1 / JANUARY 2012TABLE 3Relative risk (95% CI) of all-cause and cause-specific mortality by ever use of menopausal HT in the form of ET or EPT, overall and by age at natural menopause or age at bilateral oophorectomy in42,004 CTS participants, 1995–2007.VariableAllAge at natural menopause or age at bilateral oophorectomy