Age at period cessation and trajectories of cardiovascular risk factors across mid and later life

Objective To examine the association between age at period cessation and trajectories of anthropometry, blood pressure, lipids and glycated haemoglobin (HbA1c) from midlife to age 69 years. Methods We used data from the UK Medical Research Council National Survey of Health and Development to examine the association between age at period cessation and trajectories of systolic blood pressure (SBP), diastolic blood pressure (DBP), body mass index (BMI) and waist circumference (WC) from 36 to 69 years and trajectories of triglyceride, low density lipoprotein cholesterol (LDL-C), high density lipoprotein cholesterol (HDL-C) and HbA1c from 53 to 69 years. Results We found no evidence that age at period cessation was associated with trajectories of log triglyceride, LDL-C and HDL-C from 53 to 69 years and trajectories of SBP or DBP from 36 to 69 years, regardless of whether period cessation occurred naturally or due to hysterectomy. While we found some evidence of associations of age at period cessation with log BMI, log WC and log HbA1c, patterns were not consistent and differences were small at age 69 years, with confidence intervals that spanned the null value. Conclusion How and when women experience period cessation is unlikely to adversely affect conventional cardiovascular risk factors across mid and later life. Women and clinicians concerned about the impact of type and timing of period cessation on conventional cardiovascular intermediates from midlife should be reassured that the impact over the long term is small.


eAppendix 1 Details on confounders
We categorised participant's occupation as reported at age 53 into six classes (from professional to unskilled manual) according to the Registrar General social classification. Parity was self-reported by the participants at all data collections across adulthood and included as a continuous covariate. Women reported monthly histories of HRT use throughout follow-up and from this, HRT use (yes/no) as a timevarying covariate was included in our models. Age at menarche in years was obtained at medical examination by school doctors when participants were 14-15 years old, supplemented for those who had not reached menarche by 15 years (n=94) by retrospective reports of age at menarche by postal questionnaires when participants were aged 48 years 1 . Smoking, obtained from self-reported information at 36 years, was classified as former/current/never. Physical activity at 36 years, obtained from self-reports of frequency and duration of participation in leisure time activities was classified as inactive (reported no participation); moderately active (participated in relevant activities one to four times: in the previous month); or most active (participated in relevant activities five or more times: in the previous month 2 .

eAppendix 2 Details of model selection
Models were derived by initially examining observed data for each risk factor and plotting mean values for each risk factor over time to examine the possible shape of the trajectory.
Linear spline multilevel models were used for SBP, DBP, WC and BMI. Based on the observed data for these, we compared observed and predicted measurements for a selection of suitable models for each risk factor. We examined rates of change between time periods in order to examine whether changes between periods were similar or different. In cases where rates of change between spline periods appeared identical, the fit of models with reduced splines was explored. Final models for SBP, DBP, BMI and WC had one knot placed at 53 years resulting in two periods of change; from 36-53 and from 53-69.
This knot was selected based on examination of observed data over time and comparing model fit placed at whole years closest to mean age at clinics due to a greater density of measures). The selection of this knot point at age 53 years, also had the additional advantage of allowing comparability with the linear slopes modelled from 53 to 69 years for the blood-based biomarkers.

SBP and DBP
The models for SBP and DBP took the form of: SBP ij /DBP ij = β 0 + u 0j + (β 1 + u 1j )s ij1 + (β 2 + u 2j )s ij2 + e ij (age ij ) where for person j at measurement occasion i; β 0 represents the fixed effect coefficient for the average intercept, β 1 to β 2 represent fixed effect coefficients for the average linear slopes of each linear spline, u 0j to u 3j indicate person-specific random effects for the intercept and slopes respectively, and e ij represents the occasion-specific residuals or measurement error which was allowed to vary with age.

BMI, WC
BMI and WC were natural log transformed. The models for BMI and WC took the form of: log BMI ij /log WC ij = β 0 + u 0j + (β 1 + u 1j )s ij1 + (β 2 + u 2j )s ij2 + e ij (age ij ) where for person j at measurement occasion i; β 0 represents the fixed effect coefficient for the average intercept, β 1 to β 3 represent fixed effect coefficients for the average linear slopes of each linear spline, u 0j to u 2j indicate person-specific random effects for the intercept and slopes respectively, and e ij represents the occasion-specific residuals or measurement error which was allowed to vary with age.

Lipids and HBA1c
Triglyceride and glycated haemoglobin were natural log transformed. Lipids and HBA1c were modelled using a linear age term.The models for lipids and HBA1C took the form of: log triglycerides ij / LDL-C ij /HDL-C ij / log HBA1c ij = β 0 + u 0j + (β 1 + u 1j ) age ij1 + e ij (age ij ) where for person j at measurement occasion i; β 0 represents the fixed effect coefficient for the average intercept, β 1 represents fixed effect coefficients for the average linear slope, u 0j and u 1j indicate person-specific random effects for the intercept and slopes respectively, and e ij represents the occasion-specific residuals or measurement error which was allowed to vary with age.

eAppendix 3 Additional and sensitivity analyses
We examined the characteristics of participants included in analyses of anthropometry compared with those excluded due to missing exposure, outcome or confounder data or loss-to follow-up to better understand the role of selection bias. We examined whether pharmacologic treatment of blood In order to examine whether findings differed by type of hysterectomy, we tested whether the associations of age at period cessation with cardiovascular risk factors differed between women who had hysterectomy with bilateral oophorectomy compared with hysterectomy with conservation of at least one ovary. Trajectories of risk factors were examined in women who could have been pre-or postmenopausal at age 53. Therefore, we performed a sensitivity analysis excluding women who were still pre-menopausal at 53 years, to understand if our findings differed when the analyses were restricted to post-menopausal measures of cardiovascular risk factors.

eTable 8 Association of age at period cessation (per year increase) with blood markers from 53 to 69 years, adjusted for co-variates
Legend: HBA1c, glycated haemoglobin; HDL-C, high density lipoprotein cholesterol; HY, hysterectomy; LDL-C, low density lipoprotein cholesterol; NM, natural menopause Note that triglyceride and HBA1c are natural log transformed and all values are in log form. Δ = change per year in risk factor. * P value for the interaction of age and type at period cessation with trajectories. Adjusted for socioeconomic position, parity, time-varying hormone replacement therapy use, age at menarche, BMI at age 36, smoking at age 36, physical activity at age 36. Legend: HBA1c, glycated haemoglobin; HDL-C, high density lipoprotein cholesterol; HY, hysterectomy; LDL-C, low density lipoprotein cholesterol; NM, natural menopause Note that triglyceride and HBA1c are natural log transformed and all values are in log form. Δ = change per year in risk factor. *P value for association of type of menopause (HY compared with NM) with trajectory. Adjusted for socioeconomic position, parity, age at period cessation, time-varying hormone replacement therapy use, age at menarche, BMI at age 36, smoking and physical activity at age 36.