{"paper_id":"c820c278-4c76-4568-a672-4b950d699dfd","body_text":"Menopause is defined as the permanent cessation of menstruation, marking the loss of ovarian follicular activity\n \n 1 ,\n 2 \n . It occurs with the final menstrual period (FMP) and is typically diagnosed retrospectively after 12 months of amenorrhoea\n \n 1 ,\n 3 \n .\nAge at menopause is a marker of aging and overall health. A later age at menopause has been associated with increased life expectancy\n \n 4 \n  and reduced all-cause mortality\n \n 5 \n , as well as a lower risk of cardiovascular disease\n \n 4 ,\n 6 –\n 12 \n  and osteoporosis\n \n 13 ,\n 14 \n . It is also linked to a higher risk of breast\n \n 15 ,\n 16 \n , endometrial, and ovarian cancers\n \n 4 ,\n 17 –\n 20 \n .\nPrevious studies have mostly relied on retrospective self-reports of age at final menstrual period, which are susceptible to recall error, particularly when many years have passed since menopause\n \n 21 ,\n 22 \n . While prospective data on menstrual cycles can help capture the variability in menstrual patterns and reduce recall error, accurately identifying the age at menopause using repeated, prospective data can pose challenges.\nIn this data note, we describe our approach to estimating the age at natural menopause using repeated menstrual cycle data from participants in a UK birth cohort.\n\nThe Avon Longitudinal Study of Parents and Children (ALSPAC) is a longitudinal birth cohort that recruited pregnant women with an expected date of delivery between April 1991 and December 1992 residing in Avon, UK. The initial ALSPAC sample consisted of 14,541 pregnancies, from 14,203 unique mothers (338 mothers having two enrolled pregnancies), which resulted in 14,062 live births. As a result of the additional phases of recruitment, a further 630 mothers who did not enrol originally have provided data since their child was 7 years of age. This provides a total of 14,833 unique mothers (known as the Generation 0 or G0) enrolled in ALSPAC.\nSince recruitment, mothers, their children and partners have been followed-up through questionnaires, research clinic assessments and data linkage. Since 2014, study data have been collected and managed using REDCap (Research Electronic Data Capture), a secure, web-based platform hosted at the University of Bristol\n \n 23 \n . REDCap is specifically designed to support data capture and management in research studies.\nDetails on the representativeness, cohort profiles and recruitment have been extensively described in previous publications\n \n 24 –\n 27 \n . The\n ALSPAC study website  contains details of all data available through a fully searchable data dictionary and variable search tool (\n http://www.bristol.ac.uk/alspac/researchers/our-data ). In this study prospective data from ALSPAC G0 mothers was used.\nEthical approval for the ALSPAC study was obtained from the ALSPAC Law and Ethics Committee and local research ethics committees. These approvals cover all core study data collection, including the questionnaires and clinic data that are used in the present analysis. All questionnaire content was reviewed and approved by the ALSPAC Ethics and Law Committee.\nInitial approvals for the establishment of the cohort were granted by: Bristol and Weston Health Authority (E1808,\n Children of the Nineties: Avon Longitudinal Study of Pregnancy and Childhood (ALSPAC)) , approved 28th November 1989; Southmead Health Authority (49/89,\n Children of the Nineties – \"ALSPAC\" ), approved 5th April 1990; and Frenchay Health Authority (90/8,\n Children of the Nineties ), approved 28th June 1990. Approval details for all subsequent clinics (committee, approval number, dates) are available here:\n https://www.bristol.ac.uk/media-library/sites/alspac/documents/governance/Research_Ethics_Committee_approval_references.pdf\nInformed consent for the use of all data was obtained from participants in accordance with the recommendations of the Ethics and Law Committee at the time. The completion of a questionnaire, either on paper or online, was considered to be written consent from participants to use their data for research purposes. Participants can contact the study team at any time to retrospectively withdraw consent for their data to be used. Study participation is voluntary and during all data collection sweeps, information was provided on the intended use of data. Full details of the ALSPAC consent procedure are available on the\n study website .\nIn addition to these study-level approvals, researchers are required to submit individual level project proposals for consideration by the ALSPAC Executive Committee. The present project received such approval before data access was granted.\n\nWomen were asked a set of detailed questions about their menstrual cycles including when they last had a menstrual period (LMP), reasons for period cessation if relevant, and their use of contraception and hormone replacement therapy (HRT), at eight timepoints, via postal questionnaires or in-person clinics.\n Table 1  summarises the data collection timepoints, the number of questionnaire responses or clinic attendees, the year of completion, and the mean age at each timepoint. The average age at completion ranged from 47.4 (standard deviation [SD] 4.5) to 57.7 (SD 4.5). A complete list of the questions and response options used to determine women’s date of LMP at each timepoint i.e. their most recent menstrual period are presented in\n Table 2 . Despite minor differences in wording and response options, the questions were comparable across timepoints.\n Table 3  provides the corresponding ALSPAC variable names and source file locations. All questionnaires and clinic assessments used in ALSPAC are publicly available to view on the\n study website . The specific questionnaires used in this analysis have also been uploaded to our project’s GitHub repository [\n https://github.com/RochelleKnight/Estimating-age-of-menopause-in-mothers-in-the-ALSPAC-Study ].\n* Questionnaire MB followed a different structure for the reproductive health questions than the other questionnaires/clinics. It was therefore not used to assign dates of LMP as described under the section “\n Assigning date of LMP when self-reported date of LMP has been reported (completely or incompletely)”  however is included in the above table for completeness. See Extended Material for details on how data from Questionnaire MB were incorporated.\n†  If a participant answered ‘No’ to ‘In the last 12 months have you had a period or menstrual bleeding?’, they were then asked the question ‘Were your periods stopped by: ‘\n*More than one response was allowed to be marked\n** Questioned asked as ‘Currently using contraceptive injection or implant’\n*** Split into three questions each asking about tablets, patches or creams\nIn brief, information was obtained on the day, month and year of reported date of LMP, whether a period occurred in the last 12 months (yes/no), whether a period occurred in the last 3 months (yes/no), what caused menstruation to stop if relevant, occurrence of any surgical operations relating to reproductive organs, and if the women were taking hormonal contraceptives or HRT.\n\nWe estimated the date of LMP for each woman at each of the eight timepoints. For each woman, we ordered response timepoints by the time of questionnaire completion/clinic attendance. ‘\n Year of attendance ’ refers to the year they responded to the questionnaire/year of clinic attendance, and ‘\n date of attendance ’ refers to the date they responded to the questionnaire/date of clinic attendance. To ensure participant confidentiality, only the month and year of date of attendance are available to researchers. Therefore, we set the day to the midpoint of the month (the 15\n th ). We prioritised the clinic data as the primary source of data for women who completed a questionnaire and attended a clinic assessment in the same month. To estimate the date of women’s LMP, we created an algorithm comprising three key variables which were asked as three separate questions at all timepoints:\n1.   \n Self-reported date of LMP : date of LMP (reported as dd/mm/yyyy). Some women reported the complete date, some reported month and year, and some reported only year.\n2.   \n Whether they had a period or menstrual bleeding in the last 3 months : yes/no response indicating the presence versus absence of a period or menstrual bleeding in the last 3 months.\n3.   \n Whether they had a period or menstrual bleeding in the last 12 months : yes/no response indicating the presence versus absence of a period or menstrual bleeding in the last 12 months.\nBefore applying the LMP estimation algorithm, we sense-checked the 3- and 12-month bleeding history questions to ensure internal consistency (see Extended Material Table 1 for details on recoding rules and inconsistencies resolved).\nA flow chart of the algorithm used to estimate date of LMP at each timepoint is presented in\n Figure 1 , and the algorithm, including illustrative case examples, is detailed below. In brief, LMP dates were estimated based on the available self-reported date of LMP and/or bleeding history (period in the last 12 and/or 3 months), with consistency checks applied to ensure plausible alignment across variables.\n*subject to no self-reported LMP date at a given timepoint, the woman reported no menstrual periods in the last 12 months, and the previous timepoint occurred within the last 2 years. Else no date of LMP estimated for timepoint. Period history estimate derived from questions “In the last 12 months have you had a period or menstrual bleeding?” and “In the last 3 months have you had a period or menstrual bleeding?”.\nWhen asked to report their date of LMP, some women reported a complete date (dd/mm/yyyy), whilst others reported month and year, or only the year of their LMP.\n1. Complete date of LMP has been reported\nWhen a complete date of LMP was reported, we assigned this date as the date of LMP.\n2. Both month and year of LMP were reported, but day is missing\nWhen both month and year of LMP were reported, but the day was missing, we set the day to the midpoint of the month (the 15\n th ) and used this to assign the date of LMP.\n3. Only year of LMP was reported, and month and day were missing\nWhen only the year of LMP was reported, we combined information regarding reported period or menstrual bleeding in the last 3 and 12 months to assign the date of LMP. We did that as follows:\n3.1 Report of no period or menstrual bleeding in the last 3 and 12 months\nWhen the reported year of LMP matched the year of attendance, we did not assign a date of LMP, as this was inconsistent with no period or menstrual bleeding in the last 12 months.\n For example, if the date of attendance was 15\n th  September 2012, and the reported year of LMP was 2012, this would be inconsistent with no period or menstrual bleeding in the last 12 months.\nWhen the reported year of LMP was the year before the year of attendance, we assigned the date of LMP as the midpoint between the 1\n st  January of the reported LMP year and the date one year prior to date of attendance (\n Figure 2a ).\n For example, if the date of attendance was 15\n th  September 2012, and the reported year of LMP was 2011, the date of LMP was assigned as the midpoint between 1\n st  January 2011 and 15\n th  September 2011.\nWhen the reported year of LMP was more than one year before year of attendance, we assigned the date of LMP as 15\n th  June of the reported LMP year (\n Figure 2b ).\n For example, if the date of attendance was 15\n th  September 2012, and the reported year of LMP was 2010, the date of LMP was assigned as the 15\n th  June 2010.\n3.2. Report of period or menstrual bleeding in the last 12 months but not in the last 3 months\nWhen the reported year of LMP matched the year of attendance, and the 3-month window prior to attendance fell within the same year, we assigned the date of LMP as the midpoint between 1\n st  January of the reported LMP year and the date three months before the date of attendance (\n Figure 3a ).\n For example, if the date of attendance was 15\n th  September 2012, and the reported year of LMP was 2012, the date of LMP was assigned as the midpoint between 1\n st  January 2012 and 15\n th  June 2012.\nWhen the reported year of LMP was the year before the year of attendance and the 3-month window prior to the date of attendance did not extend into the reported LMP year, we assigned the LMP date as the midpoint between one year before the date of attendance and 31\n st  December of the reported LMP year (\n Figure 3b ).\n For example, if the date of attendance was 15\n th  September 2012, and the reported year of LMP was 2011, the date of LMP was assigned as the midpoint between 15\n th  September 2011 and 31\n st  December 2011.\nWhen the reported year of LMP was the year before the year of attendance, and the 3-month window prior to date of attendance extended into the reported LMP year, we assigned the LMP date as the midpoint between one year before date of attendance and three months before date of attendance (\n Figure 3c ).\n For example, if the date of attendance was 15\n th  February 2012, and the reported year of LMP was 2011, the date of LMP was assigned as the midpoint between 14\n th  February 2011 and 15\n th  November 2011.\nWe did not assign a date of LMP when the reported year of LMP matched the year of attendance, but the 3-month window prior to date of attendance extended into the previous year. This is because it would then be inconsistent with no menstrual bleeding in the last 3 months.\n For example, if the date of attendance was 15\n th  February 2012, and the reported year of LMP was 2012, a period lasting longer than 3 months but less than 12 months would fall in 2011, not 2012.\nWe did not assign a date of LMP when the self-reported year of LMP was more than one year before the year of attendance, as this would be inconsistent with having a period or menstrual bleeding in the last 12 months.\n For example, if the date of attendance was 15\n th  February 2012 and the reported year of LMP was 2010 or earlier, any date within that year would fall outside the 12 month window.\n3.3 Report of period or menstrual bleeding in the last 3 months\nWhen the reported year of LMP matched the year of attendance and the 3-month window prior to the date of attendance did not extend into the previous year, we assigned the LMP date as the midpoint between the date of attendance and the date three months earlier (\n Figure 4a ).\n For example, if the date of attendance was 15\n th  September 2012, and the reported year of LMP was 2012, the date of LMP was assigned as the midpoint between 15\n th  June 2012 and 15\n th  September 2012.\nWhen the reported year of LMP matched the year of attendance and the 3-month window prior to the date of attendance extended into the previous year, we assigned the LMP date as the midpoint between 1\n st  January of the reported LMP year and the date of attendance (\n Figure 4b ).\n For example, if the date of attendance was 15\n th  February 2012, and the reported year of LMP was 2012, the date of LMP was assigned as the midpoint between 1\n st  January 2012 and 15\n th  February 2012.\nWhen the reported year of LMP was the year before the year of attendance and the 3-month window prior to the date of attendance extended into the reported LMP year, we assigned the date of LMP as the midpoint between three months before the date of attendance and 31\n st  December of the reported LMP year (\n Figure 4c ).\n For example, if the date of attendance was 15\n th  February 2012, and the reported year of LMP was 2011, the date of LMP was assigned as the midpoint between 15\n th  November 2011 and 31\n st  December 2011.\nFor all other scenarios, we did not assign a date of LMP.\na. Assigned date of LMP is consistent with reported menstruation in the previous 3 and 12 months\nFor women with an assigned LMP date (as described in sections “1. Complete date of LMP has been reported” or “2. Both month and year of LMP were reported, but day is missing”), we checked consistency with responses to whether they had a period or menstrual bleeding in the last 3 and 12 months. To minimise data loss while maintaining plausibility, we applied flexible thresholds based on these responses.\nReported they had menstruated in the last 3 months:  We assumed LMP should fall within this period but allowed a ±3 month buffer to account for reporting error. Therefore, assigned dates were accepted if they fell from 3 months before the start of that period to 3 months after the date of attendance (\n Figure 5a ).\nReported they had menstruated in the last 12 months but not in the last 3 months:  A wider ±6 month buffer was applied around the expected LMP window (\n Figure 5b ).\nReported they had not menstruated in the last 12 months:  We assumed the LMP occurred at least 12 months prior to the date of attendance but allowed a 6-month buffer. Assigned LMP dates therefore had to be at least 6 months before the date of attendance (\n Figure 5c ).\nAssigned LMP dates falling outside these allowable windows were set to missing. Assigned dates of LMP that occurred after the date of attendance (i.e., future dates) were subject to additional checks (see below).\nb. Assigned date of LMP after the date of attendance (i.e., date in the future)\nSince the exact date of clinic attendance or questionnaire completion was not available to researchers, all attendance dates were standardised to the 15\n th  of the month. This standardisation could result in instances where the assigned LMP date appeared to fall after the date of attendance. For example, if the actual attendance occurred on 31\n st  May and the true LMP was also on 31\n st  May, but the attendance date was standardised to 15\n th  May, a discrepancy of 16 days would arise. To address this and minimize data loss, we allowed the assigned LMP date to occur up to 31 days after the standardised attendance date, accounting for potential misreporting of the month. If the assigned date of LMP exceeded 31 days beyond the attendance date, it was set to missing.\nc. Assigned date of LMP is prior to date of attendance at the previous timepoint\nIf the assigned date of LMP for a given timepoint occurred before the date of attendance at the previous timepoint, it was set to missing, unless no LMP date had been assigned for the previous timepoint. This approach assumes that if an LMP date falls before the date of attendance of an earlier timepoint, it is more likely to be influenced by recall error than the estimate provided at the previous timepoint. Therefore, we retained the LMP date that was most proximal to the reporting period.\nFor example, consider a woman who attended her first timepoint on 15\n th  May 2013 and reported her LMP as 17\n th  June 2008. At her second timepoint on 15\n th  October 2015, she reported her LMP as 31\n st  February 2009. Although the second timepoint’s LMP date is more recent, it predates the attendance date of the first timepoint and therefore the LMP date reported in the second timepoint was set to missing. However, if no LMP date had been assigned at the first timepoint, the LMP date from the second timepoint would be retained.\nIn cases where no self-reported date of LMP was available, we relied on information about reported menstrual bleeding in the last 3 and 12 months, as outlined below.\n4. Assigning date of LMP using responses relating to menstruation in the last 3 and 12 months\nWe assigned the date of LMP using reported period or menstrual bleeding in the last 3 and 12 months in the following two scenarios:\nNo information on self-reported date of LMP:  The woman did not report any part of the LMP date (day, month or year).\nInvalid self-reported date of LMP:  The self-reported LMP date was more than 31 days after the date of attendance.\nAs previously explained, any LMP date reported as more than 31 days after the date of attendance was set to missing. However, to minimise the number of timepoints with missing LMP dates, we assumed that a future-dated LMP, although misreported, still likely indicated that the woman was menstruating. Therefore, in both scenarios, we assigned the LMP date using responses relating to menstruation in the last 3 and 12 months, as follows:\na) Reported having a menstrual period in the last 3 months\nIf the woman reported menstruation in the last 3 months, the LMP date was assigned as the midpoint between the date of attendance and 3 months prior (calculated as 45 days prior to the date of attendance).\nb) Reported having a menstrual period in the last 12 months but not in the last 3 months\nIf the woman reported menstruation within the last 12 months but not in the last 3 months, the LMP date was assigned as the midpoint between 12 months before and 3 months before the date of attendance (calculated as 228 days prior to date of attendance).\nc) Reported having no menstrual period in the last 12 months\nIf the woman reported no menstruation in the last 12 months, an LMP date was unable to be assigned and was set as missing.\na. Assigned date of LMP is prior to date of attendance at the previous timepoint\nThe same consistency check as described above under the section ‘\n Estimating date of LMP when self-reported date of LMP has been reported (completely or incompletely)’  was also performed.\nIf an LMP date could not be assigned using self-reported LMP date or menstrual bleeding history (i.e., reports of menstrual bleeding in the last 3 or 12 months combined with the date of attendance), we used the most recent LMP from a previous timepoint to minimise missing data.\nSpecifically, we carried forward the most recent LMP from an earlier timepoint in the following scenarios:\nThe assigned LMP date at a given timepoint fell before the date of attendance of the previous timepoint.\nThere was no self-reported LMP date at a given timepoint, the woman reported no menstrual periods in the last 12 months, and the previous timepoint occurred within the last 3 years.\nIn all other cases, the LMP date remained as missing.\nThe rationale for the second scenario was to avoid misclassifying menopausal status by inferring menopause too early. Carrying forward an LMP from an earlier timepoint could misleadingly suggest a woman had reached menopause because more than 12 months had passed since her previous timepoint (and hence more than 12 months since her last recorded period). Given the variability in time gaps between timepoints—sometimes up to 10 years—such an approach could result in significant errors in estimating age at menopause.\nFor example, consider a woman who attended a timepoint on 15\n th  October 2010 and reported an LMP of 17\n th  September 2010. She next attended on 15\n th  February 2015, did not report an LMP date, but stated that she had not had a period in the last 12 months. If we assigned her 2010 LMP to this later timepoint it would imply she was postmenopausal, even though she may have continued menstruating after 2010 – information not captured due to the long gap between assessments. To address this, we only carried forward LMP dates if the previous timepoint occurred within three years of the current one, limiting the potential error. For women who reported no periods in the last 12 months and had an LMP up to three years prior, this implies that menopause likely occurred within the preceding two years — such as between ages 51 and 53 for a woman attending at age 54. We considered this level of uncertainty acceptable for classifying menopausal status, given the limitations of self-reported data. See Extended Material for details on how data from Questionnaire MB were incorporated.\n\nAge at natural menopause was determined using both our algorithm-based approach and self-reported age at menopause, whenever available.\nUsing our algorithm, we defined the final menstrual period (menopause) as occurring if a woman’s most recent LMP was more than 365 days before the date of attendance at her last assessment. Age at menopause was then defined as age at FMP. Self-reported age at menopause was derived from multiple questionnaires. Questionnaires T, V, and Y asked,\n “What was your age at your last menstrual period?” . If the reported age was at least one year prior to the respondent’s current age, it was assigned as self-reported age at menopause. Questionnaire U (completed in 2011–2012 by 4,423 women) directly asked,\n “What was your age at menopause?” . The maximum self-reported age at menopause across Questionnaires U, T, V, and Y was assigned as each woman’s self-reported age at menopause.\nTo ensure consistency between sources, we applied a validity check: self-reported age at menopause had to be no more than two years earlier than the age at the most recent LMP, otherwise it was excluded. This helped reduce instances where self-report and LMP data conflicted in implausible ways. For example, consider a women who reported a LMP within the last 12 months at age 52, meaning she was not yet classified as postmenopausal. At a later timepoint, aged 54, she reported her last menstrual period occurred at age 50 via the question\n “What was your age at your last menstrual period?” . If taken at face value, this would suggest she was already postmenopausal by age 52, contradicting her earlier LMP report. To allow for some inconsistency in recall or reporting, we accepted self-reported ages at menopause only if they were after the most recent LMP, or within two years prior to it. This threshold provided a balance between data inclusion and plausibility.\nIf a woman only had self-reported age at menopause, this value was used. Otherwise, we prioritised the assigned age at menopause from our algorithm, as it is less likely to be affected by recall bias.\n\nAs age of menopause is estimated based on menstrual bleeding patterns, we also considered factors that affect menstrual bleeding and therefore could result in an inaccurate assessment of menopausal status. These include surgery of reproductive organs, use of contraceptives and HRT, and other reasons such as chemotherapy or radiation therapy, ablation/resection, pregnancy, or breastfeeding.\nAll women were asked in questionnaires and clinics whether they had undergone any surgeries to remove their reproductive organs (see\n Table 2  for full list of options). Response options that would result in the cessation of menstruation were as follows:\nHysterectomy with bilateral oophorectomy (removal of uterus and both ovaries)\nBilateral oophorectomy (removal of both ovaries)\nHysterectomy (removal of uterus)\nHysterectomy with unilateral oophorectomy (removal of uterus and one ovary)\nWomen who reported any of these surgeries were asked to provide the date and age at which the procedure was performed. If a woman did not explicitly report undergoing surgery but provided either an age or a date of surgery, it was assumed that she had the procedure. Conversely, women who reported surgery but did not provide a date or age were excluded.\nAs a separate question, women who reported no menstrual bleeding in the last 12 months were asked to indicate the reason for cessation (see\n Table 2  for full list of options). Surgical reasons included: Surgery, Hysterectomy and Oophorectomy, without further detail. We assumed that reports of ‘Surgery’ and ‘Oophorectomy’ would result in the cessation of all subsequent menstruation. The women were not asked to report the date of surgery, hence the date and age at questionnaire completion or clinic attendance were used as proxies.\nThe overall date and age at surgery were taken as the earliest reported values. Any timepoints following surgery were censored. Estimations of date of LMP were made for timepoints preceding surgery. If self-reported age at menopause age derived from Questionnaires T, U, V and Y was greater than or equal to the surgery age, it was set to missing.\nAt all eight timepoints, women were asked whether they were currently using hormonal contraceptives or HRT.\n Table 2  summarises the specific questions and response options related to contraceptive and HRT use for all for all questionnaires and clinic assessments. Menopausal status cannot be assessed when hormonal medications are being used due to their effect on the menstrual cycle. Therefore, timepoints were censored if current use of hormonal contraceptives or HRT were reported.\nWomen who reported no menstrual bleeding in the last 12 months were asked to indicate the reason for cessation (see\n Table 2  for full list of options). Timepoints were censored if the reported cause was chemotherapy or radiation therapy; ablation/resection; pregnancy, or breastfeeding; or other reason/other medical reason.\n\nFigure 6  shows participant flow from recruitment through to analysis. After accounting for withdrawal of consent, 14,833 unique G0 women were enrolled in the ALSPAC study, of whom 7,197 responded to at least one relevant timepoint. Thirty-five women were excluded due to reporting a surgical procedure that would have ceased menstruation without providing the date or age at surgery. After applying censoring criteria, 5,949 women remained, contributing an average of 3.6 responses each. We were able to assign at least one LMP date for 5,339 of these women.\nWe derived an age at natural menopause - using both self-reported and algorithm-based methods - for 2,422 women. The overall average age at menopause was 49.4 years (SD = 4.1, range: 30–63 years, median = 50 years). Of these, 2,111 women had an algorithm-assigned age at menopause (mean = 49.6 years, SD = 4.1, range: 30–63), and 311 based on self-reported age at menopause (mean = 48.2 years, SD = 4.0, range: 34–61). A total of 1,322 women had both a self-reported and algorithm assigned age at menopause, with a correlation 0.75 between the two measures.\nWe were unable to derive an age at natural menopause for 3,588 women. At their final attended timepoint, these women had a mean age of 51.5 years (SD: 5.9, range: 34–68). Of them, 1,982 had experienced an LMP within the last 12 months. An additional 1,483 women (mean age = 56.9 years, SD = 3.1) reported that they had not had a period in the last 12 months, likely indicating they were postmenopause, however lacked sufficient data to estimate a final menstrual period.\nFigure 7  displays the distribution of age at menopause within the sample.\n\nThe UK Biobank asked women to report their age at menopause during the baseline assessment and the mean reported age was 49.8 years (median 50, SD = 5.1). The Study of Women's Health Across the Nation (SWAN Study), a longitudinal study in the US, found a median age at natural menopause of 51.4 years in its cross-sectional screener data (n= 14,620)\n \n 28 \n  and slightly higher at 52.5 years among longitudinal cohort participants (n=1,483)\n \n 29 \n . The higher estimate in the longitudinal cohort likely reflects selection factors: women who had already reached menopause before age 42 were excluded from follow-up, and attrition over time disproportionately affected women with characteristics linked to earlier menopause (e.g., smokers, lower education, poorer health). Other studies report a median self-reported age at menopause among White women from industrialized countries ranging between 50 and 52 years\n \n 30 –\n 34 \n .\nIn our sample, the mean age at menopause was 49.4 years (median 50 years), slightly lower than in other studies but still comparable. This discrepancy may be partly due to the mean age at several timepoints falling below the typical menopausal age range. As a result, women who experienced menopause earlier may have been more likely to be captured, while those who reached menopause at older ages may have been underrepresented. This could have biased the estimated distribution of menopausal age toward younger ages, potentially explaining the lower age at menopause in our sample compared to other studies.\n\nOne of the key strengths of the ALSPAC dataset is the availability of repeated data on menstrual cycles. This longitudinal information provides a detailed view of menstrual changes over time, capturing the often irregular patterns characteristic of the menopause transition. As a result, it allows for more accurate estimation of age at menopause compared to relying solely on retrospective self-reports, which are prone to recall error\n \n 21 ,\n 22 \n . Additionally, the repeated data support future work to classify woman into menopausal stages - premenopause, perimenopause and postmenopause - using the Stages of Reproductive Aging Workshop (STRAW)+10\n \n 35 \n  criteria.\nA broad challenge was the potential for incorrect estimation of age at menopause due to missing, incomplete, or inconsistent reporting of menstrual history. Responses to “\n In the last 12 months have you had a period or menstrual bleeding? ”, “\n In the last 3 months have you had a period or menstrual bleeding? ” and “\n When was your last period? ” did not always align. Although consistency checks where performed to reconcile discrepancies, misreporting of LMP may still have affected the estimation of age at menopause.\nIn some cases, we identified likely misclassification. For example, 80 women reported a period of amenorrhea lasting more than 12 months at one timepoint, suggesting menopause, but subsequently reported menstruating within the last year at a later timepoint. This highlights the limitations of relying on self-reported data to determine menopausal status. Ultimately, we used the most recent reported LMP to define age at menopause. As a result, 37 of these women do not have an estimated age at menopause, despite previously indicating a period of amenorrhea consistent with menopause.\nAnother limitation is the lack of information on when women discontinued hormonal contraceptives or HRT. While current use was recorded, there was no data on the timing of discontinuation. Because hormonal medications can obscure natural bleeding patterns, menopausal status—and particularly the timing of the final menstrual period—cannot be reliably determined while a woman is using them. We therefore excluded timepoints where hormone use was reported. However, if a woman reported hormone use at one timepoint but not at a later timepoint, we lacked information on the time since discontinuation, and therefore could not determine whether sufficient time had passed for natural menstrual cycles to resume. Without this, it is unclear whether a reported LMP might reflect bleeding prior to starting hormones, a withdrawal bleed, or a natural period. Additionally, the absence of menstruation could also reflect recent hormonal discontinuation rather than the menopause itself. Some women may have also reached menopause while still using hormonal medications, meaning the menopausal transition was not captured, and we were unable to estimate their age at menopause. These limitations may have affected the accuracy of LMP reporting and, consequently, the estimation of menopausal age.\nThe often large gaps between attended timepoints also posed challenges. While the average interval between attended timepoints was 2.1 years, some exceeded 10 years. In such cases, when the only available information was a report of no menstruation in the last 12 months, menopause likely occurred sometime during the intervening years. However, without more precise or frequent data, age at menopause could not be reliably assigned, contributing to potential misclassification and underestimation of menopausal age.\nFinally, the structure of Questionnaire Y limited our ability to assign LMP dates at that timepoint. Whereas most questionnaires asked all women about menstruation in the last 12 and 3 months and to report the date of their LMP, Questionnaire Y only asked for an LMP date if the woman reported menstruating in the last 12 months. As a result, we could only assign an LMP date for a subset of women at this timepoint—those still menstruating—limiting the usefulness of these data for estimating age at menopause and adding to the broader challenges described above.","source_license":"CC-BY-4.0","license_restricted":false}