Within-individual changes in BMI and menstrual irregularity: a cohort study using real-world data.

This cohort study was conducted using data collected from LunaLuna, a widely used smartphone menstrual tracking application in Japan with over 16 million downloads [ 9 ]. In the app, users can log their daily information such as the start/end of the menstrual bleed, daily basal body temperature, menstrual flow, pain, mood, etc [ 10 ]. Recruitment was conducted within the application. During the recruitment period from January 23rd through March 25th, 2020, active users were shown an in-app notification at their first app launch, which briefly described the study objectives, data collection procedures, and participation details. Users who clicked the survey link were directed to a study information page, and only those who provided electronic informed consent were able to access and complete the web-based questionnaires embedded within the app. Participants were asked to complete periodic questionnaires (Supplementary file 1), and responses were linked to their menstrual cycle logs. The first questionnaire (Wave 1) was administered between January 23 and March 25, 2020, followed by the second questionnaire (Wave 2) between May 14 and June 14, 2020. Each questionnaire could only be completed during its respective administration period. Follow-up recruitment was conducted via a single in-app notification shown at the first app launch of each wave to previously consented users. In all waves, questionnaires could be accessed through the app’s “Notifications” section. The questionnaires collected information on lifestyle and health conditions. Menstrual cycle logs were extracted for all available cycles recorded between January 2019 through March 2021 from users who agreed to participate. This study protocol was approved by the Institutional Review Board of the National Center for Child Health and Development (approval date: July 15, 2020, project number: 1985). A flowchart of the inclusion and exclusion of participants is shown in Fig.  1 . Among the study participants, individuals were excluded from the study if they had fewer than 3 logged cycles. To remove implausible values, the mean and standard deviation (SD) of the natural logarithm of all logged cycle lengths were calculated. Cycles outside the range of ± 4 SD were then removed, following a conventional approach used in large cohort studies to exclude extreme outliers likely resulting from erroneous inputs [ 11 , 12 ]. Fig. 1 Inclusion and exclusion criteria. This flowchart illustrates the selection process for study participants. A total of 23,570 individuals initially agreed to participate. Participants were excluded if they had fewer than 3 logged cycles pre-/post-period, and cycle logs were excluded if they deviated by more than ± 4 standard deviations from the average of all logged cycles. Of the remaining participants, 9,545 completed the second questionnaire. Further exclusions were made for individuals outside the 20–45 years age range, those with a BMI outside the 15–35 range, those currently pregnant, or those undergoing hormonal treatment. The final analytical sample included 5,444 participants Inclusion and exclusion criteria. This flowchart illustrates the selection process for study participants. A total of 23,570 individuals initially agreed to participate. Participants were excluded if they had fewer than 3 logged cycles pre-/post-period, and cycle logs were excluded if they deviated by more than ± 4 standard deviations from the average of all logged cycles. Of the remaining participants, 9,545 completed the second questionnaire. Further exclusions were made for individuals outside the 20–45 years age range, those with a BMI outside the 15–35 range, those currently pregnant, or those undergoing hormonal treatment. The final analytical sample included 5,444 participants Prior app-based studies often exclude cycles shorter than 10 days or longer than 90 days or rely on ovulation-confirmation criteria to refine the quality of the cycle logs [ 13 – 16 ]. Because the present study aimed to characterize menstrual irregularity itself rather than restrict to ovulatory or normal cycles, only minimal exclusion criteria were applied to remove clearly erroneous entries while preserving biologically possible variations in cycle length. Of 19,680 users who agreed to participate in Wave 1, 9,325 completed Wave 2. Among these, 7,493 individuals provided BMI information at age 18, Wave 1, and Wave 2. Participants were excluded if they had conditions or exposures known to affect ovulation or menstrual cycles. Specifically, individuals who were currently pregnant, as well as those using hormonal treatments, were excluded. Individuals with BMI  35 at age 18, Wave 1, or Wave 2, were also excluded. Because menstrual cycle and ovulation variability are physiologically greater during adolescence and the menopausal transition, samples were restricted to individuals aged 19 to 45 years. In Japan, the mean age at menopause is approximately 50 years [ 17 ], and most women enter the menopausal transition in their late 40 s. Therefore, the cutoff age of 45 minimizes the inclusion of perimenopausal cycles. After applying all exclusion criteria, the final analytic sample included 5,444 individuals. BMI was calculated from self-reported height and weight, reported in kg/m 2 . BMI information was obtained exclusively through the questionnaires. Participants reported recalled BMI at age 18 in Wave 2, and current height and weight were reported at both Wave 1 and Wave 2. Each participant could therefore contribute up to two contemporaneous BMI measurements plus one recalled measurement. Only participants who provided their full BMI data (age 18, Wave 1, and Wave 2) were included in the analytic sample. Individuals with BMI  35 at any point (age 18, Wave 1, or Wave 2) were excluded. Prior analyses of menstrual tracking app data have shown that, among ovulatory cycles, mean cycle length remains relatively stable within the BMI range of 15–35, but becomes more variable at extreme BMI levels [ 18 ]. Restricting BMI to this range of 15–35, therefore, reduces the likelihood that menstrual irregularity reflects the direct physiological effects of extreme BMI rather than underlying changes in ovulatory function. Because this study examined both long-term and short-term changes in BMI, two exposure variables were constructed. Long-term BMI change was defined as the difference between BMI at age 18 and BMI at Wave 2. Short-term BMI change was defined as the difference between BMI at Wave 1 and BMI at Wave 2. For the main analysis of BMI and menstrual irregularity, BMI at Waves 1 and 2 were categorized into three groups: underweight (BMI 15–18.4), normal (18.5–22.9), and overweight/obese (23–35). The definition of overweight followed the International Obesity Task Force (IOTF) recommendation for the Asian population [ 19 ]. These thresholds were selected because they better reflect metabolic risk profiles in Asian populations. For the supplementary analysis, BMI was further subcategorized following the IOTF Asian definitions: underweight (BMI 15–18.4), normal (18.5–22.9), overweight (23–24.9), and class I obesity (25–29.9) [ 19 ]. Cycle length was calculated from the start date (“menstruation started” as logged in the app) to the day before the next logged start date. Menstrual irregularity was defined as an average cycle length outside 24 to 38 days, following the FIGO definition [ 20 ]. Long-term BMI change reflects the difference in BMI at age 18 and at Wave 2. The outcome for this analysis was assessed over a fixed 12-month window, using menstrual cycle logs from April 2020 to March 2021, corresponding to the period following the distribution of the Wave 2 questionnaire. Menstrual irregularity is therefore measured after the BMI exposure window, even though the BMI change itself may span over several years. Short-term BMI change reflects the differences between BMI at Wave 1 and Wave 2. The outcome for this analysis was new-onset menstrual irregularity. Menstrual cycle logs from January 2019 to March 2020 (15 months) were used as the pre-period, and logs from April 2020 to March 2021 (12 months) as the post-period following the Wave 2 distribution. Average cycle length was calculated separately for the pre- and post-periods, and menstrual irregularity was defined as an average cycle length outside 24 to 38 days in either period. This analysis was restricted to individuals without menstrual irregularity in the pre-period, allowing for the assessment of new-onset menstrual irregularity in the post-period. Participants’ background information was collected from questionnaire responses. High-risk drinking was defined as drinking more than an average of 20 g of ethanol/day [ 21 ]. Past medical history of polycystic ovary syndrome (PCOS) and diabetes mellitus (DM) were used in the analysis since they have been shown to affect the menstrual cycle [ 22 ]. Long-term BMI change was analyzed by categorizing BMI at age 18 and Wave 2 into underweight, normal, and overweight/obese groups. Participants were classified according to whether they remained in the same category or transitioned to a different one. For each subcategory, the proportion of individuals with menstrual irregularity was calculated. Logistic regression models were used to estimate odds ratios relative to those who remained in the same BMI category. Because the time elapsed between age 18 and Wave 2 varied across participants depending on their age at enrollment, all long-term models were adjusted for current age to account for this variability. Short-term BMI change was analyzed similarly, using BMI categories at Wave 1 and Wave 2. Among participants who had regular cycles in the pre-period, new-onset menstrual irregularity in the post-period was modeled. Logistic regression models were used to estimate odds ratios for each BMI-transition category relative to those who remained in the same BMI category. Models were adjusted for age, co-residence with a child, marital status, pregnancy intention, smoking, alcohol intake, exercise habits, employment status, and past medical history, as these variables were considered potential confounders that may influence both BMI and menstrual cycles. Co-residence with a child was used as a surrogate for parity because detailed pregnancy and birth history were not collected in the questionnaire. Lifestyle factors (regular exercise, changes in exercise, and eating habits) were compared across groups using chi-square tests. All tests were two-sided, with p -values < 0.05 defined as statistically significant. Analyses were performed using STATA/SE 18.0.

A total of 126,008 cycles from 5,444 individuals were included in the study (Fig.  1 ). Table  1 presents baseline characteristics according to BMI categories at Wave 2. Compared to individuals with a normal BMI, individuals with underweight were younger, had a lower BMI at age 18, and were the only group to show a decrease in long-term BMI. Individuals with overweight/obesity, compared to those with normal BMI, were older, had a higher prevalence of pre-period menstrual irregularity, and exhibited the greatest increases in BMI in both the short- and long-term. Table 1 Participant characteristics by BMI categories at Wave 2 Characteristics Underweight ( n  = 718) Normal ( n  = 3228) Overweight/Obese ( n  = 1498) Total ( n  = 5444) Age, mean (SD), years 28.3 (6.4) 29.2 (6.5) 31.3 (6.8) 29.7 (6.7) BMI at Wave 1, mean (SD), kg/m 2 17.7 (0.9) 20.5 (1.4) 25.8 (2.8) 21.6 (3.3) BMI at Wave 2, mean (SD), kg/m 2 17.6 (0.7) 20.6 (1.2) 26.1 (2.8) 21.7 (3.4) Short-Term BMI Change, mean (SD) −0.1 (0.7) 0.1 (0.7) 0.3 (1.1) 0.1 (0.9) BMI at age 18, mean (SD), kg/m 2 18.1 (1.4) 20.3 (1.9) 23.3 (3.2) 20.8 (2.8) Long-Term BMI Change mean (SD) −0.5 (1.3) 0.3 (1.7) 2.9 (3.3) 0.9 (2.5) Pre-period Menstrual Irregularity, n(%) 49 (6.8) 224 (6.9) 166 (11.1) 29.7 (6.7) Married, n(%) 245 (34.1%) 1320 (40.9%) 727 (48.5%) 2292 (42.1%) Pregnancy Intention, n(%) 225 (31.3%) 1153 (35.7%) 621 (41.5%) 1999 (36.7%) Living with Child, n(%) 150 (20.9%) 780 (24.2%) 490 (32.7%) 1420 (26.1%) Regular Exercise, n(%) 260 (39.5%) 1407 (48.0%) 626 (45.2%) 2293 (46.1%) Alcohol Intake > 20 g/day, n(%) 111 (27.3%) 584 (29.5%) 297 (34.4%) 992 (30.5%) Current Smoker, n(%) 98 (13.6%) 296 (9.2%) 195 (13.0%) 589 (10.8%) Employment Status, n(%)  Employed without night shifts 435 (60.6%) 2008 (62.2%) 945 (63.1%) 3388 (62.2%)  Enmployed with night shifts 69 (9.6%) 324 (10.0%) 176 (11.7%) 569 (10.5%)  Unemployed/Student or unwilling to answer 214 (29.8%) 896 (27.8%) 377 (25.2%) 1487 (27.3%) Educational Attainment, n(%)  High school degree or higher 698 (97.2%) 3178 (98.5%) 1457 (97.3%) 5333 (98.0%)  Below high school degree 17 (2.4%) 48 (1.5%) 38 (2.5%) 103 (1.9%)  Unwilling to answer 3 (0.4%) 2 (0.1%) 3 (0.2%) 8 (0.1%) Past medical history, n(%)  No complication 547 (76.2) 2571 (79.7) 1128 (75.3) 1128 (22.0)  Uterine myoma 21 (2.9) 99 (3.1) 60 (4.0) 180 (3.3)  Endometriosis 13 (1.8) 58 (1.8) 18 (1.2) 89 (1.6)  Polycystic ovary syndrome 3 (0.4) 24 (0.7) 9 (0.6) 36 (0.7)  Other gynecologic diseases 52 (7.2) 254 (7.9) 128 (8.5) 434 (8.0)  Depression, or other psychiatric diseases 80 (11.1) 276 (8.6) 156 (10.4) 512 (9.4)  Sleep disorders 31 (4.3) 79 (2.5) 52 (3.0) 162 (3.0)  Diabetes Mellitus 2 (0.3) 9 (0.3) 24 (1.6) 35 (0.6)  Unwilling to answer 3 (0.4) 8 (0.3) 3 (0.2) 14 (0.3) Participant characteristics by BMI categories at Wave 2 Figure  2 shows the associations between long-term BMI change and menstrual irregularity, using individuals who remained in the same BMI category from age 18 to Wave 2 as the reference. Individuals with a normal BMI at age 18 who transitioned to overweight/obesity had higher odds of menstrual irregularity (adjusted odds ratio (aOR) 1.99, 95% confidence interval (CI) 1.51- 2.64), compared to those who remained at a normal BMI at Wave 2. In contrast, individuals with overweight/obesity at age 18 who transitioned to a normal BMI had lower odds of menstrual irregularity (aOR 0.34, 95% CI 0.18–0.64), compared to those who remained overweight/obese at Wave 2. No significant associations were observed among individuals with underweight. Supplementary Table 1 provides the population numbers for each BMI transition category, as well as the crude and adjusted odds ratios corresponding to the estimates displayed in Fig.  2 . When BMI was categorized into 5 groups at both age 18 and Wave 2, increases in BMI categories at or above the normal range were associated with higher odds of menstrual irregularity (Supplementary Table 2). Fig. 2 Long-term BMI changes and adjusted odds ratio of menstrual irregularity. The graphs depict the adjusted odds ratio (aOR) and 95% confidence intervals for menstrual irregularity across categories of BMI at age 18. The x-axis shows the corresponding aOR values for each subcategory of BMI at Wave 2. The reference group is indicated as having no change in BMI category from age 18 to Wave 2. Red points highlight significant associations ( p  < 0.001) Long-term BMI changes and adjusted odds ratio of menstrual irregularity. The graphs depict the adjusted odds ratio (aOR) and 95% confidence intervals for menstrual irregularity across categories of BMI at age 18. The x-axis shows the corresponding aOR values for each subcategory of BMI at Wave 2. The reference group is indicated as having no change in BMI category from age 18 to Wave 2. Red points highlight significant associations ( p  < 0.001) Figure  3 displays the association between short-term BMI change and new-onset menstrual irregularity. Transitioning from a normal BMI to overweight/obesity was associated with increased odds of developing menstrual irregularity in the post period (aOR 2.46, 95% CI 1.30–4.62). Transitioning to a normal BMI from either underweight or overweight/obesity was not statistically significant (aOR 0.64, 95% CI 0.21–1.96 and aOR 0.76, 95% CI 0.26–2.17, respectively). Supplementary Table 3 provides the population numbers for each BMI transition category, along with the crude and adjusted odds ratios corresponding to the estimates presented in Fig.  2 . In the 5-category analyses, only individuals who transitioned from a normal BMI to overweight showed a significant increase in the odds of menstrual irregularity (Supplementary Table 4). Fig. 3 Short-term BMI changes and adjusted odds ratio of new-onset menstrual irregularity. The graph depicts the adjusted odds ratio (aOR) and 95% confidence intervals for menstrual irregularity across categories of BMI at Wave 1. The x-axis shows the corresponding aOR values for each subcategory of BMI at Wave 2. The reference group is indicated as having no change in BMI category from Wave 1 to Wave 2. Red points highlight significant associations ( p  < 0.001) Short-term BMI changes and adjusted odds ratio of new-onset menstrual irregularity. The graph depicts the adjusted odds ratio (aOR) and 95% confidence intervals for menstrual irregularity across categories of BMI at Wave 1. The x-axis shows the corresponding aOR values for each subcategory of BMI at Wave 2. The reference group is indicated as having no change in BMI category from Wave 1 to Wave 2. Red points highlight significant associations ( p  < 0.001) Short-term transitions into or maintaining a normal BMI (“normal to normal” and “overweight/obese to normal”) were associated with healthier lifestyle patterns, including higher proportions reporting regular exercise, an increase in exercise load, and improved eating habits, compared with those who transitioned to or remained in the overweight/obese category. Among these lifestyle variables, only improved eating habits reached statistical significance for the initially overweight category (Supplementary Table 5). Individuals who remained underweight exhibited the lowest proportion of regular exercise, an increase in exercise, and modified eating habits.

Long-term and short-term changes in BMI were associated with differences in menstrual regularity. A long-term, but not short-term, transition from overweight/obesity to the normal BMI category was associated with a reduced risk of irregular cycles, whereas increases from a normal BMI to overweight/obesity (both in the short and long term) increased the risk. Maintaining a normal BMI was shown to be beneficial for individuals who were initially in the normal range. Future research is needed to clarify effective strategies for weight optimization across BMI categories, particularly among individuals with underweight, and to develop approaches that support healthy BMI maintenance for reproductive health.

This study examined the association between within-individual changes in BMI and menstrual irregularity using real-world data from more than 5,400 individuals and 126,000 cycles. Fluctuations in BMI were associated with changes in menstrual irregularity. Individuals with overweight or obesity who experienced long-term BMI reduction to the normal BMI category had significantly lower odds of menstrual irregularity, whereas transitions from normal BMI to overweight/obesity increased the risk. These findings suggest that gradual weight reduction may benefit menstrual health among individuals with overweight/obesity, and that maintaining a normal BMI may support regularity of cycles for those already within the normal range. Few studies have explored within-individual changes in body weight in relation to menstrual irregularities in the general population. Ko et al. reported that both weight gain and loss were associated with increased odds of self-reported menstrual irregularity, with the strongest effects among individuals with obesity [ 8 ]. Our findings are consistent with their results for weight gain in individuals with normal to overweight/obesity. This finding is also consistent with the well-established effects of excess adiposity on ovulatory function, mediated through mechanisms such as hyperinsulinemia and elevated peripheral estrogen production [ 23 ]. However, our results differ from Ko et al. regarding weight loss. In our study, the long-term transition from overweight/obesity to a normal BMI is associated with lower odds of menstrual irregularity. Ko et al.’s that weight loss increased irregularity may reflect unmeasured confounding, such as unintended or rapid weight loss, movement into the underweight range, or behaviors related to eating disorders, all of which are independently associated with anovulation and irregular cycles [ 24 ]. In addition, individuals who maintained a normal weight had lower odds of irregularity than those who transitioned from a normal weight to overweight/obesity, highlighting the importance of maintaining a normal weight. Taken together, the findings reinforce the reproductive health benefits of achieving and maintaining a normal BMI through gradual and sustainable weight management. In our study, no significant associations were observed between BMI changes and menstrual irregularity among individuals with underweight. Clinical studies of functional hypothalamic amenorrhea, often seen in individuals with severe underweight, show that weight restoration often leads to the resumption of menses [ 25 ], and in exercising women with menstrual disturbances, structured weight gain interventions are associated with improvement in menstrual function [ 26 ]. However, the timeframe for recovery varies widely, ranging from several months to years [ 27 ]. Moreover, weight alone does not reliably predict menstrual recovery, as some individuals experience prolonged amenorrhea after achieving a BMI in the normal range, highlighting the complexity of the pathophysiology [ 28 ]. Long-term reproductive outcomes appear more favorable. A systematic review found that individuals who recover from anorexia nervosa and achieve weight restoration show fertility, pregnancy, and childbirth rates comparable to the general population, indicating no persistent reproductive impairment following adequate treatment [ 29 ]. In a community-based setting, modest increases in BMI in the underweight may reflect gains in lean mass rather than fat mass, which may potentially be insufficient to alter ovulatory function. Further research is warranted to clarify the extent to which weight gain improves menstrual regularity among individuals with underweight in the community populations. Our findings suggest that maintaining or achieving a normal BMI may help reduce the risk of menstrual irregularity in the general population. Menstrual irregularity is closely linked to ovulatory dysfunction, a key marker of reproductive health. In our previous report, both low and high BMI were associated with longer and more variable menstrual cycles, as well as a lower proportion of biphasic (ovulatory) cycles, indicating that menstrual irregularity often reflects impaired ovulation [ 30 ]. Although infertility outcomes were not assessed in the present study, understanding how BMI changes influence menstrual regularity provides insight into broader patterns of reproductive physiology. Among individuals with obesity and infertility, weight management trials provide additional context. Randomized controlled trials in individuals with obesity and infertility have shown that lifestyle-based weight loss interventions do not improve live birth rates when compared to prompt initiation of infertility treatment [ 31 , 32 ]. However, both trials showed higher rates of natural conception in individuals assigned to the intervention before infertility treatment [ 31 – 33 ]. Report from meta-analyses further supports these observations, that lifestyle interventions improve spontaneous pregnancy and live birth when compared with minimal or no intervention, although these benefits disappear when compared with immediate access to ART [ 34 ]. Among individuals with underweight, appropriate treatment of weight restoration has been associated with the return of menses [ 25 ], and in the long term, appears sufficient to normalize fertility potential [ 29 ]. However, as stated, the restoration period is reported to vary, and weight gain does not always guarantee the return of ovulation. Importantly, both weight-loss interventions for individuals with obesity and weight-restoration programs for individuals with underweight report high dropout rates, underscoring the difficulty of sustaining behavioral change over time. A systematic review of lifestyle interventions for overweight and obese women found median dropout rates of approximately 24% [ 35 ]. Nutritional intervention study for the undernourished population reports dropout rates exceeding 40–50% over 12 months [ 26 ]. These challenges highlight that short-term changes or effective weight management after the development of infertility may not often be feasible in real-world settings, and that durable, long-term changes are likely necessary for improvements in reproductive function. In our study, long-term, but not short-term, normalization of BMI was associated with a reduced risk of menstrual irregularity. Conversely, transitions from a normal BMI to the overweight/obesity range (both long-and short-term) were associated with increased irregularity. These findings highlight the importance of sustained weight management over short-term optimization. Cycle tracking may also provide visible indicators of physiological improvement, such as normalization of cycle length or the return of ovulatory patterns, which could help reinforce weight-modification efforts. Evidence from other health domains suggests that digital self-monitoring tools, including mobile apps and wearable trackers, can enhance engagement in health-promoting behaviors [ 36 , 37 ]. However, this possibility requires further evaluation in the context of menstrual health. To our knowledge, this is the first study to investigate the association of both short- and long-term changes in BMI and menstrual cycle irregularity using real-world data. The use of prospectively logged cycles can minimize recall bias and allow for a precise characterization of menstrual irregularity, compared to traditional self-reported surveys. By evaluating BMI changes across two time scales, both fluctuations over a few months and long-term trajectories since adolescence, the study provides insight into the benefits of gradual weight management. Furthermore, unlike previous studies that assessed weight changes as continuous measures, the use of BMI categories enabled us to distinguish whether weight changes represented movement into the normal BMI range or excessive loss into the underweight range, providing more insight into the reproductive implications of weight change. Several limitations should be considered. First, selection bias may exist, as individuals who experience absent or infrequent menstrual bleeding may not utilize menstrual tracking apps, leading to underrepresentation. Generalizability may also be limited because participation required smartphone access and sufficient digital literacy. App users may differ from the broader population in economic status and health awareness. Nonetheless, smartphone ownership in Japan was 74.3% in 2021, and was considerably higher among younger adults, indicating broad but not universal coverage [ 38 ]. In addition, as BMI is reported to vary with ethnicity [ 39 ], findings may not be generalizable beyond Asian populations. Second, information on BMI was limited in several aspects. BMI was only available at age 18, and at the time of questionnaire distribution, which may not fully capture weight changes over time. For older participants, long-term changes in BMI may therefore reflect multiple unmeasured transitions. BMI was also calculated from self-reported height and weight. Although validation studies in Japanese populations have shown high correlations between self-reported and measured anthropometrics, self-reports tend to slightly underestimate the prevalence of both underweight and obesity at the population level [ 40 ]. Third, several potential confounders related to reproductive and lifestyle factors were not collected. Parity was approximated using co-residence with a child, because detailed obstetric history and breastfeeding information, both of which may influence menstrual patterns, were not available. Information on hysterectomy status was also not collected; however, because participants contributed data only by actively logging menstrual bleeding in the app, individuals without menstrual bleeding (e.g., post-hysterectomy) would be unlikely to appear in the analytic dataset. Additionally, data on total caloric intake and dietary quality were also not available. Since these factors influence both BMI and menstrual function, the absence of these variables may limit the extent to which lifestyle-related confounding can be fully accounted for. Further research incorporating more comprehensive data is warranted. Lastly, PCOS may lie on the causal pathway between weight changes and menstrual irregularity, raising the possibility of overadjustment when included as a covariate. However, only 36 participants (< 1%) reported being diagnosed with PCOS, and stratified analyses were not feasible. A sensitivity analysis excluding PCOS from the adjustment set yielded similar results, suggesting that the primary findings were robust.

The menstrual cycle is often considered a “vital sign” of overall health and reproductive health in individuals of reproductive age [ 1 ]. Irregular menstrual cycles are a common gynecological concern that may indicate underlying endocrine or metabolic disturbances, in addition to being associated with reduced quality of life [ 2 ] and adverse health outcomes [ 3 – 5 ]. While obesity independently contributes to adverse health outcomes, menstrual irregularity has also been linked to these outcomes, and these associations persist even among individuals without obesity and after adjusting for BMI [ 5 ]. These findings highlight the importance of identifying modifiable factors that contribute to menstrual irregularity. Population-based studies have reported a nonlinear, U-shaped association between BMI and irregular menstruation [ 6 , 7 ], suggesting that individuals with normal weight tend to experience the most regular menstruation. However, the effects of within-individual changes in BMI on menstrual cycles remain less studied. To our knowledge, one previous study has examined this association, reporting that both weight gain and loss were linked to higher odds of self-reported menstrual irregularity, with statistically significant effects among individuals with obesity [ 8 ]. These findings appear counterintuitive, as prior research has suggested that individuals with normal weight tend to experience the fewest menstrual irregularities. This discrepancy highlights the importance of considering changes in BMI and corresponding BMI categories, rather than weight change alone, to assess whether transitions toward a normal BMI are associated with improved menstrual regularity in the general population. This study aimed to examine the association between within-individual changes in BMI and odds of menstrual irregularity using real-world data from a menstrual tracking application. BMI changes were evaluated from both long- and short-term perspectives, encompassing variations over several years from adolescence as well as fluctuations occurring over a few months. These measures enable evaluation of how transitions between BMI categories correspond to menstrual irregularity.

Supplementary Material 1. Supplementary Material 2. Supplementary Material 1. Supplementary Material 2.