Birth weight, childhood body mass index and height and risks of endometriosis and adenomyosis

In the study we included 171,447 women. During 40 years and 5 million women-years of follow-up, 2149 women had a diagnosis with endometriosis and 1410 with adenomyosis. Among the cases of endometriosis, 950 (44.2%) were located in the ovaries, 333 (15.5%) in the pelvic peritoneum, 399 (18.6%) in other designated sites and the remaining 467 (21.7%) in unspecified sites ( Supplementary Table 1 ). The median age at diagnosis was 37 years (5–95 percentiles: 23–52 years) for endometriosis overall, 39 years (5–95 percentiles: 25–53 years) for ovarian endometriosis, 34 years (5–95 percentiles: 22–48 years) for pelvic peritoneal endometriosis, and women were generally diagnosed with adenomyosis later with a median age at diagnosis of 45 years (5–95 percentiles: 31–60 years). The median values of childhood BMI and height increased with childhood age as anticipated ( Table 1 ). Further, across the 66 years of birth included in this study, mean childhood BMI and height increased as well. In the CSHRR, the mean birthweight was 3.3kg across the three birth cohorts ( Baker et al. 2008 ). We found that childhood BMI was inversely and significantly associated with risks of endometriosis, such that girls with the highest BMI had a moderately reduced risk of endometriosis ( Figure 1 , Supplementary Table 2 ). For example, at age 7 years the HR was 0.92 (95% CI: 0.88–0.96) per BMI z-score. In addition, we consistently found that childhood BMI was inversely associated with risks of ovarian and pelvic peritoneal endometriosis. For ovarian endometriosis the HR was 0.91 (95% CI: 0.85–0.98) and for pelvic peritoneal endometriosis the HR was 0.87 (95% CI: 0.78–0.97), per BMI z-score at age 7 years. In contrast, we found very limited evidence of associations between childhood BMI and risks of adenomyosis. Although the HRs tended to be positive (HR=1.04 [95% CI: 0.98–1.10], per BMI z-score at age 7 years), they reached statistical significance only at a few ages. The associations between childhood BMI and risks of endometriosis, including specific locations, and adenomyosis, respectively, were similar at all childhood ages. In contrast with BMI, childhood height was positively associated with risks of endometriosis, whereby the tallest girls had the highest risks ( Table 2 ). At age 7 years, the HR was 1.09 (95% CI: 1.05–1.14) per height z-score. Further, we similarly found positive associations between childhood height and risks of ovarian and pelvic peritoneal endometriosis. For ovarian endometriosis the HR was 1.09 (95% CI: 1.02–1.17) and for pelvic peritoneal endometriosis the HR was 1.11 (95% CI: 1.00–1.24), per height z-score at age 7 years. We did not find strong evidence of associations between childhood height and risks of adenomyosis with a HR of 0.97 (95% CI: 0.92–1.02) per z-score at age 7 years. The associations between childhood height and risks of endometriosis, including specific locations, and adenomyosis, respectively, were similar at all childhood ages. As we detected limited indications of non-linearity in some of the associations between childhood height and risks of endometriosis and adenomyosis, respectively, risk estimates for associations with below- and above-average height are additionally presented ( Supplementary Table 3 ). There was a tendency that among girls with a below-average height z-scores, the shortest girls had the lowest risks, whereas associations were generally not observed among girls with above-average height. In the birth weight analyses including 132,270 girls, during follow-up 1780 women were diagnosed with endometriosis and 1097 with adenomyosis. Birth weight was not associated with risks of endometriosis in linear analyses or in categorical analyses ( Table 3 ). Similarly, we did not find any associations between birth weight and ovarian or peritoneal endometriosis or with adenomyosis.

The Copenhagen School Health Records Register (CSHRR) is a unique Danish database containing computerized information from school health examinations on nearly all children born from 1930 through 1996 and attending public or private schools in Copenhagen ( Baker et al. 2009 ). As part of the school-based health care program, weight and height were measured annually by school doctors or nurses at ages 7–13 years through 1983 using standardized procedures. Thereafter the examinations generally occurred only at school entry and exit. The children were measured naked or in underwear and from 1973 light clothing was permitted. Parents or guardians reported their child’s birth weight (either from recall or birth records) at the first school health examination from the birth year 1936 onwards, and these birth weights have a high validity ( Jensen et al. 2015 ). The anthropometric information was systematically recorded on individual health cards along with the child’s birth name, sex and date of birth. Beginning on 2 April 1968, personal identification numbers were assigned to all Danish residents who were alive or born thereafter ( Pedersen 2011 ). For children attending school at this time or later, the identification number was recorded on their health card, and it was retrieved for those who left school prior to this time ( Baker et al. 2009 ). Using this identification number, girls in the CSHRR were prospectively followed through record linkages to national health registers. Vital status was obtained through linkage to the Vital Statistics Register ( Pedersen 2011 ). Information on endometriosis and adenomyosis diagnoses were available from the nationwide Danish National Patient Register, which contains information on all inpatient hospital contacts since 1977 and all ambulatory outpatient contacts since 1995 ( Lynge et al. 2011 ). Using the date at the first hospital admission, endometriosis was defined according to the International Classification of Disease (ICD) codes using the 8 th revision from 1977 to 1994 (ICD-8: 625.30, 625.32–625.39) and the 10 th revision thereafter (ICD-10: N80, N80.1-N80.9). Adenomyosis was defined by the following codes: ICD-8: 625.31 and ICD-10: N80.0. In Denmark, the 9 th revision of the ICD was never used. Diagnoses of endometriosis were further sub-divided based on the location of the endometriosis, either in the ovaries, the pelvic peritoneum or elsewhere ( Supplementary Table 1 ). Women eligible for inclusion in this study were born 1930–1996, had an identification number and were alive and living in Denmark at the age of 15 years or older or January 1, 1977. Among these 175,280 women, individuals were excluded due to having a hysterectomy before age 15 years (N=2) or before 1 January, 1977 (N=3), being diagnosed with endometriosis or adenomyosis before age 15 years (N=2), having missing (N=3822) or outlying childhood height and BMI z-scores at all ages (z-score 4.5) (N=4) leaving a total of 171,447 women for analyses. In the subsample of 154,414 women born 1936–1996 available for the birth weight analyses, women with missing birth weight information (N=21,144) and with birth weights outside of the reliable range of 2.0–5.5 kilograms (kg) (N=2701) were excluded resulting in 130,569 women available for these analyses. Follow-up started on January 1, 1977, or at the age of 15 years, whichever came later, and ended on the date of a diagnosis of endometriosis or adenomyosis, hysterectomy, death, emigration, loss to follow-up, or June 30, 2017, whichever came first. Childhood BMI values were calculated. To allow comparisons across age, childhood BMI and height values were transformed into z-scores using an internal reference. For BMI z-scores, the reference was chosen from a period with a low and stable prevalence of overweight (1955–1960), whereas height z-scores were birth cohort-specific to account for secular increases in height over time. The Lambda Median Sigma method was used to reduce skewness in distributions ( Cole and Green 1992 ). Unless a measurement was taken at the exact age, z-scores were interpolated or extrapolated to exact ages within a 12-month period. Body size characteristics of the cohort are presented as medians and 5–95 th percentiles at birth and ages 7–13 years. Using Cox proportional hazard regressions, associations between birth weight, childhood BMI and height z-scores, respectively, and risks of endometriosis, including specific locations, or adenomyosis were investigated. These cause-specific hazards ratios (HR) are presented with 95% confidence intervals (CI). Age was used as the underlying time scale and all analyses were Cox stratified by birth cohort, which allows the baseline hazard to differ by birth cohort. Beyond the information on birth cohort, no other potential confounding variables were available in the CSHRR for the statistical analyses. In the analyses of endometriosis as the outcome, women with a diagnosis of adenomyosis were censored on the date of this diagnosis and vice versa. In analyses with a location-specific endometriosis as the outcome, women with a diagnosis of endometriosis at another location or with adenomyosis were censored on the date of this diagnosis. Further, using a subdistribution hazard regression model, where the outcomes were treated as competing risks, we calculated subdistribution hazard ratios ( Fine and Gray 1999 ). However, the associations between early life body size and risks of endometriosis and adenomyosis were essentially similar (results not shown). The linearity of the associations with childhood body size was assessed in models using piecewise linear splines, in comparison with a linear model, using the likelihood ratio test. We did not detect non-linearity in any of the associations between childhood BMI and endometriosis or adenomyosis using three knots positioned at z-scores of −0.68, 0 and +0.68 (all p-values ≥ 0.09), thus, associations are presented per BMI z-score in a linear model. We found limited indications of non-linear associations between childhood height and risks of endometriosis and adenomyosis. As such, besides linear estimates per height z-score, height was modelled and presented using piecewise linear splines with a single knot at a z-score of 0, as this model most appropriately represented the shape of the association. In analyses on birth weight, the linearity of the associations was also assessed using piecewise linear splines with two knots positioned at 3.25 and 3.75 kg, in comparison with a linear model, using the likelihood ratio test. Deviations from linearity were not detected in these associations (all p-values ≥ 0.08). Nonetheless, associations are presented both linearly per kilogram of birth weight and in categories (2.00–3.25, 3.26–3.75, 3.76–5.50 kg) for comparison with the literature. These categories were selected to reduce the impact of digit preference ( Baker et al. 2008 ). The assumption of proportional hazards was investigated by testing if the associations between early life body size and risks of endometriosis and adenomyosis, respectively, differed by categories of age at risk using the likelihood ratio test. No violations of the proportional hazards assumption were detected (all p-values ≥ 0.06) indicating that there are no differences in the associations across age at diagnosis. Similarly, using the likelihood ratio test, we tested for birth cohort effects by including an interaction term between the exposures and a categorical birth cohort variable (1930–1949, 1950–1969, 1970–1996). However, we did not observe any birth cohort interactions (all p-values ≥ 0.10). In the subsample of women born 1936 onwards with information on birth weight, we tested for interactions between birth weight and childhood BMI in the association with endometriosis and adenomyosis, respectively, using the likelihood ratio test. No signs of interactions were found (all p-values ≥ 0.14). This study was approved by the Danish Data Protection Agency. According to Danish law, ethical approval is not required for register-based studies.

In this large longitudinal population-based cohort study of Danish school children we found that leanness and tallness among girls were associated with increased risks of endometriosis overall and ovarian and pelvic peritoneal endometriosis. However, we did not detect associations between childhood BMI or height and risks of adenomyosis. Further, birth weight was not significantly associated with risks of endometriosis or adenomyosis. We found evidence supporting an inverse association between child BMI and risks of endometriosis overall and ovarian and pelvic peritoneal endometriosis, such that the leanest girls have the highest risks of endometriosis. We did not find associations between childhood BMI and adenomyosis risks. In our cohort, a one z-score difference in BMI between two 7-year old girls (equivalent to 2.3kg difference among girls of average-height) was associated with an 8%, 9% and 13% increased risks of endometriosis overall and ovarian and pelvic peritoneal endometriosis, respectively, for the leanest girl. Other work in this area has yielded conflicting and inconclusive results, which could partly be due to methodological constraints, as all studies were based on recalled body size and different diagnostic methods. Partially consistent with our findings, a French study using the E3N cohort ( Farland et al. 2017 ), an American study using the Nurses’ Health Study II ( Vitonis et al. 2010 ) and an American case-control study ( Hediger et al. 2005 ) provide some evidence for inverse associations between recalled childhood or adolescent body size and risks of endometriosis, which is also in agreement with the current evidence for adult body size ( Liu and Zhang 2017 ). In contrast, an Australian case-control study, also relying on recalled body size, reported positive associations between childhood overweight and endometriosis risks ( Nagle et al. 2009 ). In our study, in addition to using measured weight and height values from a larger cohort with more cases of endometriosis, we extend the previous work in this area by examining associations by endometriosis location as well as specifically investigating adenomyosis, which proved to be important, as associations differed between these diseases. In contrast with the inverse findings for childhood BMI, we found evidence that associations with height during childhood ages were positive, thus, the tallest girls had the highest risks of endometriosis. To put these findings into perspective, a one z-score difference in height between two girls who are of either average height or taller than average (equivalent to 5.2 cm difference) is associated with a 9%, 9% and 11% increased risk of endometriosis overall and ovarian and pelvic peritoneal endometriosis, respectively, for the tallest girl. Our findings for childhood height are consistent with the findings for attained adult height, where tallness increases the risks ( Farland et al. 2017 ). We found limited evidence for associations between childhood height and risks of adenomyosis. Thus, results from this study add to the current evidence by showing that height already during childhood has associations with endometriosis, but not with adenomyosis. Most previous studies have not investigated associations between anthropometrics and endometriosis by locations and there is additionally limited evidence for associations with adenomyosis. These examinations are important as the risk factor pattern likely differs. Although it has been debated whether endometriosis and adenomyosis constitute different disease entities, it is generally acknowledged that they are ( Benagiano et al. 2012 ). Nonetheless, adenomyosis is still included in the overall classification of endometriosis by the ICD codes. This fact could potentially explain why some studies do not separate internal and external ectopic growth of the endometrium (i.e. endometriosis and adenomyosis). The most widely suggested hypothesis of the origin of endometriosis is retrograde menstruation through the fallopian tubes into the pelvic cavity resulting in endometrial cells at ectopic sites, where they can implant, grow and progress ( Vercellini et al. 2014 ; Zondervan et al. 2018 ). The biological mechanisms underlying the relationship between body size and risks of endometriosis is, however, currently unknown. Oestrogens are key promoters of endometrial cellular growth and adiposity stimulates the production of oestrogens through an increased aromatase activity ( Siiteri 1987 ). However, the relative contribution of adipose tissue to the level of oestrogens is largely masked by the ovarian contribution in premenopausal women, complicating the evaluation of the possible influence of oestrogens in the association between adiposity and endometriosis. A large pre-pubertal body size is generally thought to result in an earlier age at menarche ( Juul et al. 2017 ). As an early age at menarche is consistently shown to increase risks of endometriosis, the timing of puberty is thereby unlikely to explain our findings. Thus, the inverse associations between childhood BMI and risks of endometriosis may be operating through other unknown biological or environmental mechanisms. Growth in height is determined by complex processes involving various genetic, environmental, endocrine and nutritional factors. Evidence suggests that oestrogen exposure is crucial for the acceleration in linear growth that occurs during puberty ( Drop et al. 1998 ). Thus, this is one potential mechanism contributing to the positive association between tallness and the oestrogen-related disease of endometriosis. Our novel findings of positive associations between childhood height and endometriosis risks are consistent with the current literature on attained adult height and endometriosis risks ( Farland et al. 2017 ). In addition to childhood BMI and height, we also investigated if birth weight was associated with risks of endometriosis, including specific locations, and adenomyosis. Several previous studies have assessed the association between size at birth and risks of endometriosis, however, they have yielded inconsistent findings with some studies finding a positive association between a low birth weight and endometriosis risks ( Missmer et al. 2004 ; Borghese et al. 2015 ; Gao et al. 2019 ), whereas others found no association ( Wolff et al. 2013 ; Upson et al. 2015 ). Currently, there is very limited evidence evaluating the association between birth weight and endometriosis risks by location or adenomyosis. A Swedish study including 111 women diagnosed with endometriosis found that a large size at birth was positively and strongly associated with endometriosis but not adenomyosis ( Gao et al. 2019 ). Although our study included 29 times more endometriosis cases, we did not find any indications that size at birth is related to future risks of endometriosis, whereas our null-findings on adenomyosis are consistent with the previous Swedish study. Taken together, our findings suggest that growth after foetal life may be more important for later risks of endometriosis than weight at birth. Among the major strength of our study is the prospective study design using a large cohort of practically all children attending school in the Copenhagen area during numerous decades. As the children underwent health examinations at both public and private schools, the risk of selection into the cohort is limited. The children’s heights and weights were measured, as opposed to self-reported, thus, enhancing the validity and reliability of the findings. Further, the potential for bias associated with recall of childhood body size as well as survival into adult ages for inclusion into the cohort are limited. Although birth weight was not measured, it was obtained by parental recall or from birth records and the short period of recall (approximately 7 years) has a high degree of accuracy as found both within this cohort ( Jensen et al. 2015 ) and in other studies ( O’Sullivan et al. 2000 ; Walton et al. 2000 ; Rice et al. 2007 ). Further, we were able to follow the children for an extensive period through the national health registers with virtually no loss to follow-up. As universal health care is provided without direct costs to the patients in Denmark, the risk of selection bias is also minimised. Our study also has some limitations. As our follow-up period spans numerous decades, it is evident that the diagnostic criteria and registration of endometriosis and adenomyosis have changed as well as the awareness of these diseases in the health care system. Laparoscopic visualization with histological verification is considered the gold standard for a diagnosis of endometriosis, but often a tentative diagnosis is based on clinical examination including ultrasound evaluation and symptoms ( Zondervan et al. 2018 ). Furthermore, the technique used to make a diagnosis has changed over time, due to improvements in laparoscopic surgery and ultrasound, yet the choice of the technique is also related to the location of the endometriosis or adenomyosis. Nonetheless, although the sensitivity and specificity are likely to have improved with time, we did not detect any birth cohort effects in the associations, suggesting that these changes are unlikely to explain our findings. Endometriosis and adenomyosis cases in our study were ascertained through the Danish National Patient Register, which does not contain information on histological verification of diagnoses. However, a validation study including women admitted for gynaecological surgery found a high positive predictive value (95 %) and a moderate sensitivity (46 %) for endometriosis (including adenomyosis) with some variations across the anatomical location of the disease ( Kjaergaard et al. 2002 ). Women diagnosed in a hospital-setting, as in the present study, likely represent the most severe cases, as opposed to women diagnosed at their general practitioner or gynaecologist, thus, the generalizability of our findings to less severe endometriosis or adenomyosis cases may be limited. Some women with endometriosis or adenomyosis are asymptomatic, or their symptoms may be non-specific. When combined with a lack of awareness of these diseases and limitations of diagnostic procedures, it is possible that some affected women are undiagnosed. Further, we use the age at the first diagnosis based on a hospital contact as a proxy for the age at disease onset although these diseases may have already been present for a considerable amount of time. Thus, in combination with the structure of our data resource and the availability of national health registers, the actual age at disease onset may be overestimated. Nonetheless, as we found that the associations were similar across all adult ages, and as a diagnostic delay is likely unrelated to early life body size, any potential misclassification would tend to bias our findings towards the null. We were sufficiently powered to investigate associations with ovarian and pelvic peritoneal endometriosis and adenomyosis, but not endometriosis in rare locations. Further, we have limited information on covariates of potential interest such as menstrual and reproductive factors as well as the use of oral contraceptives and hormone replacement therapy, adult anthropometrics and a family history of endometriosis. Nonetheless, most of these factors would serve as mediators rather than confounders, which are beyond the aim of this study. Girls included in this study were born from the 1930s through the 1990s, thus incorporating children born before, during and after the Second World War. However, in comparison with other European countries, the Danish population was food sufficient and childhood body size was minimally affected by the war ( Angell-Andersen et al. 2004 ). Further, during the period of 66 birth years, significant improvements in living conditions and medical advances occurred and substantial socioeconomic changes happened. The associations between early life body size and risks of endometriosis and adenomyosis did not differ by birth cohort, indicating that the associations are likely attributable to biological factors rather than other factors such as socioeconomic differences. As we included children born before and after the emergence of the obesity epidemic, our findings are likely also generalizable to contemporary populations of children. Further, children in our study were mainly of Northern European Caucasian descent, likely making our results applicable to large proportions of Western populations. In conclusion, we found that leanness and tallness during school ages were associated with increased risks of endometriosis, but not adenomyosis. As the associations differed between endometriosis and adenomyosis diagnoses, our findings highlight that investigations of risk factors should be done separately for these diseases. We did not find that birth weight, as an indicator of intrauterine growth, was associated with risks of endometriosis or adenomyosis. Taken together, these findings provide novel insights into the aetiology of endometriosis and adenomyosis and aid in the current limited understanding of the critical time window for anthropometric exposures in relation to risks of these diseases.

Endometriosis is a gynaecological oestrogen-related condition which is estimated to affect approximately 5–10% of the general female population, primarily at reproductive ages, however, it is likely underdiagnosed ( Giudice 2010 ; Vercellini et al. 2014 ; Zondervan et al. 2018 ). The disease is defined as the presence of endometrial glands and stroma-like lesions outside of the uterus. Endometriotic lesions may be present in several different locations, but the most common sites are in the ovaries and pelvic peritoneum, although endometriotic implants are also found in extra-pelvic locations ( Burney and Giudice 2012 ; Zondervan et al. 2018 ). The disease is debilitating, often chronic in nature and reduces the health and quality of life in women. Although the symptoms vary and may even be absent, the disease is often associated with reduced fertility and cyclic menstrual and pelvic pain, which can be severe ( Burney and Giudice 2012 ). Further, increasing evidence suggests that women diagnosed with endometriosis have a high risk of developing several other chronic diseases, including cardiovascular diseases and female gynaecological cancers ( Brinton et al. 2005 ; Pearce et al. 2012 ; Kim et al. 2014 ; Kvaskoff et al. 2015 ; Mogensen et al. 2016 ). Despite the impact of endometriosis on women’s lives, little is known about the aetiology and pathogenesis of this disease. Among the few established risk factors are those related to menstruation (early age at menarche and shorter cycles) and a family history of endometriosis ( Zondervan et al. 2018 ). Leanness and tallness in adulthood is also suggested to be associated with increased risks of endometriosis ( Farland et al. 2017 ; Liu and Zhang 2017 ). In the current literature, few studies have investigated if the risk factor profile differ by location of the endometriosis. Adenomyosis is a benign but often painful gynaecological disease characterized by the invasion of endometrial glands and stroma within the uterine myometrium ( Benagiano et al. 2012 ; Abbott 2017 ). As with endometriosis, adenomyosis is likely underdiagnosed. Many clinical symptoms overlap between endometriosis and adenomyosis and they are suggested to share several aetiological factors. In addition, a considerable proportion of women are diagnosed with both diseases ( Chapron et al. 2017 ). As such, it has been debated whether endometriosis and adenomyosis are part of a continuum of diseases, with adenomyosis usually being diagnosed at later ages than endometriosis, or whether they constitute two separate entities ( Abbott 2017 ). However, it is generally acknowledged that the internal and external ectopic growth of the endometrium should be regarded as different disorders ( Benagiano et al. 2012 ). Similar to endometriosis, little is known about the aetiology and pathogenesis of adenomyosis. Some studies suggest that, in contrast to endometriosis, adult overweight or obesity are positively associated with risks of adenomyosis ( Templeman et al. 2008 ; Trabert et al. 2011 ), whereas associations with adult height is unknown. Symptoms of endometriosis may present already during early adolescence, however, there is a diagnostic delay with a mean of approximately 7 years ( Nnoaham et al. 2011 ; Parasar et al. 2017 ). Thus, studies on risk factors in adult life may have missed a critical time window for endometriosis development and may additionally be affected by already existing disease. In addition, studies have shown that adenomyosis is not confined to older women but also exists among women in their early reproductive years ( Kunz et al. 2007 ; Dietrich 2010 ). Recently, evidence has emerged suggesting that early life body size may be a marker for risk of endometriosis, however, findings for weight at birth ( Missmer et al. 2004 ; Wolff et al. 2013 ; Borghese et al. 2015 ; Upson et al. 2015 ; Gao et al. 2019 ) as well as recalled childhood or adolescent body shape ( Hediger et al. 2005 ; Nagle et al. 2009 ; Vitonis et al. 2010 ; Farland et al. 2017 ) are still limited and inconsistent. Further, no studies have evaluated height early in life in relation to endometriosis risks or associations between childhood body size and risks of adenomyosis. The aim of the current study was to investigate early life indicators of risks of endometriosis and adenomyosis by examining the associations between birth weight, measured childhood body mass index (BMI; kg/m 2 ) and height at ages 7–13 years and risks of endometriosis, including specific locations, and adenomyosis using a large Danish population-based cohort.