Objective
To systematically review the literature on the association between obesity and endometrial hyperplasia or cancer in premenopausal women.
Data Sources
We searched the bibliographic databases MEDLINE, EMBASE, PubMed, and CINAHL (inception to May 5, 2015), and checked reference lists of included studies and systematic reviews.
Study Eligibility Criteria
Studies of more than 50 women with endometrial pathology diagnosed during premenopause that reported on obesity as a risk factor were eligible.
Study Appraisal and Synthesis Methods
Study identification and data extraction were independently performed by 2 authors. Where possible, data were pooled in a generic inverse variance forest plot. Heterogeneity was reported using the I 2 statistic.
Results
Nine case-control studies of moderate quality were included. Quantitative analysis of 5 studies showed a dose–response relationship of body mass index and increased risk of endometrial cancer. For studies of women with body mass index of ≥25, the pooled odds ratio was 3.85 (95% confidence interval 2.53–5.84); body mass index of ≥30 was 5.25 (4.00–6.90); and body mass index of ≥40 was 19.79 (11.18–35.03).
Conclusion
Body mass index is a consistent and leading risk factor for endometrial complex hyperplasia or cancer in premenopausal women. Body mass index should be considered when deciding to assess the endometrium in symptomatic premenopausal women.
Introduction
Endometrial cancer (EC) is the fourth most common cancer in women in the United Kingdom and the United States. Being overweight and obese has been estimated to account for 40%–45% of EC incidence in Europe and 57% in the United States. A recent Cochrane Library Editorial highlighted the increasing rates of EC and the urgent need for research into its etiology, screening, prevention and treatment.
In women of reproductive age, in the absence of ovulation, the endometrium is exposed to continuous estrogen, which can lead to endometrial hyperplasia (EH). If identified in a timely manner, EH can usually be successfully treated. Complex EH can progress to EC in up to one-fourth of women. Complex EH with atypia is associated with co-existent EC in up to one-half of women.
The leading risk factors for EH and EC include age, nulliparity, diabetes, and obesity. As complex alterations in hormonal and metabolic factors are associated with obesity, it is likely that the prolonged and constant production of estrogen from adipose tissue results in an increased risk of EH and EC. Over the past 2 decades, there has been increasing global recognition of rising levels of obesity. Previous reviews of the relationship between obesity and EC have looked at all women, and not necessarily accounted for menopausal status.
The objective of this review is to systematically review the literature that considers the association between obesity and EH and EC, specifically in premenopausal women.
Methods
Where possible, we have used the methods described in the Cochrane Handbook, and have reported according to the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines.
The protocol for this review was registered on the International Prospective Register of Systematic Reviews (PROSPERO) found at http://www.crd.york.ac.uk/prospero/ (registration number CRD42015019638). Ethics approval was not needed because this is a literature study.
Information sources and search strategy
The following electronic databases were searched, from inception to March 11, 2015 (see Appendix, Supplementary Table 1 ): MEDLINE, EMBASE, using the OVID platform, and CINAHL (Cumulative Index to Nursing and Allied Health Literature) using the EBSCO platform. We used both free text and indexed terms for endometrial cancer, endometrial hyperplasia, body mass index (BMI), and obesity. These searches strategies included filters that limited the results to observational studies. The MEDLINE and EMBASE searches were limited to observational study design by the use of specialized filters ( http://www.sign.ac.uk/methodology/filters.html#obs ) developed by the Scottish Intercollegiate Guidelines Network. Filters were adapted using the following terms to widen the results: odds ratio, logistic models, relative risk, prevalence, and survival analysis. The CINAHL search was limited by a filter designed by the Health Science Library at the University of Texas ( http://libguides.sph.uth.tmc.edu/search_filters ).
An additional search was performed in MEDLINE and EMBASE to locate any appropriate studies that may have been discarded in the original search due to the observational study filters. These terms included endometrial cancer, hyperplasia, and obesity and then the search was limited by the keywords “premenopausal” and “young.” Observational study filters were not used in this search.
Finally, the reference lists of included studies and related reviews were hand-searched. PubMed (Appendix, Supplementary Table 1 ) was also searched to locate any studies that may have been published but not yet indexed in the major databases.
The searches were not limited by publication date; however, the included studies were limited to the English language. An ENDNOTE library was used to manage the references and to remove duplicates.
Eligibility criteria
Studies were included if they assessed obesity (eg, BMI, Quetelet’s index, both measured as kg/m 2 ) and EH and/or EC specifically in women who were premenopausal at the time of diagnosis. Where menopausal status was not explicitly stated, then age <50 years was used as a proxy.
Studies that did not include a comparative group (women who did not develop EH and/or EC, or were not exposed to the primary risk factor) were excluded (for example, case series). Studies of less than 50 women were excluded. Studies in which the only obesity measure was weight were excluded. Studies that did not report menopausal status, or that reported only menopausal status at the time of study recruitment, were excluded.
Study selection and data extraction
The studies were selected and data were extracted independently by 2 authors (A.L., N.W.). Disputes were resolved by a third author (M.W.). Study characteristics included study period, study design, study size, population, setting, obesity and other risk factor(s), and primary outcome(s). Each study reported on various risk factors other than obesity; all of these were recorded.
Assessment of risk of bias
Each study was assessed by 2 authors (V.J., A.L.) using the Newcastle Ottawa Quality Assessment scale, which is a tool used for assessing quality of observational studies.
Data synthesis
Study results included the association of obesity and other risk factor(s) with the primary outcome expressed quantitatively, as a relative risk or as an odds ratio. For studies that reported quantitative data for BMI categories, data were pooled in a fixed-effect generic inverse variance forest plot. Heterogeneity was reported using the I 2 statistic. Analysis of other risk factors was not performed.
Methods
Where possible, we have used the methods described in the Cochrane Handbook, and have reported according to the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines.
The protocol for this review was registered on the International Prospective Register of Systematic Reviews (PROSPERO) found at http://www.crd.york.ac.uk/prospero/ (registration number CRD42015019638). Ethics approval was not needed because this is a literature study.
Information sources and search strategy
The following electronic databases were searched, from inception to March 11, 2015 (see Appendix, Supplementary Table 1 ): MEDLINE, EMBASE, using the OVID platform, and CINAHL (Cumulative Index to Nursing and Allied Health Literature) using the EBSCO platform. We used both free text and indexed terms for endometrial cancer, endometrial hyperplasia, body mass index (BMI), and obesity. These searches strategies included filters that limited the results to observational studies. The MEDLINE and EMBASE searches were limited to observational study design by the use of specialized filters ( http://www.sign.ac.uk/methodology/filters.html#obs ) developed by the Scottish Intercollegiate Guidelines Network. Filters were adapted using the following terms to widen the results: odds ratio, logistic models, relative risk, prevalence, and survival analysis. The CINAHL search was limited by a filter designed by the Health Science Library at the University of Texas ( http://libguides.sph.uth.tmc.edu/search_filters ).
An additional search was performed in MEDLINE and EMBASE to locate any appropriate studies that may have been discarded in the original search due to the observational study filters. These terms included endometrial cancer, hyperplasia, and obesity and then the search was limited by the keywords “premenopausal” and “young.” Observational study filters were not used in this search.
Finally, the reference lists of included studies and related reviews were hand-searched. PubMed (Appendix, Supplementary Table 1 ) was also searched to locate any studies that may have been published but not yet indexed in the major databases.
The searches were not limited by publication date; however, the included studies were limited to the English language. An ENDNOTE library was used to manage the references and to remove duplicates.
Eligibility criteria
Studies were included if they assessed obesity (eg, BMI, Quetelet’s index, both measured as kg/m 2 ) and EH and/or EC specifically in women who were premenopausal at the time of diagnosis. Where menopausal status was not explicitly stated, then age <50 years was used as a proxy.
Studies that did not include a comparative group (women who did not develop EH and/or EC, or were not exposed to the primary risk factor) were excluded (for example, case series). Studies of less than 50 women were excluded. Studies in which the only obesity measure was weight were excluded. Studies that did not report menopausal status, or that reported only menopausal status at the time of study recruitment, were excluded.
Study selection and data extraction
The studies were selected and data were extracted independently by 2 authors (A.L., N.W.). Disputes were resolved by a third author (M.W.). Study characteristics included study period, study design, study size, population, setting, obesity and other risk factor(s), and primary outcome(s). Each study reported on various risk factors other than obesity; all of these were recorded.
Assessment of risk of bias
Each study was assessed by 2 authors (V.J., A.L.) using the Newcastle Ottawa Quality Assessment scale, which is a tool used for assessing quality of observational studies.
Data synthesis
Study results included the association of obesity and other risk factor(s) with the primary outcome expressed quantitatively, as a relative risk or as an odds ratio. For studies that reported quantitative data for BMI categories, data were pooled in a fixed-effect generic inverse variance forest plot. Heterogeneity was reported using the I 2 statistic. Analysis of other risk factors was not performed.
Results
Study selection
A Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram was constructed to depict the flow of information through the different phases of the project ( Figure 1 ). It shows the number of references identified and the number that were included or excluded (references available on request). After duplicates were removed, 3 authors (A.L., N.W., M.S.) independently reviewed titles and abstracts of 3787 papers using the prespecified criteria for inclusion. The majority of papers were not eligible on title and abstract checking. Full-text manuscripts were retrieved for 103 of these papers. These papers were independently assessed, and any disagreements were resolved by advice from a fourth author (S.L.).
Study characteristics
Nine studies from 8 countries met the inclusion criteria. All studies were case-control design. A total of 1250 cases and 4957 controls were reported. For 8 studies, the primary outcome was EC, and for 1 study, complex EH. Additional information regarding the characteristics of the included studies can be found in Table 1 .
First author | Population/setting | Period | Study size | Study design | Risk factor (primary) | Risk factors (other) | Outcome (primary) |
---|---|---|---|---|---|---|---|
Dahlgren | Women aged 31–45 y with EC referred to cancer hospital in Goteborg, Sweden | 1968–1978 | 63 Cases; 498 controls | Case-control | Hormonal imbalance (ovarian dysfunction) | BMI, age, infertility, parity, marital status, abortion, diabetes, hypertension, smoking, COCP, hirsutism | EC |
Epplein | Women aged 19–51 y, members of Group Health in Seattle, WA, USA | 1985–2003 | Complex EH 134 cases; 134 controls Complex EH with atypia 58 cases; 58 controls | Case-control | Unopposed estrogen | BMI, weight 1 y before reference date, menstrual factors, parity, ethnicity, smoking, diabetes, hypertension, abnormal uterine bleeding, income | Complex EH with and without atypia |
Hosono | Women aged 20–79 y with EC diagnosed at cancer hospital in Japan | 2001–2005 | 78 Cases; 738 controls | Case-control | Self-reported BMI at age 20 y and at enrollment | Smoking, alcohol, regular exercise, age at menarche, parity, diabetes, COCP | EC |
Iatrakis | Women aged 43–48 y with EC in Athens, Greece | 1992–2004 | 81 Cases; 100 controls | Case-control | BMI | Parity, hypertension, diabetes, COCP, polycystic ovarian syndrome, menstrual cycle, personal or family history of cancer | EC |
La Vecchia | Women aged 28–74 y with EC admitted to hospital in Milan, Italy | 1983–1988 | 103 Cases; 739 controls | Case-control | Self-reported recent BMI | Socio-demographics, smoking, alcohol, caffeine, dietary intake, menstrual and reproductive variables, COCP | EC |
Parslov | Women aged <50 y with EC | 1987–1994 | 237 Cases; 538 controls | Case-control | BMI | Age at menarche, COCP, amenorrhea, parity, infertility, diabetes, hypertension, smoking, family history, abortion, ectopic pregnancy, hyperandrogenism, intrauterine contraceptive device | EC |
Thomas CASH Study | Women aged 20–54 y with EC in 8 centers, USA | 1980–1982 | 138 Cases; 1734 controls | Case-control | Self-reported BMI as an adult | Weight at age 18 y, race/ethnicity, education, COCP, parity, hypertension | EC |
Xu SECS | Women aged 30–69 y with EC from Cancer Registry in Shanghai, China | 1997–2001 | 317 Cases; 295 controls | Case-control | Self-reported BMI at various time points | Education, age at menarche, marital status, income, parity, family history, smoking, alcohol, regular exercise, COCP, energy intake | EC |
Yamazawa | Premenopausal women aged 27–53 y with EC in hospital in Chiba, Japan | 1989–2000 | 41 Cases; 123 controls | Case-control | Antipsychotic use | BMI, hypertension, COCP, parity, diabetes | EC |
Risk of bias of included studies
Overall the study quality was assessed as good; 2 studies had a high risk of bias in 1 domain each, and 3 studies had a high risk of bias in 2 domains. Most studies had some missing data. A summary of the risk of bias assessment for included studies is provided in Figure 2 (see Appendix, Supplementary Table 2 ).
Synthesis of results
All studies found that increased BMI was a risk factor for EC and EH. For 6 studies, other clinical factors were adjusted for in the analysis of the association of BMI with the primary outcome. After adjustment for other clinical factors, 1 of these 6 studies found BMI to no longer be significantly associated with the outcome. Results of the included studies are shown in Table 2 .
First author, reference | Association of BMI with outcome | Adjusted for | Significant association of other risk factors with outcome |
---|---|---|---|
Dahlgren | BMI, mean (SD) Cases: 25.7 ± 5.5 (n = 63); Controls: 23.8 ± 3.7 (n = 498); P <.001 | Not adjusted | Hirsutism: cases 26.0%; controls 9.1% Hypertension: cases 8.1%; controls 2.4% COCP use: cases 22.4%; controls 58.4% Smoking: cases 28.6%; controls 53.2% |
Epplein | Complex EH: BMI <25 (ref); 25–29.9: aOR 3.0 (1.2–7.2); 30.0–39.9: aOR 4.6 (2.1–10.3); ≥40: aOR 23.0 (6.6–79.8); P trend <.001 Complex EH with atypia: BMI <25 (ref); 25–29.9: aOR 2.3 (0.7–7.9); 30–39.9: aOR 3.7 (1.0–13.8); ≥40.0: aOR 13.0 (1.9–86.9); P trend .005 | Parity | Complex EH: parity: 0 (ref); 1: aOR a 0.5; 2: aOR 0.5; ≥3: aOR 0.3; P trend 0.02 Complex EH with atypia: parity: 0 (ref); 1: aOR 0.9; 2: aOR 0.4; ≥3: aOR 0.1; P trend 0.01 |
Hosono | BMI no change (ref); BMI change: ≤3 kg/m 2 : OR 2.1 (1.11–3.98); >3 kg/m 2 : OR 2.16 (1.02–4.55); P trend .036 | Age, smoking, alcohol, regular exercise, age at menarche, parity, diabetes, COCP | |
Iatrakis | BMI mean (SD) Cases: 34.4 ± 8.8 (n = 81); Controls: 28.3 ± 7.6 (n = 100); P <.001 | Not adjusted | Nulliparity: cases 55; controls 45 Irregular menstrual cycle: cases 30; controls 23 Polycystic ovarian syndrome: cases 3; controls 0 COCP: cases 0; controls 5 Diabetes: cases 17; controls 7 |
La Vecchia | BMI <20 (ref); 20–24: RR 1.0; 25–29: RR 0.8; 30–34: RR 1.2; ≥35: RR 1.9; χ 2 for trend 2.7; P trend .10 | Not adjusted | |
Parslov | BMI <20: cases 19/237, controls 50/538; 20–24: cases 131/237 controls; 328/538; 25–29: cases, 45/237 controls 118/538; ≥30: cases, 42/237 controls 42/538; | Age, family history, BMI, diabetes, hypertension, menarche, pregnancy, number of pregnancies, number of births, number of induced abortions, age at first birth, hyperandrogenism, amenorrhea, COCP, HRT, smoking, education | Family history: aOR 2.1 (1.1–3.8) Parity: 0 (ref); 2: aOR 0.3 (0.2–0.6); ≥3: aOR 0.2 (0.1–0.4) Age at first birth: ≤22 (ref) 23–25: aOR 0.5 (0.3–0.8); 26–29: aOR 0.1 (0.04–0.1); ≥30: aOR 0.02 (0.01–0.05) Induced abortions: 1: aOR 0.5 (0.3–0.9); ≥2: aOR 0.4 (0.2–0.8) COCP: never (ref); <1 y: aOR 0.4 (0.3–0.7); 1–5 y: aOR 0.2 (0.1–0.3); >5 y: aOR 0.2 (0.4) |
Thomas CASH Study | BMI <25 (ref); 25.0–29.9 RR 2.8 (1.5–5.3); 30.0–34.9 RR 7.0 (3.6–13.8); ≥ 35.0 RR 22.2 (11.4–43.5) | Age, race/ethncity, education, COCP, parity, hypertension | |
Xu SECS | Recent BMI ≤21 (ref); 21.1–23.3 RR 1.5 (0.9–2.5); 23.4–25.7 RR 2.0 (1.2–3.3); >25.7 RR 3.6 (2.2–5.9); p value for trend <0.01 | Age, education, years of menstruation, COCP, parity, family history, BMI at age 20 | Weight gain (kg) since age 20 <2.5 (ref); ≥15: aOR 2.2 (1.1–4.5); ≥20: aOR 2.8 (1.4–5.7) |
Yamazawa | BMI <25 (ref); ≥25 aOR 4.92 (1.68–14.4) | Obesity, nulliparity, antipsychotics use, estrogen use, diabetes, hypertension | Diabetes aOR 9.30 (1.56–55.4) Antipsychotic use aOR 5.42 (1.11–26.4) |
Risk factors other than BMI were significantly associated with EC and EH in individual studies: hirsutism, hypertension, nulliparity, irregular menstrual cycles, polycystic ovarian syndrome, diabetes, older age at first birth, family history of EC, and use of antipsychotic medications. The most consistent of these was nulliparity. Smoking, history of induced abortion, and use of combined oral contraceptive pill (COCP) were consistently found to be protective in individual studies.
Quantitative analysis of the association with BMI specifically was possible for 6 studies only. A study by Epplein et al was excluded, as the outcome EH was considered to be an entity different from EC for the purpose of this analysis. The other 3 studies were excluded for the following reasons: Dahlgren et al presented only mean BMI in women under 45 years, not calculated OR (although there was a significant difference in mean BMI (25.7 ± 5.5 in the EC group vs 23.8 ± 3.7 in the control group; P < .001). Hosono et al presented data analyzing the change in a woman’s BMI from the age of 20 until the age of diagnosis of EC, then looked at this in premenopausal women. La Vecchia et al presented only RR stratified by quintiles that had been calculated based on various anthropometric variables, and did not provide an overall analysis according to BMI for premenopausal women.
The pooled analysis by BMI category is presented in Figure 3 . A dose–response relationship of BMI and increased risk of EC was present. For the 2 studies that reported on women with BMI ≥25, the pooled OR was 3.85 (95% confidence intervals [CI], 2.53–5.84). For the 3 studies that reported on women with BMI ≥30, the pooled OR was 5.25 (95% CI, 4.00– 6.90). For the single study that reported on women with BMI ≥40, the pooled OR was 19.79 (95% CI, 11.18–35.03).