To determine the prevalence of occult uterine pathology in asymptomatic, morbidly obese women before and after bariatric surgery-induced weight loss.
To determine the prevalence of occult uterine pathology in asymptomatic, morbidly obese women before and after bariatric surgery-induced weight loss.
Prospective, blinded, non-interventional cohort.
Urban teaching hospital.
Morbidly obese women.
Endometrial biopsies were obtained at the time of Roux–en–Y gastric bypass and again 1 year later. Both the patient and the physician were blinded to the results of the initial biopsy until the conclusion of the study. Specimens were independently reviewed by two blinded pathologists.
Effect of bariatric surgery-induced weight loss on the prevalence of endometrial pathology at 1 year.
Fifty-nine women underwent an endometrial biopsy during bariatric surgery. The mean (range) age, weight, and body mass index (BMI) were 42 years (22–62 years), 127 kg (87–176 kg), and 46.8 kg/m2 (36–64.3 kg/m2), respectively. Four women had hyperplasia (three simple and one complex), for an overall prevalence of 6.8%. The prevalence among women not receiving some anti-estrogen therapy was 9.5%. Forty-six women (78%) underwent follow-up biopsy after a mean (range) weight loss of 42 kg (19–67 kg). Simple hyperplasia was identified in 3/46 women at the 1–year follow-up (6.5%). Two women had resolution of hyperplasia, two women had persistent, simple hyperplasia, and one had had a normal initial biopsy. No woman showed progressive pathology or cancer. At the end of the follow-up all but one patient had a documented resolution of endometrial pathology.
Asymptomatic morbidly obese women are at relatively high risk of harbouring occult endometrial hyperplasia. Bariatric surgery-associated weight loss reduced but did not eliminate this risk for endometrial pathology.
Increasing adiposity in women is associated with an increased conversion of adrenal androgens to estrogenic precursors, decreased sex hormone binding globulin, increased insulin-like growth factor, and decreased progesterone production as a result of impaired ovulatory function. The sum of these effects is a relative estrogen–progesterone imbalance, which has been causally implicated in the development of endometrial hyperplasia and adenocarcinoma, similar to that of estrogen-only hormone replacement therapies.
It is therefore not surprising, in light of the epidemic rates of obesity, that the incidence of endometrial cancer is rising in the USA, with over 47 000 new cases anticipated in 2012.[2-4] It is now clear that the risk of developing endometrial cancer rises nonlinearly with increasing obesity: patients 9–21 kg overweight are three-fold more likely to develop endometrial cancer, whereas women more than 23 kg overweight experience a ten-fold increase in risk.[5, 6] Despite the increased risk, neither screening nor chemoprophylaxis is currently recommended in obese women.
Endometrial hyperplasias represent premalignant states, with variable natural history and propensity for malignant transformation, based on the structural complexity and nuclear atypia. The clinical implications were first reported by Kurman et al., who identified an increasing risk of subsequent cancer diagnosis ranging from 1% in simple hyperplasia without atypia to 28% among patients with complex hyperplasia and atypia. Although the presence of atypia appears to impart the greatest risk, reclassification of hyperplasias using the more reproducible endometrial intraepithelial neoplasia (EIN) grading system showed a similar progression of risk (15% of simple hyperplasias were characterised as EIN, versus 27% for complex hyperplasia without atypia, 37% for simple hyperplasia with atypia, and 91% for complex hyperplasia with atypia), suggesting that these conditions represent a biological continuum, with progressively increasing malignant potential.
The non-surgical treatment of endometrial hyperplasia, and even some low-grade malignancies, is directed to restoring the estrogen–progesterone balance, typically through the addition of progestins. This strategy is often effective but is commonly associated with depression and secondary weight gain, which are particularly problematic in the obese population. Furthermore, endocrine therapies do little to relieve the cardiovascular, endocrine, and orthopaedic consequences of extreme obesity. Reducing endogenous estrogen levels though natural weight loss is ideal, but not easily achievable, especially among the extremely obese who are successful with diet and exercise alone less than 5% of the time. By comparison, gastric bypass surgery allows for reliable, rapid weight loss, and has been associated with the correction of endocrinopathies, including resumption of ovulation, decreased insulin resistance, and modulation of adopokine production. Interestingly, these corrections are frequently seen in advance of significant weight loss, possibly as the result of acute changes in the production and/or metabolism of gut and adipose hormones.
Taken together, these observations suggest that morbidly obese women may be at high risk of harbouring occult endometrial pathology, and that bariatric surgery-induced weight loss may have therapeutic effects. Given recent trends suggesting that the relative mortality of endometrial cancer is rising, perhaps as a function of rising obesity, this study was undertaken to provide pilot data on the prevalence of endometrial hyperplasia and/or cancer among asymptomatic morbidly obese women, and to determine whether bariatric surgery-induced weight loss alone improves these conditions without necessitating additional therapies.
After obtaining approval from the institutional review board, we recruited women from the bariatric surgery clinic to undergo two endometrial biopsies: at the time of bariatric surgery and again 1 year later. All women undergoing Roux–en–Y gastric bypass surgery who were asymptomatic by self-report (no abnormal vaginal bleeding in the previous 6 months), who had a negative pregnancy test or prior sterilisation, and who gave consent were eligible. Patients were consented for both biopsies prior to their bariatric surgery. Enrollment was limited specifically to patients undergoing Roux–en–Y gastric bypass to control for the variable physiological effects of the different bariatric surgical approaches.
In order to isolate the effect of bariatric surgery-induced weight loss alone on endometrial pathology, neither specimen was evaluated microscopically until the second biopsy was obtained, or upon withdrawal from the study. It was acknowledged that an occult malignancy could go undetected for up to 1 year in the absence of symptoms; however, as screening is not currently standard, informed enrollment in the study was felt to pose no greater oncological risk than non-enrollment. Conversely, because the protocol did not allow the immediate evaluation of the specimen, it was deemed unethical to include patients with a clinical indication for uterine biopsy (abnormal vaginal bleeding). All patients received counselling regarding the symptoms of endometrial cancer, and were instructed to follow-up with either their primary gynaecologist or with the gynaecologist associated with the study if abnormal bleeding or other symptoms developed.
We excluded women with any menstrual bleeding irregularities in the previous 6 months, or women with a personal history of endometrial cancer or hyperplasia. Post-menopausal women using hormone replacement, or patients of any age using tamoxifen, were also excluded as these agents have been associated with an elevated risk of hyperplasia (tamoxifen) or potential misuse (hormone replacement that excludes a progestin component), placing patients potentially at a higher risk for having occult pathology. Conversely, patients using oral or intrauterine contraception containing progestin were eligible, as their risk was presumed, but not known, to be lower than the untreated population.
Specimens were formalin-fixed, embedded in paraffin, sectioned, and stained using a standard haematoxylin and eosin preparation immediately after acquisition, but were stored unanalysed until the follow-up biopsies were obtained or until consent was withdrawn. Completed specimens were evaluated by two independent pathologists with expertise in gynaecological pathology, who were blinded to the patients' histories, order of specimen acquisition, and the other's assessment. Pathologists were asked to categorise the tissues as: benign, simple hyperplasia (with or without atypia), complex hyperplasia (with or without atypia), or cancer.
Upon study completion or withdrawal, all patients received both verbal and written reports detailing the biopsy findings and, if appropriate, treatment recommendations. Patients were considered to have an abnormality if either pathologist identified hyperplasia or carcinoma. All patients with abnormality on their final biopsy (second, or first if they declined a second biopsy or withdrew from the study) were counselled to follow-up with their primary gynaecologist until resolution. Patients with simple hyperplasia without atypia on the initial biopsy, but with normal endometrium on the second biopsy, were counselled again about the symptoms of endometrial cancer, and were instructed to continue with routine follow-up. Patients were considered to have insufficient tissue if neither pathologist was able to render a diagnosis; in cases where only one pathologist was able to render a diagnosis no additional biopsies were attempted. Patients with insufficient tissue on the initial specimen were not biopsied again if the second specimen was adequate and without pathological abnormality. Patients with insufficient tissue on the final sample were to be re-sampled if their initial sample demonstrated any abnormality, and were offered elective re-sampling if their initial sample was normal.
A precise estimate of the prevalence of endometrial pathology in a morbidly obese population is unknown. To determine the sample size for this study, we assumed that the prevalence of endometrial anomalies among morbidly obese patients seeking bariatric surgery would not be less than the 15% reported by Viola et al. for a mixed population of women who were symptomatic and overweight or obese, and that the prevalence in a ‘corrected’ or non-obese population would be ≤1%, consistent with the general population. Setting the alpha error level at 0.05 and the beta error level at 0.20 (for a power of 0.80), we determined that at least 45 patients would have to complete both the initial and the follow-up biopsies to detect a significant decrease in the prevalence of pathology. Our institutional review approved the enrollment of up to 75 patients with the expectation that 25–40% of patients would decline follow-up, withdraw consent, or be otherwise ineligible 1 year after the initial biopsy. The study was closed to accrual when it appeared that 50 ‘patient end points’, i.e. completed sample pairs, could safely be achieved based on the compliance rate of the initial enrollees.
Statistical analyses included the chi-square test, the Student's t–test, and the Wilcoxen test, where appropriate, and a P value of <0.05 was considered to be statistically significant. Correlation between pathologists was analysed using an unweighted Cohen's κ test. The 95% confidence intervals on point estimates were determined by the Newcombe method.
We enrolled 59 women between April 2008 and November 2009. Patients were followed for at least 1 year and the final follow-up biopsy was obtained in October 2010. Data acquisition was terminated in April 2011 (18 months after the final patient enrolled but declined a second biopsy).
Patient demographics are listed in Table 1. The median age of patients was 39 years. Sixteen patients (27%) were diabetic at enrollment: five used diet control, ten used oral agents alone (seven of these using metformin), and one required both oral agents and insulin. Ten patients (17%) were postmenopausal at the time of surgery. Almost three-quarters had documented cervical cytology screening in the previous 2 years, suggesting generally good access to health care. Almost 30% of patients were on some form of hormonal therapy, of which all but one were using either oral contraceptives or a progestin-containing intrauterine device (IUD).
|Age (years)||39 (20–60)|
|Preoperative weight (kg)||125 (87–195)|
|Preoperative BMI (kg/m2)||46 (36–64)|
|Cervical cancer screening within last 2 years|
|Use of hormonal therapy|
|Oral contraceptives||12 (20%)|
|Progestin (IUD or oral)||3 (5%)|
All patients completed laparoscopic Roux–en–Y gastric bypass without overt complication, and no long-term surgical complications were identified at 1 year. The median (range) weight loss at 12 months was 41 kg (19–67 kg). The median BMI at follow-up was 29.6 kg/m2 (range 22–46kg/m2). Compliance with a follow-up biopsy was higher than initially anticipated, with 46 of 59 patients (78%) undergoing a second biopsy; however, the compliance in the second half of the study was slightly lower than in the first half, resulting in fewer than the targeted 50 follow-up biopsies, but did allow for the required 45 paired samples. Thirteen patients did not complete a follow-up biopsy for the following reasons: elective refusal despite initial consent (eight), moved geographically (four), pregnant at 1 year (one), but none of these had abnormalities on their initial biopsy.
There was strong agreement between pathologists with regard to diagnosis (κ values = 0.88 and 0.85, respectively, for the preoperative and postoperative biopsy groups). Five preoperative specimens (8.5%) were felt to contain insufficient tissue for diagnosis, but only one (2.2%) of the postoperative specimens was inadequate. This woman had had a normal preoperative biopsy and was not re-sampled.
The prevalence of any hyperplastic abnormality on preoperative biopsy was 6.8% (95% CI 2.7–16.1%), including three patients with simple hyperplasia without atypia and one patient with complex hyperplasia without atypia. No cases of endometrial carcinoma were detected. As expected, a great majority of asymptomatic patients were without pathological abnormality. None of the women on a progestin-based therapy had a documented abnormality. Among the 42 women who were not on hormonal therapy at the time of initial biopsy, the prevalence of abnormality was 9.5% (95% CI 3.8–22.1%).
Three cases of simple hyperplasia were identified among the 46 follow-up samples, yielding an overall prevalence of 6.5% (95% CI 2.2–17.5%; P value not significant). All four patients who had an abnormality on initial biopsy were biopsied again. Of these, two patients had resolution of their abnormalities, including the patient with complex hyperplasia. The other patient who experienced a resolution of simple hyperplasia had started oral contraceptives for birth control during the follow-up window. Two patients had persistent simple hyperplasia 1 year after initial biopsy. Both reported an absence of overt or subjective symptoms. One of these was treated with oral contraceptives and had a documented resolution of hyperplasia on office biopsy. The second patient refused therapeutic intervention, but was subsequently biopsied at 18 months after bariatric surgery with resolution of her pathology. One patient who had had a normal preoperative biopsy was found to have simple hyperplasia at follow-up. This patient was asymptomatic and has to date declined intervention.
Thus, at last follow-up four of the five patients with identified endometrial pathology had a resolution of their pathology without additional surgery, and one patient remained asymptomatic but without histological confirmation of resolution. Post-hoc analysis of the patients who were not on chemoprophylaxis, and for whom a final disposition was recorded (up to 18 months after enrollment), indicated that the rate of pathology fell from 9.5 to 0% (P < 0.05).
Although the affected sample size was small, there did not appear to be any gross differences between patients with endometrial pathology and those without endometrial pathology with regards to age, preoperative weight, BMI, or weight loss (P > 0.05). The average weight loss for the three patients with persistent or new hyperplasia on follow-up sampling was 38.4 kg versus 41.6 kg for patients with a normal follow-up biopsy (P = 0.63).
Our data indicate that asymptomatic, morbidly obese women have a low but relevant risk of harbouring occult endometrial pathology. The risk of hyperplasia appears to be largely mitigated by the use of progestin-based therapy, but approached one in ten for women not receiving some form of prophylaxis in our study. Although any future clinical impact from these results would be predicated on establishing more complete risk stratification in larger populations, and the efficacy of planned treatment, our results suggest that women with extreme obesity may benefit from increased screening for endometrial pathology or potentially from prophylactic therapy, even in the absence of symptoms.
Bariatric surgery-induced weight loss appears to reduce, but not eliminate, this risk of persistent endometrial pathology. Although we did not detect a statistically significant reduction in the ‘whole population’ prevalence of endometrial pathology at the specified 1–year follow-up, there are multiple potential methodological reasons for this. First, the observed incidence of endometrial pathology in the pretreatment group was significantly lower than anticipated (6.8 versus 15%), leaving the study underpowered to assess the smaller than anticipated differences. Similarly, the inclusion of patients that were effectively receiving some form of anti-estrogenic therapy would probably have blunted the potential beneficial effects of weight loss. In a post-hoc analysis of the ‘untreated’ subpopulation there did appear to be a significant reduction in the prevalence of pathology at the conclusion of follow-up.
To date, screening asymptomatic patients irrespective of risk has been discouraged in part because most patients will present with abnormal bleeding early in the course of their disease, when the prognosis is largely favourable, making the cost-effectiveness of screening difficult to demonstrate. Recent reports, however, have demonstrated an increasing relative mortality from endometrial cancer in recent decades that, if confirmed, may improve the cost-benefit ratio of screening in higher risk populations and an earlier introduction of definitive therapy.[16, 17] Our data, and that of Viola et al., suggest that asymptomatic morbidly obese women may meet this level of risk as a general population, and certainly represent a higher risk subpopulation.
Because weight loss has numerous coincidental health benefits, weight reduction is the ideal means of reversing the estrogen imbalance. However, it is notable that although in two patients bariatric surgery-induced weight loss was associated with a resolution of the hyperplastic findings, an equal number had persistent abnormalities 1 year after surgery, and one woman had a normal initial biopsy but was determined to have hyperplasia after losing weight. One possible explanation for these apparently discrepant observations is that the ‘treatment’ duration was insufficient to normalise the endometrial milieu. Kurman and Randall have reported that the median duration of therapy required to reverse complex atypical hyperplasia or low-grade adenocarcinoma using progestins was 9 months. Given that one of the patients with persistent abnormalities ultimately experienced resolution of her hyperplasia, despite refusing therapy, it is plausible that the 1–year time point of evaluation is premature to reliably detect the reversal of hyperplasia through weight loss alone. Alternative explanations for the apparent development of hyperplasia after weight loss include: persistent obesity despite relative loss; sampling error (hyperplasia missed on the initial biopsy); or pathological ‘overcall’. Of these, pathological ‘overcall’ seems least likely because of the high rate of correlation between the blinded pathologists.
Similarly, sampling error cannot be excluded as a source of error. The Pipelle method was chosen over curettage for this pilot study, with the intention of minimising surgical risks in an asymptomatic volunteer population; however, this may not be the optimal strategy for evaluating this population. The Pipelle method is known to sample less than 5% of the endometrium, and although most studies examining the accuracy of uterine sampling have demonstrated a high degree of correlation between the pathological findings at Pipelle biopsy and at hysterectomy, there appears to be increasing discrepancies among more pathologically abnormal specimens, which may be relevant in a study such as ours.[20, 21]
Finally, persistent absolute obesity despite overall weight reduction, i.e. an improvement but not a correction of the estrogen/progestin ratio, may also account for the apparent development of new hyperplasia in one patient following bariatric surgery. Consistent with this hypothesis is the observation that none of the patients with identified abnormalities progressed, and that one of the two patients with persistent hyperplasia spontaneously resolved with additional time and weight loss.
The prevalence of endometrial pathology appears to approach 10% in untreated, asymptomatic morbidly obese women. Our results suggest a potential role for selective screening in this population, and/or for progestin-based chemoprophylaxis. The determination of both long-term effectiveness and cost-effectiveness will be required prior to making any definitive recommendations.
Bariatric surgery appears effective in reducing but not eliminating the prevalence of endometrial pathology, but prolonged follow-up (>1 year) may be required to observe maximum benefit. The additional benefits of weight loss with regard to cardiovascular, orthopaedic, and endocrine health, when juxtaposed with the poorly tolerated side effect of progestin therapy, make bariatric surgery an attractive potential option for the treatment of morbid obesity. Future studies should focus on the higher risk subgroups identified in this study, and possibly include the evaluation of symptomatic women.
None of the authors has any financial or other disclosures with regard to the content of this article.
PA contributed to the project design, specimen acquisition, and data analysis. MK contributed to the project design, specimen acquisition, and data analysis. AT contributed to the data analysis and pathological review. CS contributed to the project design, recruitment, and data analysis.
This study was approved by the Park Nicollet Institute, which is the Internal Review Board for the Methodist Hospital, St. Louis Park, Minnesota; ref. no. 03636-07–C, dated 3 November 2009.
There was no external funding for this project. Time and effort were volunteered by the authors.
The Pipelle biopsy curettes were supplied by Cooper Surgical (Trumbull, CT) at no cost.