Background
The lifetime risk of ovarian cancer is 1.9% among women with endometriosis compared with 1.3% among the general population. When an asymptomatic endometrioma is incidentally discovered on imaging, gynecologists must weigh the procedural complications and the potential for subsequent surgical menopause against future ovarian pathology or cancer.
Objective
We aimed to determine if performing unilateral salpingo-oophorectomy is a more cost-effective strategy for the prevention of death than surveillance for asymptomatic endometriomas.
Study Design
We created a cost-effectiveness model using TreeAge Pro (TreeAge Software Inc; Williamstown, MA) with a lifetime horizon. Our hypothetical cohort included premenopausal patients with 2 ovaries who did not desire fertility. Those diagnosed with asymptomatic endometrioma underwent either unilateral salpingo-oophorectomy or surveillance (ultrasound 6–12 weeks after diagnosis, then annually). Our primary effectiveness outcome was mortality, including death from ovarian cancer or surgery and all-cause mortality related to surgical menopause (± hormone replacement therapy) if the contralateral ovary is removed. We modeled the probabilities of surgical complications, occult malignancy, development of contralateral adnexal pathology, surgical menopause, use of hormone replacement therapy, and development of ovarian cancer. The costs included surgical procedures, complications, ultrasound surveillance, hormone therapy, and treatment of ovarian cancer, with information gathered from Medicare reimbursement data and published literature. Cost-effectiveness was determined using the incremental cost-effectiveness ratio of Δ costs / Δ deaths with a willingness-to-pay threshold of $11.6 million as the value of a statistical life. Multiple 1-way sensitivity analyses were performed to evaluate model robustness.
Results
Our model demonstrated that unilateral salpingo-oophorectomy is associated with improved outcomes compared with surveillance, with fewer deaths (0.28% vs 1.50%) and fewer cases of ovarian cancer (0.42% vs 2.96%). However, it costs more than sonographic surveillance at $6403.43 vs $5381.39 per case of incidental endometrioma. The incremental cost-effectiveness ratio showed that unilateral salpingo-oophorectomy costs $83,773.77 per death prevented and $40,237.80 per case of ovarian cancer prevented. As both values were well below the willingness-to-pay threshold, unilateral salpingo-oophorectomy is cost-effective and is the preferred strategy. If unilateral salpingo-oophorectomy were chosen over surveillance for premenopausal patients with incidental endometriomas, 1 diagnosis of ovarian cancer would be prevented in every 40 patients and 1 death averted in every 82 patients. We performed 1-way sensitivity analyses for all input variables and determined that there were no reasonable inputs that would alter our conclusions.
Conclusion
Unilateral salpingo-oophorectomy is cost-effective and is the preferred strategy compared with surveillance for the management of incidental endometrioma in a premenopausal patient not desiring fertility. It incurs fewer deaths and fewer cases of ovarian cancer with costs below the national willingness-to-pay thresholds.
Why was this study conducted?
Optimal management of asymptomatic endometriomas is unknown, with little evidence to guide clinical decision-making. For premenopausal women who do not desire future fertility, the risks of perioperative complications and surgical menopause must be weighed against the benefits of ovarian cancer risk reduction.
Key findings
This cost-effectiveness analysis showed that unilateral salpingo-oophorectomy was associated with fewer deaths (0.28% vs 1.50%) and fewer cases of ovarian cancer (0.42% vs 2.96%) than surveillance. Increased incremental costs were incurred by surgical intervention than sonographic surveillance ($6403.43 vs $5381.39 per case of incidental endometrioma), and these were below the national willingness-to-pay threshold for the value of a statistical life when paired with corresponding reduced mortality rates.
What does this add to what is known?
Unilateral salpingo-oophorectomy is more cost-effective than surveillance for managing incidental endometriomas in premenopausal patients not desiring fertility.
Introduction
The lifetime risk of ovarian cancer is approximately 1.9% (1 in 53) among women with endometriosis compared with 1.3% (1 in 77) for the general population. , In particular, women with ovarian endometriomas (OMAs) are more likely to develop ovarian cancer histologic subtypes of clear cell carcinoma (CCC), endometrioid adenocarcinoma (EA), and low-grade serous carcinoma. , An estimated 10% of reproductive-age women suffer from endometriosis, and approximately 17%–44% of those have OMAs. , Unfortunately, measurements of the true prevalence of endometriosis are limited by an average 10-year delay between symptom onset and diagnosis and the current requirement to undergo surgery to visualize and histologically confirm the diagnosis. ,
Endometriosis is thought to be a direct precursor lesion of CCC and EA, whereby the accumulation of somatic mutations and the ovarian microenvironment stimulate the development of histologic atypia and ultimately malignancy. In contrast, the association between endometriosis and low-grade serous ovarian carcinoma is poorly understood and may be explained in part by a diagnostic misclassification of endosalpingiosis or ectopic uterine tubal glandular epithelium. , Consequently, we will focus on the well-established correlation between ovarian endometriosis and the development of CCC and EA. Ovarian CCC and EA account for approximately 6% and 10%, respectively, of new cases of epithelial ovarian cancer in the United States. The age at diagnosis for these subtypes peaks during the fifth decade of life, and patients most often present with disease confined to the ovary (stage I). The 5-year cause-specific survival is 66% for CCC and 82% for EA, which are notably higher than that for the more common serous carcinoma.
Multiple strategies have been proposed to decrease the risk of ovarian cancer for women with endometriosis, including medication management with oral contraceptives, ultrasound surveillance of endometriomas, and surgical removal of adnexal structures. , , Specifically, ovarian excision through oophorectomy or salpingo-oophorectomy yields a protective effect against the later development of ovarian cancer, whereas removal of endometriotic cysts alone does not. ,
Conversely, bilateral oophorectomy before the natural age of menopause is associated with increased overall mortality, largely driven by cardiovascular disease. Although a strategy of unilateral oophorectomy or salpingo-oophorectomy means that 1 functional ovary remains, there is a risk of contralateral pathology resulting in subsequent removal of that ovary, causing surgical menopause. In cases of painful endometriomas or ovarian lesions in postmenopausal women, clinical judgment favors oophorectomy for symptom relief or to exclude malignancy. In women who desire future childbearing, retention of the affected ovaries is often recommended to maximize fertility. The optimal management of asymptomatic endometriomas in premenopausal women who do not desire future fertility is unknown, with little evidence to guide clinical decision-making.
The objective of this study was to determine if planned removal of the affected ovary and tube (unilateral salpingo-oophorectomy [USO]) is a more cost-effective strategy than surveillance for women with asymptomatic OMAs when fertility is no longer desired. Given the large sample size and extensive follow-up necessary in a prospective study to answer this clinical question, it is uniquely well-suited to decision analysis modeling techniques.
Materials and Methods
We created a decision analysis model using TreeAge Pro (TreeAge Software Inc; Williamstown, MA) with a lifetime horizon. Figure represents a simplified diagram of the decision tree; see Appendix for the full tree diagram. Our hypothetical cohort consisted of premenopausal patients with 2 ovaries diagnosed with an asymptomatic endometrioma who did not desire future fertility. We chose this specific scenario, as it is commonly encountered in patients diagnosed with an incidental endometrioma. We compared USO with surveillance, defined as undergoing pelvic ultrasound 6–12 weeks after diagnosis and then annually thereafter. The decision analysis tree was constructed on the basis of expert opinion with input from 6 fellowship-trained gynecologic surgeons specializing in minimally-invasive gynecologic surgery, reproductive medicine, and female pelvic medicine and reconstructive surgery. We incorporated the effects of the index treatment strategy, including surgical complications, diagnosis of occult malignancy, development of contralateral adnexal pathology, surgical menopause in the case of subsequent contralateral oophorectomy, use of hormone replacement therapy (HRT), and development of ovarian cancer. Three physician investigators reviewed primary literature to generate point estimates for each probability ( Table 1 ). , ,
| Variable | Probability | Source |
|---|---|---|
| Composite perioperative complications | 0.194 | Orlando et al, 2022 |
| Occult malignancy | 0.0015 | Nishida et al, 2000 Kuo et al, 2017 |
| Development of contralateral endometrioma after USO | 0.247 | Hidari et al, 2019 |
| Use of HRT after premature surgical menopause | 0.642 | Jang et al, 2020 |
| Ovarian cancer incidence among women with OMA and ovarian retention | 0.03328 | Saavalainen et al, 2018 SEER |
| Ovarian cancer incidence among women with OMA after USO | 0.02330 | Rice et al, 2014 Saavalainen et al, 2018 SEER |
| Surgical intervention among women with OMA undergoing surveillance | 0.127 | Froyman et al, 2019 |
| Death in the perioperative period | 0.0002 | Orlando et al, 2022 |
| Death following premature surgical menopause with HRT | 0.00404 | Parker et al, 2009 Parker et al, 2013 |
| Death following premature surgical menopause without HRT | 0.00569 | Parker et al, 2009 Parker et al, 2013 |
| Death if diagnosed with ovarian cancer | 0.503 | Heintz et al, 2006 |
We noted a gap in the literature related to the risk of perioperative complications at the time of USO for endometrioma. Our group performed a retrospective study of the Vizient clinical database to ensure that the model included reliable point estimates for perioperative complications. From 2010 to 2020, 8622 women aged 18–50 years underwent oophorectomy for endometriomas and experienced a 19.4% composite complication rate including conversion to laparotomy, blood transfusion, ileus, urinary tract injury, bowel injury, readmission, and death.
To determine the ovarian cancer incidence in a hypothetical cohort, we multiplied the general population risk of ovarian cancer (1.3%) by the standardized incidence ratio of 2.56 for women with ovarian endometriosis. The resulting 3.3% risk among women with endometriomas is notably higher than the 1.9% among those with any form of endometriosis. Undergoing unilateral oophorectomy was associated with a hazard ratio of 0.70 for ovarian cancer diagnosis in the Nurses’ Health Study and Nurses’ Health Study II, and we assumed a similar reduction in risk in our cohort.
There were limited published data related to outcomes of surveillance for patients with asymptomatic endometriomas. Two-year follow-ups from the International Ovarian Tumor Analysis Phase 5 study showed that 75 of 591 patients (12.7%) who enrolled in ultrasound surveillance for endometriomas underwent surgery during the study period. We used this estimate to represent the baseline risk of surgical intervention in women with endometriomas undergoing surveillance.
We estimated a 24.7% cumulative risk of contralateral endometrioma recurrence after unilateral oophorectomy, as noted in a 2019 retrospective study by Hidari et al. There were no published data to indicate the proportion of patients with contralateral recurrence who would choose to undergo a second oophorectomy before natural menopause. Therefore, we assumed that the proportion would be similar to those with a primary endometrioma undergoing surveillance (12.7%), because this represents individuals who, despite a desire to avoid surgery, ultimately receive surgical intervention. As this point estimate lacked data, we performed sensitivity analysis by varying the risk from 0%–100% to examine the impact on model conclusions. In both cases, factors such as enlarging lesion size or the development of concerning features would prompt sufficient concern to recommend surgery.
Our effectiveness outcomes included mortality and development of ovarian cancer. For mortality, we modeled death from ovarian cancer, surgery, and all-cause mortality related to surgical menopause with or without the use of HRT. A perioperative mortality of 0.02% was calculated from our previous analysis of the Vizient clinical database. The impact of surgical menopause on all-cause mortality was stratified by the use of HRT. In their 2009 analysis of the Nurses’ Health Study data, Parker and colleagues found that women who underwent hysterectomy with bilateral oophorectomy before the age of 50 and never used estrogen therapy had an all-cause death rate of 569 per 100,000 person-years. The authors subsequently showed that patients with premature surgical menopause who did not receive HRT had increased mortality (with a hazard ratio of 1.41) than those with ovarian conservation at the time of hysterectomy. The increased mortality rate was mitigated by the use of HRT in the study population. As we were unable to find data directly comparing patients with bilateral oophorectomy before age 50 who did and did not use HRT, we used the above hazard ratio of 1.41 to represent the increased mortality risk conferred by failure to initiate HRT. Death from ovarian cancer was calculated on the basis of an overall 5-year survival of 49.7% for 7314 patients with ovarian cancer of all stages and histologies described in the International Federation of Gynecology and Obstetrics’ (FIGO) 26th Annual Report on the Results of Treatment in Gynecological Cancer. Although endometriosis is associated with the ovarian cancer histologic subtypes of CCC, EA, and low-grade serous carcinoma, these subtypes are a small fraction of ovarian cancer overall. As our model included development of all ovarian cancer types with differential rates for those who underwent surveillance, we felt that using the FIGO Annual Report was a reasonable decision.
The costs for this model were gathered from Medicare reimbursement data and published literature. These included surgical procedures, complications, ultrasound surveillance, hormone therapy, and treatment of ovarian cancer ( Table 2 ). The cost of USO surgery was calculated from 892 hysterectomies performed at the University of Pittsburgh Medical Center in 2016, as reported in a previous cost-effectiveness analysis by Cadish et al. The average complication cost was estimated on the basis of the risk and cost of each surgical complication, including conversion to laparotomy, blood transfusion, ileus, bowel injury, and readmission in the perioperative period. The cost of HRT was estimated using one of the primary regimens recommended by the American College of Obstetricians and Gynecologists, that is, micronized estradiol-17b 1 mg daily. This assumed a mean age of 45 years at the time of premature surgical menopause and an average age of natural menopause of 51 for an average of 6 years of HRT. The cost of 1 year of treatment for ovarian cancer was calculated from the previous work of Urban et al and Cadish et al. For the surveillance arm, we modeled the pelvic ultrasound costs at 6–12 weeks after initial diagnosis and then annually thereafter for a total of 5 years. All the previously published costs were converted to 2021 US dollars. Historically, a discount rate of 3% has been applied to costs in analyses such as these. However, our lifetime horizon made this impractical because it was impossible to determine the number of years following the diagnosis of incidental endometrioma that different costs would be incurred, especially the development of contralateral pathology with possible removal of the second ovary and development of ovarian cancer. Although we could have applied the discount rate to some costs such as scheduled surveillance ultrasounds which have known timing, we felt that it was most appropriate to use a no discount rate instead of a nonuniversal discount rate, which would be inequitably applied to 1 strategy.
