To the Editor As noted by Dr Barnard and colleagues, existing evidence confirms that endometriosis is associated with an elevated risk of ovarian cancer with differences by histotype. However, despite the varying presentations of endometriosis—most notably 3 macrosurgically visualized subphenotypes (endometriomas, superficial peritoneal, and/or deep lesions)—identifying women with endometriosis who are at high risk remains elusive. To address this question, the investigators leveraged data from the Utah Population Database and defined endometriosis subphenotypes by International Classification of Diseases (ICD) codes documented from 1992 to 2019 in more than 450 000 women. A consistent challenge in registry-based studies has been the selection of a comparison group. As demonstrated by Hermens et al in the Dutch nationwide registry, women diagnosed with ovarian cancer may be concurrently diagnosed with endometriosis, biasing estimated risk compared with members of the general population, who typically have not had pelvic surgery and whose endometriosis is more likely to remain undetected, leading to an observed 29 (95% CI, 20.7-40.9) times greater risk of endometrioid ovarian cancer (incident rate per 100 000 person-years, 29.7) when they included concurrent endometriosis diagnosis, but a much lower 2.6 (95% CI, 1.5-4.5)–fold greater risk when they excluded endometriosis diagnosed within 1 year of a diagnosis of ovarian cancer (incident rate per 100 000 person-years, 4.1). Unfortunately, Barnard and colleagues did not exclude concurrent diagnoses of ovarian cancer and endometriosis (17% of endometriosis and ovarian cancer diagnoses were within <1 year). Failure to exclude these concurrent diagnoses exacerbates the impact of detection bias that appears to be present in the article’s Figure 1. Sensitivity analyses excluding concurrent diagnoses would improve the validity of the results. In addition, the ICD code–based assignment of endometriosis subphenotypes is challenging given changes in documentation across the nearly 3-decade study period, known inconsistencies in ICD subcategory utilization in clinical practice, and erroneous subtype identification (eg, 617.0/N80.0 is not superficial endometriosis but most commonly denotes adenomyosis, while deep endometriosis cannot be inferred from ICD 9/10 codes, as it is not defined by location but rather lesion depth extending >5 mm under the peritoneal surface). Furthermore, since deep lesions were combined with endometriomas, the reader is unable to distinguish the true ovarian cancer risk for women with endometriomas vs deep lesions. Ultimately, the results may encourage clinicians to advise women with endometriomas or deep endometriosis toward prophylactic bilateral oophorectomy based on an overestimation of the true risk. Such an approach disregards the growing evidence of harm caused by inducing early menopause unnecessarily, including elevated risk of cardiovascular disease, bone fracture, and Alzheimer disease. As this is an essential question for women’s health, a reanalysis of the data are needed to guide women and clinicians in understanding a realistic estimation of their ovarian cancer risk and valid attribution to an accurately defined endometriosis subtype.
