Comparing Characteristics of and Postoperative Morbidity after Hysterectomy for Endometriosis versus other Benign Indications: A NSQIP Study

A total of 29 742 people underwent hysterectomy during the study period, of which 3596 (12.1%) were performed for endometriosis. The demographic characteristics of the study population are presented in Table 1 . Patients undergoing hysterectomy for endometriosis were significantly younger (36.7% vs 23.6% received hysterectomy before the age of 40 years) and more likely to be nulliparous (23.4% vs 19.8%) than those receiving hysterectomy for other indications. Patients were predominantly White (84.1% vs 69.7%; p <.001) and less likely to be Hispanic (10.5% vs 13.3%; p <.001). Patients with endometriosis were more likely to be current smokers (16.5% vs 15.1%; p <.029), but were less likely to have comorbid diabetes (5.0% vs 6.5%; p <.001) or hypertension (16.5% vs 21.5%; p <.001). They had slightly lower body mass index than patients with other benign disease (median 29.6 kg/m 2 vs median 30.5 kg/m 2 ; p <.001). Gynecologic characteristics of the study population are presented in Table 2 . Patients with endometriosis were more likely to undergo any minimally invasive route of surgery (87.6% vs 77.8%) than patients with other benign indications, with the majority (66.3% vs 50.8%) being laparoscopic. Patients with endometriosis were much less likely to undergo vaginal hysterectomy (5.4% vs 12.6%; p <.001). Overall rates of previous abdominal (27.4%) and pelvic surgery (60.4%) were high; however, they were higher for those with hysterectomy for endometriosis (31.2% vs 26.9% [p <.001] and 67.5% vs 59.5% [p <.001], respectively). A total of 2513 cases included various concurrent lysis of adhesion procedures. Those with endometriosis were more likely to undergo concurrent lysis of adhesions with fulguration or excision (9.4% vs 2.0%; p <.001) and ureterolysis (3.1% vs 0.9%; p <.001) at the time of hysterectomy. Of all patients in the study population, 18% (n = 5347) had endometriosis at the time of surgery designated by NSQIP rather than by ICD diagnosis. More patients receiving hysterectomy for endometriosis (ICD definition) had designated endometriosis by NSQIP than those with other benign disease (64.3% vs 11.6%; p <.001), although not all did (35.7% receiving hysterectomy for endometriosis (ICD definition) did not have endometriosis by NSQIP designation). Most of these patients also had the location of endometriosis noted (n = 5346). Those with endometriosis indication for hysterectomy were more likely to have bowel (9.1% vs 3.5%; p <.001), urinary tract (6.4% vs 3.6%; p <.001), and pelvic lesions (57.1% vs 44.7%; p <.001); however, they had similar amounts of genital tract (56.2% vs 57.0%; p = .558) or “other” endometriosis lesions (7.5% vs 7.1%; p = .612) as those with other benign indications. Primary outcomes are summarized in Table 3 . The overall incidence of major morbidity (3.4%) and mortality (n = 2, 0.0%) in the study population was rare. Major morbidity was more likely in patients undergoing hysterectomy for endometriosis (adjusted OR, 1.25; 95% CI, 1.02–1.54; p = .033). On analysis of secondary outcomes, deep SSI/ organ-space infection was also more likely in patients undergoing hysterectomy for endometriosis (adjusted OR, 1.42; 95% CI, 1.12–1.80; p = .024), whereas other morbidity outcomes were the same in both populations (with the exception of transfusion being more likely in patients with other benign indications [OR, 0.58; 95% CI, 0.45–0.75; p <.001]). Overall length of stay (73.1% vs 78.6% ≥1 day; p = .983) and operative time (median 118.0 minutes [Q1–Q3, 87.0–157.0] vs median 125.0 minutes [Q1–Q3, 93.0–169.0]; p <.001) were clinically similar in patients undergoing hysterectomy for endometriosis vs other benign disease. Outcomes by NSQIP designated location of endometriosis regardless of ICD indication for hysterectomy are shown in Fig. 1 and in Supplemental Table 2 . Length of stay was clinically similar for patients undergoing hysterectomy with endometriosis lesions of the bowel (82.1%; ≥1 day) vs those with lesions of the urinary tract (69.3%; ≥1 day) or genital tract, pelvic, or other lesions (75.1% ≥1 day; p =.202). Operating time was increased for patients with bowel lesions (median 163.0 minutes [Q1–Q3, 119.2–223.0]) vs urinary tract (median 129.0 minutes [Q1–Q3, 93.5–182.0]) or other lesions (median 120.0 minutes [Q1–Q3, 90.0–164.0]; p = .001).

This was a retrospective cohort study using American College of Surgeons NSQIP data from January 2018 to December 2019 [ 18 ]. Both the general and the procedure targeted (hysterectomy) participant use files were used. This database includes both inpatient and outpatient procedures. This study was deemed exempt by the University of Minnesota Institutional Review Board. Postoperative International Classification of Diseases (ICD), 10th Revision, codes were used to identify indications for hysterectomy ( Supplemental Table 1 ). All patients aged 18 to 55 years who underwent total hysterectomy for the following benign indications were included: endometriosis, abnormal uterine bleeding, myomas, pelvic organ prolapse, and pelvic pain. These indications were mutually exclusive given that only a single ICD diagnosis code is available, and other comorbid diagnoses are limited to prespecified variables. Hysterectomies for malignant conditions, emergent cases, and cases with preoperative sepsis or blood transfusion were excluded. Current Procedural Terminology (CPT) codes were used to identify the route of total hysterectomy and to exclude nontotal hysterectomy and procedures typically indicated in the presence of malignancy. Robotic hysterectomies are included in the laparoscopic hysterectomy group, given that information separating robotic laparoscopy from straight stick laparoscopy is not available in this database owing to CPT coding. CPT codes were used to identify concurrent adhesiolysis procedures. Characteristics from the NSQIP database included year of procedure, patient age, race, ethnicity, body mass index, parity, American Society of Anesthesiologists classification, comorbidities (diabetes, current smoker, hypertension, bleeding disorders, pelvic inflammatory disease), and previous abdominal or pelvic surgery. Endometriosis characteristics from the NSQIP hysterectomy targeted data file included presence of endometriosis and location of endometriosis (bowel, genital tract, pelvis, urinary, and other). This resulted in information related to the primary study definition of endometriosis as indicated for hysterectomy (by postoperative ICD coding), which was used to separate and compare populations, and a designation of presence of endometriosis from the NSQIP hysterectomy data file, which was analyzed as a characteristic of the entire study population. These different definitions of endometriosis will be referred to as ICD- vs NSQIP-based endometriosis in the manuscript. NSQIP data on endometriosis are collected in a similar fashion to other characteristic data, via certified surgical clinical reviewers who receive standardized training to collect data from medical chart review and direct contact with patients. In addition, audits of inter-rater reliability are conducted at participating sites. Primary outcomes included 30-day postoperative major morbidity, minor morbidity, or mortality as classified by the validated Clavien-Dindo scale with grade ≥3 complications defined as major and grade ≤2 complications defined as minor [ 19 ]. Minor complications included only superficial SSI, urinary tract infection, and transfusion requirement. Secondary outcomes included the individual components of major and minor morbidity, length of stay, length of operation, reoperation, and readmission. Demographics, patient/gynecologic characteristics, and primary/secondary outcomes were summarized for all patients and by hysterectomy indication using descriptive statistics. To investigate the association between continuous characteristics and hysterectomy indication, t tests or Wilcoxon rank-sum tests were used. Chi-square tests were used for categorical characteristics. To investigate the effect of hysterectomy indication on primary outcomes, unadjusted and adjusted logistic regression models were used. Covariates included in the adjusted models were age, race, ethnicity, and surgical route. Odds ratios (ORs), 95% confidence intervals (CIs), and p values were obtained. Unadjusted logistic regression models were used for binary secondary outcomes, and Wilcoxon rank-sum tests were used for continuous secondary outcomes due to skewness of these data. Similar analyses were performed to investigate the effect of endometriosis location. All reported p values are 2 sided, and a significance level of .05 was used. Secondary outcomes were adjusted for multiple comparisons using false discovery rate correction. Statistical analyses were performed using R (v.3.6.1; R Core Team, Vienna, Austria) and SAS (v.9.4; SAS Institute Inc., Cary, NC).

This represents one of the largest studies comparing severe postoperative morbidity (Clavien-Dindo grade ≥3) after hysterectomy for patients with endometriosis vs other benign disease. Consistent with previous findings, we identified that patients undergoing hysterectomy for endometriosis were at higher odds of major morbidity (3.8%; adjusted OR, 1.25), likely driven by increased risk of deep SSI/organ-space infection (2.3%; OR, 1.42), despite the use of minimally invasive approach in most cases (87.6%). The minority of patients with endometriosis lesions of the bowel (5.9%) were at particularly high risk of major morbidity (OR, 1.65) and deep SSI/organ-space infection (OR, 2.18) and sepsis (OR, 5.94). Despite these increased risks, no other differences were found including presence of any other type of infection, wound disruption, thromboembolism, single organ failure, reoperation, readmission, length of stay, or length of operation. Previous large cohort studies comparing complication rates by route of hysterectomy sometimes include risks in the setting of endometriosis. These data are conflicting on the significance of endometriosis and risk of morbidity and are likely outdated (most patients undergoing abdominal hysterectomy) [ 13 , 20 , 21 ]. Recent studies have evaluated perioperative outcomes for various minimally invasive procedures for endometriosis in patients with predominantly American Society for Reproductive Medicine stage III and IV disease [ 7 , 8 , 14 , 16 , 17 , 22 – 24 ]. Few studies evaluate definitive surgery with hysterectomy [ 15 ]. Although the definition of morbidity varied, major morbidity ranged from 1.5% [ 22 ] up to 9.3% to 11.8% [ 7 , 16 , 24 ] in patients with any concurrent rectal procedure or up to 16% for patients with ureteral nodules [ 8 ]. Most studies identified a major morbidity rate of 4.5% to 5.4%, slightly higher than in our population (3.8%) [ 7 , 14 , 17 , 23 ]. The largest prospective cohort study by Byrne et al [ 25 ] (n = 5162) of patients with deep rectovaginal endometriosis primarily focused on pain and quality of life measures postoperatively, but did identify a complication rate of 6.8%. These studies were performed with smaller numbers of patients (n = 80−568) in single-site high-referral centers, where procedures were often performed by high-volume, fellowship-trained surgeons. Our cohort more similarly represents the general population of patients undergoing hysterectomy with nonspecialized gynecologists, likely with less severe disease than those at referral centers. Although in our study morbidity was indeed greater for patients with endometriosis, there was no difference in other important outcomes such as readmission and, surprisingly, length of operation. Similarly, despite risk of deep SSI, risk of wound disruption or reoperation was not increased. Based on our findings, we agree with Vallée et al [ 17 ] that patients with endometriosis without involvement of the digestive or urinary tract generally have an overall low risk of morbidity and may undergo hysterectomy safely. Special attention should focus on identifying patients at risk of higher stage endometriosis preoperatively and monitoring for SSI within the first postoperative month. Furthermore, referral to high volume surgeons should be considered for patients at risk of higher stage endometriosis with digestive or urinary tract involvement. Strengths of this study include the large national data set used, increasing generalizability of results. The NSQIP data set also allows detection of infrequent significant morbidity outcomes that may not be captured by single institution studies. However, the large size of the population cohort and multiple comparisons does increase the risk of type I errors; false discovery rate correction was used to adjust for this risk. Postoperative outcomes within the NSQIP database are limited to 30 days and NSQIP participating institutions. In addition, information related to robotic surgery separate from laparoscopy and surgeon training as a minimally invasive gynecologist is not available. There are several other limitations of this data set, including its retrospective design, information limited to prespecified diagnosis and procedure coding, and a lack of insight into details of NSQIP coding of endometriosis-related variables. Discrepancies between ICD and NSQIP designated endometriosis could be caused by several factors: true presence of endometriosis in patients undergoing hysterectomy for other primary disease, endometriosis not previously diagnosed but present at the time of surgery, and inaccuracy of ICD coding, NSQIP coding, or both. Neither the ICD coding nor the NSQIP designation of endometriosis includes a pathology-proven diagnosis, and information on American Society for Reproductive Medicine endometriosis staging is not available. Specific details of how the definition of deep incisional or organ-space infection was met were not available. Improvements to ICD and CPT coding of endometriosis will improve larger population-based study of this disease. Future study may investigate risk factors for deep SSI and morbidity in patients undergoing definitive surgery for endometriosis and explore the relationship between location of endometriosis lesions or concurrent procedures and other complications such as organ injury and repair using a large data set. In conclusion, this retrospective study of morbidity after hysterectomy for benign disease using a large national data base in the United States identified patients with endometriosis to be at risk of major morbidity likely driven by deep SSI/organ-space infection compared with their peers. Those with endometriosis lesions of the bowel were at particularly high risk. Overall risk of morbidity was low in the general population, and other postoperative complications were the same between groups. These results confirm previous findings that patients with endometriosis may be at higher risk of postoperative complications; however, significant morbidity is rare, and hysterectomy among the general population regardless of indication is safe.