Introduction
Approximately 600,000 hysterectomies are performed annually in the United States. Choosing the safest, most cost-effective approach is critical. Guidelines can help select appropriate candidates for vaginal hysterectomy (VH) and, when utilized, the vaginal route is successful for 90% of patients. However, these guidelines are not intended to apply to women with gynecologic cancers and have not been validated in these populations. In early-stage endometrial cancer, VH has been proposed for selected women at high surgical risk, as comorbidities such as obesity, heart disease, hypertension, diabetes, and advanced age may increase the risks associated with laparoscopy or laparotomy. While benefits to vaginal surgery include decreased operating time and postoperative complications, disadvantages include the inability to perform a thorough abdominal survey and pelvic lymphadenectomy, part of the recommended management of certain gynecologic cancers.
In January 2015, United Healthcare, a large US health insurance carrier, announced that all non-VH will require preauthorization prior to approval. This decision was based on a 2009 American Congress of Obstetricians and Gynecologists committee opinion that states that “vaginal hysterectomy is associated with better outcomes and fewer complications than laparoscopic or abdominal hysterectomy.” In the United States, 9.2% of all hysterectomies are performed for malignancy and another 2.7% are performed for endometrial hyperplasia. For most cases of gynecologic malignancy, either laparoscopic or abdominal approaches are recommended. In some precancerous conditions or very early stage malignancy, VH may be appropriate. While VH should be considered for benign indications, mandated preauthorization to pursue nonvaginal surgical approaches may delay time from diagnosis to treatment if applied to patients with a potential gynecologic malignancy. In cancer patients, a delay from diagnosis to definitive surgical treatment has been shown to adversely affect cancer outcomes. This potential delay and its impact on outcomes has prompted concern that insurance trends could carry over to the patient with a gynecologic cancer and has raised questions regarding mandated preauthorization when selecting surgical approach to hysterectomy without regard for whether the surgical indication is a malignant or premalignant condition.
The goal of this study was to determine the surgical trends over time for the route of hysterectomy in women with gynecologic cancers in Washington State from 2004 through 2013 and to determine if postoperative outcomes differed by surgical approach by comparing outcomes for women who underwent abdominal hysterectomy (AH), laparoscopic hysterectomy (LH), or total VH. We hypothesized that, among gynecologic oncology patients, postoperative complications and length of hospital stay (LOS) would differ by surgical approach, and that the extent of the advantages seen in utilizing the vaginal approach in benign cases may not apply to women with gynecologic cancer.
Materials and Methods
We performed a retrospective cohort study of all female patients age ≥18 years who underwent a hysterectomy for endometrial, ovary/fallopian tube, or cervical cancer in Washington State from 2004 through 2013. The study was approved by the Washington State Institutional Review Board (D-121712-H). We utilized the Comprehensive Hospital Abstract Reporting System (CHARS), which includes all non-federal hospital discharge data in Washington State, to identify women admitted with gynecologic cancer. The International Classification of Diseases, Ninth Revision ( ICD-9 ) codes are captured in CHARS with up to 25 diagnostic and 9 procedure codes per admission. These codes were searched to identify women admitted to 1 of 62 Washington State hospitals with a diagnosis of gynecologic cancer and to classify surgical approach to hysterectomy as AH (including total, supracervical, and radical AH), LH (which included robotic-assisted hysterectomy [RAH] since RAH was not assigned a separate ICD-9 procedural code from laparoscopic surgery until 2008), and VH ( Supplemental Table 1 ). We excluded patients with a record of admission that did not include these procedures. We excluded women undergoing pelvic exenteration.
We ascertained demographic information, medical comorbidities, surgical complexity, and perioperative complications including readmissions within 30 days from discharge from the index surgical admission. Patient demographic data included age, type of insurance (government [Medicare/Medicaid/TRICARE], private/health maintenance organization, or other [self-pay/charity care/liability and industry]), and obesity status. Patients were classified as obese if they were assigned an ICD-9 code for obesity not otherwise specified, morbid obesity, or an obese body mass index (BMI) ≥30 kg/m 2 . Obesity ICD-9 coding was modified during the time period of this study; for all years, codes for obesity not otherwise specified and morbid obesity were available, however codes indicating patient’s BMI group were modified in 2010 with a change from the initial highest category of BMI ≥40 kg/m 2 to increased specificity with codes for morbid and super obese BMIs of 40-44.99, 45-59.99, and ≥60 mg/m 2 . Because this level of detail was not available for all study years we used the pre-2010 categorization of BMI. Number and type of medical comorbidities were obtained from the index surgical hospitalization to calculate a baseline Charlson Comorbidity Index (CCI) score. Subjects received 1 point for each severe medical condition ( Supplemental Table 1 ) and points were summed to determine the overall CCI score. We modified the CCI score to exclude gynecologic cancer. Other important comorbid conditions that impact risk of surgical complications, but are not captured by the CCI score, such as hypertension, tobacco use, obstructive sleep apnea, asthma, and atherosclerotic vascular disease were ascertained from the index surgical hospitalization. Medical comorbidities used in calculating the CCI score and additional comorbid conditions with accompanying ICD-9 codes are listed in Supplemental Table 1 . Surgical complexity was determined by whether or not a lymph node sampling or dissection (LND) was performed with hysterectomy.
The outcome variables assessed were surgical trends over time, LOS measured in days, hospital readmission within 30 days of discharge from the index surgical admission, and perioperative complications overall as well as the subgroups of major and minor complications. Readmissions were assessed by linking up to 3 additional records of hospitalization using patients’ Social Security number in CHARS within 30 days after discharge from the initial surgical hospitalization. The overall complication rate was determined by summing all major and minor perioperative complications. Major and minor complications, as defined by the precise terminology in CHARS, are listed in Supplemental Table 2 .
We compared patient demographic and comorbidity characteristics by surgical approach using χ 2 testing. Surgical trends over time were assessed using linear regression to compare the mean number of cases performed by each surgical approach per year. We assessed the association between surgical approach and LOS using multivariable linear regression to estimate change in mean number of days and 95% confidence intervals (CI). The association between surgical approach and 30-day readmission was assessed using multivariable logistic regression to estimate odds ratios (OR) and 95% CIs with AH used as the referent group.
The association between surgical approach and overall perioperative complications, as well as the subgroups of major and minor complications, were assessed using multivariable logistic regression to estimate OR and 95% CIs with AH used as the referent group.
All analyses were adjusted for the a priori confounders of year of surgery, surgical complexity (LND), the patient’s preoperative medical comorbidity as measured by continuous CCI score (0-5), and cancer type (endometrial, cervical, or ovary/fallopian tube). We evaluated additional potential confounders using a data-driven, stepwise approach. Covariates assessed as potential confounders included age, race, and obesity. Confounding was defined as a 10% change in the estimated effect with addition of a variable to the linear or logistic regression model. No variables assessed met this definition and our final multivariable model was adjusted only for the prespecified a priori confounders. Statistical significance was defined as P < .05. No adjustments were made for multiple testing.
Materials and Methods
We performed a retrospective cohort study of all female patients age ≥18 years who underwent a hysterectomy for endometrial, ovary/fallopian tube, or cervical cancer in Washington State from 2004 through 2013. The study was approved by the Washington State Institutional Review Board (D-121712-H). We utilized the Comprehensive Hospital Abstract Reporting System (CHARS), which includes all non-federal hospital discharge data in Washington State, to identify women admitted with gynecologic cancer. The International Classification of Diseases, Ninth Revision ( ICD-9 ) codes are captured in CHARS with up to 25 diagnostic and 9 procedure codes per admission. These codes were searched to identify women admitted to 1 of 62 Washington State hospitals with a diagnosis of gynecologic cancer and to classify surgical approach to hysterectomy as AH (including total, supracervical, and radical AH), LH (which included robotic-assisted hysterectomy [RAH] since RAH was not assigned a separate ICD-9 procedural code from laparoscopic surgery until 2008), and VH ( Supplemental Table 1 ). We excluded patients with a record of admission that did not include these procedures. We excluded women undergoing pelvic exenteration.
We ascertained demographic information, medical comorbidities, surgical complexity, and perioperative complications including readmissions within 30 days from discharge from the index surgical admission. Patient demographic data included age, type of insurance (government [Medicare/Medicaid/TRICARE], private/health maintenance organization, or other [self-pay/charity care/liability and industry]), and obesity status. Patients were classified as obese if they were assigned an ICD-9 code for obesity not otherwise specified, morbid obesity, or an obese body mass index (BMI) ≥30 kg/m 2 . Obesity ICD-9 coding was modified during the time period of this study; for all years, codes for obesity not otherwise specified and morbid obesity were available, however codes indicating patient’s BMI group were modified in 2010 with a change from the initial highest category of BMI ≥40 kg/m 2 to increased specificity with codes for morbid and super obese BMIs of 40-44.99, 45-59.99, and ≥60 mg/m 2 . Because this level of detail was not available for all study years we used the pre-2010 categorization of BMI. Number and type of medical comorbidities were obtained from the index surgical hospitalization to calculate a baseline Charlson Comorbidity Index (CCI) score. Subjects received 1 point for each severe medical condition ( Supplemental Table 1 ) and points were summed to determine the overall CCI score. We modified the CCI score to exclude gynecologic cancer. Other important comorbid conditions that impact risk of surgical complications, but are not captured by the CCI score, such as hypertension, tobacco use, obstructive sleep apnea, asthma, and atherosclerotic vascular disease were ascertained from the index surgical hospitalization. Medical comorbidities used in calculating the CCI score and additional comorbid conditions with accompanying ICD-9 codes are listed in Supplemental Table 1 . Surgical complexity was determined by whether or not a lymph node sampling or dissection (LND) was performed with hysterectomy.
The outcome variables assessed were surgical trends over time, LOS measured in days, hospital readmission within 30 days of discharge from the index surgical admission, and perioperative complications overall as well as the subgroups of major and minor complications. Readmissions were assessed by linking up to 3 additional records of hospitalization using patients’ Social Security number in CHARS within 30 days after discharge from the initial surgical hospitalization. The overall complication rate was determined by summing all major and minor perioperative complications. Major and minor complications, as defined by the precise terminology in CHARS, are listed in Supplemental Table 2 .
We compared patient demographic and comorbidity characteristics by surgical approach using χ 2 testing. Surgical trends over time were assessed using linear regression to compare the mean number of cases performed by each surgical approach per year. We assessed the association between surgical approach and LOS using multivariable linear regression to estimate change in mean number of days and 95% confidence intervals (CI). The association between surgical approach and 30-day readmission was assessed using multivariable logistic regression to estimate odds ratios (OR) and 95% CIs with AH used as the referent group.
The association between surgical approach and overall perioperative complications, as well as the subgroups of major and minor complications, were assessed using multivariable logistic regression to estimate OR and 95% CIs with AH used as the referent group.
All analyses were adjusted for the a priori confounders of year of surgery, surgical complexity (LND), the patient’s preoperative medical comorbidity as measured by continuous CCI score (0-5), and cancer type (endometrial, cervical, or ovary/fallopian tube). We evaluated additional potential confounders using a data-driven, stepwise approach. Covariates assessed as potential confounders included age, race, and obesity. Confounding was defined as a 10% change in the estimated effect with addition of a variable to the linear or logistic regression model. No variables assessed met this definition and our final multivariable model was adjusted only for the prespecified a priori confounders. Statistical significance was defined as P < .05. No adjustments were made for multiple testing.
Results
We identified 10,117 women who underwent hysterectomy for endometrial, cervical, or ovarian/fallopian tube cancer from 2004 through 2013 at 62 hospitals in Washington State. There were 346 (3.4%) VH, 2698 (26.7%) LH, and 7073 (69.9%) AH patients ( Table 1 ). Obesity varied by surgical approach with more obese women (30.7%) undergoing LH compared to VH (26.9%) or AH (17.1%) ( P < .001). CCI score ≥2 also varied by surgical approach with 11.4% of those undergoing AH having a CCI >2 compared to 8.1% for VH and 7.9% for LH ( P < .001). Patients undergoing AH (51.4%) were most likely to have the more surgically complex procedure of LND with hysterectomy, followed by LH (49.2%) and VH (6.7%) ( P < .001).
| VH N = 346 | LH N = 2698 | AH N = 7073 | ||||
|---|---|---|---|---|---|---|
| n | (%) | n | (%) | n | (%) | |
| Cancer diagnosis a | ||||||
| Endometrial | 276 | (79.8) | 2260 | (83.8) | 4224 | (59.7) |
| Ovary/fallopian tube | 15 | (4.3) | 219 | (8.1) | 2526 | (35.7) |
| Cervix | 55 | (15.9) | 219 | (8.1) | 323 | (4.6) |
| Age, y (range) | 60 | (19–98) | 61 | (22–96) | 61 | (19–97) |
| Race a | ||||||
| White | 110 | (31.8) | 1862 | (69.0) | 2205 | (31.2) |
| African American | 1 | (0.3) | 32 | (1.2) | 64 | (0.9) |
| Hispanic | 2 | (0.6) | 25 | (0.9) | 38 | (0.5) |
| Other | 4 | (1.1) | 140 | (5.2) | 186 | (2.6) |
| Unknown | 229 | (66.2) | 639 | (23.7) | 4580 | (64.8) |
| Insurer a , b | ||||||
| Government | 145 | (41.9) | 1059 | (39.3) | 2860 | (40.4) |
| Private | 197 | (56.9) | 1543 | (57.2) | 3903 | (55.2) |
| Other | 4 | (1.2) | 96 | (3.6) | 310 | (4.4) |
| Obese a | 93 | (26.9) | 828 | (30.7) | 1207 | (17.1) |
| BMI c | ||||||
| 30–34.9 | 1 | (3.3) | 71 | (12.8) | 42 | (10.8) |
| 35–39.9 | 1 | (3.3) | 94 | (16.9) | 51 | (13.1) |
| ≥40 d | 28 | (93.4) | 390 | (70.3) | 297 | (76.1) |
| Diabetes | 62 | (17.9) | 479 | (17.8) | 992 | (14.0) |
| History of tobacco use | 35 | (10.1) | 356 | (13.2) | 745 | (10.5) |
| Asthma or OSA a | 27 | (7.8) | 260 | (9.6) | 509 | (7.2) |
| ASCVD a | 66 | (19.1) | 662 | (24.5) | 1198 | (16.9) |
| CCI score a , e | ||||||
| 0–1 | 318 | (91.9) | 2485 | (92.1) | 6270 | (88.7) |
| ≥2 | 28 | (8.1) | 213 | (7.9) | 803 | (11.3) |
| LND a | 23 | (6.7) | 1328 | (49.2) | 3632 | (51.4) |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
