Dowdy, Sean C. MD; Borah, Bijan J. PhD; Bakkum-Gamez, Jamie N. MD; Kumar, Sanjeev MD; Weaver, Amy L. MS; McGree, Michaela E. BS; Haas, Lindsey R. MPH; Cliby, William A. MD; Podratz, Karl C. MD, PhD
Patients with endometrial cancer are particularly vulnerable to adverse sequelae of surgery by virtue of their highly prevalent and complex comorbidities. Complications in patients with endometrial cancer are particularly frustrating given the lack of standardization of surgical approach (laparotomy compared with minimally invasive surgery) and procedure (lymphadenectomy and extent thereof) in practice in the United States today.1–4 In the absence of well-designed investigations defining the surgical intervention that is most beneficial to patients from an oncologic perspective, an honest assessment of the complications associated with these interventions is of critical relevance.
Large cancer registries have demonstrated that comorbidities correlate with poorer oncologic outcomes and reduced adherence to endometrial cancer treatment guidelines.5,6 However, few studies have explored the potential association between patient or process-of-care factors and surgical outcome.
The few that have done so have been handicapped by small cohorts, inconsistent and incomplete definitions of surgical complications, and assessment of a limited number of predictors. Furthermore, the effect of patient factors, process-of-care factors, and complications on cost have been only superficially studied. The objective of this investigation was to clarify these uncertainties to improve our ability to counsel patients about the potential risks of surgical treatment, to allow risk adjustment to achieve equitable reimbursement, and, most importantly, to identify controllable process-of-care factors to improve future care.
MATERIALS AND METHODS
Data from all patients surgically treated for primary endometrial cancer at Mayo Clinic in Rochester, Minnesota, between 1999 and 2008, were meticulously abstracted by a dedicated registered nurse for patient characteristics, tumor characteristics, and process-of-care variables. Patient records were periodically reviewed to ascertain information on complications, disease progression, and vital status. When information regarding survival or recurrences was not current or sufficiently detailed in medical records or tumor registries, death certificates, letters to patients or to their personal physicians, or telephone interviews were used to obtain the necessary information. All cases were the direct responsibility of a staff gynecologic oncologist and systematic practice changes were adopted only when agreed on by all staff. Quality assessment within the group was assessed on a periodic basis.7 On January 1, 2004, an algorithm for the operative management of endometrial cancer and selection of patients for lymphadenectomy was initiated based on pathologic criteria including low-grade histology, superficial myometrial invasion, and primary tumor diameter 2 cm or smaller.2,4,7 Consistent with these previous reports, approximately one third of patients were at low risk and did not require staging per protocol. Of note, minimally invasive surgery was not performed for endometrial cancer at our institution before 2006.
Complications occurring intraoperatively or within the first 30 days were categorized according to the Expanded Accordion Classification.8 The primary outcome measure in the analysis was incidence of complications grade 2 or higher.
Predictors considered in the analysis included those described in the American College of Surgeons National Surgical Quality Improvement Program platform, which contains more than 130 elements.9,10 Of the original predictors collected, 30 were chosen for analysis based on anticipated clinical relevance and prevalence greater than 3%.
Two multivariable logistic regression models were developed based on noncontrollable patient risk factors (counseling model) and on both patient and perioperative process-of-care risk factors (global model). Models predicting grade 2 or higher complications were fit ignoring preoperative hematocrit and estimated blood loss because these predictors are directly correlated with blood transfusion, a grade 2 complication. Additional models were fit to identify factors associated with complications of grade 3 or higher (preoperative hematocrit and estimated blood loss were included in this model). First, the factors each were evaluated univariately using the two-sample t test for continuous variables and the χ2 for categorical variables. Multivariable logistic models were developed using stepwise and backward variable selection methods considering the factors with P<.25 from the univariable analyses. Two-way interactions were evaluated together for the factors included in the final model. The number of factors considered was consistent with 10–15 outcome events per the total predictor degrees of freedom for the predictor variables. The c-statistic is reported as an overall measure of the model's predictive ability and is identical to the area under the receiver-operating characteristic curve.11 The value of c ranges from 0 to 1; a value of 1 indicates perfect prediction and a value of 0.5 denotes random prediction analogous to flipping a fair coin.
Cost data for the study patients were captured from Olmsted County Healthcare Expenditure and Utilization Database, a research database jointly funded by Mayo Clinic's Division of Health Care Policy and Research and the National Institute of Health–funded Rochester Epidemiology Project.12,13 Among other things, Olmsted County Healthcare Expenditure and Utilization Database captures all acute care medical costs (regardless of payer or plan) for every service and procedure received by patients examined at Mayo Clinic. This claims-based database contains information on medical resource use, associated charges, and estimated economic costs for patients receiving care at the Mayo Clinic and their associated inpatient facilities. Olmsted County Healthcare Expenditure and Utilization Database provides a standardized inflation-adjusted estimate of costs for each service or procedure provided locally in 2010 constant Medicare dollars. The algorithm adjusts for inflation and geographical wage differences to estimate the costs in nationally representative constant dollars. Although the services provided represent the clinical practice patterns of Mayo Clinic, the value of each unit of service has been adjusted to national norms by use of widely accepted valuation techniques. The multivariable analysis of 30-day all-cause cost was modeled using the generalized linear model framework.14 The modified Park test was used to establish the appropriate family of distribution for the generalized linear model framework; the test predicts the squared residuals from a tentative generalized linear model as the function of the logged predictions from the same model.14 The Pregibon link test was applied to assess the adequacy of the link function, which checks for the linearity of response variable (health care costs) on the scale of estimation.15 This test comprises performing an ordinary least squared of the cost (Y) on the predicted values (xβ) and their squares (xβ2) obtained from an initial model; if the model is linear on the scale of estimation as hypothesized, the estimated coefficient associated with squared term (xβ2) will be insignificantly different from zero.16
All calculated P values were two-sided and P<.05 was considered statistically significant. Univariable analyses and multivariable analyses of the primary outcome were performed using the SAS 9.2 software package. Stata SE 11 was used to complete the multivariable cost analyses. This investigation was approved by the Institutional Review Board of Mayo Foundation. In accordance with the Minnesota Statute for Use of Medical Information in Research, women who did not consent to the use of their medical records for research purposes were excluded.
Between 1999 and 2008, 1,415 patients underwent surgical treatment for primary endometrial carcinoma that included hysterectomy. Authorization from patients to review their records for research purposes and sufficient 30-day follow-up were available for 1,369 patients (97%). Patient, tumor, and process-of-care characteristics are shown in Table 1.
Table 1-a Summary of...Image Tools
Using the Expanded Accordion Classification, grade 2 or higher complications occurred in 398 patients (29%, 95% confidence interval 26.7–31.5); 95 (7%) experienced grade 3 or higher complications. Health care costs at 30 days were, on average, 72% higher for patients who experienced a complication compared with those who did not (mean $15,236 compared with $26,184; P<.01). As shown in Table 2, these costs increased significantly according to complication grade compared with those without any complication (P<.001 in all cases), with the greatest increase occurring with grade 3 or higher complications (those requiring an invasive intervention). The 30-day mortality rate was 0.7% (95% confidence interval 0.3–1.2).
Univariable analysis of factors evaluated for an association with grade 2 or higher complications is shown in Table 1. Two models predicting the occurrence of complications were created: a counseling model and a global model. The counseling model considered mainly patient factors to provide patients an estimation of operative morbidity before surgery. As shown in Table 3, five variables were significant in the multivariable counseling model using grade 2 or higher complications as the end point. When using grade 3 or higher complications as the end point, creatinine more than 1.5 mg/dL (odds ratio [OR] 2.9) was a strong predictor in a multivariable model that included this variable along with preoperative white blood cells and pulmonary dysfunction (P<.01; data not shown).
The intent of the global model was to identify process-of-care factors that, when controlled for by patient factors, could be modified for the purposes of quality improvement. Using grade 2 or higher complications as the end point, eight predictors were significant in multivariable modeling (Table 3). Preoperative creatinine more than 1.5 mg/dL (OR 3.0; P<.01), body mass index (BMI, calculated as weight (kg)/[height (m)]2) 50 or more (OR 2.7; P<.03), estimated blood loss (OR 1.6 per doubling; P<.001), and preoperative white blood cells (OR 1.8 per doubling; P<.03) were strong predictors of grade 3 or higher complications. Elevated creatinine was present in 11% of patients with severe complications. An existing diagnosis of diabetes had been made in 35% of the 55 patients with preoperative creatinine more than 1.5 mg/dL, and an additional 11% had preoperative hyperglycemia without a diagnosis of diabetes. However, hyperglycemia in the setting of undiagnosed diabetes was not significant in any multivariable model.
In both models surgical approach and use of para-aortic lymphadenectomy were powerful predictors of grade 2 or higher complications. Patients who underwent laparotomy were four times more likely to have a grade 2 or higher complication compared with patients who underwent minimally invasive surgery. A pelvic and para-aortic lymphadenectomy was associated with a doubling of the complication rate. Surgical complexity also was significant. Patients requiring additional major procedures related to the endometrial cancer diagnosis, such as splenectomy, bowel resection, or diaphragmatic resection (grade 4 surgical complexity), were 2.7-times more likely to experience a grade 2 or higher complication compared with those with hysterectomy and bilateral salpingo-oophorectomy only (grade 1 surgical complexity). The predicted probability of a grade 2 or higher complication from the model ranged from 0.01 to 0.97. Among the patients with a predicted probability of a grade 2 or higher complication of 10% or less, 10–20%, 20–30%, 30–40%, 40–50%, and more than 50%, the proportions of patients who actually experienced a grade 2 or higher complication were 9.3%, 13.6%, 22.8%, 34.1%, 42.8%, and 68.0%, respectively, demonstrating that our model performed well. Table 4 shows the effect that variations in patient characteristics and surgical approaches have on morbidity, using grade 2 or higher complications as the end point.
Given the strong association between surgical approach and morbidity, it was expected that corresponding costs also would be greater. In fact, multivariable analysis of modifiable risk factors (whether deep vein thrombosis [DVT] prophylaxis was provided, surgical approach and whether preoperative white blood cells were more than 11.1) in the entire cohort revealed that only surgical approach influenced 30-day cost: mean cost of laparotomy was $19,438, compared with $16,057 for laparoscopic or robotic, compared with $11,258 for the vaginal approach alone (P<.05). The highest costs were associated with conversions from minimally invasive approaches to laparotomy (mean $20,449, 95% CI $18,299–$22,598).
The cost of lymphadenectomy in patients with clinical stage I was evaluated by multivariable analysis, controlling for patient factors. Compared with patients at low risk who underwent hysterectomy alone, 30-day costs were 25% and 56% higher for patients who underwent hysterectomy and pelvic lymphadenectomy or hysterectomy plus pelvic and para-aortic lymphadenectomy, respectively (P<.01). Multivariable analysis of all significant predictors revealed that American Society of Anesthesiologists class more than 2, use of beta-blockers, a history of pulmonary dysfunction, grade 2 or higher complications, panniculectomy, estimated blood loss, and failure to use DVT prophylaxis were associated with higher 30-day costs (P<.05).
As gynecologic oncologists, our primary research efforts must no longer be limited to optimization of surgical technique, chemotherapy, and radiation therapy with the singular goal of improving cancer-related outcomes. Improving the value of care we provide, a function of oncologic outcomes, complications, and cost, must be the new “yardstick” by which we measure progress given ever-shrinking health care dollars in the setting of growing expenses. More importantly, quality improvement should be vigorously pursued because it is in the best interest of our patients.
This investigation has several important strengths. The relatively large size of our cohort permitted a valid analysis of many factors in a multivariable fashion. Predictors chosen for analysis were based on more than 130 in the National Surgical Quality Improvement Program database. Complications were reported according to the Expanded Accordion Classification, in which surgical complications are graded according to severity and level of intervention needed rather than simply comparing crude rates of “minor” and “major” complications. Our 30-day complication rate is qualified by the fact that 84% were at worst minor or moderate (grade 1 or 2). Nevertheless, this leaves room for improvement because even minor complications are associated with higher costs. Our rate of grade 3 or higher complications is comparable with other investigations (note that definitions of complications differed between investigations).17,18
The counseling model is useful to provide patients with an estimate of their personal complication rate before surgery. The global model provides an overall assessment of those variables associated with morbidity. Most striking is the fourfold increase in complications for patients who underwent laparotomy compared with minimally invasive surgery despite preservation of surgical quality. Others have reported equivalent survival and reduced length of stay with minimally invasive surgery.17,19,20 Given the increase in complication rates and 30-day costs, institutions with heterogeneous surgical approaches (including our own in the past) will be hard-pressed to justify continued routine use of laparotomy for patients with endometrial cancer.
Minimally invasive surgery fell out of the model when grade 3 and higher complications were used as the end point, suggesting that this approach was mainly responsible for reducing mild and moderate grade complications, particularly surgical site infections. In contrast, BMI 50 or higher was a significant predictor of grade 3 and higher complications, with an adjusted OR of 2.7. It is noteworthy that obesity up to a BMI of 50 (less than 30, 30–39, and 40–49 were analyzed separately) was not associated with increased morbidity in any model and provides encouragement that reducing the degree of obesity may be as important as obtaining ideal body weight.
Para-aortic lymphadenectomy was associated with a doubling in the complication rate and highlights the need to objectively define which patients derive oncologic benefit from lymphadenectomy.21,22 In contrast, we were unable to associate pelvic lymphadenectomy alone with untoward events. However, morbidity assessment was limited to 30 days; therefore, this investigation was not designed to detect long-term quality-of-life effects. Rates of lower extremity lymphedema with extended follow-up will be reported separately. In an analysis limited to clinical stage I patients to minimize bias, both pelvic and para-aortic lymphadenectomy resulted in significant cost increases; therefore, we feel strongly that unnecessary lymphadenectomy should be omitted.2
Most disappointing was the absence of modifiable process-of-care factors that could be potentially altered to improve outcomes in the global model. It appears that patient characteristics, such as American Society of Anesthesiologists class, BMI, or pulmonary dysfunction, and uncontrollable factors, such as operative complexity, dominated the model and obscured potential benefits of controllable factors, such as normothermia or mean arterial blood pressure. A strong and unexpected predictor of severe complications was preoperative creatinine more than 1.5 mg/dL, present in 11% of patients with severe complications. Renal insufficiency may be a surrogate for long-term systemic effects of hyperglycemia that impair healing or otherwise increase the risk of severe complications. Failure to use DVT prophylaxis also was predictive of cost. The use and method of DVT prophylaxis was variable in our practice in 1999, but today it is standardized across surgeons. This is one example of how quality assessment changes practice and improves patient care. Regarding the association of beta-blockers with higher cost, this is likely to be confounded by the high correlation between beta-blocker use and other high-risk characteristics.
A limitation of this investigation is that a relatively small proportion of patients underwent para-aortic lymphadenectomy via a minimally invasive approach. The extended laparotomy incision required for para-aortic lymphadenectomy would be expected to increase morbidity in and of itself. Therefore, it remains to be determined whether these models are valid when laparoscopic or robotic approaches are used exclusively. For example, we reported no transfusions or grade 3 complications in a prospective series of patients who underwent laparoscopic extraperitoneal para-aortic dissection.23
As reimbursement for surgical complications begins to be curtailed in 2012, trends of referring obese or Medicare patients to tertiary care centers may expand to those patients known to be at higher risk for even minor complications to reduce expenses. Proper risk adjustment therefore is paramount, as shown in Table 4. This report is an initial effort to define proper risk adjustment in patients with endometrial cancer. It is critical that our findings be tested in other high-volume tertiary care institutions that provide expert oncologic care to women with endometrial cancer–analyses of morbidity and cost cannot be performed without also considering oncologic outcomes. With the changing landscape of new surgical techniques and treatment algorithms, it will be necessary to continually reassess factors associated with morbidity for continued quality improvement.
1. Benedetti Panici P, Basile S, Maneschi F, Alberto Lissoni A, Signorelli M, Scambia G, et al.. Systematic pelvic lymphadenectomy vs. no lymphadenectomy in early-stage endometrial carcinoma: randomized clinical trial. J Natl Cancer Inst 2008;100:1707–16.
2. Dowdy SC, Borah BJ, Bakkum-Gamez JN, Weaver AL, McGree ME, Haas LR, et al.. Prospective assessment of survival, morbidity, and cost associated with lymphadenectomy in low-risk endometrial cancer. Gynecol Oncol 2012;127:5–10.
3. Kitchener H, Swart AM, Qian Q, Amos C, Parmar MK. Efficacy of systematic pelvic lymphadenectomy in endometrial cancer (MRC ASTEC trial): a randomised study. Lancet 2009;373:125–36.
4. Mariani A, Webb MJ, Keeney GL, Haddock MG, Calori G, Podratz KC. Low-risk corpus cancer: is lymphadenectomy or radiotherapy necessary? Am J Obstet Gynecol 2000;182:1506–19.
5. Boll D, Verhoeven RH, van der Aa MA, Lybeert ML, Coebergh JW, Janssen-Heijnen ML. Adherence to national guidelines for treatment and outcome of endometrial cancer stage I in relation to co-morbidity in southern Netherlands 1995-2008. Eur J Cancer 2011;47:1504–10.
6. Truong PT, Kader HA, Lacy B, Lesperance M, MacNeil MV, Berthelet E, et al.. The effects of age and comorbidity on treatment and outcomes in women with endometrial cancer. Am J Clin Oncol 2005;28:157–64.
7. Bakkum-Gamez JN, Mariani A, Dowdy SC, Weaver AL, McGree ME, Cliby WA, et al.. The impact of surgical guidelines and periodic quality assessment on the staging of endometrial cancer. Gynecol Oncol 2011;123:58–64.
8. Strasberg SM, Linehan DC, Hawkins WG. The accordion severity grading system of surgical complications. Ann Surg 2009;250:177–86.
9. Hall BL, Hamilton BH, Richards K, Bilimoria KY, Cohen ME, Ko CY. Does surgical quality improve in the American College of Surgeons National Surgical Quality Improvement Program: an evaluation of all participating hospitals. Ann Surg 2009;250:363–76.
10. Khuri SF, Daley J, Henderson WG. The comparative assessment and improvement of quality of surgical care in the Department of Veterans Affairs. Arch Surg 2002;137:20–7.
11. Harrell F. Regression modeling strategies: with applications to linear models, logistic regression, and survival analysis. New York (NY): Springer; 2001.
12. Dunlay SM, Shah ND, Shi Q, Morlan B, VanHouten H, Long KH, et al.. Lifetime costs of medical care after heart failure diagnosis. Circ Cardiovasc Qual Outcomes 2011;4:68–75.
13. Melton LJ III. History of the Rochester Epidemiology Project. Mayo Clin Proc 1996;71:266–74.
14. Manning WG, Basu A, Mullahy J. Generalized modeling approaches to risk adjustment of skewed outcomes data. J Health Econ 2005;24:465–88.
15. Pregibon D. Goodness of link tests for generalized linear models. Appl Stat 1980;29:15–24.
16. Basu A, Manning WG, Mullahy J. Comparing alternative models: log vs Cox proportional hazard? Health Econ 2004;13:749–65.
17. Janda M, Gebski V, Brand A, Hogg R, Jobling TW, Land R, et al.. Quality of life after total laparoscopic hysterectomy versus total abdominal hysterectomy for stage I endometrial cancer (LACE): a randomised trial. Lancet Oncol 2010;11:772–80.
18. Wright JD, Lewin SN, Deutsch I, Burke WM, Sun X, Herzog TJ. Effect of surgical volume on morbidity and mortality of abdominal hysterectomy for endometrial cancer. Obstet Gynecol 2011;117:1051–9.
19. Walker JL, Piedmonte MR, Spirtos NM, Eisenkop SM, Schlaerth JB, Mannel RS, et al.. Recurrence and survival after random assignment to laparoscopy versus laparotomy for comprehensive surgical staging of uterine cancer: Gynecologic Oncology Group LAP2 Study. J Clin Oncol 2012;30:695–700.
20. Walker JL, Piedmonte MR, Spirtos NM, Eisenkop SM, Schlaerth JB, Mannel RS, et al.. Laparoscopy compared with laparotomy for comprehensive surgical staging of uterine cancer: Gynecologic Oncology Group Study LAP2. J Clin Oncol 2009;27:5331–6.
21. Dowdy SC, Mariani A. Lymphadenectomy in endometrial cancer: when, not if. Lancet 2010;375:1138–40.
22. Todo Y, Kato H, Kaneuchi M, Watari H, Takeda M, Sakuragi N. Survival effect of para-aortic lymphadenectomy in endometrial cancer (SEPAL study): a retrospective cohort analysis. Lancet 2010;375:1165–72.
23. Dowdy SC, Aletti G, Cliby WA, Podratz KC, Mariani A. Extra-peritoneal laparoscopic para-aortic lymphadenectomy—a prospective cohort study of 293 patients with endometrial cancer. Gynecol Oncol 2008;111:418–24.
24. Aletti GD, Dowdy SC, Podratz KC, Cliby WA. Relationship among surgical complexity, short-term morbidity, and overall survival in primary surgery for advanced ovarian cancer. Am J Obstet Gynecol 2007;197:676 e1–7.
Figure. No available...Image Tools