ORIGINAL CONTRIBUTION

Laparoscopic Colorectal Resections: A Simple Predictor Model and a Stratification Risk for Conversion to Open Surgery Carlos A. Vaccaro, M.D., Ph.D.1 • Gustavo L. Rossi, M.D.1 Guillermo Ojea Quintana, M.D.1 • Enrique R. Soriano M.D., M.S.C.2 Hernan Vaccarezza, M.D.1 • Fernando Rubinstein, M.D, M.P.H.3 1 Section of Colorectal Surgery, Service de General Surgery, Hospital Italiano de Buenos Aires, Buenos Aires, Argentina 2 Clinical Research Committee, Hospital Italiano of Buenos Aires, Buenos Aires, Argentina 3 Instituto de Efectividad Clinica y Sanitaria, Buenos Aires, Argentina

BACKGROUND:  The advantages associated with the laparoscopic approach are lost when conversion is required. Available predictive models have failed to show external validation. Body surface area is a recently described risk factor not included in these models. OBJECTIVE:  The aim of this study was to develop a clinical

rule including body surface area for predicting conversion in patients undergoing elective laparoscopic colorectal surgery. DESIGN:  This was a prospective cohort study. SETTING:  This study was conducted at a single large tertiary care institution. PATIENTS:  Nine hundred sixteen patients (mean age, 63.9; range, 14–91 years; 53.2% female) who underwent surgery between January 2004 and August 2011 were identified from a prospective database. MAIN OUTCOME MEASURES:  Conversion rate was

analyzed related to age, sex, obesity, disease location (colon vs rectum), type of disease (neoplastic vs nonneoplastic), history of previous surgery, and body surface area. A predictive model for conversion was Supplemental digital content is available for this article. Direct URL citations appear in the printed text, and links to the digital files are provided in the HTML and PDF versions of this article on the journal's Web site (www.dcrjournal.com) Financial Disclosures: None reported. Correspondence: Carlos A. Vaccaro, M.D., Ph.D., Section of Colorectal Surgery, Department of General Surgery, Hospital Italiano de Buenos Aires, Buenos Aires 1424, Argentina. E-mail: [email protected] Dis Colon Rectum 2014; 57: 869–874 DOI: 10.1097/DCR.0000000000000137 © The ASCRS 2014 Diseases of the Colon & Rectum Volume 57: 7 (2014)

developed with the use of logistic regression to identify independently associated variables, and a simple clinical prediction rule was derived. Internal validation of the model was performed by using bootstrapping. RESULTS:  The conversion rate was 9.9% (91/916). Rectal disease, large patient size, and male sex were independently associated with higher odds of conversion (OR, 2.28 95%CI, 1.47–3.46]), 1.88 [1.1–3.44], and 1.87 [1.04–3.24]). The prediction rule identified 3 risk groups: low risk (women and nonlarge males), average risk (large males with colon disease), and high risk (large males with rectal disease). Conversion rates among these groups were 5.7%, 11.3%, and 27.8% (p < 0.001). Compared with the low-risk group, ORs for average- and high-risk groups were 2.17 (1.30–3.62, p = 0.004) and 6.38 (3.57–11.4, p < 0.0001). LIMITATIONS:  The study was limited by the lack of external validation. CONCLUSION:  This predictive model, including body surface area, stratifies patients with different conversion risks and may help to inform patients, to select cases in the early learning curve, and to evaluate the standard of care. However, this prediction rule needs to be externally validated in other samples (see Video, Supplemental Digital Content 1, http://links.lww.com/DCR/A137). KEY WORDS:  Body mass index; Body surface area; Conversion; Laparoscopic colorectal surgery; Obesity; Operative difficulties; Predictive model; Short-term outcomes.

L

aparoscopic colorectal surgery is becoming the standard treatment for elective colorectal resection, based on its numerous advantages.1 However, the benefits are lost when conversion to open surgery is re869

870

quired. Furthermore, conversion has been associated with higher intraoperative blood loss and transfusion requirements, a higher postoperative morbidity, and lower overall and disease-free survival.2–7 Consequently, accurate prediction of this event improves case selection for the laparoscopic approach and better informs patients about the expected postoperative outcomes. Statistical models are increasingly being used to predict outcomes in surgery. Regarding the prediction of chance of conversion to open surgery in laparoscopic colorectal procedures, models were developed at the University of Toronto8 and at the Cleveland Clinic Foundation.9 Both models, based on specific patient- and disease-related factors, showed internal validation but failed to have external validation in a cohort of patients treated at the Mayo Clinic.10 Although this limitation might be explained by institution-dependent patient and clinical factors, an additional potential explanation is the omission of additional variables related to conversion in the available models. In this regard, we recently reported the value of the body surface area (BSA) as an independent variable associated with conversion.11 Therefore, the aim of this study was to develop a multivariate model, including BSA, to predict the risk of conversion to open surgery in a large cohort of patients undergoing elective laparoscopic colorectal surgery.

METHODS Patients and Data Collection

A prospectively maintained, practice-specific database was used to identify all patients who underwent laparoscopic colorectal surgery from January 2004 to August 2011 at the Hospital Italiano de Buenos Aires, Argentina. Outcomes were prospectively recorded in an institutional review board-approved database. Data were prospectively collected relating to patient demographics (age, sex, height/weight, and a history of abdominal surgery), procedure (operative time and type of laparoscopic approach), disease (type and location), and surgeon variables (seniority [staff vs resident] and learning curve [more or less than 50 cases]). Indications for surgery included both neoplastic (adenomas and adenocarcinomas) and nonneoplastic (diverticulitis, ulcerative colitis, Crohn’s disease) disease. Procedures

Details of preoperative management and surgical technique have been previously described.11 All the procedures were performed or supervised by 1 of the 5 staff colorectal surgeons of our unit. The surgeon was blinded to the BSA calculation at the time of the decision to convert to open surgery. Straight or hand-assisted laparoscopic approaches were selected based on surgeon’s preference and both implied intracorporeal colon mobilization and ligation of the

VACCARO ET AL: VARIABLE MODEL TO PREDICT CONVERSION

main vessels. Over the course of this study, 1 surgeon performed hand-assisted laparoscopic surgery and 4 surgeons performed conventional laparoscopy routinely based on their preferences. Study End Point and Risk Factors

The main outcome was conversion to open surgery, defined as lengthening of the incision more than that required for specimen retrieval. Causes for conversion included bleeding, adhesions, and/or technical difficulties. Patient-specific, procedure-specific, and surgeon-specific variables were considered. BSA was calculated by the Mosteller formula as follows: BSA (m2) = (height in centimeters × weight in kilograms/3600)½.12 Based on data from a previous study,11 patients were classified as “large” or “nonlarge” by using a BSA cutoff value of 1.8. Operations were grouped into colonic resection (right hemicolectomy, left hemicolectomy, and sigmoid colectomy) or rectal resection (high or low anterior resection and abdominoperineal resection). Statistical Analysis

Continuous variables were reported as median and interquartile ranges. The Student t test, the Mann-Whitney U test, or ANOVA were used for comparing means, when appropriate, according to the distribution of variables. Categorical variables were compared with the χ2 test. Multivariable analysis was performed by logistic regression. Variables significantly associated with conversion in the crude analysis were added one at a time in the multivariable model, and only those significant in that context were retained in the final model. The final variable selection was based on clinical relevance and statistical significance, and individual patient variables as well as surgeon- and surgery-related variables were tested. Calibration and discrimination of the model were evaluated with the Hosmer-Lemeshow goodness-of-fit test and the area under the receiver operating characteristic curve. The coefficients and ORs derived from the multivariable analysis were subsequently used as weights to derive a simple clinical prediction rule. All statistics were 2-tailed, and a p value < 0.05 was deemed significant. For the analysis, STATA/SE version 11.1 was used. Development of the Risk Model and Validation

Patient baseline data were used to generate the probabilities of conversion to open surgery and to populate the exploratory matrix risk model. Performance measures included discriminative ability, calibration, and overall accuracy of the model. Internal validation of the coefficients was performed by using bootstrapping, with a large number of replications (500–1000) to approximate the distribution of the central estimate and to obtain bias-corrected confidence intervals of the coefficients.

871

Diseases of the Colon & Rectum Volume 57: 7 (2014)

Derivation of the Clinical Prediction Rule After obtaining the final model, the estimated ORs were used as weights to derive a simple score based on the combination of the variables. Possible values ranged from 0 to 3, depending on the number of risk factors presented by each individual. The number of observed converted cases in each group were compared with those predicted based on the score and defined 3 clinical risk subgroups based on the total score and the probability of conversion, ranging from 3% to 25%.

RESULTS A total of 916 patients (mean age, 63.9 [range, 14–91] years; 53.2% female) were identified. The overall conversion rate was 9.9% (91 cases). Demographic characteristics are depicted in Table 1. Among these, sex, BSA, disease location, and type of laparoscopic approach were related to conversion. After adjusting by logistic regression, only a BSA of ≥1.8 (as opposed to BSA

Laparoscopic colorectal resections: a simple predictor model and a stratification risk for conversion to open surgery.

The advantages associated with the laparoscopic approach are lost when conversion is required. Available predictive models have failed to show externa...
344KB Sizes 0 Downloads 3 Views