Kaplan-Meier 5-year survival curve for inadvertent splenectomy vs no splenectomy (P < .001).
Marcia L. McGory, David S. Zingmond, Evan Sekeris, Clifford Y. Ko. The Significance of Inadvertent Splenectomy During Colorectal Cancer Resection. Arch Surg. 2007;142(7):668–674. doi:10.1001/archsurg.142.7.668
To examine the frequency, predictors, and outcomes following inadvertent splenectomy during colorectal cancer resection.
Linkage of the California Cancer Registry and the California Patient Discharge Database from the Office of Statewide Health Planning and Development.
Californians undergoing colorectal cancer resection from 1995 through 2001. Inadvertent splenectomy was defined as splenectomy occurring during non-T4 or non–stage IV resection.
Main Outcome Measure
The rate of inadvertent splenectomy for the overall cohort and by tumor location (eg, splenic flexure, rectosigmoid). Multivariate risk-adjusted models identified predictors of inadvertent splenectomy and outcomes including length of stay and probability of death.
A total of 41 999 non-T4, non–stage IV colorectal cancer resections were studied. Mean age was 70.4 years; 50.4% were male; and 75.6% were non-Hispanic white. Although the overall rate of inadvertent splenectomy was less than 1%, the rate was 6% for splenic flexure tumors. A multivariate risk-adjusted model predicting inadvertent splenectomy demonstrated a statistically significant (P < .001) higher odds ratio if the tumor was located in the transverse (3.6), splenic flexure (29.2), descending (11.4), sigmoid (2.7), or rectosigmoid (2.6) regions. Using a risk-adjusted model, inadvertent splenectomy increased length of stay by 37.4% (P < .001). Perhaps most important, risk-adjusted survival analysis showed splenectomy increased the probability of death by 40% (P < .001).
To our knowledge, this is the first large study evaluating the rates and outcomes after inadvertent splenectomy. In the population-based cohort, tumor locations from the transverse colon to the rectosigmoid significantly increased the odds of inadvertent splenectomy. In addition, inadvertent splenectomy during colorectal cancer resection increased both length of stay and probability of death.
Colorectal cancer is the third most common cancer in the United States in both men and women and the third leading cause of cancer-related deaths.1 At present, the quality of care for colorectal cancer is a paramount issue in the United States. For example, the American Society of Clinical Oncology initiated the National Initiative on Cancer Care Quality in 2000 to develop potential quality measures for breast and colorectal cancer.2 In addition, the National Cancer Institute funded the Cancer Care Outcomes Research and Surveillance Consortium to evaluate the care of newly diagnosed patients with lung or colorectal cancer in a variety of health care delivery systems.3 As such, we are attempting to identify structural, process, and outcome measures that will be associated with providing high-quality surgical care for colorectal cancer.4
There are many known risks of performing colorectal cancer resection, including injury to adjacent structures. In this regard, the rate and outcomes of inadvertent splenectomy during colorectal cancer resection remain unknown. There is a paucity of literature on this topic, which is likely because of the low incidence of inadvertent splenectomy as well as publication bias since studies describing adverse events are less likely to be published in the literature. In addition, the studies published on outcomes of splenectomy during colorectal cancer resection are all case series from single institutions.5- 8 The current analysis examines a population-based data set to determine the frequency, predictors, and outcomes of inadvertent splenectomy during colorectal cancer resection.
The cohort of patients used in this analysis was identified through linkage of the California Cancer Registry (CCR) and the California Patient Discharge Database from the Office of Statewide Health Planning and Development. The CCR collects patient records from “any hospital or other facility providing therapy to cancer patients within an area designated as a cancer reporting area” (California Health and Safety Code9 enacted in 1985, mandating the establishment of a statewide system of cancer reporting). Since 1988, every cancer (with the exception of basal and squamous cell carcinoma) diagnosed in California is required by law to be reported to the CCR. Approximately 130 000 new cancer cases and 52 000 cancer deaths are reported to the CCR annually, and the entire state of California is now part of the national Surveillance, Epidemiology, and End Results database.10
The California Patient Discharge Database from the Office of Statewide Health Planning and Development collects annual data from all inpatients discharged from nonfederal hospitals licensed by the state of California. A discharge abstract is reported for each inpatient hospitalization and includes the following information: demographics (eg, age, sex, and race/ethnicity), admission and discharge dates, principal and other diagnoses and conditions present at admission (up to 24 per hospitalization), principal and other procedures and dates (up to 20 per hospitalization), and disposition of patient.11 All diagnoses and procedures are coded using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM).12
From January 1, 1995, through December 31, 2001 (with follow-up until December 31, 2002), all patients undergoing colorectal cancer resection were identified in the CCR. Inadvertent splenectomy was defined as a splenectomy occurring during resection of a non-T4 and non–stage IV colorectal cancer. T4 colorectal cancers were excluded from the analysis since it was not possible to determine from the database whether the splenectomy was inadvertent or intentional because of tumor invasion requiring en bloc resection. Stage IV tumors were excluded in the event a metastasis to the spleen was the indication for splenectomy. In addition, appendiceal carcinomas were also excluded from the analysis. The exclusion of these cases limited our sample to 41 999 colorectal cancer resections. The cohort of 241 patients undergoing inadvertent splenectomy was identified using the following ICD-9-CM codes: 41.5 for total splenectomy (98% of patients) and 41.43 for partial splenectomy (2% of patients). The sample sizes for each regression model vary because of missing data regarding the tumor location (<1% missing) for prediction of inadvertent splenectomy, type of insurance for prediction of LOS (<1% missing), and time to death for the survival analysis (<6% missing).
Clinical information included both cancer and noncancer variables. The cancer variables were tumor location (cecum [reference group] to rectum) and cancer stage (ie, American Joint Committee on Cancer, 5th edition, with stage 0 as the reference group). Noncancer clinical information included patient comorbidity, which was evaluated based on the Charlson Comorbidity Index.13 The revised Charlson Comorbidity Index variable used in our analysis excluded cancer-related comorbidities so as not to bias the parameter estimates of the multivariate regression models. The revised Charlson variable had 4 categories: 0 (reference group); 1; 2; and 3+, where 3+ included all revised Charlson scores of 3 or greater. Demographic information included age, sex, and race/ethnicity (non-Hispanic white [reference group], black, Hispanic, Asian, and other). Inadvertent splenectomy was coded as a dichotomous variable depending on whether a splenectomy procedure was performed during the hospitalization. Annual hospital volume was also included as a predictor variable with high-volume hospitals defined as the top 20% (reference group), middle volume as the middle 40%, and low volume as the lowest 40%, with respect to annual volume of colorectal cancer resections. Emergent admission was included as a dichotomous variable depending on whether the patient was admitted through the emergency department. The hypothesis was that splenectomy was performed because of accidental damage to the spleen during the colorectal cancer resection (eg, splenic flexure mobilization). In addition to inadvertent splenectomy, the outcomes of length of stay (LOS) and probability of death were also evaluated.
Descriptive statistics were performed on the 2 groups in the overall cohort: the patients who underwent colorectal cancer resection without splenectomy and those who underwent inadvertent splenectomy during colorectal cancer resection. Statistical comparisons between groups were performed using a 2-sample t test for continuous variables and a 2-sample test of proportions for categorical variables. The incidence of splenectomy overall and with regard to tumor location was also evaluated.
Three multivariate regressions were then performed. The first model used a multivariate logistic regression to predict inadvertent splenectomy. A logistic regression was used to determine the impact of various explanatory variables (eg, patient demographics, location of colorectal cancer) on the probability of splenectomy during colorectal cancer resections of non-T4, non–stage IV tumors.
The other 2 models focused on the outcomes of inadvertent splenectomy. Length of stay had a highly skewed distribution, and outliers with an LOS greater than 180 days were excluded from further analysis. A negative binomial regression was used to estimate the impact of inadvertent splenectomy on the LOS since the distribution remained skewed after removal of the outliers. As an initial comparison of survival, Kaplan-Meier survival curves were generated to compare the probability of 5-year survival between the no splenectomy vs inadvertent splenectomy groups. Furthermore, the log-rank test was used to compare the probability of survival between the 2 groups. A multivariate Cox proportional hazards model was used to evaluate the impact of inadvertent splenectomy on survival. Survival was measured in days to death or to the end of the follow-up period (December 31, 2002). All statistical analyses were performed with STATA version 8.2 (StataCorp, College Station, Texas). Statistical significance was defined as P ≤ .05.
A total of 41 999 colorectal cancer resections were identified in the CCR from January 1, 1995, through December 31, 2001. The mean age of the overall cohort was 70.4 years; 50.4% were male; and 75.6% were non-Hispanic white; 5.6%, black; 9.9%, Hispanic; and 8.4%, Asian. The majority (63.3%) of patients had a revised Charlson Comorbidity Index of 0. A comparison of the demographic and clinical characteristics for patients who underwent colorectal cancer resection without splenectomy and those who underwent inadvertent splenectomy during colorectal cancer resection appears in Table 1. There were significant differences between the patients undergoing inadvertent splenectomy with respect to increasing age, male sex, non-Hispanic white race, higher Charlson Comorbidity Index, and longer LOS (P < .01).
Although the overall rate of inadvertent splenectomy was less than 1% for the entire cohort, there were several tumor locations with a higher incidence of splenectomy. For example, inadvertent splenectomy occurred in 6% of splenic flexure tumors, 2% of descending colon tumors, and 1% of transverse colon tumors.
A multivariate risk-adjusted model predicting inadvertent splenectomy was performed on the cohort of patients with non-T4 and non–stage IV colorectal cancers. The variables of age, male sex, race/ethnicity, Charlson Comorbidity Index, tumor location, and admission through the emergency department were included as explanatory variables. Table 2 summarizes the results of the multivariate logistic regression model to predict inadvertent splenectomy. Four factors had a significant association with inadvertent splenectomy: age, male sex, Charlson Comorbidity Index, and tumor location. Tumor location was the most important factor affecting inadvertent splenectomy. For example, compared with tumors in the ascending colon, the odds of inadvertent splenectomy were 29.2-fold higher for splenic flexure tumors, 11.4-fold higher for tumors located in the descending colon, 3.6-fold higher for tumors located in the transverse colon, and more than 2-fold higher for tumors located in the sigmoid colon or rectosigmoid. These results support the descriptive findings with an increased number of splenectomies for colorectal cancer resections located in the splenic flexure, transverse, and descending colon locations. Age (odds ratio [OR], 1.02), male sex (OR, 1.46), and revised Charlson Comorbidity Index (OR, 1.55 and 1.87 for Charlson score of 2 or 3+, respectively) were also important predictors of inadvertent splenectomy.
The mean ± SD LOS for the cohort of colorectal cancer resections was 8.7 ± 6.5 days. The median LOS was slightly lower at 7 days because of some patients with an abnormally long LOS. The results of a negative binomial regression analyzing the factors affecting LOS are presented in Table 3. Other variables causing a significant percentage increase in LOS were black and other race, increasing Charlson Comorbidity Index, Medicaid and other insurance, middle- and low-volume hospitals, increasing colorectal cancer stage, and admission through the emergency department. The negative binomial regression model was used to predict LOS using a baseline model consisting of the following characteristics: mean age of the cohort (70.4 years), male sex, non-Hispanic white race, Charlson Comorbidity Index of 0, Medicare insurance, low hospital volume, colorectal cancer stage 0, and no admission through the emergency department. The predicted LOS using this baseline model was 6.5 days for the no splenectomy group vs 9.0 days for the inadvertent splenectomy group. Using the same baseline model to predict LOS for patients admitted through the emergency department increased LOS to 9.5 days and 13.0 days for the no splenectomy and inadvertent splenectomy groups, respectively.
A Kaplan-Meier 5-year survival curve for the no splenectomy and inadvertent splenectomy groups appears in the Figure. The inadvertent splenectomy group had significantly lower survival at 5 years (P<.001). The results of a Cox proportional hazards model to evaluate the impact of inadvertent splenectomy on patient survival are presented in Table 4. Patients undergoing inadvertent splenectomy during colorectal cancer resection had a 40% higher probability of death, while controlling for other covariates including patient age, sex, race/ethnicity, Charlson Comorbidity Index, hospital volume, colorectal cancer stage, and admission through the emergency department. In addition, as expected, colorectal cancer stage was also a significant predictor of survival. Patients with late-stage colorectal cancer (stage III) were significantly more likely to die (hazard ratio, 2.21) than patients with stage 0 colorectal cancer, while adjusting for the other covariates in the Cox proportional hazards model. Similarly, patients with increasing age and Charlson Comorbidity Index, middle- or low-volume hospital, and admission through the emergency department also had an increased probability of death with a hazard ratio greater than 1.
In this population-based analysis of colorectal cancers diagnosed in California during a 7-year period, we identified the frequency, predictors, and outcomes of inadvertent splenectomy during colorectal cancer resection. To identify a cohort of patients at risk for inadvertent splenectomy, we excluded all patients with T4 lesions as well as all those with stage IV colorectal cancers. We found that the overall rate of splenectomy for all colorectal cancers was less than 1%. However, the rate for inadvertent splenectomy for tumors adjacent to the spleen was 6%. Additionally, we found that there was a higher odds ratio in a multivariate risk-adjusted model for the performance of inadvertent splenectomy if the tumors were located in the transverse colon, descending colon, sigmoid colon, or rectosigmoid region. This finding is clinically believable, since for tumor resections in these locations, it is not uncommon to have to perform mobilization of the splenic flexure to perform an anastomosis with an appropriate level of tension. Splenic flexure mobilization may increase the risk of damage to the spleen and consequently increase the risk of inadvertent splenectomy. Potential mechanisms of intraoperative splenic injury include inadvertent traction injury, capsular tear due to adhesions, and, less commonly, direct injury or instrument trauma from a retractor.8,14 Inadvertent splenectomy has also been described for other intraperitoneal operations including gastrectomy, pancreatic resection, left radical nephrectomy, and abdominal vascular surgery.14- 19
In addition to identification of predictors of inadvertent splenectomy, we also evaluated the outcomes of LOS and probability of death. Performance of inadvertent splenectomy during colorectal cancer resection increased LOS by 37.4% (P < .001). Finally, and perhaps most importantly, we found that in a risk-adjusted survival analysis, the performance of an inadvertent splenectomy for non-T4 lesions and non–stage IV cancers increased the probability of death by approximately 40% (P < .001). These findings are supported by a review of the literature on iatrogenic splenic injury by Cassar and Munro,20 which showed that splenic injury caused increased operating time, blood loss, LOS, and rates of postoperative infection. Two of the studies included in the review suggest a relationship between splenectomy and increased mortality, with 3 deaths occurring in 8 patients (37% mortality) undergoing inadvertent splenectomy during surgical treatment of diverticular disease (vs 6% mortality in 342 patients without splenectomy)21 and 3 deaths occurring in 17 patients undergoing inadvertent splenectomy during abdominal vascular surgery (vs no deaths in matched controls).16
More specifically, a review of the literature on performance of inadvertent splenectomy and its effect on outcomes for colon cancer shows varied results. All studies were single-institution experiences with fewer than 70 patients in each cohort. Nonetheless, the effect of splenectomy on survival of patients with colorectal cancer was variable. In the most recent study, published in 1999 by Konstadoulakis et al,5 25 patients with colon cancer who had inadvertent splenectomy during resection of the descending colon or sigmoid lesions were studied. These patients were matched with control patients based on age, sex, Duke stage, grade, tumor location, date and type of operation, and number of blood transfusions. The patients with inadvertent splenectomy had significantly more infective postoperative complications (eg, wound infection, pneumonia, intra-abdominal abscess) than the control group. The incidence of metastases, 5-year disease-free survival, and overall survival were all lower in the splenectomy group; however, there was not a statistically significant difference between the splenectomy group and controls.5 Similarly, Varty et al6 evaluated 21 patients with a concurrent splenectomy and resection of colorectal cancer between 1970 and 1988. In comparison with a matched control group, significantly more patients in the splenectomy group experienced postoperative infective complications. The overall 5-year survival was lower in the splenectomy group (45% vs 59% in controls); however, this was not statistically significant.6 Finally, the largest prior series was a retrospective medical record review of 68 patients performed by Davis et al7 and published in 1988. There was no difference in survival at 5 years between patients undergoing splenectomy with stage B colorectal cancer (localized) and matched control subjects. However, in contrast to the prior 2 studies, there was a significantly lower survival rate for patients undergoing splenectomy with stage C (regional) colorectal cancer vs matched controls (19% vs 52%; P < .01).7 Although Konstadoulakis et al5 and Varty et al6 demonstrated lower survival rates in the splenectomy groups, it is possible these case series did not detect a statistically significant difference in survival between the patients undergoing splenectomy and matched controls because of a small sample size, which may contribute to a type II error (eg, failure to reject the null hypothesis that there is no difference in survival when an actual survival difference exists). In contrast, this population-based analysis provides a sample size of inadvertent splenectomy more than 3 times larger than any of the earlier-mentioned case series reports with which to evaluate survival in comparison with the cohort of patients who did not undergo splenectomy.
Splenectomy, in general, has become less common since the 1980s possibly because of awareness of the immunologic function of the spleen as well as the known risk, although rare, of overwhelming postsplenectomy infection (OPSI). Splenic preservation has been most extensively evaluated within the trauma literature, including operative attempts at splenic salvage (eg, splenorrhaphy), splenic artery embolization, and nonoperative management of splenic injury.22 This shift in management of traumatic splenic injuries has also likely affected the management of intraoperative injuries to the spleen with inadvertent splenectomy performed only if conservative measures (eg, splenorrhaphy, topical hemostatic agents) fail to control intraoperative bleeding. Inadvertent splenectomy during colorectal cancer resection remains a rare event as demonstrated by the occurrence in only 241 cases (<1% of colorectal cancer resections) during a 7-year period throughout the state of California. However, the results of this analysis clearly demonstrate a difference between the inadvertent splenectomy and no splenectomy groups through both prolonged LOS and increased probability of death.
For patients who do undergo an inadvertent splenectomy, the current standard of care is vaccination with polyvalent pneumococcal vaccine (with revaccination every 5 years), Haemophilusinfluenzae type b conjugate, and meningococcal polysaccharide vaccine within 2 weeks of splenectomy.23- 25 The goal of vaccination is prevention of OPSI caused by the encapsulated bacteria of Streptococcus pneumoniae, H influenzae, and Neisseria meningitidis. Although OPSI is uncommon, with an estimated lifetime risk of 5% for patients undergoing splenectomy, mortality is high. Despite these recognized guidelines and the risk of OPSI, the literature suggests that not all patients undergoing splenectomy are being appropriately immunized.23,26
As with any study, there are limitations to our investigation. First, we used an administrative database for inpatient hospitalizations linked to a cancer registry. Both databases are population based for the state of California, and it is possible that coding errors or inaccuracies could occur in either of the 2 databases. However, the outcomes of inadvertent splenectomy, LOS, and death are unlikely to have significant coding errors. Second, we performed risk adjustment using the Charlson Comorbidity Index, which is a risk-adjustment method using ICD-9-CM codes for comorbidities. Whether this is an appropriate or adequate risk-adjustment method is debated, although it does provide some quantification of comorbid disease burden within administrative databases. A third limitation of this study is the lack of clinical data. Although the postoperative infectious complications are theoretically coded in the California inpatient file, we did not evaluate these since the validity of diagnosis codes for detection of complications is less accurate and may also be subject to the problem of undercoding.27 In addition, many of the infectious problems may occur after hospital discharge, and thus, our reported frequencies might be an underestimation of the real occurrences of infectious morbidity.
There are, however, several advantages of this study. First, this study examines a population-based data source during a 7-year period, which provides a significantly larger sample size. Second, the outcome variables of performance of inadvertent splenectomy, LOS, and death that we evaluated are likely well coded in the inpatient hospitalization database. Major procedures occurring in the operating room have been shown to be reasonably well coded in administrative databases.28 In addition, LOS has been shown to have 98% agreement between administrative data and medical record review,29 while mortality outcomes rely on completion of a death certificate.
In summary, to our knowledge, this is the first large-scale study that evaluates the frequency, predictors, and outcomes after inadvertent splenectomy during colorectal cancer resection. In a population-based cohort, tumors located from the transverse colon to the rectosigmoid colon have significantly increased odds of inadvertent splenectomy. In addition, inadvertent splenectomy during colorectal cancer resection increased the LOS by 37.4% and the probability of death by 40%.
Correspondence: Marcia L. McGory, MD, Department of Surgery, David Geffen School of Medicine, University of California, Los Angeles, 10833 Le Conte Ave, 72-215 Center for Health Sciences, Box 956904, Los Angeles, CA 90095-6904 (email@example.com).
Accepted for Publication: April 19, 2006.
Author Contributions: Dr McGory had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: McGory, Zingmond, and Ko. Acquisition of data: Zingmond. Analysis and interpretation of data: McGory, Zingmond, Sekeris, and Ko. Drafting of the manuscript: McGory. Critical revision of the manuscript for important intellectual content: Zingmond, Sekeris, and Ko. Statistical analysis: McGory, Zingmond, and Sekeris. Obtained funding: Ko. Administrative, technical, and material support: McGory, Zingmond, and Ko. Study supervision: Zingmond and Ko.
Financial Disclosure: None reported.
Previous Presentation: This study was presented as a poster at the American Society of Colon and Rectal Surgeons Annual Meeting; April 2, 2005; Philadelphia, Pennsylvania.
Funding/Support: This work was supported by a Limited Project Grant from the American Society of Colon and Rectal Surgeons, grant 5U01CA086322-06 from the National Cancer Institute, and The Robert Wood Johnson Clinical Scholars Program at University of California, Los Angeles.