PAPER. Effect of Hospital Type and Volume on Lymph Node Evaluation for Gastric and Pancreatic Cancer

Transcription

1 PAPER Effect of Hospital Type and Volume on Lymph Node Evaluation for Gastric and Pancreatic Cancer Karl Y. Bilimoria, MD, MS; Mark S. Talamonti, MD; Jeffrey D. Wayne, MD; James S. Tomlinson, MD; Andrew K. Stewart, MA; David P. Winchester, MD; Clifford Y. Ko, MD, MS, MSHS; David J. Bentrem, MD Hypothesis: For gastric and pancreatic cancer, regional lymph node evaluation is important to accurately stage disease in a patient and may be associated with improved survival. We hypothesized that National Comprehensive Cancer Network (NCCN), National Cancer Institute (NCI) designated institutions, and highvolume hospitals examine more lymph nodes for gastric and pancreatic malignant neoplasms than do low-volume centers and community hospitals. Design: Volume-outcome study. Setting: Academic research. Patients: Using the National Cancer Data Base ( January 1, 2003, to December 31, 2004), patients were identified who underwent resection for gastric (n=3088) and pancreatic (n=1130 [pancreaticoduodenectomy only]) cancer. Main Outcome Measures: Multivariable logistic regression analysis was used to assess the effect of hospital type and volume on nodal evaluation ( 15 nodes). Results: Only 23.2% of patients with gastric cancer and 16.4% of patients with pancreatic cancer in the United States underwent evaluation of at least 15 lymph nodes. Patients undergoing surgery had more lymph nodes examined at NCCN-NCI hospitals than at community hospitals (median, 12 vs 6 for gastric cancer and 9 vs 6 for pancreatic cancer; P.001). Patients at highest-volume hospitals had more lymph nodes examined than patients at low-volume hospitals (median, 10 vs 6 for gastric cancer and 8 vs 6 for pancreatic cancer; P.001). On multivariable analysis, patients undergoing surgery at NCCN-NCI and high-volume hospitals were more likely to have at least 15 lymph nodes evaluated compared with patients undergoing surgery at community hospitals and low-volume centers (P.001 and P=.02, respectively). Conclusions: Nodal examination is important for staging, adjuvant therapy decision making, and clinical trial stratification. Moreover, differences in nodal evaluation may contribute to improved long-term outcomes at NCCN-NCI centers and high-volume hospitals for patients with gastric and pancreatic cancer. Arch Surg. 2008;143(7): Author Affiliations: Cancer Programs, American College of Surgeons (Drs Bilimoria, Winchester, and Ko and Mr Stewart) Chicago, Illinois; and Departments of Surgery, Feinberg School of Medicine, Northwestern University (Drs Bilimoria, Wayne, and Bentrem), and Evanston Northwestern Healthcare (Drs Talamonti and Winchester), Chicago, Illinois, and University of California, Los Angeles; and Veterans Affairs Greater Los Angeles Healthcare System (Drs Tomlinson and Ko). DESPITE RECENT ADVANCES in surgery and adjuvant therapy, the prognosis for patients with gastric and pancreatic cancer remains poor. 1 Lymph node metastases have been shown to be a powerful prognostic factor after resection of early-stage gastric 2-5 and pancreatic 6-10 malignant neoplasms. Lymphadenectomy with appropriate pathologic evaluation is a critical component of complete staging. If too few lymph nodes are examined, an accurate determination of nodal involvement is less likely because metastatic lymph nodes could fail to be examined. As a result, a patient s disease may be understaged by being improperly designated as node-negative disease. This inaccuracy in establishing the stage of disease affects prognostic information, adjuvant treatment decisions, and stratification for clinical trials. Recent efforts have attempted to quantify the number of nodes that need to be evaluated to deem a patient with gastric or pancreatic cancer free of nodal metastases with a reasonable degree of certainty These studies have demonstrated that increasing numbers of lymph nodes examined are associated with improved survival for gastric and pancreatic cancer, presumably through better staging or stage migration. 18 Although the precise number varies, current guidelines recommend resection and pathologic evaluation of at least 15 regional lymph nodes for gastric and pancreatic cancer ,15,17,19,20 An association has been demonstrated between increased hospital surgical volume and higher long-term survival rates for patients undergoing gastrectomy and pancreatectomy, but factors responsible for this relationship remain elusive Because the number of lymph nodes exam- 671

2 % of Patients With 15 Lymph Nodes Examined Gastric cancer Pancreatic cancer Year of Diagnosis Figure 1. Trends in lymph node evaluation for gastric and pancreatic cancer, 1995 to P.001 by linear 2 test. ined affects outcomes, differences in the number of lymph nodes examined may contribute to the volumeoutcome association. However, it is unknown whether nodal evaluation differs by hospital type or volume for gastrectomy and pancreaticoduodenectomy. Using the National Cancer Data Base (NCDB), patients who underwent resection for gastric and pancreatic cancer with curative intent were examined. The objectives of this study were (1) to assess lymph node evaluation for gastrectomy and pancreaticoduodenectomy using a large national cancer registry and (2) to identify patient, tumor, and hospital factors associated with examination of at least 15 lymph nodes for gastric and pancreatic cancer. To our knowledge, this is the first study to examine differences in nodal evaluation by hospital type or volume for gastric and pancreatic cancer and is the first to assess factors associated with nodal evaluation for pancreatic cancer. METHODS DATA ACQUISITION AND STUDY POPULATION Using the NCDB, patients with pancreatic and gastric cancers diagnosed from January 1, 1995, to December 31, 2004, were identified based on International Classification of Diseases for Oncology, Third Edition (ICD-O-3) 27 site codes. Gastric and pancreatic cancer cases were included if the histologic findings were consistent with adenocarcinoma based on ICD-O-3 histologic codes. Patients who underwent surgery were identified based on the Commission on Cancer s Facility Oncology Registry Data Standards 28 site-specific procedure coding. For gastric cancer, only patients undergoing gastric resection (partial, subtotal, neartotal, total, or hemigastrectomy) were included in this study. For pancreatic cancer, only patients undergoing pancreaticoduodenectomy were included in this study. For gastric and pancreatic cancer, nodal metastases have been shown to be a confounding factor because patients with node-positive disease have a higher likelihood of having more lymph nodes examined ,29 Therefore, patients were excluded if they had nodal involvement or if they had no information reported regarding nodal evaluation. Patients were also excluded if they had in situ disease or distant metastases or had received neoadjuvant chemotherapy, or neoadjuvant radiotherapy. Trends in nodal examination were assessed from January 1, 1995, to December 31, 2004, using the 2 test for trend. The Commission on Cancer classifies hospitals into academic and community centers based on case volume and access to cancer-related services and specialists. Academic centers must be primarily affiliated with a medical school or a designated National Cancer Institute (NCI) cancer center. In addition, 19 of 20 adult National Comprehensive Cancer Network (NCCN) centers and 32 of 37 NCI-designated comprehensive cancer centers report to the NCDB. The NCCN and NCI hospitals were combined for this analysis. Next, all hospitals were ranked in order of increasing mean annual hospital procedure volume, and cutoff points were selected that sorted patients into 4 approximately equal groups (quartiles). This was done separately for patients undergoing gastrectomy and pancreaticoduodenectomy. Patients were excluded if the reporting facility was not the hospital in which the surgery was performed (n=93 for gastrectomy and n=35 for pancreaticoduodenectomy). STATISTICAL ANALYSIS Categorical variables were compared using 2 tests. The Bonferroni correction was used for multiple comparisons. Medians were compared using the Mann-Whitney test. Multivariable logistic regression analysis was used to identify factors associated with evaluation of at least 15 lymph nodes. An optimal lymph node evaluation was defined for this study as resection and pathologic examination of at least 15 regional lymph nodes. Because patient-level socioeconomic data are not collected by the NCDB, median household income was ascertained through linkage with 2000 US Census Bureau data on the basis of the patient s zip code at the time of diagnosis. 30 Hospital type and volume were inserted into the model separately. The logistic regression model accounted for clustering of outcomes within hospitals using robust variance estimates; however, this had little effect on the results. 31 The Hosmer- Lemeshow goodness-of-fit test and C statistic were used to assess the fit of the model. 32 The level of statistical significance was set at P.05. All P values reported are 2-tailed. Statistical analyses were performed using commercially available software (SPSS, version 14; SPSS Inc, Chicago, Illinois; and STATA, version 9.0; Stata- Corp LP, College Station, Texas). The study protocol was approved by the Northwestern University Institutional Review Board, Chicago, Illinois. RESULTS From January 1, 1995, to December 31, 2004, the median lymph node count increased for gastric cancer from 5 to 7 and for pancreatic cancer from 4 to 7 (P.001 for both). Similarly, the proportion of patients undergoing examination of at least 15 lymph nodes increased for gastric cancer from 17.4% to 24.7% and for pancreatic cancer from 9.7% to 18.5% (P.001 for both) (Figure 1). To evaluate the most current data on nodal evaluation, patients were limited to those whose cancer was diagnosed from January 1, 2003, to December 31, 2004, yielding 3088 patients undergoing gastrectomy and 1130 patients undergoing pancreaticoduodenectomy. For gastric cancer, 11.6% underwent surgery at NCCN-NCI hospitals, 34.0% at other academic hospitals, and 54.4% at community hospitals. For pancreatic cancer, 19.0% underwent surgery at NCCN-NCI hospitals, 43.3% at other academic hospitals, and 37.7% at community hospitals. The differences in patient characteristics by whether at least 15 nodes were examined are given in Table

3 Table 1. Characteristics of Patients Undergoing Gastric or Pancreatic Cancer Resections, 2003 to 2004 Gastric Cancer a (n=3088) Pancreatic Cancer b (n=1130) Variable 15 Nodes 15 Nodes 15 Nodes 15 Nodes No. (%) of patients 2371 (76.8) 717 (23.2) 945 (83.6) 185 (16.4) Female sex, % Age, median (interquartile range), y c 72 (63-79) 69 (59-78) d 68 (59-75) 67 (56-75) Race/ethnicity, % White d Black Asian d Hispanic Other Median household income, $, % d Insurance status, % Private d Medicare d Medicaid Other or governmental Uninsured c Charlson Comorbidity Index 2, % Pathologic T stage, % d d Poorly differentiated, % Type of resection, % Total gastrectomy d Partial or other gastrectomy d Hospital type, % NCCN-NCI d d Other academic d Community d d Hospital volume quartile, % Highest d High Moderate d Low d d Abbreviation: NCCN-NCI, National Comprehensive Cancer Network National Cancer Institute. a Missing data for sex (n=2), median household income (n=195), insurance status (n=81), pathologic T stage (n=28), and grade (n=223). b Missing data for median household income (n=105), insurance status (n=56), and grade (n=189). c Compared using the Mann-Whitney test. d P.05 compared with patients having fewer than 15 lymph nodes examined using the 2 test. NODAL EVALUATION FOR GASTRIC CANCER For gastric cancer, the median number of nodes examined was 7 (interquartile range, 3-14) (Table 2 and Figure 2). The NCCN-NCI centers examined more lymph nodes than other academic and community hospitals (median, 12 at NCCN-NCI, 8 at other academic, and 6 at community hospitals; P.001 for all). Patients undergoing surgery at highest-volume hospitals had significantly more lymph nodes examined than patients at high-, moderate-, and low-volume centers (median, 10 for highest, 8 for high, 6 for moderate, and 6 for low; P.001 for all except between moderate vs low). Among all patients undergoing gastrectomy, 256 patients (8.3%) had no lymph nodes examined (4.2% at NCCN-NCI, 7.3% at other academic, and 9.8% at community hospitals; P=.005 for NCCN-NCI vs community). Of 3088 patients who underwent gastrectomy, only 717 (23.2%) had at least 15 lymph nodes examined (Table 2). Patients undergoing gastrectomy had at least 15 lymph nodes examined more frequently at NCCN-NCI centers than at other academic and community hospitals (42.3% at NCCN-NCI, 25.5% at other academic, and 17.7% at community hospitals; P.001 for all). Patients at highestvolume centers had at least 15 lymph nodes examined more frequently than patients at high-, moderate-, and lowvolume centers (34.7% for highest, 22.2% for high, 17.8% for moderate, and 16.8% for low; P.001 for all). Using a multivariable logistic regression model, factors associated with examination of at least 15 lymph 673

4 Table 2. Nodal Evaluation for Gastric or Pancreatic Cancer, 2003 to 2004 Gastric Cancer Pancreatic Cancer Variable Lymph Nodes Examined, Median No. (Interquartile Range) Proportion of Patients With 15 Lymph Nodes Examined, % Lymph Nodes Examined, Median No. (Interquartile Range) Proportion of Patients With 15 Lymph Nodes Examined, % All hospitals 7 (3-14) (4-11) 16.4 Hospital type NCCN-NCI 12 (6-20) (5-15) 27.0 Other academic 8 (4-15) a 25.5 a 7 (3-11) a 15.3 a Community 6 (3-12) a 17.7 a 6 (4-11) a 12.2 a Hospital volume quartile Highest 10 (5-18) (4-13) 19.3 High 8 (4-14) a 22.2 a 8 (3-12) 18.2 Moderate 6 (2-11) a 17.8 a 7 (4-13) a 17.0 Low 6 (3-12) a 16.8 a 6 (3-10) a 10.6 b Abbreviation: NCCN-NCI, National Comprehensive Cancer Network National Cancer Institute. a P.001 compared with NCCN-NCI hospitals or highest-volume centers using the Mann-Whitney and 2 tests. b P=.004. A 700 Gastric Cancer B 175 Pancreatic Cancer No. of Patients No. of Patients No. of Lymph Nodes Examined No. of Lymph Nodes Examined Figure 2. Numbers of lymph nodes examined for gastric cancer (A) (median, 7 nodes; 23.2% had 15 nodes examined) and pancreatic cancer (B) (median, 7 nodes; 16.4% had 15 nodes examined), 2003 to nodes were assessed adjusting for potential confounders (Table 3). Patients were more likely to have at least 15 lymph nodes examined for gastric cancer if they were younger (odds ratio [OR], 1.01; 95% confidence interval [CI], ; P=.005), female (OR, 1.44; 95% CI, ; P.001), or Asian (OR, 1.61; 95% CI, ; P =.04 compared with patients of white race/ ethnicity) or underwent total gastrectomy (OR, 1.44; 95% CI, ; P.001). Patients with gastric cancer were also more likely to have at least 15 lymph nodes examined at NCCN-NCI centers than at other academic (OR, 0.40; 95% CI, ; P=.005) and community (OR, 0.28; 95% CI, ; P.001) hospitals. Similarly, patients undergoing gastrectomy at highest-volume hospitals were more likely to have at least 15 lymph nodes examined than were patients undergoing gastrectomy at low-volume hospitals (OR, 0.44; 95% CI, ; P=.001). NODAL EVALUATION FOR PANCREATIC CANCER For pancreatic cancer, the median number of lymph nodes examined was 7 (interquartile range, 4-11) (Table 2 and Figure 2). Patients who underwent pancreaticoduodenectomy at NCCN-NCI centers had more nodes examined than patients who underwent pancreaticoduodenectomy at other academic and community hospitals (median, 9 at NCCN-NCI, 7 at other academic, and 6 at community hospitals; P.001 for all). Patients undergoing resection at highest-volume hospitals had signifi- 674

5 Table 3. Logistic Regression Model Predicting Examination of at Least 15 Lymph Nodes for Gastric or Pancreatic Cancer, 2003 to 2004 a Gastric Cancer Pancreatic Cancer Variable OR (95% CI) P Value OR (95% CI) P Value Hospital type NCCN-NCI 1 [Reference] 1 [Reference] Other academic 0.40 ( ) ( ).01 Community 0.28 ( ) ( ).001 Hospital volume quartile Highest 1 [Reference] 1 [Reference] High 0.53 ( ) ( ).97 Moderate 0.44 ( ) ( ).46 Low 0.44 ( ) ( ).02 Abbreviations: CI, confidence interval; NCCN-NCI, National Comprehensive Cancer Network National Cancer Institute; OR, odds ratio. a Adjusted for age, sex, race/ethnicity, median household income, Charlson Comorbidity Index, insurance status, pathologic T stage, grade, and type of resection. Odds ratios less than 1 indicate a lower likelihood of having at least 15 lymph nodes examined. cantly more lymph nodes examined than patients undergoing resection at low-volume centers (median, 8 for highest and 6 for low; P=.004). Among all patients undergoing pancreaticoduodenectomy, 50 (4.4%) had no lymph nodes examined (4.7% at NCCN-NCI, 5.9% at other academic, and 2.6% at community hospitals; P=.51). Of 1130 patients who underwent pancreaticoduodenectomy, only 185 (16.4%) had at least 15 lymph nodes examined. Patients undergoing pancreaticoduodenectomy had at least 15 lymph nodes examined more frequently at NCCN-NCI centers than at other academic and community hospitals (27.0% at NCCN-NCI, 15.3% at other academic, and 12.2% at community hospitals; P.001 for all). Patients at highest-volume centers had at least 15 lymph nodes examined more frequently than patients at low-volume centers (19.3% for highest and 10.6% for low, P=.04). A multivariate logistic regression model was again used to assess factors associated with evaluation of at least 15 lymph nodes. No patient or tumor factors were significantly associated with having at least 15 lymph nodes evaluated for pancreaticoduodenectomy. Patients with pancreatic cancer were more likely to have at least 15 lymph nodes examined at NCCN-NCI hospitals than at other academic centers (OR, 0.49; 95% CI, ; P=.01) and community hospitals (OR, 0.37; 95% CI, ; P.001) (Table 3). Similarly, patients treated at highestvolume hospitals were more likely to have at least 15 lymph nodes examined than patients treated at low-volume hospitals (P=.02). Hospital type and volume were the only independent predictors in the model associated with examination of at least 15 lymph nodes for pancreatic cancer. COMMENT For gastric and pancreatic cancer, the definition of an optimal lymph node evaluation remains controversial Single-institution and population-based studies have documented improved survival with resection and pathologic examination of increasing numbers of lymph nodes for gastric and pancreatic cancer. However, it is difficult to discern whether this improvement in outcomes is a result of improved staging (stage migration) or if there is a therapeutic benefit to more extensive lymphadenectomy. Alternatively, the improvement in survival may be a surrogate for better overall surgical technique, pathologic examination, perioperative care, adjuvant treatment use, or cancer surveillance activities. Regardless, the optimal number of lymph nodes required to reliably stage the disease in a patient as node negative has been found to range from 10 to 15 nodes for gastric and pancreatic cancer, and the American Joint Committee on Cancer and the NCCN have adopted this recommendation to define an optimal lymph node evaluation ,20 However, compliance with these recommendations by hospital type and volume has not been previously examined, to our knowledge. The first objective of this study was to assess lymph node evaluation for gastric and pancreatic cancer in the United States using the NCDB. Previous populationbased studies have found the median number of lymph nodes examined to be 8 for gastric cancer 11,16,29 and6to 7 for pancreatic cancer 15,17 (comparable to our median of 7 nodes for gastric and pancreatic cancer). These studies included patients with node-positive disease, and the findings indicated that the median lymph node counts were significantly higher in patients with nodal metastases compared with those with node-negative disease. This suggests that surgical resection or pathologic examination in the setting of a suspicious or positive node results in examination of more lymph nodes or that examination of increasing numbers of lymph nodes is associated with a higher likelihood of identifying a positive lymph node. Therefore, we excluded patients with node-positive disease from our analyses. The proportion of patients in our study who had at least 15 lymph nodes examined was similar to previously reported rates of 17.6% to 32% for gastric cancer and 19% for pancreatic cancer, depending on the time frame studied and whether patients with node-positive disease were included. 11,16,17,29,37 The period of our study includes an era when lymph node counts were not emphasized and were not discussed as quality metrics. Therefore, we found that the proportion of patients undergoing evaluation of at least 15 lymph nodes increased significantly from January 1, 675

6 1995, to December 31, 2004, as more emphasis has been placed on lymph node counts; however, large proportions of patients with gastric and pancreatic cancer are still inadequately staged. Numerous factors may be responsible for low nodal counts, including surgical resection, patient characteristics that may affect nodal counts, and pathologic evaluation (the method of evaluation and the use of advanced techniques). The second objective of this study was to assess other patient, tumor, and hospital factors associated with the evaluation of at least 15 lymph nodes for gastric and pancreatic cancer. Similar to our results, 2 previous studies 11,29 found that lymph node evaluation for gastric cancer is associated with patient age, sex, race/ethnicity, and type of gastrectomy (total gastrectomy vs lesser resection). However, no study (to our knowledge) has examined factors affecting the extent of lymph node evaluation for pancreatic cancer. We found no patient or tumor factors that were associated with lymph node evaluation for adenocarcinoma of the head of the pancreas. Third, we assessed the variability of lymph node evaluation by hospital type and volume. High-volume hospitals and designated cancer centers have better long-term survival rates than low-volume hospitals and non designated cancer centers However, the factors contributing to this relationship between volume and outcome remain poorly understood. Previous studies 38,39 have shown that high-volume hospitals and designated cancer centers were more likely to use surgery and multimodality therapy for pancreatic adenocarcinoma. Birkmeyer et al 21,22 demonstrated that patients undergoing high-risk cancer surgery at high-volume centers were more likely to have preoperative stress tests, invasive perioperative monitoring, and consultation with medical or radiation oncologists compared with patients undergoing high-risk cancer surgery at low-volume centers. In the present study, patients treated at high-volume hospitals and designated cancer centers were more likely to have at least 15 lymph nodes evaluated. For pancreaticoduodenectomy, hospital type and volume were the only significant predictors of having at least 15 lymph nodes examined. Multiple factors at high-volume hospitals may contribute to this finding, including the resection and pathologic examination. 34,35,40 It may be that cancer centers with a particular interest in gastrointestinal cancer surgery emphasize lymph node evaluation more than other hospitals. Singleinstitution investigations and clinical trials from centers that specialize in gastric or pancreatic cancer surgery have demonstrated higher rates of lymph node examination. 9,10,12,41,42 However, without the assistance of national cancer registry data, it is difficult to assess nodal evaluation at low-volume centers. For gastric and pancreatic cancer, this is the first study (to our knowledge) to examine differences in nodal evaluation by hospital type or volume and to specifically provide insight into nodal evaluation at low-volume and community hospitals. The results of this study should be interpreted in light of some limitations. First, all hospitals that report to the NCDB must be approved by the Commission on Cancer. Low-volume and community hospitals that submit data to the NCDB may have a higher level of cancer specialization than smaller hospitals in the United States that do not report to the NCDB. Therefore, differences in lymph node evaluation by hospital type and volume would likely be greater if all hospitals in the United States were compared. In addition, there is variation among hospitals in quality and techniques (ie, fat clearance) used in the pathologic examination of lymph nodes. Second, we used 15 lymph nodes to define an optimal nodal evaluation. Fifteen nodes is the recommended threshold for gastrectomy per the American Joint Committee on Cancer, and recent studies 9,10,12,41,42 demonstrated that, although the relationship is continuous, 15 lymph nodes is also the threshold for pancreatic cancer. Third, the analyses are limited to the variables in the database. For example, the effect of body mass index cannot be examined because cancer registries in the United States do not collect height and weight information. The NCDB also does not currently have data on surgeons. Surgical technique has an effect on the number of lymph nodes resected; however, nodal examination rates are also dependent on the diligence of the pathologist. Nodal examination should be considered a systems issue and should be examined at the hospital level. Because of similar concerns about nodal evaluation for colon cancer, the NCCN, American Society for Clinical Oncology, American College of Surgeons, and the Commission on Cancer developed a hospital-level quality measure requiring resection and pathologic examination of at least 12 regional lymph nodes, and the National Quality Forum recently endorsed the measure for quality surveillance The NCDB will serve as the monitoring mechanism for the colon nodal evaluation measure. Hospitals reporting to the NCDB will be able to examine their performance in comparison with 1440 other NCDB hospitals. Because the need to assess at least 12 lymph nodes for colon cancer has gained increasing attention during the past decade, multidisciplinary initiatives have resulted in improved lymph node examination rates for colon cancer at the local, regional, and national levels. 47 Our findings demonstrate that there is significant variability in nodal evaluation for gastric and pancreatic cancer by hospital type and volume, and a quality measure to assess nodal evaluation for these malignant neoplasms is reasonable. Multiple institutions performing standard lymphadenectomies (as opposed to extended lymphadenectomy) have demonstrated that examination of at least 15 lymph nodes is an achievable benchmark. 9,10,12,41,42 Moreover, we have shown that nodal examination rates for gastric and pancreatic cancer have improved over time. Because clinical trials are critical to improving outcomes for gastric and pancreatic cancer, it is important to stratify patients based on accurate staging, and lymph node status is perhaps the most important factor by which to stratify patients with early-stage disease. The NCDB is in the process of developing additional quality measures for other malignant neoplasms, and nodal evaluation for gastric and pancreatic cancer in the United States should be included as quality surveillance measures. CONCLUSIONS Nodal status is a powerful predictor of outcome, and every reasonable attempt should be made to assess the op- 676

7 timal number of lymph nodes to accurately stage disease in patients with gastric and pancreatic cancer. A volume-outcome association has long been recognized for cancer surgery, yet factors responsible for this relationship have remained elusive. Differences in lymph node examination may partially underlie the improved longterm survival observed at cancer centers and highvolume hospitals for gastrectomy and pancreaticoduodenectomy. Further investigation is needed to identify the reasons for better lymph node examination rates at NCCN-NCI centers and high-volume hospitals and to transfer those processes to low-volume and community hospitals. Examination of at least 15 lymph nodes is an appropriate quality surveillance measure for gastric and pancreatic cancer because it is a feasible, achievable, and modifiable factor that is associated with improved outcomes. Accepted for Publication: December 19, Correspondence: Karl Y. Bilimoria, MD, MS, Department of Surgery, Feinberg School of Medicine, Northwestern University, 251 E Huron St, Galter 3-150, Chicago, IL (k-bilimoria@northwestern.edu). Author Contributions: Study concept and design: Bilimoria, Talamonti, Tomlinson, Stewart, Ko, and Bentrem. Acquisition of data: Bilimoria, Stewart, and Ko. Analysis and interpretation of data: Bilimoria, Talamonti, Wayne, Tomlinson, Winchester, Ko, and Bentrem. Drafting of the manuscript: Bilimoria, Talamonti, and Bentrem. Critical revision of the manuscript for important intellectual content: Bilimoria, Talamonti, Wayne, Tomlinson, Stewart, Winchester, Ko, and Bentrem. Statistical analysis: Bilimoria, Stewart, Ko, and Bentrem. Obtained funding: Bilimoria. Administrative, technical, and material support: Bilimoria, Talamonti, Winchester, Ko, and Bentrem. Study supervision: Talamonti, Winchester, Ko, and Bentrem. Financial Disclosure: None reported. Funding/Support: This study was supported by the American College of Surgeons Clinical Scholars in Residence program (Dr Bilimoria) and by a grant from the Goldberg Family Charitable Trust (Dr Bilimoria). The NCDB is supported by the American College of Surgeons, the Commission on Cancer, and the American Cancer Society. Previous Presentation: This paper was presented at the Annual Meeting of the Western Surgical Association; November 6, 2007; Colorado Springs, Colorado; and is published after peer review and revision. The discussions that follow this article are based on the originally submitted manuscript and not the revised manuscript. REFERENCES 1. Jemal A, Siegel R, Ward E, Murray T, Xu J, Thun MJ. Cancer statistics, CA Cancer J Clin. 2007;57(1): Peeters KC, Kattan MW, Hartgrink HH, et al. Validation of a nomogram for predicting disease-specific survival after an R0 resection for gastric carcinoma. Cancer. 2005;103(4): Siewert JR, Böttcher K, Stein HJ, Roder JD. Relevant prognostic factors in gastric cancer: ten-year results of the German Gastric Cancer Study. Ann Surg. 1998; 228(4): Kodera Y, Yamamura Y, Shimizu Y, et al. The number of metastatic lymph nodes: a promising prognostic determinant for gastric carcinoma in the latest edition of the TNM classification. J Am Coll Surg. 1998;187(6): Kim JP, Lee JH, Kim SJ, Yu HJ, Yang HK. Clinicopathologic characteristics and prognostic factors in patients with gastric cancer. Gastric Cancer. 1998; 1(2): Geer RJ, Brennan MF. Prognostic indicators for survival after resection of pancreatic adenocarcinoma. Am J Surg. 1993;165(1): Sarela AI, Turnbull AD, Coit DG, Klimstra D, Brennan MF, Karpeh M. Accurate lymph node staging is of greater prognostic importance than subclassification of the T2 category for gastric adenocarcinoma. Ann Surg Oncol. 2003;10(7): Schmidt CM, Powell ES, Yiannoutsos CT, et al. Pancreaticoduodenectomy: a 20- year experience in 516 patients. Arch Surg. 2004;139(7): Yeo CJ, Cameron JL, Sohn TA, et al. Pancreaticoduodenectomy with or without extended retroperitoneal lymphadenectomy for periampullary adenocarcinoma: comparison of morbidity and mortality and short-term outcome. Ann Surg. 1999; 229(5): Brennan MF, Kattan MW, Klimstra D, Conlon K. Prognostic nomogram for patients undergoing resection for adenocarcinoma of the pancreas. Ann Surg. 2004; 240(2): Bouvier AM, Haas O, Piard F, Roignot P, Bonithon-Kopp C, Faivre J. How many nodes must be examined to accurately stage gastric carcinomas? results from a population based study. Cancer. 2002;94(11): Karpeh MS, Leon L, Klimstra D, Brennan MF. Lymph node staging in gastric cancer: is location more important than number? an analysis of 1,038 patients. Ann Surg. 2000;232(3): Lee HK, Yang HK, Kim WH, Lee KU, Choe KJ, Kim JP. Influence of the number of lymph nodes examined on staging of gastric cancer. Br J Surg. 2001;88(10): Pawlik TM, Gleisner AL, Cameron JL, et al. Prognostic relevance of lymph node ratio following pancreaticoduodenectomy for pancreatic cancer. Surgery. 2007; 141(5): Schwarz RE, Smith DD. Extent of lymph node retrieval and pancreatic cancer survival: information from a large US population database. Ann Surg Oncol. 2006; 13(9): Smith DD, Schwarz RR, Schwarz RE. Impact of total lymph node count on staging and survival after gastrectomy for gastric cancer: data from a large USpopulation database. J Clin Oncol. 2005;23(28): Tomlinson JS, Jain S, Bentrem DJ, et al. Accuracy of staging node-negative pancreas cancer: a potential quality measure. Arch Surg. 2007;142(8): Feinstein AR, Sosin DM, Wells CK. The Will Rogers phenomenon: stage migration and new diagnostic techniques as a source of misleading statistics for survival in cancer. N Engl J Med. 1985;312(25): Kim JP, Kim YW, Yang HK, Noh DY. Significant prognostic factors by multivariate analysis of 3926 gastric cancer patients. World J Surg. 1994;18(6): American Joint Committee on Cancer Staging. American Joint Committee on Cancer Staging Manual. 6th ed. Chicago, IL: Springer-Verlag NY Inc; Birkmeyer JD, Sun Y, Wong SL, Stukel TA. Hospital volume and late survival after cancer surgery. Ann Surg. 2007;245(5): Birkmeyer NJ, Goodney PP, Stukel TA, Hillner BE, Birkmeyer JD. Do cancer centers designated by the National Cancer Institute have better surgical outcomes? Cancer. 2005;103(3): Fong Y, Gonen M, Rubin D, Radzyner M, Brennan MF. Long-term survival is superior after resection for cancer in high-volume centers. Ann Surg. 2005;242 (4): Berger AC, Watson JC, Ross EA, Hoffman JP. The metastatic/examined lymph node ratio is an important prognostic factor after pancreaticoduodenectomy for pancreatic adenocarcinoma. Am Surg. 2004;70(3): Winchester DP, Stewart AK, Bura C, Jones RS. The National Cancer Data Base: a clinical surveillance and quality improvement tool. J Surg Oncol. 2004;85 (1): Bilimoria KY, Stewart AK, Winchester DP, Ko CY. The National Cancer Data Base: a powerful initiative to improve cancer care in the United States. Ann Surg Oncol. 2008;15(3): International Classification of Diseases for Oncology. 3rd ed. Geneva, Switzerland: World Health Organization; Facility Oncology Registry Data Standards. Chicago, IL: Commission on Cancer; Baxter NN, Tuttle TM. Inadequacy of lymph node staging in gastric cancer patients: a population-based study. Ann Surg Oncol. 2005;12(12): US Census Bureau Web site. Census /cen2000.html. Accessed January 21, Panageas KS, Schrag D, Riedel E, Bach PB, Begg CB. The effect of clustering of outcomes on the association of procedure volume and surgical outcomes. Ann Intern Med. 2003;139(8): Hosmer J, Lemeshow S. Applied Logistic Regression. New York, NY: John Wiley & Sons;

8 33. Brennan MF. Lymph-node dissection for gastric cancer. NEnglJMed. 1999;340 (12): Dicken BJ, Bigam DL, Cass C, Mackey JR, Joy AA, Hamilton SM. Gastric adenocarcinoma: review and considerations for future directions. Ann Surg. 2005; 241(1): Stojadinovic A, Brooks A, Hoos A, Jaques DP, Conlon KC, Brennan MF. An evidencebased approach to the surgical management of resectable pancreatic adenocarcinoma. J Am Coll Surg. 2003;196(6): Michalski CW, Kleeff J, Wente MN, Diener MK, Büchler MW, Friess H. Systematic review and meta-analysis of standard and extended lymphadenectomy in pancreaticoduodenectomy for pancreatic cancer. Br J Surg. 2007;94(3): Hundahl SA, Phillips JL, Menck HR. The National Cancer Data Base Report on poor survival of U.S. gastric carcinoma patients treated with gastrectomy: fifth edition American Joint Committee on Cancer staging, proximal disease, and the different disease hypothesis. Cancer. 2000;88(4): Bilimoria KY, Bentrem DJ, Ko CY, et al. Multimodality therapy for pancreatic cancer in the U.S.: utilization, outcomes, and the effect of hospital volume. Cancer. 2007;110(6): Bilimoria KY, Bentrem DJ, Ko CY, Stewart AK, Winchester DP, Talamonti MS. National failure to operate on early stage pancreatic cancer. Ann Surg. 2007; 246(2): Scott KW, Grace RH. Detection of lymph node metastases in colorectal carcinoma before and after fat clearance. Br J Surg. 1989;76(11): Pedrazzoli S, DiCarlo V, Dionigi R, et al; Lymphadenectomy Study Group. Standard versus extended lymphadenectomy associated with pancreatoduodenectomy in the surgical treatment of adenocarcinoma of the head of the pancreas: a multicenter, prospective, randomized study. Ann Surg. 1998;228(4): Bonenkamp JJ, Hermans J, Sasako M, van de Velde CJ; Dutch Gastric Cancer Group. Extended lymph-node dissection for gastric cancer. N Engl J Med. 1999; 340(12): Perlin JB, Kolodner RM, Roswell RH. The Veterans Health Administration: quality, value, accountability, and information as transforming strategies for patientcentered care. Healthc Pap. 2005;5(4): National Quality Forum Web site. Quality of cancer care performance measures: phase II. Accessed April 9, Baxter NN, Virnig DJ, Rothenberger DA, Morris AM, Jessurun J, Virnig BA. Lymph node evaluation in colorectal cancer patients: a population-based study. J Natl Cancer Inst. 2005;97(3): Gordon TA, Bowman HM, Tielsch JM, Bass EB, Burleyson GP, Cameron JL. Statewide regionalization of pancreaticoduodenectomy and its effect on inhospital mortality. Ann Surg. 1998;228(1): Bilimoria KY, Stewart AK, Palis BE, Bentrem DJ, Talamonti MS, Ko CY. Adequacy and importance of nodal examination for colon cancer in the elderly. J Am Coll Surg. 2008;206(2): DISCUSSION Anton J. Bilchik, MD, PhD, Santa Monica, California: There has been tremendous interest this year in establishing quality guidelines for cancer care. This culminated in the National Quality Forum in collaboration with the ACS [American College of Surgeons], ASCO [American Society for Clinical Oncology], NCCN, and the Commission on Cancer endorsing a 12-node minimum in colon cancer in April There have been mixed responses to this decision. On the one hand, it provides a proxy metric of quality and can be translated to other malignant neoplasms such as gastric and pancreatic cancer, as is suggested by Dr Bilimoria. It also provides important information for standardizing clinical trials. On the other hand, however, this information is in the hands of third-party payers and is already threatening reimbursement. Without fully understanding the mechanisms behind the large variations in nodal sampling, some would argue that these benchmarks are premature for the following reasons: Who is responsible? Is it the surgeon? Most surgeons say they do the same operation every time. Is it the pathologist who is relying on an experienced pathology technician to find these nodes? Or is it related to the patient s immune systems, as was recently reported in Science? So with this in mind, there are several possible interpretations of Dr Bilimoria s analysis. One, regardless of whether you are at an NCCN- or NCI-designated facility or a low- or highvolume center, 80% to 85% of patients in the United States have an inadequate operation. Two, volume is the most important factor whether you are at an academic center or not. Three, there is little difference in nodal sampling at high- or low-volume centers, 4 for gastric and only 2 for pancreas cancer. Four, lymph node sampling is actually improving over time. So which one is it? Do any of these data translate to differences in survival, and should we perhaps revisit the role of more radical lymphadenectomies in gastric and pancreas cancer? Finally, in light of recent randomized trials demonstrating a survival benefit with neoadjuvant therapy, how do you think this will affect the importance of lymphadenectomy or sampling? Dr Bentrem: Dr Bilchik noted that up to 80% may undergo an inadequate operation. I would say up to 80% do not get optimal lymph node counts. I think it is important to take a step back and look at the view from ft to see how care is delivered. You can make recommendations or create guidelines, but it is important to see how they are actually carried through. When the adjuvant therapy trial for gastric cancer came out, it was quite a noteworthy point that more than half of the patients only got a D0 lymphadenectomy, so studies to see how we are doing with established recommendations are important. With regard to the importance of center type and volume, I think it is important to consider both of those factors. We were unable to include both simultaneously in the multivariate model due to the collinearity or the overlap, but I think both are important. With regard to whether lymph node sampling is improving, we are getting a little better. Dr Hundahl s report in 2000 had a median lymph node count of 8 when looking at patients with N0 and N1 stages, so we are improving, but it is happening very slowly. Multiple prior studies for gastric and pancreatic cancer have examined the effect of nodal evaluation on survival. I did not think we needed to extend those any further. There are data from randomized trials regarding the role of extended lymphadenectomy, so we did not need to revisit that topic either. Regarding neoadjuvant therapy, you brought up an important point regarding the lessons from rectal cancer. Neoadjuvant radiation can lower lymph node counts. However, I think it is the minority of patients with pancreatic or gastric cancer who will receive neoadjuvant radiation therapy. When we looked for pancreatic cancer, it was less than 5% who received preoperative radiotherapy in the United States. With regard to the quality measures, I think these may be premature for accountability but may be appropriate for surveillance and internal hospital quality improvement. I do not think an adequate lymphadenectomy is the same thing as adequate care. Adequate staging is important; it has treatment and prognostic implications, but in terms of what is adequate care, we are a bit premature in using lymph node counts at the level of an accountability measure. Financial Disclosure: None reported. 678