P < .001 by χ2 test.
P values were 2-sided by Cochran-Armitage test for trend evaluating changes in lymph node–positive cancer over time, overall and by T stage.
Parsons HM, Tuttle TM, Kuntz KM, Begun JW, McGovern PM, Virnig BA. Association Between Lymph Node Evaluation for Colon Cancer and Node Positivity Over the Past 20 Years. JAMA. 2011;306(10):1089-1097. doi:10.1001/jama.2011.1285
Author Affiliations: Applied Research Program, National Cancer Institute, Bethesda, Maryland (Ms Parsons); Department of Surgery (Dr Tuttle), Division of Health Policy and Management (Drs Kuntz, Begun, and Virnig), and Division of Environmental Health Sciences (Dr McGovern), University of Minnesota, Minneapolis; and Masonic Cancer Center, Minneapolis (Drs Tuttle and Virnig).
Context Among patients surgically treated for colon cancer, better survival has been demonstrated in those with more lymph nodes evaluated. The presumed mechanism behind this association suggests that a more extensive lymph node evaluation reduces the risk of understaging, leading to improved survival.
Objective To further evaluate the mechanism behind lymph node evaluation and survival by examining the association between more extensive lymph node evaluation, identification of lymph node–positive cancers, and hazard of death.
Design Observational cohort study.
Setting Surveillance, Epidemiology, and End Results (SEER) program data from 1988 through 2008.
Patients 86 394 patients surgically treated for colon cancer.
Main Outcome Measure We examined the relationship between lymph node evaluation and node positivity using Cochran-Armitage tests and multivariate logistic regression. The association between lymph node evaluation and hazard of death was evaluated using Cox proportional hazards modeling.
Results The number of lymph nodes evaluated increased from 1988 to 2008 but did not result in a significant overall increase in lymph node positivity. During 1988-1990, 34.6% of patients (3875/11 200) had 12 or more lymph nodes evaluated, increasing to 73.6% (9798/13 310) during 2006-2008 (P < .001); however, the proportion of node-positive cancers did not change with time (40% in 1988-1990, 42% in 2006-2008, P = .53). Although patients with high levels of lymph node evaluation were only slightly more likely to be node positive (adjusted odds ratio for 30-39 nodes vs 1-8 nodes, 1.11; 95% CI, 1.02-1.20), these patients experienced significantly lower hazard of death compared with those with fewer nodes evaluated (adjusted hazard ratio for 30-39 nodes vs 1-8 nodes, 0.66; 95% CI, 0.62-0.71; unadjusted 5-year mortality, 35.3%).
Conclusion The number of lymph nodes evaluated for colon cancer has markedly increased in the past 2 decades but was not associated with an overall shift toward higher-staged cancers, questioning the upstaging mechanism as the primary basis for improved survival in patients with more lymph nodes evaluated.
As wide-ranging quality improvement efforts emerge throughout the health care system,1 finding mechanisms for optimizing cancer care through accurate staging and appropriate treatment has become an area of substantial interest to policy makers. Among patients surgically treated for colon cancer, several studies have demonstrated better survival for patients with more lymph nodes evaluated. Reported survival improvements among those with higher lymph node counts have approached 20% in some settings.2- 8 The proposed mechanism behind this association suggests that a more extensive lymph node evaluation reduces the risk of understaging, in which inadequate assessment may incorrectly identify a patient with node-positive disease as node negative, thus failing to identify appropriate treatment. Based on these studies, most practice organizations and consensus panels now advocate for the surgical evaluation of 12 or more lymph nodes for acceptable staging of newly diagnosed colon cancer patients,9- 14 although individual studies vary widely in their recommendations for the number of evaluated nodes necessary to accurately determine nodal status.2- 8
Recently, some studies have questioned the understaging mechanism, suggesting that efforts by payers and professional associations to increase the number of lymph nodes evaluated during colon cancer surgery may have a limited role in improving survival.7,15 On a hospital level, increasing the number of lymph nodes evaluated following colectomy for colon cancer has not been demonstrated to improve staging or survival in patients 65 years and older.7 Other studies have suggested that while patients may experience improved staging when more lymph nodes are identified, the relationship between the number of nodes evaluated, staging, and survival is not simple, with higher lymph node evaluation not necessarily leading to finding higher-staged cancers in select populations.16 Combined, these results question the hypothesis that minimizing understaging is the underlying mechanism for the relationship between lymph node evaluation and improved survival. We analyzed 20-year trends in the degree of lymph node evaluation for colon cancer and how they are associated with survival.
We used cancer data from 1988 through 2008 from the Surveillance, Epidemiology, and End Results (SEER) program (the SEER 9 Registries). Sponsored by the National Cancer Institute, SEER collects and publishes cancer incidence, treatment, and survival data from population-based cancer registries covering approximately 28% of the US population.17,18 Specifically, SEER collects information on patient age, race/ethnicity, sex, year of diagnosis, tumor registry location, tumor depth, tumor grade, number of lymph nodes evaluated, number of positive lymph nodes, and first course of treatment (not including chemotherapy). Overall and cancer-specific mortality are also reported but not recurrence. A 98% case ascertainment is mandated with annual quality-assurance studies.17
Race/ethnicity information was derived from SEER and was included in this study to account for its role as a potential confounder of the relationship between lymph node evaluation and each outcome. This study was approved by the University of Minnesota institutional review board.
Included in our study were patients older than 18 years who were diagnosed with their first invasive adenocarcinoma of the colon from January 1, 1988, through December 31, 2008. We included only patients who underwent radical resection of their colon cancer as the first course of treatment according to SEER and would therefore be eligible for nodal evaluation.
Excluded from our study were patients whose cancer was diagnosed by autopsy or first cited on the death certificate, patients who underwent preoperative irradiation (as it may reduce the ability of surgeons to perform adequate nodal evaluation),19 and patients with an unknown number of nodes examined.
Beginning in 1988, SEER has routinely recorded the number of nodes pathologically examined for each patient as well as the presence and number of positive lymph nodes (as a continuous measure). We categorized patients according to their level of lymph node evaluation in 2 ways: 12 or more lymph nodes examined (yes/no), which is generally considered an acceptable level of lymph node evaluation for determining nodal status based on several clinical guidelines10,20; and a series of smaller lymph node categories (0, 1-8, 9-11, 12-15, 15-19, 20-29, 30-39, and ≥40) that allow for more extensive evaluation of the effect of lymph node evaluation on each outcome. We additionally categorized patients according to node positivity (yes/no), with at least 1 positive lymph node recorded as an indication of lymph node–positive disease.
We evaluated differences in nodal evaluation, node positivity, and patient characteristics across years of diagnosis, which were categorized into 7 groups: 1988-1990, 1991-1993, 1994-1996, 1997-1999, 2000-2002, 2003-2005, and 2006-2008. First, we tested for trends in the proportion of patients with 12 or more lymph nodes evaluated and those with node-positive disease over time using the Cochran-Armitage test. Additionally, we evaluated differences in lymph node evaluation (using the smaller categories) and patient characteristics across time using the χ2 test.
After assessing this unadjusted relationship, we used logistic regression to examine the association between lymph node evaluation, patient demographics, tumor characteristics, and relative odds of node positivity among those with at least 1 lymph node evaluated (yes/no). Model fit was assessed using the C statistic.21 Finally, we evaluated the association between lymph node evaluation and 5-year hazard of death using Kaplan-Meier methods and Cox proportional hazards modeling. Kaplan-Meier methods were used to estimate unadjusted 5-year cumulative mortality across patient factors. Then, using patients as the unit of analysis, logistic regression and Cox models were adjusted for the level of lymph node evaluation, age group (<50, 50-59, 60-69, 70-79, ≥80 years), race (white, black, other), sex, tumor extent (American Joint Committee on Cancer [AJCC] stage or T stage), grade, tumor location, type of surgical resection according to SEER, receipt of postoperative radiation (yes/no), year of diagnosis, and registry.
There are additional factors that may explain survival that we could not include in our models. For example, we believe chemotherapy is part of the causal pathway between higher rates of lymph node evaluation and improved survival for patients with lymph node–positive disease. That is, chemotherapy does not cause more lymph nodes to be evaluated, but those with more extensive lymph node evaluation may be more likely to receive chemotherapy when indicated (eg, AJCC stage III disease), leading to improved survival.7 However, SEER does not release receipt of adjuvant chemotherapy to researchers. As a result, differences in 5-year hazards of death will necessarily reflect the composite effect of lymph node evaluation in addition to other quality indicators such as chemotherapy. To take this potential factor into account, we stratified our results by nodal status at diagnosis to examine the relationship between nodal evaluation and survival among those with either node-positive (AJCC stage III and IV) or node-negative (AJCC stage I and II) disease.
In all models, we performed several sensitivity analyses (eg, alternative category groupings, removal of nonsignificant factors, interaction analyses, colinearity, and overfitting) to ensure that the observed effects were not an artifact of our modeling decisions. Because the effect of lymph node evaluation on odds of node positivity may vary by tumor extent (eg, T stage) and patient characteristics (eg, age at diagnosis), we tested for interactions between lymph node evaluation and these factors in all multivariate models. Further, because AJCC staging classifications are dependent on nodal status to determine stage and have changed over time,22 we focused on T stage as our proxy for tumor extent when evaluating factors associated with node positivity; however, both were considered independently in the survival models. T stage is defined as the depth of bowel penetration of the tumor in the colon, with more extensive bowel penetration indicative of more advanced T stage.23 Lymph nodes are not taken into account when assigning T stage. Under all assumptions, conclusions remained unchanged. We used SAS version 9.1 (SAS Institute, Cary, North Carolina) for all analyses. All comparisons were preplanned. P values were 2-sided with a level of significance of ≤.05.
We identified 86 394 surgically treated patients diagnosed with a primary invasive colon cancer from 1988 through 2008 in the SEER program after excluding patients with cancers diagnosed by autopsy or first cited on the death certificate (n = 81), those who underwent preoperative irradiation (n = 216), and those who had an unknown number of nodes evaluated (n = 3537). Over time, the distribution of patients shifted to those younger at diagnosis, with a higher proportion of proximal and T1 tumors (Table 1). Specifically, the proportion of patients diagnosed at younger than 50 years increased from 6% (n = 715) to 9% (n = 1234) between the 1988-1990 and 2006-2008 periods (P < .001). Over the same time period, the proportion of proximal tumors in the cohort increased from 55% (n = 6191) in 1988-1990 to 62% (n = 8231) in 2006-2008 (P < .001). Additionally, these patients were diagnosed at an earlier AJCC stage, with only 16% of tumors (n = 1795) classified as AJCC stage I in 1988-1990 compared with 25% (n = 3270) in 2006-2008 (P < .001).
Lymph node evaluation for colon cancer increased markedly from 1988 to 2008 (Table 1). During 1988-1990, only 34.6% of patients (n = 3875) were receiving acceptable (≥12) lymph node evaluation (Figure 1 and Table 1). By 1994-1996, 37.9% of patients (n = 4362) had 12 or more lymph nodes evaluated, with 46.8% (n = 6175) receiving this level of evaluation in 2000-2002 and 73.6% (n = 9798) in 2006-2008 (P < .001).
Although the number of lymph nodes evaluated increased significantly over time, this change did not result in an increase in node-positive cancers over the period 1988 to 2008 (P = .53) (Figure 2). However, between T stages, there were statistically significant but clinically modest increases in the proportion of node-positive cancers as rates of lymph node evaluation increased. While patients with T2 tumors had relatively consistent rates of node positivity over time (16.8% in 1988-1990 to 19.1% in 2006-2008), those with T3 tumors had statistically significant increases in node positivity (38.8% in 1988-1990 to 49.8% in 2006-2008, P < .001) in addition to T1 and T4 tumors (Figure 2).
In our study, 96.9% of surgically treated colon cancer patients (n = 83 671) had at least 1 lymph node evaluated (Table 2). Among these individuals, multivariate analyses demonstrated that, after adjusting for patient, tumor, and initial treatment factors, those with adequate lymph node evaluation (model 1) were significantly more likely to have node-positive disease (adjusted odds ratio [OR] for ≥12 nodes vs <12 nodes, 1.13; 95% CI, 1.09-1.17). However, those with very high levels of lymph node evaluation (model 2) were only slightly more likely to have node-positive disease compared with those with few nodes evaluated (adjusted OR for 30-39 nodes vs 1-8 nodes, 1.11; 95% CI, 1.02-1.20). In addition to lymph node evaluation, younger age, higher T stage, and tumor grade were all associated with higher odds of node positivity. Patients diagnosed in later years were also more likely to have node-positive disease (adjusted OR for 2006-2008 period vs 1988-1990, 1.23; 95% CI, 1.16-1.30), after adjusting for level of lymph node evaluation.
Overall 5-year cumulative mortality was 46.9% (n = 36 435) in our study for all surgically treated patients diagnosed with AJCC stage I through IV colon cancer (Table 3). Although patients with high rates of lymph node evaluation were only slightly more likely to have node-positive disease, these patients experienced significantly lower relative hazard of 5-year death compared with those with fewer nodes evaluated (adjusted hazard ratio [HR] for 30-39 nodes vs 1-8 nodes, 0.66; 95% CI, 0.62-0.71; unadjusted 5-year mortality, 35.3%). When stratified by node positivity, patients with node-positive disease (AJCC stages III and IV: adjusted HR for 30-39 nodes vs 1-8 nodes, 0.73; 95% CI, 0.67-0.79) as well as node-negative disease (AJCC stages I and II: adjusted HR for 30-39 nodes vs 1-8 nodes, 0.54; 95% CI, 0.48-0.62) continued to experience lower relative hazard of death when more lymph nodes were evaluated. In addition to higher lymph node evaluation, distal cancer site and later year of diagnosis were also associated with lower relative hazard of death. Older age at diagnosis, black race, more advance AJCC stage, and high tumor grade were all associated with higher 5-year relative hazard of death (P < .05 for all).
In this population-based study of patients surgically treated for colon cancer from 1988 through 2008, we found marked increases in lymph node evaluation over the past 2 decades. However, this improvement in lymph node evaluation has not been associated with an increase in the overall proportion of cancers that are node positive in the population. While patients with high levels of lymph node evaluation were only slightly more likely to be node positive than those with few nodes evaluated, patients with both node-positive as well as node-negative disease had significant reductions in mortality hazard attributable to high node counts. The combination of no substantive change in the proportion of patients with positive nodes concurrent to a large secular increase in the number of lymph nodes examined and the paradoxically better survival in patients with node-negative compared with node-positive disease who have greater numbers of nodes examined suggests that upstaging cannot be the mechanism underlying the relationship between increased lymph node evaluation and colon cancer survival.
Initially, the primary mechanism for the observed association between more extensive lymph node evaluation and improved survival was believed to be upstaging. In other words, a more extensive evaluation of lymph nodes would result in more accurate identification of lymph node–positive cancers. However, in a hospital-level analysis of surgically treated colon cancer patients 65 years and older, Wong et al7 found no evidence of better 5-year survival among hospitals with higher levels of lymph node evaluation. Further, they found that regardless of the number of lymph nodes a hospital evaluated, they were equally likely to find node-positive tumors. In another study of surgically treated colon cancer patients with T3 disease, Baxter et al16 found that the odds of finding node-positive cancers increased in patients with more lymph nodes evaluated—but only to a point. While the proportion of their patients with node-positive disease increased with larger nodal counts at low levels (1-6 nodes), those with 7 nodes evaluated were as likely as patients with more than 30 nodes evaluated to be node positive—suggesting that other unmeasured factors may lead to identification of node-positive disease, potentially influencing survival. Bui et al3 further examined this relationship, noting that among patients with node-negative disease from the Ontario Cancer registry, higher lymph node counts were associated with improved survival; however, patients with node-positive disease were not simultaneously evaluated.
Our study builds on these previous studies conducted in select groups, presenting a population-based analysis of the relationship between lymph node evaluation, upstaging, and survival that is representative of adult colon cancer patients in the United States and not limited to elderly individuals or specific AJCC stages. After adjusting for patient, tumor, and primary treatment factors, we found patients with node-negative disease had lower 5-year mortality when more lymph nodes were evaluated. This effect was unexpectedly larger than that observed for patients with node-positive disease. These findings suggest that providers who evaluate more lymph nodes may provide some other unmeasured care, leading to better outcomes. Alternatively, the relationship between nodes evaluated and survival may reflect an underlying interaction between the tumor and individual, influencing survival.2,24 In other words, tumor factors may stimulate lymph nodes to enlarge, reflecting immune system recognition of the tumor and more favorable survival outcomes. Although our results cannot provide insight into which mechanism underlies the observed patterns, our findings do indicate that greater lymph node evaluation unlikely leads to improved survival primarily through the more accurate detection of node-positive disease. These findings suggest that other factors besides upstaging, such as improved surgical quality or postsurgical care, may be the driving mechanism behind the lymph node–survival relationship. As a result, implementing wide-range quality improvement initiatives to increase lymph node evaluation for colon cancer may have a limited effect on improving survival in this population.
Our results also confirm prior studies that indicate younger age,16 more advanced tumor depth,25,26 and high tumor grade16 are all important predictors of node positivity in the population. Further, we identify year of diagnosis as an additional predictor of node positivity, which may result from the adoption of new pathology or surgery techniques for harvesting lymph nodes or identifying micrometastases.27- 29 However, it is interesting to note that while the relative odds of identifying node-positive disease increased over time, the overall proportion of node-positive cancers did not significantly increase in the population despite large increases in the number of lymph nodes evaluated.
Finally, our results identify the continued inadequacy of lymph node evaluation, regardless of the role it may play in cancer staging or survival of patients. In a 2008 study of hospitals in the National Cancer Data Base, Bilimoria et al30 found that more than 45% of hospitals were still performing below guideline recommendation levels for adequate lymph node evaluation (≥12 nodes), more than 12 years after the first guideline was published. They suggested that although the proportion of hospitals consistently performing guideline-recommended staging increased markedly from the 1996-1997 period to 2004-2005, the large number of US hospitals remaining nonadherent with lymph node evaluation guidelines in 2004-2005 may indicate to some that this quality measure is not appropriate or less relevant for some cases. Our study builds on these findings and identifies that more than 25% of patients still had fewer than 12 lymph nodes evaluated in 2006-2008.
Although our study does provide further insight into the relationship between lymph node evaluation and survival in a population-based setting, we acknowledge several data-related limitations. First, SEER does not collect information on comorbidities that may have affected the ability of the surgeon to remove an adequate tissue sample for lymph node evaluation. However, patients undergoing resection should have had an underlying level of general health to undergo the procedure. Additionally, we cannot differentiate between patients diagnosed through screening vs symptomatic presentation. However, others have shown that the proportion of patients diagnosed through screening has increased over time and is consistent with the shift toward earlier stages at diagnosis over time in our study.31,32 Additionally, we are unable to determine the reason behind the extensiveness of lymph node evaluation for an individual cancer. Importantly, these results are more representative of cancers diagnosed in the US population because the data are not limited to elderly individuals and include all patients diagnosed with AJCC stage I-IV disease.
In conclusion, the number of lymph nodes evaluated for colon cancer markedly increased in the past 2 decades but was not associated with an overall shift toward higher-staged cancers, questioning the upstaging mechanism as the primary basis for improved survival in patients with more lymph nodes evaluated.
Corresponding Author: Helen M. Parsons, MPH, Applied Research Program, National Cancer Institute, Executive Plaza North, 6130 Executive Blvd, MSC 7344, Bethesda, MD 20892-7344 (firstname.lastname@example.org).
Author Contributions: Ms Parsons 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: Parsons, Tuttle, Virnig.
Acquisition of data: Parsons, Virnig.
Analysis and interpretation of data: Parsons, Tuttle, Kuntz, Begun, McGovern, Virnig.
Drafting of the manuscript: Parsons, Tuttle, Virnig.
Critical revision of the manuscript for important intellectual content: Parsons, Tuttle, Kuntz, Begun, McGovern, Virnig.
Statistical analysis: Parsons, Kuntz, Virnig.
Administrative, technical, or material support: Parsons, Tuttle, McGovern.
Study supervision: Tuttle, Begun, Virnig.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.