Relationship between age and number of lymph nodes examined.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Jakub JW, Russell G, Tillman CL, Lariscy C. Colon Cancer and Low Lymph Node Count: Who Is to Blame? Arch Surg. 2009;144(12):1115–1120. doi:10.1001/archsurg.2009.210
To identify the factors that contribute to the disparity in the number of lymph nodes examined for curative colon cancer resections.
Our prospectively accrued cancer registry was analyzed for all colon cancer resections performed in a consecutive 52-month period (January 1, 2003, through April 30, 2007).
The study was performed at an 851-bed community hospital. Seventeen surgeons performed colon resections, with the number of resections varying from 1 to 154. Ten pathologists and 3 pathology assistants evaluated the specimens.
A total of 430 patients met the inclusion criteria and underwent surgical resection. Only patients with colon cancer were included in the study; patients with rectal cancers, in situ disease only, T4 tumors, and stage IV disease at the time of diagnosis were excluded to ensure a uniform group of patients, all undergoing resection with curative intent.
Main Outcome Measures
Age of the patient; the surgeon, pathologist, and pathology technician; stage of disease; and year of surgery were analyzed.
No statistical difference was found in the number of lymph nodes retrieved based on the surgeon (P = .21), pathologist (P = .11), or pathology technician (P = .26). Age of the patient, primary site of the tumor, stage, and year of surgery were all significantly associated with number of lymph nodes retrieved (P <.001).
The origin of a low lymph node count appears multifactorial. Inadequate lymph node retrieval for colon cancer resections cannot uniformly be attributed to 1 factor, such as the surgeon.
It has been clearly documented in colon cancer that the number of lymph nodes examined has prognostic significance, especially in stage II disease.1-3 The pathologic evaluation of only a low number of lymph nodes has been shown to increase the risk of recurrence. Inadequate staging, especially for T3N0 lesions, is considered a high-risk marker and is taken into consideration when deciding on the benefit of adjuvant chemotherapy. Many patients are inaccurately diagnosed as having node-negative disease on the basis of a pathologic investigation of a suboptimal number of lymph nodes. What remains unknown and a point of intense contention is why some patients who undergo colon resection have a high lymph node count and others a low lymph node count. The surgeon, pathologist, and patient have all been held responsible for a low lymph node count. This study was designed to identify the factors that contribute to the disparity in the number of lymph nodes examined for curative colon cancer resections.
The study was approved by the internal review board of Lakeland Regional Medical Center, whose cancer registry prospectively enters all colon cancer cases diagnosed and treated at the center into a prospective database and reports the results to the National Cancer Database (NCDB). These data have been searched by internal review by the manager of oncology data services. The criteria for the search were T1-T3, any N, and M0 colon cancers that required surgical resection at the Lakeland Regional Medical Center from January 1, 2003, through April 30, 2007. Only patients with colon cancer were included; patients with rectal cancer, those who underwent an endoscopic polypectomy only, those who had in situ disease only, those who had primary tumors that directly invaded neighboring organs (T4), and those who presented with stage IV disease at the time of diagnosis were excluded. This approach was used to ensure a uniform group of patients who had all undergone resection with curative intent. The pathology reports were reviewed to identify the pathologist and pathology technician involved with the case.
The operating surgeon, pathologist, and pathology technician were identified for each case and assigned random numeric codes to deidentify the individuals. Other pertinent information collected includes patient age, sex, number of positive lymph nodes, total number of lymph nodes examined, anatomical portion of colon resected, and year of surgery.
Pathologic lymph node evaluation at our institution has been standardized. After surgical resection, the specimen is brought to the pathology suite in the operative theater. The pathologist or pathology technician grosses the fresh specimens. Lymph nodes are identified in the mesentery by palpation and dissection of the surrounding mesenteric fat and vascular structures. Lymph nodes are submitted for histologic examination in their entirety, with those larger than 5 mm being bisected. Specimens collected after hours that do not require immediate pathologic inspection are refrigerated overnight and examined in the fresh state the following morning. No lymphatic clearing solution is used.
Descriptive statistics, including means, standard deviations, and ranges for continuous measures and frequencies and proportions for categorical data, were generated for independent and dependent measures. To compare the mean number of nodes examined between low- and high-volume surgeons, an independent t test was used. To compare the nodes examined, analysis of covariance models were created by means of version 9.1.3 of SAS statistical software for Windows (SAS Institute Inc, Cary, North Carolina); number of nodes examined was fit as the outcome measure, with age, sex, surgeon, pathologist, stage, and year as independent univariate and multivariate predictors. If the effect of the predictive variable was significant, then pairwise comparisons that used the least-squares (predicted population margins or marginal means over a balanced population) means were performed to see where differences might exist. The same type of modeling was used to find a “best fit” model, in which the nonsignificant factors are removed 1 at a time until only significant factors remain.
A total of 440 patients were identified to meet the inclusion criteria; of those, 10 patients were removed from the analysis (3 patients had the gross examination performed by the surgeon, 6 had incomplete information with regard to the pathology technician, and 1 was removed as an outlier). The mean (SD) age of the patient population was 71.7 (11.9) years, ranging from 35 to 96 years. The mean (SD) number of nodes examined was 12.5 (7.5), ranging from 0 to 50. There were 192 women and 238 men. Initially, univariate models were fit for each independent measure.
Table 1 gives the sample size, mean, and standard deviation for each of the anatomical locations. The number of lymph nodes retrieved clearly is related to location of the primary tumor and anatomical segment of the colon resected (P = .001). There is a pattern of decreased number of lymph nodes reported as the site of the tumor goes from proximal to distal.
During this period, 18 surgeons performed resections for colon cancer in our institution and were included in the study; on review, 1 surgeon (surgeon 3) performed no resections that met the inclusion criteria, so our analysis included 17 different surgeons (Table 2). The number of colon resections performed by individual surgeons based on the previously defined criteria for this study varied from 1 to 154. An initial model showed a significant surgeon effect (P = .02), but closer inspection showed that this effect was owing to 1 surgeon (surgeon 6) who had performed only 1 colon cancer operation during the study period during which 36 lymph nodes were examined. Removal of this surgeon from the analysis changes the interpretation completely because the effect of the surgeon becomes nonsignificant (P = .21). All results were analyzed with surgeon 6 removed from the sample.
In an attempt to understand the effect of increased volume of colon cancer operations on the number of lymph nodes retrieved, we divided the study into 2 groups. Those who performed fewer than 10 colon cancer resections, as defined by this study, per year during the study period were considered low-volume surgeons. High-volume surgeons performed 10 or more resections per year. The mean (SD) number of nodes examined for surgeons who performed at least 10 resections per year was 12.1 (7.2) vs 13.6 (8.2) for surgeons who performed fewer than 10 resections per year. This comparison is not significant (P = .09), but it indicates that a higher number of operations are not necessarily linked to increased numbers of nodes examined. A similar analysis for only those surgeons who had performed a total of at least 10 colon cancer resections during the study period showed similar results; the effect of the surgeon was nonsignificant (P = .35).
During this same period, 10 pathologists and 3 pathology technicians examined the lymph nodes. As indicated in Table 3 and Table 4, no statistical difference was seen in the number of nodes retrieved per specimen based on the pathologist (P = .11) or the pathology technician (P = .26). The sex of the patient was also not significant (P = .47).
Univariately, there was a significant year effect (P < .001). The mean lymph node count per specimen has clearly increased during the 52 months of this study (Table 5). Breaking down the differences further, data from the year 2003 are significantly different from those of 2006 (P <.001) and 2007 (P = .002), whereas those of 2004 is different from those of 2006 (P = .008). In addition, data from 2005 are significantly different from that of 2006 (P < .001) and 2007 (P = .02). A correlation was found between age of the patient and lymph node count that is mildly negative (r = −0.19, P < .001), which means that older patients tend to have fewer nodes examined (Figure). Stage was significantly related (P < .001) to the number of nodes examined, with increasing stage associated with an increase in lymph nodes reported for each case (Table 6). Stage I was different from stages II and III (P = .02 and P < .001, respectively), and stage II was different from stage III (P < .001).
A multivariate model showed consistent results relative to the univariate findings (Table 7); these results can be seen in Table 8. Finally, a stepwise model was fit; in this model, all predictive factors start out included and are removed iteratively until only significant factors remain. Similar to both the univariate and full model results, age, primary site, stage, and year are significantly associated with number of nodes examined; all factors are significant with P<.001. The surgeon, pathologist, and pathology technician, as factors, were not statistically significant.
The minimum number of lymph nodes required to adequately stage disease has been reported to be at least 4 and up to 40.2-7 The sixth edition of the American Joint Commission on Cancer’s Cancer Staging Manual8 states, “It is important to obtain at least 7-14 lymph nodes in radical colon and rectum resections.” Twelve lymph nodes is the minimum number for adequate staging adopted by most societies. According to the NCDB, the mean number of lymph nodes removed during partial colectomy in the United States is 9.3 Any absolute cutoff is relatively arbitrary because this is a continuous variable, and increase of the number of nodes examined is associated with an improved prognosis, regardless of the absolute cutoff point used.
There is much debate and a significant degree of controversy with regard to the reason for inadequate lymph node retrieval. The surgeon, pathologist, and patient have all been held responsible for a low lymph node count. Despite this, there is an impetus to correlate the number of lymph nodes with individual physician quality indicators, such as the pay-for-performance measure. On the basis of the Surveillance, Epidemiology, and End Results data, 12 or more lymph nodes are reported in only approximately 50% of cases. However, in the National Comprehensive Cancer Center facilities, more than 90% of colon cancer patients have their disease staged with an adequate nodal evaluation.9 This finding strongly suggests that patient factors alone are not responsible for low lymph node counts, and the number of nodes reported may be a reasonable quality indicator.
We have reviewed our cancer registry and identified 440 prospectively accrued patients who underwent surgical resection for a T1-T3, any N, and/or M0 colon cancer from January 1, 2003, through April 30, 2007, at an 851-bed community hospital. Our study selected only patients with T1-T3 colon cancers from a single institution, with no evidence of distant metastatic disease at the time of resection. Patients with rectal cancers, those who underwent endoscopic polypectomy, those who underwent colectomy for polyps, those with in situ disease only, and those with T4 lesions were excluded to ensure a more homogeneous population of patients who had undergone resections with curative intent.
Numerous studies have reported on the effect of the number of lymph nodes retrieved and pathologically evaluated because it directly correlates with the prognosis of a patient and accurate staging of his or her disease. Specifically, Joseph et al,4 in an analysis of 1585 patients undergoing a right or left hemicolectomy with greater than 10 lymph nodes examined, showed a 5-year overall survival of 100% for patients whose disease is staged node-negative by greater than 30 lymph nodes examined and only 80% in those whose disease is staged node-negative by fewer than 30 lymph nodes examined. In an updated analysis of Intergroup Trial 0089, a 14% absolute improvement in 5-year overall survival was attained if more than 20 lymph nodes were identified compared with 10 or fewer.2 This apparent survival benefit is much greater than any adjuvant chemotherapy can offer for this same group of patients with node-negative disease. From these and other studies, it appears that the surgical procedure and the pathologic evaluation of the specimen play a major factor in prediction of the prognosis of a patient. This result has been shown for both node-positive and node-negative disease,10,11 with the strongest argument for adequate staging being voiced for patients with T3N0 colon cancer. In this group of patients, the role of adjuvant chemotherapy remains controversial.
A review3 of the NCDB that involved more than 56 000 patients showed that 5-year survival was truly dependent on the number of lymph nodes examined. Five-year survival of patients with T3N0 disease was 72% if patients were staged as having node-negative disease by 5 lymph nodes examined and almost 90% if more than 30 nodes were examined. There was almost a linear relationship between number of nodes examined and 5-year survival in patients with T3N0 disease. A more recent review12 of 116 995 patients with colorectal cancer found similar trends. However, only 37% of patients received adequate nodal staging. Younger patients were significantly more likely to receive adequate nodal evaluation, defined as 12 or more nodes, with 50.5% for those 50 years or younger and 35% for those 70 years or older. A number of factors were of statistical significance in our study, such as age of the patient (P < .001), primary site of the tumor (P < .001), stage (P < .001), and year of operation (P < .001).
Many reports are available with regard to the effect of the surgeon and hospital volume on outcome. Most of these reports argue that higher-volume surgeons and higher-volume centers have better outcomes. In our study, all operations were performed at 1 hospital; therefore hospital volume could not be analyzed. The effect of surgeon volume in our study was counterintuitive. Our study indicates that a higher number of curative colon cancer resections performed in a year by a surgeon is not necessarily linked to an increased number of nodes retrieved, a proposed quality indicator. No statistical difference was found in the number of nodes examined on the basis of the surgical case volume. If anything, there was a trend for lower-volume surgeons to have more nodes examined, with 12.1 for surgeons who performed 10 or more resections per year compared with 13.6 for surgeons who performed fewer than 10 (P = .09). Even if we only compare surgeons who performed at least 10 resections a year, the individual surgeon did not appear to affect the lymph node count (P = .35). Few would argue that if someone does one task often, he or she will tend to be more adept at it than someone who performs the same task less frequently. However, our higher-volume surgeons did not, on average, harvest more lymph nodes than our lower-volume surgeons.
Rieger et al13 showed that when he performed the same operation at 2 different hospitals, the number of lymph nodes identified was significantly different at 10 vs 18 (P < .001). After standardization of the pathologic evaluation at the 2 hospitals, the mean lymph node count increased from 7 to 14 in the low-lymph-node-count hospital and remained unchanged at 14 in the high-lymph-node-count hospital, which demonstrates clearly in this practice that the pathologic evaluation, rather than the surgeon, was responsible. Our results reflect the practice pattern of 17 surgeons with a wide variety of case volumes and 10 pathologists in 1 community hospital and may not be representative of practice patterns across the country.
Previous studies have shown conflicting results. In a review of 264 retrospective cases, Ostadi et al14 showed the greatest variation in the number of lymph nodes retrieved per case correlated with the pathology technician but not the surgeon. In a population-based study15 from the southern Netherlands, the pathology group that covered the hospital and not the surgeon was responsible for variation in lymph node evaluation. In our study, the individual pathologist or pathology technician had no effect on the number of nodes examined.
In summary, studies have inconsistently supported hospital volume, surgeon, pathologist, and pathology technician as factors in the number of lymph nodes removed. Two contemporary studies have used national databases to analyze factors that influence the lymph node count and have come to different conclusions. In a 2008 report of the NCDB, Bilimoria et al16 concluded that high-volume hospitals were more likely to identify 12 or more lymph nodes and that this was a modifiable factor. In the previous year, Wong et al17 had come to different conclusions on analysis of the Surveillance, Epidemiology, and End Results database. They noted that higher-volume hospitals identified more lymph nodes, but this difference was lost when other factors were controlled. The number of lymph node–positive cases was unchanged, and the number of nodes examined was not associated with staging, use of adjuvant chemotherapy, or patient survival.
More lymph nodes retrieved from right-sided lesions have previously been reported, with fewer nodes having been examined in the transverse colon.12,16 In our study, there was a clear trend of fewer lymph nodes retrieved going from proximal to distal. This is an anatomical distinction and out of the control of the surgeon. Patient age has consistently been described as influencing the number of lymph nodes retrieved,14-16 and this theory is supported by our study. Not surprisingly, having metastatic disease in the lymph nodes increased the number of nodes identified. Interestingly, the year of surgery had a strong influence on the number of lymph nodes reported, with more recent years showing a higher lymph node count. This finding is undoubtedly secondary to an increased awareness of the importance of accurate staging by all involved in the care of patients with colon cancer. This is likely multifactorial and includes pressure from the medical oncologist for an adequate lymph node count, a more conscious effort by the surgeon to provide an appropriate anatomical mesenteric resection, and provision by the pathologist and technician of a thorough pathologic inspection, searching the mesentery more vigorously for small nodes. This is clearly a positive effect of physician education through recent publications and the attention given to this topic.
We agree with the goals of quality assurance and improvement. The number of lymph nodes retrieved in a colon cancer specimen appears to be an appropriate factor to measure. However, at this point the origin of a low lymph node count appears multifactorial and not clearly defined. Our study has shown that in recent years a higher average lymph node count has been obtained. Increased awareness on the part of physicians of the role they and the hospital play in improvement of this outcome measure is a worthwhile goal; based on our findings, this awareness appears to be having an effect. However, some factors are outside the control of the surgeon, such as age of the patient and location of the primary tumor. This may need to be considered in future quality assurance measurements before basing surgeon reimbursement (pay for performance) on a single number. Recent editorials18,19 have shared similar concerns with regard to the proposed quality indicators and have concluded that use of this sole measure is an oversimplification. The surgical procedure and type of anatomical mesenteric resection must influence the number of nodes reported. However, adequate staging appears to be a team effort dependent on the surgeon, pathologist, and some factors beyond the control of either of them. Unfortunately, measurement and report of quality is often not as easy to assess as initially hoped by looking at just 1 factor.
Correspondence: James W. Jakub, MD, Mayo Clinic, 200 First St SW, Rochester, MN 55905 (firstname.lastname@example.org).
Accepted for Publication: December 2, 2008.
Author Contributions:Study concept and design: Jakub and Lariscy. Acquisition of data: Tillman. Analysis and interpretation of data: Jakub, Russell, and Lariscy. Drafting of the manuscript: Jakub, Russell, and Tillman. Critical revision of the manuscript for important intellectual content: Jakub, Russell, and Lariscy. Statistical analysis: Russell. Administrative, technical, and material support: Jakub, Tillman, and Lariscy.
Financial Disclosure: None reported.