Better recovery of lymph nodes from colorectal carcinoma resection specimens has been shown to be associated with higher survival rates for patients with TNM Stage II and Stage III tumors. It is possible that inadequate lymph node recovery and/or assessment could contribute to disparities in survival, with particular variation according to hospital volume.
Data from a population-based study that involved 33 counties in North Carolina and was conducted between April 1997 and April 2000 were available for the examination of variations in lymph node recovery and detection of positive lymph nodes according to self-reported demographic characteristics and hospital volume. The study comprised 324 patients with T2–T3N0–N1M0 colon adenocarcinoma. Logistic regression was used to determine odds ratios (ORs) associated with the recovery of fewer than seven lymph nodes and ORs associated with the detection of a positive lymph node according to hospital volume and patient characteristics.
Low-volume hospitals were more likely to recover < 7 lymph nodes compared with high- and medium-volume hospitals (low-volume vs. high-volume: adjusted OR, 1.9; 95% confidence interval [CI], 0.8–4.6; low-volume vs. medium-volume: adjusted OR, 1.7; 95% CI, 0.7–4.5) and less likely to detect positive lymph nodes. After controlling for tumor characteristics, low-volume hospitals were less than one-half as likely to detect a positive lymph node (low-volume vs. high-volume: adjusted OR, 0.3; 95% CI, 0.1–0.8; low-volume vs. medium-volume: adjusted OR, 0.4; 95% CI, 0.1–1.2).
Accurate detection and examination of lymph nodes recovered from colorectal resection specimens are essential in distinguishing Stage II (T3–4N0M0) colon tumors from Stage III (T1–4N1–3M0)1 colon tumors and ensuring the best chance of survival for the patient. Standard treatment for Stage III malignancies with lymph node involvement includes adjuvant chemotherapy, but this is not necessarily the case for Stage II malignancies.2 Patients with higher-stage disease who are incorrectly diagnosed with Stage II colon carcinoma may not receive appropriate or sufficiently aggressive treatment for the prevention of distant metastases. The percentage of patients with Stage III disease and lymph node involvement has been shown to increase with the number of lymph nodes recovered.3, 4 Therefore, the number of lymph nodes recovered during surgical resection may be an important indicator of how accurately patients' tumors are staged.
There is no consensus regarding the minimum number of lymph nodes that should be recovered to accurately stage a tumor. Studies addressing this question have recommended between 6 and 17 lymph nodes,3–6 although 3 of these 4 studies suggested ≥ 12 lymph nodes. Long-term survival rates have been positively associated with the number of lymph nodes recovered.6–10 In 2 studies, 5-year survival rates were found to be 24% (P = 0.03) and 19% (P = 0.0014) lower, respectively, for patients with Stage II colon carcinoma and < 7 lymph nodes recovered compared with patients with Stage II disease and ≥ 7 lymph nodes recovered.8, 10 In patients with lymph node–negative rectal carcinoma, 5-year survival rates were 14% and 9% lower, respectively, for patients with 0–4 and 5–9 lymph nodes recovered compared with patients with ≥ 14 lymph nodes recovered.9
As is the case with a number of other types of surgery, disparities in long-term survival after colorectal carcinoma resection according to hospital volume have been observed in the United States. Schrag et al.11 found that patients with Stage II colon carcinoma and patients with Stage III colon carcinoma at low-volume hospitals were 18% and 23% more likely, respectively, to have died within 5 years after surgery compared with patients at high-volume hospitals. A number of studies have assessed the influence of surgeon case volume on local disease recurrence and/or long-term survival rates. Three studies found a substantial disparity according to surgeon case volume after resection for rectal carcinoma,12–14 whereas one study found no disparity after colorectal resection.15
Although differences in treatment may explain some of these observed disparities, that explanation appears to be more pertinent for patients with Stage III malignancies, for whom adjuvant chemotherapy has been shown to increase survival. Alternatively, inaccurate staging may be more likely to account for the observation of poorer survival rates at low-volume hospitals. It is plausible that positive lymph nodes are detected less frequently in patients at low-volume hospitals than in patients at high-volume hospitals. We used pathology reports and other data on participants in the North Carolina Colon Cancer Study (NCCCS) to investigate variations in lymph node recovery and positive lymph node detection according to hospital volume and according to a number of patient demographic characteristics.
MATERIALS AND METHODS
Pathology reports on patients with colon carcinoma who participated in the NCCCS were used in the current analysis. The NCCCS is a population-based, case–control study with recruitment of participants from 33 counties in central and eastern North Carolina. It was designed to investigate a range of environmental and genetic risk factors for colon carcinoma. The rapid case ascertainment system of the North Carolina Central Cancer Registry (NCCCR) documented all incident cases of colon carcinoma (excluding rectal tumors) in these counties between April 1997 and April 2000. By design, 100% of African-American patients and a random sampling of white patients were recruited to assemble a study population with approximately equal numbers of African-American and white patients between the ages of 40 and 79 years.
Of those who were contacted, 85% completed an in-home interview (n = 645). The overall case response rate was 79%—7% and 14% of all eligible patients were excluded at the physician's discretion and at the patient's discretion, respectively. Study entry was denied in instances in which the patient was deceased or insufficiently healthy. The study was approved by the Institutional Review Board of the University of North Carolina (Chapel Hill, NC) or by the review boards at participating hospitals as required. Written informed consent was obtained from each patient.
Only histologically confirmed cases of invasive adenocarcinoma of the colon (not including rectal tumors) were eligible for study enrollment. The current analysis was restricted to cases with T2–T3 infiltration and no distant metastases (M0). Patients classified as having T1N0M0 or T4N0M0 disease were excluded to avoid the possibility that visual tumor characteristics would influence either the surgical procedure or the number of lymph nodes recovered.
Participants in the current study completed an extensive in-person interview conducted by a trained registered nurse interviewer. Self-reported age, race, gender, education, address, and insurance type were collected. Using participants' zip codes, an urban/rural indicator was created using 1997 U.S. Census definitions of metropolitan statistical areas.
The number of lymph nodes recovered, the number of positive lymph nodes identified, the length of the resected specimen, tumor dimensions, tumor site, and hospital identity were abstracted from patients' pathology reports. When available, one or two paraffin-embedded tumor blocks per tumor were sectioned and stained with hematoxylin and eosin. The single study pathologist (J.W.) reviewed two 5 μm sections per block and recorded diagnostic tumor characteristics such as tumor type, histologic type and grade, and microscopic depth of invasion. For the eight patients with more than one primary tumor, only the characteristics of the most advanced tumor were used. If blocks were unavailable, the slides used to make the diagnosis were retrieved and reviewed whenever possible.
Hospital volume was categorized roughly into tertiles based on the total number of cases reported by the NCCCR during the enrollment period for the NCCCS. Two hospitals that did not adhere closely to the rapid case ascertainment system were reclassified as high-volume hospitals (following their original classification as low-volume hospitals) based on other registry data. There were 9 hospitals with fewer than 35 patients during the enrollment period, 10 hospitals with 62–109 patients, and 11 hospitals with 114–280 patients.
Unadjusted differences in the mean number of lymph nodes recovered during surgery, along with corresponding 95% confidence intervals (CI), were calculated according to hospital volume, race, gender, age category, level of education, type of insurance, and urban/rural residence. Multiple regression analysis was then used to calculate mean values and 95% CIs adjusted for tumor site and tumor dimensions.
Crude and multivariable logistic regression were used to measure the association between hospital volume and the recovery of fewer than seven lymph nodes. Seven lymph nodes was chosen as the cutoff point because decreased survival has been shown to be associated with the recovery of fewer than seven lymph nodes.8, 10 Other cutoff points were examined with roughly similar results, but only the results obtained using a cutoff point of seven lymph nodes are reported. Multivariable models were used to adjust for tumor site (left-sided vs. right-sided tumors), tumor dimensions (continuous), and age (continuous). Resection specimen length was also considered, but was not included in the final models because it did not substantially change the main effect estimates. Potential correlations between recovery of fewer than seven lymph nodes and race, gender, age, level of education, type of insurance, and urban/rural residence were also evaluated. Similarly, models were adjusted for age, tumor site, and tumor dimensions.
Correlations of hospital and patient characteristics with the detection of positive lymph nodes were assessed using stratified analyses and logistic regression methods. Chi-square tests were used to assess whether the proportion of patients with positive lymph nodes was significantly associated with hospital volume, age, race, gender, education, type of insurance, or urban/rural residence. Crude and multivariable logistic regression methods were used to determine the magnitude of the association between these variables and the detection of one or more positive lymph nodes. All multivariable models controlled for tumor site, tumor grade, microscopic depth of infiltration, tumor dimensions, and age. In addition, associations with demographic characteristics were adjusted for hospital volume, which changed the main effect estimate by > 10% for some variables.
Because a consensus regarding the number of lymph nodes needed to adequately rule out the possibility of metastatic disease does not exist, the data collected were also used to examine the association between number of lymph nodes recovered and number of positive lymph nodes detected. We grouped patients according to the maximum number of lymph nodes recovered, calculated the proportion of patients with positive lymph nodes in each group, and then graphically examined the relation. We estimated the minimum number of lymph nodes required for adequate staging to be the point at which additional recovery of lymph nodes no longer appeared to improve positive lymph node detection, all the while taking into account the number of lymph nodes needed to equal the expected proportion of patients with positive lymph nodes based on data from the National Cancer Institute Surveillance, Epidemiology, and End Results (SEER) Program for the period 1975–2000.16
Six hundred forty-five patients were enrolled in the NCCCS. After exclusion criteria for histology and stage, 324 were eligible for the current analysis. Distributions of selected characteristics, overall and stratified according to hospital volume, are presented in Table 1.
Table 1. Distribution of Patient Characteristics Overall and with Stratification According to Hospital Volume
The mean number of lymph nodes recovered was 12.2 (standard deviation, 8.4; range, 0–50). Patients with positive lymph nodes had a significantly greater mean number of lymph nodes recovered compared with patients with negative lymph nodes (13.8 vs. 11.2; difference, 2.6 [95% CI, 0.8–4.5]). Consistent with data obtained from the SEER Program, 38% of patients in the current study population had positive lymph nodes (proportion with positive lymph nodes in SEER Program, 37%).16 The percentage of patients with positive lymph nodes increased with increasing lymph node recovery until approximately 14 lymph nodes were recovered, at which point the percentage of patients with positive lymph nodes approached 38%. Recovery of > 14 lymph nodes was not associated with the detection of positive lymph nodes in a larger percentage of patients, suggesting that the recovery of a minimum of 14 lymph nodes is necessary for adequate staging (Fig. 1).
Tumor site and tumor dimensions were significant predictors of the number of lymph nodes recovered. The odds ratio (OR) associated with recovering ≥ 7 lymph nodes was 3 times greater for patients with left-side lesions than for patients with right-side lesions (OR, 3.0; 95% CI, 1.8–5.0). The mean tumor dimension for patients with ≥ 7 lymph nodes recovered was 5.0 cm, compared with 3.8 cm for patients with < 7 lymph nodes recovered (P < 0.0001).
Comparisons According to Hospital Volume
The mean number of lymph nodes recovered was smallest for low-volume hospitals, although this finding was not statistically significant (low-volume, 10.2 [95% CI, 7.1–13.4]; medium-volume, 12.3 [95% CI, 10.6–13.9]; high-volume, 12.5 [95% CI, 11.3–13.6]). Low-volume hospitals were also twice as likely to recover < 7 lymph nodes compared with both high-volume (crude OR, 2.1; 95% CI, 1.0–4.5) and medium-volume hospitals (crude OR, 1.9; 95% CI, 0.8–4.3). The association was only slightly weakened by the addition of tumor site, tumor dimensions, and age into a logistic regression model (Table 2).
Table 2. Number of Unadjusted Mean Lymph Nodes Recovered and Associations with Recovery of Fewer than Seven Lymph Nodes from Colon Resection and with Detection of a Positive Lymph node
Mean no. of lymph nodes recovered (SD)
Association with the recovery of < 7 lymph nodes
Association with the detection of a positive lymph node
Ref: reference category; BMI: body mass index; SD: standard deviation; OR: odds ratio; CI: confidence interval.
Adjusted for tumor site (proximal vs. distal), tumor dimensions (continuous), and age (continuous).
Adjusted for tumor site (proximal vs. distal), tumor dimensions (continuous), tumor grade, microscopic depth of infiltration, hospital volume, and age (continuous).
Medicare (Part A)
Low (< 25)
High (≥ 30)
There was a striking difference in the proportion of patients with positive lymph nodes according to hospital volume. Although the results observed at high- and medium-volume hospitals were similar to the results expected on the basis of SEER data (43% and 36%, respectively), the percentage of patients with positive lymph nodes at low-volume hospitals was significantly lower (15%; chi-square test, P = 0.01). Even after adjusting for tumor site, tumor grade, depth of infiltration, and tumor dimensions in a logistic regression model, low-volume hospitals were three times less likely to detect positive lymph nodes compared with high-volume hospitals (Table 2).
It should be noted that low-volume hospitals were more likely than were high-volume hospitals to have specimens reviewed by contracted pathology laboratories. Approximately 40% of all cases handled by low-volume hospitals (n = 14) were reviewed by outside laboratories. Because these contracted pathology laboratories should be more experienced in reviewing colorectal resection specimens, we investigated whether the type of pathology laboratory had any effect on the detection of positive lymph nodes. Regardless of the type of laboratory contracted, patients from low-volume hospitals were still approximately two-to-three times less likely to have a positive lymph node detected compared with patients from medium- and high-volume hospitals (data not shown). However, contract pathology laboratories were disproportionately more likely to have received specimens from patients with fewer than seven lymph nodes recovered. In fact, fewer than 7 lymph nodes were submitted in 57% of all cases contracted to outside pathology laboratories by low-volume hospitals. In addition, considering all cases that were reviewed by contract pathology laboratories (n = 46), 33% of patients were found to have positive lymph nodes.
Comparisons According to Demographic Characteristics
Younger age and lower body mass index were associated with increased recovery of lymph nodes (Table 2). On average, males had fewer lymph nodes recovered compared with females (11.4 vs. 13.0; difference, 1.6 [95% CI, 0.2–3.5]) and were also more likely to have < 7 lymph nodes recovered. Although there was no association with eligibility for Medicaid, patients with Medicare insurance (Part A or B, although only data on patients with Part A insurance are presented) were more likely to have fewer than seven lymph nodes recovered compared with patients who did not have Medicare insurance, even after controlling for age at diagnosis. In contrast, the crude association indicating that patients with Medicare insurance were less likely to be diagnosed with positive lymph nodes (crude OR, 0.6; 95% CI, 0.4–1.0) was confounded by age (Table 2).
The strongest associations with lymph node recovery involved urban/rural residence and graduation from college. Patients from rural areas had a smaller mean number of lymph nodes recovered (10.3 vs. 12.8; difference, 2.5 [95% CI, 0.4–4.6]) and were substantially more likely to have < 7 lymph nodes recovered (Table 2). Stratification according to hospital volume (low vs. medium or high) suggested that rural patients at low-volume hospitals were disproportionately more likely to have fewer than seven lymph nodes recovered. Rural patients at low-volume hospitals were twice as likely (crude OR, 2.0; 95% CI, 0.7–5.2) to have < 7 lymph nodes recovered during surgery compared with rural patients at medium- and high-volume hospitals and were more than 3 times as likely as urban patients at medium- and high-volume hospitals to have < 7 lymph nodes recovered during surgery (crude OR, 3.8; 95% CI, 1.6–8.9).
Patients who were college graduates had a higher mean number of lymph nodes recovered (14.3 vs. 11.8; difference, 2.5 [95% CI, 0.08–5.01]) and were more likely to have > 6 lymph nodes recovered compared with patients with some or no college education (Table 2). College education was also associated with increased detection of positive lymph nodes. Fifty-five percent of college-educated patients had a positive lymph node detected. After controlling for hospital volume, tumor site, tumor dimensions, depth of infiltration, tumor grade, and patient age, patients with a college degree were more than twice as likely to have a positive lymph node detected compared with patients without a college degree (Table 2).
Accurate distinction between Stage II colorectal carcinoma and Stage III colorectal carcinoma relies on the thoroughness of the surgeon, appropriate specimen preparation, and vigilant examination of lymph nodes. Shortcomings in any one of these areas can lead to the increased likelihood of undetected lymph node metastases and, consequently, to a reduced chance of survival. Using archived lung carcinoma pathology specimens, Feinstein et al.17 observed a migration from lower to higher stages of lung carcinoma after reassessment of specimens with better diagnostic techniques. Within each stage group, survival increased after patients had their tumors reassigned to the proper stage, an effect termed the Will Rogers phenomenon by those authors. In the current analysis, we applied this concept indirectly by testing the hypothesis that patients at low-volume hospitals had their disease understaged as being lymph node negative more frequently compared with patients at high-volume hospitals, contributing to lower survival rates for patients at low-volume hospitals.
Although survival data were not available for the current analysis, recovery of fewer than seven lymph nodes after colorectal resection was used as a surrogate, as this characteristic has been shown to be associated with decreased survival.7, 8, 10 Our data suggest that patients at low-volume hospitals are at a higher risk of having their disease understaged compared with patients at medium- and high-volume hospitals. Both decreased surgical recovery of lymph nodes and pathologic review of the specimens may be contributing factors. As a function of the surgeon, low-volume hospitals were more likely to recover fewer than seven lymph nodes compared with medium- and high-volume hospitals. However, pathologic examination was also implicated, because even after controlling for tumor characteristics and the number of lymph nodes recovered, patients at low-volume hospitals were still less than one-half as likely to be diagnosed with lymph node–positive disease.
The observed associations between lymph node recovery and gender, college education, Medicare insurance status, and urban/rural residence were intriguing. Previous studies conducted in the United States have found that African Americans and patients with higher levels of education were more likely to have advanced colorectal carcinoma.18–20 Likewise, in the current study, college graduates were more than twice as likely to have positive lymph nodes detected. However, some of this disparity may be attributable to the more thorough recovery of lymph nodes from college-educated patients, as was observed in the current analysis. We can only speculate as to why certain demographic characteristics were associated with decreased recovery of lymph nodes, and the potential for residual confounding by variables included or not included in the analysis must also be considered. Further study is needed to determine whether these observations can be replicated and to identify factors that may account for such associations (if they persist in other study populations).
It is possible that patient comorbidity could influence the thoroughness of the surgeon in attempting to recover lymph nodes. Somewhat consistent with this assertion was the finding that increasing age was associated with decreasing lymph node recovery. However, adjustment for age in multivariable models had little influence on the results, suggesting that comorbidity, if (as expected) it is associated with patient age, may not account for the decreased recovery of lymph nodes. Examination of the variability in lymph node recovery according to the experience and specialty of the practicing surgeon would be informative, because these factors have been shown to be associated with survival after colorectal resection.11, 14, 21, 22 Similarly, the experience of the pathologist collecting the lymph node specimens and the techniques used for recovery are likely to be important factors. Unfortunately, these data were not available for the current analysis.
Because hospital volume was categorized into tertiles based on the total number of cases with colon carcinoma during the enrollment period, the number of cases handled at low-volume hospitals was quite small (n = 33). Although this can be considered a random sampling of cases, the inclusion of additional cases would have led to the improved representation of each low-volume hospital and would have increased the precision of our estimates.
The current investigation was a true population-based study, in which the randomly sampled cases represented all patients within a defined geographic area. The study cohort included substantial numbers of African-American patients and rural residents, two groups that typically are underrepresented in population-based studies. In addition, data that would have been unavailable if an administrative database had been used were obtained from patient interviews.
Although a number of studies have examined disparities in mortality according to hospital volume, few have investigated the underlying causes beyond those pertaining to the experience and specialty of the practicing surgeon. To our knowledge, the current study is the first to investigate the potential contribution of inaccurate pathologic staging to disparities in colon carcinoma survival according to hospital volume. The current analysis suggests that variations in lymph node recovery and/or detection according to hospital volume do exist. This variability in terms of pathologic evaluation may partially explain the observation of poorer survival after colon resection for patients at low-volume hospitals.