Clinical and Pathologic Factors That Predict Lymph Node Yield From Surgical Specimens in Colorectal Cancer

Abstract

BACKGROUND

The National Quality Forum endorses the recommendation of examining at least 12 lymph nodes (LNs) from colorectal cancer (CRC) specimens. However, heterogeneity in LN harvest exists. The objective of this study was to investigate the clinicopathologic factors that influence LN yield.

METHODS

The authors used the Surveillance, Epidemiology, and End Results database to identify patients who were diagnosed with stage I, II, and III CRC between 1994 and 2005. Poisson regression was used to model the number of LNs examined as a function of individual clinicopathologic factors, including age, sex, race, year of diagnosis, geographic region, anatomic site, pre-operative radiation, tumor size, tumor classification, tumor differentiation, and LN positivity.

RESULTS

In total, 153,483 patients with CRC were identified. The mean number of LNs examined (±standard deviation) was 12 (±9.3). Separate multivariate analyses revealed that age, year of diagnosis, tumor size, and tumor classification were significant predictors of LN yield for colon and extraperitoneal rectal cancers (P < .01 for all covariates). Tumor location and radiotherapy were significant predictors of LN yield in patients with colon cancer and rectal cancer, respectively. Overall LN yields increased between 2% and 3% annually.

CONCLUSIONS

Despite the increasing yields observed over time, patients with rectal cancer and older patients who had distally located, early colon cancer were less likely to meet the benchmark yield of 12 LNs. Further investigation into how LN yield is influenced by alterable factors, such as the extent of mesenteric resection and the pathologic technique, as well as nonalterable factors, such as patient age and tumor location, may reveal innovative ways to improve current staging methods.

In the United States, colorectal cancer (CRC) remains the third most common cancer in women and men combined and the second most common cause of cancer death.¹ For patients in the Surveillance, Epidemiology, and End Results (SEER) database who were diagnosed between 1996 and 2004, the overall 5-year relative survival rates (defined as the ratio of the proportion of observed survivors in a cohort of cancer patients to the proportion of expected survivors in a comparable cohort of cancer-free individuals) were 90%, 68%, and 11% for local, regional, and distant disease, respectively.² The American Joint Committee on Cancer/International Union Against Cancer (AJCC/UICC) TNM staging system currently classifies colon cancer based on the depth of penetration into the intestinal wall and the presence of lymph node (LN) metastasis. This system underscores the finding that accurate staging requires adequate LN evaluation, and many studies have demonstrated an association between the number of LNs examined and survival.^3–11 In addition, LN metastasis is highly predictive of outcome and is the major determinant of adjuvant therapy for patients with CRC. On the basis of the risk of recurrence after resection, the administration of adjuvant chemotherapy is standard practice for patients with stage III colon cancer and patients with stage II/III rectal cancer. Selective treatment is reserved for patients who have stage II colon cancer classified as high-risk, and is withheld for patients who have stage I disease.¹² Inadequate LN retrieval may result in underestimation of the presence of LN metastasis and subsequent undertreatment of patients with CRC. Such patients are at increased risk for local recurrence in the future.

In 2007, after considering the recommendations of various professional committees on CRC, the National Quality Forum endorsed the harvest of ≥12 LNs as a standard quality indicator for CRC resection specimens.¹³ In the literature, however, recommendations are conflicting. Several studies have suggested that harvesting more than 12 LNs is adequate for staging CRC.^6,14,15 One study concluded that the examination of at least 9 LNs may be enough to reliably stage LN-negative tumors.¹⁶ Other authors have concluded that there is no clear cutoff and have recommended that the pathologist harvest as many LNs as possible.^4,10

This apparent inability to reach a consensus regarding LN harvest within the literature demonstrates the complexity of accurately staging patients with CRC. Therefore, the objectives of the current study were to evaluate the impact of various epidemiologic and clinicopathologic factors on LN harvest and to describe temporal trends of LN examination in the United States.

MATERIALS AND METHODS

Database

We used the SEER database for the purposes of this study. SEER is a population-based registry sponsored by the National Cancer Institute. It collects and publishes cancer incidence and survival data from population-based cancer registries that cover approximately 26% of the US population. The SEER database includes 23% of African Americans, 40% of Hispanics, 42% of American Indians/ Alaskan Natives, 53% of Asians, and 70% of Hawaiians/ Pacific Islanders.² SEER routinely collects data on patient demographics, primary tumor site, tumor morphology, stage at diagnosis, and follow-up.

Study Cohort

Patients with stage I, II, and III primary CRC who were diagnosed between 1994 and 2005 were identified in the SEER database. All patients underwent surgical resection, and the relevant histologic codes were based on the International Classification of Diseases for Oncology, 3rd edition. Patients who had distant metastases, tumors of the appendix, or tumors of unspecified location were excluded along with patients who had colonic primaries who had received radiation. According to SEER, the rectosigmoid segment is located 10 cm to 15 cm from the anal verge, the rectum is 12 cm long, and the anal canal is the most distal 4 cm to 5 cm and extends to the anal verge.¹⁷ Rectosigmoid cancers (9.1%) were grouped with colon cancers in the current analyses. For patients with rectal cancer, the following categories of patients were excluded: those who received radiation preoperatively and postoperatively (10%), those who received intraoperative radiation (0.03%), those who received intraoperative radiation and radiation both preoperatively and postoperatively (0.04%), and those who received unknown radiation sequences (0.4%). After excluding these patients, in total, 153,483 patients with CRC remained in the study cohort.

Variables

Patient demographic variables included age, race, sex, year of diagnosis, and geographic region. The various registries within SEER were categorized into 4 groups based on time zone: west, midwest, northeast, and south. Clinical factors that were included were tumor location by anatomic site, radiation sequence, tumor size, tumor (T) classification, tumor differentiation, total number of LNs evaluated, and number of examined regional LNs that contained metastases. Coding for tumor stage was based on the SEER 1998 Extent of Disease Coding Manual. For tumor location, 5 categories were created based on anatomic site: cecum, ascending colon/hepatic flexure, transverse colon, splenic flexure/descending colon, and sigmoid colon/rectosigmoid junction. Radiation sequence was categorized as no radiation, preoperative radiation, and postoperative radiation. The category tumor differentiation was dichotomized into well differentiated and poorly differentiated, with the former including either well differentiated or moderately differentiated tumors and the latter including either poorly differentiated or undifferentiated tumors.

Statistical Analysis

The number of LNs examined per patient was the main endpoint of interest. Mean and median numbers of LNs examined according to individual patient characteristics were estimated both for patients with colon cancer and for patients with rectal cancer. Poisson regression was used to model the number of LNs retrieved as a function of demographic and clinical factors and to identify significant predictors of LN retrieval. Separate analyses were done for the colon and rectal cohorts. Regression coefficients in Poisson regression are log-risk ratios and, thus, are useful for interpretation. The antilog of the regression coefficient can be interpreted as a risk ratio corresponding to a 1-unit difference in the predictor. To account for over-dispersion, a scaling factor was used to adjust for the covariance matrix of a Poisson-based analysis.¹⁸ A P value <.05 was considered statistically significant. All analyses were performed using SAS software (version 9.2; SAS Institute Inc., Cary, NC) and R (version 2.7.1; R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Patient characteristics are summarized in Table 1. The mean age (±standard deviation [SD]) at diagnosis was 69.8 ±13.0 years for patients with colon cancer and 64.5 ±13.1 years for patients with rectal cancer. Overall, 49.8% of patients were men, and most were non-Hispanic whites (75.5%). According to the fifth edition of the AJCC staging classification, 25% of patients had stage I disease, 36.9% had stage II disease, and 33.5% had stage III disease. In 4.6% of patients, no regional LNs were examined; thus, these individuals were not staged. Tumors were classified as T1 in 15.1% of patients, T2 in 17.4%, T3 in 56.9%, and T4 in 10.9%. The overall mean tumor size was 46.1 mm. The overall mean number of LNs evaluated (±SD) was 12 ±9.3 LNs, with a mean (±SD) of 12.4 -9.4 LNs for patients with colon cancer and 10.2 ±8.8 LNs) for patients with rectal cancer.

Table 1.

Patient Characteristics

Characteristic	No. of Patients (%)
	All Patients, N=153,483	Colon Cohort, N=127,927	Rectal Cohort, N=25,556
Age at diagnosis, y
Mean±SD	68.9±13.2	69.8±13.0	64.5±13.1
Median [IQR]	71 [60–79]	72 [61–79]	65 [55–75]
Sex
Men	76,487 (49.8)	61,346 (48)	15,141 (59.3)
Race/ethnicity
White, non-Hispanic	115,840 (75.5)	96,591 (75.5)	19,249 (75.3)
Black, non-Hispanic	13,466 (8.7)	11,865 (9.3)	1601 (6.3)
Hispanic	11,231 (7.3)	9032 (7.1)	2199 (8.6)
Asian/othersa	12,570 (8.2)	10,110 (8)	2460 (9.6)
Unknown	376 (0.2)	329 (0.3)	47 (0.2)
AJCC 5 stage
I	38,305 (25)	30,984 (24.2)	7321 (28.7)
II	56,640 (36.9)	49,480 (38.7)	7160 (28)
III	51,418 (33.5)	42,839 (33.5)	8579 (33.6)
NAb	7120 (4.6)	4624 (3.6)	2496 (10)
Tumor classification
T1	23,096 (15.1)	18,795 (14.7)	4301 (16.8)
T2	26,683 (17.4)	20,356 (15.9)	6327 (24.8)
T3	86,924 (56.6)	73,762 (57.7)	13,162 (51.5)
T4	16,780 (10.9)	15,014 (12)	1766 (7)
Tumor grade of differentiation
Well	11,9002 (77.5)	99,049 (77.4)	19,953 (78.1)
Poor	27,572 (18)	23,459 (18.3)	4113 (16.1)
Unknown	6909 (4.5)	5419 (4.2)	1490 (5.8)
Tumor size, mm
Mean±SD	46.1±27.7	46.5±28.3	43.8±24
Median [IQR]	40 [30–60]	42 [30–60]	40 [30–55]
No. of lymph nodes evaluated
Mean±SD	12.0±9.3	12.4±9.4	10.2±8.8
Median [IQR]	10 [6–16]	11 [6–16]	8 [4–14]
Median positive [IQR]	2 [1–4]	2 [1–4]	2 [1–5]
Anatomic site
Cecum	28,882 (22.6)	28,882 (22.6)	NA
Ascending colon/hepatic flexure	29,119 (22.8)	29,119 (22.8)	NA
Transverse colon	11,105 (8.7)	11,105 (8.7)	NA
Splenic flexure/descending colon	11,849 (9.3)	11,849 (9.3)	NA
Sigmoid colon/rectosigmoid junction	46,972 (36.72)	46,972 (36.72)	NA
Radiation
No radiation	140,002 (91.2)	12,7927 (100)	12,075 (47.3)
Preoperative	6427 (4.2)	NA	6427 (25.2)
Postoperative	7054 (4.6)	NA	7054 (27.6)

SD indicates standard deviation; IQR, interquartile range; AJCC 5, American Joint Committee on Cancer, 5th edition; NA, not available; Well, well or moderately differentiated; Poor, poorly differentiated or undifferentiated.

^aOthers include American Indians/Alaskan Natives and unspecified race.

^bNo regional lymph nodes were examined.

Table 2 reports the mean and median number of LNs evaluated according to various patient demographic and clinicopathologic factors for the colon and rectal cohorts. The mean number of LNs examined according to sex and race was similar in both cohorts. The mean and median numbers of evaluated LNs increased in both cohorts for patients with stage III disease, higher T classification, and poorly differentiated tumors. Within the colon cohort, increased LN harvests were observed in tumors that were located in the ascending colon/hepatic flexure and transverse colon (Table 2). In addition, patients in both cohorts who were classified as LN-negative had smaller mean LN harvests compared with those who were classified as LN-positive (colon cancer, 12.3 vs 14 LNs, respectively; rectal cancer, 10.2 vs 12.9 LNs, respectively). There was a decrease in the number of LNs recovered with each 10-year incremental increase in age (Fig. 1, top).

Table 2.

Mean and Median Lymph Node Yields by Patient Characteristics

Variablea	Lymph Node Yield
	Colon Cohort		Rectal Cohort
	Mean±SD	Median	Mean±SD	Median
Sex
Men	12.2±9.4	10	10.1±8.8	8
Women	12.6±9.3	11	10.1±8.8	8
Race/ethnicity
White, non-Hispanic	12.4±9.3	11	10.1±8.8	8
Black, non-Hispanic	12.5±9.1	11	10.4±8.7	9
Hispanic	12.0±9.3	10	9.6±8.4	8
Asian/othersb	12.9±10.2	11	10.6±9.6	8
AJCC 5 stage
I	10.5±8.4	9	9.6±7.8	8
II	13.3±9.3	11	10.8±8.3	9
III	14.0±9.3	12	12.9±9.1	11
NAc	NA	NA	NA	NA
Tumor classification
T1	8.2±8.4	6	7.1±8.0	5
T2	11.4±8.6	10	9.6±8.0	8
T3	13.6±9.4	12	11.4±9.1	10
T4	13.1±9.5	11	9.7±9.0	8
Tumor grade of differentiation
Well	12.2±9.2	10	10.1±8.7	8
Poor	13.9±9.8	12	11.0±9.5	9
Anatomic site
Cecum	13.9±9.0	12	NA	NA
Ascending colon/hepatic flexure	14.8±9.8	13	NA	NA
Transverse colon	12.8±10.6	10	NA	NA
Splenic flexure/descending colon	11.9±9.8	10	NA	NA
Sigmoid colon/rectosigmoid junction	10.3±8.4	9
Lymph node status
Negative	12.3±9.1	10.0	10.2±8.1	8.0
Positive	14.0±9.3	12.0	12.9±9.1	11.0
Radiation
No radiation	NA	NA	10.1±8.7	8.0
Preoperative	NA	NA	8.3±8.0	7
Postoperative	NA	NA	12.0±9.5	10

SD indicates standard deviation; IQR, interquartile range; AJCC 5, American Joint Committee on Cancer, 5th edition; NA, not available; Well, well or moderately differentiated; Poor, poorly differentiated or undifferentiated.

^aThe variables with unknown status listed in Table 1 were excluded from the descriptive analyses.

^bOthers include American Indians/Alaskan Natives and unspecified race.

^cNo regional lymph nodes were examined.

Figure 1.

These graphs illustrate the mean number of lymph nodes (LNs) evaluated (Top) by age of disease diagnosis (in years), (Middle) by year of disease diagnosis, and (Bottom), in patients with rectal cancer, by radiation sequence.

In terms of time trends, there was an increase in the mean number of LNs examined for both colon and rectal cancers between 1994 and 2005 (Fig. 1, middle). In 2005, the mean number of LNs recovered for patients with colon cancer and rectal cancer was 14.5 and 11.7, respectively. It is noteworthy that there was a decrease in the mean age at CRC diagnosis over time. In 1994, the mean age at colon cancer diagnosis was 70.3 years and decreased to 68.8 years in 2005. The mean age at rectal cancer diagnosis was 65.6 years in 1994 and decreased to 63.2 years in 2005 (data not shown). Among the patients with rectal cancer who were classified according to radiation sequence, patients who received preoperative radiation had the smallest LN harvests (Fig. 1, bottom). In 2005, the mean number of LNs evaluated for the preoperative radiation group was 10.2.

Results from separate multivariate Poisson regression analyses for colon cancer and rectal cancer are reported in Table 3. For patients with CRC, age at diagnosis, year of diagnosis, T classification, tumor size, and positive LN status remained significant factors that affected LN harvests after adjusting for race, sex, and geographic region. In addition, tumor location by anatomic site and tumor differentiation remained significant factors for the colon cohort. In both cohorts, the mean number of LNs harvested decreased by roughly 10% for every 10-year incremental increase in age (P < .01). With increasing T classification, the average number of LNs recovered also increased. Relative to T1 tumors, the percentage increase in the mean number of LNs recovered for T2, T3, and T4 tumors was 11%, 24%, and 18%, respectively, for colon cancer (P < .001) and 7%, 20%, and 16%, respectively, for rectal cancer (P < .001). In terms of tumor differentiation, there was a 3% average increase in LN recovery between poorly differentiated or undifferentiated colon tumors compared with tumors that were well or moderately differentiated. However, no differences were observed in tumors of the rectum according to differentiation (P =.74). For LN positivity, the percentage increase in the mean number of LNs recovered from patients who had LN metastases was 5.7% in the colon cohort and 16.2% in the rectal cohort (P < .01). Anatomic site of the tumor relative to the sigmoid colon/rectosigmoid junction also was significant, indicating an increase in the average number of LNs retrieved from the cecum, ascending colon/hepatic flexure, transverse colon, and splenic flexure/descending colon (27%, 35%, 17%, and 9%, respectively; P < .001). In the rectal cohort, no difference in mean number of LNs harvested was noted between patients who received postoperative radiation and those who received no radiation (P =.22). However, in patients who received preoperative radiation, there was a 25% reduction in the mean number of LNs harvested (P < .001).

Table 3.

Multivariate Poisson Regression for the Colon and Rectal Cohorts^a

Variable	RRb	95% CI	P
Colon cohort
Age at diagnosis	0.918c	0.915–0.921	<.01
Year of diagnosis	1.023	1.022–1.024	<.01
Size of tumor	1.021c	1.020–1.022	<.01
Anatomic site
Cecum	1.272	1.258–1.286	<.01
Ascending colon/hepatic flexure	1.348	1.334–1.363	<.01
Transverse colon	1.170	1.153–1.188	<.01
Splenic flexure/descending colon	1.091	1.078–1.112	<.01
Sigmoid colon/rectosigmoid junction (reference)	1.000
Tumor classification
T1 (reference)	1.000
T2	1.116	1.095–1.138	<.01
T3	1.241	1.220–1.262	<.01
T4	1.181	1.157–1.205	<.01
Tumor grade of differentiation
Well (reference)	1.000
Poor	1.023	1.015–1.035	<.01
Lymph node status
Positive	1.057	1.048–1.066	<.01
Negative	1.000
Rectal cohort
Age at diagnosis	0.914c	0.907–0.921	<.01
Year of diagnosis	1.025	1.022–1.029	<.01
Size of tumor	1.03d	1.026–1.033	<.01
Tumor classification
T1 (reference)	1.000
T2	1.078	1.035–1.122	<.01
T3	1.197	1.151–1.245	<.01
T4	1.160	1.100–1.225	<.01
Tumor grade of differentiation
Well (reference)	1.000
Poor	1.004	0.978–1.031	.74
Radiation
Preoperative	0.752	0.728–0.776	<.01
Postoperative	1.000	0.961–1.009	.22
No radiation	1.000
Lymph node status
Positive	1.162	1.136–1.188	<.01
Negative	1.000

RR indicates relative risk; Well, well or moderately differentiated; Poor, poorly differentiated or undifferentiated.

^aThe multivariate Poisson regression also was adjusted for sex, race, and registration region.

^bThe RR of a 1-unit increase in the lymph node yield for the covariate. For example, an RR of 0.91 means that patients who had a 10-year increase in age had 0.91 times as many lymph nodes examined (ie, a 9.1% decrease).

^cEstimates the ratio of a 10-year incremental increase in age.

^dEstimates the ratio of a 10-mm incremental increase in tumor size.

DISCUSSION

The current results indicate a clear temporal trend toward increased LN yields in surgical specimens. This is true for both colon and rectal primary cancers. Despite this finding, fewer LNs are recovered in patients who have rectal cancer and among older individuals with more distally located, early colon tumors.

Because the primary outcome variable of this study was LN yield and not patient outcome, we cannot specifically comment on prognostic indicators or make recommendations regarding “best practice.” Nonetheless, the association between LN harvest and survival in colon cancer has been well documented.³ In a secondary analysis of the Intergroup Trial INT-0089 that analyzed overall survival in high-risk patients with stage II and III colon cancer, Le Voyer et al concluded that the number of LNs analyzed was a significant prognostic variable in both LN-negative and LN-positive patients.⁵ The median number of LNs examined in that study was 11 (range, 1–87 LNs). In a study that used Canadian registry data for stage II CRC, Wright et al reported that 73% of patients had <12 LNs assessed.¹⁹ In a United Kingdom national study by Tekkis et al, of 5164 patients with CRC who underwent resection between 2000 and 2002, an average of 11.7 LNs were examined per specimen.²⁰ By using SEER data, Baxter et al observed that 37% of 116,995 adults who were diagnosed with CRC between 1988 and 2001 had ≥12 LNs harvested from their resection specimens. Examining the data based on year of diagnosis, 33% of patients who were diagnosed during 1988 through 1990 had adequate LN evaluation compared with 42.5% of patients who were diagnosed during 2000 through 2001. The median number of LNs examined for all patients was 9.²¹ In the current study, we demonstrated that, between 2002 and 2005, only 49.2% of patients with CRC had at least 12 LNs harvested; however, within that subgroup, the overall mean number (±SD) of LNs examined per specimen was 13.1 ± 9.7.

Age

Age was a consistently important determinant of LN yield. The mean number of LNs harvested from colon and rectal cancer patients ≥70 years of age was 11.7 and 9.3, respectively; and, for every 10-year incremental increase in age, there was an average reduction of 9% in LN harvest in both cohorts (Table 3). This phenomenon may result from a complex interplay of patient and surgeon factors. Older patients who are considered to be high-risk operative candidates secondary to any number of life-threatening comorbidities may place undue pressure on the surgeon to perform less extensive mesenteric resections and to minimize anesthesia times and resultant postoperative morbidity and mortality. An alternate hypothesis proposed by Tekkis et al is that LNs in patients may undergo a process of involution with increasing age.²⁰

Tumor Location

Tumor location by anatomic site also influenced LN harvest. It is noteworthy that both of the population-based studies mentioned above reported a decrease in the number of LNs recovered with respect to anatomic site. Baxter et al reported that patients who had left-sided tumors were less likely to have adequate LN harvests than patients who had right-sided lesions.²¹ Wright et al reported a median number of 12 LNs and 9 LNs harvested for the right and left colon, respectively.¹⁹ The current findings were consistent with those reports. However, on further subcategorization of the right and left colon, we observed that tumors located in the ascending colon/hepatic flexure, on average, had 34% more LNs retrieved than those in the sigmoid colon/rectosigmoid junction. Because SEER registries do not record length, we can only speculate that this difference may be because right-sided colon cancer resection specimens were longer than left-sided specimens. Variant lymphatic anatomy may be another possible explanation; ie, a disproportionate number of LNs exist along the ileocolic artery, and there is a natural decline in LN numbers with more distal progression within the colonic mesentery.

Tumor Size

It has been proposed that larger tumors elicit an intense antigenic immune response within the regional LN basin, making them more visible to pathologic examination and possibly leading to increasing LN yields.¹⁹ Despite this, Wright et al did not appreciate an association between the depth of invasion and the extent of LN assessment; their comparison, however, was limited to T3 lesions versus T4 lesions. In our study, for every 1-cm increase in tumor size, there was an average increase of 2% and 3% in the number of LNs examined in colon cancer and rectal cancer specimens, respectively. Furthermore, an increase in average LN recovery was observed for individual T3 and T4 tumors versus T1 tumors (Table 3). However, an analysis of tumor classification progression from T3 to T4 in both colon cancers and rectal cancers revealed that the percentage increase in LN yield actually dropped.

Year of Diagnosis

The year of diagnosis also appears to be a factor in LN harvest (Fig. 1, middle). In the current study, we observed significant time trends associated with LN retrieval: a 2.3% annual increase for colon cancer and a 2.5% annual increase for rectal cancer (P < .01 for both) (Table 3). This finding is consistent with previous reports in the literature.¹⁵ There is a decrease in the mean age of diagnosis over time, as discussed above. This may be the result of widespread public health initiatives that promote CRC awareness or Medicare coverage of various screening modalities beginning in 2001. Furthermore, the finding that younger patients with less comorbidity are diagnosed and undergo standard surgical therapy may account for the larger LN yields reported over time. However, it is interesting to note that the trend has persisted, specifically in patients with rectal cancer, despite the use of combined-modality therapy. This may be caused by several factors. The first factor is the introduction of total mesorectal excision (TME) by Heald in 1979,²² who advocated careful preservation of the integrity of the visceral fascial envelope that contains the rectum along with its vascular supply and surrounding LNs. The second factor is a realization within the medical community of the paramount importance of LN harvest in CRC and its influence on management decisions about adjuvant therapy. This has resulted in a stalwart effort to educate surgeons and pathologists on the extent of mesenteric resection for CRC and extraction techniques for LNs. The third factor is the mounting pressure to meet consensus standards that have been set as quality indicators in cancer care.

Radiation Therapy

Adjuvant radiation therapy is another major factor associated with LN assessment.^20,23–25 In a national United Kingdom study, Tekkis et al observed mean LN harvests of 9.8, 12.1, and 11.8 in patients who received preoperative radiation, postoperative radiation, and no radiation, respectively. In their multivariate regression model, they observed a significant negative correlation in LN harvest when comparing patients who received preoperative radiation versus patients who received no radiation.²⁰ However, only 354 patients received preoperative radiation, 67 received postoperative radiation, and 4683 received no radiation. Furthermore, it is not clear whether radiation was given only to patients with rectal cancer. In our study, the mean number of LNs harvested (±SD) was 10.2 ± 8.8. In considering the effect of radiation, patients with rectal cancer were stratified according to radiation sequence (Fig. 1, bottom). It is noteworthy that we observed a 25% average reduction in LN harvest when comparing the preoperative radiation and no radiation treatment groups (P < .01) (Table 3). Although a higher mean number of LNs evaluated in the postoperative radiation group versus the no radiation group was observed across time, the difference did not reach statistical significance (P = .22).

The results of our study demonstrate that ≥12 LNs were not retrieved from most patients with CRC in the study cohort. This highlights the current controversy within the literature regarding LN yields in CRC resection specimens. By using data from the National Cancer Data Base, Swanson et al examined the relation between the number of LNs and prognosis in patients with T3N0 colon cancer and proposed examining a minimum of 13 LNs to label a T3 cancer as LN-negative.⁶

In a retrospective study investigating the relation between the number of recovered LNs and LN metastases in colorectal resection specimens, Goldstein reported that the predictive probability of identifying a single LN metastasis increased as the number of recovered LNs increased, thus suggesting that a minimum number of retrieved LNs cannot be set to accurately stage all patients.⁴ Other authors, similarly concluding that there is no clear cutoff, have argued that pathologists should harvest as many LNs as possible.^10,20

The current recommendation advocating yields of ≥12 LNs in CRC resection specimens has been established as a standard quality indicator has been questioned by a recent retrospective study of SEER-Medicare–linked data on 30,625 patients that identified no association between the number of LNs examined after colectomy and staging, the receipt of adjuvant chemotherapy, or patient survival.²⁶ Despite these findings, in another study that examined the factors that influenced compliance with CRC staging recommendations, Gagliardi et al identified several barriers to change that potentially are amenable to quality improvement, including a perceived lack of evidence for the recommendation to examine at least 12 LNs, technical issues pertaining to surgeons and pathologists, lack of communication between the 2 disciplines, and poor oversight.²⁷

Limitations and Challenges

The current study has several limitations. Data on the length of surgical specimens were unavailable; therefore, we could not control for the rate of LN recovery according to length. Nor could we control for the inherent variation in LN harvesting technique among pathologists. It is in these 2 areas (extent of mesenteric resection and technique of pathologic dissection) that a push toward standardization should take place to maximally impact LN yields in CRC and improve patient outcome. In surgery for colon cancer, adherence to strict oncologic principles requires en bloc lymphadenectomy of the adjacent lymphatic basin, removal of the involved segment of colon, and ligation of its associated vascular pedicle. In surgery for rectal cancer, the introduction and current widespread practice of TME serves as an example of how standardization impacts outcome. In a retrospective review, MacFarlane et al²⁸ reported the surgical results of TME over a 13-year period for 135 consecutive patients who were classified as high-risk according to the schema used by the North Central Cancer Treatment Group. The local recurrence rate at 5 years was 5% for TME alone (22% overall recurrence), versus 25% for conventional surgery plus radiation therapy (62.7% overall recurrence), and 13.5% for conventional surgery plus chemotherapy plus radiation therapy (41.5% overall recurrence). Moreover, additional reports have confirmed that lower rates of local recurrence, longer survival, and less long-term morbidity accompany higher rates of sphincter and autonomic nerve preservation.^29–31

With regard to pathologic technique, the National Comprehensive Cancer Network recommends that, if <12 LNs initially are identified in the specimen, the pathologist should go back and resubmit more tissue. If 12 LNs still are not identified subsequently, a comment in the report should indicate that an extensive search was performed.³² Currently, manual LN dissection is the standard at most institutions. However, a significant percentage of small LNs (<5 mm in greatest dimension) harbor metastases from CRC. Emergent strategies to identify these LNs have been studied widely. Herrera and Villar-real retrospectively analyzed surgical specimens from 27 consecutive patients with rectal cancer using a clearing technique that identified LNs as small as 1 mm. In that study, 72 of 345 LNs identified in Dukes C tumors had metastases. Of these, 56 metastases (78%) occurred in LNs that measured ≤5 mm. Those authors concluded that the use of LN clearing methods has the potential to enhance staging in rectal cancer.³³ In contrast, Andreola et al retrospectively studied 50 consecutive patients with primary adenocarcinoma of the rectum who underwent sphincter-sparing proctectomy, TME, and coloanal anastomosis. Manual dissection was used to search for meta-static LNs. In that study, 1793 LNs were identified. Of those, 174 contained metastases, and 79 (45.4%) of the involved LNs measured <5 mm in greatest dimension.³⁴ A novel perspective was provided in a recent study by West et al, who pathologically assessed the quality of colon cancer surgery by evaluating the extent of variation in the plane of surgical resection, the amount of tissue removed, and its association with survival. Favorable outcome data from studies evaluating TME for rectal cancer formed the basis of their query regarding colon cancer. Specimens were photographed and graded according to the plane of mesocolic dissection. Tissue morphometry was done on 253 tumors. Univariate and multivariate analyses were performed to ascertain association with 5-year survival. Three hundred ninety-nine specimens were analyzed, and marked variations were noted in the proportion of each plane of surgery (24% muscularis propria, 44% intramesocolic, 32% mesocolic). The mean cross-sectional tissue area outside the muscularis propria and the distance from the muscularis propria to the mesocolic resection margin were significantly higher in mesocolic plane surgery. On univariate analysis, mesocolic plane surgery was associated with a 15% overall survival advantage at 5 years compared with surgery within the muscularis propria plane; the advantage rose to 27% in patients with stage III disease. These findings were not upheld, however, in the multivariate model for all patients. The authors concluded that improving the plane of dissection might improve survival, especially in patients with stage III disease. In addition, given the incorporation of assessment of quality of surgery into the National Cancer Research Institute trial of fluoropyrimidine, oxaliplatin, and targeted receptor preoperative neoadjuvant chemotherapy for colon cancer in advanced resectable colon cancer, consideration should be given to routine pathologic auditing as a cost-effective way to rapidly improve outcomes in patients with stage III colon cancer. It is noteworthy that no association was observed between the number of LNs retrieved by the pathologist and the plane of surgery achieved; thus, that study offers a potential foundation for future standardized pathologic analysis in CRC.³⁵

In addition, because SEER-Medicare–linked data were not used, hospital information was unavailable, and we could not control for hospital or surgeon volume in our analyses. Also, because of our large study cohort of more than 100,000 patients, our analyses were over-powered to identify a factor as significant. Therefore, statistical inference based on this large dataset should not rely heavily on the P values.

Despite these limitations, the strengths of our study include the use of SEER data, which are audited extensively for accuracy and completeness and have a high level of quality control. In addition, this study reflects community standards and practices in the United States. Finally, we studied actual LN yields and the associated factors that had an impact on retrieval as opposed to modeling the probability of harvesting at least 12 LNs based on a single cutoff value.

In conclusion, the current results confirmed that LN retrieval from CRC surgical specimens is the result of interplay between patient demographic and clinicopathologic factors. Patient factors like age, sex, and ethnicity and tumor factors like location, grade, and tumor size cannot be changed to improve LN yields. However, other factors, such as the extent of mesenteric resection and pathologic technique, can be changed to improve LN yields, thus improving the accuracy of surgical staging, appropriate administration of adjuvant therapy, and overall prognosis. Despite the increases observed over time, LN yield in rectal cancer remains a challenge, as does LN yield in older patients who have early distal tumors. Tumors with these characteristics are less likely to be amenable to the benchmark of 12 harvested LNs. In view of these findings, further long-term studies using externally validated predictive models of LN yield are warranted to improve staging. Continuing emphasis should be placed on refining surgical and pathologic practice to increase overall LN yields.