Role of lymph node yield and lymph node ratio in predicting outcomes in non‐metastatic colorectal cancer

Background Lymph node yield (LNY) of 12 or more in resection of colorectal cancer is recommended in current international guidelines. Although a low LNY (less than 12) is associated with poorer outcome in some studies, its prognostic value is unclear in patients with early‐stage colorectal or rectal cancer with a complete pathological response following neoadjuvant therapy. Lymph node ratio (LNR), which reflects the proportion of positive to total nodes obtained, may be more accurate in predicting outcome in stage III colorectal cancer. This study aimed to identify factors correlating with LNY and evaluate the prognostic role of LNY and LNR in colorectal cancer. Methods An observational study was performed on patients with colorectal cancer treated at three hospitals in Melbourne, Australia, from January 2010 to March 2016. Association of LNY and LNR with clinical variables was analysed using linear regression. Disease‐free (DFS) and overall (OS) survival were investigated with Cox regression and Kaplan–Meier survival analyses. Results Some 1585 resections were analysed. Median follow‐up was 27·1 (range 0·1–71) months. Median LNY was 16 (range 0–86), and was lower for rectal cancers, decreased with increasing age, and increased with increasing stage. High LNY (12 or more) was associated with better DFS in colorectal cancer. Subgroup analysis indicated that low LNY was associated with poorer DFS and OS in stage III colonic cancer, but had no effect on DFS and OS in rectal cancer (stages I–III). Higher LNR was predictive of poorer DFS and OS. Conclusion Low LNY (less than 12) was predictive of poor DFS in stage III colonic cancer, but was not a factor for stage I or II colonic disease or any rectal cancer. LNR was a predictive factor in DFS and OS in stage III colonic cancer, but influenced DFS only in rectal cancer.


Introduction
Oncological resection of colorectal cancer involves removing the tumour with clear margins and harvesting of draining lymph nodes. Adequate lymph node harvest allows accurate staging and minimizes the risk of understaging nodal positive disease. The AJCC/UICC recommend a minimum of 12 lymph nodes should be identified in colorectal cancer specimens 1 -4 . Although lymph node yield (LNY) has been used as a surrogate indicator of adequate surgical resection, a substantial proportion of resections  per cent) still fall into the low LNY category based on population studies 5 -7 . Low LNY has been associated with poorer survival outcome in stage II and III colorectal cancer 7 -10 . A systematic review 11 concluded that increased LNY was associated with improved survival in stage II-III colonic cancer. LNY was also shown in a large population study to be an independent prognostic factor in node-negative colorectal cancer irrespective of T category 12 . This study 12 concluded that stage migration alone contributed to the observed survival difference. Despite a rise in median LNY over recent years, the    Values in parentheses are 95 per cent confidence intervals. *Linear regression using lymph node yield (less than 12 or 12 or more) as a dichotomous variable. pAJCC, pathological stage according to the AJCC; CRT, chemoradiotherapy.
percentage of stage III colorectal cancers has not increased proportionately 9,13 , and at 20 years the Surveillance, Epidemiology, and End Results (SEER) data showed no correlation between the proportion of positive lymph nodes and LNY despite an increase in LNY, suggesting that stage migration alone may not account for improved survival in patients with colorectal cancer 14 .
The prognostic significance of low LNY (less than 12) is debatable in locally advanced rectal cancer treated with neoadjuvant therapy, as LNY is known to be reduced in this setting 14 -19 . Several studies 20 -22 have found no survival difference between low and high LNY groups in this setting. Some studies 16,23 -25 have suggested that lymph node ratio (LNR) should be used for prognostication rather than LNY in stage III colorectal cancer, as it is more representative of tumour burden.
This study aimed to examine the correlation between clinical variables and LNY, and to examine the prognostic significance of LNY and LNR in stage 0 (complete pathological response, pCR) and stage I-III colorectal cancer. Oncological resection involved en bloc resection of tumour with clear margins and high ligation of vascular pedicles, which ensured adequate lymphadenectomy. Neoadjuvant therapy, either long-course chemoradiotherapy (nCRT) or short-course radiotherapy, was routinely offered to patients with locally advanced rectal cancer (T3-4 or node-positive disease). Patients with pCR after neoadjuvant therapy were included in the study. pCR was defined by the absence of residual tumour cells in the surgical specimen, as described previously 27 . Patients who presented with distant metastasis and/or synchronous colorectal cancer were excluded.
Management decisions were based on multidisciplinary team meetings held before surgery. Resected specimens were examined by the pathology department at the hospital in which the resection took place (the small number of specimens from Avenue Hospital were outsourced to a pathology service). Tumours were staged according to AJCC guidelines (7th edition) 1   as fewer than 12 nodes in the resected specimen. If the initial LNY was less than 12, a fat clearance technique using Carnoy's solution was performed routinely in two of the three sites to increase the yield. LNR was defined as the ratio of metastatic lymph nodes to the total number of lymph nodes examined. LNR was classified into four tiers based on rectal cancer data published by Danish Colorectal Cancer Group 5 : LNR 1, less than 0⋅08; LNR 2, 0⋅08 to less than 0⋅25; LNR 3, 0⋅25 to less than 0⋅50; LNR 4, 0⋅5-1⋅0.
All patients had routine follow-up every 3-6 months for the first 2 years with serial measurement of carcinoembryonic antigen. CT of the chest, abdomen and pelvis was performed annually in addition to colonoscopy as deemed clinically appropriate. After 2 years, follow-up was usually annual, with CT, estimation of blood tumour markers and colonoscopy for metachronous disease performed at the discretion of the treating surgeon and in line with national guidelines. In data analyses, follow-up was defined as the  time from the date of primary surgery to a patient event, such as disease recurrence or death. There was no minimum duration of follow-up. Follow-up information was derived from the colorectal neoplasia database and patient hospital records. After surgery for the primary tumour, patients who developed local recurrence or metastasis were no longer considered disease-free for statistical analysis.
Patients who died at any time, for any reason, after surgery were counted as deaths in overall survival analysis.

Statistical analysis
Data analysis was performed using Stata ® 14 (StataCorp, College Station, Texas, USA). Linear regression (ordinary least squares) was used to analyse the association of LNY and LNR with clinical variables (histological grade, overall stage, tumour site and use of neoadjuvant therapy if rectal cancer). The effects of LNY and LNR on disease-free (DFS) and overall (OS) survival Values in parentheses are 95 per cent confidence intervals. LNRQ, lymph node ratio quartile.
were investigated using survival analysis techniques (Cox regression, Kaplan-Meier survival analysis and log rank tests). Independent prognostic factors were identified in both univariable and multivariable analyses (Cox regression). χ 2 tests were used in additional analyses between groups. The significance level was set at 5 per cent, and terms were included in the models when the P value was below this level. Robust standard error estimates were obtained to account for the lack of independence between observations for patients with more than one procedure. P < 0⋅050 (two-tailed) was considered statistically significant.  Table 1.
In multivariable analysis using linear regression, LNY below 12 was associated with increasing age, open surgery, robotic surgery (performed by 5 surgeons in 2012-2016) and elective surgery ( Table 3). High LNY (12 or above) was associated with stage I-III versus stage 0 tumours. Patients with rectal cancer who had received neoadjuvant therapy showed a higher LNY for stage III pAJCC cancers alone and when in receipt of short-course radiotherapy ( Table 3).
Of the 1320 (83⋅9 per cent) of 1573 patients with colorectal cancer for whom follow-up data were available, 130 (9⋅8 per cent) had disease progression within 5 years. Univariable survival analysis showed no statistical difference in DFS between low LNY (less than 12) and high LNY (12 or more) (  5 Kaplan-Meier analysis of disease-free survival in patients with stage III a colonic and b rectal cancer according to lymph node ratio quartile: Q1, 0 to less than 0⋅0825; Q2, 0⋅0825 to less than 0⋅25; Q3, 0⋅25 to less than 0⋅5; Q4, 0⋅5-1. a P < 0⋅001, b P = 0⋅011 (log rank test) when 12 or more lymph nodes were collected ( Table 4). A higher risk of progression was also evident for stage II or III compared with stage I tumours. No patient with stage 0 (pCR) progressed. After identifying an association between rectal cancer and low LNY (below 12), colonic and rectal cancer subdivisions were analysed further. DFS for colonic cancer was worse only for stage III tumours when data were stratified by low versus high LNY: stage I, P = 0⋅355; stage II, P = 0⋅751; stage III, P = 0⋅002 (log rank test) (Fig. 1). In contrast, there were no differences in DFS for rectal cancer when different stages were compared: stage 1, P = 0⋅493; stage II, P = 0⋅912; stage III, P = 0⋅058 (log rank test) (Fig. 2). OS for colonic cancer was worse only for stage III tumours when data were stratified by low LNY versus high LNY: stage I, P = 0⋅697; stage II, P = 0⋅881; stage III, P = 0⋅004 (log rank test) (Fig. 3). There were no differences in OS for rectal cancer when different stages were compared: stage I, P = 0⋅448; stage II, P = 0⋅469; stage III, P = 0⋅951 (log rank test) (Fig. 4).

Role of lymph node ratio
Higher LNR was associated with poorer DFS in multivariable analyses of colorectal cancer ( Table 5). Risk of disease progression increased with increasing LNR quartile (0⋅25 and above). When colonic and rectal cancer were considered separately, LNR was an independent predictor of DFS in stage III colonic cancer, with patients having poorer DFS with LNRs greater than 0⋅25 ( Table 6). LNR of 0⋅5 or above was also a risk factor for patients with stage III rectal cancer ( Table 6 and Fig. 5b).

Discussion
LNY is an important prognostic factor in patients with stage II-III colorectal cancer, but evidence for stage I colorectal cancer is unclear 8,11,28 -31 . Although there are studies reporting differences in survival between low and high LNY in patients with Dukes' A/stage I colorectal cancer 10,15 , some indicate that LNY alone has no effect on survival outcome 32,33 . The present study highlights that a substantial number of patients with stage I and II colorectal cancer have a low LNY. Nonetheless, both DFS and OS in stage I and II colonic and rectal cancers in the present cohort were excellent and not influenced by LNY. Low LNY in stage III colon cancer did, however, result in poorer DFS and OS. Low LNY had no effect on DFS or OS for rectal cancer.
LNR is generally considered more reliable as a prognostic tool in stage III colorectal cancer 5,16,23,34 . A systematic review and meta-analysis 25 involving 33 984 patients with stage III colorectal cancer described numerous different cut-off LNR values. The authors concluded LNR was a more reliable prognostic factor than total number of positive nodes 25 . In the present study, a higher LNR resulted in worse DFS and OS in colorectal cancer. In addition, a high LNR correlated with a poorer DFS for both stage III colonic and rectal cancers. In stage III colonic cancer a high LNR also correlated with a poorer OS, whereas there was no correlation between LNR and OS in stage III rectal cancer. A high LNR in rectal cancer following neoadjuvant therapy resulted in worse DFS.
Neoadjuvant therapy is the current standard of care for locally advanced rectal cancer (T3-4 or node-positive disease). A reduction in LNY secondary to radiotherapy is well recognized and the prognostic role of LNY in this setting is debatable. Miller and colleagues 18 examined seven cohort studies and found a significant reduction in mean LNY in the nCRT group compared with that in patients who did not undergo nCRT (range 7-53 per cent). Despite some studies 35,36 showing that low LNY in rectal cancer treated with neoadjuvant therapy may be associated with poorer survival, there is emerging evidence 20,21,37 that low LNY in this clinical setting may not necessarily confer a poorer prognosis. A systematic review 38 of 11 studies showed that although preoperative radiotherapy for locally advanced rectal cancer was associated with a lower LNY, most studies did not identify any adverse outcome in the low LNY group. This is consistent with the present findings in the subgroup of patients with locally advanced rectal cancer treated with nCRT.
In the present study, 23⋅5 per cent of all colorectal cancer resections were below the current benchmark of a minimum harvest of 12 lymph nodes; this is comparable to contemporary data from specialist centres 22,39,40 , but lower than that from other population-based studies 6,7,41 . The majority of the variation in reported LNY is likely to be due to non-modifiable patient-specific factors, including tumour characteristics 6 . Factors such as older age, early tumour stage, type of operation (left-sided/rectal operation compared with right-sided) and use of neoadjuvant therapy in locally advanced rectal cancer have all been shown to be associated with low LNY 7,28,40,42 -44 . This was reflected in the present study, where lower LNY rates were associated with rectal cancer, elective surgery, stage 0 and I tumours, older patient age, left-sided surgery and neoadjuvant treatment. Studies 43,45 have shown that operations performed by colorectal specialists in high-volume centres are more likely to have a higher LNY than those performed by non-specialists. It is possible that both surgeon and pathologist factors contributed to the LNY observed in the present study. In addition, fat clearance techniques can enhance lymph node assessment 46,47 .
The present study has a number of limitations. Its observational nature is mitigated by the existence of a specific database for the collection of patient information.
Although it is possible that there may have been differences in the way resection specimens were processed, the pathology departments followed the same AJCC guidelines. There was no statistical difference between the three hospitals in terms of LNY. A proportion of patients were lost to follow-up -around 16⋅1 per cent over the study interval of 6 years. It is not known whether DFS and OS data would be different with full patient follow-up (1573) in contrast to the 1320 patients for whom follow-up data were available.
The principle of oncological resection is to create clear margins. Together with high ligation of the vascular pedicle, adequate lymph nodes can be harvested in the corresponding drainage basin. This allows accurate pathological staging of colorectal cancer that guides adjuvant chemotherapy. Nonetheless, despite following this principle, low LNY can be associated with earlier-stage colorectal cancer as demonstrated in the present study. Low LNY in these clinical settings did not confer a poorer prognosis provided an adequate oncological resection and thorough examination of the surgical specimen had been performed.
These results support the use of LNR as a prognostic tool in stage III colorectal cancer to inform surveillance and treatment discussions (for instance with regard to adjuvant chemotherapy) following primary surgery. In addition, the use of LNR rather than LNY seems more logical in patients with rectal cancer treated with neoadjuvant therapy, given the tendency for a significant reduction in LNY.