Lymph node yield in right colectomy for cancer: a comparison of open, laparoscopic and robotic approaches

Introduction

Lymph node (LN) status is the single most important predictor of survival in nonmetastatic colorectal cancer 1, 2. Multiple studies have demonstrated a relationship between the number of LNs retrieved during colectomy and survival 3-9. Although the American Joint Committee on Cancer (AJCC) recommends a minimum of 12 nodes for adequate staging, an optimal cutoff has yet to be established 5, 10, 11. The Intergroup 0089 trial showed that LN yields of more than 20 and 35 were associated with better disease-specific survival in node-negative and node-positive patients, respectively 5. Similarly, the INTACC trial suggested an optimal LN cutoff of 18, although the survival difference was only seen in node-negative patients 12. Using data from the SEER database, Cserni et al. 13 found that LN yields of up to 40 may be associated with the greatest probability of survival; they showed that overall survival increases linearly with each additional retrieved negative node between 1 and 40.

Multiple factors, including surgical technique, may affect LN yield in resections for colon cancer 9, 10, 14-16. To date, studies comparing robotic, laparoscopic and open colectomies have reported conflicting results regarding LN yield 17-20. As the paradigm for oncological resections shifts towards increasing LN yield, it is important to assess the role of surgical technique in optimizing LN retrieval. The purpose of this study, therefore, was to compare the pathological outcomes after open, laparoscopic and robotic right colectomy (O-RC, L-RC and R-RC, respectively) for cancer. We hypothesized that with its increased dexterity and visualization, R-RC would lead to a higher LN yield and improved mesocolic excision.

Method

All patients who had RC for cancer between 1 January 2012 and 31 December 2014 were identified in the institutional database by CPT and ICD-O codes. Patients were included if they had undergone surgery for a colonic neoplasm and had been operated on by one of six specialist colorectal surgeons. Exclusion criteria were: age < 18 years, previous colectomy for cancer and pathology other than adenocarcinoma or carcinoid. Patients undergoing additional operative procedures during the index surgery were included as long as these did not involve other intestinal anastomoses. This study was approved by the Institutional Review Board of Memorial Sloan Kettering Cancer Center.

The electronic medical record was reviewed by two members of the research team for clinical, demographic, operative and pathological details. Cases were categorized as O-RC, L-RC or R-RC resections on an intention-to-treat basis, according to the CPT code listed for the surgery. In L-RC and R-RC cases, enlarging the extraction site incision for any purpose besides extraction of the specimen was considered a conversion to O-RC. Some surgeons, by convention, mobilized the colon using a minimally invasive approach and then exteriorized the segment for the purposes of vascular ligation, resection and anastomosis; these cases were not considered conversions to O-RC. Resections were considered ‘extended RC’ if the surgeon designated them as such. Information about the ligated vascular pedicle was also gathered from the operative report.

Pathological evaluation of the resected specimen was performed according to standard procedure. Specifically, specimens are sent to pathology fresh right after removal from the patient. The pathologist then opens the bowel, takes tissue samples for tumour procurement purposes, pins the specimen to a supportive vehicle and places the specimen in formalin right away. The fixation time is usually 6 h to overnight. All attached pericolic soft tissues are thoroughly examined for LNs with possible involvement and all such LNs are submitted for microscopic examination. Gross pathological examination of specimens is performed by trained gross room assistants. During the study period, the same group of assistants handled all the cases. In our practice we do not routinely use fat-resolving agents for LN dissection. Microscopic examination of all colorectal resection specimens was performed by specialist gastrointestinal pathologists at our institution. The differentiation of positive LNs from tumour deposits followed current recommendations 21. The surgical pathology report was reviewed for details including total specimen length, LNs retrieved, location of tumour from the proximal edge of the specimen and AJCC (seventh edition) tumour stage.

Univariable analyses using analysis of variance (ANOVA), chi-square and nonparametric alternatives were used to compare characteristics and outcomes between O-RC, L-RC and R-RC cohorts. Tukey's range test was used for multiple comparisons between the techniques. There were two primary end-points for this study: total LN yield and the LN to length of surgical specimen (LN-LSS) ratio.

Multivariable modelling using a mixed effects model was used to predict LN yield and LN-LSS ratio. Model selection began with stepwise selection of relevant clinical, demographic and operative variables using sas version 9.4 (SAS Institute, Cary, North Carolina, USA). Several variables posed a high likelihood of collinearity, including tumour distance from the proximal resection margin, extended vs standard RC, ligated vessels and specimen length. Correlation matrices were used to choose between variables that could lead to collinearity, and to exclude potentially intermediate variables, whose inclusion would lead to over-adjustment. The final model consisted of both those variables that had been selected by stepwise regression and those determined to be relevant by a priori knowledge. Surgeons and pathologists were considered random effects in order to control for differences in technique and skill. Variance inflation factor and residual analyses were used to verify the robustness of the model. Analyses were conducted on an ‘intention-to-treat’ basis with a P-value of < 0.05 being considered statistically significant. A sensitivity analysis excluding Stage IV disease was performed. An ‘as-treated’ analysis was also conducted in which L-RC and R-RC cases that were converted to O-RC were re-categorized as O-RC cases.

Results

A total of 181 O-RC, 163 L-RC and 119 R-RC cases met the inclusion criteria. Demographics and tumour characteristics are listed in Table 1. Patients in these cohorts differed in terms of gender and AJCC stage, with patients in the open group more likely to be female and to have Stage IV disease compared with L-RC and R-RC patients. Mismatch repair (MMR) status by immunohistochemistry was determined in 294 patients. The proportion of patients with MMR-deficiency was not significantly different amongst the study groups.

There were significant differences in operative details between groups (Table 2). Patients undergoing O-RC were more likely to have had concurrent major procedures (41% vs 12% in L-RC and 5% in R-RC), most of them liver resections. Compared with R-RC, patients undergoing L-RC were more likely to have conversions to O-RC, and accordingly, to have ligation of the vascular pedicle done in an extracorporeal fashion after termination of laparoscopy. R-RC patients had almost exclusively intracorporeal vascular pedicle ligations and were the most likely to have undergone preoperative tattooing of the tumour.

The colectomies were performed by six colorectal surgeons. The frequencies of O-RC, L-RC and R-RC cases performed by each surgeon are listed in Table 3. Five of the surgeons performed each of the three approaches at least once. Two of the surgeons accounted for nearly 90% of R-RCs performed. The rates of conversion to O-RC ranged from 0 to 47% among the six surgeons.

Postoperative complications are listed in Table 4. Overall, 106 patients (23%) experienced a postoperative complication, with Clavien–Dindo Grade III or higher-grade complications occurring in 34 (7%) patients. The rate of superficial and deep surgical-site infections was 8%, and the rate was significantly higher for O-RC patients than for L-RC or R-RC patients (P < 0.01). The overall rate of anastomotic leaks and intra-abdominal abscesses was 6%, and again the rate was significantly higher for O-RC patients than for L-RC or R-RC patients (P < 0.01).

Pathological outcomes are shown in Table 5. The LN yield was highest among patients undergoing R-RC with a mean of 34, compared with the lowest LN yield in patients undergoing O-RC, in which a mean of 28 LN were recovered. Univariable analysis showed significant differences in LN yield between the cohorts (P < 0.01). Tukey's range test was used for multiple comparisons, and showed that R-RC resulted in a significantly greater LN yield than L-RC and O-RC (both P < 0.01). Robotic RC resulted in the highest LN-LSS ratio, approaching 1. L-RC cases produced a mean ratio of 0.91, and O-RC had the lowest ratios, with a mean of 0.83. These differences were significant on univariable analysis, but in multiple comparison analysis only the R-RC vs O-RC comparison was statistically significant (P < 0.01).

In multivariable analysis, R-RC was associated with a higher LN yield, with approximately four more nodes harvested compared with O-RC and L-RC (Table 6). These findings were even more pronounced in sensitivity analyses excluding Stage IV disease. Location of the tumour (greater distance from the proximal margin) and increased tumour size were also associated with increased LN yield, whereas older age was associated with decreased LN yield. A preoperative tattoo mark was associated with a higher LN yield and concurrent surgical procedures were associated with lower LN yield, though these findings were not significant after exclusion of Stage IV cancers.

Compared with R-RC, O-RC was associated with a significantly lower LN-LSS ratio in multivariable analyses, while L-RC showed a lower estimate, but this difference was not significant (Table 7). A preoperative tattoo mark increased the ratio by 0.09. Older age, increased tumour distance from the proximal resection margin and higher body mass index (BMI) were associated with a lower LN-LSS ratio.

An additional analysis was performed with O-RC as the reference group. In this analysis, L-RC and O-RC were associated with similar LN yields (P = 0.99), while both L-RC and R-RC cases were associated with significantly increased LN-LSS ratios (P = 0.03 and < 0.01 for L-RC and R-RC, respectively). For both outcomes, we performed sensitivity analyses in which minimally invasive cases that were converted to open surgery were reclassified as O-RC. The results of these sensitivity analyses did not alter the findings of the intention-to-treat analyses.

Discussion and conclusions

The findings of this study indicate that R-RC for colon cancer is associated with higher LN yield than both O-RC and L-RC. In addition, we found that minimally invasive colectomy is associated with higher LN-LSS ratios, with the highest ratio seen in R-RC cases.

Our study confirms previous observations from our own institution; preoperative tattooing is associated with a higher LN yield, but MMR status is not associated with LN yield 15, 22. Increasing age was associated with a decrease in both LN yield and LN-LSS ratio, while increasing tumour size was associated with an increase in LN yield only. The significance of these findings is unclear given the small magnitude of the associations between these factors and the outcomes of interest.

Lymph node yield can be increased either by harvesting more paracolic nodes through a longer surgical specimen or by increasing the uptake of intermediate and central nodes through higher ligation of the vascular pedicle, or both. Therefore, LN yield alone cannot be considered a surrogate of complete mesocolic excision (CME) with central vascular ligation (CVL). By dividing the number of LNs retrieved by the length of the surgical specimen, the LN-LSS ratio may provide an indirect indication of the completeness of the mesocolic excision. In the case of a longer specimen without a higher vascular ligation, we would expect little change in the LN-LSS ratio since the specimen length would be increasing concurrently with LN yield. However, a resection with a higher ligation of the vascular pedicle would be expected to increase this ratio, assuming an equivalent length of resection.

Few studies to date have reported LN yield after R-RC. A recent systematic review of the six largest studies comparing R-RC and L-RC included only two studies with pathological data, and in only one could the LN yield after RC be readily abstracted 23. A prospective trial from Korea comparing 35 patients having R-RC with the same number of patients having L-RC reported equivalent LN yields, namely 29.9 and 30.8, respectively 24. As the length of the surgical specimen was longer in the R-RC group, the LN-LSS ratio is likely to have been lower in the R-RC group as well. This discrepancy with our study cannot be simply attributed to differences in patient population, because race and BMI have not been previously shown to be independently associated with LN yield. It is possible that the differences may be attributable to differences in surgical technique, in both R-RC and L-RC. For example, in the Korean study, the mesenteric dissection was performed from lateral to medial, while in our centre, the operation typically starts with central vascular control followed by a medial to lateral approach.

While the absolute number of LNs may be associated with a survival advantage, recent evidence suggests that the completeness of the mesocolic excision also has prognostic implications 25, 26. A CME has been associated with a larger LN yield by removing the centrally located nodes close to the named vascular pedicles. Institutional case series, and more recently a population-based study, have found an association between complete CME with CVL and improved survival in patients with colon cancer 27-29. In a Danish study, this survival benefit was associated with higher rates of intra-abdominal organ injuries and nonsurgical complications compared with conventional colon resections 30. In our study, the overall rate of complications was comparable to those in the two arms of the Danish study and no intra-abdominal organ injuries or superior mesenteric vein injuries occurred.

An important limitation of this study is its retrospective design. Measuring the length of the vascular pedicle at the time of surgery could have provided a separate method for assessing central vascular control. It is likely that only examination of the surgical specimen could establish which nodal groups (paracolic, intermediate, apical) were excised. In addition, multiple surgeons, pathologists and pathology assistants were involved in this study. As in all studies of this nature, such confounders are difficult to eliminate. However, by using a mixed-effects model, with surgeon and pathologist as random variables, we accounted for the potential variability introduced by differences in technical proficiency.

In conclusion, R-RC is associated with a higher LN yield than O-RC and L-RC, and a higher LN-LSS ratio than O-RC. Our study suggests that the improved visualization and dexterity provided by the robotic platform may facilitate the performance of a CME and CVL. These findings should be validated in a prospective study with confirmation of the height of the vascular pedicle ligation.

Conflict of interest

None of the authors had any conflict of interest.