Laparoscopic surgery facilitates administration of adjuvant chemotherapy in locally advanced colon cancer: propensity score analyses

Synopsis

In this multicenter propensity score analyses of 424 pT4N0-2M0 colon cancer patients who underwent curative intent surgery, laparoscopic surgery compared to open was associated with reduced postoperative morbidity and a higher percentage of patients receiving adjuvant chemotherapy, which translated into improved survival.

Introduction

Resection of locally advanced (T4) colon cancer stage is still regarded as a relative contraindication for laparoscopic surgery, because radicality of the resection (R0) might be jeopardized, thereby impacting the long-term oncological outcome.¹

Laparoscopic surgery for colon cancer has become widely accepted and implemented in routine practice for localized disease (pT1-3). Both favorable short-term and long-term benefits have been demonstrated,^2–5 including less postoperative complications, shorter hospital stay, lower risks of adhesion-related small bowel obstruction, and incisional hernia compared to open surgery.⁶ In the multicenter randomized COLOR trial,⁷ the noninferiority of laparoscopic surgery for pT1-4 colon cancer in terms of oncological outcomes (3-year disease-free survival, DFS) was suggested. However, clinically suspected tumor invasion of adjacent structures (cT4b stage) and emergency surgery were exclusion criteria. Due to a conversion rate of about 50% for the remaining T4 tumors in the COLOR data, the authors presumed the open approach to be most appropriate for T4 colon cancer.

Experience with laparoscopic surgery for T4 tumors increases, and the concerns on achieving radicality of resection may be outdated. Accumulating series of laparoscopic resections of T4 colon cancer are published,^8–18 and two recent meta-analyses^19,20 show comparable rates of R0 resections and long-term oncological outcomes after laparoscopic versus open surgery. However, only a limited body of evidence is available and is restricted to retrospective series with substantial allocation bias.^19,20 Also, emergency cases and multivisceral resections (MVR, pT4b) were often excluded in the previous series.

The aim of this multicenter cohort study was to evaluate long-term oncological outcomes after curative intent laparoscopic surgery for pT4 colon cancer and compare it to open surgery using propensity score analyses in order to correct for allocation bias.

Methods

Patients and databases

In this cohort study, four prospectively maintained databases of the University Hospital Leuven, the Dutch (teaching) St. Antonius Hospital, and two Dutch university medical centers (Radboud UMC and Amsterdam UMC) were combined (centers 1–4). The database of center 1 contained all consecutive patients who underwent resection of colon cancer between January 2004 and July 2013. The databases of centers 2 and 3 included all pT4 colorectal cancer patients undergoing surgery between January 2000 and December 2013, and between January 2000 and December 2007, respectively. The database of center 4 included all pT4 colon cancer patients who underwent surgery between January 2004 and December 2014. Patients undergoing a curative intent pT4N0-2M0 primary colon cancer resection were included. Patients with a macroscopic incomplete (R2) resection or with an inadequate pathological or surgical report were excluded. Adjuvant chemotherapy was administered according to the national protocols: indicated in high-risk stage II and stage III, consisting of 5-flourouracil or capecitabine and preferably combined with oxaliplatin (FOLFOX or CAPOX) since 2005, for a total duration of 6 months. Follow-up was performed according to the national protocol and data were collected until April 2018.^21,22 This study was waived from the review of the medical ethics boards, since the prospective data collection did not interfere with the psychological integrity of the patients. The study is reported in accordance with the STROBE checklist.

Variables and outcomes

The pT4 stage was subdivided into pT4a and pT4b according to TNM7:²³ pT4a refers to tumors perforating the visceral peritoneum and pT4b refers to tumors directly invading other organs or structures. MVR could be either limited or extensive: limited additional resection included the abdominal wall, the omentum, or the ovaries, and extensive resection was defined as resections including the pancreas, spleen, kidney, liver, stomach, bladder, ureters, uterus, or additional bowel segments.

The primary outcome measures were DFS and overall survival (OS). Secondary outcomes were radicality of resection, classified as R0 (radical resection) and R1 (microscopically irradical resection) resections, administration of adjuvant chemotherapy, and postoperative complications (<30 days or in hospital). Postoperative complications were registered if the Clavien–Dindo²⁴ (CD) score was 2 or higher. Postoperative morbidity was defined as CD 2–4, and postoperative mortality as CD5.

Statistical analyses

The laparoscopic converted procedures were included in the laparoscopic group according to “intention-to-treat principle”. For each patient, a propensity score was calculated, displaying the probability of receiving laparoscopic surgery based on a multivariable logistic regression model including (potential) confounding baseline variables. Potential confounders were variables only affecting outcome measures, and true confounders were variables affecting both choice for surgical approach and outcome measures.²⁵ Confounders included age, emergency setting, tumor location (right/left), MVR (none/limited/extended), N-stage, and histopathology. The choice for variables included in the model was based on identified (potential) confounders in the previous series.^8,9,11,19 Due to multicollinearity between MVR and T4 subcategory (T4a/T4b), the latter was not included in the propensity score calculation.

The propensity score was used in three different manners to adjust for group differences, all with distinct advantages and disadvantages.^25,26 Firstly, the propensity score was used as a linear predictor in Cox regression analyses (referred to as “Regression adjusted”), from which an estimated effect was derived. In this way, the entire cohort could be included in the analysis. In order to minimize baseline differences before statistical comparison, the propensity score was used for 1:1 matching with a caliper width of 0.02 and using matching without replacement (referred to as “Matched cohort”). Besides comparison of the two matched cohorts, the excluded cohort of open surgery after matching was analyzed and compared to the included matched open cohort. Thirdly, the propensity score was used to subdivide patients into four strata based on quartiles of the propensity scores, creating an equal distribution of confounding variables within the four strata (referred to as ‘Strata 1–4ʹ). For each stratum, baseline and outcome variables were compared between laparoscopic and open subgroups. This method uses the entire cohort and creates clinically relevant subgroups from a surgical perspective. In addition, sensitivity analyses were performed for MVR and emergency surgery.

Differences in baseline characteristics between subgroups were assessed using a chi-square test for categorical variables or a Fisher’s exact test in case of low counts (<5). For normally distributed continuous variables, mean and SD were reported. The survival analyses were performed using Kaplan–Meier and Cox regression analyses. A p-value of <0.05 was considered statistically significant in the multivariable analyses. For statistical analyses, PASW Statistics, version 24 (SPSS inc., Chicago, IL) was used.

Results

Patients

A total of 424 patients that underwent macroscopic complete (R0/R1) resection of primary pT4N0-2M0 colon cancer in one of the four participating centers were included (Figure S1). Mean age was 69 years (SD 12) and 51% of patients were male. Laparoscopic surgery was performed in 131 patients (31%), of which 33 procedures were converted (25%). A MVR was performed in 171 patients (40%) and emergency surgery in 60 patients (14%). Median follow-up was 48 months (IQR 22–60). Baseline characteristics are displayed in Table 1. The following baseline characteristics were significantly different for the laparoscopic and open groups, respectively: emergency setting (5% vs 18%, p<0.001), MVR (31% vs 45%, p=0.014), and T4 subcategory (T4a: 81% vs 66%, p=0.002).

Table 1 Baseline characteristics of the entire cohort for the laparoscopic and the open group

Numbers of performed resections between the centers were 164 (39%) in center 1, 113 (27%) in center 2, 90 (21%) in center 3, and 57 (13%) in center 4 with significantly different laparoscopy rates (42%, 24%, 3%, and 56%, respectively; p<0.001) and conversion rates (38%, 11%, 67%, and 6%, respectively; p=0.001). No differences in case mix were found between centers 2, 3, and 4, while center 1 showed higher rates of elective surgery and lower rates of MVRs and pT4b tumors (Table S1).

Unadjusted analysis of outcome measures

The R0 resection rate was higher in the laparoscopic group as compared to the open group (99% vs 93%, p=0.010). The postoperative morbidity rate was lower after laparoscopic surgery (27% vs 46%, p<0.001) and postoperative mortality rates were comparable (2% vs 3%, p=0.357). Adjuvant chemotherapy was more often administered after laparoscopic surgery (60% vs 44%, p=0.004). Laparoscopy was significantly associated with better 5-year DFS (52% vs 38%, p=0.008) and 5-year OS (68% vs 57%, p=0.023) (Tables 2 and S2, Figures 1 and 2).

Table 2 Short-term oncological outcomes after laparoscopic and open surgery

Figure 1 Propensity score distribution, disease-free survival (DFS), and overall survival (OS) in the entire cohort and the propensity matched cohort.Abbreviations: Laparoscopic, laparoscopic approach; Open, open approach; DFS, disease-free survival; OS, overall survival; cum, cumulative; open (excl), open approach, excluded from the matched cohort; open (incl), open approach, included from the matched cohort.

Figure 2 Disease-free (DFS) and overall survival (OS). Abbreviations: Lap, laparoscopic approach; MVR, multiviscersal resection; Open, open approach; unadjusted, the unadjusted (entire) cohort; DFS, disease-free survival; OS, overall survival; HR, hazard ratio; CI, confidence interval.

Regression adjusted analysis

The distributions of propensity scores are displayed in Figure 1. After adjusting for the propensity score in regression models, R0 resection rate (OR 0.16; 95% CI 0.02–1.25) as well as postoperative mortality (OR 0.73; 95%CI 0.14–3.68) were comparable between the two approaches, while postoperative morbidity was lower (OR 0.48; 95% CI 0.30–0.77) and the rate of administration of adjuvant chemotherapy higher (OR 1.94; 95% CI 1.25–3.03) after laparoscopic surgery. Five-year DFS was significantly better in the laparoscopic group (HR 0.70; 95% CI 0.52–0.94), without significant benefit for 5-year OS (HR 0.73; 95%CI 0.50–1.06) (Figure 2).

Propensity matched analysis

In Figure 1, the distributions of propensity scores are displayed for the propensity matched cohort. Within the propensity matched cohort (n=262), no baseline differences were found between the laparoscopic and open groups (Table 3), apart from the center. The excluded open group (n=162) after 1:1 matching contained more emergency cases as compared to the included open group (29% vs 5%, p<0.001) as well as more MVRs (57% vs 29%, p<0.001) and more pT4b tumors (44% vs 21%, p<0.001).

Table 3 Baseline characteristics of the matched cohort for the laparoscopic and the open group

Laparoscopic surgery as compared to open surgery in the 1:1 matched cohort revealed higher R0 resection rates (99% vs 95%, p=0.030), lower postoperative morbidity (27% vs 72%, p<0.001), a trend toward lower postoperative mortality (2% vs 8%, p=0.051), and comparable rates of administration of adjuvant chemotherapy (60% vs 54%, p=0.301). Five-year DFS was significantly better in the laparoscopic group (52% vs 40%, p=0.038), as well as 5-year OS (68% vs 57%, p=0.044) (Figure 1). Survival rates of patients that were excluded from the open group after propensity matching were comparable with the included open group (5-year DFS 37%, p=0.664 and 5-year OS 57%, p=0.914).

Propensity score–based strata

After dividing the entire cohort into four strata based on quartiles of the propensity scores, with stratum 1 containing the lowest propensity for laparoscopic surgery, the R0 resection rates were higher in all strata (although nonsignificant), postoperative morbidity, mainly CD2 complications, was lower in stratum 4 (22% vs 43%, p=0.031), adjuvant chemotherapy higher in strata 2 and 3 and survival outcomes favorable in all strata (although nonsignificant) after laparoscopic surgery vs open (Table 2 and Figure 2). Baseline characteristics of the four strata are provided in Table S3–S6. Stratum 1 contains all emergency cases as well as the highest percentage of extended MVRs (50%). Stratum 2 also includes a high percentage of MVRs (limited MVR 30%, extended MVR 47%), whereas the MVR rates of strata 3 and 4 are 16% and 0%, respectively.

Sensitivity analyses

After sensitivity analyses, no subgroup of patients (eg emergency setting) could be identified in which oncological outcomes of laparoscopic surgery were inferior to outcomes after open surgery (Table 2 and Figure 2). In the “no-MVR” subgroup, DFS and OS were significantly better, and in the extended MVR and elective cases, more adjuvant chemotherapy was administered after laparoscopic surgery as compared to open surgery. For elective cases also, a significant better DFS was seen after laparoscopy.

Adjuvant chemotherapy

Post hoc analyses showed that after adjusting for adjuvant chemotherapy in the Cox regression models, the beneficial effect of laparoscopic surgery on DFS and OS did not remain significant, neither in the unadjusted nor in the regression adjusted and the 1:1 matched cohort. In every multivariable model, adjuvant chemotherapy was significantly associated with improved DFS (HR 0.46; 95% CI 0.35–0.61, HR 0.54; 95% CI 0.41–0.71 and HR 0.37; 95% CI 0.24–0.56, respectively) and OS (HR 0.49; 95% CI 0.33–0.72, HR 0.44; 95% CI 0.31–0.62 and HR 0.31; 95% CI 0.19–0.51, respectively), both in stage II and stage III subgroups (data not shown).

Discussion

In this multicenter series of 424 pT4N0-2M0 colon cancer patients who underwent curative intent surgery, laparoscopic surgery was associated with reduced postoperative morbidity and a higher percentage of patients receiving adjuvant chemotherapy. The multivariable models revealed an independent association of adjuvant chemotherapy with improved survival, but not surgical approach. This suggests that laparoscopic surgery facilitates the administration of adjuvant chemotherapy by improving postoperative recovery and reducing complication rates, with ultimate gain in survival.

A recently published Chinese expert series²⁷ similarly observed higher survival rates after laparoscopic surgery (5-year DFS 57% vs 40%, p=0.053, 5-year OS 61% vs 47%, p=0.060). The increased likelihood of receiving adjuvant chemotherapy due to better recovery from minimally invasive surgery has also recently been observed. Lee et al reported higher rates and less delay to the initiation of adjuvant chemotherapy after laparoscopic surgery for stage III colon cancer in a propensity matched cohort of 66,266 patients, ultimately resulting in improved survival. But this is a new insight that has not previously been described in the literature.²⁸ In randomized controlled trials, such as the COLOR trial,⁷ comparable rates of adjuvant chemotherapy and survival after laparoscopic and open surgery are likely related to strict inclusion of relatively fit patients with small tumors, being operated upon in the elective setting. Therefore, randomized controlled trials have restricted external validity. We hypothesize that more extended application of laparoscopy by experienced surgeons in routine practice is now going to shed a different light on the impact of minimally invasive surgery. The most pronounced differences in administration of adjuvant chemotherapy between laparoscopy and open surgery were found in the extended MVR subgroup (71% vs 44%) and in the emergency setting (60% vs 39%), suggesting that minimizing surgical trauma is only becoming clinically relevant in such high-risk patients. Two series including substantial numbers of T4b cases also reported higher rates of adjuvant chemotherapy after laparoscopic as compared to open surgery (63% vs 37%, p=0.047 and 71% vs 57%, p=0.513).^13,16

The main reason why T4 colon cancer has been considered a contraindication for laparoscopic surgery was the concern about achieving complete resection with negative margins, especially in the case of local ingrowth (pT4b). In this multicenter series, a 100% R0 resection rate in 40 patients requiring MVR was found. Four studies assessed outcomes after laparoscopic resection of pT4b colon cancer in relatively small cohorts of 15–23 patients.^{9,11,13,20,29} R0 resection rates were 83%, 93%, 96%, and 100%, respectively. For experienced surgeons in open MVR who also completed the learning curve for laparoscopic colon surgery, the laparoscopic approach might eventually facilitate dissection in locally advanced cases. This is related to the magnified view at high resolution with a 30-degree camera that enables optimal visualization from different angles, which, for example, might improve dissection from the retroperitoneum below a bulky tumor as compared to median laparotomy. It is often considered that tactile feedback is missing in laparoscopy, but there is still an indirect sense of firmness of tissues. Furthermore, high-quality MVR has become more and more dependent on thorough preoperative anatomical planning of the surgical planes of dissection. Performing minimally invasive surgery increases the skills in intraoperative recognition of (disturbed) anatomy. The proven better hemostatic dissection with less blood loss also favors the laparoscopic approach. Even if a larger incision is needed for extraction of a bulky pT4b tumor, laparoscopy might be the preferred approach by reducing surgical stress response and because of flexibility in the site of extraction to reduce the risk of incisional hernia (eg Pfannenstiel).

Occurrence of severe postoperative complications is an important factor that precludes the use of adjuvant chemotherapy.³⁰ Postoperative infection requiring re-intervention (CD3) or ICU admission (CD4) were highest in the extended MVR and emergency subgroups undergoing open surgery (31% and 28%, respectively). This likely reduced the possibilities of administration of adjuvant chemotherapy in these subgroups. Moreover, it has been increasingly described in the literature that postoperative complications have an independent adverse effect on oncological outcomes, possibly related to upregulation of inflammatory cytokines that enhance tumor progression.^31–33

Previously published series comparing laparoscopic and open surgery for pT4 colon cancer often address the issue of allocation bias with more advanced cases in the open subgroup contained. In this analysis, we have attempted to overcome this bias by using propensity scores and by not excluding MVR and emergency cases. Propensity scores control for baseline differences inherent to observational studies.²⁵ However, the present analysis might still be subject to allocation bias. We have only roughly subdivided MVRs in “limited” and “extended”, while still the more complex extended MVRs could be assigned to the open subgroup. Furthermore, the number of emergency cases were limited. We did not correct for effects such as tumor size, presence of intraabdominal infections, presence of adhesions, body mass index, and comorbidity. Consequently, hidden confounding by allocation to laparoscopic or open surgery (selecting patients) could have occurred for which propensity score analyses do not correct. Nonetheless, we have implemented the propensity scores using different manners and results were comparable amongst all analyses. This represents the best achievable evidence, as new RCTs comparing surgical approach in T4 colon cancer are unlikely to be conducted. Previous RCTs comparing surgical techniques have struggled with differences in completion of learning curves amongst the participating surgeons.^34–36 Rouanet et al proposed to compare a series of expert centers that apply different surgical techniques as an alternative for RCTs, in this way overcoming the problems of learning curves and allocation bias.³⁷ Our study approaches this methodology to some extent, by including two centers (1 and 4) with relatively high laparoscopic surgery rates (42% and 56%), and one participating hospital applying open surgery in 97% of cases, with comparable case mixes. However, the conversion rate of 38% in center 1 suggests that the learning curve has not been fully completed during the study period.

Another limitation of the study is the retrospective data collection with difficulties in the uniformity of defining variables and potential underestimation of the postoperative complication rate. However, postoperative complication rates among the four hospitals were comparable or even higher than previously published.³⁸

Conclusion

In conclusion, this multicenter cohort study showed that laparoscopic surgery for pT4 colon cancer was associated with a lower risk of postoperative morbidity and a higher chance of receiving adjuvant chemotherapy. Receiving adjuvant chemotherapy was independently associated with better survival, which indicates an indirect impact of surgical approach on oncological outcome. However, these results might not be generalizable due to the level of experience in advanced laparoscopic surgery. Centralization of care for pT4 colon cancer should be aimed for, and future series of expert centers performing laparoscopic surgery for T4 colon cancer, especially in the more technically challenging cases, should further confirm our hypothesis.