Introduction

Surgery for colonic diverticulitis can be challenging due to recurrent inflammation and often complicated nature of the cases requiring surgery (fistula, abscess, and stricture). Those patients often have adhesions, bulky mesentery, and loss of tissue planes. As such, colectomy for diverticulitis can be more difficult than colectomy for cancer [1]. Some studies suggest higher conversion rate with laparoscopy for diverticulitis in comparison with other diseases [1,2,3,4].

The robotic platform has many technical advantages over conventional laparoscopy (improved visualization, articulating instruments, and stable camera platform) and can be helpful for dissection in complex cases such as complicated diverticulitis [5,6,7,8]. Earlier studies comparing robotic colectomy for diverticulitis to laparoscopic approach were limited to small size, single-institution case series [6,7,8]. More recently, large case-matched studies comparing robotic with laparoscopic colectomy using national databases were performed [9,10,11,12,13]. Those studies showed that the robotic approach was associated with lower conversion rate [9, 10] and shorter hospital stay [9,10,11,12] with longer operative time [10, 12, 13] and no morbidity or mortality benefits [9,10,11,12,13]. None of those studies specifically compared robotic with laparoscopic colectomy for diverticulitis [9,10,11,12,13] or accounted for the case complexity characterized by disease severity and concomitant procedures performed at the time of colectomy. Disease severity and concomitant procedures could be potential confounders of the association between the index procedure and operative time or postoperative complications [14].

The purpose of this study is to compare elective robotic and laparoscopic left-sided/sigmoid colectomy performed for diverticulitis using case-matched analysis that accounts for surgical complexity.

Materials and methods

We used the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) Targeted Colectomy Database that was available to our institution (2012–2014) to identify patients undergoing colectomy for diverticulitis as defined by the International Classification of Disease Ninth Edition (ICD-9) diagnosis codes (562.11, 562.13). Patients who had emergent surgery, open colectomy, preoperative sepsis, right side colectomy, or total colectomy were excluded. The final cohort included patients undergoing elective laparoscopic or robotic left side colectomy with current procedural terminology (CPT) codes (44140, 44141, 44143, 44144, 44145, 44146, 44147, 44204, 44206, 44207, and 44208) for diverticulitis. Concomitant procedures were identified from any additional CPT codes that were reported in conjunction with the primary index procedure.

Age, gender, race, body mass index (BMI), functional status, the American Society of Anesthesiologists (ASA) class, smoking, steroid use, bleeding disorders, cardiac disease (congestive heart failure or history of coronary artery disease), renal disease (renal failure or being on dialysis), pulmonary disease (dyspnea or chronic obstructive pulmonary disease), weight loss, preoperative laboratory values, and intraoperative details (ostomy creation, hand-assisted technique, robotic vs. laparoscopic approach, surgical wound class, presence of colonic fistula, and concomitant procedures) were reported. Primary study outcomes were 30-day mortality, overall morbidity, and major morbidity. Secondary outcomes included operative time, hospital length of stay, conversion to open, readmission rate, and unplanned return to the OR. Major morbidity was defined as (organ/space infection, wound dehiscence requiring return to OR, postoperative intubation, pulmonary embolism, acute renal failure or dialysis, stroke, cardiac arrest, myocardial infarction, bleeding, sepsis, septic shock, leak, concomitant splenectomy, or ureteric repair). Any morbidity was defined as major morbidities (listed above) plus unplanned return to operating room, superficial wound infection, deep wound infection, pneumonia, failure to wean off the ventilator > 48 h, deep vein thrombosis, urinary tract infection, or ileus). Hospital stay was defined as prolonged if a patient stayed more than 5 days, which is the cutoff for 75th percentile of hospital stay in the cohort. Readmission was reported only if it was related to the operation.

Chi-squared test and Fisher’s exact test were used to compare categorical variables, while Student t test or Wilcoxon rank sum test were used to compare continuous variables.

Propensity scores derived from multivariate logistic regression model for performing robotic vs. laparoscopic surgery were used to match the two groups (1:1) using the nearest neighbor method without replacement. Preoperative characteristics and lab values as well as intraoperative details were used in the propensity score logistic regression model. It was decided a priori that splenectomy and ureteric repair or reconstruction were not included in the matching process, because those procedures are likely performed for intraoperative inadvertent events. In the matched cohort, bivariate analysis was used to compare the primary and secondary outcomes between the two approaches.

For the entire non-matched cohort, supplemental multivariate logistic regression analysis was performed to test the effect of surgical approach (robotic versus laparoscopic) on mortality, major morbidity, and overall morbidity. Multivariate linear regression analysis was used to test the effect of surgical approach on hospital length of stay and operative time. The results of this supplemental analysis were provided as online tables.

All statistical analyses were performed using Stata, version 11 (College Station, TX, USA). Statistical significance was judged at p < 0.05.

Results

We identified 441 robotic and 6776 laparoscopic cases that met our inclusion criteria. Patient demographics and preoperative characteristics were comparable between the two groups. Mean patient age was 57 years for both groups, and mean BMI was 29. The robotic cohort was more likely to have albumin < 3.5 g/dl (11.1% vs. 6.9%; p = 0.003), but it is important to note that 40% of patients had missing albumin levels (Table 1).

Table 1 Comparison of baseline characteristics of the cohort before propensity score matching
Full size table

Stoma creation was performed in less than 4% of the cases and that was similar between the two groups (p = 0.870). Hand assistance (42.98% vs. 35.83%, p = 0.003) and concomitant procedures (38.1% vs. 30.2%, p = 0.001) other than splenic flexure takedown were more commonly performed in the laparoscopic group (Table 2).

Table 2 Comparison of intraoperative details for the cohort before propensity score matching
Full size table

The rate of diverticular fistulas was similar between the two groups (laparoscopic 1.69% vs. robotic 1.59%). Splenic flexure takedown (49.26% vs. 41.5%, p = 0.002) and ureteric stent placement (23.57% vs. 18.59%, p = 0.017) were also more common in the laparoscopic group; however, performing splenectomy (0.45% vs 0.05% p < 0.002) and small bowel repair (1.13% versus 0.41%, p = 0.029) were more common in the robotic group. Other procedures such as lysis of adhesions, cholecystectomy, hernia repair, small bowel resection, and gastrectomy were more often performed in the laparoscopic group, but the difference was not statistically significant (Table 2).

Out of 441 patients undergoing robotic procedures, 439 were matched (1:1) to patients undergoing laparoscopic resection on preoperative characteristics, laboratory values (Table 3), and concomitant intraoperative procedures (Table 4).

Table 3 Comparison of baseline characteristics of the propensity score–matched cohort
Full size table
Table 4 Comparison of intraoperative details for the propensity score–matched cohort
Full size table

In the case-matched cohort, the robotic approach was associated with lower conversion rate to open (7.5% vs. 14.3%, p = 0.001) as well as shorter hospital stay (3.89 days vs. 4.75 days, p < 0.001) and longer operative time (219.26 ± 95.66 min vs.188.87 ± 82.28 min, p < 0.001). There was a suggestive association with lower overall morbidity (14.4% vs. 19.2%, p = 0.058), but it did not reach statistical significance. Major morbidity, mortality, individual complications, return to the OR, and readmissions were similar between the two groups (Table 5).

Table 5 Postoperative outcomes in the propensity score–matched cohort
Full size table

In multivariate analysis of the entire cohort before propensity score matching, robotic colectomy was associated with better overall morbidity (OR = 0.72, 95% CI = 0.55–0.96) and suggestive association with better major morbidity (OR = 0.69, 95% CI = 0.46–1.04). Robotic colectomy was also associated with an average 33-min increase in operative time and 0.75-day shorter hospital stay in comparison with the laparoscopic approach (Supplemental Table 1).

Discussion

We report that elective robotic colectomy for diverticulitis is associated with lower morbidity than laparoscopic colectomy in multivariate analysis, but not in a case-matched cohort that accounts for disease complexity and intraoperative details. To our knowledge, this is the first study from the NSQIP database to compare robotic with laparoscopic colectomy for diverticulitis while adjusting for disease complexity and intraoperative details. Only few studies specifically investigated the use of the robot for diverticulitis [6,7,8]. In addition, most studies from national databases grouped this patient population with other diseases across all types of colectomy [9,10,11,12,13, 15]. Many prior studies failed to show differences in morbidity and mortality between robotic and laparoscopic colon resection [6,7,8,9,10,11,12,13,14,15,16,17]; however, very few of those studies stratified their analysis by the disease process and type of resection [10, 15,16,17], and the outcome of robotic procedures for diverticulitis is largely unknown. Dolejs et al. stratified their case-matched analysis by type of colon resection, but not the disease process and found that robotic low anterior resection was associated with less septic complications (3.1 vs. 1.6%) without differences in overall mortality or morbidity [10]. Similarly, Al-Mazrou et al. found robotic colectomy for various diagnoses to be associated with lower septic complications (4% vs. 2.6%) [11].

Similar to these studies, we report that the robotic approach is not associated with significantly lower overall morbidity rate in case-matched cohort of patients undergoing colectomy for diverticulitis. However in our unmatched cohort, robotic colectomy was associated with better overall morbidity. Conversion rate [18] and performing concomitant procedures [14] are independent predictors of morbidity and could have contributed to the difference in morbidity rates on multivariate analysis for the non-matched cohort. It is likely that the effect of these details did not persist after matching due to their low frequency of occurrence and larger sample size is needed to validate these findings if difference in overall morbidity exists.

Robotic surgery was associated with longer operative time in our study, which is consistent with earlier randomized controlled studies [19] and retrospective comparative analysis [10, 12, 13, 15, 16, 20]. Longer operative time can be attributed to the learning curve of new technique [21], the need for extra time to dock and undock the robot [21] as well as performing concomitant procedures or more complex cases [14]. Performing concomitant procedures is associated with longer operative time; therefore, it is necessary to adjust for disease-related concomitant procedures when comparing operative time of different interventions. The differences in operative time persist in our study even when case complexity and intraoperative details (ostomy creation, presence of fistula, and type of concomitant procedures) are accounted for. These findings suggest that the learning curve and docking/undocking time are perhaps better explanations of the longer operative time associated with robotic cases [21]. Another potential explanation for longer operative time for robotic cases is that using the robot may encourage surgeons to persist despite the complexity of the dissection resulting in lower conversion rate but at a longer time. Details about the operator case volume and experience as well as breakdown of robot-specific times (console time, docking time, and total operative time) are not available in the NSQIP database, so it is not feasible to clearly define whether those factors are contributing to longer operative time in this study.

Shorter hospital stay and lower conversion rate were associated with robotic procedures in our study, which is consistent with other reports [9,10,11,12, 16]. These positive outcomes after using the robot are encouraging and might offset the time and cost associated with robotic procedures [22].

In the unmatched cohort, robotic surgery was associated with higher rate of splenectomy and lower rate of splenic flexure takedown than laparoscopic surgery. In the case-matched cohort, splenectomy was still more common in the robotic group, but that did not reach statistical significance. Splenectomy in the context of colectomy for diverticulitis is most likely related to traction injuries. The loss of haptic feedback and the learning curve associated with using the robot might explain concomitant splenectomy in the robotic group. Furthermore, mobilizing the splenic flexure at time of robotic sigmoidectomy using the earlier models of the robot (Da Vinci Si) can sometimes be challenging, because the robot arm position at time of sigmoidectomy could limit the reach from the left lower quadrant and the pelvis to the left upper quadrant. Splenic flexure mobilization under such circumstances might be a daunting task that is avoided unless absolutely necessary, which may explain the lower rate of splenic flexure mobilization in the robotic group. In the same context, enterotomy repair was also more common in the robotic group. It is not clear whether those are iatrogenic enterotomies or they are related to the disease process (such as the presence of coloenteric fistula); however, these findings are consistent with a prior national study [17]. In a study of the National Inpatient Sample database, Yeo et al. found that robotic procedures were associated with higher rate of iatrogenic complications (such as intraoperative bleeding events or bowel puncture). While this is an important finding in both studies, they are rare events that did not translate into an observed increase in overall morbidity or mortality.

Both ureteric stent placement and ureteric reconstruction were less common in the robotic group, but the latter was not significant. Ureteric stent placement could be related to surgeon preference or the complexity of diverticular disease. Also, better visualization of vital structures at time of dissection and lack of tactile feedback in robotic procedures may limit the usefulness of stents and are therefore less often utilized.

This study is limited by the retrospective nature of the study design; however, the NSQIP database is a robust clinical data that is validated in prior studies [23]. Second, rational for concomitant procedures is not captured, and it is not clear whether they are related to intraoperative inadvertent events or planned combined procedures. As such procedures that are likely related to inadvertent event at time of surgery for diverticulitis (ureteric reconstruction and splenectomy) were included as complications. Enterotomy repair and small bowel resection could be related to inadvertent event or the disease process (coloenteric fistula), so sensitivity analysis was performed to include them as complications and as concomitant procedures without any significant change in the outcomes demonstrated between groups.

In conclusion, robotic surgery for left side diverticulitis is associated with shorter hospital stay and lower conversion rate to open than conventional laparoscopy at the expense of longer operative time. There was a trend of lower overall morbidity with the robotic approach that did not reach statistical significance after controlling for operative complexity and performance of concomitant procedures. Controlled prospective studies are needed to confirm these findings.