Comparison of complication and conversion rates between robotic-assisted and laparoscopic rectal resection for rectal cancer: which patients and providers could benefit most from robotic-assisted surgery?

Abstract

Aims: To compare (1) complication and (2) conversion rates to open surgery (OS) from laparoscopic surgery (LS) and robotic-assisted surgery (RA) for rectal cancer patients who underwent rectal resection. (3) To identify patient, physician, and hospital predictors of conversion.

Materials and methods: A US-based database study was conducted utilizing the 2012–2014 Premier Healthcare Data, including rectal cancer patients ≥18 with rectal resection. ICD-9-CM diagnosis and procedural codes were utilized to identify surgical approaches, conversions to OS, and surgical complications. Propensity score matching on patient, surgeon, and hospital level characteristics was used to create comparable groups of RA\LS patients (n = 533 per group). Predictors of conversion from LS and RA to OS were identified with stepwise logistic regression in the unmatched sample.

Results: Post-match results suggested comparable perioperative complication rates (RA 29% vs LS 29%; p = .7784); whereas conversion rates to OS were 12% for RA vs 29% for LS (p < .0001). Colorectal surgeons (RA 9% vs LS 23%), general surgeons (RA 13% vs LS 35%), and smaller bed-size hospitals (RA 14% vs LS 33%) have reduced conversion rates for RA vs LS (p < .0001). Statistically significant predictors of conversion included LS, non-colorectal surgeon, and smaller bed-size hospitals.

Limitations: Retrospective observational study limitations apply. Analysis of the hospital administrative database was subject to the data captured in the database and the accuracy of coding. Propensity score matching limitations apply. RA and LS groups were balanced with respect to measured patient, surgeon, and hospital characteristics.

Conclusions: Compared to LS, RA offers a higher probability of completing a successful minimally invasive surgery for rectal cancer patients undergoing rectal resection without exacerbating complications. Male, obese, or moderately-to-severely ill patients had higher conversion rates. While colorectal surgeons had lower conversion rates from RA than LS, the reduction was magnified for general surgeons and smaller bed-size hospitals.

Keywords:

• Rectal resection
• rectal cancer
• robotic-assisted
• laparoscopic
• conversion
• complication

Introduction

Colorectal cancer affects 40.1 per 100,000 people in the US¹, with an estimated number of new cases reaching 95,520 for colon cancer and 39,910 for rectal cancer in 2017². Treatment of rectal cancer requires a multidisciplinary approach, often including a combination of chemotherapy, radiation and surgery^3,4.

Open surgery (OS) remains the mainstream surgical approach for rectal resection. However, minimally invasive laparoscopic surgery (LS) for colorectal cancer has been performed for decades⁵. LS is performed using rigid instruments, provides the surgeon with a two-dimensional view via an assistant-controlled camera, and is associated with poor ergonomics, fulcrum effect, and tremor enhancement⁶. LS is also technically challenging due to limited space within the pelvis and the need for autonomic nerve preservation⁵.

Robotic-assisted surgery (RA) aims to overcome the difficulties associated with LS by improving visualization and providing the surgeon with wristed instruments, allowing an improved range of movement and enhanced dexterity^6–9. Developed in the early 1990s, robotic-assisted surgery (da Vinci Surgical System, Intuitive Surgical, Inc., Sunnyvale, CA) is commonly used in procedures like radical prostatectomy and hysterectomy, but was only adopted for use in colorectal surgery in 2001⁹. As an emerging minimally invasive option for colorectal cancer patients who are surgical candidates, many prior investigations comparing RA to LS in rectal cancer resection were single-center, descriptive studies^8,10, often with small sample sizes^6,7. The majority of the prior research has focused primarily on patient-level predictors of conversion from RA and LS to OS and not provider-level factors^11,12.

We conducted an analysis of rectal cancer patients who underwent rectal resection using a large-scale, national hospital administrative database with the objectives of (1) comparing surgical complication rates between LS and RA; (2) comparing conversion rates from LS and RA to OS; and (3) identifying opportunities to improve the quality and efficiency of healthcare delivery by exploring predictors of conversion. To our knowledge, this is the first study with both RA and LS to explore surgeon- and hospital-level predictors of conversion in rectal cancer resection.

Materials and methods

Study design

A US-based real-world database analysis was conducted for rectal cancer patients who underwent a rectal resection between January 1, 2012 and December 31, 2014. To be included in the study, patients had to be at least 18 years of age and identified as having RA or LS.

Data sources

Data for this study were obtained from the Premier Healthcare Data (hereafter, Premier). Premier includes hospital administrative data for all payers, and is an alliance of US community and teaching hospitals that are non-profit¹³. The Premier database is geographically diverse and currently contains 40% US hospital discharges¹⁴. As an aggregated, de-identified, HIPAA compliant database, no institutional review board approval is required for analysis and publication.

Study sample

International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) procedures for rectal resections including anterior and low anterior resections (ICD-9-CM codes 48.62, 48.63, or 48.69) and ICD-9-CM diagnoses for rectal cancer (ICD-9-CM codes 154.0, 154.1, 154.2, 154.3, or 154.8) were used to identify rectal resection procedures for rectal cancer. ICD-9-CM codes were also used to identify conversion to OS (V64.41) and complications (Supplementary Material: Appendix A). Complications were assessed during hospitalization and through 30 days post-discharge; specifically for the intra-operative, initial hospitalization, post-operative 30 day, and perioperative periods, where perioperative is defined as the period from the day of rectal resection procedure through 30 days post-discharge. The specific complications and the corresponding ICD-9-CM codes are provided in Supplementary Material: Appendix B. Surgical modality was identified with ICD-9-CM codes, current procedural terminology (CPT) codes, and text strings available in the Premier database, which are described in the Supplementary Material (Appendix C). Patients identified in the RA or LS group were analyzed in their respective groups, even if they were converted to OS (intent-to-treat analysis).

The inclusion and exclusion criteria and the study sample are summarized in Figure 1. Medicare severity diagnosis related groups (MS-DRG) on the same discharge record with either a rectal resection or a major small and large bowel procedure were included in the study sample.

Figure 1. Flow chart of study sample inclusion and exclusion criteria. Abbreviations. APR, all patient refined; CC, complication or comorbidity; DRG, diagnosis-related group; ICD-9-CM, International Classification of Diseases, Ninth Revision, Clinical Modification; LOS, length of stay; LS, laparoscopic surgery; MedPAR, Medicare provider analysis and review; MCC, major complication or comorbidity; MS, Medicare severity; OR, operating room; R&B, room and board; RA, robotic-assisted surgery. (a) ICD-9-CM codes 48.62, 48.63, or 48.69; included procedures conducted January 2012–November 2014. (b) ICD-9-CM codes 154.0, 154.1, 154.2, 154.3, or 154.8 occurring on the same discharge record as the rectal resection. (c) MS-DRG 329 (major small and large bowel procedure with MCC), 330 (major small and large bowel procedure with CC), 331 (major small and large bowel procedure without CC/MCC), 332 (rectal resection with MCC), 333 (rectal resection with CC) or 334 (rectal resection without CC/MCC) occurring on the same discharge record as the rectal resection. (d) Non-elective is defined as emergency, newborn, trauma center, other, invalid code, and not provided. (e) Discharge status of “LEFT AGAINST MEDICAL ADVICE”, “STILL A PATIENT–EXPECTED TO RETURN”, “DISCH/TRANS TO COURT/LAW ENFORCE” . (f) Chemo- or radiotherapy performed during the initial hospitalization. (g) After enforcing prior exclusion criteria, minimal records were excluded due to missing OR time or $0 in R&B (LS: 5%; RA: 5%). (h) The bottom and top 1% by modality were evaluated and excluded as outliers. (i) Room and board (R&B) per diem costs that were below or above the bottom and top 1% of the 2011 Medicare Provider Analysis and Review (MedPAR) data for R&B per diem costs were excluded. Costs were trimmed to remove extreme values based on MedPAR data. Provider-specific R&B cost-to-charge ratios were applied to the MedPAR R&B charges prior to identifying the R&B outlier threshold.

Exclusion criteria were: (1) rectal resection procedure reported as non-elective, or conducted in the outpatient setting; (2) all patient refined DRG (APR-DRG) severity of extreme, in order to remove patients that had additional medical concerns that may influence the outcomes; (3) chemotherapy or radiation therapy administered during the hospitalization; (4) surgery not performed on admission day, in order to remove patients that had additional medical concerns that may have delayed surgery or influenced other outcomes; (5) a discharge status that could impact the outcomes (left against medical advice, transferred to law enforcement); (6) missing operating room (OR) time; (7) $0 room and board (R&B) costs; and (8) outliers in each modality for OR time, length of stay (LOS), R&B costs, and total costs.

Records with missing data and outliers were excluded. The bottom and top 1% of records for OR time, LOS, and total costs—by modality—were considered outliers. Additionally, R&B per diem costs that were below or above the bottom and top 1% of the 2011 Medicare Provider Analysis and Review (MedPAR) data for R&B per diem costs were also excluded from our sample. Provider-specific R&B cost-to-charge ratios were applied to the MedPAR R&B charges prior to identifying the R&B outlier threshold.

Statistical analysis

Propensity score (PS) matching (1–1) was conducted with the Greedy SAS scoring algorithm¹⁵. PS match was performed matching patients who had RA surgery to those who had LS. Covariates considered in the model included patient demographic and clinical characteristics such as: age categories (18–64, ≥ 65), gender, race, primary payer, the MS-DRG categories of major complication or comorbidity (MCC), complication or comorbidity (CC), or without MCC/CC, body mass index (BMI) categories, and Charlson Comorbidity Index (CCI) categories (0–3, 4–6, 7–10, > 10). Additionally, provider characteristics such as hospital facility, surgeon specialty, physician volume of the applicable ICD-9-CM procedure codes performed between 2010–2014, and year of rectal resection procedure were also considered in the model.

Descriptive statistical comparisons between RA and LS were examined using Fisher’s exact test or Chi-square test for unmatched categorical measures and McNemar’s test for matched categorical measures; t-test and paired t-test were used for unmatched and matched continuous variables, respectively. Conversion rates by surgical modality were computed for the PS-matched population. Statistical comparisons between RA and LS were examined using Fisher’s exact test.

Predictors of conversion were identified with stepwise logistic regression in the unmatched RA and LS sample. Covariates considered in the stepwise model included surgical modality (LS or RA), age, gender, race, primary payer, APR-DRG severity, BMI categories, CCI, geographic region, urban/rural, number of beds, teaching hospital, surgeon specialty and year of rectal resection procedure. Physician volume was not considered in the model due to the inability to link a given surgeon across hospitals. Entry into the model was allowed if p < .15. Selected variables were removed from the model if p > .10. Analyses were performed using SAS software version 9.4 (SAS Institute Inc., Cary, NC).

Results

Patient sample

Figure 1 shows that, with the inclusion and exclusion criteria applied, there were 758 and 1,192 eligible patients for RA and LS, respectively. The comparisons of the surgical modalities prior to the match are provided in the Supplementary Material (Appendix D). Prior to PS matching, patients in the RA sample were less severe compared to the LS sample, based on the MS-DRG and APR-DRG distributions. Further, prior to PS matching, general surgeons performed 52% of the total rectal resection surgeries. Within the over 700 Premier hospitals, RA was performed less frequently in rural, non-teaching hospitals and by general surgeons.

With PS matching, 77% of the RA patients were matched to 79% of the LS patients (n = 533 per group; Figure 1); thereby creating balanced groups based on patient, surgeon and hospital level characteristics (Table 1). Although the MS-DRG distribution was similar, LS had more patients with a DRG with CC, whereas RA had more patients with a DRG with MCC. Of note, because all patients had a diagnosis of rectal cancer, the minimum CCI score was 2. Additionally, no meaningful difference in procedure volume was observed between surgical modalities within a physician specialty (colorectal surgeon: RA 51%, LS 50%; general surgeon: RA 42%, LS 46%; Table 1). The RA patients had significantly lower average LOS as compared to the LS group (RA 5.1 vs LS 5.7 days; p = .0006).

Table 1. Study sample demographic and baseline clinical characteristics for PS-matched sample.

Characteristic	Robotic-assisted surgery (n = 533)	Laparoscopic surgery (n = 533)	p-value^a
Patient characteristics
Age category, n (%)			.8461
18–64	314 (58.9)	311 (58.3)
≥65	219 (41.1)	222 (41.7)
Gender, n (%)			.6225
Male	303 (56.8)	311 (58.3)
Female	230 (43.2)	222 (41.7)
Race, n (%)			.4200
White	389 (73.0)	382 (71.7)
Black	41 (7.7)	39 (7.3)
Other	103 (19.3)	112 (21.0)
Primary payer, n (%)			.2232
Medicaid	42 (7.9)	29 (5.4)
Medicare	207 (38.8)	197 (37.0)
Self-pay/other	35 (6.6)	46 (8.6)
Commercial	249 (46.7)	261 (49.0)
MS-DRG categories, n (%)			<.0001
DRG is MCC (329 or 332)	33 (6.2)	23 (4.3)
DRG is CC (330 or 333)	227 (42.6)	261 (49.0)
DRG is w/o CC/MCC (331 or 334)	273 (51.2)	249 (46.7)
BMI categories^b, n (%)			.6883
<30	469 (88.0)	467 (87.6)
30–39	37 (6.9)	41 (7.7)
≥40	27 (5.1)	25 (4.7)
CCI categories, n (%)			.8919
2	283 (53.1)	276 (51.8)
3	110 (20.6)	108 (20.3)
4–7	131 (24.6)	141 (26.5)
>7	9 (1.7)	8 (1.5)
Provider or setting characteristics
Provider site of rectal resection^c, n (%)			<.0001
Procedure physician specialty, n (%)			.0670
Colon/Rectal Surgery (CRS)	272 (51.0)	267 (50.1)
General Surgery (GS)	222 (41.7)	243 (45.6)
Other	39 (7.3)	23 (4.3)
Physician volume^d, n (%)			.5271
>5	164 (30.8)	168 (31.5)
≤5	369 (69.2)	365 (68.5)
Year of rectal resection, n (%)			.4250
2012	164 (30.8)	170 (31.9)
2013	201 (37.7)	191 (35.8)
2014	168 (31.5)	172 (32.3)

Abbreviations. BMI, body mass index; CC, complication or comorbidity; CCI, Charlson comorbidity index; DRG, diagnosis related group; MCC, major complication or comorbidity; MS, Medicare severity; SD, standard deviation.

^aP-values compare robotic-assisted to laparoscopic surgery and are from a paired t-test for continuous variables and McNemar’s test or an extension thereof for categorical variables. The variables selected in the stepwise regression for the PSM included race, MS-DRG categories, BMI categories, provider site of rectal resection, and year of rectal resection.

^bBMI categories derived based on ICD-9-CM codes. <30 category: 278.02 or no presence of an ICD-9-CM code for BMI; 30–39 category: 278.0, 278.00, 278.03, or V85.30-V85.39; ≥40 category: 278.01 or V85.41-V85.45.

^cProvider sites are not listed as there were ∼297 provider sites prior to the PSM.

^dPhysician volume derivation is based on the volume of applicable ICD-9-CM procedure codes a physician performed between 2010–2014.

Complications in the matched sample

Within the post-match sample, complication rates during the initial hospitalization did not differ between RA and LS (18% vs 19%, p = .7991; Table 2). Similarly, perioperative complication rates did not differ between RA and LS (29% vs 29%, p = .7784; Table 2). All subjects with a post-operative complication also had a perioperative complication. The RA patients had significantly higher ileostomy rates as compared to the LS group (temporary ileostomy: 3.6% vs 1.3%, p = .0186; ileostomy (not otherwise specified): 10.7% vs 6.4%, p = .0116). Of note, colostomy was not among the top ten concomitant ICD-9-CM procedures, occurring less than 1.3% of the time. The anastomosis rates did not differ between RA and LS (intestinal anastomosis: 8.4% vs 8.1%, p = .8185; anal anastomosis: 5.1% vs 5.3%, p = .8907). The anastomotic leak rates were also similar between RA and LS (3.0% vs 3.8%, p = .4795).

Table 2. Outcomes in the PS-matched sample.

Outcome	Robotic-assisted surgery (n = 533)	Laparoscopic surgery (n = 533)	p-value^a
Conversion to open surgery^b	62 (11.6%)	157 (29.5%)	<.0001
Complications
Intraoperative complications^c	9 (1.7%)	15 (2.8%)	.2207
Complications during the initial hospitalization^d	98 (18.4%)	101 (18.9%)	.7991
Post-operative complications to 30 days^e	157 (29.5%)	153 (28.7%)	.7784
Perioperative complications^f	157 (29.5%)	153 (28.7%)	.7784

Abbreviations. PS, propensity score.

^aP-values compare robotic-assisted to laparoscopic surgery and are from the McNemar’s test.

^bPresence of ICD-9-CM code V64.41 in the intraoperative period.

^cPresence of ICD-9-CM codes 998.11, 998.2, E870.0, and E874.0.

^dComplications occurring during the initial hospitalization, which includes both intraoperative complications and the complications listed in Supplementary Appendix A.

^eIncludes complications listed in Supplementary Appendix A during the initial hospitalization and through the calendar month following discharge (i.e. intraoperative complications are not included).

^fComplications occurring in the initial hospitalization (including intra-operative complications) through the calendar month following discharge.

Conversions in the matched sample

Post-match, overall conversion rate to OS was 12% for RA vs 29% for LS (p < 0001; Table 2). Conversions to OS occurred more frequently in hospitals with fewer than 600 beds. However, RA patients were significantly less likely to be converted across all bed size categories as compared to LS. Additionally, physicians specializing in colorectal or general surgery were significantly less likely to convert to OS from RA than LS. The reduction in conversions for general surgeons using RA vs LS was 22% (RA 13% vs LS 35%; p < .0001) and for colorectal surgeons using RA vs LS was 14% (RA 9% vs LS 23%; p < .0001; Table 3).

Table 3. Conversion characteristics by robotic-assisted and laparoscopic surgery in the PS-matched sample.

Characteristic	Robotic-assisted surgery		Laparoscopic surgery		% Difference (LS – RA)	p-value^a
	Total	Converted n (%)	Total	Converted n (%)
Patient characteristics
Gender
Male	303	41 (13.5)	311	99 (31.8)	18.3	.0045
Female	230	21 (9.1)	222	58 (26.1)	17.0	<.0001
BMI categories^b
<30	469	46 (9.8)	467	132 (28.3)	18.5	<.0001
30–39	37	10 (27.0)	41	12 (29.3)	2.2	1.0000
≥40	27	6 (22.2)	25	13 (52.0)	29.8	.0430
APR-DRG severity
Minor	244	14 (5.7)	225	54 (24.0)	18.3	<.0001
Moderate	236	35 (14.8)	242	78 (32.2)	17.4	<.0001
Major	53	13 (24.5)	66	25 (37.9)	13.4	.1660
Physician or hospital characteristics
Number of hospital beds
1–400	163	23 (14.1)	174	57 (32.8)	18.6	<.0001
401–600	180	23 (12.8)	182	56 (30.8)	18.0	<.0001
>600	190	16 (8.4)	177	44 (24.9)	16.4	<.0001
Procedure physician specialty
CRS	272	25 (9.2)	267	62 (23.2)	14.0	<.0001
GS	222	28 (12.6)	243	85 (35.0)	22.4	<.0001
Other	39	9 (23.1)	23	10 (43.5)	20.4	.1528

Abbreviations. APR, all patient refined; BMI, body mass index; CRS, colon/rectal surgeon; DRG, diagnosis related group; GS, general surgeon; LS, laparoscopic surgery; PS, propensity score; RA, robotic-assisted surgery. ^aP-values compare robotic-assisted to laparoscopic surgery and are from Fisher’s Exact test. ^bBMI categories derived based on ICD-9-CM codes. < 30 category: 278.02 or no presence of an ICD-9-CM code for BMI; 30–39 category: 278.0, 278.00, 278.03, or V85.30-V85.39; ≥40 category: 278.01 or V85.41-V85.45.

Predictors of conversion to OS

The predictive variables selected from the stepwise regression analysis for conversion to OS included surgical modality (LS or RA), gender, APR-DRG severity, BMI categories, number of beds and surgeon specialty. Per the odds ratio (OR), LS surgeries were 4.21-times (95% confidence interval [CI] = 3.180–5.574; p < .0001) more likely to convert to OS than RA surgeries. Males were 1.48-times (p = .0009) more likely to convert than females and, as APR-DRG severity increased, the odds of conversion increased (moderate: OR = 1.32; severe: OR = 1.54). Similarly, the odds of conversion increased as BMI increased (BMI 30–39: OR = 1.59; BMI ≥40: OR = 1.85). Conversion was more likely to occur in hospitals with fewer than 600 beds (1–400 beds: OR = 1.50; 401–600 beds: OR = 1.82) and also if the procedure was performed by a physician specializing in general surgery (OR = 1.54) (Table 4).

Table 4. Predictors of conversion to open surgery in the unmatched sample.

Characteristic, n (%)	Number of observations for the row^a	No conversion (n = 1,495)	Conversion (n = 454)	Estimate	Odds ratio (95% CI)	p-value
Intercept				–3.3915		<.0001
Procedure type
Laparoscopic	1,191	809 (67.9)	382 (32.1)	1.4374	4.21 (3.180–5.574)	<.0001
Robotic-assisted	758	686 (90.5)	72 (9.5)	Reference
Gender
Male	1,125	835 (74.2)	290 (25.8)	0.3896	1.48 (1.173–1.858)	.0009
Female	824	660 (80.1)	164 (19.9)	Reference
APR-DRG severity
Minor	817	669 (81.9)	148 (18.1)	Reference
Moderate	888	660 (74.3)	228 (25.7)	0.2773	1.32 (1.020–1.707)	.0349
Severe	244	166 (68.0)	78 (32.0)	0.4346	1.54 (1.082–2.204)	.0166
BMI categories
<30	1,697	1,333 (78.6)	364 (21.4)	Reference
30–39	148	98 (66.2)	50 (33.8)	0.4640	1.59 (1.068–2.368)	.0223
≥40	104	64 (61.5)	40 (38.5)	0.6134	1.85 (1.176–2.899)	.0077
Number of beds
1–400	863	632 (73.2)	231 (26.8)	0.4048	1.50 (1.126–1.996)	.0056
401–600	504	374 (74.2)	130 (25.8)	0.5979	1.82 (1.325–2.495)	.0002
>600	582	489 (84.0)	93 (16.0)	Reference
Physician specialty
Colon/Rectal Surgery (CRS)	813	677 (83.3)	136 (16.7)	Reference
General Surgery (GS)	1,022	747 (73.1)	275 (26.9)	0.4304	1.54 (1.198–1.974)	.0007
Other^b	114	71 (62.3)	43 (37.7)	1.2430	3.47 (2.200–5.460)	<.0001

Abbreviations. APR, all patient refined; CC, complication or comorbidity; CCI, Charlson comorbidity index; CI, confidence interval; DRG, diagnosis related group; MCC, major complication or comorbidity.

Note: The model predicts the probability of a conversion. Covariates entered the model in a stepwise fashion in the order presented. Entry into the model was allowed if p < .15. Selected variables were removed from the model if p > .10. Variables considered in the model for the stepwise regression included: procedure type, age, gender, race, primary payer, APR-DRG severity, BMI categories, CCI, geographic region, urban/rural, number of beds, teaching hospital, physician specialty, and year of rectal resection. Physician volume was not considered in the model due to the inability to link a given surgeon across hospitals.

^aPercentages are based on the number of observations for the row.

^bIncludes surgical sub-specialties other than colon/rectal surgery and general surgery.

Discussion

This analysis examined complication rates and conversion from RA and LS to OS among patients diagnosed with rectal cancer that underwent rectal resection. Our results suggest that RA has comparable complication rates compared to LS, and a significantly lower conversion rate compared to LS. Hence, without exacerbating patient complications, RA appears to offer a higher probability of completing a successful minimally invasive operation than LS. Furthermore, the current study identified physician- and hospital-level characteristics that may potentially be used to recognize rectal cancer patients in whom completion of minimally invasive LS surgery may prove difficult; specifically, rectal resection procedures performed by non-colorectal surgeons or at hospitals with <600 hospital beds. Such surgeons and facilities may particularly benefit from RA, recognizing that improving the quality and efficiency of healthcare delivery is particularly important among the more severe (and oftentimes, costly) patient populations, such as those with rectal cancer.

In contrast to previous investigations that were often single-center, descriptive studies with limited sample size^8,10, the Premier database represents a more heterogeneous, unselected, real-world population, which provides the opportunity to analyze a more generalizable dataset and provides real-world evidence. In our study, we demonstrated that the complication rate was comparable between RA and LS for the initial hospitalization, 30 days post-discharge, and perioperatively, which is consistent with prior prospective and retrospective studies^10,16,17, including RObotic vs LAparoscopic Resection for Rectal Cancer (the ROLARR Trial); 2015 Annual meeting of the American Society of Colon and Rectal Surgeons [ASCRS], May 30–June 3, 2015, Boston, MA; Presentation by Alessio Pigazzi, MD, PhD, Associate Clinical Professor, University of California, Irvine, CA.

Furthermore, our study demonstrated that the conversion rate was significantly lower for RA as compared to LS, which is consistent with a systematic review by Araujo et al.⁷, who reported lower conversion rates for those undergoing RA (0–9.4%), as compared to LS (0–22%); as well as with a meta-analysis by Sun et al.¹⁸, who also reported a lower conversion rate with RA as compared to LS (OR = 0.07; 95% CI = 0.02–0.31; p = .0004). However, the conversion rates from the present study for both RA (11.6%) and LS (29.5%) were higher than previously reported, based on single-center prospective studies (RA 0% vs LS 0–10.5%)^8,10, a propensity score matched study using the National Cancer Database (NCDB; RA 9.5% vs LS 16.4%)¹⁹, a study using the Healthcare Cost and Utilization Project National Inpatient Sample (HCUP NIS; RA 5.38% vs LS 13.38%)⁹ and the ROLARR Trial (RA 8.1% vs LS 12.2%).

The lower conversion rates reported in previous studies are mainly from studies conducted in academic hospitals^8,20,21, which are likely to be large centers with many specialized surgeons and a high case volume.

Another factor that might have influenced the conversion rate is the different patient populations. Specifically, the present study included exclusively rectal resection for rectal cancer defined by both ICD-9 diagnosis and procedure codes, whereas other studies either included all colorectal procedures²², or consisted of rectal cancer patients with the majority of patients in early stages¹⁹. Masoomi et al.²³ found, in multivariate regression analysis, that malignant pathology significantly increased the risk of conversion nearly two-fold (adjusted odds ratio of 1.90), which supports the higher conversion rates reported in the present study (RA 11.6% vs LS 29.5%). In addition, Allaix et al.²⁴ and Clancy et al.¹² found that conversion to OS commonly occurred in patients with a more advanced tumor, which further suggests that cancer operations are technically more challenging and are associated with greater uncertainty.

Conversion also impacts the patient journey. A lower conversion rate may be indicative of better long-term outcomes, because conversions have been found to be associated with disease recurrence and mortality^12,25–27. A recent meta-analysis of 15 studies with 5,293 colorectal cancer patients¹² reported conversion from LS to OS was associated with increased risk of perioperative blood transfusion, both 30-day and overall mortality, and long-term disease recurrence¹². Prior studies suggest that patient-related factors that may influence conversion from minimally invasive colorectal surgery to OS include – amongst others – higher BMI^11,12,28, male gender^12,28, tumor location^12,28, locally advanced disease²⁸, T3/T4 tumor¹², and lymph node-positive disease¹².

In this study, predictors of conversion included patient, physician, and hospital characteristics; specifically, predictive factors associated with conversion to OS were LS, male gender, higher APR-DRG severity, higher BMI, admission to hospitals with fewer beds, and general surgery surgical specialty. Because it was conducted in academic centers with colorectal surgeons, the ROLARR Trial was unable to explore hospital and physician predictors of conversion. Nonetheless, consistent with our findings, the ROLARR Trial demonstrated that conversion to OS was lower for RA compared to LS among male patients (RA 8.7% vs LS 16.0%), obese patients (RA 18.9% vs LS 27.8%), and those undergoing lower anterior resection (RA 7.2% vs LS 13.3%). These findings consistently suggest consideration of RA in patients with these risk factors in the interest of reducing the likelihood of conversion to OS, and minimizing the risk management implications from conversion in an environment of bundled payment^29–31, where hospitals are increasingly placed at financial risk. With emerging value-based payment methods, providers will be rewarded for quality and efficiency of care.

Similar to the current physician specialty landscape in which general surgeons outnumber colorectal surgeons³², our results prior to matching showed that general surgeons performed a larger proportion of the total rectal resections (52%). In the PS matched population, colorectal surgeons experience lower conversion rates with RA in comparison to LS (RA 9% vs LS 23%). This difference in conversion rate is magnified when comparing conversion rate for general surgeons (RA 13% vs LS 35%). These results suggest that, while both colorectal and general surgeons may benefit from use of RA, general surgeons may benefit to a greater extent.

A similar trend was observed with respect to hospital bed size in that, while small and large hospitals appear to benefit from RA, smaller bed size hospitals may benefit to a greater degree. Additionally, the lower conversion rate in hospitals with more beds may reflect facility procedure volume or processes and procedures. Prior studies of colorectal procedures have demonstrated that higher institutional procedure volume is associated with lower rates of conversion from LS to OS^33–35.

Limitations

The use of a nationwide database may have led to sampling error. However, this is mitigated by the use of the Premier database, which is geographically diverse and currently contains ∼40% of US hospital discharges¹⁴, allowing for generalizability to the US hospital market. Another potential limitation is selection bias, which was minimized through PS matching on known and measured confounders to focus on similar patients between the surgical modalities. Prior to the PS matching, primary payer, APR-DRG severity, MS-DRG, BMI, and patient and provider characteristics were statistically significantly different between RA and LS (Supplementary Material: Appendix D). While MS-DRG remained statistically significantly different following the PS match (Table 1), the RA group had more patients with MCC, suggesting that the results may be conservative for RA. Further, complications could not be classified by Clavien-Dindo classification, because intervention/treatment information specific to a complication cannot be retrieved from this database. As with other oncology publications using Premier data, the PS matching in this study could not control for tumor characteristics such as histology, grade, stage, and distal tumor location, because these tumor characteristics are not captured in the Premier database³⁶. These tumor characteristics may have guided procedure selection, whereby RA may have been used in more complex cases (higher stage and grade tumors)³⁷ due to the improved range of motion, enhanced dexterity, and ability to reach deep within the pelvis. As such, the results reported herein may again be conservative for RA.

Further, surgeon volume has been shown to reduce conversion in both LS and RA modalities¹¹. However, in the Premier database, a given physician cannot be linked across hospitals; as such, physician volume is likely under-estimated and unreliable and, therefore, was omitted from the predictors of conversion analysis. In addition, while ICD-9-CM diagnosis code V64.41 (“laparoscopic surgical procedures converted to open procedure”) has been used consistently in other database studies^21,38–41, the code may be under-reported because it is not directly linked to reimbursement. Nonetheless, any under-reporting should be comparable between the RA and LS groups. Of note, ICD-9-CM V64.41 converts directly to ICD-10-CM Z53.31 (“laparoscopic surgical procedure converted to open procedure”). ICD-10-CM codes were enforced as of October 2015, following the December 31, 2014 cut-off date in the present study for rectal cancer patients who underwent a rectal resection. Lastly, there is the potential for discharge coding errors in the Premier discharge data. However, codes are used for billing, so these will likely be sufficiently complete; coding errors should not differentially impact the surgical modalities, and these are unlikely to substantially affect the findings.

Conclusions

Compared to LS, RA has a comparable complication rate and, when a surgeon chooses to complete a rectal cancer resection procedure using a minimally invasive approach, RA offers a higher probability of completing a successful resection. In addition to LS, predictors of conversion included non-colorectal surgeon and smaller bed-size hospitals. While colorectal surgeons had lower conversion rates from RA than LS, the reduction was magnified for general surgeons and smaller bed-size hospitals. Male, obese, or moderately-to-severely ill patients may also benefit more from RA in terms of reduced conversion to OS. RA may be particularly beneficial in these patients and for providers with these risk factors.

Transparency

Declaration of funding

This study was sponsored by Intuitive Surgical, Inc (Sunnyvale, CA).

Declaration of financial/other relationships

SJA, SD, and RB served as consultants to Intuitive Surgical, Inc. through their employment at Covance. As salaried employees of Covance they did not receive any direct compensation. MH did not receive funding support for this study, but received compensation from Intuitive Surgical as a consultant faculty for training and education for robotic-assisted surgery. EL, SM, and ST are employees of Intuitive Surgical, Inc. Intuitive Surgical is a manufacturer of the da Vinci® Surgical System, which is used to perform various surgical procedures, including the robotic-assisted rectal resection procedure described in this study. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Previous presentations

The study was presented at the AcademyHealth 2016 Annual Research Meeting, Boston, MA, June 26–28, 2016.

Acknowledgments

The authors would like to acknowledge the contributions of Ali Andreasen for her clinical and analytic support, as well as Karolina Badora for her editorial support.