The impact of level II evidence on surgical practice: Dual agent bowel prep for elective colorectal surgery

Introduction

Retrospective cohort studies are the most commonly published and cited original articles in major surgical journals.¹ Ideally, level I evidence guides clinical decision-making, but randomized trials are expensive, resource-consuming. and at times impractical. In these settings, findings from well-designed retrospective cohort studies represent the next best option.² Improvements in retrospective cohort studies are often dependent on data reliability. and the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) offers high-quality data about surgical patients from across the US.³ Thus, retrospective studies using NSQIP-derived data are frequently published in surgical journals to inform better practice.

Due to its reliability and breadth of membership, ACS-NSQIP registries are well-positioned to answer questions about national practice trends. A long-debated example of one such trend is the utilization of preoperative dual mechanical and antibiotic bowel preparation (DBP) before elective colectomy. DBP is an easily implementable intervention, and its use remains a debated topic among general surgeons. The results of this discourse were well-demonstrated in 2015, when four NSQIP based studies were published documenting the advantages of DBP in elective colorectal surgery after 40 years of controversy.^{4 9} These studies were the largest and most comprehensive of their kind, sparking further discussion among the surgical community and corroborating the findings of two major randomized trials published decades prior.^8,9

We aimed to evaluate the impact of high-quality retrospective data on surgical practice across the US. To do this, we examined the utilization rates of an implementable intervention for a common operation during the years surrounding the publication of the largest and most comprehensive retrospective studies strongly supporting its use. Ultimately, by leveraging a well-curated database, we hope to raise awareness regarding the use of evidence-based practice in surgery.

Methods

Data source, design, setting, and participants

The 2013 to 2018 ACS-NSQIP targeted colectomy participant user files were the source of all data in our analysis. Further, owing to its inclusion of only de-identified entries, our study was exempt from institutional review board review. We selected patients who underwent elective colorectal resection without concomitant colostomy creation for chronic diverticular disease, non-obstructive cancer, or non-malignant polyps, and who had complete preoperative bowel prep documentation. We included patients with American Society of Anesthesiologists (ASA) physical status classifications I to III and excluded ventilator dependent patients or those with disseminated cancer, concomitant infectious diagnoses, incomplete planned operative details, or who underwent emergency procedures.

Primary outcome and exposure

Patients who underwent resection from 2013 to 2015 and 2016 to 2018 constituted the “prepublication” and “postpublication” cohorts, respectively. Those who underwent preoperative dual mechanical and oral antibiotic bowel preparation comprised the “DBP” cohort, and those who did not (either no bowel preparation, oral antibiotic preparation only, or mechanical preparation only) constituted the “non-DBP” cohort.

Statistical methods

Descriptive statistics were calculated for demographic and clinicopathologic variables. Per year prepublication and postpublication DBP implementation rates were calculated. Per-year trends of proportional increase in postpublication DBP utilization were evaluated using the Cochran-Armitage Trend Test. Differences between prepublication and postpublication cohorts were evaluated first using univariable logistic regression and then subsequently using multivariable logistic regression while including statistically significant variables from the univariable model. The discriminatory ability of our model was evaluated by using the concordance statistic (c-statistic). We further evaluated the model by randomly dividing our cohort into 2 groups, with a randomly selected 90% of the sample constituting our training set, and the remaining 10% constituting the test set. We derived our second model using the same approach (logistic regression) as above, and evaluated its ability with our test set, obtaining the area under curve as it pertains to the model. Analyses were conducted using R version 3.6.3 (R Foundation, Vienna, Austria) and graphics were created using GraphPad Prism 8.1.1 (GraphPad Software, San Diego, CA).

Results

Overall, 60,693 patients (23,694 prepublication and 36,999 postpublication) met inclusion criteria. Most patients in the post publication cohort (Table I), regardless of non-DBP or DBP receipt, were aged between the years of 50 to 64 (37.1% and 39.7%, respectively), female (52.1% and 52.0%), white (62.9% and 77.3%), were deemed to have Class III ASA status (53.2% and 49.0%), underwent resection for cancer (58.4% and 54.3%), underwent resection through a planned minimally invasive approach (85.8% and 88.9%), and did not have a low anastomosis as a part of their operation (71.6% and 66.5%). Prepublication and postpublication DBP implementation rates were 39.7% and 60.0% (odds ratio [OR] 2.27, 95% confidence interval [CI] 2.20–2.35; P < .001), respectively. While DBP implementation was most popular in both prepublication and postpublication cohorts, the use of mechanical bowel prep and no bowel prep reduced significantly between pre and postpublication cohorts [Table II, P < .001) between the time periods of interest. There was no difference in the use of oral antibiotic bowel preparation only between prepublication and postpublication cohorts (Table II, P = .85). Yearly DBP implementation rose from 33.9% in 2013 to 64.8% in 2018, at a rate of approximately 6.2% per year (Fig). This trend of increase was statistically significant from 2016 through 2018 (P < .001). Among the postpublication cohort, older patients (age 65–74 years: OR 0.90, 95% CI 0.84–0.97;≥:75: OR 82, 95% CI 0.75–0.89), Asian-Americans (OR 0.82, 95% CI 0.73–0.92), smokers (OR 0.89, 95% CI 0.84–0.95), and patients undergoing colectomy for non-malignant polyps (OR 0.84, 95% CI 0.78–0.90) or cancer (OR 0.93, 95% CI 0.88–0.99) were less likely to receive DBP, while patients planned to undergo minimally-invasive procedures (OR 1.27, 95% CI 1.19–1.35), or procedures that involved a low anastomosis (OR 1.21, 95% CI 1.15–1.27) were more likely to receive DBP (Table I). The c statistic of the multivariable model was 0.60, and the area under curve of our model derived using a training set and evaluated using a test set was 0.59.

Table 1.

Characteristics among patients who underwent elective colon surgery from 2016 to 2018

No.(%)	Non-DBP* (N = 14,817)	DBP† (N = 22,182)	Unadjusted		Adjusted
			OR and 95% CI	P value	OR and 95% CI	P value
Age, y
18–49	1,769 (11.9)	3,150 (14.2)	1.00	N/A	1.00	N/A
50–64	5,492 (37.1)	8,796 (39.7)	0.90 (0.84–0.96)	.002	0.94 (0.88–1.01)	.11
65–74	4,251 (28.7)	6,138 (27.7)	0.81 (0.76–0.87)	<.001	0.90 (0.84–0.97)	.01
≥75	3,305 (22.3)	4,098 (18.5)	0.70 (0.65–0.75)	< .001	0.82 (0.75–0.89)	<.001
Sex
Female	7,725 (52.1)	11,528 (52.0)	0.99 (0.95–1.03)	.75	N/A
Race
White	9,315 (62.9)	17,144 (77.3)	1.00	N/A	1.00	N/A
African-American	1,127 (7.6)	1923 (8.7)	0.93 (0.86–1.00)	.06	0.97(0.90–1.05)	.45
Asian-American	523 (3.5)	801 (3.6)	0.83 (0.74–0.93)	.001	0.82 (0.73–0.92)	.001
Other / unspecified	3,852 (26.0)	2314 (10.4)	N/A	N/A	N/A	N/A
Hispanic ethnicity
No	654 (4.4)	1,144 (5.2)	1.00	N/A	N/A
Yes	10,656 (71.9)	19,431 (87.6)	1.04 (0.94–1.15)	.41
Unspecified	3,507 (23.7)	1,607 (7.2)	N/A	N/A
ASA class
I	337 (2.3)	452 (2.0)	1.00	N/A	1.00	N/A
II	6,602 (44.6)	10,858 (48.9)	1.23 (1.06–1.42)	.01	1.08 (0.93–1.25)	.32
III	7,878 (53.2)	10,872 (49.0)	1.03 (0.89–1.19)	.70	1.00 (0.86–1.16)	1.00
Diabetic
Yes	2,575 (17.4)	3,453 ( 15.6)	0.88 (0.83–0.93)	<.001	0.94 (0.89–1.00)	.04
Smoker
Yes	2,240 (15.1)	3,123 (14.1)	0.92 (0.87–0.98)	.005	0.89 (0.84–0.95)	<.001
BMI
< 18.5	219 (1.5)	260 (1.2)	0.80 (0.66–0.96)	.02	0.85 (0.70–1.03)	.09
18.5–24.9	3,696 (24.9)	5,499 (24.8)	1.00	N/A	1.00	N/A
25.0–29.9	5,193 (35.0)	7,584 (34.2)	0.98 (0.93–1.03)	.50	0.96 (0.91–1.02)	.15
≥30	5,655 (38.2)	8,790 (39.6)	1.04 (0.99–1.10)	.11	0.98 (0.93–1.04)	.53
Unspecified	54 (0.4)	49 (0.2)	N/A	N/A	N/A	N/A
Indication for surgery
Chronic diverticular disease	3,411 (23.0)	6452 (29.1)	1.00	N/A	1.00	N/A
Non-malignant polyp	2,685 (18.1)	3594 (16.2)	0.71 (0.66–0.76)	<.001	0.84 (0.78–0.90)	<.001
Cancer	8,660 (58.4)	12045 (54.3)	0.74 (0.70–0.77)	<.001	0.93 (0.88–0.99)	.02
Planned surgical approach
Open	2,100 (14.2)	2,455 (11.1)	1.00	N/A	1.00	N/A
Minimally-invasive	12,717 (85.8)	19,727 (88.9)	1.33 (1.25–1.41)	<.001	1.27 (1.19–1.35)	<.001
Low anastomosis
No	10,616 (71.6)	14,748 (66.5)	1.00	N/A	1.00	N/A
Yes	4,201 (28.4)	7,434 (33.5)	1.27 (1.22–1.33)	<.001	1.21 (1.15–1.27)	<.001

CI, confidence interval; DBP, dual mechanical and antibiotic bowel preparation; OR, odds ratio.

^†Non-DBP: no bowel prep, mechanical bowel prep alone, or oral antibiotic prep alone.

^*DBP: mechanical bowel prep and oral antibiotic prep.

Table 2.

Patients who underwent dual bowel preparation, antibiotic bowel preparation only, mechanical bowel preparation only, and no bowel preparation

No.(%)	2013–2015	2016–2018	P values
Dual bowel preparation	9,411 (39.7)	22,182 (60.0)	Reference
Oral antibiotic bowel preparation only	904 (3.8)	2,114 (5.7)	0.85
Mechanical bowel preparation only	7,644 (32.3)	6,476 (17.5)	<0.001
No bowel preparation	5,735 (24.2)	6,227 (16.8)	<0.001

Fig.

Utilization of dual prep among patients who underwent elective surgery. DBP: mechanical bowel prep and oral antibiotic prep. Non-DBP: no bowel prep, mechanical bowel prep alone, or oral antibiotic prep alone. *Denotes year of presentation at the American College of Surgeons Clinical Congress. DBP, dual mechanical and antibiotic bowel preparation

Discussion

Here, we demonstrate the utilization rate of a well-known, widely available, and beneficial therapy, DBP, which underscores the need for a comprehensive evaluation of evidence-based practice implementation in surgery. Despite a gradual improvement in nationwide implementation, over one-third of patients did not receive DBP after the publication of four large, independent retrospective cohort studies supporting its use. Certainly, clinicians must make the decision about individual care, but our findings suggest evidence-based care is being triaged to certain patient populations, perhaps based upon perceived risks. In the broader sense, this study documents that surgeons do heed level II evidence that populates surgical journals, but adherence and adoption vary. While little prior literature documents practice changes based on the publication of level II data, these findings are reassuring when compared with clinician-reported behavior relative to outcomes from randomized trials.¹⁰ Ultimately, it appears that surgeons decide whether to administer care based on a combination of their understanding of current evidence, training, practice habits, and patient selection, though we concede there is likely a multitude of factors that shape practice landscapes.^{11 14}

This study highlights that while publication of data and associated initiatives likely have some influence over clinician behavior, they may be insufficient for widespread change in practice. Although some surgeons may be resistant to altering their approach in the context of novel findings on methodologic grounds or objections to the generalizability of published data, there ultimately seems to be an incomplete understanding of how or why surgeons apply evidence-based practices to their individual practices.¹⁰ For example, DBP utilization has fluctuated since its popularization in the 1970s, and these peaks and nadirs occurred largely in the absence of any new evidence-based findings.¹⁵ While it seems unlikely that a modern surgeon is wholly unfamiliar with DBP, there may be some who lack detailed understanding of its benefits or who simply doubt the results of the corresponding studies. Some surgeons may opt against DBP use to avoid its associated nausea and vomiting or watery diarrhea, which can then translate into dehydration and potential electrolyte abnormalities.^16,17 Surgeons may want to spare older or otherwise more frail patients these potential risks. Further, intolerance to the preoperative DBP regimen may lead to incomplete preparation and copious watery stool within the colon, which also may increase rates of intraoperative spillage and subsequent infection rates.^18,19 Bowel preparation also has been associated with the disruption of the native colonic mucosa and subsequent local inflammation, which may be associated with higher complication rates.²⁰ Reluctance specifically toward the use of oral antibiotic bowel preparation may be related to a perceived increased risk for Clostridium diffidle infection, though studies suggest that these concerns may be unfounded.^4–7,21 Additionally, surgeons may believe that DBP is unindicated or unnecessary in the setting of their already low infection rates among their own cases^16,17 Furthermore, some may not doubt its effectiveness in some patients but choose to selectively employ DBP for higher risk cases to mitigate the odds of a poorer outcome. Obstacles to widespread implementation of evidence-based practices may be overcome through their integration into preoperative checklists, and our findings do suggest the concomitant popularization of DBP alongside minimally-invasive cases, potentially through the incorporation into enhanced recovery protocols. However, historical experience with venothromboembolic and intravenous antibiotic prophylaxis suggests that while somewhat effective, this strategy is neither immediate nor sufficient.²²

While we feel that our study provides insight into national practice patterns, there are certainly limitations. Despite training and certification examinations that hold NSQIP clinical reviewers to a high standard, only information entered into the medical record can be populated into the NSQIP database. Thus, patients who were perhaps prescribed DBP preoperatively but who did not receive it or underwent an incomplete prep would be erroneously recorded within NSQIP. Importantly, we were unable to determine why patients did not receive DBP. Additionally, the increase of DBP use cannot be singularly attributed to the 2015 publication of these 4 retrospective studies. Specifically, the publication of concurrent prospective studies and meta-analyses could have influenced clinician- and hospital-level decision making regarding the use of DBP.^{11 14}

Fundamentally, incorporation of the best available evidence into daily practice is critical to optimizing patient care. Further, although our findings regarding DBP utilization are encouraging, there remains room for improvement. More generally, we believe that a focus on educating surgical trainees about evidence appraisal will help with applying research to practice. These efforts may be enhanced by efforts of surgical leaders, similar to the model of transferring knowledge from literature to practice suggested previously in the public health sphere.²³ This point is perhaps best highlighted when considering the publication of the Mechanical and Oral Antibiotic Bowel Preparation versus No Bowel Preparation for Elective Colectomy (MOBILE) trial.²⁴ While the trial found that surgical site infections were more common in patients who underwent NBP than in those who underwent DBP (11% vs 7%), this difference was not found to be statistically significant.²⁴ The applicability of these data have been questioned, particularly with respect to the clinical and economic implications of even a 4% difference in surgical site infections rate (regardless of statistical significance), the relatively low percentage of higher-risk low anastomoses, and the relatively high percentage of minimally-invasive cases when compared with prior cohorts.^6,7,25,26 We share these concerns, and argue that individual patient risk-benefit ratio should lead surgeons toward DBP irrespective of the lack of statistical significance in the MOBILE trial. We posit that enhanced commitment to educating graduating general and colorectal surgeons about data appraisal alongside redoubled efforts to assess current literature on behalf of practicing surgeons must be pursued to broadly affect practice patterns.