Ulcerative colitis (UC) is characterized by relapse and remission. Up to 25% of patients with UC will experience an acute flare severe enough to require hospitalization.1,2 Despite advancements in medical therapy, 20% to 40% of patients with UC who are hospitalized at least once for medical treatment eventually undergo definitive colectomy.2–4 Although it is preferred to perform a colectomy for UC in the elective setting,5,6 a substantial number of patients will need a colectomy emergently for acute UC flare refractory to medical therapy.7,8
The optimal timing of colectomy in refractory UC remains controversial. Prior studies have shown that delaying colectomy even 2 days after admission for UC in patients who are refractory to medical treatment is associated with increased postoperative complications, hospital costs, and lengths of stay.9–13 However, these conclusions have limited generalizability because of the nature of single-institutional case series or the potential residual confounding of large billing-derived data sets. Current guidelines recommend early consideration of colectomy for acute severe UC,14 but the exact criteria for and timing of such surgery is unclear. Not only is the inpatient trajectory of each patient with acute severe UC difficult to predict, but clinical factors such as immediate preoperative acuity, chronic steroid or immunosuppressant use, and sepsis also influence postoperative outcomes in patients undergoing nonelective surgery. Previous large data set analyses have not controlled for these factors, limiting their conclusions regarding optimal timing for early intervention.13
The purpose of our study was to compare outcomes of early versus delayed emergency colectomy in patients admitted with an acute UC flare while controlling for clinically important preoperative risk factors and markers of patient acuity. Based on prior findings, we hypothesized that we would observe an independent association between immediate colectomy and decreased postoperative complications and mortality rates.
MATERIALS AND METHODS
Data Source and Case Identification with Primary Exposure Variable
We identified all adult patients undergoing a total abdominal colectomy for UC in the American College of Surgeon’s National Surgical Quality Improvement Program (NSQIP) Participant Use Data File from January 1, 2005 to December 31, 2015. Patients with UC were identified by using the International Classification of Diseases, 9th Revision (ICD-9) diagnosis codes listed as the operative indication (ICD-9 code: 55.6). A total abdominal colectomy was identified by Current Procedural Terminology codes (Open: 44150, 44151, 44155, 44156, 44157, 44158; Laparoscopic: 44210, 44211, 44212). We excluded ileal pouch procedures because of their unlikely use in the emergent setting. The NSQIP abstracting guidelines maintain an extensive definition of an “emergent” case (see Supplemental Digital Content – Appendix, http://links.lww.com/DCR/A801), and cases coded as nonemergent were excluded.
Patients were classified as having an “early” surgical intervention if the procedure day was 2 days or less following admission. We based this definition on prior studies and our own exploratory analysis demonstrating a similar trend of increasing complications after 2 to 3 days of admission in NSQIP data.9–13 All other surgical interventions were classified as “late.” We defined these terms based on the perspective of the consulting surgeon rather than the patient’s clinical time course, which is not readily obtainable from NSQIP data. Therefore, in the analysis below, we performed sensitivity analysis of findings using all admissions versus admissions excluding acute inpatient transfers. Ultimately, we reported results with all admissions given statistically identical primary outcomes associations with covariates, the perspective described here, and the larger sample size affording additional subanalyses.
NSQIP Clinically Abstracted Variables
Patient demographics included sex and age categorized into quartiles. The NSQIP also provides the most recent preoperative laboratory values and clinically abstracted, standardized comorbidities that have been previously described and can be found in the information accompanying the annual publication of the Participant Use File.15,16 Laboratory values were stratified by relative distributions within the study population, assigning a value of “high” if greater than 2 SDs from the population mean, and a value of “low” if less than 2 SDs from the population mean. An exception to distributional stratification was made for 3 well-established and clinically highly relevant laboratory tests (hematocrit, white blood cell count, and albumin) using widely published sex- and age-adjusted thresholds.17 In-hospital mortality, postoperative complications, and postoperative length of stay were also obtained from the NSQIP data set by using similarly standardized definitions.15,16
Because of the unique availability of preoperative laboratory values and the clinically abstracted, standardized comorbidities available in NSQIP data, we first compared the early versus late emergent total abdominal colectomy for UC populations by the complete set of preoperative predictors, as well as postoperative outcomes, available in the Participant Use File. In this initial exploratory analysis, we also tested for statistically significant trends in the year-to-year complication rate using simple logistic regression; the lack of any observed trend or statistical significance led us to not pursue temporal effects of the study time period from further analysis. We also evaluated stratifying the analysis by procedures with a bowel-bowel anastomosis and without, but we decided to exclude this variable because of its presumed role as a mediator rather than a confounder of complications and existing power limitations. Because of the high potential for covariate differences between the exposure groups, we performed propensity score 1:1 matching with replacement on the following covariates: sex, age quartile, ASA class tertile, operative approach, chronic (>10 days) steroid (or other immunosuppressant) use, preoperative transfusion needs, acute renal failure, preoperative sepsis and systemic inflammatory response syndrome, and weight loss greater than 10% in the past 6 months.
We selected these predictors through consensus-driven discussions between the authors to identify important factors when assessing the degree of toxicity and operative readiness during hospitalization for acute UC. When considering different measures for certain aspects of physiology (eg, assessing the oxygen-carrying capacity of blood with the last hematocrit laboratory value versus the abstractor’s indication of the need for a preoperative blood transfusion), we selected the option that we believed was more durable and more demonstrably true of worse disease state. Balance plots were used to qualitatively compare pre- and postmatched populations (data not reported). Continuous variables were compared using Wilcoxon-Mann-Whitney tests of medians. Binary and categorical variables were compared using χ2 tests of proportions. We then performed doubly robust logistic regression on the matched sample to examine independent risks of mortality and complications. Regression models were constructed using a priori surgeon and gastroenterologist consensus of predictors with well-established associations to outcomes of interest. We considered including any variables that remained unbalanced after propensity score matching at a p value significance of less than 0.05, but ultimately left these values out of our final model because of the risk that the remaining differences were demonstrative of lead-time bias and were more descriptive of more time in hospital rather than rational predictors of postoperative outcomes. We also tested significant covariates (eg, sepsis, preoperative transfusion) for interaction effects with the primary explanatory variable of operation timing. A p value of less than 0.05 was considered statistically significant for interpretation of multivariable regressions. Missing data were excluded from regression analysis if less than 10% for any variable within the sample. All analyses were performed by using Stata 14.2 (StataCorp, College Station, TX). This study design was reviewed and approved by the Johns Hopkins Medicine Institutional Review Board.
We identified 508 total abdominal colectomies that, after propensity score matching, yielded an effective sample size of 573 through 1:1 matching with replacement. Of the total population, 50.1% underwent colectomy within 2 days of admission. Median time to colectomy was 1 hospital day in the early group versus 6 (interquartile range, 4–9) hospital days in the delayed group (p < 0.001). Overall, 30-day mortality was 6.1% in the unmatched groups and 10.1% after matching. There were no differences in propensity-matched variables in the early versus late surgery groups by sex, age category, surgical approach, operative acuity by ASA class, steroid (or other immunosuppressant) use, weight loss, acute transfusion needs, preoperative sepsis rate, or preoperative renal failure (Table 1). After matching, those in the early colectomy group still had a worse functional status (functional independence: 81.6% versus 87.0%, p = 0.014), higher rates of hypertension requiring medication (39.4% versus 22.4%, p = <0.001), lower rates of preoperative dyspnea (11.1% versus 19.6%, p < 0.001), lower rates of chronic obstructive pulmonary disease (8.4% versus 14.7%, p = 0.018), and lower rates of ventilator dependence (9.8% versus 15.7%, p = 0.032). Preoperative laboratory values were not statistically different, with the exception for a greater prevalence of low serum albumin in the delayed group (94.7% versus 88.6%, p = 0.011), low hematocrit in the delayed group (94.0% versus 87.2%, p = 0.006), high platelets in the delayed group (7.3% versus 3.5%, p = 0.048), and high prothrombin time in the delayed group (6.8% versus 2.6%, p = 0.030) (Table 2).
When comparing early versus late groups, an early colectomy was associated with a lower mortality rate (4.9% versus 20.3%, p < 0.001) and lower complication rate (64.5% versus 72.0%, p = 0.052). Reduced complication occurrences that contributed to a lower overall complication rate in the early group included wound infections, postoperative sepsis, prolonged postoperative mechanical ventilation, unplanned reintubation, postoperative renal failure, postoperative cardiac arrest, and postoperative bleeding (Table 3). There was no difference in postoperative lengths of stay or operative duration.
Multivariable logistic regression with propensity weighting of mortality on preoperative risk factors demonstrated that early colectomy is associated with an 82% decrease in the odds of death compared with delayed colectomy (p < 0.001) (Table 4). Regression of morbidity on preoperative risk factors demonstrated that early colectomy is associated with a 35% decrease in the odds of a complication compared to delayed colectomy (p = 0.034) (Table 5).
We examined patients with UC who underwent emergent total abdominal colectomy, comparing those who were operated on within 48 hours of admission versus those operated on in a delayed fashion. When controlling for other factors, those operated on within 48 hours reduced their odds of a complication by 35% and their odds of death by 82%. The outcomes benefit seen in the early group was maintained even when factoring potential differences in acuity presentation by factors such as sepsis, chronic steroid use, and chronic malnutrition.
The optimal timing threshold of urgent colectomy for refractory UC remains unknown, and prior studies have been constrained by the lack of generalizability in institutional series and potential confounding in administrative billing data. The purpose of this study was to use the increasing availability of emergent colectomy data in the clinically abstracted NSQIP Participant Use file to further explore the relationship between immediate and delayed operation interventions for refractory UC requiring urgent surgery.
The benefits of early surgery when urgent total abdominal colectomy is required for acute UC is well known, but recent studies continue to narrow the interval of how early is early enough. “Early” has historically been defined as colectomy following a short 7- to 10-day course of pharmacologic rescue therapy,4,18 but recent evidence suggests that much shorter time intervals between acute presentation and colectomy may provide substantial reductions in the risk of morbidity.6,9,10,13 Evidence is mounting that, for those patients with acute UC who will require colectomy, even an overnight delay may be associated with clinically significant worse surgical outcomes.
Our findings provide additional support for early surgical intervention in those patients with acute UC who will ultimately require total abdominal colectomy on the same admission. An important limitation of prior studies has been that ascertaining the effect of the ubiquitous features of acute UC patient presentation (eg, hemodynamic stability, steroid use) on surgical timing. Importantly, our conclusions are sustained even when factoring in the preoperative clinical condition of the patient. Although some laboratory values remained unbalanced in the analytic sample, we believe the variable selection methodology described in Materials and Methods addressed any clinical effect of these differences by choosing similar variables that were successfully matched upon.
The limitations of this study are 2-fold: analytical and clinical. First, data quality is limited by the NSQIP data being prospectively collected for quality improvement rather than specifically for clinical decision making. These data may be susceptible to between-institution reporting differences as well as heterogeneity in data capture abilities of the individual clinical reviewers. For example, more than 25% of those without sepsis or systemic inflammatory response syndrome still had leukocytosis. We carefully tested our models with and without such variables to ensure no material impact on our conclusions. Similarly, although the NSQIP administrators maintain in-depth definitions of variables, the current definition of “emergent” can be interpreted differently by different abstractors, and there is the added complexity of interpreting how a surgeon may book a case and his or her intended urgency.19 Finally, differentiating rescue medication regimens is not currently available in NSQIP data. If occurring, the lack of complete or accurate data capture should be equally likely across all records and, therefore, would favor the null hypothesis.
Importantly, although the use of propensity score matching may help account for covariate outliers, the methodology is equally as susceptible as conventional regression methods for residual confounding due to unobserved differences in these populations.20 In addition, propensity score matching enriches analytical samples so that between-group imbalances are minimized, but it also limits the generalizability of overall point estimates. For example, the mortality rates reported here, although statistically comparable between early and late groups, cannot be directly compared with other studies. Because of the ethical and practical difficulties of performing a randomized controlled trial of timing of emergent colectomy in this patient population, data from a high-quality outcomes data set such as NSQIP likely represent the best compromise between data quality and feasibility. The NSQIP data continue to grow, and we anticipate the opportunity to revisit the data in the future to ask additional questions with the increasing potential sample size (eg, day-by-day delay in complication rates).
The second consideration of limitations of this study is the limited usefulness of its findings for the UC stakeholders outside a surgical context. This study is not able to inform the critical transition point when the multidisciplinary care team acknowledges the failure of medical therapy for acute UC and begins to favor surgical intervention. The NSQIP data provide time from admission to surgery, but we are not able to report on the sequence of key events (eg, admission for medical therapy, surgical consultation) that occurs between these 2 points. An important question that cannot be answered by our study design is how the surgeon and multidisciplinary team make an evidence-based decision to abort further medical interventions in favor of surgery. Such a management question is critical for improving care of UC, but will likely need large multi-institutional data sets of acutely presenting patients with UC before further investigation. What we can suggest is that, once it is clear to the team that medical rescue is not possible, any further delay in colectomy may confer substantially worst postoperative outcomes. Such a recommendation is increasingly common at critical surgical decision-making junctures that rely on observational studies and meta-analyses for current professional society guidelines. For example, guidelines for surgical intervention for Clostridium difficile colitis are based on a similarly designed study that noted worse outcomes with further delay if surgery was inevitable.21–23
Patients with UC undergoing emergent total abdominal colectomy had substantially lower mortality and lower rates of complication when comparing those operated on immediately versus those operated on more than 48 hours after admission. Even short delays in surgical intervention once medical therapy has failed may be contributing to substantially worse postoperative outcomes.
The authors thank Joseph K. Canner, M.S., for biostatistical and data management support.
1. Doherty GA, Cheifetz AS. Management of acute severe ulcerative colitis
. Expert Rev Gastroenterol Hepatol. 2009;3:395–405.
2. Dinesen LC, Walsh AJ, Protic MN, et al. The pattern and outcome of acute severe colitis. J Crohns Colitis. 2010;4:431–437.
3. Ananthakrishnan AN, McGinley EL, Saeian K. Outcomes of weekend admissions for upper gastrointestinal hemorrhage: a nationwide analysis. Clin Gastroenterol Hepatol. 2009;7:296–302e1.
4. Aratari A, Papi C, Clemente V, et al. Colectomy
rate in acute severe ulcerative colitis
in the infliximab era. Dig Liver Dis. 2008;40:821–826.
5. Tøttrup A, Erichsen R, Sværke C, Laurberg S, Srensen HT. Thirty-day mortality after elective and emergency total colectomy
in Danish patients with inflammatory bowel disease: a population-based nationwide cohort study. BMJ Open. 2012;2:e000823.
6. Patel SS, Patel MS, Goldfarb M, et al. Elective versus emergency surgery
for ulcerative colitis
: a National Surgical Quality Improvement Program analysis. Am J Surg. 2013;205:333–337.
7. Ozawa T, Ishihara S, Tanaka T, et al. Clinical characteristics and postoperative complications of patients undergoing emergency surgery
for ulcerative colitis
. Hepatogastroenterology. 2015;62:853–858.
8. Kaplan GG, Seow CH, Ghosh S, et al. Decreasing colectomy
rates for ulcerative colitis
: a population-based time trend study. Am J Gastroenterol. 2012;107:1879–1887.
9. Coakley BA, Telem D, Nguyen S, Dallas K, Divino CM. Prolonged preoperative hospitalization correlates with worse outcomes after colectomy
for acute fulminant ulcerative colitis
. Surgery. 2013;153:242–248.
10. Randall J, Singh B, Warren BF, Travis SP, Mortensen NJ, George BD. Delayed surgery for acute severe colitis is associated with increased risk of postoperative complications. Br J Surg. 2010;97:404–409.
11. Bartels SA, Gardenbroek TJ, Bos L, et al. Prolonged preoperative hospital stay is a risk factor for complications after emergency colectomy
for severe colitis. Colorectal Dis. 2013;15:1392–1398.
12. Saha SK, Panwar R, Kumar A, et al. Early colectomy
acute severe ulcerative colitis
improves operative outcome. Int J Colorectal Dis. 2018;33:79–82.
13. Leeds IL, Truta B, Parian AM, et al. Early surgical intervention for acute ulcerative colitis
is associated with improved postoperative outcomes. J Gastrointest Surg. 2017;21:1675–1682.
14. Ross H, Steele SR, Varma M, et al; Standards Practice Task Force of the American Society of Colon and Rectal Surgeons. Practice parameters for the surgical treatment of ulcerative colitis
. Dis Colon Rectum. 2014;57:5–22.
15. Bilimoria KY, Liu Y, Paruch JL, et al. Development and evaluation of the universal ACS NSQIP surgical risk calculator: a decision aid and informed consent tool for patients and surgeons. J Am Coll Surg. 2013;217:833–42.e1.
17. Stedman’s Medical Dictionary. 2016.Philadelphia: Lippincott Williams & Wilkins.
18. Zer M, Wolloch Y, Dintsman M. Pitfalls in the surgical management of fulminating ulcerative colitis
. Dis Colon Rectum. 1972;15:280–287.
19. Mullen MG, Michaels AD, Mehaffey JH, et al. Risk associated with complications and mortality after urgent surgery vs elective and emergency surgery
: implications for defining “quality” and reporting outcomes for urgent surgery. JAMA Surg. 2017;152:768–774.
20. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70:41–55.
21. Ferrada P, Velopulos CG, Sultan S, et al. Timing and type of surgical treatment of Clostridium difficile-associated disease: a practice management guideline from the Eastern Association for the Surgery of Trauma. J Trauma Acute Care Surg. 2014;76:1484–1493.
22. Gash K, Brown E, Pullyblank A. Emergency subtotal colectomy
for fulminant Clostridium difficile colitis–is a surgical solution considered for all patients? Ann R Coll Surg Engl. 2010;92:56–60.
23. Halabi WJ, Nguyen VQ, Carmichael JC, Pigazzi A, Stamos MJ, Mills S. Clostridium difficile colitis in the United States: a decade of trends, outcomes, risk factors for colectomy
, and mortality after colectomy
. J Am Coll Surg. 2013;217:802–812.
Colectomy; Emergency surgery; Refractory; Ulcerative colitis
Supplemental Digital Content
© 2019 The American Society of Colon and Rectal Surgeons