More than 62 million computed tomography (CT) scans currently are being performed each year in the United States compared with approximately three million in 1980.1 Accompanying this rise in CT scans has been in an increase in the concern for cancer risk as a result of ionizing radiation.2 The sensitivity and specificity of CT imaging for postoperative abdominal complications has been reported to be greater than 90%.3,4
In gynecologic surgery, where major postoperative morbidity occurs in up to 4% of cases, CT imaging is considered an essential tool in the diagnosis of postoperative complications.5–9 Bowel perforation and anastomotic leaks, which are reported in 0.8–6.7% of gynecologic oncology cases, are devastating and life-threatening complications associated with a mortality rate of up to 16%.10–12 In many instances, CT imaging may increase cost, delay time to reoperation, and be of low yield in the face of strong clinical suspicion for bowel injury.13 Clinical criteria may be as effective as imaging in diagnosing certain postoperative bowel complications, whereas CT may modulate the diagnosis of other postoperative bowel complications where clinical suspicion is low.
Given concerns with the overuse of CT scans, it is essential to thoroughly evaluate the use of CT imaging in the triage of patients suspected to have postoperative bowel-related complications. This study was designed to investigate whether clinical criteria can be sufficiently used to predict bowel-related complications in postoperative low-, medium-, and high-risk gynecologic patients and to examine the diagnostic accuracy of CT in evaluating bowel complications in these subgroups.
MATERIALS AND METHODS
A retrospective cohort review of patients who underwent abdominal or vaginal gynecologic surgical procedures that involved entry into the peritoneal cavity between January 2, 2008, and December 30, 2010, at the Mayo Clinic, Rochester, Minnesota, was undertaken. The study was approved by the institutional review board. Patients were identified using the Surgical Information Recording System database and patients who had not authorized use of their medical records for research purposes were excluded. Procedures included in the study were hysterectomy (robotic, laparoscopic, vaginal, and abdominal); bilateral or unilateral salpingo-oophorectomy (robotic, laparoscopic, vaginal, and abdominal); laparotomy for adhesiolysis or exploration; retropubic bladder suspension procedures; abdominal sacrocolpopexy; abdominal or robotic myomectomy; diagnostic laparoscopy; surgical cytoreductive surgery or staging for cervical, ovarian, or endometrial cancer; and pelvic exenteration. Patients who underwent hysteroscopic procedures, dilatation and curettage, vulvar surgery, wound exploration, or examination under anesthesia were excluded. Patients meeting these criteria were crossreferenced with an institutional radiology database to identify all patients who underwent CT imaging of the abdomen, pelvis, or abdomen and pelvis within 42 days after their surgery. Patients who underwent CT imaging solely for disease follow-up (eg, malignancy) or for a reason unrelated to the gynecologic procedure were excluded.
For the purpose of this report, the cohort was restricted to those patients with CT scan indications suggestive of a bowel-related complication, which included any of the following: unexplained abdominal pain or distention, unexplained leukocytosis, fever, tachycardia or hypotension, suspected intestinal obstruction, suspected bowel leak or intestinal perforation, unexplained nausea or vomiting, unexplained diarrhea, prolonged ileus, and unexplained abnormal electrolytes. Gastrointestinal contrast was administered routinely in all patients suspected to have a bowel complication as long as it was deemed clinically safe and appropriate. Rectal contrast was not administered in our patient cohort. The reporting of the study is in accordance with the STrengthening the Report of OBservational studies in Epidemiology (STROBE) statement.14
Demographic and clinical data were abstracted from electronic medical records including age, parity, body mass index, medical comorbidities, history of prior abdominal or pelvic radiation, history of prior abdominal or pelvic surgery, and preoperative albumin and hemoglobin. Operative reports were reviewed to determine the type of surgery, presence and extent of adhesions, length of surgery, number of bowel resections if performed and their location, and pathologic diagnoses. Clinical signs and symptoms of potential complications in the postoperative period and before CT imaging were recorded for each patient. The indication(s) for the CT scan, as outlined previously, were also abstracted. The patient's final outcome, defined as confirmed bowel-related complication (compared with no bowel-related complication), was identified through CT scan reports or findings from any subsequent surgical interventions. Surgical complexity was classified as either laparotomy performed for any indication (exploratory laparotomy with or without lysis of adhesions or bowel resection, laparotomy for severe endometriosis, gynecologic cancer staging procedure, cytoreduction, or exenteration) or minimally invasive surgery (robotic, laparoscopic, or vaginal). Validation of data collection was performed in a sample of 10 random patients (kappa 0.99, 95% confidence interval [CI] 0.96–1.0). All data were subjected to edit checks by the project statistician (A.L.W.) and any data inconsistencies were rechecked and resolved.
Statistical analysis was performed using SAS 9.3. Variables were compared between patients with and without a confirmed bowel-related complication using the two-sample t test for age; the Wilcoxon rank-sum test for body mass index, albumin, and hemoglobin; and the χ2 or Fisher’s exact test for all categorical variables. Multivariable logistic regression models were fit based on considering the factors identified in the univariate analysis (parsimonious model) as well as additional factors suggested in the literature (full model). Associations were summarized by calculating the odds ratio and corresponding 95% CIs from the model estimates. The c-index, equal to the area under the receiver operating characteristic curve, was used to summarize a model's overall predictive ability. All calculated P values were two-sided and P values <.05 were considered statistically significant.
The parsimonious model was used to derive a prediction tool to predict the occurrence of a bowel-related complication. Points were assigned for each variable relative to the magnitude of the association in the multivariable model by assigning the variable with the highest β coefficient a total of 10 points. A patient's pretest probability and risk score were computed based on their total points and patients were divided into three risk groups (low, medium, and high risk). The diagnostic accuracy of the CT (negative compared with positive) was determined overall and separately for the three risk groups. An exact method for a binomial proportion was used to derive 95% CIs for the sensitivity and specificity measures and the 95% CIs for the positive and negative likelihood ratios were computed according to Simel et al.15 Lastly, pretest probabilities and corresponding 95% CIs were estimated based on fitting a logistic regression model with a single covariate defining the three risk groups and posttest probabilities were estimated after adding the CT finding (negative compared with positive) as an additional covariate to this model.
From January 2, 2008, through December 30, 2010, a total of 609 patients were identified for review. Among these patients, 431 women underwent abdominal or vaginal gynecologic surgical procedures that involved entry into the peritoneal cavity who met the surgical procedure inclusion criteria. Of these 431 patients, 311 patients had a total of 409 CT scans for a suspected complication within 42 days of surgery. Table 1 lists all 431 cases by surgical complexity. On further review of the medical records, 205 patients had a total of 240 CT scans performed specifically for suspected bowel-related complications within 42 days of surgery and the results here are based on the first CT scan for these 205 patients. The mean time from the surgery to the first CT was 12.4 (standard deviation 10.1) days. Among the 205 included patients, indications for performing CT scans included the following: unexplained abdominal pain or distention (n=100); unexplained leukocytosis, fever, tachycardia, or hypotension (n=35); suspected intestinal obstruction (n=35); unexplained nausea or vomiting (n=22); suspected bowel leak or intestinal perforation (n=19); unexplained diarrhea (n=2); prolonged ileus (n=2); and unexplained abnormal electrolytes (n=1). The numbers exceed 205 because nine patients had two indications and one patient had three indications.
Among the 205 patients, 38 (18.5%) had a confirmed bowel-related complication including bowel obstruction (n=22), bowel perforation (n=10), and anastomotic leak (n=6). Table 2 compares the demographic and clinical data between the patients with and without a bowel-related complication. Compared with patients who were not diagnosed with a bowel-related complication, the 38 patients with a bowel-related complication were on average older (mean [standard deviation] age of 61.3 [13.7] compared with 55.3 [15.9] years, P=.033), more likely to have a history of coronary artery disease (7.9% compared with 1.2%, P=.045), and had a laparotomy for any indication (92.1% compared with 66.5%, P=.002). Patients with bowel-related complications were more likely to present with unexpected output contents from the drain, wound, or stoma (8.4% compared with 0.6%, P<.001) as summarized in Table 3. Additionally, the percentage of patients with a bowel resection tended to be higher in patients with a bowel-related complication (12 [(31.6%] compared with 30 [18.0%], P=.061). Of note, in the subgroup of 114 patients with a gynecologic cancer diagnosis, the only factor significantly different between those with and those without a bowel-related complication was unexpected output from the drain, wound, or stoma (25% compared with 0%, P<.001). Among the 91 patients undergoing benign gynecologic surgery, patients with bowel-related complications were more likely to have a history of coronary artery disease (14.3% compared with 0%, P=.022) and were more likely to have a laparotomy (85.7% compared with 35.1%, P<.001) compared with patients without a bowel-related complication. The median overall operative time was 206 minutes (interquartile range 127–301). Among the 38 patients with bowel-related complications, median operative time was 243 (interquartile range 136–322) minutes in comparison to 200 (interquartile range 123–290) minutes for the 167 patients without a bowel-related complication (Wilcoxon rank-sum test, P=.12).
A parsimonious multivariable model was fit considering the significant variables outlined previously along with a full multivariable model that also included bowel resection (compared with no) and a gynecologic cancer diagnosis (compared with benign), which are two factors known to increase a patient's risk of bowel-related complications (Table 4). A comparison of the receiver operating characteristic curves is displayed in Figure 1; the estimate for the predictive ability of the parsimonious model based on the area under the curves was 0.73 (95% CI 0.65–0.81). Given the number of bowel-related complications in the cohort (n=38), the parsimonious model with four variables was selected for further use to determine the pre-CT (pretest) probability of bowel-related complications (top portion of Fig. 2). Points were assigned to each variable relative to the magnitude of the association in the multivariable model and patients were categorized into three risk categories of low (0–1 point), medium (2–4 points), and high (5+ points) based on their total points. The median time interval between initial surgery and CT scan in low-risk, medium-risk, and high-risk patients was 10 (interquartile range 5–22), 10 (interquartile range 6–14), and 5 (interquartile range 4–7), respectively.
Using information acquired from CT scans performed on the 205 patients, CT results indicated the possibility of a bowel complication (positive test) in 37 patients. Computed tomography results had an overall sensitivity of 89.5% (34 of 38, 95% CI 75.2–97.1) and a specificity of 98.2% (164 of 167, 95% CI 94.8–99.6) yielding a positive likelihood ratio of 49.8 (95% CI 16.1–153.7) and a negative likelihood ratio of 0.11 (95% CI 0.04–0.27). The use of the CT within each of the risk categories is summarized in Table 5. Among the 57 patients in the low-risk category, three patients with confirmed bowel-related complications would have been missed as a result of low clinical probability had CT scan not been performed (time interval between surgery and CT 3, 6, and 8 days, respectively). In this low-risk category, the diagnostic accuracy of CT was the highest (sensitivity 100% [three of three], 95% CI 29.2–100; specificity 98% [53 of 54], 95% CI 90.1–100). In comparison, among the 13 patients in the highest-risk category based on their pretest probability, the sensitivity of CT was lower at 70% (95% CI 34.8–93.3), and CT was negative for bowel complications in three patients who were subsequently confirmed to have serious complications (one with anastomotic leak and two with bowel perforations).
When CT diagnoses were included in the previously selected model based on clinical predictors, the area under the curve of receiver operating characteristic increased to 0.99 (95% CI 0.98–1.00) (Fig. 1). Figure 2 summarizes the scoring algorithm and the effect of the CT findings on the pretest probabilities.
Computed tomography imaging is currently considered the gold standard in the diagnosis of the majority of postoperative abdominal complications with sensitivity and specificity rates greater than 90%.3,4 In some cases, however, such as suspected anastomotic leaks or bowel perforation, CT may not add any diagnostic information over and above clinical history and examination in guiding postoperative surgical management. We hypothesized that a set of clinical criteria may enable physicians to circumvent the unnecessary use of CT scans.
Findings of this study affirm the role of CT imaging as a critical tool in the evaluation and management of the postoperative patient as shown in previous studies.4,16–18 However, the diagnostic accuracy of CT imaging can be maximized through its application to patients with a low clinical suspicion of bowel complications. We noted that the sensitivity of CT imaging was lower in patients with a higher pretest probability based on clinical risk factors. In these patients in whom there is a high clinical suspicion of a bowel complication, we suggest that a negative CT may not always be accurate, and immediate intervention may be beneficial in avoiding delays in the surgical diagnosis and management of serious bowel complications.
Bowel complications, particularly bowel leaks, can be devastating in patients of any age and particularly patients undergoing extensive cytoreductive surgery. Although one could argue that the exemption of CT imaging may place patients at risk of incurring complications of an unnecessary reoperation, a decision analysis model would be needed to determine the costs and benefits of such an approach. Findings of this study can be beneficial in modifying decision-making and directing physicians toward safer and more conservatives approaches to care.
In our study, three high-risk patients (one with anastomotic leaks and two with bowel perforation) were misdiagnosed using CT imaging and clinical criteria would have been sufficient in the decision to pursue surgical intervention. It is plausible that a negative CT scan in high-risk patients (by clinical criteria) delayed surgical intervention and provided false reassurance in these patients. Similarly, delay in performing CT scans in low-risk patients (by clinical criteria) in whom CT scan is most accurate (100% sensitivity, 98% specificity) (median time interval between initial surgery and CT scan 10 days, interquartile range 5–22 days) may explain the diagnosis of bowel complications in three patients had CT imaging not been performed. Using our proposed risk-scoring system, physicians may be better able to identify patients in whom the added use of CT imaging can be garnered.
There are limited data in the gynecologic surgery literature on the use of CT imaging as well as the predictors of bowel complications. However, some studies have evaluated risk factors for postoperative complications after gynecologic surgery. Erekson et al19 developed a prediction tool to determine patients' risk of major postoperative complications after benign gynecologic surgery, in which patients were stratified into three risk subgroups. In their model, which was intended for use in the preoperative setting, patient comorbidities including American Society of Anesthesiologist score and preoperative variables were predictive of postoperative complications. In contrast to our study, all postoperative complications were considered together and CT imaging criteria were not included in the analysis.
Limitations of our study include its retrospective nature, which may bias patients' categorizing into risk groups based on clinical pretest probability. However, we relied on details from each patient's medical records to define indications for CT imaging within 42 days of surgery without taking into consideration patients' final outcome. In addition, inconsistencies in documentation of clinical signs and symptoms may have limited the performance of CT criteria. Despite this limitation, we were able to identify a group of clinically relevant criteria that stratify patients into risk categories. Furthermore, having a population of patients with an inherently high risk of postoperative complications (those who underwent CT imaging within 42 days of surgery) restricted our comparison groups. We circumvented this limitation by considering only those patients in whom a clinical suspicion of a bowel complication was present, and we subsequently analyzed patients with confirmed postoperative bowel complications and those without such complications.
In patients with a low clinical probability of a postoperative bowel-related complication (based on absence of unexpected drainage from a drain, wound, or stoma, age younger than 45 years, minimally invasive surgery, and absence of history of coronary artery disease), positive CT findings can accurately determine patients requiring surgical intervention. However, when clinical probability of postoperative complication is high, CT may fail to accurately identify patients with serious complications. Our proposed algorithm is a valuable mechanism of improving patient care and limiting the use of unnecessary resources.
1. Brenner DJ, Hall EJ. Computed tomography—an increasing source of radiation exposure. N Engl J Med 2007;357:2277–84.
2. Schattner A. Computed tomographic pulmonary angiography to diagnose acute pulmonary embolism: the good, the bad, and the ugly: comment on ‘The prevalence of clinically relevant incidental findings on chest computed tomographic angiograms ordered to diagnose pulmonary embolism’. Arch Intern Med 2009;169:1966–8.
3. Gayer G, Jonas T, Apter S, Amitai M, Shabtai M, Hertz M. Postoperative pneumoperitoneum as detected by CT: prevalence, duration, and relevant factors affecting its possible significance. Abdom Imaging 2000;25:301–5.
4. Adibe OO, Muensterer OJ, Georgeson KE, Harmon CM. Severity of appendicitis correlates with the pediatric appendicitis score. Pediatr Surg Int 2011;27:655–8.
5. Stany MP, Farley JH. Complications of gynecologic surgery. Surg Clin North Am 2008;88:343–59, vii.
6. Mason A, Goldacre M, Meddings D, Woolfson J. Use of case fatality and readmission measures to compare hospital performance in gynaecology. BJOG 2006;113:695–9.
7. Quesnel-Garcia Benitez C, Aviles-Cabrera RN. Accidents and incidents in gynecologic surgery [in Spanish]. Ginecol Obstet Mex 2010;78:238–44.
8. Sharp HT, Swenson C. Hollow viscus injury during surgery. Obstet Gynecol Clin North Am 2010;37:461–7.
9. Erekson EA, Yip SO, Ciarleglio MM, Fried TR. Postoperative complications after gynecologic surgery. Obstet Gynecol 2011;118:785–93.
10. Matthiessen P, Hallbook O, Andersson M, Rutegard J, Sjodahl R. Risk factors for anastomotic leakage after anterior resection of the rectum. Colorectal Dis 2004;6:462–9.
11. McArdle CS, McMillan DC, Hole DJ. Impact of anastomotic leakage on long-term survival of patients undergoing curative resection for colorectal cancer. Br J Surg 2005;92:1150–4.
12. Mourton SM, Temple LK, Abu-Rustum NR, Gemignani ML, Sonoda Y, Bochner BH, et al.. Morbidity of rectosigmoid resection and primary anastomosis in patients undergoing primary cytoreductive surgery for advanced epithelial ovarian cancer. Gynecol Oncol 2005;99:608–14.
13. Allen AM, Antosh DD, Grimes CL, Crisp CC, Smith AL, Friedman S, et al.. Management of ileus and small-bowel obstruction following benign gynecologic surgery. Int J Gynaecol Obstet 2013;121:56–9.
14. von Elm E, Altman DG, Egger M, Pocock SJ, Gotzsche PC, Vandenbroucke JP. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet 2007;370:1453–7.
15. Simel DL, Samsa GP, Matchar DB. Likelihood ratios with confidence: sample size estimation for diagnostic test studies. J Clin Epidemiol 1991;44:763–70.
16. Balthazar EJ, Gordon RB. CT of appendicitis. Semin Ultrasound CT MR 1989;10:326–40.
17. Hulnick DH, Megibow AJ, Balthazar EJ, Naidich DP, Bosniak MA. Computed tomography in the evaluation of diverticulitis. Radiology 1984;152:491–5.
18. Frager D, Medwid SW, Baer JW, Mollinelli B, Friedman M. CT of small-bowel obstruction: value in establishing the diagnosis and determining the degree and cause. AJR Am J Roentgenol 1994;162:37–41.
© 2013 by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. All rights reserved.
19. Erekson EA, Yip SO, Martin DK, Ciarleglio MM, Connell KA, Fried TR. Major postoperative complications after benign gynecologic surgery: a clinical prediction tool. Female Pelvic Med Reconstr Surg 2012;18:274–80.