Predictors of Gastrointestinal Complications in Cardiac Surgery

Abstract

Gastrointestinal problems are infrequent but serious complications of cardiac surgery, with high rates of morbidity and mortality. Predictors of these complications are not well developed, and the role of fundamental variables remains controversial. In a retrospective review of our cardiac surgery experience from July 1991 through December 1997, we found that postoperative gastrointestinal complications were diagnosed in 86 of 4,463 consecutive patients (1.9%). We categorized these 86 patients into 2 groups—Surgical and Medical—according to the method of treatment used for their complications. In the Medical group, 9 of 52 patients (17%) died; in the Surgical group, 17 of 34 (50%) died. By logistic multivariate analysis, we identified 8 parameters that predicted gastrointestinal complications: age greater than 70 years, duration of cardiopulmonary bypass, need for blood transfusions, reoperation, triple-vessel disease, New York Heart Association functional class IV, peripheral vascular disease, and congestive heart failure. Postoperative re-exploration for bleeding was a predictor specific to the Surgical group. Use of an intraaortic balloon pump was markedly higher in the Gastrointestinal group than in the Control group (30% vs 10%, respectively), as was the use of inotropic support in the immediate postoperative period (27% vs 5.6%).

Our results suggest that intra-abdominal ischemic injury is a likely contributing factor in most gastrointestinal complications. In turn, the ischemia is probably caused by hypoperfusion due to low cardiac output, hypotension due to blood loss, and intra-abdominal atheroemboli. The derived models are useful for identifying patients whose risk of gastrointestinal complications after cardiac surgery may be reduced by clinical measures designed to counter these mechanisms.

Gastrointestinal (GI) complications are an infrequent (0.3% to 3%) but serious consequence of cardiac surgery that includes cardiopulmonary bypass. ^1–20 Despite their high associated rates of morbidity and mortality, GI complications are among the least studied postoperative complications of cardiac surgery. It has been well established, however, that the severity of these complications varies widely; consequently, the associated treatments and outcomes are likewise diverse.

Results of previous studies ^14–20 have suggested a variety of demographic variables, risk factors, and surgical parameters that might raise the risk of GI complications after cardiac surgery and cardiopulmonary bypass. However, there is no definitive consensus among these studies.

We hypothesized that the variance in the reported predictors of GI complications after cardiac surgery is, for the most part, a byproduct of one or more of the following: 1) applying only univariate statistical analyses, 2) investigating a limited number of patients with complications, and 3) combining surgically and medically treated GI complications as nondistinct groups. Accordingly, the primary objective of this study was to use multivariate analysis to determine the independent predictors of GI complications after cardiac surgery in a large patient population. We also explored the ways in which these predictors vary, depending on the type and severity of the complication. These multivariate models and comparisons may provide guidelines for identifying patients who are at risk of developing GI complications and may lead to modified diagnostic and therapeutic approaches in such patients.

Patients and Methods

Patient Groups. We retrospectively reviewed 4,463 consecutive cardiac surgery patients, all of whom underwent surgery at our institution between 1 July 1991 and 31 December 1997. Of these, 3,545 patients underwent isolated coronary artery bypass grafting (CABG). Cardiopulmonary bypass was performed using standard techniques, and normothermia was used in most patients (93%). The database was queried for demographic parameters, preoperative comorbidities, cardiovascular history, catheterization data, surgical variables, and postoperative complications. The patients were then categorized into 2 groups: those who developed documented postoperative GI complications (GI group, n = 86), and those who did not (Control group, n = 4,377). The GI group was divided into subgroups, the Surgical group (n = 34) and the Medical group (n = 52), based on the treatment approach for their GI complications.

Statistical Methods. Univariate analysis for categorical variables was conducted using either the χ² test or the Fisher's exact test, depending on applicability (SigmaStat; Jandel Scientific; San Raphael, Calif). We analyzed continuous variables with either the unpaired Student's t-test or the nonparametric Mann-Whitney test, depending on normality. Multivariate analysis was performed using stepwise logistic regressions (SAS Institute Inc.; Cary, NC). We used a P value of less than 0.2 for covariate inclusion and a P value of less than 0.05 for covariate retention.

Results

A total of 113 GI diagnoses were identified in 86 of the 4,463 patients (1.9%) who underwent cardiac surgery (Table I). The overall mortality rate among these 86 patients was 30% (26 deaths), compared with 2.5% (109 deaths) in the Control group comprising 4,377 patients (P <0.001). Other postoperative complications were also more frequent in the GI group (Table II), which contributed to their increased morbidity rate and prolonged hospitalization (Table I, P <0.001). The following perioperative variables were associated with increased mortality in the GI group: preoperative renal failure, excessively long duration of bypass, use of intraaortic balloon pump (IABP) and multiple-system failure (Table III).

Table I. Patient Outcome in Control, GI, Surgical, and Medical Groups

Table II. Postoperative Data for Control and GI Groups

Table III. Univariate Predictors of Surviving versus Nonsurviving Patients with GI Complications

GI Complications

Multiple GI diagnoses were present in 26 patients (30%), and these were associated with increased need for surgery and a higher mortality rate (Table IV).

Table IV. Profile of GI Complications and Corresponding Surgery and Mortality

Medical Group. Medical treatment was applied in 52 of the 86 patients (60%) with GI complications. Five types of GI complications were present in this group (Table IV). In descending order of incidence, these complications were upper GI bleeding, pancreatitis, paralytic ileus, liver dysfunction, and lower GI bleeding. The mortality rate in this group was 17% and was the highest in patients with liver failure (80%).

Surgical Group. Surgical treatment was used in 34 of the 86 patients (40%) with GI complications. Three types of GI complications were encountered in the surgically treated patients (Table IV): intestinal ischemia, cholecystitis, and perforated duodenal ulcer. Five patients had intestinal ischemia that was documented at autopsy, because death preceded diagnosis and surgical treatment. Two patients had liver failure, and 6 had pancreatitis. The mortality rate in the Surgical group was 50% and was the highest in patients who had intestinal ischemia (68%).

Univariate Analysis

Parameters distinguishing the Medical and Surgical groups by univariate analysis are summarized in Table V. Patient data and operative data for GI and Control patients are compared in Table VI. Patients in the GI group were an average of 8 years older than patients in the Control group: 61% in the GI group were more than 70 years old, in contrast with 27% of the Control group. Women were over-represented in the group with GI complications (42% vs 33%), and this difference approached statistical significance (P = 0.1). Excluding smoking history, diabetes, obesity, and hyperlipidemia, other concomitant factors all occurred more frequently in the GI group than in the Control group. Most notable were preoperative renal failure (including dialysis), peripheral vascular disease, cerebrovascular disease, chronic obstructive pulmonary disease (COPD), congestive heart failure, inotropic support, triple-vessel disease, and New York Heart Association (NYHA) functional class IV.

Table V. Univariate Differences between Surgical and Medical Groups

Table VI. Univariate Risk Factors for GI Complications in Cardiac Surgery

Patients who underwent isolated valve procedures were equally represented in the GI (7/86; 8.1%) and Control (339/4377; 7.8%) groups; however, combined procedures (for example, CABG and other, or valve and other) were more prevalent in the groups with GI complications. Reoperation, emergency surgery, prolonged time on cardiopulmonary bypass, longer cross-clamp time, hypothermia, use of inotropic support immediately post-op, blood transfusions, use of IABP, and a left ventricular ejection fraction of 30% or less were all associated with GI complications in univariate analysis.

Multivariate Model Analysis

Results of the multivariate, stepwise, logistic regression analysis of GI complications for the entire patient series are summarized in Table VII. The 8 independent predictors of GI complications were as follows: age greater than 70 years, duration of cardiopulmonary bypass, postoperative blood transfusions, reoperation, NYHA functional class IV, triple-vessel disease, history of peripheral vascular disease, and history of congestive heart failure.

Table VII. Independent Predictors of GI Complications

Of the 3,545 patients who underwent CABG alone, 52 had GI complications (1.5%). For these patients, the multivariate predictors of GI complications were similar to those of the entire cardiac surgery series, excluding peripheral vascular disease, congestive heart failure, and NYHA functional class IV (Table VII). Age greater than 70 years, prolonged cardiopulmonary bypass, and postoperative blood use were independent predictors of GI complications in both the Medical and the Surgical groups (Table VII). Re-exploration for bleeding was an independent predictor of GI complications specific to the Surgical group. Note that the number of independent predictors possible in any of the above multivariate logistic models is limited by the number of patients in the GI complications group. The predictors found for the CABG, Surgical, and Medical subgroups were probably influenced by this limitation, and the final variables in each of the models (Table VII) are likely the most important predictors.

Discussion

The incidence of GI complications in the present patient series and the associated high morbidity and mortality rates are comparable to those of previous studies. ^1–20 Also similar was our observation of higher mortality rates in the surgically treated patients than in the medically treated patients (Table I). Factors associated with mortality in these patients were preoperative renal failure, prolonged time on cardio-pulmonary bypass, cross-clamp time, and multiple-system failure (Table III). The different outcomes of patients in the Surgical and Medical treatment groups emphasize the need for early diagnosis and initiation of treatment.

Results of published studies diverge widely with regard to the function of fundamental variables, such as age, type of surgery, and use of cardiopulmonary bypass, in the development of GI complications. ^14–20 Logistic multivariate analysis of our patient series (Table VII) identified 8 parameters as predictors of GI complications for the entire series of cardiac surgery patients: 1) age greater than 70 years, 2) duration of cardiopulmonary bypass, 3) postoperative blood transfusions, 4) reoperation, 5) triple-vessel disease, 6) NYHA functional class IV, 7) peripheral vascular disease, and 8) congestive heart failure. Only parameters 1 through 5 were predictors of GI complications after isolated coronary bypass. Re-exploration for bleeding was of statistical significance in the Surgical group. The differences in the multivariate models for the entire series compared with the subgroups may, in part, result from the smaller numbers of patients in the subgroups.

Christenson's ¹⁶ and Perugini's ¹⁷ groups applied the multivariate analyses necessary to determine independent predictors of GI complications following cardiac operations. In their CABG series of 79 patients with GI complications, Chistenson's group ¹⁶ found that hypertension, postoperative low cardiac output, preoperative NYHA functional classes III and IV, left ventricular ejection fraction less than 40%, postoperative hypovolemic hypotension, reoperation, age of 70 years or more, and emergency surgery were independent predictors of postoperative GI complications after CABG. On the basis of their findings, those authors postulated that splanchnic hypoperfusion is likely a common cause of most or all GI problems.

After applying univariate analysis to data from a smaller group of patients undergoing cardiac surgery, Moneta and colleagues ¹⁸ similarly concluded that visceral hypoperfusion was a crucial factor in the development of postoperative GI complications. Their results also implicated prolonged cardiopulmonary bypass as a predisposing factor for GI complications. Johnston and coworkers ¹⁹ found that longer cardiopulmonary bypass times, valve surgery, and the use of IABP and vasopressors predisposed patients to upper GI complications.

In our study, duration of cardiopulmonary bypass was an important factor in the development of GI complications for the cardiac surgery group as a whole and for patients undergoing CABG. Moreover, the duration of bypass was generally longer for patients in the Surgical group (median, 130 min) than for those in the Medical group (median, 107 min). This suggests that cardiopulmonary bypass may affect the severity of the GI complication. We also suggest that duration of bypass explains the apparently higher incidence of GI complications after combined CABG and valve procedures in comparison with isolated CABG.

Opinions regarding the relationship of cardiopulmonary bypass to GI complications are not uniform in the literature. ^16–20 Spotnitz and coworkers ²⁰ found a direct association between the duration of bypass and the occurrence of GI complications. They linked duration of bypass to a higher rate of GI bleeding, which we also documented in our series. In contrast, Christenson's group ¹⁶ did not find a significant difference in the duration of bypass between patients with and without GI complications after isolated CABG.

The mechanisms by which cardiopulmonary bypass can lead to GI complications have yet to be ascertained. Visceral vasoconstriction that occurs during bypass may contribute to this problem. ^21,22 In 1993, Ohri and coworkers ²² demonstrated that transcellular transport in the small intestine was impaired and gut permeability was increased during cardiopulmonary bypass in 41 patients who had undergone cardiac operations. These findings coincided with mucosal hypoperfusion and ischemia that increased the likelihood of colonization and infection. ²¹

The duration of cardiopulmonary bypass, congestive heart failure, blood transfusions, re-exploration for bleeding, and other factors univariately associated with GI complications (such as use of inotropic support and IABP) are often correlated with low cardiac output and systemic hypotension. Under such conditions, the likelihood of splanchnic hypoperfusion increases. The higher incidence of both triple-vessel disease and peripheral vascular disease in the GI group suggests that advanced arteriosclerotic disease increases the risk of visceral ischemia and, hence, the risk of GI complications. In our opinion, the presence of severe and diffuse arteriosclerosis, combined with decreased splanchnic perfusion, presents the greatest risk of ischemic visceral injury. This may be exacerbated by the use of a wide range of vasoconstrictors to correct hypotension and low cardiac output. ^23–26

Atheroemboli may also lead to acute intra-abdominal complications after open-heart surgery. ^26,27 Moreover, the use of intraaortic balloon pumps may be a source of atheroemboli in patients who have severe aortoiliac arteriosclerotic disease. The occurrence of multiple emboli in the extremities and viscera of patients after intraaortic balloon counterpulsation has been documented and must be considered as a contributing factor. 16,19,26,27 In our series, these devices were used 3 times as often in the GI group as in the Control group.

Blauth and associates ²⁷ found a direct correlation between age, atherosclerosis of the ascending aorta, and atheroemboli. Incremental risk factors for atheroemboli were peripheral vascular disease and atherosclerosis of the ascending aorta. Moreover, those authors found that atheroemboli were most prevalent in the mesenteric circulation (75% of affected patients); next were the brain and the kidneys. Such frequent involvement of the mesenteric circulation indicates a possible role for atheroemboli in GI complications after cardiopulmonary bypass.

The derived multivariate model is useful for estimating the risk of GI complications in individual patients. On the basis of the hypothesized mechanism, specific clinical measures can be implemented in high-risk patients and may reduce the incidence of GI complications. These are some possible measures:

• Consider the severity of ventricular dysfunction and the complexity of the surgical case as indicators of prolonged cardiopulmonary bypass, postoperative low cardiac output, and hypotension.

• Preoperatively identify severe arteriosclerosis of the abdominal aorta and its branches.

• Use vasoconstrictors and intraaortic balloon pumping judiciously in patients who have severe abdominal aortic disease, because the incidence of GI complications may be related to the duration of these measures.

• Exercise particular caution regarding early re-exploration of high-risk patients in case of postoperative bleeding, to avoid prolonged periods of hypotension and blood transfusions.

• Attempt to reduce the cardiopulmonary bypass time and increase efforts to perform off-pump myocardial revascularization in high-risk patients. As the number of off-pump cases increases, the possible impact of cardiopulmonary bypass in the development of this complication may become clearer.

Finally, we recommend that when faced with high-risk patients who match the described profile, physicians should alert the patients and their families to this potential complication and its inherent high mortality rate.