In-Hospital Nonvariceal Upper Gastrointestinal Bleeding following Cardiac Surgery: Patient Characteristics, Endoscopic Lesions and Prognosis

Background. Nonvariceal upper gastrointestinal bleeding (NVUGIB) can occur following cardiac surgery, with sparse contemporary data on patient characteristics and predictors of outcome in this setting. Aim. To describe the clinical and endoscopic characteristics of patients with NVUGIB following cardiac surgery and characterize predictors of outcome. Methods. Retrospective review of 131 consecutive patients with NVUGIB following cardiac surgery from 2002 to 2005. Demographic characteristics, therapeutic management, and predictors of outcomes were determined. Results. 69.5% were male, mean age: 68.8± 10.2 yrs, mean Parsonnet score: 24.6 ± 14.2. Commonest symptoms included melena (59.4%) or coffee ground emesis (25.8%). In-hospital medications included ASA (88.5%), heparin (95.4%, low molecular weight 6.9%), coumadin (48.1%), clopidogrel (22.9%), and NSAIDS (42%). Initial hemodynamic instability was noted in 47.1%. Associated laboratory results included hematocrit 26 ± 6, platelets 243 ± 133 109/L, INR 1.7 ± 1.6, and PTT 53.3 ± 35.6 s. Endoscopic evaluation (122 patients) yielded ulcers (85.5%) with high-risk lesions in 45.5%. Ulcers were located principally in the stomach (22.5%) or duodenum (45.9%). Many patients had more than one lesion, including esophagitis (28.7%) or erosions (26.8%). 48.8% received endoscopic therapy. Mean lengths of intensive care unit and overall stays were 10.4 ± 18.4 and 39.4 ± 46.9 days, respectively. Overall mortality was 19.1%. Only mechanical ventilation under 48 hours predicted mortality (O.R = 0.11; 95% CI = 0.04−0.34). Conclusions. This contemporary cohort of consecutive patients with NVUGIB following cardiac surgery bled most often from ulcers or esophagitis; many had multiple lesions. ICU and total hospital stays as well as mortality were significant. Mechanical ventilation for under 48 hours was associated with improved survival.


Introduction
Upper gastrointestinal bleeding (UGIB) remains an important complication following cardiac surgery since, although infrequent, these complications are clinically relevant because of their associated mortality [1].Many GI complications are reported, with only few studies addressing specifically upper GI bleeding in this context [1][2][3].Moreover, the generalizability of past study findings is limited by differences between the supportive cares administered over a decade ago versus today.Indeed, data are limited on endoscopic findings, the role of modern endoscopic treatment, and the effectiveness of currently approved pharmacological approaches such modern endoscopic hemostasis and adjuvant PPI administration [4].
The primary objective of this study was thus to report on the clinical and endoscopic presentations of patients undergoing cardiac surgery while identifying the impact of modern-day supportive care, including endoscopic therapy.Additionally, we attempted to determine independent predictors of mortality in this highly selected patient population, that we hypothesized would be more dependent on underlying comorbidity than the actual severity and cause of bleeding.

Study Design and Selection of Participants.
We reviewed the care and outcomes of patients who presented with an episode of nonvariceal upper gastrointestinal bleeding (NVUGIB) complicating cardiac surgery between January 2002 and July 2005.Patients were identified through an institutional database maintained by the Department of Cardiovascular Surgery in which all admissions are entered prospectively.In addition, patient information was also abstracted from a retrospective review of medical charts.We collected pre-, intra-, and postoperative variables, including postoperative complications.Three trained experienced research staff and a clinical monitor used a standardized glossary of terms that included definitions of all variables entered into the study.

Study Population.
We included patients presenting with NVUGIB within 30 days following cardiac surgery between Ulcers 3 January 2002 and July 2005.NVUGIB was confirmed through written documentation by a health care professional of hematemesis and/or coffee ground vomiting, melena or hematochezia on rectal examination, the recovery of a bloody nasal gastric aspirate, or a combination thereof [5].We excluded patients initially assessed at another institution for the present episode of bleeding and subsequently transferred to the participating site, patients who were not already in hospital at the time of bleeding, and patients not undergoing cardiac surgery.Approval was obtained from our institutional research ethics board.

Data Collection.
Recorded data included the following variables: immediate prehospitalization characteristics such as demographic information, symptoms, signs, and laboratory data at the onset of bleeding, and the Parsonnet score [6].We also noted information from the hospital stay, including details about the cardiac surgery, endoscopic findings (including the likely cause of bleeding) and hemostasis, supportive treatment and administered pharmacotherapy.Total and intensive care unit (ICU) lengths of stay were also abstracted.Additional outcomes of interest were the development of Clostridium difficile-(C.difficile-) associated diarrhea (defined as diarrhea accompanied by a positive C. difficile toxin assay), sternal infection, heart failure, and the need for mechanical ventilation and its duration.Bleeding outcomes that were recorded included the need for blood transfusions and the need for surgery related to the episode of NVUGIB.

Statistical Analysis.
Descriptive data were generated for independent and dependent variables.All categorical data were expressed as proportions and 95% CI.All continuous data were expressed as means ± standard deviations.Logistic regression modeling was carried out to identify independent predictors of mortality using possible clinically relevant variables associated with mortality on univariable analysis (adopting a threshold inclusion P value of 0.15).All statistical analyses were carried out using the SAS software version 9.2 (SAS Institute, Cary, NC, USA).   3.

Study Population
The results of univariable analysis are presented in Tables 1 and 2 for presenting demographics in the form of historical and laboratory characteristics on presentation as well as outcomes amongst patients who survived and died.Significant differences were noted for the presence of chronic renal failure (P = 0.008), heart failure (P = 0.024), and blood urea nitrogen (P = 0.037).
When introducing characteristics associated with death in multivariable analysis, the final model included age, platelets, and mechanical ventilation.The sole significant Ulcers

Discussion
This study describes the endoscopic spectrum of lesions in patients presenting with an upper GI bleed following cardiac surgery.The vast majority (85.5%) had ulcers on endoscopy either in the esophagus, stomach, duodenum, or at multiple sites.In addition, a substantial proportion (28.7%) was found to have esophagitis.GI bleeding events were associated with prolonged ICU and overall hospital stays.GI bleeding in this contemporary setting was also associated with a significant mortality rate (19.1%).
Although rare in this setting, GI complications have recently been reported in 53% of a series of 8709 consecutive patients [7].The most common manifestation is that of UGI bleeding [8], noted in 16%.The mortality of GI complications in the postcardiac surgery patient has long been noted to be high [9].Our findings of a 19.1% mortality confirm that this is still true today and in fact remains very elevated, even when compared to an unselected group of patients with in-hospital onset of nonvariceal UGI bleeding in the same country for whom the observed mortality rate was 11% [10].
Visceral hypoperfusion is often quoted as the main pathophysiological factor leading to GI complications, including bleeding, in this setting with additional contributing factors relating to patient comorbidities [11].Indeed, the average age of the patients is elevated and in our study was 68.8 yrs ± 10.2 years, which is similar to what others have reported [7].An antecedent history of ulcer disease, although not noted in our study, has been reported to be a risk factor for the development of UGI bleeding in past analyses [12].
In a series of 4892 patients undergoing open heart surgery, 18 developed upper gastrointestinal bleeding, all of whom were receiving antiplatelet or anticoagulant medication at the time of the bleed [13], a finding approximating results from our own series.
Halm et al. assessed the effect of Helicobacter pylori infection as a risk factor for UGI bleeding after cardiac surgery, failing to find a significant association [14].On the other hand, patients with UGI bleeding have usually experienced a significantly longer duration of cardiopulmonary bypass and aortic cross-clamp time [14].An analysis of 1477 cardiac surgery patients focusing on the broader issue of postoperative GI complications concluded that the use of a left internal mammary artery also seemed to be a risk factor [1].The role of on-and off-pump surgery in the development of postoperative GI complications remains controversial with disparate findings in the literature [15][16][17].
Our finding of a protective association between a shortened course of ventilatory support (under 48 hours) and subsequent mortality may reflect lesser systemic involvement which has also been associated with a decreased incidence of associated stress related mucosal disease reported in this subgroup of patients; indeed, independent risk factors for GI bleeding in an ICU population include respiratory support for greater than 48 hours in addition to the presence of a coagulopathy [18].Some have suggested a lesser incidence of bleeding duodenal ulcers in postoperative vascular patients [19].However, in most reports, commonly reported lesions at endoscopy have included, as in our series, gastroduodenal ulcers or erosions and esophagitis [12,[20][21][22].These findings are similar to those noted for in-hospital and outpatient patients presenting with an episode of NVUGIB [5,10].Subsequent management therefore utilizes similar therapeutic approaches including the performance of early endoscopy, therapeutic hemostasis, and pharmacotherapy in appropriate patients [4].This management scheme, however, must be even more greatly individualized and guided by a cautious assessment of the cardiorespiratory status of the postcardiac surgery patient.The increased performance of endoscopic hemostasis exceeding the proportion of patients with highrisk stigmata may be due to a more aggressive attitude in this group of patients that is not evidence-based, or an under reporting attributable to the retrospective nature of the data collection.
A randomized trial that initially assessed patients with endoscopy and followed these with postoperative endoscopic studies suggested the benefit of PPIs in this highly selected patient group [2], a finding echoed by a recent underpowered observational study [23].The observed low rate of prophylactic coprescription of antisecretory agents is in part explained by widely publicized fear of side effects such as the development of C. difficile-associated diarrhea [24].
In conclusion, in this contemporary large series of old and sick consecutive patients presenting with an upper GI bleed following cardiac surgery, many patients had multiple endoscopic lesions, with mostly ulcers (85.5%) and esophagitis (28.7%) noted.The ICU and total hospital stays were markedly prolonged, with significant mortality.Only duration of mechanical ventilation less than 48 hours was associated with an improved survival.

Table 2 :
Laboratory data at presentation.