Abstract
Patients with end-stage renal disease (ESRD) have a five times higher risk of gastrointestinal bleed (GIB) and mortality than the general population. Aortic stenosis (AS) has been associated with GIB from intestinal angiodysplasia. In this retrospective analysis, we obtained data from the 2012 and 2019 National Inpatient Sample. The primary outcome of interest was all-cause in-hospital mortality and risk factors of mortality in patients with ESRD with GIB with aortic valve disorders especially AS. We identified all patients (≥18 years of age) with ESRD () and analyzed based on discharge diagnosis of valvular heart disease () in patients with GIB compared with those without GIB (). Survey statistical methods accounting for strata and weighted data were used for analysis using survey packages in R (version 4.0). Baseline categorical data were compared using Rao-Scott chi square test, and continuous data were compared using Student’s t-test. Covariates were assessed using univariate regression analysis, and factors with value less than 0.1 in the univariate analysis were entered in the final model. The univariate and multivariable associations of presumed risk factors of mortality in ESRD with GIB patients were performed by Cox proportional hazards model censored at length of stay. Propensity score matching was done using MatchIt package in R (version 4.3.0). 1 : 1 nearest neighbour matching was done with propensity scores estimated through logistic regression, in which occurrence of GIB, valvular lesions, and AS was regressed according to other patient characteristics. Among patients with ESRD with valvular heart diseases, AS was found to be associated with increased risk of GIB (; 95% CI 1.003–1.008; ). ESRD patients with AS showed increased risk of lower GIB (; 95% CI 1.01–1.06; ), colonic angiodysplasia (; 95% CI 1.01–1.05; ), stomach and duodenal angiodysplasia (; 95% CI 1.02–1.06; ), need for blood transfusion add pressors as compared to those without AS. However, there was no increased risk of mortality (; 95% CI 0.95–0.99; ).
1. Introduction
Gastrointestinal bleeding (GIB) accounts for over half a million admissions annually in the United States [1]. Patients with end-stage renal disease (ESRD) have a five times higher risk of GIB and mortality than the general population [2, 3]. An incidence of GIB of 161 per 1000 person-years has been reported from initiation of hemodialysis (HD) [4]. Angiodysplasia is acquired from small lesions (generally <5 mm). The etiology of angiodysplasia is multifactorial ranging from age-related degeneration of connective tissue and decreased gastric mucosal perfusion in aortic stenosis (AS), and inhibition of von Willebrand factor (vWF) leads to increased angiogenesis [5].
Previous studies have shown that angiodysplasias are the most common cause of GIB in HD patients and that 47% of Chronic Kidney Disease (CKD) patients had angiodysplasias compared with 17% in controls [6]. Similarly, cardiac valvular lesions have been implicated in increased risk of GIB from using antithrombotic medications like aspirin, antiplatelets, anticoagulants, and nonsteroidal anti-inflammatory drugs [7]. Moreover, AS has also been reported to be associated with GIB from intestinal angiodysplasia and has been termed Heyde’s syndrome (HS).
In the present literature, the risk and mortality of GIB in patients with cardiac valvular lesions and ESRD have been explored separately; however, their concurrent effect remains uninvestigated. To our knowledge, this is the first study to evaluate whether the concurrence of ESRD and valvular lesions leads to increased mortality. We additionally aim to assess morbidity and predictors of mortality in the above-mentioned patient population.
2. Materials and Methods
2.1. Study Data
In this retrospective analysis, we obtained data from the 2012 and 2019 National Inpatient Sample (NIS), sponsored by the Agency for Healthcare Research and Quality, as a part of the Healthcare Cost and Utilization Project.
2.2. Study Design
Patients’ identifications are not revealed in these data. Thus, an institutional review board approval was not required for this study. We identified all patients (≥18 years of age) who had a discharge diagnosis of ESRD () using their respective ICD-9/CM and ICD-10-CM codes. Among the ESRD patients, we excluded the ones without GIB. All patients with upper or lower GIB were identified based on ICD-9/ICD-10 codes. We excluded patients with age <18 years. For baseline characteristics, we used patient demographics (age, race, and sex), relevant comorbidities: coronary artery disease (CAD) or CAD equivalent, peripheral vascular disease (PVD), stroke, liver disease, malignancy, hypertension (HTN), diabetes mellitus (DM), chronic lung disease, tobacco smoking, alcohol use, blood thinner use (anticoagulants/antithrombotic/antiplatelets), nonsteroidal anti-inflammatory drug use, and congestive heart failure (CHF; Table 1). We further analyzed patients with ESRD with GIB for outcomes and risk factors. We additionally analyzed patients with ESRD with GIB based on discharge diagnosis of presence () or absence () of valvular heart disease focusing on aortic valve disorders, especially AS.
2.3. Statistical Analysis
Survey statistical methods accounting for strata and weighted data were used for analysis using survey packages in R (version 4.0). Categorical variables are summarized as percentages, and continuous variables are presented as . Baseline categorical data were compared using Rao-Scott chi square test, and continuous data were compared using Student’s t-test. Identification of risk factors was done using logistic regression models with inverse probability weighing for complex survey data. Covariates were assessed using univariate regression analysis, and factors with value less than 0.1 in the univariate analysis were entered in the final model. The univariate and multivariable associations of presumed risk factors of mortality in ESRD with GIB patients were performed by Cox proportional hazards model censored at length of stay (Supplementary Table 1). Propensity score matching was done using MatchIt package in R (version 4.3.0). 1 : 1 nearest neighbour matching was done with propensity scores estimated through logistic regression, in which occurrence of GIB, valvular lesions and AS were regressed according to other patient characteristics. Subsequently, outcomes like mortality, mechanical ventilation, need for pressors, need for blood transfusion, and occurrence of angiodysplasias among others were studied in the matched model using logistic regression. Data extraction with ICD-9 and ICD-10 codes was done using Python (version 3.9.1). Statistical analysis was carried out using R (version 3.6.2 R Foundation for Statistical Computing, Vienna, Austria).
2.4. Outcomes
The primary outcome of interest was all-cause in-hospital mortality and risk factors of mortality in patients with ESRD with GIB with aortic valve disorders especially AS.
The secondary outcomes included the length, shock requiring vasopressors, acute respiratory failure, and mechanical ventilation. Complications were identified using their respective ICD-9-CM and ICD-10-CM codes.
3. Results
A total of 1,707,452 patients with ESRD were analyzed. Out of this total, 532,061 patients were on HD. Of the total number of patients, 82,330 patients were diagnosed with GIB over the course of their hospitalization. Comparison of baseline characteristics of ESRD patients with and without GIB is summarized in Table 1. Mean age of patients admitted with GIB was significantly higher than those without GIB (). Length of stay, mortality, need for mechanical ventilation, and extracorporeal membrane oxygenation (ECMO) were also significantly higher in patients with GIB. A propensity match analysis between the groups (Supplementary Table 2; Figure 1) also showed a significantly higher risk of mortality (; 95% CI 1.031–1.036; ), need for mechanical ventilation (; 95% CI 1.065–1.073; ), need for pressors (; 95% CI 1.003–1.008; ), and increased length of stay ( vs. ; ; Supplementary Table 3).
Among ESRD patients with GIB, while increasing age was associated with higher mortality, gender and race did not seem to have any associated risks (Supplementary Table 2). Comorbidities like prior myocardial infarction (MI; ; 95% CI 1.51–1.72;), atrial fibrillation (; 95% CI 1.12–1.24; ), chronic liver disease (; 95% CI 1.56–1.73; ), and malignancy (; 95% CI 1.24–1.43; ) showed increased risk of mortality, whereas presence of valvular heart diseases, DM, and HTN was not associated with increased risk (Supplementary Table 2). Interestingly, patients presenting with hematemesis had a 1.5 times higher risk of mortality than those presenting with melena ( vs. ; 95% CI 1.46–1.99 vs. 95% CI 0.58–0.8; ; Supplementary Table 2).
In ESRD patients, there was increased risk of GIB with increasing age (Supplementary Table 4). On multivariate analysis (Supplementary Table 4; Figure 2), Asians were found to be at highest risk for GIB when compared with Caucasians (; 95% CI 1.009–1.013;). Patients with prior MI (; 95% CI 1.012–1.015; ), malignancy (; 95% CI 1.007–1.011; ), liver disease (; 95% CI 1.04–1.06; ), atrial fibrillation (; 95% CI 1.009–1.012; ), and long-term use of Non-steroidal Anti-inflammatory Drugs (NSAIDs) (; 95% CI 1.035–1.055; ) had increased risk of developing GIB (Supplementary Table 4). Among patients with valvular heart diseases, AS was found to be associated with increased risk of GIB (; 95% CI 1.003–1.008; ).
Out of 82,330 patients with ESRD and GIB, 6521 had valvular heart disease. Comparison of baseline characteristics between the two groups is summarized in Table 2. Patients with valvular heart disease were older ( vs. ; ). Proportion of female patients was also higher among patients with valvular heart disease when compared with those without valvular disorders (47.5% vs. 44.9%; ). At baseline, there was no significant difference in mortality (8.6% vs. 9.3%; ) and length of stay ( days vs. days; ) between patients with and without valvular heart disease (Table 3). Propensity score match analysis showed an increased risk for angiodysplasia of colon (; 95% CI 1.01–1.03; ) and angiodysplasia of stomach and duodenum (; 95% CI 1.01–1.03; ) in patients with valvular heart disease with GIB when compared with patients without valvular heart disease (Table 3). There was no increased risk of mortality (; 95% CI 0.98–0.995; ), need for pressors (; 95% CI 0.98–0.996; ), or need for mechanical ventilation (; 95% CI 0.97–0.99; ) with the presence of valvular disorders (Table 3). However, patients with valvular heart disease had higher need for blood transfusions (; 95% CI 1.04–1.07; ).
Among patients with valvular heart disorders and GIB, subgroup analysis was carried out for patients with AS (Tables 4, 5, and 6). Propensity score match analysis showed increased risk of lower GIB (; 95% CI 1.01–1.06; ), increased risk of colonic angiodysplasia (; 95% CI 1.01–1.05; ), stomach and duodenal angiodysplasia (; 95% CI 1.02–1.06; ), and need for blood transfusion (; 95% CI 1.02–1.09; ). However, there was no increased risk of mortality (; 95% CI 0.95–0.99; ), need for mechanical ventilation (; 95% CI 0.95–0.99; ), or pressor requirement (; 95% CI 0.99–1.01; ). There was also no significant difference in length of stay between ESRD and GIB patients with and without AS ( vs. ; ).
4. Discussion
We investigated whether the concurrence of ESRD and valvular lesions leads to increased risk and mortality from GIB. The following key findings can be drawn from this retrospective study of patients with ESRD. First, patients with ESRD with valvular heart disorders have a higher risk of GIB from angiodysplasias compared with the ones with ESRD only. Moreover, patients with ESRD and AS have higher risk of lower gastrointestinal bleeding (LGIB) and angiodysplasias compared with patients with ESRD only. However, among patients with ESRD and GIB, there is no significant difference in mortality and length of stay between patients with and without valvular heart disease including AS.
The association of AS with increased risk of GIB from angiodysplasias is well established and is termed HS. The prevalence of GIB related to HS has been described as 1–3% of patients with AS [8]. The pathophysiology for increased risk of GIB in both ESRD and AS has been linked to acquired vWF deficiency. The proposed pathophysiology of acquired vWF deficiency in both AS and ESRD provokes increased structural transformation of the vWF from the globular to elongated form and, hence, increased proteolysis by metalloproteinases. The shearing forces also result in increased activation of metalloproteinases as per some experimental studies, thereby increasing the lysis of vWF. This is one proposed mechanism that, even in the presence of normal vWF multimeters, there can be increased risk of bleeding, which is also what our results portrayed [9].
Hence, this can explain the increased risk of bleeding from angiodysplasia of colon (; 95% CI 1.01–1.03; ) and angiodysplasia of stomach and duodenum (; 95% CI 1.01–1.03; ) when compared with patients without valvular heart disease. Consistent with these results, our study concluded that ESRD patients with valvular heart disorders have a higher risk of GIB from angiodysplasias compared with the ones with ESRD only. Furthermore, patients with ESRD and AS have higher risk of LGIB and angiodysplasias compared with patients with ESRD only. Moreover, patients in our study with valvular heart disease had a higher need for blood transfusions (; 95% CI 1.04–1.07; ). However, among patients with ESRD and GIB, there is no significant difference in mortality and length of stay between patients with and without valvular heart disease including AS despite increased risk of GIB.
The etiology of GIB in ESRD patients is multifactorial, including antiplatelets, unfractionated heparin, uremic platelet dysfunction, gastrointestinal angiodysplasia, and erosive ulcers [10]. In patients with CKD, the most common cause of upper gastrointestinal bleed (UGIB) is peptic ulcer disease (PUD). However, the most common cause of LGIB is angioectasias followed by diverticular diseases and ischemic colitis [11]. According to previous studies, ESRD patients on Hemodialysis (HD) have higher incidence of angiodysplasias (19–32%) compared with the general population (5–6%) [11]. The reason for this observed high prevalence of angiodysplasias remains unclear. It is hypothesized that these lesions are not more common, rather more frequently observed as a result of bleeding more in the setting of uremic platelets and vWF deficiency [12].
Prior studies have shown an increased mortality for ESRD patients hospitalized with UGIB, with the first-month mortality rate of 13.7% [13]. Furthermore, in a study by Garlapati et al., patients with ESRD who developed LGIB had 2.5 times higher risk of in-hospital mortality than those without LGIB after propensity matching [2]. Compatible with the above studies, our results portrayed that ESRD with GIB had a higher all-cause mortality compared with the ESRD without GIB (; 95% CI 1.031–1.036; ).
A retrospective study analyzing the impact of comorbidities on GIB has shown that advanced age (age >65 years), renal failure, liver disease, and cardiac arrhythmias are highest predictors of GIB [14]. All these parameters are well-recognized risk factors for rebleeding and death and form components of widely accepted and validated risk-stratification scores that have been used in GIB [15]. In congruence with these findings, our study confirmed these predictors of mortality, including advanced age, atrial fibrillation, prior MI, stroke, CHF, liver disease, and malignancy. Other predictors shown in our study included Hispanic race and history of hematemesis. However, in contrast to these earlier publications, Cox analysis for mortality predictors showed statistically significant higher risk of mortality from hematemesis, whereas no increased risk from melena (, 95% CI 1.45–1.99; ). The association remains unclear as both melena and hematemesis are signs of UGIB. However, the bleeding from the right side of the colon and small bowel may also represent melena, and it is possible that the presentations from these were gradually progressive, allowing time for work-up and management without abrupt hemodynamic compromise.
Our study had several strengths that portrayed consistent results. The NIS is the largest publicly available inpatient healthcare database designed to produce U.S. regional and national estimates of inpatient utilization, access, cost, quality, and outcomes. The NIS covers more than 97% of the U.S. population, containing data from around 7 million hospital stays yearly. It estimates about 35 million hospitalizations nationally. The use of NIS thus provided a large sample size contributing overall to a high-power study.
Where most of the findings of our study confirmed the findings from prior literature, there were limitations associated, which may explain some of the unexpected results. Despite concluding that ESRD patients with valvular heart disorders have a higher risk of GIB and thus need for transfusion, especially in relation to AS and angioectasia based on the described pathophysiology, increased risk of mortality was not shown. One major limitation that might explain the unchanged mortality was the characteristic of the database used in this study—in the registry, a considerable number of aortic valve disorders were not specified, and hence, they did not contribute to the total pool of AS. Hence, the appropriate number of AS could not be reported that might have led to GIB leading to a significant increase in mortality. More studies with precise data specification are warranted to further explore the effect of ESRD and valvular disorders like AS on mortality secondary to GIB. According to prior data, cardiovascular disease has been implicated as the most common and first cause of death in patients with ESRD on HD [16]. This might be an important contributing factor to the unchanged mortality since the combined effect of ESRD and valvular disorders including AS may have been preceded by other causes of death, such as cardiovascular. A future study to evaluate an age expectancy in patients with valvular heart disease and ESRD needs to be evaluated as the major cause of mortality in patients with ESRD is from high cardiovascular disease like MI.
5. Conclusion
In conclusion, our study revealed significantly higher mortality, length of stay, need for mechanical ventilation, and ECMO in ESRD patients with GIB. We also concluded that ESRD patients with valvular heart disorders have a higher risk of GIB from angiodysplasias compared with the ones with ESRD only. Furthermore, patients with ESRD and AS have higher risk of LGIB and angiodysplasias compared with patients with ESRD only. Moreover, patients in our study with valvular heart disease had a higher need for blood transfusions. However, among patients with ESRD with GIB, there is no significant difference in mortality and length of stay between patients with and without valvular heart disease including AS despite increased risk of GIB.
Data Availability
The data used to support the findings of this study are included within the article.
Disclosure
The abstract has been accepted to be presented at the Digestive Disease Week (DDW) conference.
Conflicts of Interest
The author(s) declare(s) that they have no conflicts of interest.
Supplementary Materials
Contains supplementary Tables 1–4 referenced in the manuscript text. (Supplementary Materials)