Gastrointestinal bleeding is a common medical emergency with an estimated incidence of 48 to 160 events per 100,000 adults, accounting for approximately 300,000 hospital admissions per year in the United States.1–3 Despite major advances in the management of non-variceal upper gastrointestinal bleeding (NVUGB) in the last decades, its related mortality rate continues to be considerable at 5 to 10%.4,5 Peptic ulcer bleeding is still the most common cause and is responsible for approximately 31 to 67% of all cases, followed by erosive disease, esophagitis, malignancy, and Mallory-Weiss tears.3

Different studies have shown that serum albumin levels have prognostic value for clinical complications in different scenarios including elective surgery, surgical oncology, hospital stay in patients admitted to internal medicine or pediatrics, hospital mortality in stroke patients and patients with major trauma, among other conditions.6–8 Despite the prognostic value of serum albumin in different settings, its clinical value in patients with NVUGB has not been widely evaluated in prospective studies.9 We previously evaluated a cohort of 1,067 patients and found that the serum albumin level upon admission was an independent predictor of in-hospital mortality.10 We also analyzed a group of patients with chronic liver disease (CLD) and NVUGB, and again, hypoalbuminemia was an independent predictor of mortality.4 However, the previously mentioned studies included patients with comorbidities such as CLD, end-stage chronic renal disease (ESRD), and neoplasia, which have been clearly associated with hypoalbuminemia.11–13 These entities are confounding factors associated with final outcomes in these studies.

In this prospective study, we sought to examine the role of serum albumin upon admission in relation to clinical course and in-hospital mortality in patients with NVUGB with no related CLD, ESRD, or neoplasia.

All patients admitted to our hospital with NVUGB during the period of August 2010 to August 2014 were evaluated. Inclusion criteria were adult patients with confirmed NVUGB receiving endoscopic therapy within the first 24h of admission. Patients that met the established criteria were prospectively recorded in a database. We excluded patients transferred to another institution before completing follow-up, patients with CLD, ESRD (on dialysis or not), those with neoplasia as a cause of bleeding, or patients with the presence of cancer at another site.

During the study period, 444 patients with upper gastrointestinal bleeding were evaluated. Bleeding was deemed to be of non-variceal origin in 235 of the patients. Of this population, 33 had CLD, 7 were diagnosed with some form of cancer, and 10 presented with ESRD. They were excluded from the study. A final total of 185 patients, 105 men and 80 women with a mean age of 59.1 ± 19.9 years, were included. Mean serum albumin upon admission was 2.9 ± 0.9g/dl, and 132 patients (71.4%) had hypoalbuminemia (< 3.5g/dl). Overall in-hospital mortality was 7.0%. The global medical and endoscopic findings, as well as outcomes, are shown in Table 1. ROC curve analysis was performed with the serum albumin value upon admission, resulting in an AUROC of 0.738. A value of 3.1g/dl was found to be the discriminating figure between deceased patients and survivors with a sensitivity of 92.3%, specificity of 44.2%, PPV of 10.6%, and NPV of 98.8%. Based on this serum albumin cutoff value, there were differences in the clinical endoscopic variables and outcomes between those with an albumin level ≥ 3.2g/dl and those with a level < 3.2g/dl (Table 2).

We performed a secondary data analysis, comparing the overall performance of serum albumin upon admission and the Rockall score for identifying the mortality rate. No statistical difference for predicting mortality was found between the serum albumin AUROC (0.738) and the total Rockall score (0.715) (fig. 1).

Figure 1.

ROC curve of serum albumin values upon admission and Rockall score for prediction of mortality. AUROC Rockall score: 0.715, AUROC Albumin: 0.738, difference between AUROCs: 0.023, p = NS.

(0.1MB).

The serum albumin level in the setting of NVUGB is an in-hospital mortality marker. To the best of our knowledge, there is only one previous retrospective analysis that specifically addresses the role of serum albumin levels in NVUGB9, although several other studies have reported the association of hypoalbuminemia and mortality.4,10 This is the first prospective report addressing the value of serum albumin upon admission as a risk factor for in-hospital mortality in patients with NVUGB.

Several studies have shown that an alteration in albumin plasma levels has prognostic value in relation to different clinical conditions and scenarios.6–8 Our research team recently published a cohort of 1,067 patients with NVUGB in whom a serum albumin level upon admission < 2.6g/dl, the Rockall score, the presence of rebleeding, and hospital stay duration were independent predictors of in-hospital mortality.10 However, we included patients with CLD, ESRD, and neoplasia. These three comorbidities have been previously identified as predictors of a low albumin level and poor outcome.11–13 In addition, it has been stated that serum albumin presents oxidative damage in cirrhosis and an alteration in its structure and turnover.12,17,18 ESRD has been shown to have a higher risk for gastrointestinal bleeding, with the likely contribution of several factors (glomerular filtration rate, age, Helicobacter pylori infection, and albumin level).19 Finally, hypoalbuminemia in the setting of malignancy without gastrointestinal bleeding has been extensively associated with poor survival.20

When the demographic and clinical variables, the etiology of bleeding, and the mortality rate in our population were compared with other studies in Italy (PNED),5 Canada (RUGBE),21 the United States (AIMS65),22 and Spain,23 we noted that our patients tended to be younger (59.1 ± 19.9 years) than those in the Italian PNED (68 ± 16 years), Canadian RUGBE (66 ± 17 years), American (AIMS65) (75 years, IQR1-3, 60-83), and Spanish (68 ± 15 years) studies. In addition, our study had similar results in regard to the etiology of bleeding, the rebleeding rate, and hospital stay. However, overall mortality in our patients was higher (7.0% vs PNED 4.5%, RUGBE 5.4%, AMIS65 3.2-2.7%, and the Spanish study 2.5-1%). This may be related to the major presence of hypoalbuminemia (71.4%) due to subclinical comorbidities such as CLD caused by nonalcoholic steatohepatitis,24,25 the presence of subclinical malnutrition, and the high prevalence of comorbidity, cardiovascular disease, and diabetes in our group (Table 2).

When comparing our cohort with a previous report from Taiwan,9 we also found that our patients were younger, and the number of men was smaller. However, in our study, similar to the Taiwanese study, patients with normal serum albumin compared with patients in the hypoalbuminemia group were older, they presented with a major prevalence of comorbidities, lower levels of hemoglobin, and elevated blood urea nitrogen and creatinine upon admission. In addition, these patients had longer hospital stay, greater blood transfusion requirements, and higher rebleeding and mortality rates. We did not find any difference between groups in the type of endoscopic lesion responsible for bleeding.

There was no statistically significant difference when albumin values were analyzed and compared with the total Rockall score. This could suggest that serum albumin upon admission has a clinical value as a marker similar to that of the Rockall score for classifying the risk for in-hospital mortality in NVUGB patients without CLD, ESRD, or cancer (fig. 1). The high sensitivity and negative predictive value enabling those patients (with a serum albumin ≥ 3.2g/dl) with a very low likelihood of poor outcome to be distinguished, is remarkable.

Even though albumin plays an important role in maintaining normal physiologic processes and it has been described as a marker of worse prognosis, treatment with its replacement in patients with hypoalbuminemia has not led to better survival in other settings, such as in critically ill patients and those with sepsis.26,27 In the setting of peptic ulcer bleeding, one prospective study demonstrated that albumin administration in patients with hypoalbuminemia, compared with a control group (also with hypoalbuminemia), shortens the duration of hospital stay (9 vs 15 days, p = 0.02). However, no significant difference was found in the rate of rebleeding and mortality.28 Therefore, based on these pieces of evidence and given the higher presence of comorbidity in the hypoalbuminemia group in our study, it appears that albumin itself is not a determinant of mortality, but a surrogate marker of a worse baseline condition.

One of the strengths of our study is the fact that it is a prospective cohort study with predefined inclusion criteria in patients with NVUGB receiving early endoscopic therapy that do not have evident CLD, ESRD, or cancer. We also found that the serum albumin level (< 3.2g/dl) upon admission was a marker of increased in-hospital mortality. The main limitations of our study were that we did not evaluate the nutritional status of the patients, which may be an important factor causing hypoalbuminemia,29 and we were not able to assess 30-day mortality, which may have resulted in an underestimated mortality, since the overall hospital mortality rate was not significantly higher than that reported in other studies.

In conclusion, hypoalbuminemia appears to be an important surrogate marker of poor clinical condition, which subsequently suggests a poor outcome in NVUGB, with an overall performance for identifying mortality similar to that of the Rockall score.

No financial support was received in relation to this article.

The authors declare that there is no conflict of interest.