Cancer, platelet distribution width, and total protein levels as predictors of rebleeding in upper gastrointestinal bleeding

Study design and population

This study was a retrospective, observational, and single-center study conducted at the Health Sciences University, Istanbul Training and Research Hospital among patients consecutively admitted to the emergency department over 2 years. Institutional Review Board approval was obtained before the study by Istanbul Training and Research Hospital Ethics Committee (no: 1100).

Selection criteria

The inclusion criteria were (a) over 18 years of age and (b) a clinical diagnosis of UGIB. UGIB was identified clinically in patients presenting with complaints suggesting UGIB after confirmation of diagnosis by examination and/or endoscopy. Trauma patients, patients who arrested within the first 24 h, patients with active lower gastrointestinal bleeding, patients with gastrointestinal cancer, or patients with insufficient information on the hospital system were excluded.

Data collection

Patient demographics, chief complaints for emergency admission, blood pressure and heart rate at presentation, underlying diseases, history of medication (i.e., non-steroidal anti-inflammatory drugs [NSAIDs], steroids, anticoagulants, and antiplatelets), rectal examination findings, nasogastric tube contents, first laboratory results, Forrest score, bleeding origin, blood transfusion status, rebleeding rate, in-hospital mortality data, and 30-day mortality data were collected through the hospital information system.

Patients presenting to the emergency department with direct manifestations of UGIB (i.e., hematemesis or melena), as well as patients with indirect signs of blood loss (i.e., syncope, shock, dizziness, anemia, tachycardia, or pallor), were included if they were diagnosed with UGIB clinically and/or using endoscopy.

Patient hemodynamic status was reported as stable or unstable according to the first recorded data. Patients with systolic blood pressure above 90 mmHg and pulse rate lower than 100 beats/min was defined as stable, and those with systolic blood pressure above 90 mmHg or pulse rate greater than 100 beats/min was defined as unstable.

All patients were managed according to the UGIB treatment protocol of the hospital. Every patient in the study underwent endoscopy within 24 h, and the findings were collected in terms of Forrest classification (active spurting bleeding (1a), active oozing bleeding (1b), vessel exposure (2a), adherent clot (2b), and clean base ulcer (3)) (Forrest, Finlayson & Shearman, 1974) and the etiology of the bleeding.

Rebleeding was defined as hematemesis, hematochezia, and/or melena with hemodynamic instability or decreased hemoglobin to at least 2 g/dl within 24 h after the initial bleeding episode within 7 days after successful endoscopic therapy.

In-hospital mortality was defined as death during the index hospital stay, and 30-day mortality was defined as all-cause mortality within 30 days after the first emergency admission.

The primary outcome of the study was a rebleeding rate within 7 days after the first successful endoscopic intervention.

Statistical analyses

Statistical analysis was performed using the SPSS 27.0 program for Windows (SPSS, Chicago, IL, USA). The normal distribution of variables was determined using the Shapiro–Wilk test. For descriptive statistics, continuous variables were presented as median values (interquartile range (IQR)), and nominal and categorical variables were presented as percentages or counts. Univariate analyses were performed using Fisher’s exact test, χ² test, and independent sample t-test. Logistic regression analysis was performed to verify the results of the univariate analysis according to rebleeding rate using odds ratio (OR) and 95% confidence interval (CI). A 5-fold cross-validation was performed to evaluate model performance. The area under receiver operating characterics (AUROC) was performed for internal validation analysis. An alpha value of <0.05 was considered statistically significant (adjusted for multivariate analysis).

Characteristics of patients

As shown on Fig. 1; a total of 495 patients with UGIB over the age of 18 years were identified, and 124 were excluded due to early arrest (n = 46) or insufficient information (n = 78).

As shown in Table 1, a total of 371 patients (253 males, 118 females; median age 61 years, IQR 48–82) were analyzed. The patients were divided into two groups: the rebleeding and non-rebleeding patient groups. A total of 316 (85.2%) patients were in the non-rebleeding group, whereas 55 (14.8%) patients were included in the rebleeding group. The major chief complaints that led the patients to emergency were bloody and coffee-ground vomitus (n = 181, 48.8%) and dark tarry stool (n = 123, 33.2%). The rebleeding group showed high proportions of bloody and coffee-ground vomitus (p = 0.004). A total of 62 patients (16.7%) showed no bleeding. No patient with dizziness, palpitations, abdominal pain, or syncope encountered rebleeding. The most common co-morbidities were hypertension (n = 121, 32.6%) and diabetes mellitus (n = 61, 16.4%). Patients with cancer had a significantly high rate of rebleeding (p < 0.001). Among the groups, no significant differences in terms of NSAID, antiplatelet, or anticoagulant medication were found. In the rebleeding group (n = 14, 25.5%), a higher proportion of patients showed unstable hemodynamics at presentation than in the non-rebleeding group (n = 55, 17.4%); however, the difference was insignificant (p = 0.157). No significant difference in rectal examination findings was found among the groups; however, the rebleeding group had a higher proportion of patients with blood in the gastric tube than the non-rebleeding group (p = 0.019).

In the laboratory findings at presentation (Table 2), mean cell hemoglobin concentration (p = 0.023), red cell distribution width (RDW; p = 0.003), neutrophil count (p = 0.013), lymphocyte count (p = 0.004), platelet distribution width (PDW; p = 0.003), albumin (p < 0.001), total protein (p = 0.006), and neutrophil-to-lymphocyte ratio (p < 0.001) were significantly different between groups. Both in-hospital mortality and 30-day mortality were significantly higher in the rebleeding group (p < 0.001).

Endoscopic findings

As shown on Table 3, in this study, the most common endoscopic finding was gastritis (n = 101, 27.2%). Interestingly, the most common endoscopic finding in the rebleeding group was peptic ulcers (n = 21, 38.1%), and the proportion was significantly higher than that in the non-rebleeding group (p = 0.012).

Multivariate analysis

Given that significant differences were found between the groups in the above-mentioned parameters through univariate analysis, a multivariate stepwise forward logistic regression analysis was conducted (Table 4). The Hosmer-Lemeshow goodness of fit test showed a χ² value of 6.94 and a p value of 0.543. The model explained 24,4% of variance in rebleeding (Nagelkerke R²). The performance of the model yielded an accuracy of 84,9% and area under curve (AUC) value of 0.754 (95% CI [0.683–0.826]). Analysis revealed that PDW (p = 0.006, OR = 1.31, 95% CI [1.08–1.59]), total protein (p = 0.045, OR = 0.51, 95% CI [0.27–0.99]), and cancer (p = 0.015, OR = 3.64, 95% CI [1.28–10.3]) were independent predictive variables for rebleeding in UGIB. A 5-fold cross validation was performed to analyze the model (Table 5).

Limitations

This study has several limitations. It is a single center, observational, retrospective study. Due to the nature of the condition; it is impossible to determine the exact time bleeding has started and time elapsed between the start of bleeding and presentation to emergency department may vary among patients. The sample size is small (n = 371) and the data is imbalanced (14.8% in the rebleeding group) for a precise multivariate analysis. Without a separate validation set to replicate these findings; it is hard to conclude PDW, total protein and cancer as independent prognostic factors.