AIMS65 scoring system is comparable to Glasgow-Blatchford score or Rockall score for prediction of clinical outcomes for non-variceal upper gastrointestinal bleeding

From January 2013 to June 2016, patients for the UGI bleeding who visited the emergency medical center of the Sanggye Paik Hospital, Seoul, South Korea were reviewed retrospectively in this study. UGI bleeding was confirmed by the hospital staff based on the presence of hematemesis or melena. All patients underwent emergency upper endoscopy within 24 h. Patients with esophageal or gastric variceal bleeding were excluded. Cases of iatrogenic post-procedural bleedings after endoscopic resection for gastric tumors were also excluded. The data of the patients were collected in the electronic medical record. Patients were excluded if the data required for calculation of relevant risk stratification scores were unavailable. The data were reviewed by two endoscopy specialists (M.S.K. and J.C.) for all cases.

Patients received the same treatment according to emergency bleeding protocols. All patients received intravenous proton pump inhibitor infusion before upper endoscopy. Upper endoscopy was performed by endoscopy specialists within the first 24 h. Endoscopic therapies were applied as following conditions: (1) peptic ulcers with actively bleeding (type Ia), oozing hemorrhage (type Ib), or a nonbleeding visible vessel (type IIa); (2) Dieulafoy’s lesion (3) angiodyplasia (4) and any lesions with active bleeding. Adherent blood clot (type IIb), flat spot (type IIc), clean base ulcer (type III), or lesions with no active bleeding stigmata including acute gastric mucosal lesion or Mallory-Weiss tear were initially treated medication without endoscopic therapy.

Endoscopic therapies included injection of epinephrine, hemostatic forcep electrocoagulation, argon plasma coagulation, or application of endoscopic clips. In cases of failed endoscopic hemostasis, transarterial embolization was preferred over surgery. When rebleeding occurred after successful endoscopic therapy, endoscopic therapy was initially preferred. Surgery as salvage therapy was performed when embolization was unsuccessful.

Transfusion was required if hemoglobin was below 8 g/dL. ICU admission was considered in the following conditions: if the patients have any symptoms of confusion or altered mentality; presence of hemodynamic instability (systolic blood pressure < 90 mmHg, need for vasoactive drugs); or severe comorbid illness including heart failure, chronic renal failure, liver cirrhosis, or chronic lung disease. Decision of ICU admission was made finally by ICU unit attending physicians.

The following data were collected through electronic medical record review: age, sex, symptoms of visiting emergency centers (hematemesis, melena, shock, syncope, or altered mentality), vital sign (heart rate, systolic blood pressure), the Glasgow Coma Scale, mental status, medications that contribute to bleeding (aspirin, clopidogrel, warfarin, nonsteroidal anti-inflammatory drugs, other antithrombotic agents). Laboratory findings (hemoglobin, albumin level, blood urea nitrogen, PT, and INR), and endoscopic findings were also collected. The AIMS65 score, RS, PRS, and GBS were obtained based on the data. Components of the each scoring system were described in Table 1. Primary outcome was in-hospital mortality and secondary outcomes were composite clinical outcomes of in-hospital mortality, intensive care unit (ICU) admission; rebleeding; blood transfusion requirement; and endoscopic, radiologic, or surgical intervention.

The performance of each scoring system was evaluated with a receiver operating characteristic (ROC) curve, calculation of area under the curve (AUC) with 95% confidence intervals. In a ROC curve the true positive rate (sensitivity) is plotted in function of the false positive rate (100-specificity) for different cut-off points of a parameter. Each point on the ROC curve represents a sensitivity/specificity pair corresponding to a particular decision threshold. The cut-off value of each scoring system is determined by the ROC curve with the most specificity and sensitivity. The risk prediction scoring system was divided into two groups, high-risk group and low-risk group. For each scoring systems, a cutoff point was calculated that maximized the sum of sensitivity and specificity in predicting the primary and secondary outcomes. Fischer’s exact test was used for categorical variables. All confidence intervals are described as two-sided binomial 95% confidence intervals. The cut-off value 0.05 considered to be the threshold for statistical significance. In a Bonferroni correction to adjust for the multiple comparisons in the scoring systems (in which the P values were divided by 4 for the comparison with an alpha level of 0.05), the corrected P values were 0.0125 for the combination group versus. Data were analyzed using SPSS Statistics for Windows (version 23.0; IBM Corp., Armonk, NY, USA). AUCs were compared with the DeLong method by using MedCalc program (version 16.8.4; MedCalc Software, Mariakerke, Belgium).

A total of 578 patients were enrolled, of which consecutive 512 patients were included in the final analysis. Of these, 66 patients were excluded from the study for the following reasons: 30 patients did not have sufficient data for the study; 16 had iatrogenic post-procedural bleedings (endoscopic submucosal dissection for gastric tumors (n = 11), endoscopic mucosal resection (n = 2), endoscopic sphincterotomy (n = 3)), and 20 patients were lost to follow-up (Fig. 1).

The median age was 64 (range, 48–80 years) years old, and 71.9% patients were men. Among patients, 397 (69.7%) patients had comorbidities and 327 (36%) were taking anti-platelet agents or anti-coagulant medications on admission. The most common symptom of visiting the emergency center was melena (27.5%). The common causes of bleeding were gastric ulcer (32.8%), duodenal ulcer (20.5%), Mallory-Weiss tear (13.1%), and acute gastric mucosal lesion (12.9%) (Table 2).

Seventeen of the 512 patients (3.3%) died. Their median age was 70.24 (range 43–93) years old. The causes of death were uncontrolled bleeding (n = 5), complications due to cirrhosis (n = 5), sepsis due to pneumonia (n = 4), renal failure (n = 2), and cerebral infarction (n = 1). In the cases of 5 deaths due to uncontrolled bleeding, two patients underwent angiographic embolization and two patients underwent angiography followed by surgery. One patient died of active duodenal ulcer bleeding and hypovolemic shock during endoscopic therapy.

All but one of the 17 patients who died had comorbidities. There was no difference between survivors and non-survivors in the use of anticoagulants. The mortality increased with increasing AIMS65 score, although death occurred in 1 patient who scored 0 on the AIMS65. Mortality was seen in 1/161 (0.6%) for AIMS65 0, 1/201 (0.5%) for AIMS65 1, 6/104 (5.8%) for AIMS65 2, 6/36 (16.7%) for AIMS65 3, 3/9 (33.3%) for AIMS65 4, 0/1 for AIMS65 5. The AUC values of each test were: AIMS65 = 0.84 (95% confidence interval (CI), 0.81–0.88), PRS = 0.74 (95% CI, 0.70–0.78), RS = 0.75 (95% CI, 0.71–0.79), and GBS = 0.72 (95% CI, 0.68–0.76). With regard to AUC value, there was a trend suggesting that the AIMS65 scoring system (0.84) seemed more accurate than the GBS system (0.72) for predicting mortality (P = 0.07) (Table 3) (Fig. 2).

The cut-off values for the endpoints of the risk stratification scores from the AIMS65, PRS, RS, and GBS were obtained when the cut-off value that maximized the sum of the sensitivity and the specificity was obtained. Sensitivity of AIMS65 was from 41.5% (CI 19.4–54.4, p < 0.001) to 88.2% (CI 63.6–98.5, p < 0.0001), which was similar to other scoring systems, ranging from 71.3 (CI 68.1–76.6, p < 0.001) to 78.6 (CI 74.1–82.6, p < 0.001), respectively. Cut-off as a value separating risk levels (high vs. low risk) for death was above 2 points on the AIMS65 and 8 points on the GBS, and 1 point on the AIMS65 and 11 points on the GBS for rebleeding. The cut-off values for ICU admission were 2 points on the AIMS65, 10 points on the GBS, 4 points on the PRS, and 5 points on the RS (Table 4).