Introduction: Bleeding from the upper gastrointestinal tract (UGIB) is a frequent emergency in which rapid evaluation and treatment are needed to reduce complications and deaths. Over time, care has shifted away from surgery and toward endoscopic therapy, a change that matters most in low-resource hospitals where surgical capacity is limited. Aim: We set out to characterize patients treated for UGIB at a tertiary hospital in Peshawar, Pakistan, from January 2024 to December 2025 in terms of presentation, bleeding source, endoscopic treatment, and outcomes, and to determine which factors predicted admission to the intensive care unit (ICU). Methods: In this retrospective review, we analyzed 235 patients who had endoscopy for UGIB, using demographic, clinical, etiological, and treatment data taken from hospital records. Rebleeding within 4 weeks, in-hospital death, and readmission within 30 days were the primary outcomes; ICU admission, hospital stay, transfusion requirement, repeat endoscopy, interventional radiology (IR), and surgery were secondary outcomes. Predictors of ICU admission were examined with binary logistic regression. Results: Patients had a mean age of 45.5 ± 17.1 years, and 61.3% were men. Duodenal ulcer was the single most frequent bleeding source (28.1%), whereas gastric and esophageal varices together accounted for 40.9% of cases. The commonest procedures were band ligation (19.6%), adrenaline injection (16.2%), and hemoclip placement (12.8%). Rebleeding within 4 weeks affected 18.7% of patients and occurred more often with variceal than with non-variceal bleeding (28.1% vs. 12.2%, p = 0.004). Readmission within 30 days occurred in 16.6%, again more often after variceal bleeding (32.3% vs. 5.8%, p < 0.001). Repeat endoscopy was performed in 15.3%, IR in 1.7%, and surgery in 0.9%. ICU care was needed in 23.4%; on average, patients stayed 5.31 ± 2.60 days and received 2.17 ± 1.72 units of packed red cells, and 1.3% died in hospital. On multivariable analysis, ICU admission was more likely with lower admission hemoglobin (OR 0.68; 95% CI 0.52–0.89; p = 0.004), faster heart rate (OR 1.03; 95% CI 1.01–1.05; p = 0.013), variceal etiology (OR 2.25; 95% CI 1.16–4.37; p = 0.017), and female sex (OR for male sex 0.46; 95% CI 0.24–0.89; p = 0.020). Conclusion: In this resource-limited tertiary care setting, an endoscopy-based approach to UGIB was associated with low rates of surgery, IR, and in-hospital death. Variceal bleeding formed the highest-risk group, with more rebleeding and readmission, pointing to the need for structured secondary prophylaxis and follow-up. Admission hemoglobin and heart rate are simple bedside measures that may help identify patients who need intensive care.
Acute bleeding from the upper gastrointestinal tract (UGIB) ranks among the commonest emergencies in gastroenterology. It carries a considerable burden of morbidity, recurrent bleeding, blood transfusion, prolonged hospitalization, and death. Presentation ranges from minor mucosal oozing to massive hemorrhage with hemodynamic instability, so prompt resuscitation, risk assessment, and timely endoscopy are the core of modern management. Guidelines now advise that upper endoscopy be carried out within the first 24 hours after presentation, once the patient is hemodynamically stable, and that lesions with high-risk stigmata be treated endoscopically [1–3].
Approaches to UGIB have evolved markedly in recent decades. Patients with uncontrolled or recurrent bleeding, particularly from peptic ulcers or large varices, were once treated mainly with surgery. Endoscopy has since become the principal therapeutic tool, and care is now largely endoscopy-centered [2–5]. For most patients, endoscopic hemostasis is now the initial treatment of choice, using adrenaline injection, hemoclips, band ligation, sclerotherapy, thermal coagulation, or a combination of these. In non-variceal bleeding, repeat endoscopy is the usual first step when bleeding recurs; transcatheter arterial embolization is reserved for cases in which endoscopy fails or is unavailable, and surgery for patients in whom other options have failed [5–8].
This shift is particularly relevant to resource-limited tertiary care hospitals, where surgical care is constrained by limited operating room time, higher perioperative risk, cost, late presentation, and shortages of specialist staff. Previous studies indicate that an efficiently run emergency endoscopy service can lower the need for surgery in these settings [8–11]. Interventional radiology (IR) provides a further minimally invasive rescue option, but in many low- and middle-income countries it is not readily available [12]. Real-world data on the outcomes of endoscopic management in resource-constrained hospitals are therefore needed to guide service planning and reduce UGIB-related morbidity and mortality [13]. National audits further suggest that how endoscopy services are organized, for example whether endoscopy is available after hours and whether standard endoscopic therapy is applied consistently, influences the quality of UGIB care [14].
Our objectives were to profile patients treated for UGIB at our hospital in terms of their background and presentation, to identify the leading causes of bleeding, to document the endoscopic treatments used, and to assess rebleeding, in-hospital death, readmission within 30 days, intensive care unit (ICU) admission, repeat endoscopy, IR, and surgery. A further aim was to determine which factors were independently related to ICU admission. Such local data may help reduce reliance on surgery and strengthen UGIB care pathways in similar resource-limited settings.
This retrospective observational study was carried out at a tertiary care hospital in Peshawar, Pakistan, using records for the period between January 2024 and December 2025. Over these 2 years, the hospital performed 1,152 upper gastrointestinal endoscopies.
Patients qualified for inclusion if they were 16 years of age or older, had been admitted with UGIB, and had undergone upper endoscopy. Those with suspected lower gastrointestinal bleeding or incomplete outcome data were excluded, which left 235 patients for analysis.
From hospital records, we obtained demographics, presenting features (melena, hematemesis, and heart rate), admission hemoglobin, endoscopic findings, treatment, blood transfusion, and outcomes. A heart rate of more than 100 beats per minute (bpm) was considered tachycardia. The bleeding source was recorded as variceal (esophageal or gastric varices), peptic ulcer (duodenal or gastric), Mallory–Weiss tear, esophagitis, Dieulafoy lesion, angiodysplasia, or gastric malignancy. For comparative analyses, esophageal and gastric varices were grouped together as variceal bleeding.
All patients underwent endoscopy. Endoscopic treatment was chosen according to the lesion found and included adrenaline injection, hemoclips, heater probe thermocoagulation, band ligation, cyanoacrylate injection, argon plasma coagulation (APC), and combination therapy. Selected patients with variceal bleeding also received intravenous terlipressin, either alone or with band ligation.
Three primary outcomes were assessed: rebleeding, death in hospital, and readmission within 30 days. Rebleeding meant a clinically significant recurrence of bleeding after initial hemostasis had been achieved, occurring up to 4 weeks after the index endoscopy. In-hospital mortality covered deaths from any cause before discharge from the index admission, and readmission covered any unplanned return to hospital within 30 days of discharge. As secondary outcomes, we recorded ICU admission, duration of hospitalization, the number of red cell units transfused, repeat endoscopy, IR intervention, and surgery.
Statistical analyses were run in IBM SPSS Statistics (version 27.0; IBM Corp., Armonk, NY, USA), and figures were produced in Python (version 3.13) with the Matplotlib library (version 3.11). Continuous data are summarized as mean ± standard deviation (SD), with median and interquartile range (IQR) added where helpful, and categorical data as number and percentage. Because several continuous variables were skewed, nonparametric tests were applied: the Mann–Whitney U test when two groups were compared, and the Kruskal–Wallis H test when three or more were compared. Categorical comparisons used the chi-square test with Yates’ correction, replaced by Fisher’s exact test whenever an expected cell frequency was less than 5. The strength of the relationship between admission hemoglobin and units transfused was expressed as Pearson’s r.
Independent predictors of ICU admission were identified with a binary logistic regression model, into which six covariates (age, admission hemoglobin, heart rate, units transfused, variceal etiology, and sex) were entered simultaneously. Associations are expressed as odds ratios (ORs) and accompanying 95% confidence intervals (CIs). The omnibus chi-square test, Nagelkerke R2, and the Hosmer–Lemeshow test were used to evaluate the model. Missing values were not imputed, so only patients with complete covariate data contributed to the model. Each outcome was additionally compared between the variceal and non-variceal bleeding groups. Significance was set at a two-sided p < 0.05.
Ethical approval for the study was granted by the Ethical Review Committee for Medical and Biomedical Research, Combined Military Hospital, Peshawar, Pakistan (approval No. 284/2026). Records were anonymized before any analysis.
Demographic and Baseline Characteristics
Over the 2-year period, 235 patients with acute UGIB were studied: 144 men (61.3%) and 91 women (38.7%), giving a male-to-female ratio of 1.58:1 (Table 1). Their mean age was 45.5 ± 17.1 years (median 47 years, IQR 32–56; range 16–85).
The mean heart rate at admission was 102.0 ± 17.3 bpm, and 126 of the 230 patients with a recorded pulse (54.8%) were tachycardic. Mean admission hemoglobin was 8.24 ± 2.08 g/dL, indicating moderate anemia on average. Because the data were collected retrospectively, hemoglobin was missing for 12 patients (5.1%) and pulse for 5 (2.1%). Melena was the most common presenting feature (n = 142, 60.4%), followed by hematemesis (n = 137, 58.3%); at least 44 patients had both.
Table 1. Characteristics of Patients at Presentation (N = 235).
|
Characteristic |
Value |
Missing, n (%) |
|
Demographics |
||
|
Age, years, M ± SD |
45.5 ± 17.1 |
0 (0.0) |
|
Median [IQR] |
47 [32–56] |
— |
|
Range |
16–85 |
— |
|
Male sex, n (%) |
144 (61.3) |
0 (0.0) |
|
Female sex, n (%) |
91 (38.7) |
0 (0.0) |
|
Clinical presentation |
||
|
Hemoglobin on admission, g/dL, M ± SD |
8.24 ± 2.08 |
12 (5.1) |
|
Heart rate, bpm, M ± SD |
102.0 ± 17.3 |
5 (2.1) |
|
Tachycardia (>100 bpm), n (%) a |
126 (54.8) |
5 (2.1) |
|
Melena, n (%) |
142 (60.4) |
0 (0.0) |
|
Hematemesis, n (%) |
137 (58.3) |
0 (0.0) |
Note. Data are presented as mean ± SD, median [IQR], or n (%). IQR, interquartile range; bpm, beats per minute. a Percentage of patients with a recorded heart rate (n = 230).
Etiology of Upper Gastrointestinal Bleeding
The leading single cause of bleeding was duodenal ulcer (n = 66, 28.1%), followed by gastric varices (n = 51, 21.7%) and esophageal varices (n = 45, 19.1%) (Table 2). Taken together, variceal bleeding accounted for 96 cases (40.9%), making it the largest etiological group. Gastric ulcer (n = 27, 11.5%) and Mallory–Weiss tear (n = 22, 9.4%) were the next most frequent causes. Less common causes were esophagitis (n = 8, 3.4%), Dieulafoy lesion (n = 6, 2.6%), angiodysplasia (n = 5, 2.1%), and gastric malignancy (n = 5, 2.1%).
Admission hemoglobin differed significantly across etiological groups (Kruskal–Wallis H(8) = 15.70, p = 0.047).
Table 2. Distribution of Bleeding Sources (N = 235).
|
Etiology |
n |
% |
Cumulative % |
|
Duodenal ulcer |
66 |
28.1 |
28.1 |
|
Gastric varices |
51 |
21.7 |
49.8 |
|
Esophageal varices |
45 |
19.1 |
68.9 |
|
Gastric ulcer |
27 |
11.5 |
80.4 |
|
Mallory–Weiss tear |
22 |
9.4 |
89.8 |
|
Esophagitis |
8 |
3.4 |
93.2 |
|
Dieulafoy lesion |
6 |
2.6 |
95.7 |
|
Angiodysplasia |
5 |
2.1 |
97.9 |
|
Gastric malignancy |
5 |
2.1 |
100.0 |
|
Total |
235 |
100.0 |
— |
Note. Variceal bleeding (gastric and esophageal varices combined) accounted for 96 cases (40.9%).
Endoscopic Interventions
All patients underwent diagnostic endoscopy, and treatment was tailored to the lesion found (Table 3). Band ligation was the most frequent procedure (n = 46, 19.6%) and was used mainly for esophageal variceal bleeding. Adrenaline injection was used in 38 patients (16.2%) with peptic ulcers that were actively bleeding or carried signs of recent bleeding; hemoclips were applied in 30 (12.8%), and adrenaline injection combined with heater probe thermocoagulation in 24 (10.2%). The six patients (2.6%) with a Dieulafoy lesion received adrenaline injection plus hemoclips.
Among patients with variceal bleeding, 18 (7.7% of the cohort) received cyanoacrylate injection for gastric varices, 16 (6.8%) received band ligation with intravenous terlipressin for active bleeding with hemodynamic compromise, and 16 (6.8%) received intravenous terlipressin alone. Twenty-three patients (9.8%), mostly with Mallory–Weiss tears or esophagitis in whom spontaneous hemostasis was expected, were managed conservatively after endoscopic assessment, and a further 10 hemodynamically stable patients (4.3%) required no specific intervention. APC was used in 5 patients (2.1%) with angiodysplasia, and palliative APC in 3 (1.3%) with gastric malignancy.
Table 3. Endoscopic and Therapeutic Interventions (N = 235).
|
Intervention |
n |
% |
Main indication |
|
Band ligation |
46 |
19.6 |
Esophageal varices |
|
Adrenaline injection |
38 |
16.2 |
Peptic ulcer |
|
Hemoclip application |
30 |
12.8 |
Peptic ulcer |
|
Adrenaline + heater probe |
24 |
10.2 |
Actively bleeding ulcer |
|
Conservative management |
23 |
9.8 |
Mallory–Weiss tear, esophagitis |
|
Cyanoacrylate injection |
18 |
7.7 |
Gastric varices |
|
Band ligation + IV terlipressin |
16 |
6.8 |
Active variceal bleeding |
|
IV terlipressin alone |
16 |
6.8 |
Gastric varices |
|
No intervention |
10 |
4.3 |
Hemodynamically stable |
|
Adrenaline + hemoclip |
6 |
2.6 |
Dieulafoy lesion |
|
APC |
5 |
2.1 |
Angiodysplasia |
|
Palliative APC |
3 |
1.3 |
Gastric malignancy |
|
Total |
235 |
100.0 |
— |
Note. Percentages are of the total cohort. Band ligation includes procedures for both esophageal and gastric varices. Conservative management refers to endoscopic assessment followed by medical therapy alone; no intervention refers to hemodynamically stable patients who required no specific treatment. APC, argon plasma coagulation; IV, intravenous.
Blood Transfusion Requirements
On average, patients received 2.17 ± 1.72 units of packed red cells (range 0–8). The higher the admission hemoglobin, the fewer units were transfused, a strong inverse association (r = −0.713, p < 0.001; Figure 1B).
Clinical Outcomes
ICU admission. Fifty-five patients (23.4%) required ICU admission. Compared with patients managed on the ward, they had a lower mean admission hemoglobin (7.55 ± 2.14 g/dL vs. 8.46 ± 2.03 g/dL; p = 0.012) and a faster heart rate (109.4 ± 18.1 bpm vs. 99.8 ± 16.6 bpm; p < 0.001; Table 4). ICU admission rates were 28.1% in variceal and 20.1% in non-variceal bleeding, a difference that was not statistically significant on univariable analysis (χ2(1) = 1.60, p = 0.206). After adjustment for other covariates, however, variceal etiology was independently associated with ICU admission (see below).
Length of stay. Patients spent a mean of 5.31 ± 2.60 days in hospital. Those admitted to the ICU stayed significantly longer than ward patients (7.98 ± 2.71 vs. 4.48 ± 2.06 days, p < 0.001; Figure 1A). Length of stay also differed across the five most frequent etiological groups (Kruskal–Wallis H(4) = 13.27, p = 0.010).
Rebleeding. Forty-four patients (18.7%) rebled within 4 weeks. Rebleeding was more frequent in variceal than in non-variceal bleeding (28.1% vs. 12.2%; χ2(1) = 8.41, p = 0.004). Rebleeding rates were 29.4% for gastric varices and 26.7% for esophageal varices, compared with 18.2% for duodenal ulcer, 11.1% for gastric ulcer, and 4.5% for Mallory–Weiss tear (Figure 1C).
Readmission. Thirty-nine patients (16.6%) were readmitted within 30 days. Readmission was considerably more common after variceal bleeding (32.3% vs. 5.8%; χ2(1) = 27.00, p < 0.001).
Repeat endoscopy, interventional radiology, and surgery. Repeat endoscopy was required in 36 patients (15.3%), IR intervention in 4 (1.7%), and surgery in 2 (0.9%).
In-hospital mortality. Three patients (1.3%) died in hospital. All three had severe hemodynamic compromise at presentation and required ICU admission; two had variceal bleeding and one had a non-variceal source. Mortality was therefore higher among ICU patients (5.5% vs. none on the ward; p = 0.012, Fisher’s exact test).
Table 4. Clinical Characteristics and Outcomes by ICU Admission Status.
|
Variable |
Total (N = 235) |
ICU (n = 55) |
Ward (n = 180) |
p |
|
|
Continuous variables |
|||||
|
Hemoglobin, g/dL |
8.24 ± 2.08 |
7.55 ± 2.14 |
8.46 ± 2.03 |
0.012* |
|
|
Heart rate, bpm |
102.0 ± 17.3 |
109.4 ± 18.1 |
99.8 ± 16.6 |
<0.001** |
|
|
Units transfused |
2.17 ± 1.72 |
2.58 ± 1.84 |
2.04 ± 1.67 |
0.151 |
|
|
Hospital stay, days |
5.31 ± 2.60 |
7.98 ± 2.71 |
4.48 ± 2.06 |
<0.001** |
|
|
Categorical variables |
|||||
|
Rebleeding within 4 weeks |
44 (18.7) |
18 (32.7) |
26 (14.4) |
0.004* |
|
|
Readmission within 30 days |
39 (16.6) |
16 (29.1) |
23 (12.8) |
0.008* |
|
|
In-hospital mortality |
3 (1.3) |
3 (5.5) |
0 (0.0) |
0.012*a |
|
Note. Values are given as mean ± SD or n (%). Continuous variables were tested with the Mann–Whitney U test and categorical variables with the Yates-corrected chi-square test, except where indicated. ICU, intensive care unit; bpm, beats per minute. a Fisher’s exact test. * p < 0.05. **p < 0.001.
Predictors of ICU Admission
Overall, the regression model was significant (χ2(6) = 25.65, p < 0.001; Nagelkerke R2 = 0.16), and calibration was acceptable (Hosmer–Lemeshow p = 0.622). Table 5 shows the four variables that remained independently associated with ICU admission. For every 1 g/dL rise in admission hemoglobin, the odds of ICU admission fell by 32% (OR 0.68; 95% CI 0.52–0.89; p = 0.004); conversely, every 1 g/dL drop raised the odds by roughly 47%. The odds increased by about 2.5% with each additional beat per minute of heart rate (OR 1.03; 95% CI 1.01–1.05; p = 0.013) and were more than twice as high with variceal etiology (OR 2.25; 95% CI 1.16–4.37; p = 0.017). Men had lower odds of ICU admission than women (OR 0.46; 95% CI 0.24–0.89; p = 0.020). Neither age nor units transfused predicted ICU admission.
Table 5. Multivariable Predictors of ICU Admission (Binary Logistic Regression).
|
Predictor |
B |
OR |
95% CI |
Wald χ2 |
p |
|
Age (per year) |
−0.006 |
0.994 |
0.974–1.015 |
0.30 |
0.582 |
|
Hemoglobin (per g/dL) |
−0.388 |
0.68 |
0.52–0.89 |
8.13 |
0.004* |
|
Heart rate (per bpm) |
0.025 |
1.03 |
1.01–1.05 |
6.14 |
0.013* |
|
Units transfused (per unit) |
−0.174 |
0.84 |
0.64–1.11 |
1.52 |
0.218 |
|
Variceal etiology |
0.810 |
2.25 |
1.16–4.37 |
5.71 |
0.017* |
|
Male sex |
−0.775 |
0.46 |
0.24–0.89 |
5.39 |
0.020* |
Note. Model χ2(6) = 25.65, p < 0.001; Nagelkerke R2 = 0.16; Hosmer–Lemeshow p = 0.622. Non-variceal etiology and female sex served as the reference categories. B, regression coefficient; OR, odds ratio; CI, confidence interval; ICU, intensive care unit; bpm, beats per minute. * p < 0.05.
Comparison by Bleeding Source
Of the 235 patients, 96 (40.9%) had variceal and 139 (59.1%) had non-variceal bleeding (Table 6). Rebleeding (χ2(1) = 8.41, p = 0.004) and readmission (χ2(1) = 27.00, p < 0.001) were both significantly more frequent in the variceal group. By contrast, the two groups were similar in ICU admission (28.1% vs. 20.1%; p = 0.206), hospital stay (5.84 ± 2.91 vs. 4.95 ± 2.34 days; p = 0.082), units transfused (2.44 ± 1.81 vs. 1.99 ± 1.63; p = 0.148), and in-hospital mortality (2.1% vs. 0.7%; p = 0.569).
Table 6. Clinical Outcomes Stratified by Bleeding Source.
|
Outcome |
Variceal |
Non-variceal |
Test statistic |
p |
|
ICU admission, n (%) |
27 (28.1) |
28 (20.1) |
χ2(1) = 1.60 |
0.206 |
|
Hospital stay, days, M ± SD |
5.84 ± 2.91 |
4.95 ± 2.34 |
U = 5,821 |
0.082 |
|
Rebleeding within 4 weeks, n (%) |
27 (28.1) |
17 (12.2) |
χ2(1) = 8.41 |
0.004* |
|
Readmission within 30 days, n (%) |
31 (32.3) |
8 (5.8) |
χ2(1) = 27.00 |
<0.001** |
|
In-hospital mortality, n (%) |
2 (2.1) |
1 (0.7) |
Fisher’s exact |
0.569 |
|
Units transfused, M ± SD |
2.44 ± 1.81 |
1.99 ± 1.63 |
U = 5,925 |
0.148 |
Note. Continuous outcomes were compared by two-tailed Mann–Whitney U tests; categorical outcomes by the chi-square test with Yates’ correction or, if any expected count was below 5, by Fisher’s exact test. ICU, intensive care unit. * p < 0.05. ** p < 0.001.
UGIB is a potentially life-threatening condition, and delays in diagnosis or treatment can have serious consequences. This study shows the central role of endoscopy in managing UGIB in a resource-limited tertiary care setting. All 235 patients underwent endoscopy, which served as both the main diagnostic and the main therapeutic tool, in keeping with the wider move toward endoscopy-first management.
The endoscopic techniques used included adrenaline injection, hemoclips, band ligation, cyanoacrylate injection, APC, and combination therapy. Repeat endoscopy was needed in 15.3% of patients, IR in 1.7%, and surgery in only 0.9%. This low surgical rate is consistent with earlier reports that therapeutic endoscopy reduces the need for surgery in UGIB [13,15]. Because the study had no comparison group, however, the low surgical rate cannot be attributed to endoscopic therapy alone.
The rebleeding rate of 18.7% was higher than the 13% reported in the 2007 UK national audit [16]. The most likely explanation is the high proportion of variceal bleeding in our cohort (40.9%; gastric varices 21.7%, esophageal varices 19.1%). As in previous work, patients with varices rebled significantly more often than those with non-variceal lesions, consistent with the higher risk of recurrence associated with portal hypertension and chronic liver disease [5]. Our 4-week follow-up window, which captures events after discharge, may also have produced a higher rate than studies that report in-hospital rebleeding only.
Timely endoscopy is central to the management of UGIB. The guideline of the European Society of Gastrointestinal Endoscopy (Gralnek et al.) advises performing endoscopy within 24 hours of presentation once the patient has been resuscitated, so that the bleeding source can be identified and treated promptly and more invasive interventions avoided [17]. Every patient in our cohort had endoscopy during the index admission; however, we did not assess whether the timing of endoscopy influenced outcomes.
Almost a quarter of patients (23.4%) required ICU admission. Compared with ward patients, they had lower hemoglobin and faster heart rates, stayed in hospital longer, rebled more often, and were more likely to die. After multivariable adjustment, lower hemoglobin, higher heart rate, and variceal etiology were independently associated with ICU admission. These are simple bedside variables that could help identify high-risk patients early, although they should complement rather than replace validated tools such as the Glasgow–Blatchford score. Variceal etiology was not significantly associated with ICU admission in the unadjusted comparison but became significant after adjustment, suggesting that group imbalances in hemoglobin, heart rate, and sex had obscured the unadjusted association. The finding that women had higher adjusted odds of ICU admission was unexpected and needs cautious interpretation; unmeasured differences in comorbidity, bleeding severity, or referral patterns could explain it.
The mean admission hemoglobin was 8.24 g/dL, and the mean transfusion requirement was 2.17 units. As expected, lower admission hemoglobin was strongly associated with a greater transfusion requirement, which supports early assessment of anemia severity in patients with UGIB.
IR was rarely used (1.7%). Transcatheter arterial embolization is an effective rescue option when endoscopic hemostasis fails [18], but its availability in low- and middle-income settings is limited by a lack of trained interventional radiologists, equipment, and funding [12]. The low rate of IR in our cohort probably reflects both effective endoscopic control and this limited access, and it suggests that endoscopic therapy, as recommended by current guidelines, was sufficient for most patients [17,19].
In-hospital mortality was low (1.3%), well below the 10% reported in the 2007 UK national audit [16]. All three deaths occurred in patients with severe hemodynamic instability who required ICU care. This is consistent with the Rockall risk model, in which shock, age, and comorbidity are among the main predictors of death after UGIB, and it suggests that disease severity at presentation, rather than failure of endoscopic treatment alone, was the main determinant of mortality [20]. The low mortality should, however, be read in light of the relatively young cohort (mean age 45.5 years) and the fact that deaths were recorded only during the index admission.
Overall, these findings support the role of therapeutic endoscopy in managing UGIB in this setting. Surgery and IR were rarely needed, and outcomes were generally favorable for both variceal and non-variceal sources. Even so, the variceal group continued to face a greater risk of rebleeding and readmission. The high readmission rate in this group may reflect the chronic, relapsing course of portal hypertension and possible gaps in secondary prophylaxis after discharge. These patients require close follow-up, with repeat endoscopy and band ligation when needed, suitable drug therapy, and long-term secondary prevention using non-selective beta-blockers combined with endoscopic surveillance [5,21].
Our study has some limitations. Being retrospective and conducted at a single center, it may not be generalizable to other settings, and it depends on the completeness of documentation; admission hemoglobin and pulse were missing for some patients. Second, without a comparison group, we cannot establish that endoscopic therapy caused the low rates of surgery and death. Third, we could not calculate validated risk scores (Glasgow–Blatchford or Rockall), and comorbidities and the timing of endoscopy were not analyzed. Fourth, the regression model included units transfused, a treatment variable that is strongly correlated with admission hemoglobin and may itself follow the decision to admit a patient to the ICU; with 55 events and six predictors, the model may also be prone to overfitting. Fifth, the small number of deaths did not allow analysis of mortality predictors. Finally, readmissions to other hospitals may have been missed. Despite these limitations, the study provides real-world data on UGIB management and outcomes from a resource-limited setting in Pakistan, where such data remain scarce.
Therapeutic endoscopy was the cornerstone of UGIB management in this resource-limited tertiary care center. All 235 patients underwent endoscopic evaluation, and treatment with band ligation, adrenaline injection, hemoclips, thermal coagulation, or combination therapy was accompanied by low rates of surgery (0.9%) and IR (1.7%). In-hospital mortality was 1.3%, despite marked anemia and frequent tachycardia at presentation.
Variceal bleeding accounted for 40.9% of cases and was associated with significantly higher rates of rebleeding and readmission than non-variceal bleeding. Lower hemoglobin, higher heart rate, variceal etiology, and female sex were independent predictors of ICU admission. These findings support early risk stratification and close follow-up, including structured secondary prophylaxis, for patients with variceal bleeding. Confirmation in prospective multicenter studies that use validated risk scores is still required.