Predictors of the need for surgery in upper gastrointestinal bleeding in a resource constrained setting: the Pietermaritzburg experience

This review from a tertiary centre in South Africa aims to describe the spectrum and outcome of upper gastrointestinal bleeding (UGIB) and identify risk factors for surgical management and death.

This was a retrospective review of a prospectively entered database of all adults presenting with UGIB between December 2012 and December 2016. Demographics, presenting physiology, risk assessment scores, outcomes of endoscopy endo-therapy and surgery were reviewed. Comparisons were made between patients who required operative therapy and those who did not, and between survivors and non-survivors.

During the review period, 632 patients were admitted with suspected UGIB. Out of these, 406 (64%) had an identifiable potential source of bleeding and 226 (36%) had no identifiable potential source of UGIB. The latter were excluded from further analysis. Of the 406 patients with a potential source of haemorrhage, there were 249 males (61%) and 157 females (39%). Nine of these were expedited directly to the operating room and never underwent an endoscopy. Of the 397 (98%) who had upper endoscopy 107 (26%) had endotherapy. Forty-six patients (11%) required surgery. They had significantly higher shock index (SI), increased need for transfusion, higher international normalised ratio (INR) and higher serum lactate than the non-operative group. Nine patients went to the operating room without endoscopy. Of the 46 patients who required surgery, 37 underwent an attempt at endoscopic intervention. Transfusion and transfusion volume increased the probability of requiring a laparotomy (p = 0.015) and (0.003) respectively. The independent predictors of need for operation were a raised shock index or serum lactate and Forrest Ia and Ib ulcers. Thirty-nine patients died, giving a mortality rate of 9.6%; ten had a gastric ulcer and 16 had a duodenal ulcer. Survival was significantly higher in the non-operative group (93.1% versus 68.2%; p < 0.001). The odds ratio for mortality in the laparotomy group is 6.73, 95% CI (3.15-14.17). Receiver operator curve (ROC) analysis showed that the pre-endoscopic Rockall score (PER), total Rockall score (TR) and the SI were poor predictors of mortality.

Patients with UGIB in our setting are younger than in high-income countries (HIC) and a larger number fail endoscopic therapy and require open surgery. The mortality in this subset is very high. Detailed analysis of failed endotherapy has the potential to reduce mortality.