Introduction Upper gastrointestinal variceal bleeding (UGIVB) is associated with high mortality rate.
Aim To determine predictors of outcome (42-day survival and 5-day rebleeding rate) of patients admitted with severe UGIVB.
Methods Retrospective study of all patients admitted to liver intensive therapy unit (LITU) in our centre with severe UGIVB 01/01/2000 to 31/03/2008.
Results 177 patients, 157 analysed. Male 90/157 (57%). Median age was 49 years (IQR 39–57). Alcohol related liver disease 83/157 (53%). Cirrhosis was confirmed in 136 patients (87%). Median MELD was 15 (IQR 11–20). All patients referred 126/177 (71%) failed endoscopic haemostasis at referring hospital (RH). Median RBC transfusion at RH was 9 units (IQR 6–15) with 80% had Sengstaken-Blakemore tube inserted prior to transfer. Median LITU length of stay (LOS) 6 days (IQR 3–15). Median hospital stay was 24 days (IQR 13–40). LITU survival was 70% (133/157). Hospital survival 61% (96/157). Survival at 42 days was 60% (94/157). Factors determining 6-week survival, survivors to non-survivors, were Day (D) 1 MELD (13 vs 19), D5 MELD (13 vs 25), ▵ (D5-D1) MELD (0 vs 4), D1 UKELD (Na incorporated MELD) (53 vs 57), D5 UKELD (52 vs 56) and APACHE II score (16 vs 24), p<0.0001 (all). AUROC curve analysis showed the following prognostic indicators performed well: D1 MELD=0.75, D5 MELD=0.84, ▵MELD=0.74 and D1 APACHE II score=0.78. Median number of endoscopies was 2 (IQR 1–3). Endoscopy performed within 24 h of LITU admission in 94%. Bleeding source was oesophageal varices 96/145 (66.2%), gastric varices 12/145 (8.3%) and ectopic varices 2/145 (1.4%). Active bleeding identified in 93/144 (65%). Five day rebleeding rate was 13% (20/153). Four had failure of control of bleeding. There was no significant difference in hospital survival (62% vs 65%), LITU survival (72% vs 65%), LITU LOS in days (6 vs 8) and hospital LOS in days (24 vs 32) comparing those who achieved primary haemostasis to rebleeders (P=NS). Admission haemoglobin, platelet count, transfusion requirement, MELD and UKELD did not predict rebleeding (p=NS). TIPSS was performed on 10%. Seven patients (35%) of those who rebled within 5 days had TIPSS. TIPSS insertion did not influence hospital survival (p=0.46).
Conclusion Outcome of severe UGIVB is determined by severity of liver disease and organ failure assessed by MELD/UKELD and APACHE II, respectively. However, these factors did not predict rebleeding. Endoscopic haemostasis was achieved in the majority of patients. TIPSS did not influence outcome in this selected cohort of UGIVB.