Acute pancreatitis in a high HIV prevalence environment: analysis of prevalence, demographics, prognosticators and outcomes.

Abstract

Background It is unclear what is the true prevalence of HIV related acute pancreatitis and whether diagnostic and prognostic markers used in patients without HIV infection are as effective in HIV related pancreatitis and if morbidity is worse in HIV infected patients. Methods Using a prospective, descriptive design, HIV prevalence was compared in trauma and acute pancreatitis patients. Serum amylase was used to diagnose acute pancreatitis. Prognostication was by CRP, BISAP, Glasgow and APACHE II scores at 24 hours. Sensitivity, specificity and AUC were compared in predicting a severe outcome in acute pancreatitis. Complications and mortality were compared in 238 HIV+ve and HIV-ve patients admitted to 2 regional hospitals in Durban between August 2013 and October 2015. One hundred and eighty one patients were admitted with trauma. Results Between August 2013 and October 2015, 238 patients were admitted with acute pancreatitis and 181 with trauma. HIV infection was higher in patients with acute pancreatitis (38% vs 16%) (p=0.001) and they were also older (40 vs 33 years) (p=0.001). Fifty three percent of HIV +ve patients were female and 65% of the HIV-ve patients were male in the pancreatitis cohort and 59% of the trauma and pancreatitis patients were on Highly Active Antiretroviral Therapy. The prevalence of gallstone (27% vs 30%), alcohol (41% vs 52%), dyslipidaemia (0% vs 3%) and idiopathic (6% vs 14%) aetiologies were similar in HIV+ve and HIV-ve patients and a drug related aetiology (24% vs 0%) (p<0.001) was more prevalent in HIV related acute pancreatitis. CRP was more effective in predicting severe disease in HIV-ve patients (AUC= 0.75) and patients with CD4 counts of ≥ 200 cells/mm3 (AUC=0.73) and not HIV+ve patients (AUC= 0.59) or patients with counts below 200 cells/mm3 (AUC= 0.46). The BISAP system had similar efficacy with AUC of 0,71 and 0.74 in HIV-ve and HIV+ve patients respectively, was poor in CD4 count < 200 cells/mm3 (AUC=0.68) and good in CD4 count> 200 cells/mm3 (AUC=0.9). The Glasgow score was of similar efficacy in HIV-ve (AUC = 0.72) and HIV+ve patients (AUC=0.78) and better in patients with CD4 count < 200 cells/mm3 (AUC=0.83) and CD4 count ≥ 200 cells/mm3 (AUC=0.81). The APACHE II had uniform efficacy in both HIV-ve and HIV+ve patients (AUC >0.8) and both CD4 count ranges (AUC > 0.80). Septic complications occurred in 10(8%) of HIV-ve patients and 4(4%) HIV+ve patients. There was no difference in morbidity (25% vs 33%) and mortality (6% vs 6%). Conclusions HIV infections is more prevalent in acute pancreatitis than in a hospital trauma cohort which represented the general population. The APACHE II system was the most accurate in predicting morbidity and CRP least accurate. The outcomes were similar in HIV+ve and HIV-ve patients but the statistical assumptions in calculating the sample size, given the low frequency of morbidity and mortality observed in this study may have resulted in an alpha error.