l-Arginine, a semi-essential amino acid, is the sole substrate for endothelial nitric oxide synthase (eNOS). Reduced NO availability is hypothesized to be a key component in the pathophysiology of a range of complications observed in sickle cell disease (SCD) and other haemolytic disorders (Gladwin & Kato, 2005). Ornithine is metabolized from arginine by arginase, released from lysed red cells and damaged liver cells into the plasma. Ornithine competes for endothelial uptake with arginine, thus reducing its availability. Reduced global arginine bioavailability (GAB), measured as low ratios of plasma l-arginine to l-ornithine or l-ornithine plus l-citrulline, is associated with mortality in adult American SCD patients (Morris et al, 2005) and cardiovascular disease in the general population (Tang et al, 2009). In cerebral malaria, low plasma arginine is associated with case fatality (Lopansri et al, 2003). Whether GAB is associated with adverse outcomes in African SCD patients is not known. We hypothesized that steady-state GAB would be lower in patients who had died compared to matched controls followed up for the same time period. This was a retrospective, case control study nested within a prospective urban cohort of patients with confirmed haemoglobin SS at Muhimbili National Hospital, Dar-es-Salaam, Tanzania. Archived samples collected between July 2007 and June 2009 were available for analysis of plasma amino acids. Ethics approval was granted from Muhimbili University of Health and Allied Science (reference MU/RP/AEC/VOL XI/33) and the London School of Hygiene & Tropical Medicine (reference 5158). Recruitment of patients, clinical procedures and data collection for this cohort have been described (Makani et al, 2009). Blood samples were collected at scheduled clinic visits and classified as steady-state in the absence of pain, fever (temperature >37·4°C), or current/recent malaria infection (malaria rapid tests and microscopy). To further ensure true steady-state, samples were excluded if hospitalization or death occurred within a month of sample collection. EDTA plasma was separated from whole blood and stored at −80°C within 2 h of blood collection at the clinic. Separate blood sample tubes were used for routine complete blood counts (Pentra 60, Horiba ABX, Kyoto, Japan) and in some cases for blood chemistry (Roche Cobas Mira, New York, NY, USA or Abbott Architect, New York, NY, USA). Amino acid profiles were measured on a Biochrom 30 amino acid analyser (Biochrom Instruments, Cambridge, UK) using a lithium high-resolution column optimized for physiological fluid separations by the Neurometabolic Unit, National Hospital for Neurology and Neurosurgery, London, UK. Unpaired Student’s t-tests were used to compare amino acid concentrations and GAB between cases and controls. Associations between continuous variables and possible confounding of the relationship between amino acid ratios and group by age, sex and body mass index (BMI) were investigated using linear regression models. On 1st July 2009, there had been a total of 40 deaths from 1060 patients in active follow-up during the period July 2007–June 2009. Archived steady-state samples were available for 12 patients who had died (cases) for whom samples from controls matched for age, sex, duration of follow-up and month of sample collection were available, thus minimizing possible extraneous sources of variation. One case sample was insufficient for analysis. Patient characteristics by group are shown in Table I. None of the patients had symptoms of, or were known to be positive for, human immunodeficiency virus, although routine testing was not conducted. Previous hospitalizations ranged from 0 to 5 for cases, with pain and anaemia being the most frequent causes (14 out of 23 hospitalizations). Four controls had one previous hospitalization each. Cases had significantly lower BMI at time of amino acid analysis (Table I). Table I Patient characteristics and steady state plasma amino acid concentrations in SCA patients who subsequently died compared to age and sex matched SCA patients who were alive as of 1st July 2009. There was a trend for lower taurine levels in cases and GAB was significantly lower (Table I) which remained significant when adjusting for BMI as a potential co-factor, although there were no significant associations between GAB and age, sex, haemoglobin or BMI. Full amino acid profiles can be seen in Table SI. Laboratory markers of haemolysis and liver function at time of amino acid analysis and enrolment were available for some patients (Table SII). Markers of haemolysis, total and unconjugated bilirubin were available in 21/23 and 19/23 patients at enrollment and 14/23 and 9/23 at time of amino acid analysis, whilst lactate dehydrogenase was available in eight and 14 patients at enrollment and time of amino acid analysis, respectively. There was no evidence of differences in these haemolytic markers between cases and controls (Table SII) and nor was there evidence for associations with GAB. In contrast, conjugated bilirubin at enrollment was significantly higher in cases compared to controls (32·3 μmol/l vs. 15·6 lmol/l, P = 0·005, n = 20) but no differences were observed for aspartate transaminase or alkaline phosphatase with smaller numbers of data points. Interestingly, there were significant associations between higher GAB ratios and lower conjugated bilirubin when measured at enrollment (Arg:Orn ratio, regression coefficient = −0·006, P = 0·021 and Arg:Orn + Cit ratio, −0·005, P = 0·021, n = 20) (Fig 1), with a similar association, not quite reaching statistical significance, at time of amino acid analysis (Arg:Orn ratio: coefficient = −0·009, P = 0·07; and Arg:Orn + Cit ratio: coefficient = −0·006, P = 0·055, n = 9). Fig 1 Scatterplot of GAB (plasma ratio of arginine to ornithine) and conjugated bilirubin in cases (●) and controls (○). In line with the previous literature in adults (Morris et al, 2005), the ratios for plasma GAB were lower in adolescents who subsequently died in this small but carefully controlled study. Healthy, non-SCA children from urban Dar-es-Salaam, Tanzania, had higher arginine concentrations (120 ± 10 μmol/l; Lopansri et al, 2003); and Arg:Orn ratios [1·35 ± 0·49 (unpublished observations, E. Mwaikambo & D. Granger)] compared to our patients. The lack of difference in haemolytic markers between groups of patients in the current analysis may result from small sample size, but could reflect different causes of severe disease and death in these relatively young African patients compared to older American patients. Increased steady-state conjugated bilirubin may represent cholestatic liver dysfunction, and/or upregulated haem-oxygenase (HO-1). Lower GAB, through decreased NO bioavailability, could reduce blood flow and oxygen supply to the liver, directly affecting its function and increasing the compensatory HO-1 pathway (Jison et al, 2004). Further studies looking at the relative importance of infection, blood flow, oxygen supply and HO-1 activity on total and conjugated bilirubin levels in SCD patients are warranted. The preliminary finding of lower taurine levels in those who died is intriguing as this amino acid has a protective effect where there is oxidative stress. Confirmation of these findings in a larger sample and closer investigation of the pathways of disease could form the basis of development of novel dietary interventions appropriate to low income settings.