Background: Gilbert's syndrome is a common inherited disorder of bilirubin metabolism, characterised by mild, unconjugated hyperbilirubinaemia. However, the effect of Gilbert's syndrome on the disposition of some drugs can lead to unexpected toxicity. We tested the hypothesis that patients undergoing myeloablative conditioning and haemopoietic cell transplantation would have different mortality outcomes depending on whether or not they had laboratory evidence of Gilbert's syndrome., Methods: In this retrospective cohort study, we used clinical and laboratory data of patients who had haemopoietic cell transplantation from Jan 1, 1991, to Dec 31, 2011. Patients were included if they had received high-dose conditioning regimens of cyclophosphamide plus total body irradiation (CY/TBI), busulfan plus cyclophosphamide (BU/CY), busulfan plus melphalan plus thioTEPA (BUMELTT), or melphalan before transplant. Patients were excluded if their original consent forms to report transplant outcomes were not signed, if consent was withdrawn, or if they were a prisoner. Patients with Gilbert's syndrome were defined as having laboratory values before the start of conditioning therapy for unconjugated serum bilirubin concentrations of at least 17·1 μmol/L (≥1 mg/dL), normal conjugated serum bilirubin, and no evidence of hepatitis, cholestasis, or haemolysis. We assessed the association of Gilbert's syndrome with overall mortality and non-relapse mortality using adjusted Cox regression models at day 200 after transplantation., Findings: Our study cohort was 3379 patients-1855 (55%) allograft and 1524 (45%) autograft recipients. 211 (6%) patients had Gilbert's syndrome and 3168 (94%) did not have this condition. Most patients were adults (median age 45·8 years [IQR 33·2-55·5]) with haematological malignancies. For overall mortality 664 (20%) patients had died by day 200 after transplant (47 [22%] of 211 who had Gilbert's syndrome vs 617 [19%] of 3168 who did not have Gilbert's syndrome), and for non-relapse mortality 499 (92%) patients had died before relapse was recorded (38 [18%] who had Gilbert's syndrome vs 461 [15%] who did not have Gilbert's syndrome). The effect of Gilbert's syndrome on the risk of overall mortality and non-relapse mortality by transplant day 200 varied between the conditioning regimens and donor groups. In patients conditioned with a myeloablative regimen that contained busulfan (n=1131), those with Gilbert's syndrome (n=60) were at a significantly increased risk of death and non-relapse mortality by day 200 compared with those without Gilbert's syndrome (n=1071; hazard ratio [HR] 2·30, 95% CI 1·47-3·61, p=0·00030; and 2·77, 1·71-4·49, p<0·0001). In patients who received CY/TBI or melphalan conditioning regimens, those with Gilbert's syndrome had similar outcomes to those without Gilbert's syndrome (overall mortality at day 200 HR 0·90, 95% CI 0·60-1·34, p=0·60; non-relapse mortality at day 200: 0·90, 0·56-1·45, p=0·65). Analyses of causes of death and busulfan disposition provided no mechanistic explanation for the differences in mortality., Interpretation: Overall mortality and non-relapse mortality at day 200 after transplant were significantly worse in patients with Gilbert's syndrome who received busulfan-containing myeloablative conditioning regimens, compared with non-Gilbert's syndrome patients. Patients with Gilbert's syndrome should receive busulfan-containing myeloablative conditioning regimens with caution., Funding: US National Institutes of Health., (Copyright © 2016 Elsevier Ltd. All rights reserved.)