Background: Patients with myelodysplastic syndrome or acute myeloid leukaemia who are thrombocytopenic and unable to receive disease-modifying therapy have few treatment options. Platelet transfusions provide transient benefit and are limited by alloimmunisation. Eltrombopag, an oral thrombopoietin receptor agonist, increases platelet counts and has preclinical antileukaemic activity. We aimed to assess the safety and tolerability of eltrombopag for the treatment of thrombocytopenia in adult patients with advanced myelodysplastic syndrome, secondary acute myeloid leukaemia after myelodysplastic syndrome, or de-novo acute myeloid leukaemia. Methods: We did this multicentre, randomised, placebo-controlled, double-blind, phase 1/2 trial at 37 centres in ten countries in Europe, east Asia, and the Americas. Patients aged 18 years or older who had relapsed or refractory disease or were ineligible for standard treatments; had platelet counts of less than 30 × 109 platelets per L; had 10-50% bone-marrow blasts; or were platelet transfusion dependent were randomly assigned (2:1), via a telephone-based interactive voice-response system (GlaxoSmithKline Registration and Medication Ordering System) with a permuted-block randomisation schedule (block size of three), to receive once-daily eltrombopag or matching placebo dose adjusted from 50 mg to a maximum dose of 300 mg. Randomisation was stratified by presence of poor-prognosis (complex) karyotype (presence of at least three abnormalities, or chromosome 7 abnormalities, vs absence) and bone-marrow blast count (<20% vs ≥20%). Patients and study personnel were masked to treatment allocation. The primary endpoint was safety and tolerability, including adverse events, non-haematological laboratory grade 3-4 toxic effects, and changes in bone-marrow blast counts from baseline. Analysis was by intention to treat. This trial is registered at ClinicalTrials.gov, number NCT00903422. Findings: Between May 14, 2009, and May 9, 2013, we randomly assigned 98 patients to receive either eltrombopag (n=64) or placebo (n=34). 63 (98%) patients in the eltrombopag group and 32 (94%) patients in the placebo group had adverse events. The most common adverse events were pyrexia (27 [42%] vs 11 [32%]), nausea (20 [31%] vs 7 [21%]), diarrhoea (19 [30%] vs 6 [18%]), fatigue (16 [25%] vs 6 [18%]), decreased appetite (15 [23%] vs 5 [15%]), and pneumonia (14 [22%] vs 8 [24%]). Drug-related adverse events of grade 3 or higher were reported in six (9%) patients in the eltrombopag group and four (12%) patients in the placebo group. Increases in the proportion of peripheral blasts did not differ significantly between groups. Haemorrhage of grade 3 or higher was reported in ten (16%) patients given eltrombopag and nine (26%) patients given placebo. 21 (33%) patients receiving eltrombopag and 16 (47%) patients receiving placebo died while on treatment. No deaths in patients receiving eltrombopag and two deaths in patients receiving placebo were regarded as treatment related. Post-baseline bone-marrow examinations were done in 40 (63%) patients in the eltrombopag group and 17 (50%) patients in the placebo group. The most common reason for no examination was death before the scheduled 3 month assessment. There were no differences between median bone-marrow blast counts or proportions of peripheral blasts between groups. Interpretation: Eltrombopag doses up to 300 mg daily had an acceptable safety profile in patients with advanced myelodysplastic syndrome or acute myeloid leukaemia. The role of eltrombopag in these patients warrants further investigation. Funding: GlaxoSmithKline.