Cheng 2011.

Methods	Randomised, double‐blind, placebo controlled trials. Phase III. 6‐week study. Multicenter (75 sites in 23 countries). Stratification by concomitant ITP medication (yes or no), splenectomy (yes or no), and the baseline platelet count (>15,000 per cubic millimeter versus ≤ 15,000 per cubic millimeter).
Participants	Chronic ITP, a history of ITP for at least 6 months, PLT≤ 30 × 109/l, one or more prior treatments for ITP, the dose of maintenance immunosuppressive regimens, primarily glucocorticoids, must have been stable for at least 1 month, age ≥ 18 years, n = 197
Interventions	Eltrombopag 50 mg p.o qd for 6 months versus placebo
Outcomes	Primary: percentage of responders Secondary: median platelet counts, the proportion of patients who responded at 75% or more of assessments, mean cumulative weeks of response, mean maximum weeks of continuous response, bleeding symptoms, reduction of baseline treatment for immune thrombocytopenia, and use of rescue treatment; HR‐QoL Instrument and Domain Scores; adverse event
Notes	Results first received: July/2009 Followed‐up period: 6 months Study supported by GlaxoSmithKline The protocol was developed by the principal investigators and employees of the sponsor. Data were collected and analysed by the sponsor. All authors had access to the primary data and vouch for the completeness and accuracy of the data and analyses. Interpretation of the data and decisions related to the content of the report were made through collaboration among all authors. The corresponding author had final responsibility for the decision to submit for publication.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	A ‐ Adequate. Quote: "The randomisation schedule was computer generated with an in‐house validated randomisation system, and was generated by a member of the GlaxoSmithKline statistics and programming group, who held a statistical consultation role for the rest of the trial."
Allocation concealment (selection bias)	Low risk	A ‐ Adequate. Quote: "All randomisations were done with RAMOS, an automated interactive voice recognition telephone randomisation and drug ordering system."
Blinding (performance bias and detection bias) Physician assessed	Low risk	Quote: "Double blind". "Patients, investigators, and those assessing the data were masked to allocation."
Blinding (performance bias and detection bias) Patient assessed	Low risk	Quote: "Double blind". "Patients, investigators, and those assessing the data were masked to allocation."
Incomplete outcome data (attrition bias) Efficacy outcomes	Low risk	83% eltrombopag group and 89% placebo group completed the trial. Judged a low risk of bias as efficacy analysis included all patients treated.
Incomplete outcome data (attrition bias) Safety outcomes	Low risk	83% eltrombopag group and 89% placebo group completed the trial. Judged a low risk of bias as safety analysis included 99.5% patients treated.
Selective reporting (reporting bias)	Low risk	Quote: "Each patient was assessed for response (defined as a platelet count of 50000 ‐ 400000 per μL) at each assessment during the 6‐month treatment period; the primary endpoint was the odds of response to eltrombopag versus placebo during this period. As described above, any assessment, irrespective of platelet count, was regarded as a non‐response if it occurred during a period of rescue treatment. Bleeding was assessed with the WHO bleeding scale (grade 0, no bleeding; grade 1, petechiae; grade 2, mild blood loss; grade 3, gross blood loss; grade 4, debilitating blood loss). For the secondary endpoint of median platelet counts, all platelet counts were included in the analysis irrespective of whether the patient received rescue treatment. Rescue treatment was defined as new treatment for chronic immune thrombocytopenia, an increased dose of baseline treatment, platelet transfusion, or splenectomy. The acute recall version of the short form‐36, version 2 (SF‐36v2) questionnaire was used to measure health‐related quality of life at baseline and at weeks 6, 14, and 26 or on discontinuation of study drug. Measurement of fatigue with the vitality domain of the SF‐36v2 was supplemented by use of the fatigue subscale of the functional assessment of chronic illness therapy (FACIT)‐fatigue questionnaire. The e? ect of bleeding on quality of life was assessed with a six‐item subset from the functional assessment of cancer therapy‐thrombocytopenia (FACT‐Th6) questionnaire. Changes to physical and mental energy and social motivation were assessed with the short form of the motivation and energy inventory (MEI‐SF) questionnaire. Adverse event reports (graded with the National Cancer Institute Common Terminology Criteria for Adverse Events, version 3.0) and laboratory assessments were completed at each on‐treatment and post‐treatment visit. On the basis of preclinical findings, ocular examinations were done at baseline and regularly during the study."
Other bias	Unclear risk	Funded by drug company. There is evidence that industry‐sponsored trials may overestimate the treatment effect (Bhandari 2004)