STOP 1998.

Study characteristics
Methods	Multicentre randomised controlled trial conducted in 12 centres in the USA and Canada in children 2 ‐ 16 years of age with HbSS or HbSβº thalassaemia. Screening began in January 1995 and ended in November 1996. The trial was to run to December 1998 but was stopped in September 1997.
Participants	Inclusion criteria: children 2 to 16 years of age and who had been given a diagnosis of sickle cell anaemia or sickle β0 thalassaemia at high risk of stroke with a blood flow velocity of at least 200 cm per second on 2 TCD trials. Exclusion criteria: history of stroke, had an indication for or contraindication to long‐term transfusion, were receiving other treatments that affected the risk of stroke, were infected with the human immunodeficiency virus (HIV), had been treated for seizures, were pregnant, or had a serum ferritin concentration above 500 ng per millilitre. Participant flow: screened: N = 1934; eligible: N = 206 randomised: N = 130 Transfusion: N = 63 Sex: male: 31 (49%) Age: mean (SD): 8.2 (3.2) years HbS% mean (SD): 87 (10) HbF% mean (SD): 8.0 (5.2) Alpha thalassaemia: 14 (22%) TCD velocity: mean (SD): 223 (27) cm/sec Lesions on initial MRI: N (%) participants: 19 (31%) Standard care: N = 67 Sex: male: 29 (43%) Age: mean (SD): 8.4 (3.3) years HbS% mean (SD): 87 (7) HbF% mean (SD): 9.4 (5.0) Alpha thalassaemia: 7 (9%) TCD velocity: mean (SD) : 223 (28) cm/sec Lesions on initial MRI N (%): participants 25 (38%) Phenotypes: not reported (trial included only HbSS or HbSβº)
Interventions	Transfusion: N = 63 In the transfusion arm the goal was to reach an HbS concentration < 30 per cent of total haemoglobin within 21 days without exceeding a haemoglobin concentration of 120 g/L and a hematocrit of 36%. Exchange or simple transfusion were allowed: 63% were simple transfusions, 12% were exchange; 25% a combination of simple and exchange. Red cells were delivered in a volume of approximately 10 to 15 mL per kg of packed cells per transfusion. Red cell component: leucocyte‐depleted, negative for haemoglobin S. Red cell matching: ABO, Rh and Kell antigens. Iron chelation: none. Potential participants with a ferritin level above 500 ng/mL were excluded from the trial. The intention was to exclude any child with a significant iron burden before initiation of treatment, thus avoiding clinically significant iron overload during the trial. Standard care: N = 67
Outcomes	Primary outcome: cerebral infarction and intracranial haemorrhage. Secondary outcomes: death, transfusion‐related adverse events.
Notes	Funding: supported by Cooperative Agreements (U10 HL 52193 and U10 HL 52016) with the National Heart, Lung, and Blood Institute. Declarations of interest: none published. Trial registration: no registration found. Mean (SD) length of follow‐up: transfusion arm: 21.0 (5.7) months; Standard care: 18.3 (7.0) months. Power calculation: "Estimates of stroke risk for patients randomized to standard care were obtained by fitting an exponential model to the follow‐up of TCD (1) patients follow‐up, it was estimated that 47% of patients in this group should develop stroke on study. Assuming transfusion prevents 70% of these strokes, 14% of the patients randomized to transfusion should have strokes on study. Taken together, these values imply that a sample size of 46 per treatment arm should provide the desired statistical power of 90% to detect a 70% reduction in stroke incidence at a type I error rate of 0.05 for a two‐sided test". Analysis: 4 interim analyses and one final analysis were planned. the date of the first analysis changed from 20 months to 14 months after recruitment began.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"The DCC developed permuted blocks within which treatment allocations were randomly and evenly assigned. The blocks themselves were randomly assigned to each of the 12 Centers."
Allocation concealment (selection bias)	Low risk	"After telephone verification that the patient was eligible and acquisition of written parental consent, the DCC ran a short randomization program and provided the Investigator with the trial group assignment." " Permuted blocks are used to blind Investigators to the potential treatment assignment of each patient while preserving approximate balance within and across Centers. The DCC provided the Clinical Center Investigator and patient with the treatment assignment."
Blinding of participants and personnel (performance bias) All outcomes	High risk	Clinicians and participants were unblinded.
Blinding of outcome assessment (detection bias) Stroke and TIA	Low risk	A panel of physicians with no knowledge of the children’s treatment assignments who were not affiliated with the trial centres determined whether an event was a stroke. The primary end points were cerebral infarction and intracranial haemorrhage. "The protocol was intended to identify all neurologic events. A panel of physicians with no knowledge of the children’s treatment assignments who were not affiliated with the trial centres determined whether an event was a stroke.".
Blinding of outcome assessment (detection bias) All outcomes apart from stroke or all‐cause mortality	High risk	Unblinded trial.
Incomplete outcome data (attrition bias) All outcomes	Low risk	An ITT analysis was used, despite 12 participants crossing over between groups (2) or withdrawing from the trial (10). Reasons were provided. 10 participants from the transfusion group withdrew from the trial because of problems with compliance (n = 4), multiple alloantibodies (n = 1), ineligibility (n = 1) or other unspecified reasons (n = 4). Two participants from the standard care group crossed over to the transfusion group, one on the second day due to diagnosis of subacute intracerebral hematoma and the other after 12 months for treatment of leg ulcers.
Selective reporting (reporting bias)	Unclear risk	No protocol available and no prospective trial registration.
Other bias	Unclear risk	"Four interim analyses and one final analysis were planned, with the Lan–DeMets approximation of the O’Brien–Fleming stopping boundary. The date of the first analysis was changed from 20 months to 14 months after recruitment began." "Because of the high rate of stroke in the standard‐care group and the significant effect of transfusion found at the second interim analysis, the data safety and monitoring board recommended that the trial be stopped 16 months before the planned date of December 1998 so that transfusion could be offered to children in the standard‐care group." There was an imbalance between the number of participants with alpha thalassaemia trait between treatment arms (22% transfusion arm versus 9% standard treatment arm).