ECLAP 2004.

Methods	Randomised controlled trial.
Participants	518 people with polycythaemia vera (210 females, 308 males; mean age 61 years: a total of 26 percent of participants were 70 years of age or older). Participants were eligible if they had no clear indication for aspirin treatment and no clear contraindication to it, were able to provide written informed consent, and had no clinically significant co‐existing conditions.
Interventions	Low‐dose aspirin, 100 mg per day (N = 253) versus placebo (N = 265). All participants who were recruited received other recommended treatments: Phlebotomy, cytoreductive drugs, and standard cardiovascular drugs were given as required.
Outcomes	The two pre‐defined primary composed efficacy endpoints were: non‐fatal myocardial infarction, non‐fatal stroke, or death from cardiovascular causes; and non‐fatal myocardial infarction, non‐fatal stroke, pulmonary embolism, major venous thrombosis, or death from cardiovascular causes. The secondary endpoints consisted of fatal or non‐fatal cerebrovascular events, fatal or non‐fatal cardiac events, minor thrombotic complications (including atypical cerebral or visual symptoms of ischaemia, erythromelalgia, and thrombophlebitis), and major and minor thrombotic complications as previously defined. Safety was assessed by examining rates of fatal and non‐fatal major haemorrhage (any haemorrhage requiring transfusion, hospitalisation, or both), minor haemorrhage, and any adverse events leading to discontinuation of treatment.
Notes	Cytoreductive therapy was used in the following percentages of participants: ‐ Phlebotomy: 175 participants (69.2%) in the aspirin group and 197 participants (74.3%) in the placebo group. ‐ Any cytoreductive drug (radioactive phosphorus, hydroxyurea, busulfan, chlorambucil, pipobroman, interferon alpha): 149 participants (58.9%) in the aspirin group and 145 participants (54.7%) in the placebo group. After a planned interim safety analysis (in December 2000), the steering committee was informed that fewer centres than expected were recruiting effectively; that after the planned two years of recruitment, the rate of randomisation was reduced to nearly zero; that an impractically long follow‐up period would be required to accumulate the number of events needed to reach the pre‐defined rate of end points; and that no additional support for the trial could be obtained. For these reasons, the study was stopped, and follow‐up of participants who had undergone randomisation was completed during the next 12 months. These decisions were made with the advice and consent of the data and safety monitoring board and were communicated to the investigators, who were monitored to ensure that they conducted a final follow‐up visit. We obtained updated follow‐up information after September 1, 2001, for 92% of the participants who had undergone randomisation, for a total duration of follow‐up of 1478 person‐years.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	'... A double‐blind, placebo‐controlled design was used. A total of 253 participants were randomly assigned to receive aspirin (100 mg daily), and 265 were randomly assigned to receive placebo ... Participants were assigned to treatments with the use of a program based on the biased‐coin algorithm, which allowed for stratification according to centre ...'
Allocation concealment (selection bias)	Low risk	'... randomisation was centralised and was performed over the telephone ...'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Inability to judge whether there was selective loss to follow‐up or study withdrawal (6%). No further data were provided by the authors.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	'... A double‐blind, placebo‐controlled design was used. A total of 253 participants were randomly assigned to receive aspirin (100 mg daily), and 265 were randomly assigned to receive placebo ... Participants were assigned to treatments with the use of a program based on the biased‐coin algorithm, which allowed for stratification according to centre ...'
Blinding of outcome assessment (detection bias) All outcomes	Low risk	'... The validation of the clinical events included in the primary endpoints was ensured by an ad hoc committee of expert clinicians who were unaware of the treatment‐group assignments. Each event was validated independently by two evaluators, and disagreement between evaluators was assessed by the chairman of the study ...'