The prevention of progression of arterial disease and diabetes (POPADAD) trial was a multicentre, randomised, double blind, placebo controlled trial. We used a 2×2 factorial design30
to examine the efficacy and safety of aspirin plus antioxidant compared with aspirin alone, antioxidant alone, and placebo.
The interventions were daily aspirin 100 mg or placebo tablet, plus antioxidant or placebo capsule. The antioxidant capsule contained α-tocopherol 200 mg, ascorbic acid 100 mg, pyridoxine hydrochloride 25 mg, zinc sulphate 10 mg, nicotinamide 10 mg, lecithin 9.4 mg, and sodium selenite 0.8 mg. We selected this particular mix following advice from experts in antioxidants, with approval of the Medicines’ Control Agency for use as a nutritional aid in clinical studies. The placebo tablet and capsule were identical in appearance to the active tablet and capsule.
Participants were recruited by the Royal College of Physicians, Edinburgh, East of Scotland Diabetic Registry Group, which covers diabetic clinics in hospitals in the south east of Scotland and in central Scotland, and from clinics in the west of Scotland. The coordinating centre was Ninewells Hospital, Dundee. Sixteen hospital centres participated in the trial, supported by 188 primary care groups.
Inclusion criteria were adults of either sex, aged 40 or more, with type 1 or type 2 diabetes who were determined as having asymptomatic peripheral arterial disease as detected by a lower than normal ankle brachial pressure index (≤0.99). We selected a higher cut-off point (0.99 v 0.9) for the trial as it is recognised that calcification in the vessels of people with diabetes can produce normal or high ankle brachial pressure indexes, even in the presence of arterial disease. We excluded people with evidence of symptomatic cardiovascular disease; those who use aspirin or antioxidant therapy on a regular basis; those with peptic ulceration, severe dyspepsia, a bleeding disorder, or intolerance to aspirin; those with suspected serious physical illness (such as cancer), which might have been expected to curtail life expectancy; those with psychiatric illness (reported by their general practitioner); those with congenital heart disease; and those unable to give informed consent.
Potentially eligible participants attending the diabetes clinics during the enrolment period were invited to join the trial. If they agreed they were screened for eligibility. We scrutinised each patient’s hospital and general practice records for evidence of symptomatic cardiovascular disease. We excluded those not currently free from symptoms of vascular disease at the time of this screening visit. Ankle brachial pressure was measured in asymptomatic patients using a standardised technique.31
Training was provided to nurses without experience of measuring this pressure. Patients with an ankle brachial pressure index of less than 1.00 were eligible for inclusion.
After providing written informed consent, patients were randomly assigned to one of four treatment groups: aspirin plus antioxidant, aspirin plus placebo, antioxidant plus placebo, or double placebo. The allocation sequence used randomised permuted blocks of eight and was computer generated by the trial statisticians. To ensure allocation concealment an independent pharmacist packaged the drugs into numbered containers. Recruiting nurses dispensed the trial drugs on the day of randomisation, under the authority of the consultant or primary care doctor. Participants also received standard therapy as appropriate (for example, statins, β blockers) at the discretion of the investigator and other responsible clinicians. We emphasised to the investigators the use of appropriate background cardiovascular risk reduction therapy according to current international guidelines. The participants, research nurses, and staff involved in providing care were blinded to group assignment.
Follow-up evaluations were done every six months. At these visits we recorded outcome events, adverse events, and interventions. The results of electrocardiography were recorded at the baseline visit and annually thereafter. The electrocardiograms were reviewed manually for evidence of silent myocardial infarction on the basis of criteria from the Minnesota code. When relevant, a copy of data from hospital admissions was obtained for use by the committee deciding on the presence or absence of a specific end point using predefined criteria. For example, any electrocardiograms obtained from a hospital admission in the case of a suspected myocardial infarction were also coded using the criteria of the Minnesota code so that the appearances of the electrocardiograms were considered in a uniform manner along with clinical and biochemical data in reaching a decision. All primary and secondary end points were adjudicated on a blinded basis by the committee.
We used two hierarchical composite primary end points: death from coronary heart disease or stroke, non-fatal myocardial infarction or stroke, or above ankle amputation for critical limb ischaemia; and death from coronary heart disease or stroke. Definitions for these events were according to the World Health Organization criteria for the diagnosis of coronary events and strokes (fatal and non-fatal).
The main secondary end points were all cause mortality, non-fatal myocardial infarction, and occurrence of other vascular events, including stroke, transient ischaemic attack, coronary or peripheral arterial bypass surgery, coronary or peripheral arterial angioplasty, development of angina, claudication, or critical limb ischaemia.
The event rate for similar end points in patients with asymptomatic peripheral arterial disease in the Edinburgh artery study32
was 4% per annum. The event rate in the literature at the time of the start of the trial for patients with diabetes was between twofold and threefold that of the population without diabetes.1 2 3
We originally planned to recruit 1600 participants and follow-up each for four years. If only one treatment was effective this would provide 90% power to detect a 25% relative reduction in a four year event rate of 28% (8% per annum) as significant at the 5% level. This equates to 392 events occurring during the trial. With this sample size but both treatments equally effective, so slightly lower overall event rates, 343 events would be expected in the four years of follow-up. This would still provide greater than 80% power to detect for each treatment the same relative reduction in event rate as significant.
A slower than expected recruitment rate and lower event rates led to ongoing consideration of recruitment and termination dates by the data monitoring and ethics committee, trial steering committee, and funding body. Eventually 1276 patients were recruited and the final power calculations, undertaken in 2003, projected that if follow-up continued until June 2006 then 256 events would be expected to occur during the trial. This would give 73% power to detect a 25% relative reduction in event rate and 89% power to detect a 30% reduction in event rate if only one treatment was effective.
The statistical analysis followed the plan determined at the start of the trial. The primary and secondary end points were measures of survival. Accordingly, we used a Cox proportional hazards model as the primary method of analysis. We assessed the interventions by fitting terms corresponding to aspirin, antioxidants, and the interaction between these treatments. As there was no evidence of interaction we dropped this term, allowing the overall effect of each intervention to be assessed. We assessed the assumption of proportionality of hazards and we used Kaplan-Meier plots for the survival experience by treatment group. Specific adverse events were assessed using logistic regression with terms corresponding to aspirin, antioxidants, and the interaction between these treatments. As we found no evidence of interaction we dropped this term, allowing the overall effect of each intervention to be assessed. All analyses were done on an intention to treat basis, with two tailed tests of significance used throughout.
The trial was designed, planned, and executed by the trial steering committee in collaboration with the UK Medical Research Council. The trial was done in accordance with good clinical practice regulations.33
Experienced research nurses collected the data, which were entered and analysed by the Medical Statistics Unit, University of Edinburgh. The progress of the study was monitored throughout by the data monitoring and ethics committee. This committee met at six monthly or yearly intervals depending on the stage of the trial. No formal stopping rules were used. The principle employed was that early termination for efficacy or futility would only occur if the evidence was assessed to be strong enough to influence practice.