HIV Articles  
The PROactive study: some answers, many questions
  Criticisms and response by study author.
The Lancet, Volume 366, Number 9493, 08 October 2005
Hannele Yki-Jarvinen
Panel: Pros, cons, and unknowns of pioglitazone
Good news

Pioglitazone reduces composite of all-cause mortality, non-fatal myocardial infarction, and stroke in patients with type 2 diabetes with pre-existing cardiovascular disease by about 16%
Bad news
Incidence of oedema not attributable to heart failure is four times greater, and that of heart failure two times greater, than reduction in incidence of cardiovascular events by pioglitazone
Body weight increases more than with other antihyperglycaemic therapies, including insulin
What is prognosis of heart failure?
Who is at greatest risk of developing oedema not attributable to heart failure and who is at risk of developing heart failure?
Is it safe to use insulin with pioglitazone?
Why does pioglitazone reduce cardiovascular events?
The insulin-sensitising glitazones, which are selective ligands of the nuclear transcription factor peroxisome proliferator-activated receptor _ (PPAR_), ameliorate the basic problem of insulin resistance and have therefore been thought to reduce the risk of cardiovascular disease in patients with type 2 diabetes.1 Glitazones are currently approved for treatment of hyperglycaemia. These drugs lower glucose concentrations by ameliorating insulin resistance, especially in the liver, but their mechanism of action, at least in vitro, involves changes in the expression of hundreds of genes. This action is ligand-specific, which implies that results obtained with one glitazone might not be applicable to others. Findings in mice have given mixed messages about the effects of PPAR_ agonists on atherogenesis and on colon and bladder cancer, raising questions about the safety and potential benefits of the drugs in human beings. Despite such uncertainties, glitazones account for 21% of oral antihyperglycaemic drugs used in the USA and 5% in Europe.2 This fact exemplifies the power of marketing compared with evidence-based medicine in guiding treatment practices.
The PROactive study,3 reported in today's Lancet, addresses the role of pioglitazone in the prevention of macrovascular events in patients with type 2 diabetes. The inclusion criteria of the study were brave, as 5238 patients who had extensive evidence of macrovascular disease participated. A third of the patients were treated with insulin. The patients were already using many drugs known to reduce the risk of cardiovascular disease-85% were on antiplatelet medications, 70% on blockers of the renin-aldosterone-angiotensin axis, and 43% on statins. Exclusion criteria included concentrations of serum alanine aminotransferase elevated 2·5-fold above the upper limit of normal. This principle might have excluded patients who respond best to glitazones, as the main mechanism of action of these drugs is a reduction in the fat content of the liver and in hepatic insulin resistance.1
The primary endpoint in PROactive was a composite that included both disease endpoints, such as death, myocardial infarction, and stroke, and procedure endpoints such as coronary and leg revascularisations. The secondary endpoint only included disease endpoints. Pioglitazone did not affect the number of patients reaching the primary endpoint, but significantly reduced the number that reached the secondary endpoint. Inclusion of procedure endpoints in the primary endpoint was unfortunate, since procedure endpoints are less specific and less sensitive than disease endpoints, and bias treatment effect towards the null value of 1. In PROactive, pioglitazone reduced all disease endpoints, whereas the number of coronary revascularisations was the same in both groups and the number of leg revascularisations greater in the pioglitazone than the placebo group. Paradoxically, it was also unfortunate that recruitment was faster than anticipated and the decision to close the trial (after 34·5 instead of 48 months) was endpoint driven, because this might have reduced the likelihood of achieving a positive result with respect to the primary endpoint, assuming that the Kaplan-Meier curves continued to diverge as a function of time.
Pioglitazone reduced the number of primary composite endpoints by 58, but its use was associated with an increase in oedema not attributable to heart failure (221 events more in the pioglitazone than the placebo group) and in heart failure (115 events more in the pioglitazone than the placebo group; figure). There was also a significant increase in the incidence of pneumonia, which is sometimes misdiagnosed as heart failure, a marginally significant increase in bladder cancer (p=0·069) -although after exclusion of tumours judged by blinded review to be unrelated to the treatment, the difference was not significant (p=0·309)-and a significant decrease in the number of breast cancers in the pioglitazone group compared with placebo. From the patient's perspective, is it better to have healthy arteries in the heart than a failing heart? The prognosis of heart failure is particularly poor in patients with type 2 diabetes. When presenting the results, the investigators emphasised that heart failure was not a centrally adjudicated event and that increased reporting of heart failure was at least partly caused by a diagnostic bias because of increased oedema. The cardiologists were keen to comment that heart failure induced by pioglitazone is not as dangerous as other types of heart failure, although no data are presented to support this comment in the study. Weight gain was 4 kg greater with pioglitazone than placebo; four times greater than would be expected on the basis of improved glycaemic control.4 Common sense would suggest that anything that causes weight gain increases the risk of heart failure. Even if half the patients were misclassified, the number of heart failures (115/2=58) would still equal the number of primary endpoints that were reduced by pioglitazone (n=58). Since combination therapy with glitazones and insulin is contraindicated in Europe because of an increased risk of heart failure, it would have been important, by stratification at randomisation, to know whether heart failure was more common in patients who used insulin than in those who did not. The article lists hazards associated with relevant baseline characteristics for the main secondary endpoint.3 For the practising physician it would be more important to identify predictors of heart failure.
Figure. Effect of pioglitazone compared with placebo on primary endpoint and on incidence of oedema diagnosed as heart failure in PROactive ACS=acute coronary syndrome. MI=myocardial infarction.


Overall, PROactive is an important study that leaves us with some good news, some bad news, and some unknowns (summarised in the panel). The clinician, of course, wants to know who should be treated with pioglitazone. Unfortunately, the study does not provide such answers. It showed that pioglitazone is beneficial in patients with type 2 diabetes and pre-existing macrovascular disease who do not develop heart failure.
PROactive study
The Lancet, Volume 367, Number 9504, 07 January 2006
The Lancet 2006; 367:25-26
RR Holman a, R Retnakaran a, A Farmer a and R Stevens
The PROactive study1 has provided an exploratory estimate of the relative risk reduction attributable to pioglitazone for a composite endpoint of all-cause mortality, non-fatal myocardial infarction, and stroke. Although this secondary endpoint was not prespecified and is not significant when adjusted for multiple testing,2 it is of interest because of the possibility that the observed 16% relative risk reduction might not be mediated through established macrovascular risk factors such as blood pressure, lipids, and glycaemia.
We have used the United Kingdom Prospective Diabetes Study (UKPDS) Outcomes Model,3 which can simulate health outcomes for patients with type 2 diabetes on the basis of their initial characteristics and changes in risk factors over time, to assess expected PROactive outcomes given the reported risk factor changes. We computer-generated a cohort matched for baseline characteristics (age, ethnic origin, sex, body-mass index, glycosylated haemoglobin [HbA1c], lipids, blood pressure, smoking status, and peripheral vascular disease) to the PROactive study participants.1,4 The outcomes model then estimated likely 3-year differences in rates of secondary endpoints and congestive heart failure that would be associated with the noted 0·5% decrease in HbA1c, 3 mm Hg reduction in systolic blood pressure, 0·10 mmol/L increase in HDL cholesterol, and 4·0 kg increase in bodyweight compared with placebo. Because the current model does not simulate repeat macrovascular events, outcomes were estimated for the subgroup of patients in PROactive with baseline ischaemic heart disease but not previous myocardial infarction or stroke.
The relative risk reduction of 13% obtained with respect to the secondary endpoint was well within the 95% CI (2-28%) seen in PROactive. Sensitivity analyses adjusting for baseline differences between PROactive study groups, applying changes in HbA1c, systolic blood pressure, HDL cholesterol, and weight in proportion to baseline values, or applying treatment effects gradually over the first year, gave similar results (relative risk reductions of 14%, 12%, and 11%, respectively). The model-estimated 11% favourable relative risk reduction in congestive heart failure associated with the reported risk factor changes contrasts sharply with the 39% relative risk increase (p=0·007) in heart failure needing hospital admission seen with pioglitazone in the PROactive study.
The non-significant 10% relative risk reduction seen with pioglitazone for the PROactive primary cardiovascular endpoint was substantially less than the 20% or more on which Dormandy and colleagues based their power calculation.4 Our analysis supports the explanation that any macrovascular benefits seen reflect the modest improvements obtained in established risk factors, with little evidence that changes seen previously in novel risk factors with pioglitazone have any substantive effect. More worryingly, the estimated macrovascular benefits are offset by an increased risk of heart failure and concerns about increased peripheral revascularisation rates.
In the primary prospective analysis of 753 overweight patients randomly assigned metformin or dietary treatment after diagnosis of type 2 diabetes in the UKPDS,5 metformin was associated with a 39% relative risk reduction in myocardial infarction (p=0·010) and 36% relative risk reduction in all-cause mortality (p=0·011), effects not thought to be mediated through established risk factors. Metformin is the only antidiabetic agent that has been shown definitively to reduce macrovascular risk in overweight type 2 diabetic patients5 and it remains the first-line agent of choice recommended by most treatment guidelines.
RS has received a speaker's honorarium from Takeda. The other authors declare that they have no conflict of interest.
PROactive study
The Lancet, Volume 367, Number 9504, 07 January 2006
John S Yudkin a b and Nick Freemantle b
John Dormandy and colleagues1 claim that pioglitazone significantly reduces the incidence of cardiovascular endpoints in diabetic patients with known cardiovascular disease. However, in concentrating on the "main secondary endpoint" they ignore the statistically neutral primary outcome. One assumes that, had the 10% reduction in primary endpoint events been significant, Dormandy and colleagues would have felt no need to emphasise the analysis of the main secondary endpoint. Since there was apparently no provision in the statistical analysis plan for multiple primary endpoints, the result on the main secondary endpoint should be regarded as hypothesis-generating or exploratory rather than groundbreaking proof.2
The argument raised in the accompanying Comment3 that including procedural endpoints within the primary endpoint produces a bias against pioglitazone are incorrect, and post hoc: the finding of a nominally significant result if these endpoints are excluded does not equate to bias. In a review of the uses and abuses of composite outcomes in clinical trials, we identified that such procedural endpoints are more generally associated with an increased likelihood of achieving a significant primary outcome (odds ratio 2·24, 95% CI 1·15-4·34).4
Thus although Dormandy and colleagues might have been bold (or idiosyncratic) to include leg outcomes in their primary endpoint, it seems doubly so to exclude leg outcomes and coronary revascularisation procedures from their main secondary endpoint. The meta-analysis of statins, published in the same issue,5 shows a parallel 23-24% reduction in major coronary events and in coronary revascularisation per mmol/L reduction in LDL cholesterol. If the same were not the case for glitazones, it would require postulating a very different process of reversing pathophysiology (or very different pattern of interventions in eastern Europe)-something which would seem curious for Dormandy and colleagues to do a priori. Indeed, less frequent use of revascularisation in contributing countries would not necessarily indicate a reduced effect of an intervention, since it is the relation between the rate of events in each randomised group which is of interest. Clearly, however, a lower revascularisation rate would lead to a correspondingly lesser contribution from that component of the primary outcome than in more interventionist countries.
An appropriate conclusion from PROactive is that glitazones reduce cardiovascular event rates with a point estimate of around 10-15%, but with a confidence interval including zero-a result in line with the equally uncertain reduction in macrovascular events seen in the United Kingdom Prospective Diabetes Study (UKPDS).5 Judging from the way in which the results were presented at the European Association for the Study of Diabetes in Athens in September, 2005, and from the website (http://www.proactive-results.c...), there is a risk that the marketing divisions of Takeda and Eli Lilly will use these questionable results mercilessly in their promotional material.
NF has received funding for research and consultancy from several pharmaceutical and device companies that manufacture products for diabetes, including Takeda Pharmaceuticals, which sponsored PROactive.
PROactive study - Authors' reply
The Lancet, Volume 367, Number 9504, 07 January 2006
John Dormandy a, on behalf of the PROactive Writing Committee
Mark Goldstein raises the issue of bladder cancer in PROactive. The incidence of cancers overall was similar between groups, although imbalances in some categories of cancer were seen. For example, the incidence of breast cancer was significantly higher in the placebo group, whereas the incidence of bladder cancer was non-significantly higher with pioglitazone. Owing to the expected variability associated with the small number of cases in each cancer category, we did not conclude that pioglitazone protected against breast cancer. Likewise, it was not reasonable to conclude that pioglitazone causes bladder cancer. The original concerns about PPAR _/_ agonists causing bladder tumours are based on findings in male rats, which are probably related to species-specific and sex-specific microcrystal formation rather than genotoxicity.1
Pierre-Jean Guillausseau, John Yudkin, and Nick Freemantle, and Rury Holman and colleagues are concerned by our interpretation of the main secondary composite endpoint of all-cause mortality, myocardial infarction, and stroke. As described the PROactive report, this endpoint was prespecified in the statistical analysis plan and submitted to the Food and Drug Administration before unblinding. It showed a 16% relative risk reduction with pioglitazone (p=0·027). We agree that in isolation this finding would not allow any definite conclusion to be drawn. However, the primary composite endpoint, which also included silent myocardial infarction, acute coronary syndrome, major leg amputation, and coronary and leg revascularisation, also showed a reduction with pioglitazone-10% relative risk reduction (p=0·095). Indeed, there were fewer events in the pioglitazone group for each component individually, except for amputation, which had the same frequency in both groups, and leg revascularisation, which was commoner in the pioglitazone group.
The accompanying Comment,2 quite reasonably in our view, draws attention to the fact that all the components of the main secondary endpoint are disease-related, rather than procedure-related. The one disease-related component of the primary endpoint not included in the main secondary endpoint is acute coronary syndrome. If this is added to the main secondary composite endpoint, the relative risk reduction with pioglitazone is 17% (p=0·0103). The totality of the trial evidence, coupled with a very plausible mechanism of action (increase in HDL cholesterol, decreased triglyceride, and decreased HbA1c), justifies the conclusion that pioglitazone reduces the composite of all-cause mortality, non-fatal myocardial infarction, and stroke.
Yudkin and Freemantle write, "One assumes that, had the 10% reduction in primary endpoint events been significant, Dormandy and colleagues would have felt no need to present the analyses of the main secondary endpoint". On the contrary, it is the effect on mortality, myocardial infarction, and strokes that would be of most interest to patients and the regulatory authorities. Presumably, Yudkin and Freemantle would have been content to conclude that pioglitazone is effective in preventing macrovascular events if the 10% reduction in primary endpoints had been significant, even if the main secondary endpoint of all-cause mortality, myocardial infarction, and stroke had shown no significant difference. Would the main secondary endpoint then have assumed a "clinical, if not statistical" significance?
The PROactive study cannot be compared to Steno-2, as suggested by Peter Gaede and colleagues. Steno-2 involved a different population, and looked at a multifactorial intervention in an open study over more than twice as long a period. We agree that the reduction in blood pressure was probably part of the explanation for the beneficial effect seen in PROactive, but the improvement in lipid profile, better control of glycaemia, and some of the other known actions of pioglitazone, probably also played a part. The data are being analysed further to try to elucidate the relative contribution of the different effects of pioglitazone.
Holman and colleagues draw attention to the increase in reported heart failure cases, also highlighted in the Comment. Heart failure episodes and hospital admissions for heart failure were unadjudicated events and the diagnostic criteria were not specified in the protocol. The diagnoses were made by the investigators, who might have been sometimes misled by the known increase in oedema with pioglitazone. We are surprised by the comparison in the Comment of first events for endpoints avoided, on the positive side, versus all episodes of heart failure, on the negative side. Nevertheless, the Executive Committee is further exploring the issue of heart failure by setting up an independent commission of experts to review, in a blinded way, all source material for patients reported to have had a fatal or non-fatal heart failure at any time during the trial.
  icon paper stack View Older Articles   Back to Top