| |
Glitazones: 3 commentaries/editorials in The Lancet
|
| |
| |
The Lancet Oct 2 2007
Comment
Thiazolidinediones, deadly sins, surrogates, and elephants
John GF Cleland a and Stephen L Atkin b
a. Department of Cardiology, Castle Hill Hospital, University of Hull, Hull HU16 5JQ, UK
b. Department of Diabetes, Hull York Medical School, Hull, UK
The heated debate about the safety of thiazolidinediones has made the scientific community and the public uneasy, which is good because complacency is a deadly sin, akin to sloth. This debate was triggered by a meta-analysis of 42, mainly short-term, trials that investigated the effects of rosiglitazone on glycaemic control in more than 27 000 patients, most of whom had type 2 diabetes mellitus.1 The analysis suggested that rosiglitazone increased the relative risk of myocardial infarction or death from cardiac causes by about 50%, although this risk represented an absolute difference of less than 0¥15%. Had no substantial outcome study been underway, this meta-analysis would have been an important stimulus to do one; however, one such study is underway.2 The scare caused by the meta-analysis was widely aired in the medical and lay press. Such publicity could have damaged the conduct of the trial and rendered the outcome neutral, which would provide no proof of safety because stopping rosiglitazone would mean both treatment groups would end up on the same treatment. Accordingly, the steering committee and sponsor of this trial had little choice but to publish an interim analysis. Rates of myocardial infarction and death were low and much the same in patients assigned to rosiglitazone or comparator drug, but there were too few events to prove the absence of excess risk, and the incidence of heart failure doubled in the rosiglitazone group. These findings were similar to those of a large trial that compared pioglitazone and placebo.3
In today's Lancet, Rodrigo Lago and colleagues report a meta-analysis of seven trials of rosiglitazone and pioglitazone in more than 20 000 patients with, or at high risk of developing, diabetes.4 This analysis also noted a doubling in the incidence of heart failure but suggested that neither agent is associated with increased cardiovascular mortality, although the confidence interval cannot exclude a 25% increase.
Lago and colleagues' report included far fewer trials than the previous meta-analysis of rosiglitazone.1 It is not certain which set of trial selection criteria was least prone to bias. When presented with small trials, journals have a bias against publishing negative results but will often publish positive ones. Any resulting meta-analysis will reflect the same bias.
The outcome measures assessed in a meta-analysis should be chosen with care. Myocardial infarction is not necessarily a definitive outcome measure because only a few patients who have myocardial scars consistent with infarction report any event.5 Avoidance of the pain and worry of myocardial infarction is an appropriate goal of treatment, but subsequent disability and mortality can be measured directly. Ankle swelling is usually not due to heart failure and is certainly not a valid endpoint. Thiazolidinediones can cause fluid retention that might result in peripheral and pulmonary oedema but there is little evidence that such drugs affect cardiac function adversely.6 Patients with oedema caused by thiazolidinediones do not seem to have an unfavourable prognosis,3 although symptoms and signs might only resolve when these drugs are stopped. Why should fluid retention be blamed on the heart rather than the kidneys? The diagnosis of heart failure was generally not robust in studies of thiazolidinediones and much of the adverse prognosis of "diastolic" heart failure may be due to non-cardiovascular comorbidity rather than cardiac dysfunction.7
However, all the meta-analyses consistently fail to spot the elephant in the room. Treatments should be effective rather than merely innocuous. Improved glycaemic control is not a surrogate for effective care of patients who have diabetes, which should be to reduce disability and increase lifespan. Surrogate markers of treatment effects have not done well in many cardiovascular disease areas, including hyperlipidaemia,8,9 hypertension,10 or heart failure.11 In the RECORD study,2 more than 90% of patients had no important cardiovascular event in 3¥75 years of follow-up. Having no symptoms and low rates of events leads to difficulties in designing trials to show real benefits to patients. However, such difficulty suggests the absence of an important role for that intervention in present medical practice.
Government should bear the brunt of the criticism for the approval of pointless drugs. Patients and clinicians have always emphasised improvement of symptoms, reduction of disability, and delaying of death as much as safety. The regulatory authorities need greater emphasis on ensuring that drugs have effects that are clinically relevant, both in their actions and extent, without stifling innovation in an industry that is valuable to society. Short patent-life contributes to the development of expensive low-efficacy drugs as companies scramble to make money for profit and to invest in further research. Longer patents, akin to that provided by copyright for an author or a song, would encourage long-term investment in high-quality clinical research. Short-term trials with large numbers of patients designed to show statistically significant but clinically spurious differences would be replaced by trials with fewer patients designed to show substantial and clinically relevant long-term benefits. Trials should also show that such benefits wane if treatment is stopped, to ensure that long-term treatment is effective for chronic relatively low morbidity, such as diabetes and hypertension. The main problem with this arena of debate is the use of surrogate markers and surrogate forms of proof, such as meta-analysis. Let us ensure that the sins of our fathers are not meted out to future generations.
JGFC has: a British Heart Foundation grant for a study of metformin and gliclazide in ventricular function in diabetes; has funding for research into heart failure and carvedilol from GlaxoSmithKline; has received funding from Roche, AstraZeneca, Pfizer, Sanofi, and Servier; has received speaker's honoraria and funding from Takeda for research on candesartan; was part of a GlaxoSmithKline advisory board on rosiglitazone-induced fluid retention; and has worked on a safety committee reviewing adverse events with rosigiltazone. SLA has: a British Heart Foundation grant for a study of metformin and gliclazide in ventricular function in diabetes; has received speaker's honoraria from GlaxoSmithKline and Takeda; and has been sponsored GlaxoSmithKline and Takeda for attendence at the American Diabetes Association.
Editorial
Ensuring drug safety: lessons from the thiazolidinediones
In today's Lancet, Rodrigo Lago and colleagues add to the ongoing analysis of the safety of thiazolidinediones for control of hyperglycaemia. Initial concerns about the cardiovascular risks of one of the thiazolidinediones, rosiglitazone, erupted in a firestorm of controversy when a meta-analysis by Nissen and Wolski was published online in the New England Journal of Medicine in May. Nissen and Wolski found that rosiglitazone (Avandia, GSK) was associated with a significantly increased risk of myocardial infarction, and a risk of death from cardiovascular causes that was of borderline statistical significance. The authors acknowledged that their analysis was limited by the public availability of trial results and a lack of access to patient-level data, and that the pooled studies were not designed to assess cardiovascular outcomes. But they concluded that the risks of rosiglitazone use in diabetes should be carefully considered by both doctors and their patients.
Response to these findings was swift. An editorial that accompanied the paper, while acknowledging the findings' "fragility", called for urgent regulatory action; GSK vigorously defended its product, saying that studies show Avandia's cardiovascular profile to be comparable to other oral antidiabetes agents; and this journal counselled against a rush to judgment, advising patience until the final results of RECORD, a phase III trial designed to assess cardiovascular outcomes, were available. Media coverage was extensive. A US Congressional hearing was hastily called, as was a Food and Drug Administration (FDA) Advisory Committee meeting. The Congressional hearing was a spectacular display of partisan agendas and bipartisan ignorance. The FDA meeting resulted in a 22 to 1 vote to keep Avandia on the market, and to add a "black box" warning on the label of the risks of the drug's use in patients with congestive heart failure. If this sequence of events sounds familiar, it is because a nearly identical path was trodden when the safety of selective cyclo-oxygenase 2 (COX-2) inhibitors was called into question. Then, as now, the question was about the drugs' cardiovascular effects.
Other analyses of the thiazolidinedione have followed the NEJM paper, including two papers that appeared in JAMA on Sept 12, one on the risk of cardiovascular events with pioglitazone, the only other available agent in the thiazolidinedione class (Actos, Takeda), and one assessing long-term cardiovascular risk with rosiglitazone. Here, both drugs were associated with an increased risk of heart failure, though without an associated increase in mortality. This is the same conclusion reached by Lago and colleagues: patients taking thiazolidinediones seem to have a higher risk of congestive heart failure, but do not have a higher risk of death from cardiovascular causes.
Is there then a bottom line to all these bits of evidence? What should doctors and patients do? Is there in fact enough good evidence on which to decide anything? It seems that the jury is still out for the thiazolidinediones as a class. But there are many interim take-home messages. Some of these are highlighted by Comments in today's issue. First, it must be remembered that meta-analysis is a technique with important limitations. And the studies on which the thiazolidinedione meta-analyses are based have thus far all involved surrogate markers; the studies were not designed to assess cardiovascular outcomes, but rather improved glycaemic control. This outcome, as Victor Montori and colleagues note, is not a patient-centred one. The current clinical emphasis on glucose control (as measured by HbA1c) skirts the outcomes that matter most to patients-microvascular and macrovascular complications, quality of life, and survival.
These commentators highlight issues that must be taken into account in the ongoing debate about thiazolidinediones. Future trials ought to be designed with these issues firmly in mind. Further, it is no secret that the regulatory system is also in urgent need of repair. Manufacturers must do-in a timely fashion-postmarketing studies that assess the long-term safety of their drugs, and regulatory agencies must hold manufacturers' feet to the fire to ensure that these are performed, performed properly, thoroughly evaluated, and made available to guide decisions about prescribing. Agencies like the US FDA must have the resources and authority to close what is now a potentially dangerous gap. Unless limitations on the understanding, analysis, and communication of drug safety issues are addressed, the thiazolidinediones might simply become the latest in a series of preventable drug disasters.
Comment
Patient-important outcomes in diabetes-time for consensus
Victor M Montori a, Gunjan Y Gandhi a and Gordon H Guyatt b
a. Knowledge and Encounter Research Unit, Division of Endocrinology, Diabetes, Metabolism, Nutrition, and Internal Medicine, Mayo Clinic College of Medicine, Rochester, MN 55905, USA
b. CLARITY Research Group, Department of Clinical Epidemiology and Biostatistics, Faculty of Health Sciences, McMaster University, Hamilton, ON, Canada
The epidemic of type 2 diabetes and its resulting complications-cardiovascular, renal, ophthalmological, and neurological-has generated widespread alarm. Despite the alarm, the resultant flurry of investigation, and an interval of 50 years since the introduction of the first oral agent for the treatment of diabetes, we remain uncertain if any antihyperglycaemic drug can favourably affect key patient-important outcomes, including morbidity, mortality, and quality of life. Why this profound ignorance? One key reason is that diabetes trials have focused on the effect of interventions on glucose control (eg, HbA1c) rather than on patient-important outcomes.1
Measurement of HbA1c purportedly captures the effect of therapy on diabetes complications: the lower the HbA1c achieved, the lower the risk of diabetes-related complications. The Diabetes Control and Complications Trial (DCCT) offers the best evidence in support of this putative causal link.2 DCCT established that, in patients with type 1 diabetes, the near-physiological replacement of insulin led to lower HbA1c concentrations and reductions in the risk of microvascular complications (retinopathy, neuropathy, and nephropathy) and, with less certainty, reductions in the risk of macrovascular complications.3 In this context, HbA1c may be a credible surrogate endpoint.
Unfortunately, HbA1c loses its validity as a surrogate marker when patients have a constellation of metabolic abnormalities, when the most common complications are macrovascular, and when the treatments have multiple poorly understood effects. This situation characterises type 2 diabetes.
In the UK Prospective Diabetes Study, insulin and sulfonylureas achieved similar HbA1c reductions as did metformin, but failed to achieve the same reduction in the risk of myocardial infarction and other important diabetes complications.4 Indeed, sulfonylureas and glitazars might reduce HbA1c and increase cardiovascular risk.5,6 Recent evidence7 that treatment of diabetes with rosiglitazone increases cardiovascular risk has further reduced the credibility of HbA1c as an adequate surrogate.
Available antihyperglycaemic agents have effects that go beyond glycaemic control. Thus revelations of a paradox between lower HbA1c concentrations and increased risk of coronary events should not surprise. Both clinicians and experts must acknowledge that it matters how (ie, with what agent or agents) we achieve glucose-control targets.
Medical therapeutic history is littered with instances in which reliance on surrogate outcomes has provided misleading results. Encainide and flecainide, antiarrhythmic drugs that suppressed malignant-looking arrhythmias, increased mortality. Oestrogen-replacement therapy favourably affected HDL-cholesterol and carotid plaque, but did not prevent cardiovascular events. Torcetrapib increased HDL-cholesterol and modified vascular morphology,8 but increased mortality. Despite these lessons, trialists continue to err in relying on surrogate outcomes to substitute for effects on patient-important outcomes.
The direct measurement of patient-important outcomes in diabetes trials remains uncommon. Only one in five randomised trials in diabetes published in top general and specialty journals was powered to measure such outcomes.9 The future seems even grimmer. Clinical trial registries reveal that only 14% of randomised trials in diabetes will assess patient-important outcomes as primary endpoints. Even large trials miss the opportunity to measure these outcomes: the ADOPT trial followed up over 4000 patients for 4 years but was powered to assess when patients will need to add further therapy to maintain glycaemic control, rather than whether therapy improves patients' wellbeing, limits their signs and symptoms, or prolongs their lives.10
The medical community is increasingly aware of the need to engage patients with chronic conditions in clinical decisions.11 A conscientious patient with diabetes would like to choose drugs that maximise benefit (reduce complications) and minimise burden (route of administration, need for self-monitoring, cost), side-effects (weight gain), and efficacy failure (hypoglycaemia). But if an informed asymptomatic patient asks whether they will be better off if they follow a regimen that will reduce HbA1c by 0¥5%, the lack of high-quality evidence leaves us unable to provide a satisfactory response. Clinicians and policymakers who view quality-of-care criteria that focus on reductions in unsubstantiated surrogate markers should share patients' well-warranted incredulity.
What is the way forward? Surrogate outcomes allow smaller, shorter, and cheaper trials, provide a faster offering of more choices to patients and clinicians, save research money, and allow new drugs more rapid access to market. The apparent benefits are, however, a mirage and the apparent savings represent false economy. Any savings are quickly overwhelmed by costs associated with potentially ineffective or even harmful (yet heavily advertised) expensive therapies, and the incremental costs of treating the harms these interventions might cause. Patients and society may end up paying dearly for drugs that cause more harm than good.
The medical community should insist that we invest the resources needed to do trials that ascertain the effect of interventions on patient-important outcomes. This policy will prevent the premature dissemination of therapies that ultimately prove harmful, facilitate patients' participation in decisionmaking, and speed the day when we can confidently offer treatments that will provide long-term benefit to patients with diabetes.
We declare that we have no conflict of interest.
|
|
| |
| |
|
|
|
