Intended for healthcare professionals


Licensing drugs for diabetes

BMJ 2010; 341 doi: (Published 06 September 2010) Cite this as: BMJ 2010;341:c4805
  1. Richard Lehman, senior research fellow1,
  2. John S Yudkin, emeritus professor of medicine2,
  3. Harlan Krumholz, Harold H Hines Jr professor of medicine3
  1. 1Department of Primary Care, University of Oxford, Oxford OX3 7LF
  2. 2University College London, London
  3. 3Yale University, New Haven, CT, USA
  1. edgar.lehman{at}

    Surrogate end points are not enough, robust evidence of benefits and harms is needed

    The rosiglitazone (Avandia) story sounds deceptively similar to that of other drugs that have been withdrawn in recent years because of emerging safety concerns, notably rofecoxib (Vioxx). But several of the lessons are importantly different and raise fundamental questions about the criteria for licensing new drugs for diabetes and the way that clinicians use evidence about such drugs to incorporate them into clinical practice.

    The additional 80% relative risk of myocardial infarction attributed to rosiglitazone at recent hearings of the US Food and Drug Administration1 2 is probably comparable to that of several non-steroidal anti-inflammatory drugs that are still on the market.3 The scandal of rosiglitazone lies in the fact that we were prescribing the drug to reduce the serious consequences of type 2 diabetes, the most important of which is myocardial infarction. Yet because no licensing body demanded evidence of effect on cardiovascular risk at the time of licensing, 10 years after the release of rosiglitazone we still cannot accurately quantify the harm to which we were exposing our patients. As clinicians we did not increase this risk knowingly, but ignorantly, because we were focused on the wrong end point: glycated haemoglobin (haemoglobin A1c). We had lost sight of the main reason for treating this complex and progressive disease, which is not to reduce glycaemia but to prevent complications.

    False promise

    Rosiglitazone was licensed because it lowers glycated haemoglobin by about 1 percentage point. Since the UK Prospective Diabetes Study (UKPDS) findings were published in 1998,4 the ostensibly evidence based mantra has been that lowering blood sugar on this scale is clinically important because it reduces the risk of microvascular complications by 25% and might also reduce the risk of myocardial infarction. But a succession of major studies of intensive glucose lowering, involving some 27 000 patients with established type 2 diabetes, has shown neutral, or deleterious, effects on cardiovascular mortality4 5 6 7 8; furthermore, it had no significant effect on the risks of either blindness or renal failure (figure).8 9 Expressed in terms that we could tell our patients, it seems that if 100 people with type 2 diabetes are treated with a regimen that lowered their haemoglobin A1c concentrations by 1%, then over the next five years 99 of them will have the same risk of cardiovascular disease, blindness, or renal failure as they would have had anyway. Unless, that is, the drug being used to reduce haemoglobin A1c itself increases the risk of adverse outcomes such as myocardial infarction, heart failure, and fractures. The point is that all drugs have many effects, and we cannot infer how a drug affects patient risk on the basis of how it modifies a single biomarker.10


    Estimated effect of reducing glycated haemoglobin by 1% over five years on rates of macrovascular and microvascular events and mortality in 27 049 patients with type 2 diabetes8

    When rosiglitazone first appeared in1999, doctors in both primary and secondary care were strongly disposed to welcome a new drug for type 2 diabetes, and it was vigorously promoted to a receptive market. We knew the limits of what could be achieved using the existing treatments with metformin, sulfonylureas, and insulin. Ironically, it was the disappointing lack of effect of lowering blood sugar on macrovascular complications in the UKPDS that made some doctors wonder if we might not do better using a new drug that would help to achieve glycaemic control and was promoted as having a range of potentially beneficial effects on other markers of cardiovascular risk.

    Our mistake then—and we have yet to put it right—was that we did not demand better proof before we embarked on mass medication of a large group of patients who looked to us for advice and treatment. We let them down then; and we will let them down again if we do not require better proof of long term benefit from drugs for type 2 diabetes in the future.

    Licensing requirements

    Media attention has focused largely on the role of rosiglitazone’s manufacturer GlaxoSmithKline in this process, but some of the blame must also lie with clinicians for not insisting on better proof of long term benefits, not to mention safety, from glucose lowering drugs. Drug companies work within the limits of the regulatory system, and we should be deeply worried that even in the wake of extensive professional and public concern about rosiglitazone, the drug licensing agencies in the United States and Europe still decline to demand proof of cardiovascular safety before licensing drugs for type 2 diabetes.

    The Food and Drug Administration has recently stipulated that phase III trials of new drugs for diabetes should show that the excess relative risk of cardiovascular events is clearly less than 80%.11 Given that a typical reduction in glycated haemoglobin of 1% might at best reduce the relative risk of myocardial infarction by 10% and is unlikely to affect microvascular complications,8 this degree of permitted risk is almost incredible. In fact we need firm proof not just of long term cardiovascular safety but of long term cardiovascular benefit, because this is the most important aim of treating type 2 diabetes.

    Drug companies argue that this requirement will delay the emergence of new drugs by up to 10 years, and that it risks undermining the whole system of global commercial support for research, professional education, patient support, and even clinical care. Several promising new drug classes that reduce glycaemia have appeared in recent years. But now we know that tighter glycaemic control below a certain level does not prevent blindness or renal failure in the short term and has uncertain effects in the long term, do we really need a whole arsenal of drugs to lower blood sugar—unless they can be shown to improve outcomes? A change in the regulatory system is not optional but essential for safe treatment of patients with type 2 diabetes. If insisting on pre-marketing proof of safety delays new drugs, this is the price of patient safety and must be paid, perhaps by means of extended patents on such drugs.

    The story of how rosiglitazone was developed and marketed will keep investigative journalists at work for some time to come. For clinicians and patients the lesson of rosiglitazone is much simpler and more fundamental. It is that we need to be absolutely certain that our long term treatments for type 2 diabetes are not causing the harm they are meant to prevent. And if the regulatory bodies do not insist on definite evidence of greater benefit than harm to patients, rather than changes in surrogate end points, they are failing in their basic purpose.


    Cite this as: BMJ 2010;341:c4805


    • doi:10.1136/bmj.c4848
    • doi:10.1136/bmj.c4812
    • Competing interests: All authors have completed the unified competing interest form at (available on request from the corresponding author) and declare no support from any organisation for the submitted work; no financial relationships with any organisation that might have an interest in the submitted work in the previous three years; and no other relationships or activities that could appear to have influenced the submitted work.

    • Provenance and peer review: Commissioned; not externally peer reviewed.