Search tips
Search criteria 


Logo of bmjThis ArticleThe BMJ
BMJ. 2007 December 8; 335(7631): 1185.
PMCID: PMC2128670
Life and Death

The road to hell . . .

Iona Heath, general practitioner, London

Too often, evidence from clinical trials is being shamelessly extrapolated across time, population subgroup, and condition

The quality improvement movement within medicine is awash with worthy good intentions. The trouble is that those in the driving seat are often in a hurry and have an unfortunate tendency to cut important corners. If, as a result, the principles and methods of science are ignored, the good intentions become self defeating, unrealisable, and dangerous. This is sadly what seems to be happening. Good intentions are being enacted through a series of interventions that begin to represent something of a nadir in the ways in which research is translated into practice.

We are witnessing a degradation of knowledge, which results from its bureaucratic application to whole populations. Too often, evidence from clinical trials is being shamelessly extrapolated across time, across population subgroup, and across condition. Again and again, efforts are concentrated on crude process measures, while clinical outcomes that are genuinely significant for patients because they reduce or delay suffering or prolong life are ignored. Thirdly, the present state of clinical evidence systematically neglects the reporting of harms (

Let us examine these trends in relation to the care of type 2 diabetes in those belonging to that rapidly increasing proportion of the population who are aged over 80. Within the Quality and Outcomes Framework of the UK general practice contract and within similar systems in an increasing number of countries, doctors are incentivised to reduce patients’ blood levels of glycosylated haemoglobin below 7.5%, whatever their age. The application of this target to the older people exemplifies every stage of the perversion of knowledge that I am trying to describe. There is almost no trial information available which is directly relevant to the care of those who are more than 80 years old. The deficit is glossed over by applying information gained from trials involving much younger people to a completely different population subgroup. Yet older people, because of their necessarily more limited life expectancy, have much less opportunity to benefit from interventions aimed at reducing the risk of future complications. On the other hand, because of increasing physical frailty and being much more subject to multiple and compounding conditions, older patients are at greater risk of the adverse effects of polypharmacy and these risks, unlike the potential benefits, are incurred immediately.

An even more fundamental problem is that the level of glycosylated haemoglobin is a reductive and simplistic process measure which correlates poorly with clinically significant complications of type 2 diabetes in any age group and not at all for people over 80. The emphasis on these levels seems to derive from an extrapolation across conditions. In 1993, the results of the Diabetes Control and Complications Trial (DCCT) documented the potential and substantial benefits of tight control of blood glucose levels in patients with type 1 diabetes. The authors warned explicitly against using their findings in the care of those with type 2 diabetes, but that is precisely what has happened in the intervening years. Any use of the DCCT findings in the care of older people is particularly shocking given that the average age of the trial participants was 27 years.

The results of the United Kingdom Prospective Diabetes Study (UKPDS), published in 1998, have been used to support the emphasis on tight control of blood glucose levels in type 2 diabetes. However, as McCormack and Greenhalgh have so carefully argued (BMJ 2000;320:1720-3, doi: 10.1136/bmj.320.7251.1720), the study in fact showed no clinically important benefit from the control of blood glucose with sulphonylureas and insulin over more than 10 years. Differences were detected in process measures such as albuminuria and the progression of retinopathy, but there were no differences in the outcomes that are important to patients: renal failure, blindness, and loss of visual acuity. The only effective agent was metformin, which is undoubtedly beneficial through mechanisms that seem poorly understood and are not completely explained by the effect on blood glucose levels.

Many treatment trials underemphasise the possibility of harm. To take the UKPDS as a single example, harms were recorded per year, whereas benefits were expressed over 10 years. The effect of this kind of reporting is to imply that harms are minimal in comparison to potential benefits. This is not always the case. The most frequent harm caused by intensive treatment of blood glucose levels is hypoglycaemia and some degree of hypoglycaemia was noted in 37% of those receiving the most intensive treatment. This is likely to be particularly devastating in older people because of their already increased vulnerability to falls, and the exclusion of older people from trials means that harms which are relevant to them may be significantly underestimated.

There seems little doubt that the treatment of type 2 diabetes in very old people should be carefully adjusted to their individual circumstances and preferences and not dictated by a futile insistence on ever tighter control of blood glucose. Every working day, thoughtful clinicians, subject to crudely implemented programmes of quality improvement or payment by results, are being confronted not only by the mismatch between the imperatives of such programmes and the needs and preoccupations of their patients, but also by the disturbing possibility that they may actively be harming their oldest and frailest patients. Maybe, if we as doctors have the courage to learn from our mistakes, the Quality and Outcomes Framework will, in the end, provoke a reaction sufficient to make it the cause of positive change and more integrity in the ways in which research is applied in clinical practice. Good intentions are not enough.

Maybe the Quality and Outcomes Framework will, in the end, provoke a reaction sufficient to make it the cause of positive change and more integrity in the ways in which research is applied in clinical practice

Articles from The BMJ are provided here courtesy of BMJ Publishing Group