Playing the numbers game

Repercussions continue from the recent removal of the antidiabetic drug rosiglitazone from the European Union market.1 Whether this treatment's protracted, unwelcome presence resulted from regulatory failure, differential interpretation of the relevant data or wishful thinking by clinicians is debatable. But deep in this story lies a flawed fascination with numbers. Ever since publication of the United Kingdom Prospective Diabetes Study (UKPDS) in 1998, it has been used to support a mantra that the greater the lowering of glycated haemoglobin (HbA1c) in people with type 2 diabetes, the greater the benefit.2 This surrogate measure has been used to support a treatment strategy that has helped to get drugs licensed, driven guideline development and even determined the amount that GPs get paid. And yet in the UKPDS, the degree of HbA1c reduction was not related to measures that matter most - reductions in the rate of macrovascular events, that is, patient-oriented outcomes. Somehow, though, the surrogate number became the main outcome of interest and synonymous with ‘benefit‘.

Used appropriately, surrogate markers play a key role in therapeutic management, by faithfully indicating current clinical status or helping to predict outcome. Obvious examples in routine use include glucose, cholesterol and blood pressure. But the problem comes when surrogate markers are used instead of patient-orientated outcomes without sufficient evidence that the two are strongly linked. This seems to happen often in early assessments of some new drugs: here surrogate marker evidence that can be accumulated relatively easily and quickly dominates considerations of the treatment's worth. There needs to be greater realisation that playing the numbers game doesn't automatically help patients (and, as with rosiglitazone, might actually cause harm). In particular, blindly chasing a numerical surrogate target with multiple unproven drug interventions may risk more than it achieves. It should go without saying that, …