Article Text

Download PDFPDF

‘Hard,’ ‘soft’ and ‘surrogate’ endpoints in diabetes
  1. John S Yudkin1,
  2. Emma M Eggleston2,3
  1. 1Division of Medicine, University College London, London, UK
  2. 2Department of Population Medicine, Harvard Medical School and Harvard Pilgrim Health Care Institute, Boston, Mass, USA
  3. 3Division of Endocrinology, Diabetes and Hypertension, Brigham and Women's Hospital, Boston, Mass, USA
  1. Correspondence to Professor John S Yudkin, Division of Medicine, University College London, 28, Huddleston Road, London N7 0AG, UK; j.yudkin{at}ucl.ac.uk

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Since the publication of the UK Prospective Diabetes Study (UKPDS),1 much of the discourse around the management of type 2 diabetes has been dominated by an emphasis on reduction in levels of glycated haemoglobin (HbA1c). This emphasis is ubiquitous in clinical guidelines, in quality measures for healthcare organisations, and in metrics for reimbursement. Yet, recent evidence suggests the need for a shift in this paradigm. This is particularly the case in terms of the ability of lowering HbA1c to predict cardiovascular benefit, but we argue that it also applies to the persistent faith in glycated haemoglobin reduction as a marker of microvascular complication risk. We have previously argued that because HbA1c is a poor surrogate disease marker for cardiovascular disease (CVD), it is important to engage patients with diabetes in decision making based on discussion of the risks and benefits of treatment on outcomes that have clinical meaning to them.2 While recent guidelines now call for its use as a treatment measure to be tailored to the individual patient,3 we argue now that there remains a reluctance to apply this in practice. We argue that the reasons for this reluctance are several, and include an overconfidence in the impact of HbA1c reduction on the early markers of diabetic microvascular disease, which we term ‘soft endpoints’. We outline the paucity of trial data on the impact of HbA1c reduction on the clinically relevant ‘hard endpoints’ of vision loss and renal failure,2 and of clinical trial data in younger patients or demographically diverse populations.

Successive reductions in diagnostic thresholds for type 2 diabetes4 mean that at any given time, many people have a condition that is arguably less an asymptomatic disease than a risk factor for future disease. This is even more so for the growing numbers of people with ‘pre-diabetes’.5 Combined with ever more rigorous treatment targets, there is the increasing likelihood that side effects of treatment may produce more symptoms, both in the immediate and long term, than the risk factor/disease itself. This is particularly the case in a disease for which the prevalence is highest in those over 65 years of age,6 a group least likely to benefit from tight control and more likely to suffer harm from overtreatment and polypharmacy.

The evidence for intensified glucose control in type 2 diabetes has largely been based on the findings of the UKPDS.1 This study showed a nearly significant 16% reduction in the incidence of myocardial infarction, and a 25% reduction in microvascular risk with a 0.9% lower HbA1c over 10 years. It has been used to justify a virtually universal target HbA1c of 7.0%. In the last 4 years, however, the advent of three other large glucose-lowering studies7–9 has put into question the benefits of intensified glucose lowering for CVD risk reduction, and with it, the appropriateness of blanket targets. In summarising the clinical trial findings to date, the recently published American Diabetes Association/European Association for the Study of Diabetes (ADA-EASD) Position Statement on managing hyperglycaemia3 agrees that every HbA1c reduction of ∼1% may be associated with a 15% relative risk reduction in non-fatal myocardial infarction, but without benefits on stroke, or on cardiovascular or all-cause mortality. They noted suggestions from these studies that patients without overt CVD, with lower baseline HbA1c, and with shorter duration of disease (perhaps because of a possible ‘legacy effect’ of tight control early in the disease) could have benefited from the more intensive strategies. Moreover, because intensive glucose lowering produced benefit on retinal and renal endpoints and neuropathy, the ADA-EASD Position Statement states that intensive treatment is thus warranted in most patients to reduce the incidence of microvascular disease, but that the accumulated data suggest that not everyone benefits from aggressive glucose management. It therefore follows that it is important to individualise treatment targets according to age, diabetes duration, presence of CVD, life expectancy and vulnerability to treatment side effects. Most importantly, the desires and values of the patient must be considered, with shared decision making based on synthesis of best available evidence from the literature. The Position Statement also points out that using the percentage of diabetic patients who are achieving an HbA1c <7.0% as a quality indicator, as promulgated by various healthcare organisations, is inconsistent with the emphasis on individualisation of treatment goals. This emphasis on individualised care and shared decision making contrasts with the prevailing approach in Europe and the USA, in which HbA1c targets are generally applied by healthcare organisations and regulatory bodies without distinction between populations at differing risk of adverse outcomes from diabetes and its treatment.

By calling into question the validity of HbA1c as the primary driver in diabetes care, and emphasising individualisation of treatment targets, this Position Statement is making a bold leap which has profound implications for patient care. The Statement also took the important step of questioning the value of HbA1c as a surrogate endpoint for drug registration. It is a curious phenomenon that as a specialty, cardiology has been far quicker than has diabetology at grasping the need for studying clinically relevant, rather than surrogate, endpoints for drug registration. It is inconceivable to contemplate a company trying to license a new HDL-raising compound solely on the basis of its effects on HDL-cholesterol levels. While such studies will have impacts on the lead-in period for registration of any new product, the saga of rosiglitazone10 demonstrated that without such an approach, both risks to patients and costs can be enormous. Following the experience with rosiglitazone, the US Food and Drug Administration recognises that there can be no assumptions about full reversibility of glycaemic impact on CVD with glucose lowering, and now require hard cardiovascular endpoint studies for new diabetes (and obesity) drugs.11 The UK's National Institute of Health and Clinical Excellence (NICE) has taken a different approach, as is clear, for instance, in its consideration of weekly prolonged-release exenatide.12 Despite commenting that ‘…while HbA1c was considered an appropriate surrogate marker in type 2 diabetes…, long-term efficacy and safety data with weekly prolonged-release exenatide, particularly relating to cardiovascular outcomes, are needed,’ the Evidence Review Group of NICE assessed prolonged-release exenatide as providing cost-effective treatment for type 2 diabetes. The Group came to this conclusion using epidemiological data from the UKPDS to justify the estimate that a 1% difference in HbA1c is associated with a 16% difference in risk of myocardial infarction, and a 37% difference in risk of microvascular disease. Thus, despite the reservations about cardiovascular safety, this Group appears not only to accept the concept of reducing levels of HbA1c as benefitting the incidence of diabetes complications, but to do so using the epidemiological relationships with outcomes rather than the effects on the relationships between the two from interventional studies. Alternately stated, they considered, and then dismissed, the problems inherent in using HbA1c as a surrogate endpoint.

As we have seen, the arguments around HbA1c targets for clinical care, quality appraisal, reimbursements and for HbA1c lowering for drug registration, are increasingly dependent upon expected benefits on microvascular endpoints. These, in turn, are based both on epidemiologial relationships (such as those seen in the epidemiological analyses of UKPDS data13), and on the impact of HbA1c lowering on surrogate endpoints (primarily microalbuminuria and retinal photographs)—which back up the epidemiology. Data from the UKPDS in patients with type 2 diabetes1 suggested a decrease in risk of 25% in retinopathy and nephropathy for every reduction of 1% in HbA1c with intensified treatment. As we have noted,2 because of the low incidence of serious visual impairment, blindness or end-stage renal disease,14 ,15 these studies predomnantly used subclinical markers of eye and kidney disease (retinal photographs or presence of microalbuminuria) - or an unblinded clinical procedure (photocoagulation). Because of the low incidence in these clinical trials of symptomatic complications, or those relevant to patients, such as vision loss or renal failure (doubling of creatinine or nephropathy) meta-analyses of the major studies of intensive glucose lowering in type 2 diabetes show that even 27 000 people treated for 4.4 years provide inadequate power to show benefit on these hard microvascular endpoints.14 ,15 In their recent meta-analysis of glycaemic control in type 2 diabetes, Coca et al15 showed that intensive glucose lowering had markedly greater proportional impact on micro- and macroalbuminuria than on doubling of serum creatinine, end-stage renal failure or renal death. Similarly, the suggestion that slowing of deterioration in retinal morphology is more favourably impacted than that of visual deterioration is reinforced by recent analyses of the Action to Control Cardiovascular Risk in Diabetes Eye Study,16 which showed that retinal photography scales dramatically overestimated the impact of glucose lowering and of fenofibrate on moderate vision loss. The implications are, first, that the question of microvascular benefits of intensive glucose lowering in type 2 diabetes remains unresolved17; and second, that if such benefits exist, the absolute risk reductions are likely to be small, both because of the low absolute incidence of the complications (at least in the types of patients studied in the clinical trials to date) and the smaller proportional reduction on these endpoints when compared with their surrogates.

Our arguments should not be interpreted as a case for ignoring glycaemic control. Because the relationship between level of glycaemia and microvascular risk is log-linear, the greatest absolute benefits, at the lowest risk, will be achieved by improving control of people with the highest levels of HbA1c. The best data exploring this relationship comes from the Diabetes Control and Complications Trial,18 and while this was conducted in people with type 1 diabetes, the observations are likely to be relevant to all those with diabetes. A 1% reduction in HbA1c from 10% to 9% over 6 years reduced the rate of progression of retinopathy from 8 per 100 to 5.5 per 100, while the same 1% improvement from 7.5% to 6.5% produced a fall in retinopathy progression from around 2.5 per 100 to 1.5 per 100. The corollary of improving control at these two levels is a threefold greater increase in the risk of severe hypoglycaemia by treatment intensification at the lower HbA1c level, such that the 1 per 100 gain in 6-year retinal morphology needs to be set beside the incidence of over 70 events per 100 patient-years consequent upon intensified control. In this context, it behoves us to involve fully the individual person in making decisions and choices, as the perceptions and values of patients, and the ‘dysutilities’ of the disease or treatment complications, may differ widely.19 Moreover, informed decision making requires the presentation of absolute risks and risk reductions, ideally based on hard endpoints, rather than the usual sort of patient information which is all too common—‘This treatment will reduce your chance of kidney disease by 25%.’ Finally, because of the limitations of even the largest clinical trials in terms of duration of treatment and paucity of hard endpoints, carefully designed modelling studies, using and extrapolating from such data, will be important for this purpose.

There are important differences between the subjects studied in randomised trials and patients with diabetes in the community at large. The incidence, morbidity and mortality of diabetes varies significantly by factors including socioeconomic status and race/ethnicity.20 ,21 Subjects in the majority of trials to date are Caucasian and are, by definition, receiving care for their diabetes as part of a clinical trial. In addition, even though prevalence of diabetes is some 10 times higher in those over 65 years of age compared with people under 45 years of age, half or more of all patients with diabetes in the UK and the USA are under the age of 65 years,6 ,20 with increasing numbers of patients developing diabetes as young adults or even in their teens.22 Furthermore, the age of onset of diabetes is substantially lower in Asian populations, and several of these populations have particularly high prevalence, and rapidly increasing incidence, of diabetes.23 These patients will have diabetes for much of their lives, and the balance between risks and benefits of glycaemic control on hard outcomes which develop over decades is unknown.

We conclude that it will still be some time before answers can be provided as to whether the true benefits of tight glycaemic control for the hard eye and kidney endpoints are greater than those seen to date, but merely with a long lag period, and whether these benefits vary according to patient characteristics. In the interim, choices about lifelong treatments require participatory decision making between clinicians and their patients based on discussions encompassing harms of treatment as well as benefit, and framed in terms of both absolute and relative risk of patient-relevant clinical endpoints.24 For these purposes, approaches such as probability modelling, might be better able to estimate both the likely impacts of different treatment strategies on symptomatic endpoints, and the risks for any individual, using a combination of absolute risk levels and relative risk reductions.25 These could then be used to personalise treatment decisions according to the dysutilities attributed by the patient to particular strategies or disease states,19 with the option of intensive control being available to all patients with diabetes in consultation with their doctors. Finally, recognising the impact on the drug pipeline, we argue that it might be challenging to introduce similar requirements for microvascular endpoint studies as for cardiovascular. Nevertheless, rigorous modelling studies need to recognise differences in the proportional impact of glycaemia between observational and interventional studies, between shorter and longer studies, and between ‘soft’ and ‘hard’ endpoints.

References

Footnotes

  • Contributors John S Yudkin and Emma Eggleston both contributed to the conceptualisation and writing of the manuscript and approved the final version.

  • Funding This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None.

  • Provenance and peer review Commissioned; externally peer reviewed.