The present analysis using the ADVANCE trial demonstrates that patient-specific characteristics at baseline can be used to quantify the anticipated individual effects of intensive vs standard glucose control in terms of risk reduction for major vascular events and risk increase for severe hypoglycaemia. The estimated net benefit was positive in the majority of patients, depending upon the individual weight assigned to the beneficial and adverse effects of treatment.

 

The ADVANCE trial was designed to evaluate the efficacy of gliclazide-based intensive glucose control in a broad sample of patients with type 2 diabetes. The strategy of intensive treatment yielded an overall 10% relative risk reduction of the primary composite outcome of major micro- and macrovascular events. The effect was largely driven by a reduction in renal events by 21%, whereas the reduction in major cardiovascular events was only 6% [29]. However, meta-analyses substantiate the claim that intensive glucose control provides modest but significant protection from macrovascular disease [7, 8]. At the same time, intensive glucose control in ADVANCE was associated with a more than doubled risk of severe hypoglycaemia, although the average incidence was lower compared with similar trials [6].

 

A recent Cochrane review summarised absolute effects of intensive treatment and showed that the average NNTB5 ranged from 32 to 142 for microalbuminuria and from 117 to 150 for myocardial infarction. In addition, the average NNTH5 ranged from 15 to 52 for intensive glucose control [6]. However, these estimates of beneficial and harmful effects are the averages for entire study populations and are not informative of the treatment effect for individual patients [15]. Therefore, the latest joint statement from the ADA/EASD stressed the importance of a patient-centred approach including the consideration of diabetes duration, history of macrovascular disease and the individual’s hypoglycaemia risk [30]. The present study aimed to provide further details on the potential benefits and harms of treatment intensification for individual patients.

 

We show that prediction models based on multiple characteristics can quantify the estimated beneficial effects on major vascular events and adverse effects on severe hypoglycaemia of a gliclazide-based regime of intensive glucose control in individual patients. At a group level, treating a larger proportion of patients decreased the incidence of major vascular events and increased the incidence of severe hypoglycaemia. However, this is not informative of the risk to benefit ratio of intensive glucose control at a patient-specific level. Despite shared risk factors for both outcomes, they did vary in strength and produced variations in net estimated treatment effect for individual patients. Further, we observed a different relationship between baseline HbA1c level and risk of severe hypoglycaemia in patients assigned to standard treatment compared with intensive treatment. In patients receiving intensive glucose control, the risk of severe hypoglycaemia increased with higher levels of HbA1c, whereas risk slightly decreased in patients receiving standard therapy. A higher risk of hypoglycaemia in patients with poorer glycaemic control was also demonstrated in the ACCORD trial [31]. In ADVANCE, the increased risk was confined to the intensive treatment arm, probably representing the use of more complex glucose-lowering strategies, including more frequent use of insulin, in an attempt to achieve the lower glycaemic target in this group.

 

Due to the combined and multiplicative effect of multiple risk factors, no single characteristic could be used to identify patients with a positive effect. Therefore, we developed a risk score incorporating multiple patient-specific variables to tailor the assessment of the risks and benefits of treatment for the individual patient. For some patients, the benefits of intensive glucose control will largely offset treatment disadvantages (see Text box). On the other hand, for patients with a smaller or absent net benefit of intensive treatment a less stringent treatment target might be considered. Further, individual treatment decisions are critically dependent on the relative appraisal of beneficial and adverse treatment effects. The present analysis indicated that the majority (85%) of patients in ADVANCE derived a net benefit from intensive glucose control when positive and negative effects were assumed to be equally important. However, the comparison of the immediate risk of severe hypoglycaemia in terms of clinical importance and relevance with, for example, the long-term prevention of end stage renal disease is difficult for both patients and clinicians. Some patients are likely to attach greater weight to avoiding severe hypoglycaemia in relation to, for example, their occupation or the presence of hypoglycaemia unawareness. At a group level, a selective treatment strategy aimed at treating the 85% patient subgroup with an estimated net benefit had no effect on the overall incidence of major vascular events or severe hypoglycaemia. Thus, selective treatment according to conflicting risks may benefit the individual patient while achieving similar population level effects compared with treating everyone.

Importantly, we showed that intensive glucose control was not associated with the risk of death from any cause. There was a similar, albeit non-significant, relative risk reduction of death with intensive treatment across different levels of estimated mortality risk as evaluated by the continuous interaction terms between treatment and risk [32, 33]. The interaction of intensive treatment with HbA1c level on risk of death, as reported in a post hoc analysis of the ACCORD trial, was not significant in this study [34]. Hence, the present analyses did not find any evidence of an excess risk of death with intensive glucose control, not even for the potentially most vulnerable patients who are at highest risk of death.

Some limitations need to be considered with respect to this study; first, follow-up data under randomised therapies were only available for a median of 5.0 years, while treatment is often lifelong. In addition, treatment targets for participants were fixed at study enrolment whereas in clinical practice treatment goals are often revised at repeated outpatient visits. Nevertheless, in ADVANCE the intensive treatment target was gradually reached over a period of 3 years. Hence, this paper provides estimates of the individual benefits and harms of adhering to a specific treatment target in the longer term. Second, the present comparison used a composite outcome of major beneficial effects, comprising individual components with a potentially different clinical impact. However, clinical usefulness and interpretation will benefit from such a summary outcome of positive vascular effects. For example, the development of macroalbuminuria might go unnoticed to a patient but does confer a three to fivefold increased risk of (cardiovascular) mortality [35]. Similarly, we focused on the main adverse effect of treatment; effects on weight change were not incorporated, yet earlier concerns about weight gain were largely refuted for the gliclazide-based regimen used in ADVANCE [36]. Also, mild hypoglycaemia (which can, for example, result in avoiding exercise and increased worry about hypoglycaemia that affects daily life) was not incorporated into this study since it is not related to the risk of vascular events or death [11, 37, 38]. Third, the analyses and estimates related to this study apply to those patients who were eligible for inclusion and were treated with a gliclazide-based regimen. At the end of follow-up the most frequently prescribed drugs were gliclazide, metformin and insulin. Although the use of sulfonylureas is decreasing, they are firmly embedded in the current guidelines [30, 39]. Further, novel therapies such as incretin-based drugs and sodium-glucose cotransporter-2 inhibitors may demonstrate different safety profiles and result in different risk to benefit ratios.

 

Fourth, in the absence of available external risk algorithms for the outcomes of interest, we derived new prediction models that are likely to perform optimistically in the sample from which they were derived [24]. Therefore, we evaluated the amount of over-optimism and provided adjusted effect estimates and performance measures. Finally, individual estimates are accompanied by larger uncertainty margins. However, once the causal effects of treatment have been established at a group level, the point estimate is the most likely approximation of the true effect for the individual patient and is most useful to inform medical decisions.

 

In conclusion, the individual effects of intensive glucose control in terms of reducing the risk of major vascular events and increasing the risk of severe hypoglycaemia can be quantified using a multivariable risk algorithm.