Registry Data Support 'J-Curve' CV Risk Theory for Hba1c in Diabetes

Shelley Wood

September 16, 2011

September 16, 2011 (Lisbon, Portugal) — Researchers say the "J-curve" pattern of cardiovascular risk as it relates to HbA1c levels seen in the ACCORD and ADDITION studies can also be found among patients with diabetes in Sweden.

Presenting findings from almost 33 000 patients with type 2 diabetes here at the European Association for the Study of Diabetes (EASD) 2011 Meeting, Dr Carl Johan Östgren (Linköping University, Sweden) said the findings reinforce the view that intensive glucose lowering increases the risk of cardiovascular events.

"Future diabetes guidelines might include a minimum HbA1c value for patients subjected to treatment with glucose-lowering agents with the potential of inducing hypoglycemia," he said.

Östgren et al's study linked electronic medical records for 32 871 patients with type 2 diabetes over aged 35 years to national registers of deaths, hospital discharges, and prescriptions. Linking the different records enabled them to examine rates of a composite end point of major cardiovascular events (acute MI, heart failure, stroke, or CV death) according to HbA1c levels.

J-Curve "Strongest" in Least Educated Patients

In Cox regression analyses that used age and systolic blood pressure as covariates, the hazard ratio for the composite end point was lowest for HbA1c levels in the range of 6% to 7%, but rose below 6% and above 7%. A similar U- or J-shape was seen for both patients taking insulin and those taking oral antidiabetic agents, although the curve was sharper in the insulin group.

In a secondary analysis that stratified results by education level--grade school, secondary school, or upper secondary or tertiary education, Östgren et al reported that the J-curve shape persisted only for those in the lowest education group.

When asked why he had chosen to zero in on education levels, Östgren explained that this was one type of information available in the databases they accessed and is at least one marker of socioeconomic status, suggesting that efforts to educate lower-socioeconomic groups might be one way of reducing cardiovascular risk in those with diabetes. He acknowledged, however, that this was only one of many issues that could be addressed in these patients, and that education itself could be a marker for a number of different factors.

The Problem With Registries

Östgren's presentation drew fire during the question and answer session following his talk, with some members of the audience pointing out that the higher event rate in the low-HbA1c group likely just reflects a more ill group of patients, with more comorbidities. Another observer pointed to the wide confidence intervals surrounding the furthest reaches of the J-curve and scolded Östgren for not presenting the p values for these low- and high-range patients.

In an interview with heartwire , Dr Daniel Witte (Steno Diabetes Center, Gentofte, Denmark), session moderator, pointed out that a "perfectly flat curve" would actually fall within the J-shape presented by Östgren, making it impossible to know for sure whether a true curve exists.

"You can't exclude the possibility of a fully flat 'curve' statistically," Witte said. "That's one of the problems of showing a figure like this, is that most people are very sensitive, visually, to these curves, and might overinterpret the findings."

The bigger issue, he added, was that while a number of registry and observational studies have shown this kind of a J-shaped curve, "the big problem is assessing whether that's causal or not, because people have low HbA1c for a reason, often related to a disease process. We can't say, on the basis of this kind of evidence, that to lower HbA1c to that level is dangerous. Clinical trial evidence is the only sort of evidence we can use to ultimately make the case that there is a causal association.

Many diabetes experts--Witte among them--believe that it was not a specific lower HbA1c cut-point per se that increased cardiovascular risk in ACCORD; rather, the speed at which the reduction in HbA1c was achieved. Witte believes future guidelines are more likely to caution clinicians against lowering HbA1c too swiftly, rather than singling out a specific number beyond which glucose lowering would be deemed harmful.

Witte also dismissed the analysis of education level, pointing out that this is just one of countless factors that would need to be addressed. "Studies based on registries have so many cases that they are adequately powered to find small differences in almost anything. Generally, it is methodologically better to first define a physiological process that would be underlying this J-curve, then you would have to have a lot more explanation as to why risk might be different according to education level," he said.

Still, Witte acknowledged that registry studies such as this one serve to provide a snapshot and help with "developing ideas" that can then be tested in clinical trials.

In this large, Swedish dataset, for example, he commented "it would be interesting to see if they can look at change in the population over time, to see which patients have HbA1c levels that are stable, which ones are moving up the curve or moving down, and whether that has an association with different incidence rates for CVD."

Östgren disclosed being a speaker and consultant for AstraZeneca (which sponsored the study), Merck Sharp & Dohme, NovoNordisk, Sanofi-Aventis, Bristol-Myers Squibb, Lilly, Pfizer, and Boehringer-Ingelheim.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.