Pre-diabetes as a contributor to stroke

BMJ 2012; 344 doi: http://dx.doi.org/10.1136/bmj.e3285 (Published 07 June 2012) Cite this as: BMJ 2012;344:e3285
  1. Jonathan R Treadwell, senior research analyst
  1. 1Technology Assessment Group, ECRI Institute, Plymouth Meeting, PA 19462, USA
  1. jtreadwell{at}ecri.org

Preventive measures should be considered

Numerous studies have suggested an overall association between pre-diabetes and future stroke, but several confounding factors (such as age, obesity, and hypertension) make it impossible to know the exact size of this association.

Pre-diabetes and “not pre-diabetes” cannot be randomly assigned, so the best that we can hope for is an observational design that controls for all important cardiovascular risk factors. But researchers disagree about which factors need to be controlled for. For example, one study controlled for seven factors,1 whereas another controlled for 11 (only three of which were also controlled for by the first study).2 Unsurprisingly, the two studies yielded markedly different estimates of the association between impaired glucose tolerance and future stroke (14% increased risk in the first, but 71% in the second).

The linked meta-analysis by Lee and colleagues (doi:10.1136/bmj.e3564) examined data from 15 prospective studies that enrolled more than 700 000 participants.3 Importantly, all studies reported multivariate analyses that controlled (in some way or other) for various risk factors for stroke.

Despite great variability in the methods and quality of the 15 studies, the overall message of the meta-analysis is that pre-diabetes is one of many contributing causes of stroke. The immediate implication for clinicians is to consider preventive measures for those with pre-diabetes, both to prevent progression to frank diabetes (lifestyle modifications) and to prevent stroke itself (drugs).

The reviewers encountered several problems when interpreting the data. For example, how should we define “pre-diabetes”? Some analyses defined it on the basis of fasting plasma glucose of 100-125 mg/dL (5.6-6.9 mmol/L), which aligned with the 2003 definition of the American Diabetes Association. Others set the bar a little higher at 110-125 mg/dL (6.1-6.9 mmol/L), which aligned with the association’s 1997 definition. Still others defined it on the basis of a glucose tolerance test (impaired glucose tolerance; glucose 7.8-11 mmol/L). And yet others used a combination of fasting plasma glucose and glucose tolerance. The reviewers chose to combine these last two categories, and only this combined category consistently predicted stroke.

A second problem was that some studies may have inadvertently included patients who had frank diabetes. Specifically, seven studies defined pre-diabetes on the basis of fasting plasma glucose only, so some of their patients may have had levels of glucose intolerance consistent with diabetes. When the authors excluded these seven studies they found no statistically reliable association between fasting plasma glucose and stroke.

In addition to the clinical implication that preventive measures should be considered for patients with pre-diabetes, the review underscores two points. Firstly, it is too simple to categorise people as having “diabetes,” “pre-diabetes,” or “not diabetes.” A natural continuum of glucose management is the biological reality, and we must be wary of putting people into boxes. And the same point applies to “statistical significance,” which is a box-like approach to data analysis. What we have are continuously measured associations, with varying degrees of trustworthiness.

The second lesson from the current review is that meta-analysis is not just for randomised trials that compare treatments. Lee and colleagues performed a meta-analysis of non-randomised trials of prognostic associations.3 They are to be commended for thinking outside the box of what meta-analysis can achieve. The clinical question must drive the research methods.


Cite this as: BMJ 2012;344:e3285


  • Research, doi:10.1136/bmj.e3564
  • Competing interests: The author has completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declares: no support from any organisation for the submitted work; no financial relationships with any organisations that might have an interest in the submitted work in the previous three years; no other relationships or activities that could appear to have influenced the submitted work.

  • Provenance and peer review: Commissioned; not externally peer reviewed.