In reply

Dr Khoo has suggested another possible mechanism via obstructive sleep apnea (OSA) to explain the direct relationship between fasting plasma glucose (FPG) level and increasing risk of stroke outcomes in men from the Aerobics Center Longitudinal Study. Certainly, substantial evidence supports the strong relationship between FPG level and increased risk of cardiovascular diseases, as discussed in our article and others recently published in Mayo Clinic Proceedings.^1-3 Clearly, OSA is related to obesity/metabolic syndrome,⁴ and our data do not exclude the possibility that OSA may be related to stroke. Although we do not have sufficient information on OSA, we have data on unexplained shortness of breath while sleeping. Across normal FPG levels, impaired fasting glucose levels, and undiagnosed diabetes mellitus, there were 428 (1.2%), 67 (1.2%), and 25 (2.2%) men who reported having shortness of breath while sleeping (trend P=.01). However, additional adjustment of this variable in our final model did not significantly change the observed relationships between FPG level and each stroke outcome. Excluding these 520 men with shortness of breath while sleeping also had no effect on the observed associations. Although we could not perform a more formal assessment of OSA in our large cohort, these additional analyses suggest that OSA is not the obvious cause of the increased stroke risk that we observed in our large cohort of asymptomatic men with FPG values of 110 mg/dL and higher.

In regard to Dr Khoo's suggestion of examining follow-up FPG data to demonstrate whether improvement of FPG levels in those patients at risk would in fact decrease stroke incidence, we agree that this is an important point, but this is an acknowledged limitation of our study. Like many other prospective epidemiologic studies, information about changes in levels of key risk factors (FPG in this case) or changes in other relevant health behaviors were not available to us during the intervening years of follow-up for the full cohort. It is possible that these factors changed over time for many participants. Many of the men initially classified by impaired fasting glucose level at study baseline may have been diagnosed as having diabetes or were treated during the follow-up period. The question then becomes how such changes might influence the results of our study. We believe that this type of misclassification would lead to an underestimation of the magnitude of the association for stroke outcomes evaluated and would also make it less likely that we would find significant results. Future studies with detailed OSA information and follow-up FPG values are warranted to clarify whether fasting hyperglycemia is an important predictor of stroke outcomes or is just a surrogate marker of another underlying condition that independently contributes to an increased risk of cerebrovascular events.