The relationship between serum glucose regulation and cognitive processing efficiency in non-insulin dependent diabetes mellitus

A small but growing literature points to the existence of subtle cognitive deficits in otherwise healthy diabetic patients. These deficits have been attributed to peripheral neuromotor dysfunction (Skenazy & Bigler, 1984), attentional disruption (Holmes, 1986), defects in both verbal and visual learning and memory (Mooradian et al., 1988; Perlmuter et al., 1984), and problems with abstract reasoning (Francheschi et al., 1984). Although it is well known that peripheral and central neural conduction velocity is slowed in diabetes (Hansen & Ballantyne, 1977; Pozzessere et al., 1988), no existing neuropsychological studies address the relative contributions of speed and accuracy to the obtained deficits in cognition. In the present study, the neuropsychological performance of 26 patients with non-insulin-dependent diabetes mellitus (NIDDM) was compared to that of an equivalent control group of 19. Patients with poorly-controlled NIDDM (fasting glucose = 225.0 $\pm$ 66.1 mg/dl) were treated with an oral hypoglycemic agent. Those responding well to treatment discontinued their participation after eight weeks. Nonresponders participated for an additional eight weeks on both the oral agent and insulin. Ten of the patients and six of the control subjects participated in this second phase. All subjects received neuropsychological testing at entry and at Weeks 2 and 8 of the patients' treatment. Those in the second phase were also tested at Weeks 10 and 16. The cognitive measures selected to assess groups' speed and accuracy were: a 4-choice reaction time test, the Trail Making Tests with several alternate forms, and a matching to sample paradigm administered both with and without limits on the time available for a response. As predicted, the patients' speed was significantly poorer than that of controls on virtually all measures, while their accuracy was generally unimpaired and subjects' scores were often significantly related to both glucose and HgbA$\sb1$. While across-subjects analyses seemed to indicate that there was improvement in patients' performance relative to controls, this was not borne out by within-subjects analyses of the rate at which groups changed over time. Post-hoc hypotheses regarding the failure to detect a treatment effect and recommendations for future research are outlined.