We thank Kubiak et al. (1) for their thoughtful appraisal of our study relating to memory response in adults with type 2 diabetes following carbohydrate ingestion (2). We agree entirely with their comment that the underlying origins of memory impairment in this population is poorly understood and that lack of consensus on standardized neuropsychologic testing procedures may, in part, be contributing to this confusion (3). Clearly, as Kubiak et al. point out, a major contributor to the variance in cognitive performance observed in this population is the high prevalence of other risk factors for cognitive decline, including cardiovascular disease, hypertension, and depression (4,5), making it challenging to isolate the potential contribution of diabetes per se.
We sought to explore cognitive function in adults with type 2 diabetes by perturbing the system through the administration of glucose, which is a treatment commonly used in studies of cognitive aging to explore the system’s plasticity in the face of underlying age-related deficits (6,7). A major advantage of applying this approach to the type 2 diabetic population is that changes observed in response to the challenge were unlikely to be directly attributable to vascular complications. Our data provided evidence for cognitive deficits, primarily related to declarative memory function, following the ingestion of 50 g of glucose in the form of rapidly absorbed carbohydrate foods (bagel and juice). We then argued that this impairment was consistent with observations in healthy senior adults, in whom moderate elevations in blood glucose resulted in memory enhancement and more extreme increases in blood glucose were associated with deficits (what is often referred to as an inverted-U dose-response relationship) (8). Based on this argument, we concluded that adults with type 2 diabetes likely responded to a glucose challenge in a manner comparable with that of older adults, with the caveat that they were more likely to attain levels of hyperglycemia associated with cognitive impairment given their underlying disease. Clearly, this is a conclusion requiring further verification.
One issue of concern raised by Kubiak et al. is that we include alternate versions, developed by us and others (9), of standardized neuropsychologic tests, although we apply these versions using standardized methodology. Our within-individual design, i.e., requiring multiple testing of subjects, necessitates their use. While we do not provide the precise details used to develop these alternate versions, we previously directed readers to those publications that we relied on to do so. Importantly, we have never stated or implied that these alternate versions should be used clinically from a diagnostic perspective; rather they are only used as experimental research tools. We agree that the use of these alternate versions potentially adds unwanted variance to our measures, but disagree that this detracts from the results obtained. Rather, the additional variance contributed by the use of alternate test versions makes it more, not less, difficult to observe change following glucose ingestion or associations between subject characteristics and performance levels. We address test version variability through the random allocation of different test versions both between and within subjects to uniformly distribute this additional variance (as much as possible) throughout the data, thereby minimizing potential bias associated with their use.
Another concern expressed was the limited number (n = 19) of subjects in our study, which is in essence an extension of those concerns related to test version variability. While not commented on in the original publication, this sample size was based on power analyses drawing on our results in healthy senior adults receiving glucose in the form of carbohydrate foods and including the same alternate versions of the neuropsychologic tests (9). Thus we believe our study to be statistically sound. Nevertheless, as with all studies, the extension of the results to the broader population is complicated by the fact that subjects willing to participate in experimental procedures are somewhat unique and, in this sense, differ from the more heterogeneous population typically observed in clinical practice—a factor of importance in all studies drawing on human volunteers.
Finally, Kubiak et al. comment on concerns related to interference when multiple tests probing declarative memory are used. This factor is not discussed in this work but is addressed by us previously (9) in studies conducted on healthy senior adults. This comment raises multiple issues of interest. The first is that the exact nature of the declarative memory deficits observed in adults with type 2 diabetes remains largely unexplored. Clearly, multiple components of cognitive function are recruited and contribute to performance on end measures of delayed verbal recall; yet the precise deficit, potentially including interference and inhibitory control, remains largely unexplored. Admittedly, we observed deficits in our study on the second, not the first, verbal recall test used, thus raising the possibility that interference is an important contributor and one requiring further exploration. Yet this does not detract from the fact that performance on this second test was poorer when subjects were tested following carbohydrate ingestion compared with when they were tested following placebo (water) ingestion. The second issue relating to the comment by Kubiak et al. is the degree to which one controls for external factors influencing cognitive performance. Clearly, our data suggest that the fed/fasted state of the individual may be an important contributor to observed variance. Similarly, time of day of testing is another recognized contributor to within-individual variance in cognitive function and shifts in peak performance times occur in states, such as aging, wherein disruptions to circadian sleep rhythms are apparent (10). This is clearly a pattern disruption to which adults with type 2 diabetes may be especially vulnerable. Yet rarely do authors address when during the day testing occurred and whether a fixed time of day was used, as we did in our studies. All of these factors are likely important and contribute to the “unexplained” variance in regression models. Some will view this as exciting opportunities for new exploration, whereas others will see this as reason to discount “explained” variance in function.
Certainly, it is essential that as we wade through conflicting data regarding the origins of cognitive deficits in adults with type 2 diabetes, be they primarily or secondarily associated with the high prevalence of other risk factors in this population, that we not lose sight of the shared and common interest, which is helping those with type 2 diabetes prevent or minimize their risk of cognitive dysfunction. Kubiak et al. state that we were daring to conclude that transitory food-induced hyperglycemic episodes could be associated with acute cognitive deficits in this population. We continue to stand by our conclusions but recognize that further studies will either support or refute our conclusions.
