Lina Palaiodimou, MD
Diabetes mellitus (DM) affects about 422 million people and is one of the leading causes of death worldwide (World Health Organization). More importantly, the burden of type 2 diabetes (T2D) has been rising relentlessly in all countries in the past three decades. However, it is estimated that a significant percentage of cases of T2D remain undiagnosed. DM is one of the major modifiable risk factors for stroke. In addition, it has been associated with adverse outcomes after stroke, including higher mortality, poorer neurological and functional outcomes, longer hospital stay, higher readmission rates, and stroke recurrence. Another outcome, the post-stroke cognitive function, and its relationship with DM, are being evaluated in the STROKOG collaboration study.
Lo et al. for the STROKOG collaboration present a meta-analysis of individual participant data (IPD) derived from seven international post-stroke cohorts with the aim to investigate the relationship between T2M and prediabetes with cognitive impairment after stroke.
In their study, diagnosis of DM and prediabetes (or impaired fasting glucose) was performed based on the fasting glucose level (FGL), which was measured post-stroke. Diagnosis of DM was also supported by a previous history of DM or the use of antidiabetic treatment. It has been demonstrated before that the temporal profile of post-stroke hyperglycemia is probably biphasic; there is one early phase of hyperglycemia (peaking at 8 hours, in all diabetic and half of non-diabetic patients), followed by a delayed phase, 48–88 hours post-stroke in 78% of diabetic and 27% of non-diabetic patients (Allport, 2006). In the study by Lo et al., blood glucose was measured on average 10 days post-stroke and, therefore, glucose levels and the subsequent DM diagnosis are considered not to be confounded by the stress-related hyperglycemia observed in the acute phase of stroke. However, the authors proceeded to additional sensitivity analyses either by evaluating HbA1c values (which are independent from post-stoke hyperglycemia) or by reducing FGL values by 20% (a percentage that corresponds to the assumed stress-related glucose elevation after stroke).
Another important methodological aspect of the study by Lo et al. is that cognitive function was assessed with various detailed neuropsychological test batteries, which were harmonized among the different cohorts and standardized adjusting for age, sex and educational level. Five separate cognitive domains (attention, memory, language, perceptual motor and executive function) and global cognition, as well, were evaluated. Moreover, neuropsychological testing was performed 3-6 months after stroke, a period during which cognitive performance is considered to have been stabilized after a cerebrovascular event.
The relationship of glucose status and cognitive function was investigated through consecutive models, which were used for adjusting for age, sex, education, type of stroke, stroke location, ethnicity, vascular risk factors and depression. In addition, the authors tested further possible interactions between glucose and age, gender and ethno-racial groups and their association with cognitive performance.
Since previous studies investigating high glucose levels and cognitive impairment had provided conflicting results, the results presented by Lo et al. were enlightening. Data were provided for 1,601 stroke patients worldwide. Twenty-eight percent of the patients were finally excluded due to missing glucose values. However, the authors evaluated the differences between the patients excluded and those finally included in the analysis, and they proved that despite several different baseline characteristics, global cognition did not differ between the two groups. Therefore, it is unlikely that the percentage of excluded patients, albeit significant, would have affected the results.
In their study, Lo et al. proved that stroke patients with DM had poorer cognitive function in all tested domains compared to the stroke patients with normal blood glucose. Characteristically, DM patients had more than half a standard deviation lower global cognition in comparison. Attention deficits were the more prevalent ones. However, prediabetes was not associated with cognitive impairment compared to normoglycemic stroke patients. It should also be noted that the patients with normal glucose values had poorer cognitive function compared to stroke-free controls of the cohort studies included, not surprisingly though, since those patients had suffered a cerebrovascular event.
The aforementioned association between glucose status and cognitive impairment seemed to be independent from stroke location, depression, ethno-racial groups and assumed post-stroke hyperglycemia. Gender and age were proved to have an interesting interaction to the glucose-cognition association in stroke patients with DM. Women with T2D were significantly more affected in the domains of memory, perpetual motor and executive function compared to men. In addition, memory of younger individuals with T2D was more negatively affected after stroke.
Certain limitations are well acknowledged by the authors. Data about pre-stroke cognitive performance were not available, so cognitive impairment may have pre-existed is some cases. In addition, data were derived from studies that were performed in different periods, as early as 1988 (EpiUSA). Another potential shortcoming could be the fact that the patient assignment in any of the three glucose status groups (T2M, prediabetes and normoglycemic patients) was based on a single measurement of fasting plasma glucose, testing for hyperglycemia. Neither hypoglycemia nor glycemic variability (which are the other components of dysglycemia) were evaluated. Interestingly, one previous study has proved that glycemic variability, as indicated by higher values of mean absolute glucose (MAG) in an acute stroke population, was correlated with post-stroke cognitive impairment after 3 months in the subgroup of DM patients (Lim, 2018). The evaluation of all aspects of dysglycemia and their correlation with cognitive function would be an intriguing subject for future studies.
Despite the limitations, the study by Lo et al. has a strong design and presents straight-forward results; T2M, but not prediabetes, is related to poorer cognitive performance of patients 3 to 6 months after stroke. Such results underscore the need for interventions to prevent the progression of prediabetes to diabetes in stroke patients and warn about possible inattention of diabetic stroke patients in diabetes self-care due to cognitive dysfunction.