Mild Cognitive Impairment in African Americans Is Associated with Differences in EEG Theta/Beta Ratio

Theta/beta ratio (TBR)

The ratio of amplitudes in the theta frequency band (around 4–7 Hz) relative to the beta band (around 13–30 Hz) has long been hypothesized to reflect cognitive functioning, particularly in attention-deficit hyperactivity disorder [33]. Originally, Lubar [33] proposed that it reflected trait arousal. Subsequent research has complicated this picture [34, 35], but in cognitively normal participants it has been associated with executive functioning, particularly attentional control [36–38], and cognitive processing capacity [39]. Both executive function and cognitive processing capacity are compromised in MCI and AD [40–42], so to the extent that TBR reflects these constructs, we would expect a higher TBR in MCI than neurologically intact controls.

To reiterate, in normal aging there is an increase in the amplitude of slow waves and decrease in fast waves, this trend is exaggerated in MCI, the TBR is presumably influenced by both phenomena, and TBR is associated with several cognitive functions compromised in MCI. Despite these indications that TBR would likely distinguish cognitively normal aging from MCI, there have been few investigations of TBR specifically in AD or MCI. Brayet et al. [43] investigated a closely related measure of slow (delta and theta) to fast (alpha and beta) waves and found that this slow/fast ratio discriminated between MCI and control participants, particularly when derived from a sleeping state. Schmidt et al. [44] found that the ratio of alpha to theta distinguished participants with probable AD from control participants. Miao et al. [45] similarly observed greater TBR for MCI and AD relative to controls which persisted across spectral decompositions computed in time windows of various widths. Dimitraidis et al. [46] studied a dietary intervention for MCI and found that TBR was reduced in people who showed improvements on neuropsychological evaluations.

Theta/beta ratio asymmetry (TBRA)

The majority of the studies of hemispheric asymmetries in EEG frequency bands have focused on the alpha frequency band and frontal region [47]. Higher-order cognitive processes are not only lateralized to different hemispheres (e.g., verbal processing to left hemisphere and visuospatial processing to the right hemisphere) but are also related to hemispheric asymmetries of the EEG spectral alpha and beta power when EEG is recorded during the performance of attentional, working, and long-term memory and learning tasks. Ray and Cole [48] found that parietal beta asymmetry was associated with cognitive demands, while temporal beta asymmetry was associated with the emotional valence of a task. Papousek et al. [49] showed that changes of working memory performance were related to changes of cortical asymmetry in resting conditions before the execution of tasks. Such findings are consistent with the idea that spontaneously occurring alterations of cortical asymmetry obtained in resting conditions can predict changes in lateralized cognitive performance.

It has been empirically validated that the beta frequency band may be more important in studies on the lateralization of cognitive processes than alpha asymmetry [48–51]. It also has been suggested that EEG beta power directly reflects activity of cortical regions effectively involved in higher-order cognitive processes [48, 52].

While TBR has been under-researched in MCI, TBRA has not been researched in any prior study of cognitive performance, using either a convenience sample of college students or a clinical population, to our knowledge. Wang et al. [53] investigated TBR laterally and found that left frontal TBR (measured at F3) but not right frontal TBR (F4) was predicted by negative emotion differentiation. Therefore, there are good reasons to wonder if resting state TBRA might serve as a meaningful index of brain health.

From a neuropathological perspective, neurodegeneration is not typically symmetric, although which hemisphere degenerates more rapidly is idiographic [54]. However, the side of greater degeneration is predictive of deficits [55, 54]. Assumptions that AD neuropathology is uniform across the brain and across the individuals was shown to be incorrect by a recent study [56]. They showed that during asymptomatic, preclinical AD, older adults, early tau deposition follows multiple trajectories and may involve different cortical regions. These authors reported about three main subtypes: asymmetrical left, precuneus predominant, and asymmetrical right. Results showed that preclinical AD participants with these three cortical tau PET patterns were younger and showed worse executive functioning than participants with preclinical AD and typical medial temporal lobe PET pattern.

Previous work has highlighted the potential importance of functional comparisons between MCI and normal groups through comparison of pre- to post-task aftereffects [6] and frontal asymmetry [7]. Here, we extend this analysis to an index of cognitive and emotional functioning that has been relatively under-investigated in MCI, the TBR and hemispheric asymmetries in TBR. The following hypotheses were tested:

Hypothesis 1: TBR is higher in MCI than in cognitively normal controls.

Hypothesis 2: TBR is more asymmetric in MCI than in cognitively normal controls.

Participants

The institutional review board of Wayne State University approved all procedures. Participants were recruited from the volunteer pool of Healthier Black Elders Center (HBEC), which is a collaboration between the Institute of Social Research at Wayne State Institute of Gerontology and at the University of Michigan and from the Michigan Alzheimer’s Disease Research Center (MADRC). Members of HBEC who self-reported a recent change in memory on a screening questionnaire were recruited into the study. A total of 99 participants met all inclusion criteria in the final sample. Forty-one (35 female, 6 male) were diagnosed with MCI (see below) and the remaining 58 (53 female, 5 male) serving as cognitively normal controls. The mean age for participants with MCI was 73.7 (s = 7.2) years, while for controls it was 71.1 (s = 6.2) years. This difference was not statistically significant, t(97) = 1.95, p = 0.103.

Diagnosis of MCI was by consensus conferences that included neurologists, neuropsychologists, nurses, and technicians, who reviewed neuropsychological assessments that included the National Alzheimer’s Disease Cooperating Centers (NACC) standardized evaluation battery and the Unified Dataset (UDS) [57]. Diagnoses excluded seizure disorders, psychiatric or sleep disorders, AIDS, learning disability, head injury with loss of consciousness exceeding 30 minutes, previous diagnosis of dementia, radiation therapy of the brain in the past year, history of substance use disorder, or major medical illness in the previous five years. Of those diagnosed with MCI, 13 were diagnosed as non-amnesic and 28 as amnesic, but all analyses reported here collapse across that distinction because these group sizes are unlikely to provide adequate statistical power to make distinctions between them. Everyone scored at or above 25 on the MMSE.

Neuropsychological evaluation

Participants were tested with the National Alzheimer’s Disease Coopering Centers (NACC) Unified Dataset (UDS) test battery [57]. A consensus conference that included neurologists, neuropsychologists, nurses, and technicians then reviewed UDS results. Diagnoses were made after applying exclusion criteria, including no other psychiatric or sleep disorders, head injury with loss of consciousness greater than 30 min, seizure disorder, learning disability, AIDS, radiation therapy of the brain in the past year, diagnosis of dementia, history of substance use disorder, or major medical illness in the previous five years.

EEG recording

EEG data were recorded with a Brain Vision Inc. 64-channel amplifier and Acti Cap (active electrode) system with modified International 10–20 System placement of electrodes. Online data were sampled at 500 Hz (32-bit resolution) with a 0.1–70 Hz bandpass filtering, 12 dB roll-off per octave on each side, referenced to location FCz and grounded at AFz. Impedances were kept below 10 kΩ and balanced over all channels within 5 kΩ.

Participants were seated in a chair adjusted for height, facing a computer in a dimly lit room. Resting state EEG was measured for a minimum of 3 min with eyes open and 3 min with eyes closed, with time extending up to half a minute beyond 3 min only if noisy epochs were observed, to ensure the availability of sufficient rsEEG data to obtain a minimum of 60 artifact-free 2-s epochs. This was done twice, once before and once after a motion direction discrimination task.

The motion direction discrimination task is the same one previously reported [6]. Participants indicated the direction of dots moving left or right with a button press of left or right index fingers.

Data analysis

Data analyses were conducted with SPSS 28 (IBM, Inc.). Plots were generated in JASP [64]. The multivariate approach to repeated measures was used in all ANOVAs [65]. This approach does not assume sphericity, instead taking advantage of correlations among repeated measures to increase the statistical power to detect differences between repeated conditions [66]. Analyses for both TBR and TBR asymmetry were 2 (group)×2 (timepoint)×4 (location) mixed ANOVAs. To follow up an unpredicted interaction between timepoint and location, we performed two post-hoc contrasts of the effect size of timepoint at different locations. Specifically, we compared the effect of timepoint on Oz to the average effect at Fz and Cz, and similarly the effect of timepoint on Pz to the effect at Fz and Cz. We chose this approach rather than the more common approach of testing the effect of timepoint separately at each electrode, because a pattern of significant and non-significant effects may not itself be significant [67] and therefore may not explain an interaction, but the contrasts directly test whether the effect size of one variable differs as a function of the other variable [68]. We followed this up with a parallel evaluation of TBR asymmetry.

TBR along the central axis

Figure 1 illustrates the topography of TBR across the scalp, separated by MCI status and timepoint. Table 1 gives cell means and confidence intervals for TBR. Figure 2A illustrates the mean TBR along the midline as a function of MCI status. There was a main effect of MCI status, F(1,95) = 4.44, p = 0.038, η p 2 = 0.045. From Fig. 2A, it is clear that MCI participants had a higher TBR than controls. Figure 2B shows TBR as a function of location and MCI status. There was no main effect of timepoint, F(1,95) = 0.924, p = 0.339, η p 2 = 0.01. There was a main effect of location, F(3,93) = 8.85, p < 0.001, η p 2 = 0.22, as well as an interaction between location and timepoint, F(3,93) = 3.34, p = 0.023, η p 2 = 0.097. From Fig. 2B, it is clear that TBR was relatively unaffected at frontal locations but increased in the post-task period in parietal and occipital regions, with central sites being more ambiguous. Post hoc contrasts confirmed this pattern: the effect of timepoint was significantly greater at Oz than at the average of the other locations, F(1,96) = 10.08, p = 0.002, while for Pz with respect to the mean of Fz and Cz, F(1,96) = 3.82, p = 0.057. No other interactions were significant.

OPEN IN VIEWER

Fig. 1

Topography of TBR as a function of MCI status and timepoint. The topographies show a higher TBR overall in MCI relative to controls, as well as an increased TBR over posterior regions post-task in both groups. Although this task after-effect appears most obvious for the MCI pre-post topographies, the interaction between MCI status and timepoint was not significant, indicating that any difference in the size of this effect as a function of MCI status is not reliable.

OPEN IN VIEWER

Fig. 2

Illustration of main effects and interactions. A) The main effect of MCI status on TBR. B) The main effect of location on TBR. Significant differences between timepoints are marked with an asterisk. C) The main effect of location on TBRA. Significant differences between groups are marked with an asterisk. Bars represent standard errors. CTR, neurologically intact controls; MCI, mild cognitive impairment.

We collapsed across categories of MCI (amnesic, N = 28, and non-amnesic, N = 13) because these sample sizes were not adequate to distinguish between them. However, concern that this distinction might be driving the hemispheric asymmetries led us to re-run this analysis excluding non-amnesic MCI (naMCI) participants. The pattern of results was the same, with the same significant main effects and interaction, indicating that this asymmetry was not driven predominantly by naMCI participants.

TBR asymmetry

Figure 2C presents means for the analysis of TBR asymmetry. Cell means for the ANOVA analyzing TBR asymmetry are presented in Table 2. There was a main effect of location, F(3,93) = 8.744, p < 0.001, η p 2 = 0.22. There was no main effect of timepoint, F(F(1,95) = 0.234, p = 0.63, η p 2 = 0.002. There was no main effect of MCI status on asymmetry, F(1,95) = 0.698, p = 0.406, η p 2 = 0.007, but MCI status did interact with location, F(3,93) = 5.082, p = 0.003, η p 2 = 0.141. From Fig. 2C, it can be seen that TBR is higher on the right in frontal locations but higher in the left (negative) toward posterior locations, and this is similar for both MCI and controls except for the frontal location F3-F4, where MCI participants are much more rightward-leaning than controls, who straddle 0 or no net asymmetry. Post-hoc contrasts (contrasting the group effect at frontal vs. all other locations) confirmed this pattern, F(1,95) = 12.288, p < 0.001.

Limitations

This was a secondary analysis of correlational data, so it has all the usual limitations of correlational data. Additionally, because this was a secondary analysis, the new phenomena we report here should be replicated in an independent sample before they are given high confidence. Furthermore, some authors have questioned the utility of TBR or any ratio between frequency bands on the grounds that it is a less-precise measure of aperiodic activity [35]. Nonetheless, TBR has been well-established as a reliable diagnostic marker in conditions like attention-deficit hyperactivity disorder and even excessive electronic screen time in infants [71], which could improve TBR’s appeal for use in a clinical setting.

Finally, our sample consisted of community-dwelling African American older adults, a group under-represented in AD/MCI research. While we have no reason to suspect that findings from this study would differ in other racial or ethnic groups, we cannot be certain about the generalizability of these findings until our methods are replicated in other populations.

Conclusions

Despite the limitations of this study, the results suggest that TBR indexes important parameters of brain functioning in healthy aging that differ in mild cognitive impairment. With further validation in large and diverse samples, TBR offers a compelling opportunity for clinical evaluation following previously established methodologies in attention-deficit hyperactivity diagnostic applications. It would be beneficial to our understanding if future research was able to tie structural and metabolic brain data to differences in TBR between MCI and healthy controls.