Independent Contributions of Cortical Gray Matter, Aging, Sex and Alcoholism to K-Complex Amplitude Evoked During Sleep

Abstract

Study Objectives:

The amplitude of the N550 component derived from the averaged evoked K-complex decreases with normal aging and with alcoholism. The study was designed to determine whether these declines are related to the extent of cortical or subcortical shrinkage.

Setting:

Research sleep laboratory and MR imaging facility

Participants:

26 abstinent long-term alcoholic men, 14 abstinent long-term alcoholic women, 18 control men, and 22 control women.

Measurements and Results:

MRI data collected at 3T were analyzed from alcoholic and control men and women previously reported to have significantly different evoked delta activity during sleep. Segmented and parcellated MRI data collected at 3T were compared between these groups and evaluated for correlation with evoked K-complex amplitude measured at FP1, Fz, FCz, Cz, CPz, and Pz. Cortical gray matter and regional subcortical tissue volumes entered as predictors into stepwise multiple regression identified cortical gray matter as a unique significant predictor of evoked K-complex at all sites. Age added independent variance at 5 of the 6 sites, while alcoholism and sex added independent variance at frontal sites only.

Conclusions:

These data support recent intracranial studies showing cortical generation of K-complexes by indicating that cortical, but not subcortical volume contributes to K-complex amplitude. Establishing the extent of the relation between cortical volume and K-complex amplitude provides a mechanistic understanding of sleep compromise clinically relevant to normal aging, alcoholism, and likely other conditions affecting cortical volume and integrity.

Citation:

Colrain IM; Sullivan EV; Rohlfing T; Baker FC; Nicholas CL; Padilla ML; Chanraud S; Pitel AL; Pfefferbaum A. Independent contributions of cortical gray matter, aging, sex and alcoholism to K-complex amplitude evoked during sleep. SLEEP 2011;34(6):787-795.

INTRODUCTION

K-complexes are large delta frequency EEG events ubiquitous to NREM sleep. These delta wave events have the unusual characteristic of occurring spontaneously, yet they can also be evoked deliberately by stimuli delivered by different modalities.^1,2 Animal and human experimental evidence indicates that K-complexes are single instances of the delta EEG activity that predominates in slow wave sleep (SWS).^3–6 Thus the evoked K-complex provides an experimentally controlled probe of delta EEG production ability. To the extent that it reflects cortical integrity, the robustness of the evoked K-complex is a marker of the potential of the cerebral cortex to produce a synchronized response.² Because the production of K-complex is thought to reflect a sleep protective process^2,5,7 and to relate to slow oscillation in cortical currents⁸ and slow wave delta EEG activity,⁶ recently shown to have a role in supporting learning and memory,⁹ the K-complex can be used as a marker of functionally important sleep-specific brain processes.

In adults, the scalp distribution of the prominent negative (N550) component of the averaged evoked K-complex shows a bilaterally symmetrical distribution, with the amplitude typically maximal at Fz and then diminishing systematically at more posterior locations.² Similar scalp distributions have been reported for spontaneous K-complexes and delta activity marking slow wave sleep.¹⁰ The frontal distribution is supported by positron emission tomography¹¹ and high density EEG array¹² studies, implying a role for frontal cortex in the generation of delta activity during sleep.

Despite lack of definitive determination of an intracranial generator of the K-complex, converging evidence suggests that it is a cortical phenomenon. Animal studies¹³ and intracranial recordings in human epileptic patients¹⁴ show polarity inversion of voltage associated with the K-complex between superficial (negative voltage) and deeper cortical layers (positive voltage) that would not be seen if the K-complex were the result of a subcortical generator such as the cingulate gyrus, thalamus, hippocampus, or basal ganglia, as previously hypothesized.^7,15,16 The large amplitude of the K-complex (and the N550) is typically an order of magnitude greater than even relatively large event-related potential components such as the P300. This implies a need for large numbers of healthy neurons to be involved in its generation and suggests a relationship between cortical gray matter volume and N550 amplitude.

Factors affecting brain integrity also affect components of K-complexes. The amplitude of the prominent negative component of the averaged evoked K-complex (N550) declines with age in healthy men and women.^17–19 Alcohol dependence also has a clear negative effect on N550 amplitude²⁰ that is more pronounced at frontal sites and represents an exacerbation of normal aging effects.²¹ We have previously proposed that the effects of normal aging on the evoked K-complex may relate to the MRI-identified age-related decreased gray matter volume salient in prefrontal cortex.^19,22,23 We extended this position to explain the effects of alcoholism on evoked K-complex amplitude, given the presumed multi-site influence of alcoholism on brain tissue notable in superior prefrontal cortex both postmortem and in vivo.^{20,21,24–34} The alcoholism-related cortical shrinkage occurs over and above that occurring in normal aging.^32,35

Here, we tested the hypothesis that N550 amplitude in the averaged evoked K-complex is related to gray matter volume in the cortex in general and the frontal cortex in particular, using data from alcoholic and control subjects we previously demonstrated to have age-¹⁹ and alcoholism-related decline in N550 amplitude.²¹

METHODS

Subjects

EEG and MRI data were obtained from abstinent, long-term alcoholic men (n = 26) and women (n = 14) and control men (n = 18) and women (n = 22) (Table 1). These 80 were a subset of the 84 subjects whose sleep evoked potential data were previously described²¹ and who had MRI data (not previously reported), collected within a week of the sleep studies. All procedures were approved by the Institutional Review Board of SRI International, and all subjects provided written informed consent for study participation.

Table 1.

Participant characteristics presented as means (standard deviations) for alcoholic and control men and women. Data are presented for age, age of onset and duration of drinking in years, estimated lifetime alcohol consumption (EAC) in kg, and the period since last drink prior to scanning (Sobriety) in days.

	Age	EAC*	Sobriety	Age of Onset	Duration of Drinking
Women
Control	51.6 (10.1)	26.9 (37.5)
Alcoholic	49.0 (9.6)	823.8 (327.4)	260.2 (209.6)	26.3 (7.5)	22.7 (9.4)
Men
Control	51.1 (10.3)	89.9 (115.0)
Alcoholic	50.2 (7.5)	1502.4 (844.8)	135.9 (147.0)	22.7 (5.6)	27.5 (8.8)

^*Alcoholic women vs. control women, P < 0.001; alcoholic men vs. control men, P < 0.001; alcoholic men vs. alcoholic women, P < 0.01.

Alcoholics had significantly higher estimated lifetime alcohol consumption than sex-matched controls for both men (Z = 5.59, P < 0.001) and women (Z = 5.00, P < 0.001) (based on the Mann-Whitney test). Alcoholic men had significantly higher estimated lifetime alcohol consumption than alcoholic women (Z = 2.89, P = 0.003) and showed a trend to have a shorter period of sobriety before testing than alcoholic women (Z = 1.79, P = 0.076). Neither alcoholism age of onset nor years spent drinking differed between sexes. A sex-by-diagnosis (alcoholic, control) ANOVA showed the alcoholic and control men and women to be of similar age.

Magnetic Resonance Imaging and Analysis

MRI data were acquired on a 3T General Electric (Milwaukee, WI) Signa human MRI scanner (gradient strength = 40 mT/m; slew rate = 150 T/m/s; software version VH3). Structural data were acquired with an Inversion Recovery Prepared SPoiled Gradient Recalled echo sequence (IRPrepSPGR; FOV = 24 cm, TI = 300 ms, TR/TE = 6.5/1.54 ms, thick = 1.25mm, slices = 124).

Gray matter and tissue volumes were derived for 45 bilateral supratentorial regions defined in the SRI24 atlas^36,37 (http://nitrc.org/projects/sri24) and propagated to each subject via nonrigid registration. The 45 regions were measured in each hemisphere with homologous hemispheric regions summed to produce 45 bilateral values. The regions defined were as outlined in.³⁸ For data reduction and to address our hypothesis regarding large regional volumes of cortex and generators of the K-complex, we derived cortical lobar values by summing appropriate areas, yielding cortical volumes of frontal, sensorimotor, temporal, parietal and occipital gray matter (Figure 1A). Thus, a variable was derived for frontal gray matter volume by summing the values from the 13 individual bilateral regions measured in the frontal lobes. Other composite values were similarly derived, with a sensorimotor cortex variable consisting of the sum of gray matter volume from the precentral gyrus, postcentral gyrus, and Rolandic operculum. For subcortical structures, tissue volumes were calculated by adding gray and white matter bilateral volumes for the thalamus, hippocampus, basal ganglia (caudate, pallidus, and putamen combined), and cingulate gyrus (anterior, middle, and posterior combined).

Figure 1.

(A) Surface-rendered brain from the SRI24 atlas³⁷ (http://nitrc.org/projects/sri24), showing the cortical regions for gray matter volume evaluation. The frontal lobe is shown in red, the sensorimotor cortex in green, the parietal lobe in yellow, the occipital lobe in light blue, and the temporal lobe (minus the temporal pole) in dark blue. The bottom right of the panel shows a 2-dimensional representation of a coronal slice showing an outline of the parcellated cortical surface. (B) Cortical gray matter volumes expressed as a proportion of ICV for control and alcoholic men and women. Error bars reflect standard deviations around each mean value. (C) Subcortical tissue volumes expressed as a proportion of ICV for control and alcoholic men and women. Error bars reflect standard deviations around each mean value. *indicates a significant effect of alcoholism diagnosis (P < 0.01).# indicates a significant effect of sex (P < 0.01).

The measurement approach has been validated against manually delineated brain structures, showing for example, a 0.96 correlation between the 2 approaches for the corpus callosum.³⁹ Further validation of the automated technique for thalamus, basal ganglia, cingulate and hippocampus was conducted using 127 control cases, 18-85 years, from our laboratory database. Parcellation convergent validity was assessed by computing the correlation of each ROI with intracranial volume (ICV), creating ICV-adjusted regional volume with linear regression of each ROI volume on ICV, and correlating ICV-adjusted volume with age. The predictions were that (1) all ROIs would show a positive correlation with ICV; (2) ICV-adjusted volumes of the thalamus, basal ganglia and cingulate cortex would be negatively correlated with age; and (3) ICV-adjusted volume of the hippocampus would not be correlated with age. ICV was indeed positively correlated with measurements of all structures (thalamus r = 0.374; basal ganglia, r = 0.531; cingulum, r = 0.749; hippocampus, r = 0.700, P < 0.0001 in all cases). Age was negatively correlated with volumes of the thalamus (r = −0.726), basal ganglia, (r = −0.383) and cingulate (r = −0.417) (P < 0.0001 in all cases), but not with the hippocampus (r = 0.014, P = 0.879).

Sleep Evoked Potentials

Detailed methods for evoked K-complex acquisition and analysis have been published.²¹ Briefly, auditory stimuli (1000 Hz pure tones presented for 50 ms (2 ms rise time) at 80 dB(A)) were presented binaurally throughout a single night of recording. K-complexes evoked by the stimuli during stage 2 sleep were counted and averaged at 7 scalp sites (FP1, FP2, Fz, FCz, Cz, CPz, Pz). All subjects had a prior adaptation night used to screen for sleep disorders and an additional night of recording of sleep EEG without stimulus presentation.⁴⁰

Statistical Analysis

All statistical analyses were conducted using SPSS-18. The effects of ICV, sex, age, and alcoholism on cortical gray matter volume were assessed using MANCOVA or ANCOVA, using ICV and age as covariates and sex and diagnosis as between-group factors. When an overall MANCOVA model displayed significant age or diagnosis effects, univariate ANCOVA values for each gray matter volume were then evaluated for diagnosis, age, or both on each cortical region entered into the model. The same approach was taken to evaluate tissue volumes from the anterior, middle, and posterior cingulate gyrus; and to evaluate tissue volumes from subcortical structures (thalamus, hippocampus, and basal ganglia).

Stepwise multiple regression analyses were conducted to predict N550 amplitude at electrode sites FP1, Fz, FCz, Cz, CPz, and Pz (FP2 data were generally equivalent amplitudes to FP1²¹ and were thus not evaluated separately), using a series of models in which variables were forced into the models (SPSS “ENTER” command). Factors were considered to add independent variance if the significance of the r² change was at P < 0.05, following their addition to the model. Model 1 contained gray matter volumes for frontal, sensorimotor, temporal, parietal, and occipital cortices. Model 2 added the tissue volumes from the anterior, middle, and posterior portions of the cingulate gyrus. Model 3 added the tissue volumes from the thalamus, hippocampus, and basal ganglia. Model 4 added alcoholism diagnosis. Model 5 added age, and Model 6 added sex as predictor variables.

RESULTS

N550 Amplitude

As previously reported,²¹ the probability of eliciting a K-complex was significantly less in alcoholic men (0.35) and women (0.46) than in control men (0.54) and women (0.62). N550 amplitude from these subjects show significant effects of alcoholism diagnosis and a significant diagnosis-by-electrode site interaction.²¹ Specifically, alcoholics had smaller N550 amplitudes than controls at frontal but not more posterior sites. Although men had significantly smaller N550 amplitudes overall than women, N550 amplitudes were significantly smaller in older than younger subjects regardless of diagnosis or sex. N550 data at Fz were also analyzed to determine the affect of family history of alcoholism (data available from 79 subjects). A positive family history was present in the majority of alcoholics (25/40) and a minority of controls (8/31), however, family history did not show any effect on N550 amplitude (F_1,70 = 0.049, P = 0.825).

Regional Brain Volumes

Cortical gray matter volume

MANCOVA for cortical gray matter volume (GMV) revealed significant effects of ICV (F_5,70 = 11.12, P < 0.001), age (F_5,70 = 10.46, P < 0.001), diagnosis (F_5,70 = 3.71, P = 0.005), and sex (F_5,70 = 3.90, P = 0.004), with no diagnosis-by-sex interaction (P = 0.339). In the total sample of 80 subjects, the effect of ICV was significant for all regions, as was the effect of age, with older age being associated with smaller volumes of the frontal, sensorimotor, parietal, occipital, and temporal regions (all P ≤ 0.001). After correction for age, sex, and ICV, a diagnosis of alcoholism was associated with smaller volumes in the sensorimotor (P < 0.001) and temporal (P = 0.003) regions. After correction for age, diagnosis, and ICV, women had larger volumes in the temporal region than men (P = 0.004) (see Figure 1B).

Cingulate gyrus tissue volume

MANCOVA for cingulate tissue volumes revealed a significant effect of ICV (F_3,72 = 24.23, P < 0.001), with no effect of age, alcoholism diagnosis, or sex. The ICV effect was significant for all 3 cingulate regions (P < 0.001).

Subcortical tissue volume

MANCOVA for subcortical tissue volume revealed significant effects of ICV (F_3,72 = 11.48, P < 0.001), age (F_3,72 = 10.47, P < 0.001), and diagnosis (F_3,72 = 4.80, P = 0.004) (see Figure 1C), with neither a sex effect (F_3,72 = 2.08, P = 0.111) nor sex-by-diagnosis interaction effects (F_3,72 = 1.44 P = 0.237). ANCOVA results indicated that older age (F_1,74 = 31.37, P < 0.001) and a diagnosis of alcoholism (F_1,74 = 14.61, P < 0.001) were associated with smaller thalamic volume.

Regression modeling of N550 using brain, age, sex and alcoholism as predictors

The output of regression models for all 6 electrode sites are presented in Table 2, with the results of the final Model (Model 6) presented in Figure 2. Cortical gray matter volume (Model 1) was a significant predictor of N550 amplitude at all electrode sites (FP1, Fz, FCz, and Cz, P < 0.05; CPz and Pz, P < 0.01). Adding tissue volumes from the anterior, middle, and posterior cingulate to the Model (Model 2) did not significantly increase the proportion of explained variance for N550 at any electrode site. Likewise, there was no significant change in any of the electrode site models following the addition of tissue volumes from the hippocampus, thalamus, and basal ganglia (Model 3). Diagnosis of alcoholism (Model 4) added significant independent variance for models of N550 at FP1 (P = 0.001) and Fz (P = 0.010). This diagnosis effect is evident in the distribution of N550 amplitudes at FP1 and Fz (Figure 2). Age added significant independent variance to the models (Model 5) for all sites other than FP1 (Fz, FCz, Cz, CPz P < 0.01; Pz, P < 0.05). Finally, sex added significant independent variance to the final Model (Model 6) at FP1, Fz, and FCz (P < 0.05).

Table 2.

Output of stepwise regression modeling of N550 amplitude at each electrode site

FP1
Model	Variables	R2	Model Statistics	R2 Change	Change Statistics
1	Frontal, Temporal, Sensorimotor, Parietal, and Occipital GMV	0.157	F5,73 = 2.72, P = 0.026
2	1 + Anterior, Middle, and Posterior Cingulate Tissue Volume	0.227	F8,70 = 2.57, P = 0.016	0.070	F3,70 = 2.12, P = 0.105
3	2 + Thalamus, Hippocampus, and Basal Ganglia Tissue Volume	0.268	F11,67 = 2.23, P = 0.022	0.041	F3,67 = 1.26, P = 0.295
4	3 + Alcoholism Diagnosis	0.373	F12,66 = 3.28, P = 0.011	0.105	F1,66 = 11.06, P = 0.001
5	4 + Age	0.392	F13,65 = 3.23, P = 0.001	0.019	F1,65 = 2.08, P = 0.161
6	5 + Sex	0.437	F14,64 = 3.55, P < 0.001	0.045	F1,64 = 5.11, P = 0.027
Fz
Model	Variables	R2	Model Statistics	R2 Change	Change Statistics
1	Frontal, Temporal, Sensorimotor, Parietal, and Occipital GMV	0.171	F5,74 = 3.06, P = 0.014
2	1 + Anterior, Middle, and Posterior Cingulate Tissue Volume	0.239	F8,71 = 2.79, P = 0.010	0.068	F3,71 = 2.12, P = 0.106
3	2 + Thalamus, Hippocampus, and Basal Ganglia Tissue Volume	0.286	F11,68 = 2.48, P = 0.011	0.047	F3,68 = 1.50, P = 0.224
4	3 + Alcoholism Diagnosis	0.354	F12,67 = 3.05, P = 0.002	0.067	F1,67 = 6.96, P = 0.010
5	4 + Age	0.424	F13,66 = 3.74, P < 0.001	0.071	F1,66 = 8.09, P = 0.006
6	5 + Sex	0.479	F14,65 = 4.27, P < 0.001	0.055	F1,65 = 6.85, P = 0.011
FCz
Model	Variables	R2	Model Statistics	R2 Change	Change Statistics
1	Frontal, Temporal, Sensorimotor, Parietal, and Occipital GMV	0.181	F5,72 = 3.19, P = 0.012
2	1 + Anterior, Middle, and Posterior Cingulate Tissue Volume	0.232	F8,69 = 2.60, P = 0.015	0.051	F3,69 = 1.52, P = 0.218
3	2 + Thalamus, Hippocampus, and Basal Ganglia Tissue Volume	0.291	F11,66 = 2.46, P = 0.012	0.059	F3,66 = 1.82, P = 0.151
4	3 + Alcoholism Diagnosis	0.314	F12,65 = 2.48, P = 0.010	0.023	F1,65 = 2.19, P = 0.144
5	4 + Age	0.392	F13,64 = 3.18, P = 0.001	0.079	F1,64 = 8.28, P = 0.05
6	5 + Sex	0.448	F14,63 = 3.65, P < 0.001	0.055	F1,63 = 6.28, P = 0.015
Cz
Model	Variables	R2	Model Statistics	R2 Change	Change Statistics
1	Frontal, Temporal, Sensorimotor, Parietal, and Occipital GMV	0.177	F5,73 = 3.18, P = 0.012
2	1 + Anterior, Middle, and Posterior Cingulate Tissue Volume	0.221	F8,70 = 2.51, P = 0.018	0.044	F3,71 = 1.34, P = 0.270
3	2 + Thalamus, Hippocampus, and Basal Ganglia Tissue Volume	0.269	F11,67 = 2.27, P = 0.020	0.048	F3,68 = 1.49, P = 0.226
4	3 + Alcoholism Diagnosis	0.270	F12,66 = 2.07, P = 0.031	0.002	F1,67 = 0.15, P = 0.705
5	4 + Age	0.384	F13,65 = 3.16, P = 0.001	0.113	F1,66 = 12.14, P = 0.001
6	5 + Sex	0.402	F14,64 = 3.12, P = 0.001	0.018	F1,65 = 1.98, P = 0.164
CPz
Model	Variables	R2	Model Statistics	R2 Change	Change Statistics
1	Frontal, Temporal, Sensorimotor, Parietal and Occipital GMV	0.186	F5,73 = 3.34, P = 0.009
2	1 + Anterior, Middle and Posterior Cingulate Tissue Volume	0.219	F8,70 = 2.45, P = 0.021	0.033	F3,70 = 0.98, P = 0.406
3	2 + Thalamus, Hippocampus and Basal Ganglia Tissue Volume	0.268	F11,67 = 2.23, P = 0.023	0.049	F3,67 = 1.48, P = 0.227
4	3 + Alcoholism Diagnosis	0.268	F12,66 = 2.02, P = 0.036	0.000	F1,66 = 0.05, P = 0.833
5	4 + Age	0.347	F13,65 = 2.66, P = 0.005	0.079	F1,656 = 7.86, P = 0.007
6	5 + Sex	0.355	F14,64 = 2.52, P = 0.006	0.008	F1,645 = 0.80, P = 0.374
Pz
Model	Variables	R2	Model Statistics	R2 Change	Change Statistics
1	Frontal, Temporal, Sensorimotor, Parietal, and Occipital GMV	0.191	F5,74 = 3.49, P = 0.007
2	1 + Anterior, Middle, and Posterior Cingulate Tissue Volume	0.218	F 8,71 = 2.48, P = 0.020	0.027	F3,71 = 0.83, P = 0.483
3	2 + Thalamus, Hippocampus, and Basal Ganglia Tissue Volume	0.278	F11,68 = 2.38, P = 0.015	0.060	F3,68 = 1.87, P = 0.143
4	3 + Alcoholism Diagnosis	0.279	F12,67 = 2.16, P = 0.024	0.001	F1,67 = 0.07, P = 0.788
5	4 + Age	0.335	F13,66 = 2.56, P = 0.006	0.057	F1,66 = 5.62, P = 0.021
6	5 + Sex	0.339	F14,65 = 2.38, P = 0.010	0.004	F1,65 = 0.38, P = 0.541

Figure 2.

The relationship between N550 amplitude measured at FP1, Fz, FCz, Cz, CPz, and Pz and the standardized predicted values from Model 6 (frontal, sensorimotor, temporal, parietal, and occipital gray matter volumes; anterior, middle, and posterior cingulate tissue volumes; thalamus, hippocampus, and basal ganglia tissue volumes; age, sex; and alcoholism diagnosis). Men are represented by triangles and women by circles. Alcoholics are represented as filled symbols and controls as open symbols. In each panel, the line shows the linear regression relationship between the standardized model predictor and N550 for all subjects combined. Note that N550 is a negative component and thus more negative amplitudes on the X axis and more negative Z scores on the Y axis each indicate larger responses.

Additional analyses were conducted to directly test the role of cortical GMV as a mediator between observed effects of age and alcoholism on N550 amplitude. These analyses provide an additional direct test of the hypothesis that age and alcoholism effects on N550 are due to age and alcoholism effects on cortical gray matter volume. Three regression relationships were considered using N550 amplitude at Fz, the site where N550 is typically maximal and where both age and alcoholism effects are present. The first assessed the relationships between each of age or alcoholism and cortical GMV (relationship between the “mediator” and independent variable). The second tested the relationships between each of age or alcoholism and N550 amplitude (relationship between the dependent and independent variables). The third tested the relationships between N550 and the combination of GMV and age or GMV and alcoholism diagnosis. GMV could be considered to be the mediating variable in the age/N550 or alcoholism/N550 relationships if the significance in the first 2 relationships was lost when GMV was added to the independent variable in the model, or even if the regression coefficient was reduced.

For each case of the independent variable being either age (r = 0.489, P < 0.001) or alcoholism (r = 0.521, P < 0.001) regression relationships in which GMV was added to the independent variable were highly significant and greater than the observed relationships between either age (r = 0.411, P < 0.001) or alcoholism (r = 0.400, P < 0.001) and N550 amplitude alone. Thus the hypothesis of GMV being a mediator variable cannot be supported.

DISCUSSION

We previously hypothesized that age- and alcoholism-related declines in evoked and spontaneous delta activity during sleep are due to decreases in frontal cortical gray matter volume that occur with age and alcoholism.^{17,19–21,30,40} The present data, combining sleep-specific delta measurement and structural MRI from the same subjects, provide the first direct test of this hypothesis, a test that largely failed to support the hypothesis. While the data indicate that 15% to 19% of the variance of evoked delta amplitude can be explained by cortical gray matter volume, the relations are not specific to frontal cortex. Importantly, age, alcoholism, and sex each contributed to delta amplitude variance over and above their respective effects on cortical gray matter, and gray matter was not shown to be a significant mediating variable in either the age/N550 or alcoholism/N550 relationships.

Delta activity has long been viewed as a marker of brain changes in aging,⁴¹ alcoholism,⁴² and other psychiatric disease,⁴³ but may be limited by measurement parameters. Nonetheless, Feinberg and colleagues⁴⁴ demonstrated that developmental changes in delta EEG amplitude, cortical metabolic rate, and synaptic density taken from different data sets all varied in a manner best described by the same statistical model, a Gamma distribution. His work and that of others⁴⁵ supports delta amplitude as being related to the number of cortical neurons and their interconnections.

Our previous studies have shown strong relations between N550 amplitude measured at a broad range of scalp sites and aging. In a study of 70 normal healthy subjects aged 19 to 78 years, we reported that in both men and women, increasing age was associated with decreasing N550 amplitude with linear regression, explaining between 49% and 59% of the N550 variance depending on electrode site. Data from Fz, for example, showed a reduction of approximately 14.5 μV per decade.¹⁹ In another large study, we were able to show significantly smaller N550 amplitude in middle-aged alcoholic men and women when compared to controls.²¹ In this case, differences were limited to frontal electrode sites with alcoholics having N550 amplitudes approximately 19 μV smaller than controls at FP sites, 16 μV smaller at Fz, 10 μV smaller at FCz, and not different at more posterior sites. In both papers, we hypothesized that the N550 effects were probably mediated by aging or alcoholism effects on cortical gray matter, but in the absence of MRI data this hypothesis was speculative.

Studies assessing relations between cortical volume and EEG phenomena in the same subjects are rare.⁴⁶ In an evaluation of waking EEG and MRI data, Whitford et al. reported significant correlations between regional slow-wave power and regional gray matter volume in adolescents.⁴⁷ However, to our knowledge the present study is the first to test relations between brain volumes from parcellated structural MRI and sleep EEG data collected from the same subjects. Moreover, it does so using a novel measure of delta activity. The amplitude of the N550 component in the averaged evoked K-complex represents the voltage capable of being produced when the brain is in an appropriate microstate to generate delta. Careful parametric work has revealed it to be an all-or-nothing response when stimulus characteristics are held constant⁴⁸ and that its amplitude is stable over time within a night.⁴⁹ It is thus in a sense a measure of delta generation capacity that is suitable for comparison to tissue volume as a brain structural index of EEG generation capacity.

Recent studies have supported earlier reports in humans¹⁴ and animals¹³ in showing the probable role of the cortex in generating K-complexes. Cash et al. recently reported a consistent and uniform increase over baseline in delta activity associated with K-complexes in electrodes covering the lateral surface of the inferior frontal gyrus and parts of precentral and postcentral gyri.⁵⁰ Using linear microelectrode arrays placed perpendicular to the cortical surface, they also observed current sources that they interpreted as reflecting a passive sink in Layer I and an active source in Layer III of the cortex. Wennberg also indicated cortical generation of evoked and spontaneous K-complexes, with polarity inversions observed in subcortical white matter, thalamus or cingulate tissue, relative to cortical surface recordings.⁵¹ Recent functional magnetic resonance imaging (fMRI) data⁷ demonstrated widespread increased blood oxygen level dependent (BOLD) signals associated with evoked K-complexes relative to baseline EEG periods in the superior frontal, inferior frontal, middle/superior temporal, and cingulate gyri in the left hemisphere, and the middle frontal gyrus insula, and superior, middle, and inferior temporal gyri in the right hemisphere, indicative of a widespread generation source. The data from the intracranial and fMRI studies support the notion that the K-complex is a large, widespread scalp recorded response, generated by synchronized activity in large areas of cortex. These findings are consistent with the current study.

The present data reveal the persistence of significant effects on evoked K-complex amplitude and cortical gray matter volume in alcoholic men sober on average for four and half months and in alcoholic women sober on average for nearly nine months. Whether this effect is permanent or can change with either continued sobriety or relapse to alcoholism requires longitudinal assessment.

The additional independent variance associated with sex at frontal sites is of particular interest as women tend to have larger evoked and spontaneous delta activity despite having smaller ICV and smaller absolute volume of cortical gray matter. Women do show significantly larger temporal lobe gray matter volume than men after correction for ICV but no differences in other regions. The sex effect on the model is seen clearly in the upper panels of Figure 2, where women show more negative (larger) N550 components and correspondingly more negative model Z scores than men.

Observed correlations between N550 amplitude and gray matter volume in the present data highlight the importance of functional gray matter for the production of large amplitude evoked delta activity during sleep. The addition of significant independent variance explained by age and alcoholism highlight that their observed effects on N550 amplitude are at least in part due to their impact on changes in the brain other than those observed in cortical gray matter volume. The lack of significant relations between N550 amplitude and cingulate, thalamic, hippocampal, or basal ganglia volumes support recent studies indicating that these structures probably do not play a role in K-complex generation.

The amplitude of N550 seen in the averaged evoked K-complex is a special case of an evoked delta wave. Delta activity in general is produced by burst firing of groups of neurons driven by slow frequency currents, leading to oscillations in the membrane voltages of cells, taking them closer to or further away from their burst firing threshold potential.⁵² While Massimini and colleagues indicated that delta waves started in prefrontal and orbitofrontal regions, they modeled the slow oscillation as a travelling wave that moved in a rostro-caudal direction across the scalp, implying widespread cortical involvement in the subsequent generation of delta EEG.¹² To produce a large N550, the burst firing needs to be synchronized such that large numbers of neurons not only burst fire, but do so at the same time relative to the stimuli evoking the response. In addition to the number of neurons available, the integrity of white matter pathways connecting neuronal groups will thus also play a major role. While significant, the observed correlations between gray matter volume and N550 amplitude leave large proportions of unexplained variance. It is reasonable to speculate that this unexplained variance could well be accounted for by differences in white matter integrity which is also negatively affected by alcohol dependence and normal aging.^53–59 Further, it is possible that white matter integrity might act as the physiological mediator underlying the independent variance in delta amplitude explained by age and alcoholism and delta amplitude. Tests of these hypotheses will require further study with diffusion tensor imaging.

A limitation of naturalistic observational studies of alcoholism is that it is difficult to disentangle differences that occur as a function of alcohol dependence from those that might reflect premorbid predispositions. It is thus possible that the alcoholics may have had smaller delta amplitude prior to them abusing alcohol. This pattern is seen in other evoked responses such at the P300 which shows evidence of being a marker of alcoholism predisposition.⁶⁰ While this possibility remains for the K-complex, the lack of an effect of family history of alcoholism on N550 amplitude would appear to make it less likely. Future studies addressing the heritability of the K-complex could also inform this question.

The slow oscillations associated with K-complex generation⁸ have been implicated in the decrease of synaptic strength potentiated during prior wakefulness, a fundamental process of synaptic homeostasis essential for learning and brain cellular energy regulation.⁹ This, together with the emerging consensus view^5,7,61 that K-complexes reflect sleep protective mechanisms, highlights the role of cortical gray matter in supporting these essential sleep functions and their likely degradation with normal aging, alcoholism, and likely other conditions producing cortical gray matter volume shrinkage.

DISCLOSURE STATEMENT

This was not an industry supported study. The authors have indicated no financial conflicts of interest.