Department of Applied Mathematics and Statistics, State University of New York, Stony Brook, NY 11794-3600, USA, 1 National Institute on Drug Abuse, 6001 Executive Boulevard, Bethesda, MD 20892-9561, USA, 2 National Institute on Alcohol Abuse and Alcoholism, 6000 Executive Boulevard, Bethesda, MD 20892-7003, USA and 3Chemistry and 4Medical Departments, Brookhaven National Laboratory, Upton, NY 11973-5000, USA
* Author to whom correspondence should be addressed at: Email zhu{at}ams.sunysb.edu
(Received 3 April 2003; first review notified 9 May 2003; in revised form 20 October 2003; accepted 27 October 2003)
![]() |
ABSTRACT |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
![]() |
INTRODUCTION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Indeed, after normalizing for global effects, studies have reported marked change in the patterns of normalized metabolic activity (decreases in posterior areas of the brain and relative increases in striatum during intoxication when compared with sobriety). Here we assess if the changes in the normalized metabolic measures correlate better with the behavioural effects of alcohol than the absolute measures. To address the problem of multiple comparisons when performing correlational analysis between regional measures and behavioural effects, we used principal component analyses for dimension reduction and canonical correlations to summarize the relationship between two sets of measurements (e.g. metabolic measures and cognitive tests). We predicted that while for certain behavioural effects the association with the metabolic changes would be linear, for others that relationship could be non-linear. For this purpose, we assessed regional brain glucose metabolism with PET and FDG in 20 healthy subjects who were tested at baseline and during alcohol intoxication. The results on the metabolic effects of alcohol have been previously published (Wang et al., 2000).
![]() |
SUBJECTS AND METHODS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
PET scan
Subjects received two PET scans with FDG on two separate days within 1 week of each other. For the females, the studies were done in the mid-luteal phase (1622 days after the onset of menstruation). On the first day, subjects drank a placebo (100 ml of diet noncaffeinated soda) 4050 min prior to FDG administration. On the second day, subjects drank a mixture of 95% alcohol (0.75 g/kg) with diet soda added up to 100 ml, 4050 min prior to FDG. The two scans for a given subject were done at the same time of day (±1 h). The subjects were placed in the scanner with their eyes open and their ears unplugged, in a dimly lit room with minimal noise. PET scans were performed with a Siemens HR+ tomograph (resolution 4.5 x 4.5 x 4.5 mm full width half-maximum, 63 slices). Metabolic images were computed as described previously. Thirteen composite brain regions (frontal, parietal, temporal, occipital cortices, basal ganglia, thalamus, limbic system, midbrain, cerebellum, insular, anterior cingulate gyrus, posterior cingulate gyrus, paracentral) were extracted. Measurement of global brain metabolism was obtained by averaging the values from all of the regions of interest.
Motor, behavioural and cognitive evaluation
Motor coordination was evaluated at the beginning and at the end of the study. The following items were assessed: gait (walk heel-to-toe in a straight line for 20 feet); the Romberg test (subject touches the tip of his nose with his fingers while keeping his eyes closed); rhythm (tapping the back and the front of the hand with the other hand) and equilibrium (eyes closed, arms 90 degrees up in front and then stand on one foot for 30 s). For gait and rhythm, the response variable was time required to complete each task and for equilibrium and Romberg the response variable was the number of errors. Subjects' responses under placebo or alcohol influence were rated with respect to baseline (measures prior to placebo or alcohol) as 0 = no change, 1 = minimal change, and 2 = marked changed.
Before placebo or alcohol and at 20, 40, 55, 80 and 140 min after initiation of placebo or alcohol drinking, subjects were asked to evaluate on an analog scale (rated 010) their subjective perception of intoxication (feeling drunk), sleepiness, dizziness and high. Cognitive effects of alcohol were evaluated using the Stroop tests [reading colour names (Stroop read)], describing the colour (Stroop colour), and reading colour names coloured with discrepant colours (Stroop interference), the Controlled Oral Word Association test (COWA), the Symbol Digit Modality test (SDMT), and arithmetic calculations (Woods et al., 1992). The behavioural effects of alcohol were greatest at 80 min after alcohol consumption and therefore measurements at this time point after alcohol or placebo consumption were adopted in the ensuing analyses.
Statistical analysis
Correlational analyses were performed to evaluate the relationship between alcohol's effect on behavioural, cognitive and motor functions and alcohol-induced changes in relative regional brain metabolism. These include the usual canonical correlation to evaluate the linear relationship between two sets of variables and a novel measure, the canonical correlation in the polynomial space, to gauge possible nonlinear relationship between two sets of variables. Principal component analysis was performed for dimension reduction prior to the correlational analyses.
Canonical correlations
There are four sets of variables: metabolism of 13 brain regions of interest (ROIs), six cognitive tests, four motor functional measurements and four behavioural evaluations. These would result in a total of 182 Pearson product moment correlations between changes in metabolism and other measures. Multiple-test corrections such as the Bonferroni adjustment would wipe out virtually any test significance. Besides, variables within the same class are often highly correlated which renders the correlations redundant. Consequently, we adopted the canonical correlations to gauge the relationship between two sets of variables directly. Canonical correlation is essentially the Pearson correlation between the linear combination of variables in one set and the linear combination of variables from another set. The pair of linear combinations having the largest correlation is determined first. Next, the pair of linear combinations having the largest correlation among all pairs uncorrelated with the initially selected pair is identified, and so on. The pairs of linear combinations are called the canonical variables, and their correlations are called the canonical correlations. The first canonical correlation, which is often the only significant one as in our case, is usually adopted to describe the inter-class correlation. Here we will report the first canonical correlation, its test statisticWilks' Lambda (), the equivalent F-statistic and the P-value.
Principal component analysis (PCA)
The sample size in a PET study is intrinsically small (20 here), which would frequently render the degrees of freedom insufficient to detect any significant canonical correlation. We adopted two measures for preliminary dimension reduction. First, only variables that changed significantly ( = 0.05, 2-sided) after alcohol intoxication were included. Second, PCA were performed on the remaining variables in each set for further dimension reduction. The major PCs were selected as follows. For each class, the first few PCs that would account for at least 60% of the variations were selected. Pearson correlations of the selected PCs were obtained and PCs on the behavioural, cognitive or motor functions that were not significantly correlated with PCs of the metabolic measures or vice versa were dropped. Canonical correlations were obtained using the remaining PCs.
Canonical correlations in the polynomial space
To detect possible nonlinear relationship we have included powers of variables in each set, and termed the resulting canonical correlations between the extended variable sets as the canonical correlations in the polynomial space. The usual test of significance for a canonical correlationWilks' Lambda (), is still valid as long as the set of variables without the polynomial terms is jointly normal (Kshirsagar, 1972
). This is a simple and yet effective measure because any function can be written as a polynomial referred to as the Taylor series. Historically, two previous attempts were made to extend the canonical correlations to account for possible nonlinear relationships. The first was by Gregg et al. (1992)
who developed a semi-linear canonical correlation to measure the effect of opioids on the central nervous system. However, it would only allow variables in one set to be non-linear. More recently, Hsieh (2000)
has developed non-linear canonical correlations using neural networks. His method is more general but difficult to implement and interpret because the neural net would not reveal the explicit nonlinear relationship under investigation. In contrast, our method is explicit and expedient. It could be further generalized to include interactions among variables in each set. However, since the principal components are orthogonal to each other, the interaction terms are not necessary here.
Multiple-test correction. To correct for multiple tests, we set the family-wise error rate to be 0.05 and adopted the improved Bonferroni procedure based on the ordered P-values (Simes, 1986). In summary, let P(1)
P(2)
P(k), be the ordered P-values for testing hypotheses H0 = {H(1), H(2)
H(k)}. Then H0 is rejected if P(j) < j*
/k for any j = 1 ... k. All P-values reported are 2-sided.
![]() |
RESULTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
|
|
Correlational analysis using PCs
With the seven remaining PCs, the first canonical correlations between relative metabolic measures and the other three measures were: (1) metabolism and motor: 0.76 ( = 0.43, F = 3.96, d.f. 4, 30; P = 0.0107); (2) metabolism and behaviour: 0.59 (
= 0.65, F = 4.51, d.f. 2,17; P = 0.0269); and (3) metabolism and cognition: 0.50 (
= 0.74, F = 1.19, d.f. 4, 30; P = 0.3348). Further examination revealed that the first PC of the metabolism had a non-linear relationship with the second PC of the cognitive functions. Therefore a quadratic term of the first PC of metabolism was added and the resulting first canonical correlation in the polynomial space was 0.76 (
= 0.40, F = 2.67, d.f. 6, 28; P = 0.0357). This non-linear correlation is more reflective of the true nature of the underlying relationship and thus supersedes the corresponding linear canonical correlation.
In summary, the three ordered P-values are 0.0107, 0.0269 and 0.0357. They are smaller than the corresponding threshold values 0.0167, 0.0333, and 0.05 in the ordered P-value test (k = 3, = 0.05). Therefore we conclude that changes due to alcohol influence in brain regional metabolism are significantly correlated with the corresponding changes in motor function, cognitive function and behavioural measurements.
![]() |
DISCUSSION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Cognitive and metabolic measures the nonlinear brain
The brain is said to be nonlinear (Frackowiak et al., 1997). Thus, the conventional measures for linear relationships including the usual canonical correlations would be insufficient to detect all possible relationships. We have developed a simple new measure, the canonical correlations in the polynomial space, to detect possible nonlinear relationships. The fact that the first PC of the relative metabolism has a nonlinear relationship with the second PC of the cognitive functions confirmed our suspicion. By adding a quadratic term of the first PC of metabolism, the resulting first canonical correlation in the polynomial space between metabolism and cognition increased from 0.50 (P = 0.3348) to 0.76 (P = 0.0357), a big leap from an insignificant linear relationship to a significant nonlinear relationship. This is the first ever documentation of a non-linear relationship between metabolism and cognition according to our knowledge. This non-linear association suggests that the effects of alcohol on cognitive function may not be evident until after large changes in metabolic activity (shifts in activity in frontal and anterior cingulate versus activity in cerebellum) have occurred. The association between alcohol-induced changes in activity in frontal cortex and anterior cingulate was expected since these brain regions are involved in executive functions, working memory, decision making and attention, all of which were required to perform the cognitive tasks (Duncan and Owen, 2000
; Krawczyk, 2002
). Though the cerebellum has been traditionally considered to be involved with motor coordination increasing, evidence from imaging studies points to its involvement in cognitive operations (Habas, 2001
). It is suggested that this is in part achieved via its neuroanatomical connections with the frontal cortex (Andreasen et al., 1998
). Indeed, in this study it was the contrast in the activity between the frontal and anterior cingulate and that in cerebellum that accounted for the association with cognitive performance.
Behavioural and metabolic measures
The significant canonical correlation was mainly due to the association between the second PC (contrast between metabolism in basal ganglia and that in temporal insula) and the behavioural measures that reflected the subjective perception of intoxication (drunk, dizzy, sleep and high). The striatum, which includes the nucleus accumbens, is one of the brain regions directly implicated in the reinforcing effects of drugs of abuse. Specifically it is believed that the ability of drugs of abuse, including alcohol, to increase DA in nucleus accumbens underlies their reinforcing effects (Di Chiara and Imperato, 1988; Koob and Bloom, 1988
). Indeed, imaging studies have reported striatal activation during drug intoxication (Breiter et al., 1997
; Stein et al., 1998
). The insula is also a brain region that had previously been shown to be activated during alcohol intoxication (Sano et al., 1993
). In fact, alcohol-induced increases in CBF in temporal cortex have been linked to its reinforcing effects (Ingvar et al., 1998
). Moreover, animal studies (Collins et al., 1996
; Lyons et al., 1998
; Crews et al., 2001
) consistently documented the temporal insula as one of the major brain regions affected by acute alcohol intoxication.
Motor and metabolic measures
The significant canonical correlation was largely due to the association between motor functions and the second PC of relative metabolism (contrast between metabolism in basal ganglia and that in the temporal insula). This suggests an involvement of these brain regions in the motor-incoordinating effects of alcohol. Indeed the effects of alcohol on basal ganglia have been associated with its motor effects (Williams-Hemby and Porrino, 1994; Dar, 1998
). Though to our knowledge the effects of alcohol on the temporal insula have not been linked with the motor impairing effects of alcohol, imaging studies have documented the role that the temporal insula has on learning of new motor sequences (Ghaem et al., 1997
).
In summary, the results from this study suggest that alcohol-induced deterioration in cognitive and motor function as well as its behavioural effects are linked with changes in patterns of brain activity rather than changes in specific brain regions. Specifically, the contrasting effects of alcohol in basal ganglia versus the insula appear to be involved in the perception of 'feeling drunk', and the contrasting effects in cerebellum versus those in frontal and parietal cortices are involved in its motor-incoordinating effects. The cognitive effects were also linked with a contrast function between frontal and anterior cingulate versus cerebellum though this relationship was non-linear, which suggests that a threshold effect may be necessary in order to observe impairment during intoxication.
![]() |
ACKNOWLEDGEMENTS |
---|
![]() |
REFERENCES |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Bates, M. E. (2000) Utility of component-process approaches for understanding complex alcohol-related behavior within an executive functioning framework: comment on Giancola (2000). Experimental and Clinical Psychopharmacology 8, 598600.[CrossRef][ISI][Medline]
Breiter, H. C., Gollub, R. L., Weisskoff, R. M., Kennedy, D. N., Makris, N., Berke, J. D., Goodman, J. M., Kantor, H. L., Gastfriend, D. R., Riorden, J. P., Mathew, R. T., Rosen, B. R. and Hyman, S. E. (1997) Acute effects of cocaine on human brain activity and emotion. Neuron 19, 591611.[ISI][Medline]
Collins, M. A., Corso, T. D. and Neafsey, E. J. (1996) Neuronal degeneration in rat cerebrocortical and olfactory regions during subchronic "binge" intoxication with ethanol: possible explanation for olfactory deficits in alcoholics. Alcoholism: Clinical and Experimental Research 20, 284292.[ISI][Medline]
Crews, F. T., Braun, C. J., Ali, R. and Knapp, D. J. (2001) Interaction of nutrition and binge ethanol treatment on brain damage and withdrawal. Journal of Biomedical Science 8, 134142.[CrossRef][ISI][Medline]
Curtin, J. J., Patrick, C. J., Lang, A. R., Cacioppo, J. T. and Birbaume, N. (2001) Alcohol affects emotion through cognition. Psychological Science 12, 527531.[CrossRef][ISI][Medline]
Dar, M. S. (1998) Involvement of kappa-opioids in the mouse cerebellar adenosinergic modulation of ethanol-induced motor incoordination. Alcoholism: Clinical and Experimental Research 22, 444454.[ISI][Medline]
Di Chiara, G. and Imperato, A. (1988) Drugs abused by humans preferentially increase synaptic dopamine concentrations in the mesolimbic system of freely moving rats. Proceedings of the National Academy of Sciences of the United States of America 85, 52745278.[Abstract]
Duncan, J. and Owen, A. M. (2000) Common regions of the human frontal lobe recruited by diverse cognitive demands. Trends in Neurosciences 23, 475483.[CrossRef][ISI][Medline]
Eckardt, M. J., Rohrbaugh, J. W., Rio, D. E. and Martin, P. R. (1990) Positron emission tomography as a technique for studying the chronic effects of alcohol on the human brain. Annals of Medicine 22, 341345.[ISI][Medline]
Frackowiak, R. S. J., Friston, K. J., Dolan, R. J. and Mazziotta, J. C. (1997) Human brain function. Academic Press, London.
Ghaem, O., Mellet, E., Crivello, F., Tzourio, N., Mazoyer, B., Berthoz, A. and Denis, M. (1997) Mental navigation along memorized routes activates the hippocampus, precuneus, and insula. Neuroreport 8, 739744.[ISI][Medline]
Giancola, P. R. (2000) Executive functioning: a conceptual framework for alcohol-related aggression. Experimental and Clinical Psychopharmacology 8, 576597.[CrossRef][ISI][Medline]
Gregg, K. M., Varvel, J. R. and Shafer, S. L. (1992) Application of semilinear canonical correlation to the measurement of opioid drug effect. Journal of Pharmacokinetics and Biopharmaceutics 20, 611635.[ISI][Medline]
Habas, C. (2001) The cerebellum: from motor coordination to cognitive function. Review Neurology 157, 14711497.
Hsieh, W. W. (2000) Nonlinear canonical correlation analysis by neural networks. Neural Network 13, 10951105.[CrossRef][ISI]
Ingvar, M., Ghatan, P. H., Wirsen-Meurling, A., Risberg, J., Von Heijne, G., Stone-Elander, S., Ingvar, D. H. (1998) Alcohol activates the cerebral reward system in man. Journal of Studies on Alcohol 59, 258269.[ISI][Medline]
Koob, G. F. and Bloom, F. E. (1988) Cellular and molecular mechanism of drug dependence. Science 242, 715723.[ISI][Medline]
Krawczyk, D. C. (2002) Contributions of the prefrontal cortex to the neural basis of human decision making. Neuroscience Biobehavior Review 26, 631664.[CrossRef][ISI]
Kshirsagar, A. M. (1972) Multivariate Analysis. Marcel Dekker, New York.
Lau, M. A., Pihl, R. O. and Peterson, J. B. (1995) Provocation, acute alcohol intoxication, cognitive performance, and aggression. Journal of Abnormal Psychology 104, 150155.[CrossRef][ISI][Medline]
Lemon, J., Chesher, G., Fox, A., Greeley, J. and Nabke, C. (1993) Investigation of the "hangover" effects of an acute dose of alcohol on psychomotor performance. Alcoholism: Clinical and Experimental Research 17, 665668.[ISI][Medline]
Lyons, D., Miller, M. D., Hedgecock-Rowe, A. A., Crane, A. M. and Porrino, L. J. (1998) Time-dependent effects of acute ethanol administration on regional cerebral blood flow in the rat. Alcohol 16, 213219.[CrossRef][ISI][Medline]
Meng, Z. H., Anwer, J. and Dar, M. S. (1997) The striatal adenosinergic modulation of ethanol-induced motor incoordination in rats: possible role of chloride flux. Brain Research 776, 235245.[CrossRef][ISI][Medline]
Quillian, W. C., Cox, D. J., Kovatchev, B. P. and Phillips, C. (1999) The effects of age and alcohol intoxication on simulated driving performance, awareness and self-restraint. Age and Ageing 28, 5966.[Abstract]
Sano, M., Wendt, P. E., Wirsen, A., Stenberg, G., Risberg, J. and Ingvar, D. H. (1993) Acute effects of alcohol on regional cerebral blood flow in man. Journal of Studies on Alcohol 54, 369376.[ISI][Medline]
Simes, R. J. (1986) An improved Bonferroni procedure for multiple tests of significance. Biometrika 73, 751754.[ISI]
Stein, E. A., Pankiewicz, J., Harsch, H. H., Cho, J. K., Fuller, S. A., Hoffmann, R. G., Hawkins, M., Rao, S. M., Bandettini, P. A. and Bloom, A. S. (1998) Nicotine-induced limbic cortical activation in the human brain: a functional MRI study. American Journal of Psychiatry 155, 10091015.
Volkow, N. D., Mullani, N., Gould, L., Adler, S. S., Guynn, R. W., Overall, J. E. and Dewey, S. (1988) Effects of acute alcohol intoxication on cerebral blood flow measured with PET. Psychiatry Research 24, 201209.[CrossRef][ISI][Medline]
Volkow, N. D., Hitzemann, R., Wolf, A. P., Logan, J., Fowler, J. S., Christman, D., Dewey, S. L., Schlyer, D., Burr, G. and Vitkun, S. (1990) Acute effects of ethanol on regional brain glucose metabolism and transport. Psychiatry Research 35, 3948.[Medline]
Volkow, N. D., Wang, G.-J., Begleiter, H., Hitzemann, R., Pappas, N., Burr, G., Pascani, K., Wong, C., Fowler, J. S. and Wolf, A. P. (1995) Regional brain metabolic response to lorazepam in subjects at risk for alcoholism. Alcoholism: Clinical and Experimental Research 19, 510516.[ISI][Medline]
Wang, G-J., Volkow, N. D., Franceschi, D., Fowler, J. S., Thanos, P. K., Scherbaum, N., Pappas, N., Wong, C., Hitzemann, R. and Felder, C. (2000) Regional brain metabolism during alcohol intoxication. Alcoholism: Clinical and Experimental Research 24, 822829.[ISI][Medline]
Williams-Hemby, L. and Porrino, L. J. (1994) Low and moderate doses of ethanol produce distinct patterns of cerebral metabolic changes in rats. Alcoholism: Clinical and Experimental Research 18, 982988.[ISI][Medline]
Woods, J. H., Katz, J. L. and Winger, G. (1992) Benzodiazepines; use, abuse and consequences. Pharmacological Reviews 44, 151347.[ISI][Medline]