- Email: [email protected]
Electrophysiological indices predict resumption of drinking in sober alcoholics
Alcohol, Vol. 10, pp. 89-95, 1993
0741-8329/93 $6.00 + .00 Copyright©1993PergamonPress Ltd.
Printed in the U.S.A. All rightsreserved.
Electrophysiological Indices Predict Resumption of Drinking in Sober Alcoholics SUSAN WAGNER
GLENN, l RAJITA SINHA* AND OSCAR A. PARSONS
Center f o r Alcohol and Drug Related Studies, University o f Oklahoma Health Sciences Center, Oklahoma City, OK 73104 *Department o f Psychiatry, Yale University School o f Medicine, New Haven, C T 06519 Received 2 A p r i l 1992; A c c e p t e d 4 S e p t e m b e r 1992 GLENN, S. W., R. SINHA AND O. A. PARSONS. Electrophysiologicalindicespredict resumption of drinking in sober alcoholics. ALCOHOL 10(2) 89-95, 1993. -Eighty-nine alcoholics and 54 nonalcoholic controls were tested on measures of late component event-related potentials (ERPs) using a visual "oddball" stimulus task. The alcoholics had just completed inpatient alcoholism treatment programs and were 21-45 days sober. Approximately 13 months later, subjects returned for retesting; alcoholics were classified as resumers or abstainers based on their drinking patterns during the intertest interval. Using the ERP measures from the initial testing session, alcoholics differed significantly from controls in the multivariate analysis and on P300 amplitude (P3A). Resumer alcoholics showed significantly longer N200 latencies (N2L) than abstainer alcoholics. Discriminant function analyses predicting resumer/abstainer status from N2L, P3A and NIA indicated a 63e/0 prediction rate, X2 = 5.67, p < 0.02. Addition of N2L to previously tested psychological and social predictor variables indicated an increase in the amount of variance explained. The results support a biopsychosocial model for understanding and predicting relapse in chronic alcoholics. Alcoholism
Evoked potentials
ERPs and alcoholism
Biopsychosocial model
Prediction of resumption
and marital status, history of multiple previous hospitalizations, depression, and childhood hyperactivity symptoms have all been implicated as predictor variables for resumption o f drinking (4,7,14,19). In addition, cognitive impairment has been shown to also have specific predictive capabilities. Alcoholics who resume drinking manifest neuropsychological deficits relative to those alcoholics who stay sober (1,15). These deficits are present at the time of initial treatment and testing and remain as long as 14 months (7,15). Note, then, that both psychological and social variables have successfully served as variables for predicting relapse. Because our understanding of the development and progression of alcoholism has been enhanced by the introduction of a biopsychosocial approach, the investigation of biological variables which may also predict relapse poses an interesting investigative question. It has been suggested that as an indicator of cerebral impairment, late components of event-related brain potentials (ERPs) may be even more sensitive measures than psychometric tests (5). Evoked potential measures have fairly consistently been found to be impaired in chronic alcoholics (e.g., 5,22,29) and have been correlated with performance deficits on neuropsychological tests (16). It would follow, then, that evoked potential responses, like neuropsychologieal performance, may also serve as predictors of resumption of drinking, and may provide additional information concerning the
A L C O H O L abuse is one of the nation's foremost health problems, as over 10.5 million adults suffer from alcoholism and millions o f others are directly or indirectly affected by its consequences (2). Treatment of alcoholism has only been partially successful, however, as normative outcome statistics indicate relapse rates as high as 80% over a I-year period (11). The identification of those persons most likely to relapse or resume drinking following treatment, is of considerable interest from several standpoints. First, if those persons most vulnerable to relapse could be identified prospectively, perhaps special treatment programs addressing specific relapse issues could be designed and implemented. Second, recognition of those prone to relapse may aid follow-up programs in providing more individual attention and in specifically targeting those most in need. Third, a priori identification of potential resumers based on differences between resumers and abstainers in cognitive, behavioral, or physiological characteristics will further our understanding of concurrent or premorbid factors which may contribute to the development or perpetuation of alcoholism. It is this third issue which we address in the present study. Previous research has indicated that certain psychological or psychosocial variables seem to characterize or help identify individuals who resume drinking following alcoholism treatment. Variables such as employment instability, residential
Requests for reprints should be addressed to Susan Wagner Glenn, Ph.D., Center for Alcohol and Drug Related Studies, University of Oklahoma Health Sciences Center, 800 NE 15th, Suite 410, Oklahoma City, OK 73104. 89
90
GLENN, SINHA AND PARSONS
characteristics of those individuals who are prone to relapse. These issues are addressed in the current study by answering the following questions: 1. Do ERPs differentiate alcoholics from control subjects? 2. Do ERPs measured following treatment predict resumer/ abstainer status one year later? 3. Do ERPs add to the prediction of resumer/abstainer status achieved by psychosocial variables and neuropsychological performance? METHOD
Subjects The alcoholic sample consisted of 53 men and 36 women recruited from community inpatient alcoholism treatment centers in and around the Oklahoma City metropolitan area. The control sample consisted of 23 men and 31 women recruited by newspaper and word of mouth. Alcoholic subjects met the National Council on Alcoholism (13) or the DSM-III-R (6) criteria for alcoholism and had been detoxified for 3 to 6 weeks prior to testing. Exclusion criteria for all subjects (alcoholics and controls) included a borderline intelligence or lower (as measured by the Shipley Vocabulary Scale; 25), a history of neurological disease or trauma, current psychoactive medications, history o f major psychiatric or medical disorder, or substantial history of abuse of drugs other than alcohol. Subjects were screened prior to testing for auditory acuity (those with hearing loss over 15 db were excluded) and color vision (subjects with red-green color blindness were excluded). Subjects gave informed consent and received monetary compensation for their participation.
Event-Related Potentials The ERP paradigm used has been described in detail in our earlier work (16) and is a modification of the paradigm employed by Porjesz and Begiefter (21). ERPs were recorded in an oddball dual-modality target selection paradigm in which auditory stimuli (1000 or 1500 I-Iz 50 ms tone pips) and visual stimuli (red or green 1 in circles) were presented in random order with ISis ranging from 0.5 to 1.5 s. Auditory stimuli were presented by earphones and visual stimuli were presented on a CRT screen 1 meter in front of the subject. There were 4 ERP runs with a minimum of 400 trials per run such that each of the 4 stimuli served as a "target" during one run. The sequence of the runs was randomly preselected for each subject. The total number of stimuli in each modality (visual or auditory) was approximately equal in each run. However, the target stimuli for each run was the oddball or rare stimulus in that it was presented 20% of the time for the stimuli of that modality, 80% of the stimuli in that particular modality were composed of nontarget stimuli. The subject's task was to count to himself the number of target stimuli (attended trials) while disregarding the nontarget, same modality stimuli as well as the nontarget, different modality stimuli (nonattended trials). After each 100 trials the subject was asked to provide the number of target stimuli which had been presented thus far. If the count given was it.correct, the subject was given the correct count and instructed to strive for accuracy and to continue counting from the corrected count. Although this procedure did not allow for the discrimination of ERPs for correctly identified versus incorrectly identified targets, most subjects made very few, if any, counting errors and thus the overwhelming majority of the target trials represented correctly identified targets.
Electrode sites used for recording included Cz, Pz, and Oz, F3 and F4. Reference electrodes were linked earlobes. The Grass Model 78-B polygraph was calibrated with a 20 Hz, 50 microvolt peak-to-peak input signal, which resulted in an output value of _+2.5 volts. Analog filters of the Grass 7P511E amplifiers were set at one-half down at 0.3 Hz and 100.0 Hz. A / D conversion was done with 1024 bit resolution of _+5 volt input at the board. Immediately before stimulus presentation, each electrode was sampled 250 times over a 500 msec interval to provide a baseline voltage for that particular trial. During this half-second period, if the sampled EEG exceeded 75% of the output of the amplifier (or 75 mv peak to peak), the trial was reset and the baseline sampling restarted. The same screening process was repeated for the poststimulus period in that EEG output exceeding 75 microvolts peak-topeak during the 500 msec following stimulus presentation resulted in the trial being rejected and repeated later in the experimental run. In this way, trials were kept free from artifacts secondary to eyeblinks, eye movements or gross physical movements, or high frequency alpha EEG activity which could mask ERP responses. The total number of rejected trials or the average experimental run time did not differ between alcoholic and control subjects. A peak detection algorithm was applied to the averaged trial vectors for each stimulus condition for electrodes Oz, Pz, and Cz to obtain automatic computation of latencies and amplitudes of the NI, N2, and P3 ERP components. The N l , N2, and P3 amplitudes reported (NIA, N2A, and P3A) are absolute voltages measured from the 1/2 s prestimulus baseline. Because preliminary analyses indicated that there were no significant differences in the amplitudes or latencies of these three components between the attended target trials of the two visual or auditory stimuli, the scores were combined over the two stimuli in each modality. Our preliminary analyses and subsequent studies (16,18) also indicated that the greatest differences between alcoholics and controls were found in the visual modality, particularly with the Pz electrode, therefore, only the averaged scores for the attended target trials of the visual modality at Pz are presented in the current study.
ERP Measures The amplitudes and latencies of the ERP measures used have been associated with various aspects of the cognitive process. The N1 wave (N100) can be elicited by stimuli delivered in various modalities and typically has a poststimulus latency of 120 msec in normal healthy adults. Given stimuli presented in an attended modality and a nonattended modality, the NI amplitude is increased by both target and nontarget stimuli in the attended modality and thus reflects a measure of modality selection. The Nd, or negative difference wave, represents the difference in amplitudes between the N 1 components of the target and nontarget stimuli and is considered a measure of selective attention and further processing of the target stimuli within the relevant modality (12,18,24). The N2 component of the ERP (N200) has a typical latency of 200-250 msec poststimulus and is reliably evoked by infrequent stimuli placed randomly in a stream of frequent stimuli. The N2 wave is thought to reflect a mismatch process associated with the orienting reflex (24). N2 has also been found to be related to stimulus complexity, and is sensitive to the nature of cognitive tasks (5,9). The P3 (P300) component is a large endogenous event with positive polarity which peaks at 300-500 msec poststimulus.
ELECTROPHYSIOLOGICAL INDICES OF RESUMPTION
It is often evoked in an oddball target selection task in which the subject is required to respond to infrequent target stimuli occurring in a series of nontarget frequent stimuli. The P3 amplitude is increased by stimuli that are novel, task relevant, and are somehow important to the subject or require the updating of working memory. For further discussion of these late component measures of the ERP, see Patterson et eLI.(17), Johnson (I0), and Pritchard (23).
91 TABLE 1 FOLLOW-UP DATA FOR ALCOHOLICS NOT RETESTED n
~'e
Follow-Up
13 3
(33.3°70) (7.7070)
6
(15.4o'/o)
4
00.3070)
3
(7.7We)
2
(5.107o)
6
(15.407o)
l
(2.5%)
1
(2.507e)
Lost immediately after initial testing session Sober for two months, then moved out of town or were put into jail Sober for three months, then moved, were put into jail, or resumed drinking and refused to participate with followup calls. Sober for four months, then moved or resumed drinking and refused to continue with followup, Sober 6-7 months, then moved or resumed drinking and refused to continue with followup. Sober 9-10 months, then moved or refused to continue with followup. Sober 12 months, but then moved or refused to return for followup testing. Sober for 12 months, then resumed heavy drinking and moved out of state. Light drinking in the first month, progressed to moderate drinking for the next 12 months, but refused to participate in followup testing.
Procedures The subjects in this study were part of a larger study investigating the neuropsychological, psychosociai, as well as electrophysiological consequences of alcohol abuse. Following completion of inpatient alcoholism treatment, alcoholic subjects were brought to the Center for a day of testing and interviews. Control subjects were also tested at the Center following an initial screening session. All subjects were given an Intoxilyzer test prior to beginning the testing session. Any person with breath alcohol readings over 0.01 was rescbeduled. Subjects participated in the electrophysiological tests in the morning, receiving the oddball counting task, as described previously, in addition to an auditory evoked brainstem response task and a visual pattern reversal response task. Including placement of electrodes, task set-up time, and short breaks between tasks, the electrophysiologicai procedures took approximately 3-3.5 h to complete. The subjects were then given a lunch break. The afternoon testing session consisted of a neuropsychological test battery and interviews concerning drinking patterns, affective symptomatology, childhood behaviorai disorders, and psychosociai variables. The results of the afternoon testing session have been reported elsewhere (7,15). All of the testing procedures were repeated approximately l year later (average for the three groups, 13.7 months), but the predictive measures reported in the current study are the results of only the initialtesting session. (The results of the E R P measures at the retesting session are reported elsewhere; 8.) During the intertest period, alcoholic subjects were contacted monthly by phone and were questioned concerning current drinking behaviors. Collaterals, or friends or relatives designated by the subject, were also contacted by experimenters on a regular basis and provided corroboration of the subject'sdrinking status in 9 6 % of the cases. If subjects or collaterals could not be reached during a particular month, a letter was sent to the subject requesting that they phone the research center and/or provide forwarding address information if a change of residence had taken place. O n the basis of these monthly contacts with subjects and the drinking histories obtained in the follow-up testing session, subjects who consumed at least I0 oz of ethanol (approximately 17 drinks) during the 6 months prior to retesting (approximately 8-14 months after initial testing) were classified as resumers (n = 36); the remainder were classified as abstainers (n = 53). Of the 126 alcoholic subjects receiving all of the test measures in the initial testing session, 89 were retested, yielding a retest or follow-up rate of 71%. Data identifying the time frame and various reasons for failure to complete the followup portion of the study are provided for the alcoholics who were not retested (See Table 1). Control subjects were not contacted on a monthly basis but were sent birthday and Christmas cards and were contacted by phone as the time for retest approached. Of the 76 control subjects tested initially, 54 returned for retesting, yielding a follow-up rate of 71%. Combining controls, abstainers and
resumers, 112 (67.9%) of the men and 90 (74.4%) of the women were retested (X2" = 1.05, p = ns). Demographic and neuropsychologicai performance differences between subjects who returned for retest and those who did not are discussed in Parsons et al. (15). The only differences in ERP measures between the two groups o f subjects were on Nd (for controls) and P3A (for alcoholics) (Table 2). RESULTS
The results of the demographic data are shown in Table 3. Controls, resumers, and abstainers did not differ in age, education, or intertest interval. Resumers and abstainers did not differ in years o f alcoholism or ounces of absolute ethanol consumed in the 6 months prior to treatment. By selection procedures, the two groups did differ in the amount of absolute ethanol consumed in the 6 months prior to retest. Significant differences were found in the gender distribution of resumers and abstainers. Twenty-eight alcoholic women (77.8%) were abstainers, 8 (22.2%) were resumers. Of the 53 men alcoholic subjects, 25 (47.1%) were abstainers and 28 (52.8%) were resumers 0c2 = 8.34, p < 0.01). A 3 × 2 multivariate analysis (MANOVA) was conducted initially to examine the main effects of group (controls, abstainers, and resumers), sex, and group × sex on the ERP dependent variables (N1A, Nd, N2A, P3A, N2L, and P3L). A significant main effect of group was found (Wilkes' Lambda F -- 2.00, p = 0.0248), but there was no significant main effect of sex or group × sex interaction. Given these results, univariate analyses were next conducted to examine the effect of group on each of the individual ERP measures. The results of these analyses can be seen in Table 4. Significant effects of group were found for N2L and P3A, with a trend toward significance for N I A . Post-hoc examination of the means using Duncan's Multiple Range Test indicated that the group effect was primarily due to differences between resumers versus abstainers and controls for N2L and to differences between controls versus resumers and abstainers for
92
GLENN, SINHA AND PARSONS TABLE 2 ERP VARIABLES (MEAN +_ SD) FOR SUBJECTS RETESTED VERSUS SUBJECTS NOT RETESTED
NIA Nd* N2A N2L P3At P3L
Controls Retested (n = 54)
Controls Not Retested (n = 22)
Alcohols Retested (n = 89)
Alcohols Not Retested (n = 37)
- 4.63 (2.45) 1.20 (2.06) - 0.14 (3.47) 277.33 (26.14) 10.86 (3.36) 394.35 (29.77)
- 5.08 (4.07) 2.47 (3.19) - 0.34 (3.49) 269.58 (24.60) 10.14 (4.03) 400.80 (28.57)
- 3.91 (2.07) 1.01 (1.75) 0.58 (2.89) 283.26 (26.38) 9.24 (3.27) 394.28 (30.33)
- 4.39 (3.38) 1.36 (2.99) 0.19 (3.42) 282.01 (29.56) 11.11 (3.79) 402.18 (31.49)
*For controls retested versus not retested, F(1, 75) = 4.06, p < 0.05. tFor alcoholics retested versus not retested, F(1,130) = 8.17, p < 0.01.
P3A. O n N1A, controls a n d a b s t a i n e r s were n o t significantly different f r o m each o t h e r a n d abstainers a n d resumers were not significantly different, b u t resumers were significantly different f r o m controls. In order to examine the efficacy o f the E R P variables in predicting r e s u m e r / a b s t a i n e r status, a d i s c r i m i n a n t f u n c t i o n analysis was c o n d u c t e d using t h e three E R P variables which h a d yielded significant or near-significant g r o u p effects (N2L, P 3 A , a n d N1A). Using this three v a r i a b l e model, 32 (61.5%)
o f abstainers a n d 21 (65.6*70) o f resumers were correctly identified for a n overall prediction rate o f 6 3 . 1 % (×2 = 5.67, p < 0.02). T h e t h i r d experimental question e x a m i n e d the possible increase in predictive capacity which m i g h t be gained by the inclusion o f E R P variables with neuropsychological a n d psychosocial variables which have been s h o w n to predict r e s u m e r / a b s t a i n e r status. In a previous study (7), we examined the role o f 5 variables in predicting r e s u m p t i o n o f drink-
TABLE 3 MEANS (+ SD) OF DEMOGRAPHIC AND DRINKING VARIABLES Controls (n = 54)
Abstainers (n = 53)
Resumers (n = 36)
Age
36.5 (9.9)
39.2 (9.8)
36.3 (8.0)
Years o f education
13.7 (2.0)
12.9 (1.6)
13. l (2. i)
13.9 (5.2)
12.9 (3.3)
14.3 (3.2)
0.41 (0.50)
0.98 (0.51)
I. 14 (0.71)
Interest interval (months) Drug index* Oz. Absolute Etohl" 6 months prior to test Oz. Absolute Etoh~ 6 months prior to retest Years Alcoholic
43.2 (64.2) 53.5 (101.2) -
1632.1 (1220.9) 0.27 (0.83)
1823.0 (1294.4) 598.8 (766.5)
11.0
11.7
(8.1)
(7.2)
*Drug Index is measure in which subjects were asked if they had ever tried any of 10 classes of drugs (e.g., amphetamines, barbiturates, halhicinogenics, etc). The number given represents the mean number of affirmative responses for subjects in each group. For the ANOVA, F(2, 113) = 18.38, p < 0.001; Resumers = Abstainers > Controls. tF(2, 113) = 50.92, p < 0.001; Resumers = Abstainers > Controls. ~ ( 2 , 113) = 29.13,p < 0.001; Resumers > Abstainers = Controls.
93
ELECTROPHYSIOLOGICAL INDICES OF RESUMPTION TABLE 4 CONTROLS, ABSTAINERS,AND RESUMERS ON ERP VARIABLES
NIA~ NdA N2A N2L P3A P3L
Controls
Abstainers
Resumers
(n = 54)
(n = 53)
(n = 36)
F*
-4.63 (2.47), 1.20 (2.08)° - 0.12 (3.50)d 277.89 (26.05)e 10.89 (3.37) 394.39 (30.02)s
-4.13 (1.91),.b 1.08 (1.63)c 0.61 (2.95)d 277.99 (25.05), 9.28 (3.47)f 392.35 (27.84)~
- 3.57 (2.28)b 0.89 (1.96)c 0.54 (2.82)d 291.03 (26.70) 9.17 (3.00)f 397.03 (33.74)g
2.42
.0926
0.28
n.s.
0.85
n.s.
p
3.45 .0344 4.56
.0120
0.27
n.s.
Means (+ SD) are presented. Means with the same subscript letters are not significantly different by Duncan's Multiple Range Test. *F's and p's represent ANOVA between the three groups (df = 1,142). tERP measures are from Pz electrode and are the average responses of the visual modality responses to the target stimuli. ing: depressive symptomatology, neuropsychological performance, psychosocial maladjustment, previous treatment history, and childhood attention deficit disorder symptomatology. The subjects used were the same as in the present study with the addition of several alcoholics who did not receive the ERP testing. Using the 5 mentioned variables in a discriminant function analysis, 79% of abstainers and 67%0 of resumers were correctly identified for an overall prediction rateof75%0 (X2 = 22.1,p < 0.001). In order to examine th~ utility of the inclusion of the present study's ERP variables into this predictive model, the N2L measure was chosen as the ERP measure which best differentiated resumers and abstainers. A discriminant function analysis was then conducted using the 5 psychosocial and neuropsychological variables described in the previous study and N2L. Seventy-four percent of abstainers and 69% of resumers were correctly identified using this 6 variable model for an overall predictive rate of 71.5% (X2 = 16.1, p < 0.001). Stepwise regression equations predicting resumer/abstainer status indicated an R" of .26 or that 26% o f the variance was explained using the original 5 variable model; the R: increased to .31 with the addition of N2L to the equation, indicating some degree o f shared variance among the variables. Further examination of the 6-variable equation revealed that the depressive symptomatology, psychosocial maladjustment and N2L were the significant contributors to the equation, F s (3, 86) = 4.6 to 13.8, p's = 0.034 to 0.001, together contributing .25 to the total R 2 of .31. Using only these three significant variables, a final discriminant function analysis was conducted. Seventytwo percent o f abstainers and 69% of resumers could be identified usin[g this 3-variable model for an overall prediction rate o f 7 1 % , x = 14.7,p < 0.001. DISCUSSION In answer to the first experimental question, ERPs successfully differentiated alcoholics from controls as seen in the significant group main effect and in the P3A and N2L measures. Visual examination of the ERP variables illustrates that controls generally showed greater amplitudes and shorter latencies than their alcoholic counterparts, even on measures
which did not show statistical significance. These results replicate those of previous studies which have examined ERP changes in chronic alcoholics (5,22,29) and provide further support for findings of central nervous system changes as a result of chronic alcohol ingestion. The second experimental question concerned the predictive capacity of evoked potential responses in accurately determining resumer/abstainer status. Using three of the ERP measures, N2L, P3A, and N1A, 61.5%0 of abstainers and 65.6%0 of resumers could be correctly identified. These findings are significant, considering that the ERP measures were recorded an average of 13 months before the resumer/abstainer outcome classification was obtained. In previous studies, ERP measures have been found to be indices sensitive to various effects of alcohol such as alcoholization (acute changes), tolerance, withdrawal, and long-term brain dysfunction (22). With the findings of the present study, predictive capabilities may be added to the list of potential uses and indications for ERPs and may suggest new applications for these measures in future studies. The a priori identification of resumers and abstainers based on ERP measures provides support for the hypothesis that those individuals who are likely to relapse can be characterized by variables indicative of greater nervous system dysfunction in addition to psychosocial and neuropsychological variables. Impaired ERPs have been found to be related to greater cognitive deficits (18,27) and are thought to reflect deficits in the integrity of the information processing system (29). In the present study, N2L was the ERP measure which most clearly differentiated resumers from abstainers. Porjesz et al. (22) specifically investigated the N2 component of ERPs in an alcoholic population. Using an easy/difficult line discrimination task, the authors found delayed N2 latencies in alcoholics relative to controls; the differences were particularly apparent for the easy discrimination part of the task. The authors reported that the alcohol-related deficit in N2L reflected an impairment in (a) the time taken for stimulus evaluation processes and (b) the ability of alcoholics to adequately differentiate between easy and difficult tasks, suggesting that alcoholics adopt an undifferentiated mode of responding regardless of task requirements. Generalizing from these laboratory
94
GLENN, SINHA AND PARSONS
findings to the complexities involved in relapse is difficult. However, it is possible that the N2L deficit in resumers compared to abstainers reflects a relatively greater impairment in the ability to evaluate and differentiate between stimuli, and, even more speculatively, a concomitant inability of resumers to learn to adopt new modes of responding required of new situational tasks (i.e., not returning to drinking after release from treatment). Obviously, replication of the N2L findings in resumers versus abstainers needs to be achieved before the above speculation can be seriously considered. Impairment in ERPs has been implicated as a marker for the development of alcoholism, as the ERP changes described for chronic alcoholics have been found in the young sons of alcoholic fathers (3,28) as well as in nonalcoholic adults with a family history of alcoholism ( F H + ) (17). Pfefferbaum et al. (20) reported that family history of alcoholism, rather than alcohol consumption, was the primary determinant of reduced amplitudes in alcoholic men, and supported the hypothesis that P3 amplitude reduction may be a biological marker of premorbid risk for alcoholism rather than a sequela of alcohol abuse. Furthermore, F H + alcoholics have been found to have reduced amplitudes on certain ERP measures when compared with FH - alcoholics (17,22), suggesting that genetic or familial factors may contribute to the degree of impairment seen in alcoholic populations. The ERP differences between resumers and abstainers in the present study, however, could not be attributed familial alcoholism effects. The percentage of FH + subjects was approximately equal in the resumers and abstainers (resumers were 69.407o F H + , abstainers were 60.4070 F H + , X2 = 0.765, p = ns). For the ERP variables, no significant differences were found between F H + and F H - subjects in either the control, abstainer, or resumer groups, providing no support
for familial differences as a primary factor in ERP impairment. It is possible, however, that the ERP differences between resumers and abstainers reflect premorbid differences in factors other than familial alcoholism, such as a greater premorbid vulnerability to maladaptive psychological and psychosocial factors among resumers or represent those who are perhaps more sensitive to the consequences of chronic alcohol abuse. The present study examined the extent to which addition of ERP variables could improve prediction of resumer/abstainer status from cognitive and psychosocial variables. Discriminant function analyses incorporating N2L with previously tested measures of depressive symptomatology, neuropsychological performance, psychosocial maladjustment, childhood attention deficit disorder symptoms, and previous treatment history did not improve predictive capacity over the model in which N2L was not included. Using only the three variables which contributed significantly to the regression equation (depressive symptomatology, psychosocial maladjustment, and N2L), a predictive rate of 71°70 was achieved, a rate comparable to the 5- or 6-variable models. These three variables are representative of three general lines of research: N2L (biological), depressive symptomatology (psychological), and psychosocial maladjustment (social). These findings suggest that the best model for understanding relapse prediction may be one which incorporates variables from all three aspects of a biopsychosocial model of alcoholism. ACKNOWLEDGEMENTS Preparation of this article was supported by the National Institute of Alcohol Abuse and Alcoholism Grant IROIAA06153 awarded to Oscar A. Parsons.
REFERENCES 1. Abbott, M. W.; Gregson, R. A. Cognitivedysfunctionin the prediction of relapse in alcoholics. J. Stud. Alcohol 42:230-243; 1981. 2. Alcohol and Health. Sixth special report to the U.S. Congress. Department of Health and Human Services. National Institute of Alcohol Abuse and Alcoholism. Rockville, MD; 1990. 3. Begleiter, H.; Porjesz, B.; Bihari, B.; Kissin, B. Event-related brain potentials in boys at risk for alcoholism. Science 225:14931496; 1984. 4. Caster, D. U.; Parsons, O. A. Locus of control in alcoholics and treatment outcome. J. Stud. Alcohol 38:2087-2095; 1977. 5. Ciesielski, K. T.; Madden, J. S.; Bligh, J. G.; Schopflocher, D. Long-term brain impairment in chronic alcoholics: N2-P3 cognitive potentials in a template-matching memory task. Alcohol and Alcoholism 20:403-408; 1985. 6. Diagnostic and Statistical Manual of Mental Disorders (3rd ed,. rev). Washington, DC: American Psychiatric Press; 1987. 7. Glenn, S. W.; Parsons, O. A. Prediction of resumption of drinking in posttreatment alcoholics. Internatl J. of the Addictions 26: 237-254; 1991. 8. Glenn, S. W.; Sinha, R.; Parsons, O. A. Assessment of recovery of electrophysiological and neuropsychological functions in chronic alcoholics. Submitted for publication. 9. Hatter, H. R.; Guido, W. Attention to pattern orientation: Negative cortical potentials, reaction time, and the selection process. Electroencephal. Clin. Neurophysiol. 49:461-475; 1980. 10. Johnson, R. A triarchic model of P300 amplitude. Psychophysiology 23:367-384; 1986. 11. Marlatt, G. A.; Gordon, J. R. Relapse prevention. New York: Guilford Press; 1985. 12. Naatanen, R.; Picton, T. The NI wave of the human electric and magnetic response to sound: A review and analysis of the component structure. Psychophysiology 24:375-425, 1987.
13. National Council on Alcoholism. Criteria for the diagnosis of alcoholism. Am. J. Psychiatry 129:2; 1972. 14. O'Leary, M. R.; Rohsenow, D. J.; Chancy, E. F. The use of multivariate personality strategies in predicting attrition from alcoholism treatment. J. Clin. Psychology 40:190-193; 1979. 15. Parsons, O. A.; Schaeffer, K. W.; Glenn, S. W. Does neuropsychological test performance predict resumption of drinking in posttrcatment alcoholics? Addictive Behaviors 15:297-307; 1990. 16. Parsons, O. A.; Sinha, R.; Williams, H. L. Relationships between neuropsychological test performance and event-related potentials in alcoholic and nonalcoholic samples. Ale: Ctin. Exp. Res. 14: 746-755; 1990. 17. Patterson, B. W.; Williams, H. L.; McLean, G. A.; Smith, L. T.; Schaeffer, K. W. Alcoholism and family history of alcoholism: Effects on visual and auditory event-related potentials. Alcohol 4:265-274; 1987. 18. Patterson, B. W.; Sinha, R.; Williams, H. L.; Parsons, O. A.; Smith, L. T.; Schaeffer, K. W. The relationship between neuropsychological and late component evoked potential measures in chronic alcoholics. Internatl. J. Nenroscience 49:319-327; 1989. 19. Penick, E. C.; Powell, B. J.; Bingham, S. F.; Liskow, B. I.; Miller, N. S.; Read, M. R. A comparative study of familial alcohofism. J. Stud. Alcohol 48:136-146; 1987. 20. Pfefferbaum, A.; Ford, J.; Mathalon, D. Event-related potentials in alcoholic men: P3 amplitude reflects family history but not alcohol consumption. Ale: Clin. Exp. Res. 15:839-850; 1991. 21. Porjesz, B.; Begleiter, H. Brain dysfunction and alcohol. In: Kissin, R.; Bagleiter, H., ed. The pathogenesis of alcoholism: Biological factors. New York: Plenum Press; 1983:415-483. 22. Porjesz, B.; Besleite~, H. Evoked brain potentials and alcoholism. In: Parsons, O. A.; Butters, N.; Nathan, P. E. ed. Nenropsy-
ELECTROPHYSIOLOGICAL INDICES OF RESUMPTION
23. 24.
25. 26.
chology of alcoholism: Implications for diagnosis and treatment. New York: Guilford Press; 1987:45-63. Pritchard, W. S. Psychophysiology of P300. Psych. Bull. 89:506540, 1981. Ritter, W.; Ford, J. M.; Gaillard, A. W. K.; Harter, M. R.; Kutas, M.; Naatanen, R.; Polich, J.; Renault, B.; Rohrbaugh, J. Cognition and event-related potentials I: The relation of negative potentials and cognitive processes. Ann. NY Acad. Sci. 425:2438; 1984. Shipley, W. C. A self-administering scale for measuring intellectual impairment and deterioration. J. Psychol. 9:371-377; 1940. Sinha, R.; Bernardy, N.; Parsons, O. A. Long term test-retest
95 reliability of event-related potentials in normals and alcoholics. Submitted for publication; 1992. 27. Skerchock, J. A.; Cohen, J. Alcoholism, organicity and eventrelated potentials. In: Karrer, R. ed. Brain and information: Event-related potentials. Ann. NY Acad. Sci. 425:623-628; 1984. 28. Whipple, S. C.; Parker, E. S.; Noble, E. P. An atypical neurocognitive profile in alcoholic fathers and their sons. J. Stud. Alcohol 49:240-244; 1988. 29. Williams, H. L. Evoked brain potentials and alcoholism: Questions, hypotheses, new approaches. In: Parsons, O. A.; Butters, N.; Nathan, P. E. ed. Neuropsychology of alcoholism: Implications for diagnosis and treatment. New York: Guilford Press; 1987:103-128.
0741-8329/93 $6.00 + .00 Copyright©1993PergamonPress Ltd.
Printed in the U.S.A. All rightsreserved.
Electrophysiological Indices Predict Resumption of Drinking in Sober Alcoholics SUSAN WAGNER
GLENN, l RAJITA SINHA* AND OSCAR A. PARSONS
Center f o r Alcohol and Drug Related Studies, University o f Oklahoma Health Sciences Center, Oklahoma City, OK 73104 *Department o f Psychiatry, Yale University School o f Medicine, New Haven, C T 06519 Received 2 A p r i l 1992; A c c e p t e d 4 S e p t e m b e r 1992 GLENN, S. W., R. SINHA AND O. A. PARSONS. Electrophysiologicalindicespredict resumption of drinking in sober alcoholics. ALCOHOL 10(2) 89-95, 1993. -Eighty-nine alcoholics and 54 nonalcoholic controls were tested on measures of late component event-related potentials (ERPs) using a visual "oddball" stimulus task. The alcoholics had just completed inpatient alcoholism treatment programs and were 21-45 days sober. Approximately 13 months later, subjects returned for retesting; alcoholics were classified as resumers or abstainers based on their drinking patterns during the intertest interval. Using the ERP measures from the initial testing session, alcoholics differed significantly from controls in the multivariate analysis and on P300 amplitude (P3A). Resumer alcoholics showed significantly longer N200 latencies (N2L) than abstainer alcoholics. Discriminant function analyses predicting resumer/abstainer status from N2L, P3A and NIA indicated a 63e/0 prediction rate, X2 = 5.67, p < 0.02. Addition of N2L to previously tested psychological and social predictor variables indicated an increase in the amount of variance explained. The results support a biopsychosocial model for understanding and predicting relapse in chronic alcoholics. Alcoholism
Evoked potentials
ERPs and alcoholism
Biopsychosocial model
Prediction of resumption
and marital status, history of multiple previous hospitalizations, depression, and childhood hyperactivity symptoms have all been implicated as predictor variables for resumption o f drinking (4,7,14,19). In addition, cognitive impairment has been shown to also have specific predictive capabilities. Alcoholics who resume drinking manifest neuropsychological deficits relative to those alcoholics who stay sober (1,15). These deficits are present at the time of initial treatment and testing and remain as long as 14 months (7,15). Note, then, that both psychological and social variables have successfully served as variables for predicting relapse. Because our understanding of the development and progression of alcoholism has been enhanced by the introduction of a biopsychosocial approach, the investigation of biological variables which may also predict relapse poses an interesting investigative question. It has been suggested that as an indicator of cerebral impairment, late components of event-related brain potentials (ERPs) may be even more sensitive measures than psychometric tests (5). Evoked potential measures have fairly consistently been found to be impaired in chronic alcoholics (e.g., 5,22,29) and have been correlated with performance deficits on neuropsychological tests (16). It would follow, then, that evoked potential responses, like neuropsychologieal performance, may also serve as predictors of resumption of drinking, and may provide additional information concerning the
A L C O H O L abuse is one of the nation's foremost health problems, as over 10.5 million adults suffer from alcoholism and millions o f others are directly or indirectly affected by its consequences (2). Treatment of alcoholism has only been partially successful, however, as normative outcome statistics indicate relapse rates as high as 80% over a I-year period (11). The identification of those persons most likely to relapse or resume drinking following treatment, is of considerable interest from several standpoints. First, if those persons most vulnerable to relapse could be identified prospectively, perhaps special treatment programs addressing specific relapse issues could be designed and implemented. Second, recognition of those prone to relapse may aid follow-up programs in providing more individual attention and in specifically targeting those most in need. Third, a priori identification of potential resumers based on differences between resumers and abstainers in cognitive, behavioral, or physiological characteristics will further our understanding of concurrent or premorbid factors which may contribute to the development or perpetuation of alcoholism. It is this third issue which we address in the present study. Previous research has indicated that certain psychological or psychosocial variables seem to characterize or help identify individuals who resume drinking following alcoholism treatment. Variables such as employment instability, residential
Requests for reprints should be addressed to Susan Wagner Glenn, Ph.D., Center for Alcohol and Drug Related Studies, University of Oklahoma Health Sciences Center, 800 NE 15th, Suite 410, Oklahoma City, OK 73104. 89
90
GLENN, SINHA AND PARSONS
characteristics of those individuals who are prone to relapse. These issues are addressed in the current study by answering the following questions: 1. Do ERPs differentiate alcoholics from control subjects? 2. Do ERPs measured following treatment predict resumer/ abstainer status one year later? 3. Do ERPs add to the prediction of resumer/abstainer status achieved by psychosocial variables and neuropsychological performance? METHOD
Subjects The alcoholic sample consisted of 53 men and 36 women recruited from community inpatient alcoholism treatment centers in and around the Oklahoma City metropolitan area. The control sample consisted of 23 men and 31 women recruited by newspaper and word of mouth. Alcoholic subjects met the National Council on Alcoholism (13) or the DSM-III-R (6) criteria for alcoholism and had been detoxified for 3 to 6 weeks prior to testing. Exclusion criteria for all subjects (alcoholics and controls) included a borderline intelligence or lower (as measured by the Shipley Vocabulary Scale; 25), a history of neurological disease or trauma, current psychoactive medications, history o f major psychiatric or medical disorder, or substantial history of abuse of drugs other than alcohol. Subjects were screened prior to testing for auditory acuity (those with hearing loss over 15 db were excluded) and color vision (subjects with red-green color blindness were excluded). Subjects gave informed consent and received monetary compensation for their participation.
Event-Related Potentials The ERP paradigm used has been described in detail in our earlier work (16) and is a modification of the paradigm employed by Porjesz and Begiefter (21). ERPs were recorded in an oddball dual-modality target selection paradigm in which auditory stimuli (1000 or 1500 I-Iz 50 ms tone pips) and visual stimuli (red or green 1 in circles) were presented in random order with ISis ranging from 0.5 to 1.5 s. Auditory stimuli were presented by earphones and visual stimuli were presented on a CRT screen 1 meter in front of the subject. There were 4 ERP runs with a minimum of 400 trials per run such that each of the 4 stimuli served as a "target" during one run. The sequence of the runs was randomly preselected for each subject. The total number of stimuli in each modality (visual or auditory) was approximately equal in each run. However, the target stimuli for each run was the oddball or rare stimulus in that it was presented 20% of the time for the stimuli of that modality, 80% of the stimuli in that particular modality were composed of nontarget stimuli. The subject's task was to count to himself the number of target stimuli (attended trials) while disregarding the nontarget, same modality stimuli as well as the nontarget, different modality stimuli (nonattended trials). After each 100 trials the subject was asked to provide the number of target stimuli which had been presented thus far. If the count given was it.correct, the subject was given the correct count and instructed to strive for accuracy and to continue counting from the corrected count. Although this procedure did not allow for the discrimination of ERPs for correctly identified versus incorrectly identified targets, most subjects made very few, if any, counting errors and thus the overwhelming majority of the target trials represented correctly identified targets.
Electrode sites used for recording included Cz, Pz, and Oz, F3 and F4. Reference electrodes were linked earlobes. The Grass Model 78-B polygraph was calibrated with a 20 Hz, 50 microvolt peak-to-peak input signal, which resulted in an output value of _+2.5 volts. Analog filters of the Grass 7P511E amplifiers were set at one-half down at 0.3 Hz and 100.0 Hz. A / D conversion was done with 1024 bit resolution of _+5 volt input at the board. Immediately before stimulus presentation, each electrode was sampled 250 times over a 500 msec interval to provide a baseline voltage for that particular trial. During this half-second period, if the sampled EEG exceeded 75% of the output of the amplifier (or 75 mv peak to peak), the trial was reset and the baseline sampling restarted. The same screening process was repeated for the poststimulus period in that EEG output exceeding 75 microvolts peak-topeak during the 500 msec following stimulus presentation resulted in the trial being rejected and repeated later in the experimental run. In this way, trials were kept free from artifacts secondary to eyeblinks, eye movements or gross physical movements, or high frequency alpha EEG activity which could mask ERP responses. The total number of rejected trials or the average experimental run time did not differ between alcoholic and control subjects. A peak detection algorithm was applied to the averaged trial vectors for each stimulus condition for electrodes Oz, Pz, and Cz to obtain automatic computation of latencies and amplitudes of the NI, N2, and P3 ERP components. The N l , N2, and P3 amplitudes reported (NIA, N2A, and P3A) are absolute voltages measured from the 1/2 s prestimulus baseline. Because preliminary analyses indicated that there were no significant differences in the amplitudes or latencies of these three components between the attended target trials of the two visual or auditory stimuli, the scores were combined over the two stimuli in each modality. Our preliminary analyses and subsequent studies (16,18) also indicated that the greatest differences between alcoholics and controls were found in the visual modality, particularly with the Pz electrode, therefore, only the averaged scores for the attended target trials of the visual modality at Pz are presented in the current study.
ERP Measures The amplitudes and latencies of the ERP measures used have been associated with various aspects of the cognitive process. The N1 wave (N100) can be elicited by stimuli delivered in various modalities and typically has a poststimulus latency of 120 msec in normal healthy adults. Given stimuli presented in an attended modality and a nonattended modality, the NI amplitude is increased by both target and nontarget stimuli in the attended modality and thus reflects a measure of modality selection. The Nd, or negative difference wave, represents the difference in amplitudes between the N 1 components of the target and nontarget stimuli and is considered a measure of selective attention and further processing of the target stimuli within the relevant modality (12,18,24). The N2 component of the ERP (N200) has a typical latency of 200-250 msec poststimulus and is reliably evoked by infrequent stimuli placed randomly in a stream of frequent stimuli. The N2 wave is thought to reflect a mismatch process associated with the orienting reflex (24). N2 has also been found to be related to stimulus complexity, and is sensitive to the nature of cognitive tasks (5,9). The P3 (P300) component is a large endogenous event with positive polarity which peaks at 300-500 msec poststimulus.
ELECTROPHYSIOLOGICAL INDICES OF RESUMPTION
It is often evoked in an oddball target selection task in which the subject is required to respond to infrequent target stimuli occurring in a series of nontarget frequent stimuli. The P3 amplitude is increased by stimuli that are novel, task relevant, and are somehow important to the subject or require the updating of working memory. For further discussion of these late component measures of the ERP, see Patterson et eLI.(17), Johnson (I0), and Pritchard (23).
91 TABLE 1 FOLLOW-UP DATA FOR ALCOHOLICS NOT RETESTED n
~'e
Follow-Up
13 3
(33.3°70) (7.7070)
6
(15.4o'/o)
4
00.3070)
3
(7.7We)
2
(5.107o)
6
(15.407o)
l
(2.5%)
1
(2.507e)
Lost immediately after initial testing session Sober for two months, then moved out of town or were put into jail Sober for three months, then moved, were put into jail, or resumed drinking and refused to participate with followup calls. Sober for four months, then moved or resumed drinking and refused to continue with followup, Sober 6-7 months, then moved or resumed drinking and refused to continue with followup. Sober 9-10 months, then moved or refused to continue with followup. Sober 12 months, but then moved or refused to return for followup testing. Sober for 12 months, then resumed heavy drinking and moved out of state. Light drinking in the first month, progressed to moderate drinking for the next 12 months, but refused to participate in followup testing.
Procedures The subjects in this study were part of a larger study investigating the neuropsychological, psychosociai, as well as electrophysiological consequences of alcohol abuse. Following completion of inpatient alcoholism treatment, alcoholic subjects were brought to the Center for a day of testing and interviews. Control subjects were also tested at the Center following an initial screening session. All subjects were given an Intoxilyzer test prior to beginning the testing session. Any person with breath alcohol readings over 0.01 was rescbeduled. Subjects participated in the electrophysiological tests in the morning, receiving the oddball counting task, as described previously, in addition to an auditory evoked brainstem response task and a visual pattern reversal response task. Including placement of electrodes, task set-up time, and short breaks between tasks, the electrophysiologicai procedures took approximately 3-3.5 h to complete. The subjects were then given a lunch break. The afternoon testing session consisted of a neuropsychological test battery and interviews concerning drinking patterns, affective symptomatology, childhood behaviorai disorders, and psychosociai variables. The results of the afternoon testing session have been reported elsewhere (7,15). All of the testing procedures were repeated approximately l year later (average for the three groups, 13.7 months), but the predictive measures reported in the current study are the results of only the initialtesting session. (The results of the E R P measures at the retesting session are reported elsewhere; 8.) During the intertest period, alcoholic subjects were contacted monthly by phone and were questioned concerning current drinking behaviors. Collaterals, or friends or relatives designated by the subject, were also contacted by experimenters on a regular basis and provided corroboration of the subject'sdrinking status in 9 6 % of the cases. If subjects or collaterals could not be reached during a particular month, a letter was sent to the subject requesting that they phone the research center and/or provide forwarding address information if a change of residence had taken place. O n the basis of these monthly contacts with subjects and the drinking histories obtained in the follow-up testing session, subjects who consumed at least I0 oz of ethanol (approximately 17 drinks) during the 6 months prior to retesting (approximately 8-14 months after initial testing) were classified as resumers (n = 36); the remainder were classified as abstainers (n = 53). Of the 126 alcoholic subjects receiving all of the test measures in the initial testing session, 89 were retested, yielding a retest or follow-up rate of 71%. Data identifying the time frame and various reasons for failure to complete the followup portion of the study are provided for the alcoholics who were not retested (See Table 1). Control subjects were not contacted on a monthly basis but were sent birthday and Christmas cards and were contacted by phone as the time for retest approached. Of the 76 control subjects tested initially, 54 returned for retesting, yielding a follow-up rate of 71%. Combining controls, abstainers and
resumers, 112 (67.9%) of the men and 90 (74.4%) of the women were retested (X2" = 1.05, p = ns). Demographic and neuropsychologicai performance differences between subjects who returned for retest and those who did not are discussed in Parsons et al. (15). The only differences in ERP measures between the two groups o f subjects were on Nd (for controls) and P3A (for alcoholics) (Table 2). RESULTS
The results of the demographic data are shown in Table 3. Controls, resumers, and abstainers did not differ in age, education, or intertest interval. Resumers and abstainers did not differ in years o f alcoholism or ounces of absolute ethanol consumed in the 6 months prior to treatment. By selection procedures, the two groups did differ in the amount of absolute ethanol consumed in the 6 months prior to retest. Significant differences were found in the gender distribution of resumers and abstainers. Twenty-eight alcoholic women (77.8%) were abstainers, 8 (22.2%) were resumers. Of the 53 men alcoholic subjects, 25 (47.1%) were abstainers and 28 (52.8%) were resumers 0c2 = 8.34, p < 0.01). A 3 × 2 multivariate analysis (MANOVA) was conducted initially to examine the main effects of group (controls, abstainers, and resumers), sex, and group × sex on the ERP dependent variables (N1A, Nd, N2A, P3A, N2L, and P3L). A significant main effect of group was found (Wilkes' Lambda F -- 2.00, p = 0.0248), but there was no significant main effect of sex or group × sex interaction. Given these results, univariate analyses were next conducted to examine the effect of group on each of the individual ERP measures. The results of these analyses can be seen in Table 4. Significant effects of group were found for N2L and P3A, with a trend toward significance for N I A . Post-hoc examination of the means using Duncan's Multiple Range Test indicated that the group effect was primarily due to differences between resumers versus abstainers and controls for N2L and to differences between controls versus resumers and abstainers for
92
GLENN, SINHA AND PARSONS TABLE 2 ERP VARIABLES (MEAN +_ SD) FOR SUBJECTS RETESTED VERSUS SUBJECTS NOT RETESTED
NIA Nd* N2A N2L P3At P3L
Controls Retested (n = 54)
Controls Not Retested (n = 22)
Alcohols Retested (n = 89)
Alcohols Not Retested (n = 37)
- 4.63 (2.45) 1.20 (2.06) - 0.14 (3.47) 277.33 (26.14) 10.86 (3.36) 394.35 (29.77)
- 5.08 (4.07) 2.47 (3.19) - 0.34 (3.49) 269.58 (24.60) 10.14 (4.03) 400.80 (28.57)
- 3.91 (2.07) 1.01 (1.75) 0.58 (2.89) 283.26 (26.38) 9.24 (3.27) 394.28 (30.33)
- 4.39 (3.38) 1.36 (2.99) 0.19 (3.42) 282.01 (29.56) 11.11 (3.79) 402.18 (31.49)
*For controls retested versus not retested, F(1, 75) = 4.06, p < 0.05. tFor alcoholics retested versus not retested, F(1,130) = 8.17, p < 0.01.
P3A. O n N1A, controls a n d a b s t a i n e r s were n o t significantly different f r o m each o t h e r a n d abstainers a n d resumers were not significantly different, b u t resumers were significantly different f r o m controls. In order to examine the efficacy o f the E R P variables in predicting r e s u m e r / a b s t a i n e r status, a d i s c r i m i n a n t f u n c t i o n analysis was c o n d u c t e d using t h e three E R P variables which h a d yielded significant or near-significant g r o u p effects (N2L, P 3 A , a n d N1A). Using this three v a r i a b l e model, 32 (61.5%)
o f abstainers a n d 21 (65.6*70) o f resumers were correctly identified for a n overall prediction rate o f 6 3 . 1 % (×2 = 5.67, p < 0.02). T h e t h i r d experimental question e x a m i n e d the possible increase in predictive capacity which m i g h t be gained by the inclusion o f E R P variables with neuropsychological a n d psychosocial variables which have been s h o w n to predict r e s u m e r / a b s t a i n e r status. In a previous study (7), we examined the role o f 5 variables in predicting r e s u m p t i o n o f drink-
TABLE 3 MEANS (+ SD) OF DEMOGRAPHIC AND DRINKING VARIABLES Controls (n = 54)
Abstainers (n = 53)
Resumers (n = 36)
Age
36.5 (9.9)
39.2 (9.8)
36.3 (8.0)
Years o f education
13.7 (2.0)
12.9 (1.6)
13. l (2. i)
13.9 (5.2)
12.9 (3.3)
14.3 (3.2)
0.41 (0.50)
0.98 (0.51)
I. 14 (0.71)
Interest interval (months) Drug index* Oz. Absolute Etohl" 6 months prior to test Oz. Absolute Etoh~ 6 months prior to retest Years Alcoholic
43.2 (64.2) 53.5 (101.2) -
1632.1 (1220.9) 0.27 (0.83)
1823.0 (1294.4) 598.8 (766.5)
11.0
11.7
(8.1)
(7.2)
*Drug Index is measure in which subjects were asked if they had ever tried any of 10 classes of drugs (e.g., amphetamines, barbiturates, halhicinogenics, etc). The number given represents the mean number of affirmative responses for subjects in each group. For the ANOVA, F(2, 113) = 18.38, p < 0.001; Resumers = Abstainers > Controls. tF(2, 113) = 50.92, p < 0.001; Resumers = Abstainers > Controls. ~ ( 2 , 113) = 29.13,p < 0.001; Resumers > Abstainers = Controls.
93
ELECTROPHYSIOLOGICAL INDICES OF RESUMPTION TABLE 4 CONTROLS, ABSTAINERS,AND RESUMERS ON ERP VARIABLES
NIA~ NdA N2A N2L P3A P3L
Controls
Abstainers
Resumers
(n = 54)
(n = 53)
(n = 36)
F*
-4.63 (2.47), 1.20 (2.08)° - 0.12 (3.50)d 277.89 (26.05)e 10.89 (3.37) 394.39 (30.02)s
-4.13 (1.91),.b 1.08 (1.63)c 0.61 (2.95)d 277.99 (25.05), 9.28 (3.47)f 392.35 (27.84)~
- 3.57 (2.28)b 0.89 (1.96)c 0.54 (2.82)d 291.03 (26.70) 9.17 (3.00)f 397.03 (33.74)g
2.42
.0926
0.28
n.s.
0.85
n.s.
p
3.45 .0344 4.56
.0120
0.27
n.s.
Means (+ SD) are presented. Means with the same subscript letters are not significantly different by Duncan's Multiple Range Test. *F's and p's represent ANOVA between the three groups (df = 1,142). tERP measures are from Pz electrode and are the average responses of the visual modality responses to the target stimuli. ing: depressive symptomatology, neuropsychological performance, psychosocial maladjustment, previous treatment history, and childhood attention deficit disorder symptomatology. The subjects used were the same as in the present study with the addition of several alcoholics who did not receive the ERP testing. Using the 5 mentioned variables in a discriminant function analysis, 79% of abstainers and 67%0 of resumers were correctly identified for an overall prediction rateof75%0 (X2 = 22.1,p < 0.001). In order to examine th~ utility of the inclusion of the present study's ERP variables into this predictive model, the N2L measure was chosen as the ERP measure which best differentiated resumers and abstainers. A discriminant function analysis was then conducted using the 5 psychosocial and neuropsychological variables described in the previous study and N2L. Seventy-four percent of abstainers and 69% of resumers were correctly identified using this 6 variable model for an overall predictive rate of 71.5% (X2 = 16.1, p < 0.001). Stepwise regression equations predicting resumer/abstainer status indicated an R" of .26 or that 26% o f the variance was explained using the original 5 variable model; the R: increased to .31 with the addition of N2L to the equation, indicating some degree o f shared variance among the variables. Further examination of the 6-variable equation revealed that the depressive symptomatology, psychosocial maladjustment and N2L were the significant contributors to the equation, F s (3, 86) = 4.6 to 13.8, p's = 0.034 to 0.001, together contributing .25 to the total R 2 of .31. Using only these three significant variables, a final discriminant function analysis was conducted. Seventytwo percent o f abstainers and 69% of resumers could be identified usin[g this 3-variable model for an overall prediction rate o f 7 1 % , x = 14.7,p < 0.001. DISCUSSION In answer to the first experimental question, ERPs successfully differentiated alcoholics from controls as seen in the significant group main effect and in the P3A and N2L measures. Visual examination of the ERP variables illustrates that controls generally showed greater amplitudes and shorter latencies than their alcoholic counterparts, even on measures
which did not show statistical significance. These results replicate those of previous studies which have examined ERP changes in chronic alcoholics (5,22,29) and provide further support for findings of central nervous system changes as a result of chronic alcohol ingestion. The second experimental question concerned the predictive capacity of evoked potential responses in accurately determining resumer/abstainer status. Using three of the ERP measures, N2L, P3A, and N1A, 61.5%0 of abstainers and 65.6%0 of resumers could be correctly identified. These findings are significant, considering that the ERP measures were recorded an average of 13 months before the resumer/abstainer outcome classification was obtained. In previous studies, ERP measures have been found to be indices sensitive to various effects of alcohol such as alcoholization (acute changes), tolerance, withdrawal, and long-term brain dysfunction (22). With the findings of the present study, predictive capabilities may be added to the list of potential uses and indications for ERPs and may suggest new applications for these measures in future studies. The a priori identification of resumers and abstainers based on ERP measures provides support for the hypothesis that those individuals who are likely to relapse can be characterized by variables indicative of greater nervous system dysfunction in addition to psychosocial and neuropsychological variables. Impaired ERPs have been found to be related to greater cognitive deficits (18,27) and are thought to reflect deficits in the integrity of the information processing system (29). In the present study, N2L was the ERP measure which most clearly differentiated resumers from abstainers. Porjesz et al. (22) specifically investigated the N2 component of ERPs in an alcoholic population. Using an easy/difficult line discrimination task, the authors found delayed N2 latencies in alcoholics relative to controls; the differences were particularly apparent for the easy discrimination part of the task. The authors reported that the alcohol-related deficit in N2L reflected an impairment in (a) the time taken for stimulus evaluation processes and (b) the ability of alcoholics to adequately differentiate between easy and difficult tasks, suggesting that alcoholics adopt an undifferentiated mode of responding regardless of task requirements. Generalizing from these laboratory
94
GLENN, SINHA AND PARSONS
findings to the complexities involved in relapse is difficult. However, it is possible that the N2L deficit in resumers compared to abstainers reflects a relatively greater impairment in the ability to evaluate and differentiate between stimuli, and, even more speculatively, a concomitant inability of resumers to learn to adopt new modes of responding required of new situational tasks (i.e., not returning to drinking after release from treatment). Obviously, replication of the N2L findings in resumers versus abstainers needs to be achieved before the above speculation can be seriously considered. Impairment in ERPs has been implicated as a marker for the development of alcoholism, as the ERP changes described for chronic alcoholics have been found in the young sons of alcoholic fathers (3,28) as well as in nonalcoholic adults with a family history of alcoholism ( F H + ) (17). Pfefferbaum et al. (20) reported that family history of alcoholism, rather than alcohol consumption, was the primary determinant of reduced amplitudes in alcoholic men, and supported the hypothesis that P3 amplitude reduction may be a biological marker of premorbid risk for alcoholism rather than a sequela of alcohol abuse. Furthermore, F H + alcoholics have been found to have reduced amplitudes on certain ERP measures when compared with FH - alcoholics (17,22), suggesting that genetic or familial factors may contribute to the degree of impairment seen in alcoholic populations. The ERP differences between resumers and abstainers in the present study, however, could not be attributed familial alcoholism effects. The percentage of FH + subjects was approximately equal in the resumers and abstainers (resumers were 69.407o F H + , abstainers were 60.4070 F H + , X2 = 0.765, p = ns). For the ERP variables, no significant differences were found between F H + and F H - subjects in either the control, abstainer, or resumer groups, providing no support
for familial differences as a primary factor in ERP impairment. It is possible, however, that the ERP differences between resumers and abstainers reflect premorbid differences in factors other than familial alcoholism, such as a greater premorbid vulnerability to maladaptive psychological and psychosocial factors among resumers or represent those who are perhaps more sensitive to the consequences of chronic alcohol abuse. The present study examined the extent to which addition of ERP variables could improve prediction of resumer/abstainer status from cognitive and psychosocial variables. Discriminant function analyses incorporating N2L with previously tested measures of depressive symptomatology, neuropsychological performance, psychosocial maladjustment, childhood attention deficit disorder symptoms, and previous treatment history did not improve predictive capacity over the model in which N2L was not included. Using only the three variables which contributed significantly to the regression equation (depressive symptomatology, psychosocial maladjustment, and N2L), a predictive rate of 71°70 was achieved, a rate comparable to the 5- or 6-variable models. These three variables are representative of three general lines of research: N2L (biological), depressive symptomatology (psychological), and psychosocial maladjustment (social). These findings suggest that the best model for understanding relapse prediction may be one which incorporates variables from all three aspects of a biopsychosocial model of alcoholism. ACKNOWLEDGEMENTS Preparation of this article was supported by the National Institute of Alcohol Abuse and Alcoholism Grant IROIAA06153 awarded to Oscar A. Parsons.
REFERENCES 1. Abbott, M. W.; Gregson, R. A. Cognitivedysfunctionin the prediction of relapse in alcoholics. J. Stud. Alcohol 42:230-243; 1981. 2. Alcohol and Health. Sixth special report to the U.S. Congress. Department of Health and Human Services. National Institute of Alcohol Abuse and Alcoholism. Rockville, MD; 1990. 3. Begleiter, H.; Porjesz, B.; Bihari, B.; Kissin, B. Event-related brain potentials in boys at risk for alcoholism. Science 225:14931496; 1984. 4. Caster, D. U.; Parsons, O. A. Locus of control in alcoholics and treatment outcome. J. Stud. Alcohol 38:2087-2095; 1977. 5. Ciesielski, K. T.; Madden, J. S.; Bligh, J. G.; Schopflocher, D. Long-term brain impairment in chronic alcoholics: N2-P3 cognitive potentials in a template-matching memory task. Alcohol and Alcoholism 20:403-408; 1985. 6. Diagnostic and Statistical Manual of Mental Disorders (3rd ed,. rev). Washington, DC: American Psychiatric Press; 1987. 7. Glenn, S. W.; Parsons, O. A. Prediction of resumption of drinking in posttreatment alcoholics. Internatl J. of the Addictions 26: 237-254; 1991. 8. Glenn, S. W.; Sinha, R.; Parsons, O. A. Assessment of recovery of electrophysiological and neuropsychological functions in chronic alcoholics. Submitted for publication. 9. Hatter, H. R.; Guido, W. Attention to pattern orientation: Negative cortical potentials, reaction time, and the selection process. Electroencephal. Clin. Neurophysiol. 49:461-475; 1980. 10. Johnson, R. A triarchic model of P300 amplitude. Psychophysiology 23:367-384; 1986. 11. Marlatt, G. A.; Gordon, J. R. Relapse prevention. New York: Guilford Press; 1985. 12. Naatanen, R.; Picton, T. The NI wave of the human electric and magnetic response to sound: A review and analysis of the component structure. Psychophysiology 24:375-425, 1987.
13. National Council on Alcoholism. Criteria for the diagnosis of alcoholism. Am. J. Psychiatry 129:2; 1972. 14. O'Leary, M. R.; Rohsenow, D. J.; Chancy, E. F. The use of multivariate personality strategies in predicting attrition from alcoholism treatment. J. Clin. Psychology 40:190-193; 1979. 15. Parsons, O. A.; Schaeffer, K. W.; Glenn, S. W. Does neuropsychological test performance predict resumption of drinking in posttrcatment alcoholics? Addictive Behaviors 15:297-307; 1990. 16. Parsons, O. A.; Sinha, R.; Williams, H. L. Relationships between neuropsychological test performance and event-related potentials in alcoholic and nonalcoholic samples. Ale: Ctin. Exp. Res. 14: 746-755; 1990. 17. Patterson, B. W.; Williams, H. L.; McLean, G. A.; Smith, L. T.; Schaeffer, K. W. Alcoholism and family history of alcoholism: Effects on visual and auditory event-related potentials. Alcohol 4:265-274; 1987. 18. Patterson, B. W.; Sinha, R.; Williams, H. L.; Parsons, O. A.; Smith, L. T.; Schaeffer, K. W. The relationship between neuropsychological and late component evoked potential measures in chronic alcoholics. Internatl. J. Nenroscience 49:319-327; 1989. 19. Penick, E. C.; Powell, B. J.; Bingham, S. F.; Liskow, B. I.; Miller, N. S.; Read, M. R. A comparative study of familial alcohofism. J. Stud. Alcohol 48:136-146; 1987. 20. Pfefferbaum, A.; Ford, J.; Mathalon, D. Event-related potentials in alcoholic men: P3 amplitude reflects family history but not alcohol consumption. Ale: Clin. Exp. Res. 15:839-850; 1991. 21. Porjesz, B.; Begleiter, H. Brain dysfunction and alcohol. In: Kissin, R.; Bagleiter, H., ed. The pathogenesis of alcoholism: Biological factors. New York: Plenum Press; 1983:415-483. 22. Porjesz, B.; Besleite~, H. Evoked brain potentials and alcoholism. In: Parsons, O. A.; Butters, N.; Nathan, P. E. ed. Nenropsy-
ELECTROPHYSIOLOGICAL INDICES OF RESUMPTION
23. 24.
25. 26.
chology of alcoholism: Implications for diagnosis and treatment. New York: Guilford Press; 1987:45-63. Pritchard, W. S. Psychophysiology of P300. Psych. Bull. 89:506540, 1981. Ritter, W.; Ford, J. M.; Gaillard, A. W. K.; Harter, M. R.; Kutas, M.; Naatanen, R.; Polich, J.; Renault, B.; Rohrbaugh, J. Cognition and event-related potentials I: The relation of negative potentials and cognitive processes. Ann. NY Acad. Sci. 425:2438; 1984. Shipley, W. C. A self-administering scale for measuring intellectual impairment and deterioration. J. Psychol. 9:371-377; 1940. Sinha, R.; Bernardy, N.; Parsons, O. A. Long term test-retest
95 reliability of event-related potentials in normals and alcoholics. Submitted for publication; 1992. 27. Skerchock, J. A.; Cohen, J. Alcoholism, organicity and eventrelated potentials. In: Karrer, R. ed. Brain and information: Event-related potentials. Ann. NY Acad. Sci. 425:623-628; 1984. 28. Whipple, S. C.; Parker, E. S.; Noble, E. P. An atypical neurocognitive profile in alcoholic fathers and their sons. J. Stud. Alcohol 49:240-244; 1988. 29. Williams, H. L. Evoked brain potentials and alcoholism: Questions, hypotheses, new approaches. In: Parsons, O. A.; Butters, N.; Nathan, P. E. ed. Neuropsychology of alcoholism: Implications for diagnosis and treatment. New York: Guilford Press; 1987:103-128.