Addictive Behaviors, Vol. 15, pp. 297-307, 1990
0306-4603/90 $3.00 + .130 Copyright © 1990 Pergamon Press plc
Printed in the USA. All rights reserved.
DOES NEUROPSYCHOLOGICAL TEST PERFORMANCE PREDICT RESUMPTION OF DRINKING IN POSTTREATMENT ALCOHOLICS? OSCAR A. PARSONS, KIM W. SCHAEFFER, and SUSAN W. GLENN Universityof Oklahoma Health Sciences Center Abstract -- The predictionof resumptionof drinking in posttreatmentalcoholics was investigated as a function of five possible confoundingvariables: depression, anxiety, childhood symptoms of attentiondeficit and conductdisorders and family history of alcoholism. Male and female detoxified alcoholics (n = 103) in inpatienttreatmentprograms were administereda neuropsychologicalbattery and retested as outpatients 14 monthslater; peer nonalcoholics(n = 73), giventhe same battery, had a similar intertest interval. Alcoholics who resumed drinking (N = 41) performed significantly poorer on an overall neuropsychologicalperformanceindex than abstainers(N = 62) who performed significantlypoorer than nonalcoholics.Stepwise multipleregression equationsusing the variables noted above revealed that depressive symptoms, ADD and the performance index were the only variablesto enter the prediction(R2 = .26, p < .001); depressionaccountedfor mostof the variance. At retest all three groups improved significantly,but not differentially,and were as significantly different at retest as at initial testing. Implicationsof these results are discussed.
The neuropsychological performance of alcoholic patients who are in inpatient alcoholism treatment programs is impaired relative to nonalcoholic samples (Bergman, 1987; Grant, 1987; Parsons, 1987). After treatment, some alcoholics remain abstinent (abstainers) while others resume drinking (resumers). The identification of the factors or variables involved in alcoholics' relapse is one of the major problems in current research on alcoholism (Marlatt & Gordon, 1985; Nathan & Skinstad, 1987). A relevant research question is whether neuropsychological performance at the time of treatment can predict posttreatment drinking status. There have been mixed answers to this research question. Some researchers have found that alcoholic patients who were tested during treatment and who later resumed drinking performed poorer than comparably tested alcoholic patients who did not resume drinking (Abbott & Gregson, 1981; Fabian & Parsons, 1983; Gregson & Taylor, 1977; Yohman, Parsons, & Leber, 1985). However, other researchers have found that neuropsychological performance measured at the time of treatment did not reliably predict posttreatment drinking status (Donovan, Kivilahan, & Walker, 1984; Eckardt et al., 1988). There are many possible reasons for .such discrepancies. Variations among samples of alcoholics in subject characteristics such as depression, anxiety, family history of alcoholism and childhood histories of hyperkinesis and attention deficit disorders, if not controlled or investigated, could lead to inconsistent results. The possible interfering effects of anxiety and depression upon cognitive functioning have long been noted in the clinical literature. Both of the history variables have been found in our laboratories to be related to neuropsychological deficits in alcoholics (DeObaldia, Parsons, & Yohman, 1983; Schaeffer, Parsons, & Errico, 1988). Accordingly, our first aim addressed the question of whether alcoholics' neuropsychological test performance, measured near the end of treatment, predicts posttreatment resumption of This research was supported, in part, by the National Institute on Alcohol Abuse and Alcoholism Grant 1R01AA06135 to Oscar A. Parsons. We appreciate the help of Mark Everett, M.A., Rajita Sinha, M.S., Larry Stevens, Ph.D., and Janice Turner, M.S., in various phases of the project. Requests for reprints should be sent to Oscar A. Parsons, Ph.D., Oklahoma Center for Alcohol and Drug Related Studies, Rogers Building, Suite 410, 800 N.E. 15th St., Oklahoma City, OK 73104. 297
298
OSCARA. PARSONS,KIM W. SCHAEFFER,and SUSANW. GLENN
drinking and whether this relationship is affected by taking into account measures of affective symptomatology, self-reported childhood behavioral disorders and family history of alcoholism. The second aim was to investigate whether abstainers improve their neuropsychological performance relative to resumers after treatment or to nonalcoholic controls. Presumably, if alcohol abuse leads to neuropsychological deficits, abstinence should result in improved neuropsychological functioning. In support of such reasoning, a number of test-retest studies have concluded that abstinent alcoholics over time do recover some cognitive functioning (Parsons & Leber, 1981; Schau, O'Leary, & Chaney, 1980). However, many of these studies have not included a retested nonalcoholic group, to control for practice effects, or have not compared abstinent alcoholics with those who resumed drinking. In our laboratories we have conducted two such studies. Fabian and Parsons (1983) found that retested (after 1.8 years) groups of female resumer and abstainer alcoholics and nonalcoholic controls all improved significantly in neuropsychological test performance but did not differ in their improvement scores. Yohman et al. (1985) conducted a similar experiment with alcoholic men using a somewhat different neuropsychological battery. At retest (13 months later), resumers, abstainers and nonalcoholic controls did not differ in amount of improvement. In both studies alcoholics performed significantly poorer than their controls on both initial testing and retest; also, resumers tended to have lower levels of performance than abstainers at both testings. These results, if cross-validated, raise some interesting questions as to the causal relationship between alcohol ingestion and neuropsychological functioning and the time course for recovery of cognitive functions in abstinent alcoholics. Such a cross-validation was attempted in the present study with some methodological improvements: the number of subjects was increased and both male and female alcoholics and their nonalcoholic controls were administered the same neuropsychological test battery. METHOD
Subjects Phase 1. Alcoholics and nonalcoholics were studied in a test-retest design in which the same battery of tests, inventories and interviews was administered on both occasions. Initial testing (Phase 1) was done with alcoholics while they were completing or had just completed an inpatient treatment program at one of ten treatment units in the Oklahoma City area. Participation was strictly voluntary; those who expressed interest in the presentations or flyers distributed by the project coordinator and research assistants composed the alcoholic sample. All alcoholic subjects were abstinent from alcohol for at least 21 days but not more than six weeks before being tested. Nonalcoholic peers were recruited by word of mouth and by classified ads in local newspapers. Subjects were not accepted into the study if they had any of the following: a Shipley Institute of Living (Shipley, 1940) vocabulary age less than 12.7 (borderline intelligence or lower); any condition that could affect central nervous system function (other than alcoholism for the alcoholics); a major psychiatric or medical disorder; unconsciousness for more than one hour; or an uncorrected defect in vision or hearing. All alcoholic subjects met the National Council on Alcoholism (1972) criteria for the diagnosis of alcoholism. Conversely, all nonalcoholics had never met such criteria. All subjects gave informed consent and were paid for participation. Phase 1 testing was completed for 76 male alcoholics, 48 male nonalcoholics, 67 female alcoholics, and 48 female nonalcoholics. Subjects were given a detailed interview in which their history of consumption of alcohol and other drugs was noted in addition to other demographic information. Data relating to family history of alcoholism were also obtained in this interview. Subjects were asked if any
Predicting drinking resumption
299
Table 1. Means ( -+ SD) of relevant variables Resumer Male n
Abstainer
Female
Male
Nonalcoholic
Female
Male
Female
28
13
30
32
29
44
Age
37.3 ---8.1
36.1 _+10.1
37.7 -+- 10.1
38.7 -+9.5
36.4 ±9.2
36.1 ±10.4
Education
13.3 -+2.3
12.9 -+1.8
13.3 -+1.5
13.0 -+1.8
14.2 -+2.0
13.2 -+1.7
Intertest Interval (Months)
14.0 ---3.4
15. I ---4.3
12.4 -+3.4
13.2 -+2.8
13.3 -+2.1
14.4 ±5.7
Years of Alcoholism
10.7 -+ 8.2
8.9 ± 4.8
12.1 ± 8.4
10.6 _+7.3
1837.1 ± 1282.2
1979.7 ± 1261.6
1754.2 ± 1119.8
692.4 -+823.4
370.4 -+361.9
.2 -+ .5
Oz of Absolute Alcohol (6 months prior to test) Oz of Absolute Alcohol (6 months prior to retest)
---
---
1503.3 ± 1347.4
65.7 ±75.8
17.5 ±27.1
.4 -+ 1.0
81.6 -+ 127.4
17.6 -+32.7
primary or secondary relative had been diagnosed or treated for alcoholism or manifested behaviors associated with alcohol abuse, for example, arrests for driving under the influence, loss of job or divorce due to alcohol abuse, and so on. However, the rating of family history positive (FH+) for alcoholism was assigned only to those alcoholics who had at least one primary family relative (mother, father, sister, or brother) who was an alcoholic or alcoholic abuser. Contacts with collaterals verified these ratings in 95% of the subjects. Approximately half of the subjects in each group were (FH+); the remaining subjects were family history negative (FH - ). Phase 2. Twelve to 16 months after the initial testing, as many of the subjects who would cooperate were retested (Phase 2). Detailed records as to their drinking behavior were obtained on a monthly telephone contact basis between test and retest and at the retest interview. There were no abstinence requirements for the period before retesting but all subjects were required to pass a breathalyzer test (BAC of less than 0.02) on the retesting day. The several subjects who had BACs above 0.02 were rescheduled for testing. Alcoholic subjects were classified as "resumers" if their absolute ethanol intake exceeded 10 oz (approximately 17 drinks) over the six months prior to their retest; all other alcoholic subjects were classified as "abstainers." We chose this time period for two reasons. First, we could make a direct comparison of drinking intake information gathered at initial testing (also covering the preceding 6 months). Second, we found that a sizable minority of our alcoholics resumed drinking for several months after treatment but then stopped and were sober for the last six months before retesting. Contact with patient's collaterals verified whether or not the patient had been drinking in the past 6 months in 96% of the cases. There were a total of 176 subjects retested in Phase 2:41 were resumer alcoholics (28 males and 13 females), 62 were abstainers (30 males and 32 females) and 73 were nonalcoholics (29 males and 44 females). Subjects comprising the attrition sample were not retested for several reasons: of the 40 alcoholics who were not retested, 52.5% (21) had
300
OSCAR A. PARSONS, KIM W. SCHAEFFER. and SUSAN W. GLENN
moved or dropped out of contact with family, friends or collaterals and could not be located; 15% (6) were incarcerated at the time of retest; 27.5% (11) refused to participate in the monthly follow-ups or the retesting session; 2.5% (1) moved out of state and could not afford to come back for testing; and 2.5% (1) reported being too ill to return for retest. Of the controls who were not retested, 82.5% (19) had moved away; 8.7% (2) were unable to locate, left no forwarding information, and 8.7% (2) refused to participate in the retest session. Seventy-two percent of the alcoholics were retested compared to 76% of the controls; 77% of the females were retested compared to 70% of the males and 74% of the F H + compared to 72% of the FH - . None of the chi-squares for these comparisons approached significance. Similar analyses were conducted considering the percentages of subjects in each of the eight cells (e.g., alcoholic, female, F H + , nonalcoholic, male, F H - ) ; again there were no significant chi-squares. Finally, there were no significant differences in the distribution of F H + and F H - alcoholics in the Resumer versus Abstainer groups. The resumer group had 68% F H + and 32% F H - , the abstainer group had 61% F H + and 39% F H - respectively (X2(1) = .526, p = .47). The mean age and educational levels of the groups at retest are presented in Table 1. Group × Gender analyses of variance (ANOVAs) for age and education revealed no significant effects. Also presented in Table 1 are the mean intertest intervals, ounces of absolute alcohol drunk in the six months before Phase 1 testing and ounces of absolute alcohol drunk in the six months before Phase 2 retesting. Alcoholic groups did not differ significantly on years of alcoholism [F(I, 98) = .94, p = .33]. The 3 groups did not differ on intertest interval IF(2, 170) = 2.47, p = .09]. As would be expected, both resumers and abstainers differed from nonalcoholics on total ounces of ethanol consumed in the six months before initial testing, but the two alcoholic groups did not differ from each other by Duncan's tests (p > .05). At the Phase 2 retesting, resumers (by definition) had drunk significantly higher total ounces of alcohol over the previous 6 months than abstainers and nonalcoholics [/7(2, 170) = 30.09, p = .0001]; the latter two groups did not differ from each other by Duncan's tests (p > .05). Tests and Procedure At both Phase 1 and Phase 2 testings, subjects were administered the tests in a three-hour session with a 15-min break in the middle of the testing session. The order of the tests was randomized for each subject. The tests were administered according to standardized instructions by trained research assistants. The neuropsychological tests were grouped into the following test clusters: verbal, learning and memory, problem-solving, perceptual-motor, and an overall performance index was calculated. The verbal cluster consisted of Shipley Institute of Living verbal age (Shipley, 1940) and the Information and Comprehension subtests of the Wechsler Adult Intelligence Scale -- Revised (Wechsler, 1981). The learning and memory cluster was comprised of the logical and figural portions of the Wechsler memory scale (Russell, 1975), the Symbol-Digit Paired Associates Test (Ryan & Butters, 1980), and the Face-Name Learning Test (Schaeffer & Parsons, 1987). The problem-solving cluster consisted of the Booklet Category Test (DeFillipis & McCampbell, 1979), the Levine Hypothesis Test (Levine, 1966), the Conceptual Level Analogy Test (Willner, 1970), the Word Finding Test (Reitan, 1972), the Shipley Institute of Living Abstraction Age (Shipley, 1940), and the Block Design subtests of the WAIS-R (Wechsler, 1981). The perceptual-motor cluster consisted of the Digit-Symbol subtest of the WAIS-R, the Grooved Pegboard (Dominant and Nondominant hands) and Trail Making test A and B of the Rennick Repeatable Battery (Rennick, Russell. Kempler, & Schwartz, 1972). The performance index was comprised of
Predicting drinking resumption
301
the 11 tests on which the alcoholics performed significantly poorer (p < .05) than the controls. These included all of the tests mentioned above except for the tests in the verbal cluster, the nondominant hand of the Grooved Pegboard, Trail Making Test A, the Block Design subtest of the WA1S-R, and the Levine Hypothesis test. Cluster scores were derived by first obtaining t scores (M = 50 ± 10) for each subject on each test based on the distributions of the Phase 1 raw scores for the combined groups. The t scores of the tests within a cluster were averaged to obtain a cluster score and an overall performance index for each subject. Mean cluster scores were then obtained for each group. In addition to the neuropsychological tests, the Beck Depression Inventory (BDI) (Beck, Ward, Mendelson, Mock, & Erbaugh, 1961), the State Anxiety Inventory (SAI) (Spielberger, Gorsuch, & Lushene, 1970), and the Hyperkinesis/Minimal Brain Damage (Hk/ MBD) questionnaire (Tarter, McBride, Buonpane, & Schneider, 1977) were administered at Phase 1 and, with the exception of the Hk/MBD, at Phase 2. The Hk/MBD questionnaire is a 50-item list of behaviors reported to be present in children who are classifiable as having attention deficit disorders, and/or conduct disorder. Subjects are asked to check " y e s , " " n o , " or "unsure" for behaviors that were present before the age of 13. Alcoholics report significantly more of these behaviors than controls and their reports have satisfactory reliability and validity (DeObaldia & Parsons, 1984). Alcoholics who score higher on the Hk/MBD scale were found to have lower scores on the abstracting subscale of the Shipley (DeObaldia et al., 1983). RESULTS Before attempting to answer the research questions, a preliminary analysis of the two phases of the study was undertaken to determine whether the retested subjects differed from those who were not retested on any of the initial testing demographic or experimental variables. A groups (alcoholics vs. nonalcoholics) x gender x FH x phase/subjects (subjects retested vs. those who were not) ANOVA was performed on each of the independent and dependent variables. The only significant differences due to the main effect of phase/subjects were on age IF(l, 224) = 17.58, p = .0001], education [F(1, 224) = 7.01, p = .001], and verbal cluster scores [F(1, 224) = 5.38, p = .02]. Further, none of the group x phase/subjects, gender x phase/subjects and FH x phase/subjects interactions were significant. Thus, any effects that are present due to groups, gender or FH, would not be differentially affected by subject attrition. With respect, then, to the major independent experimental variables, the retested subjects appear to be representative of the total sample. The only differences obtained on the initial testing demographic and experimental variables indicate that our retested subjects are older, better educated and have better verbal skills than those subjects who were not retested. Turning to our aims, we first investigated which of the variables measured at initial testing (Table 2) were related to posttreatment drinking. A group (Resumers, Abstainers, Nonalcoholics) x gender analysis of variance was performed. (In these analyses we omitted the FH variable because including it caused some of the cell ns to drop below acceptable levels; it is included in subsequent multiple regression analyses.) The group main effect was significant for the following response variables: verbal cluster F(2, 170) = 7.25, p < .001; learning and memory cluster, F(2, 170) = 8.06, p < .001; perceptual-motor cluster F(2, 170) = 6.22, p < .002; problem-solving cluster, F = 2, 170) = 7.23, p < .001; performance index, F(2, 170) = 9.84, p < .001; BDI, F(2, 170) = 40.27, p < .001; SAI, F(2, 170) = 7.80, p < .001 ; and the Hk/MBD questionnaire, F(2, 170) = 33.16, p < .001. None of the group x gender interactions were significant in any of the analyses. Duncan's Multiple Range Tests for significant group differences (alpha level o f p ~ .05)
49.5 (7.55)
52.4 (7,78)
Abstainers (N = 62)
Nonalcoholics (N = 72)
54.4 (7.18)
51,1 (7.94)
47.8 (7.95)
52.7 (6.87)
49.3 (7.54)
46.3 (7.62)
54.9 (6.35)
52.2 (6.4)
49.5 (5.26)
52.1 (6.30)
49.8 (7.78)
46.7 (8.77)
56.9 (6.09)
52.7 (8.73)
50.6 (7.74)
Phase 2
Prob-Solv
Phase 2 Phase I Phase 2 Phase I
L &M
52.3 (5.95)
49.6 t7.23)
46.4 (8.69)
Phase I
54.3 (6.31)
50.3 (8.17)
47.3 (6.48)
Phase 2
Percept-Motor
52.6 (5,73)
49.5 (6.75)
46. I (7.04)
Phase I
55.5 (5.53)
51.9 (6.70)
49.2 (5.35)
Phase 2
Performance
3,13 (3.26)
5.75 (5.12)
11,5 (7.61)
Phase I
3.86 (5.24)
4.77 (5.59)
12.9 (9.87)
46.5 (7.09)
49.1 (8.94)
54.4 (12.3)
Phase 2 Phase I
Beck
46,3 (7,42)
47.7 (8.95)
57.5 (12.7)
Phase 2
AI
'~Column means connected by the same vertical line are not significantly different. All other column means differ significantly (Duncan's tests p
0306-4603/90 $3.00 + .130 Copyright © 1990 Pergamon Press plc
Printed in the USA. All rights reserved.
DOES NEUROPSYCHOLOGICAL TEST PERFORMANCE PREDICT RESUMPTION OF DRINKING IN POSTTREATMENT ALCOHOLICS? OSCAR A. PARSONS, KIM W. SCHAEFFER, and SUSAN W. GLENN Universityof Oklahoma Health Sciences Center Abstract -- The predictionof resumptionof drinking in posttreatmentalcoholics was investigated as a function of five possible confoundingvariables: depression, anxiety, childhood symptoms of attentiondeficit and conductdisorders and family history of alcoholism. Male and female detoxified alcoholics (n = 103) in inpatienttreatmentprograms were administereda neuropsychologicalbattery and retested as outpatients 14 monthslater; peer nonalcoholics(n = 73), giventhe same battery, had a similar intertest interval. Alcoholics who resumed drinking (N = 41) performed significantly poorer on an overall neuropsychologicalperformanceindex than abstainers(N = 62) who performed significantlypoorer than nonalcoholics.Stepwise multipleregression equationsusing the variables noted above revealed that depressive symptoms, ADD and the performance index were the only variablesto enter the prediction(R2 = .26, p < .001); depressionaccountedfor mostof the variance. At retest all three groups improved significantly,but not differentially,and were as significantly different at retest as at initial testing. Implicationsof these results are discussed.
The neuropsychological performance of alcoholic patients who are in inpatient alcoholism treatment programs is impaired relative to nonalcoholic samples (Bergman, 1987; Grant, 1987; Parsons, 1987). After treatment, some alcoholics remain abstinent (abstainers) while others resume drinking (resumers). The identification of the factors or variables involved in alcoholics' relapse is one of the major problems in current research on alcoholism (Marlatt & Gordon, 1985; Nathan & Skinstad, 1987). A relevant research question is whether neuropsychological performance at the time of treatment can predict posttreatment drinking status. There have been mixed answers to this research question. Some researchers have found that alcoholic patients who were tested during treatment and who later resumed drinking performed poorer than comparably tested alcoholic patients who did not resume drinking (Abbott & Gregson, 1981; Fabian & Parsons, 1983; Gregson & Taylor, 1977; Yohman, Parsons, & Leber, 1985). However, other researchers have found that neuropsychological performance measured at the time of treatment did not reliably predict posttreatment drinking status (Donovan, Kivilahan, & Walker, 1984; Eckardt et al., 1988). There are many possible reasons for .such discrepancies. Variations among samples of alcoholics in subject characteristics such as depression, anxiety, family history of alcoholism and childhood histories of hyperkinesis and attention deficit disorders, if not controlled or investigated, could lead to inconsistent results. The possible interfering effects of anxiety and depression upon cognitive functioning have long been noted in the clinical literature. Both of the history variables have been found in our laboratories to be related to neuropsychological deficits in alcoholics (DeObaldia, Parsons, & Yohman, 1983; Schaeffer, Parsons, & Errico, 1988). Accordingly, our first aim addressed the question of whether alcoholics' neuropsychological test performance, measured near the end of treatment, predicts posttreatment resumption of This research was supported, in part, by the National Institute on Alcohol Abuse and Alcoholism Grant 1R01AA06135 to Oscar A. Parsons. We appreciate the help of Mark Everett, M.A., Rajita Sinha, M.S., Larry Stevens, Ph.D., and Janice Turner, M.S., in various phases of the project. Requests for reprints should be sent to Oscar A. Parsons, Ph.D., Oklahoma Center for Alcohol and Drug Related Studies, Rogers Building, Suite 410, 800 N.E. 15th St., Oklahoma City, OK 73104. 297
298
OSCARA. PARSONS,KIM W. SCHAEFFER,and SUSANW. GLENN
drinking and whether this relationship is affected by taking into account measures of affective symptomatology, self-reported childhood behavioral disorders and family history of alcoholism. The second aim was to investigate whether abstainers improve their neuropsychological performance relative to resumers after treatment or to nonalcoholic controls. Presumably, if alcohol abuse leads to neuropsychological deficits, abstinence should result in improved neuropsychological functioning. In support of such reasoning, a number of test-retest studies have concluded that abstinent alcoholics over time do recover some cognitive functioning (Parsons & Leber, 1981; Schau, O'Leary, & Chaney, 1980). However, many of these studies have not included a retested nonalcoholic group, to control for practice effects, or have not compared abstinent alcoholics with those who resumed drinking. In our laboratories we have conducted two such studies. Fabian and Parsons (1983) found that retested (after 1.8 years) groups of female resumer and abstainer alcoholics and nonalcoholic controls all improved significantly in neuropsychological test performance but did not differ in their improvement scores. Yohman et al. (1985) conducted a similar experiment with alcoholic men using a somewhat different neuropsychological battery. At retest (13 months later), resumers, abstainers and nonalcoholic controls did not differ in amount of improvement. In both studies alcoholics performed significantly poorer than their controls on both initial testing and retest; also, resumers tended to have lower levels of performance than abstainers at both testings. These results, if cross-validated, raise some interesting questions as to the causal relationship between alcohol ingestion and neuropsychological functioning and the time course for recovery of cognitive functions in abstinent alcoholics. Such a cross-validation was attempted in the present study with some methodological improvements: the number of subjects was increased and both male and female alcoholics and their nonalcoholic controls were administered the same neuropsychological test battery. METHOD
Subjects Phase 1. Alcoholics and nonalcoholics were studied in a test-retest design in which the same battery of tests, inventories and interviews was administered on both occasions. Initial testing (Phase 1) was done with alcoholics while they were completing or had just completed an inpatient treatment program at one of ten treatment units in the Oklahoma City area. Participation was strictly voluntary; those who expressed interest in the presentations or flyers distributed by the project coordinator and research assistants composed the alcoholic sample. All alcoholic subjects were abstinent from alcohol for at least 21 days but not more than six weeks before being tested. Nonalcoholic peers were recruited by word of mouth and by classified ads in local newspapers. Subjects were not accepted into the study if they had any of the following: a Shipley Institute of Living (Shipley, 1940) vocabulary age less than 12.7 (borderline intelligence or lower); any condition that could affect central nervous system function (other than alcoholism for the alcoholics); a major psychiatric or medical disorder; unconsciousness for more than one hour; or an uncorrected defect in vision or hearing. All alcoholic subjects met the National Council on Alcoholism (1972) criteria for the diagnosis of alcoholism. Conversely, all nonalcoholics had never met such criteria. All subjects gave informed consent and were paid for participation. Phase 1 testing was completed for 76 male alcoholics, 48 male nonalcoholics, 67 female alcoholics, and 48 female nonalcoholics. Subjects were given a detailed interview in which their history of consumption of alcohol and other drugs was noted in addition to other demographic information. Data relating to family history of alcoholism were also obtained in this interview. Subjects were asked if any
Predicting drinking resumption
299
Table 1. Means ( -+ SD) of relevant variables Resumer Male n
Abstainer
Female
Male
Nonalcoholic
Female
Male
Female
28
13
30
32
29
44
Age
37.3 ---8.1
36.1 _+10.1
37.7 -+- 10.1
38.7 -+9.5
36.4 ±9.2
36.1 ±10.4
Education
13.3 -+2.3
12.9 -+1.8
13.3 -+1.5
13.0 -+1.8
14.2 -+2.0
13.2 -+1.7
Intertest Interval (Months)
14.0 ---3.4
15. I ---4.3
12.4 -+3.4
13.2 -+2.8
13.3 -+2.1
14.4 ±5.7
Years of Alcoholism
10.7 -+ 8.2
8.9 ± 4.8
12.1 ± 8.4
10.6 _+7.3
1837.1 ± 1282.2
1979.7 ± 1261.6
1754.2 ± 1119.8
692.4 -+823.4
370.4 -+361.9
.2 -+ .5
Oz of Absolute Alcohol (6 months prior to test) Oz of Absolute Alcohol (6 months prior to retest)
---
---
1503.3 ± 1347.4
65.7 ±75.8
17.5 ±27.1
.4 -+ 1.0
81.6 -+ 127.4
17.6 -+32.7
primary or secondary relative had been diagnosed or treated for alcoholism or manifested behaviors associated with alcohol abuse, for example, arrests for driving under the influence, loss of job or divorce due to alcohol abuse, and so on. However, the rating of family history positive (FH+) for alcoholism was assigned only to those alcoholics who had at least one primary family relative (mother, father, sister, or brother) who was an alcoholic or alcoholic abuser. Contacts with collaterals verified these ratings in 95% of the subjects. Approximately half of the subjects in each group were (FH+); the remaining subjects were family history negative (FH - ). Phase 2. Twelve to 16 months after the initial testing, as many of the subjects who would cooperate were retested (Phase 2). Detailed records as to their drinking behavior were obtained on a monthly telephone contact basis between test and retest and at the retest interview. There were no abstinence requirements for the period before retesting but all subjects were required to pass a breathalyzer test (BAC of less than 0.02) on the retesting day. The several subjects who had BACs above 0.02 were rescheduled for testing. Alcoholic subjects were classified as "resumers" if their absolute ethanol intake exceeded 10 oz (approximately 17 drinks) over the six months prior to their retest; all other alcoholic subjects were classified as "abstainers." We chose this time period for two reasons. First, we could make a direct comparison of drinking intake information gathered at initial testing (also covering the preceding 6 months). Second, we found that a sizable minority of our alcoholics resumed drinking for several months after treatment but then stopped and were sober for the last six months before retesting. Contact with patient's collaterals verified whether or not the patient had been drinking in the past 6 months in 96% of the cases. There were a total of 176 subjects retested in Phase 2:41 were resumer alcoholics (28 males and 13 females), 62 were abstainers (30 males and 32 females) and 73 were nonalcoholics (29 males and 44 females). Subjects comprising the attrition sample were not retested for several reasons: of the 40 alcoholics who were not retested, 52.5% (21) had
300
OSCAR A. PARSONS, KIM W. SCHAEFFER. and SUSAN W. GLENN
moved or dropped out of contact with family, friends or collaterals and could not be located; 15% (6) were incarcerated at the time of retest; 27.5% (11) refused to participate in the monthly follow-ups or the retesting session; 2.5% (1) moved out of state and could not afford to come back for testing; and 2.5% (1) reported being too ill to return for retest. Of the controls who were not retested, 82.5% (19) had moved away; 8.7% (2) were unable to locate, left no forwarding information, and 8.7% (2) refused to participate in the retest session. Seventy-two percent of the alcoholics were retested compared to 76% of the controls; 77% of the females were retested compared to 70% of the males and 74% of the F H + compared to 72% of the FH - . None of the chi-squares for these comparisons approached significance. Similar analyses were conducted considering the percentages of subjects in each of the eight cells (e.g., alcoholic, female, F H + , nonalcoholic, male, F H - ) ; again there were no significant chi-squares. Finally, there were no significant differences in the distribution of F H + and F H - alcoholics in the Resumer versus Abstainer groups. The resumer group had 68% F H + and 32% F H - , the abstainer group had 61% F H + and 39% F H - respectively (X2(1) = .526, p = .47). The mean age and educational levels of the groups at retest are presented in Table 1. Group × Gender analyses of variance (ANOVAs) for age and education revealed no significant effects. Also presented in Table 1 are the mean intertest intervals, ounces of absolute alcohol drunk in the six months before Phase 1 testing and ounces of absolute alcohol drunk in the six months before Phase 2 retesting. Alcoholic groups did not differ significantly on years of alcoholism [F(I, 98) = .94, p = .33]. The 3 groups did not differ on intertest interval IF(2, 170) = 2.47, p = .09]. As would be expected, both resumers and abstainers differed from nonalcoholics on total ounces of ethanol consumed in the six months before initial testing, but the two alcoholic groups did not differ from each other by Duncan's tests (p > .05). At the Phase 2 retesting, resumers (by definition) had drunk significantly higher total ounces of alcohol over the previous 6 months than abstainers and nonalcoholics [/7(2, 170) = 30.09, p = .0001]; the latter two groups did not differ from each other by Duncan's tests (p > .05). Tests and Procedure At both Phase 1 and Phase 2 testings, subjects were administered the tests in a three-hour session with a 15-min break in the middle of the testing session. The order of the tests was randomized for each subject. The tests were administered according to standardized instructions by trained research assistants. The neuropsychological tests were grouped into the following test clusters: verbal, learning and memory, problem-solving, perceptual-motor, and an overall performance index was calculated. The verbal cluster consisted of Shipley Institute of Living verbal age (Shipley, 1940) and the Information and Comprehension subtests of the Wechsler Adult Intelligence Scale -- Revised (Wechsler, 1981). The learning and memory cluster was comprised of the logical and figural portions of the Wechsler memory scale (Russell, 1975), the Symbol-Digit Paired Associates Test (Ryan & Butters, 1980), and the Face-Name Learning Test (Schaeffer & Parsons, 1987). The problem-solving cluster consisted of the Booklet Category Test (DeFillipis & McCampbell, 1979), the Levine Hypothesis Test (Levine, 1966), the Conceptual Level Analogy Test (Willner, 1970), the Word Finding Test (Reitan, 1972), the Shipley Institute of Living Abstraction Age (Shipley, 1940), and the Block Design subtests of the WAIS-R (Wechsler, 1981). The perceptual-motor cluster consisted of the Digit-Symbol subtest of the WAIS-R, the Grooved Pegboard (Dominant and Nondominant hands) and Trail Making test A and B of the Rennick Repeatable Battery (Rennick, Russell. Kempler, & Schwartz, 1972). The performance index was comprised of
Predicting drinking resumption
301
the 11 tests on which the alcoholics performed significantly poorer (p < .05) than the controls. These included all of the tests mentioned above except for the tests in the verbal cluster, the nondominant hand of the Grooved Pegboard, Trail Making Test A, the Block Design subtest of the WA1S-R, and the Levine Hypothesis test. Cluster scores were derived by first obtaining t scores (M = 50 ± 10) for each subject on each test based on the distributions of the Phase 1 raw scores for the combined groups. The t scores of the tests within a cluster were averaged to obtain a cluster score and an overall performance index for each subject. Mean cluster scores were then obtained for each group. In addition to the neuropsychological tests, the Beck Depression Inventory (BDI) (Beck, Ward, Mendelson, Mock, & Erbaugh, 1961), the State Anxiety Inventory (SAI) (Spielberger, Gorsuch, & Lushene, 1970), and the Hyperkinesis/Minimal Brain Damage (Hk/ MBD) questionnaire (Tarter, McBride, Buonpane, & Schneider, 1977) were administered at Phase 1 and, with the exception of the Hk/MBD, at Phase 2. The Hk/MBD questionnaire is a 50-item list of behaviors reported to be present in children who are classifiable as having attention deficit disorders, and/or conduct disorder. Subjects are asked to check " y e s , " " n o , " or "unsure" for behaviors that were present before the age of 13. Alcoholics report significantly more of these behaviors than controls and their reports have satisfactory reliability and validity (DeObaldia & Parsons, 1984). Alcoholics who score higher on the Hk/MBD scale were found to have lower scores on the abstracting subscale of the Shipley (DeObaldia et al., 1983). RESULTS Before attempting to answer the research questions, a preliminary analysis of the two phases of the study was undertaken to determine whether the retested subjects differed from those who were not retested on any of the initial testing demographic or experimental variables. A groups (alcoholics vs. nonalcoholics) x gender x FH x phase/subjects (subjects retested vs. those who were not) ANOVA was performed on each of the independent and dependent variables. The only significant differences due to the main effect of phase/subjects were on age IF(l, 224) = 17.58, p = .0001], education [F(1, 224) = 7.01, p = .001], and verbal cluster scores [F(1, 224) = 5.38, p = .02]. Further, none of the group x phase/subjects, gender x phase/subjects and FH x phase/subjects interactions were significant. Thus, any effects that are present due to groups, gender or FH, would not be differentially affected by subject attrition. With respect, then, to the major independent experimental variables, the retested subjects appear to be representative of the total sample. The only differences obtained on the initial testing demographic and experimental variables indicate that our retested subjects are older, better educated and have better verbal skills than those subjects who were not retested. Turning to our aims, we first investigated which of the variables measured at initial testing (Table 2) were related to posttreatment drinking. A group (Resumers, Abstainers, Nonalcoholics) x gender analysis of variance was performed. (In these analyses we omitted the FH variable because including it caused some of the cell ns to drop below acceptable levels; it is included in subsequent multiple regression analyses.) The group main effect was significant for the following response variables: verbal cluster F(2, 170) = 7.25, p < .001; learning and memory cluster, F(2, 170) = 8.06, p < .001; perceptual-motor cluster F(2, 170) = 6.22, p < .002; problem-solving cluster, F = 2, 170) = 7.23, p < .001; performance index, F(2, 170) = 9.84, p < .001; BDI, F(2, 170) = 40.27, p < .001; SAI, F(2, 170) = 7.80, p < .001 ; and the Hk/MBD questionnaire, F(2, 170) = 33.16, p < .001. None of the group x gender interactions were significant in any of the analyses. Duncan's Multiple Range Tests for significant group differences (alpha level o f p ~ .05)
49.5 (7.55)
52.4 (7,78)
Abstainers (N = 62)
Nonalcoholics (N = 72)
54.4 (7.18)
51,1 (7.94)
47.8 (7.95)
52.7 (6.87)
49.3 (7.54)
46.3 (7.62)
54.9 (6.35)
52.2 (6.4)
49.5 (5.26)
52.1 (6.30)
49.8 (7.78)
46.7 (8.77)
56.9 (6.09)
52.7 (8.73)
50.6 (7.74)
Phase 2
Prob-Solv
Phase 2 Phase I Phase 2 Phase I
L &M
52.3 (5.95)
49.6 t7.23)
46.4 (8.69)
Phase I
54.3 (6.31)
50.3 (8.17)
47.3 (6.48)
Phase 2
Percept-Motor
52.6 (5,73)
49.5 (6.75)
46. I (7.04)
Phase I
55.5 (5.53)
51.9 (6.70)
49.2 (5.35)
Phase 2
Performance
3,13 (3.26)
5.75 (5.12)
11,5 (7.61)
Phase I
3.86 (5.24)
4.77 (5.59)
12.9 (9.87)
46.5 (7.09)
49.1 (8.94)
54.4 (12.3)
Phase 2 Phase I
Beck
46,3 (7,42)
47.7 (8.95)
57.5 (12.7)
Phase 2
AI
'~Column means connected by the same vertical line are not significantly different. All other column means differ significantly (Duncan's tests p
