COGNITIVE PERFORMANCE DEFICIT REGRESSED ON ALCOHOL DOSAGE

Transcription

1 COGNITIVE PERFORMANCE DEFICIT REGRESSED ON ALCOHOL DOSAGE Robert S. Kennedy, Robert L. Wilkes, Essex Corporation, Orlando, Florida, U.S.A.; and Gene G. Rugotzke, Wyoming Public Health Laboratory, Cheyenne, Wyoming, U.S.A. Summary. The purpose of this experiment to was relate cognitive performance deficits to circulating blood alcohol levels. Blood alcohol concentration (BAC) was carefully monitored in 18 subjects across four BAC treatments (0.00, 0.05, 0.10, & 0.15). The cognitive performance tests were administered when subjects were at the target BAC. Additionally, field sobriety measures were obtained for comparison. Results indicated that performance decrements were proportional to BAC on eight of the nine cognitive tests employed and on all of the field sobriety measures. It was also determined that the cognitive tests add variance to the prediction of BAC that is unique from variance predicted by the field sobriety measures. Although motor skills such as braking and steering are important elements in driving, behaviorally complex elements including perception, cognition, information processing, and decision making are also critically involved. More information in the scientific literature is available for the former than the latter, perhaps because of the difficulties associated with assessment of the more complex mental functions. For the past 10 years we have been engaged in the development of a multi-purpose, microbased battery of cognitive tests. Presently, a menu of 30 tests is available which possess what we consider to be the requisite metric properties of stability and reliability. It was our hypothesis that performance deficits on tests in this battery may be correlated with circulating blood alcohol levels. The secondary purpose of this experiment was to seek additional validation of the U.S. Department of Transportation recommended field sobriety test (National Highway Traffic Safety Administration, 1984). METHODS Subjects. Eighteen male subjects, 21 years of age or older, attending Casper College, Casper, Wyoming served as subjects. Acceptable candidates were those indicating some, but not excessive, experience with alcohol, no past history of chronic dependency of any type, good general health, and indications of low risk for future alcohol-based problems. Experimental Design. Subjects were randomly assigned to a series of three BAC treatments and one placebo condition over four separate testing sessions. Each subject was tested for performance decrements at 0.00, and approximately 0.05, 0.10, and 0.15 BACs. Order effects were controlled through the serial use of Latin Square randomization techniques (Edwards, 1985, pp ). Each subject served as his own control with performance measures completed both prior to and after the treatment. Blood alcohol concentrations were closely monitored by breath testing procedures until prescribed treatment levels were attained. Corresponding whole blood, blood sera, and urine measures were then obtained. Double-blind procedures were employed across all testing sessions to control for experimenter and subject expectancy effects. 354

2 Blood alcohol concentration was manipulated by administering alcoholic drinks mixed from orange, tomato or fruit juice and grain alcohol (95% alcohol) with 2.5 ml drops of rum extract floated on top. Drinks were premixed for each subject with weight and BAC treatment condition determining the proportions of alcohol. Proportions of grain alcohol and juice were combined to raise a subject BAC slightly above the target level permitting monitoring on the descending limb of the BAC curve. The amount of grain alcohol in milliliters was calculated using a condensed version of the Widmark Equation: ml grain alcohol = (200/190) (30) (subject weight in pounds) (target BAC ). If the target was 0.00, then no grain alcohol was used in the placebo preparation. Performance was assessed with both the Automated Performance Test System (APTS) and standard law enforcement field sobriety tests. The APTS battery was composed of nine subtests, previously evaluated relative to repeated-measures selection criteria. The nine subtests are listed in Figure 1 and the repeated-measures selection criteria are fully discussed elsewhere (Kennedy, Wilkes, Lane, & Homick, 1985). Collectively, the nine subtests have been demonstrated to identify four separate factors including: motor speed, symbol manipulation/reasoning, cognitive processing speed, and speed of response selection. Three common procedures for roadside testing of intoxication were also employed. The procedure included the Gaze Nystagmus test (GN), Walk-and-Turn test (W&T), and the One-Leg Stand test (OLS). The testing and scoring procedures have been standardized and are described in more detail elsewhere (National Highway Traffic Safety Administration, 1984). Apparatus. Testing with the APTS was conducted with eight NEC PC8201A microcomputers. RESULTS Statistical analyses were performed using measures from three distinct sources: the Automated Performance Test Battery, biochemical data collection, and performance on field sobriety tests. The test battery variables included number correct, average response latency, and percent correct scores for each test administered, but only number correct is reported here. Biochemical measures consisted of whole blood, blood serum, urine, and two breath measures (initial and final), all recorded at each blood alcohol level treatment administered (i.e., from 0.0 to 0.15 BAC). The interrelationships of these measures are reported elsewhere (Rugotzke, Wilkes & Kennedy, 1989), and in this report, whole blood was employed as the sole criterion. Field sobriety test scores were obtained for three tests: Gaze Nystagmus, One-Leg Stand, and Walk and Turn. Automated Performance Test System (APTS) Measures. Figure 1 shows the mean performances during the time when the alcohol dosage was at the four prescribed levels for the nine performance tests. Graphically, it may be seen that there are generally monotonic relationships between mean performance and BAC with the greatest change occurring for the.15 BAC level. In order to maximize statistical power, the performance data obtained for the four experimental trials were analyzed in a repeated measures ANOVA framework. Each subject was considered to have received exposures to four treatment levels (.00,.05,.10, and.15 BAC). To enhance interpretability each test was analyzed separately 355

3 yielding a total of nine analyses for APTS. Two problems attend such a strategy: 1) the individual blood alcohol concentrations may contain additional predictive power and 2) multiple comparisons do not provide protection for the type I error rate; that is, by testing the same sample over and over we increase the chance of finding a difference where none exists. Analyses to account for the first problem are covered below. To offset the increase in type I bias, we selected a higher than usual alpha level for acceptance of significance, in this case a cut-off of.001 was adopted. The resulting probability values for the APTS tests are shown underneath each curve in Figure 1. Only one test did not show a significant decrement from placebo at our selected alpha level - Grammatical Reasoning. The remaining APTS tests were significant in excess of P <.001 level. Next, we sought a method of analysis which would elucidate whether there was additional strength in individual levels of measured alcohol concentration x magnitude of the performance deficit. To this end, EACH subject's score for EACH test was regressed against (correlated with) HIS alcohol concentration (whole blood) value measured at the time of his performance. Thus, an individual predictive validity (in the form of a Pearson product moment correlation coefficient) was obtained for all tests and for each subject. In order to average these individual predictive relationships, the obtained correlations were converted to Fisher's Z, and a group average was calculated. The obtained value then was returned to a Pearson correlation and it was considered that this average correlation over all the subjects would index the strength of the predictive relationship between the two measures. Table 1 shows correlations of the nine tests rank ordered by strength of relationship. It may be seen that performances on Code Substitution, Manikin, the two Tapping tests and Reaction Time followed most closely the increasing blood alcohol levels. This ordering is consistent with the ANOVA where Grammatical Reasoning was not shown to reveal a statistically significant decrement with alcohol dosage. Based on this table, the average predictive correlations of the APTS tests were entered into a multiple regression equation for purposes of determining the best combination of tests from this battery with respect to the predicted criterion of blood alcohol level over the four administrations. Only the Code Substitution test was entered into the prediction equation, none of the other tests providing additional unique variance. The correlation adjusted for shrinkage was r =.52; p <.01. FIELD SOBRIETY MEASURES Descriptives of field sobriety scores were reviewed for the Gaze Nystagmus, One-Leg Stand, Walk and Turn tests and two combinatorial scores (National Highway Traffic Safety Administration, 1984). "Combo I" is the simple sum of walk and stand scores and "Combo II" add gaze scores to the sum. All measures exhibited a monotonic relationship to the BAC levels (P <.001). As with the performance tests, correlations between each subject's field sobriety score and his blood alcohol level were calculated. As with the APTS test scores, high correlations would imply a strong predictive relationship between the blood alcohol level and the performance deficit. It may be seen in Table 1 that all measures show positive correlations and the gaze measure shows the highest 356

4 predictive relationship for blood concentration levels. with the steeper slope of this measure. This is consistent As with the APTS scores, the average correlations between field sobriety measures and blood alcohol levels, as well as the intercorrelations between them, were cast into a multiple regression analysis. This analysis provided a multiple R which was equal to the correlation of the Gaze measures alone (R.80; p <.0001); the other variables not adding unique variance at a statistically significant (p =.076) level. Combinations of APTS and Field Sobriety Measures. We attempted to improve predictive power by administering several tests from two different batteries. Both of these batteries have been developed to tap human behavior capabilities which can be expected to be involved in the driving task. Therefore, as a test of this notion we sought to determine whether the addition of test(s) from the two batteries could be combined into composite scores to improve predictive power. We found a multiple R =.86, which adjusted for shrinkage accounts for 71% of the variance - a hefty value. The two tests combined were significantly (P =.02) better than each alone. DISCUSSION It is our view that mental tests can provide an indication of the onset, duration, and severity of impairment in operational performance which may be due to environmental hazards or toxic chemicals. The advent of microcomputers can expand the potential for assessment over paper-and pencil media by permitting more rapid, diverse, and accurate assessment of capabilities. When taken singly, eight of the nine tests produced significant differences in connection with the disparate blood alcohol levels, and functional relationships were essentially monotonic. A multiple regression analysis suggested that most of the tests were behaving similarly; using only two or three tests can capture most of the variance attributable to alcohol that was explained by the longer battery. The greatest reduction in performance occurred between.10 and.15 BAC, and the relatively abrupt nature of this change implies that sharp cut-offs in cognitive performance occur at that point, and future studies should focus on this breakpoint and explore its functional shape since it has important implications for agencies with regulatory responsibilities. Not surprisingly, this breakpoint coincides with the legal limit on driving while intoxicated (DWI) or driving under the influence (DUI) used in most of the United States. The present study did not surface any evidence which suggests that this is an inappropriate break point, at least from a measurement standpoint. The case for the field sobriety tests was clearcut. All tests revealed monotonic relationships to the blood alcohol levels, and the size of the decrements were proportional to the concentrations. When compared to the APTS tests the Field Sobriety measures were shown to provide considerable common variance but there were predictive relationships in each which were not shared. From these findings we infer that improved predictions of the adverse effects of alcohol on driving behavior are likely to be obtained by combining the two batteries, at least the best score from each battery. We would suggest that the Field Sobriety tests, the accepted BEHAVIORAL measures of DUI, are sampling the output of brainstem motor control centers whereas the tests of the APTS 357

5 Table 1. Average Correlation (Within Subjects) for APTS and Field Sobriety Measures with Blood Alcohol Levels APTS Correlation Code Substitution Manikin Reaction Time.626 Preferred Tapping Sternberg Nonpreferred Tapping Math Processing Pattern Comparison Grammatical Reasoning Field Sobriety Gaze.795 Stand.712 Walk

6 battery would appear to be more cortically dependent. Therefore, the combination of both elements would appear to converge better on the types of behaviors which are known to be disrupted by alcohol consumption. REFERENCES Edwards, A. L. (1985). York: Harper & Row. Experimental design in psychological research. New Kennedy, R. S., Wilkes, R. L., Lane, N. E., & Homick, J. L. (1985). Preliminary evaluation of a microbased repeated-measures testing system (Tech. Rep. No. EOTR 85-1). Orlando, FL: Essex Corporation. National Highway Traffic Safety Administration. (1984). testing. Washington, DC: Author. Improved sobriety Rugotzke, G. G., Wilkes, R. L., & Kennedy, R. S. (1989). Reliability and validity of blood alcohol concentrations related to measured performance decrement. Paper presented at the 11th International Conference on Alcohol, Drugs, and Traffic Safety, Chicago, IL. FIGURE CAPTION Figure 1. Effects of three graded dosages of alcohol compared to placebo for the microcomputer tests. ucns Iff s I h t t a i t l c a l Pracan ln g K > P rtiirn Ciaptrlacm CS i Coil h k t l t i t l c n Q = C r u w t l c t l fetacnlng Id = lunlkln IDI = Short f a n Hamm BB = PrtfitT»! Hind Itp jln j KPI : NoirFrefnTai Hull lipping II = Jour Ctalcs lu c tlo n I l n 359