"Review of Recent Literature " Othmer, Kaiser and Othmer

EEG Biofeedback Training for Attention Deficit Disorder:
A Review of Recent Controlled Studies and Clinical Findings
Siegfried Othmer, Ph.D., David Kaiser, Ph.D., and Susan F. Othmer, B.A.
June, 1995

Introduction and Summary
One of the fastest-growing applications of biofeedback at the present time is the use of EEG biofeedback for the remediation of attention deficits ( Attention Deficit Hyperactivity Disorder, ADHD), related behavioral disorders, and specific learning disabilities. This is happening largely on the basis of the continuing work of Joel and Judith Lubar, of Michael Tansey, and of a growing core of clinicians. The knowledge base is expanding primarily through increasing clinical use of the technique, rather than through controlled research. The lack of contemporary, large, suitably controlled studies has, however, inhibited acceptance within the larger psychological, psychiatric, and educational communities.
Efficacy of the EEG training for ADHD has been ascribed variously to remediation of the underlying condition of physiological underarousal manifesting in hyperactivity (Lubar, 1976); to addressing the motor component of hyperactivity by changing the set-point of the motor system with training at sensorimotor cortex [the same mechanism proposed for motor seizures (Sterman, 1980)], or by training the supplementary motor area responsible for the initiation of movement (Tansey, 1990); and to remediation of disregulation of arousal manifesting variously in inattention or behavioral disinhibition (Othmer, 1994). The validity of one of these mechanisms does not rule out validity for another. Indeed, they may each be responsible for addressing some aspect of ADHD symptomatology.
In the following, some recent studies will be briefly reviewed, and a statistical analysis of data coming out of current clinical practice will be presented. The clinical findings leave the matter of efficacy for ADHD beyond any reasonable doubt. The results are so robust that they cannot be attributed to a placebo effect, or other nonspecific effect of the training. However, the controlled studies to date have been much more ambiguous in their outcomes. Some possible explanations of these differences are presented.

Return to Introduction * Recent Findings * Clinical Findings * Results * Discussion

Recent Findings of Controlled Studies
Several studies have surfaced recently which meet criteria for controlled studies, and which therefore are beginning to fill the vacuum. Michael Linden has performed two studies which have both documented statistically significant shifts in IQ scores, along with favorable behavioral changes, using training protocols derived from Lubar. In his first study, which involved a small group of 9 experimental subjects and 9 controls, Linden found an increase in I.Q. scores of 10 points, with a statistical significance of p < .05, using the Kaufman Brief Intelligence Test, or K-BIT (Kaufman, 1990) (Linden, in press). He also found an improvement in parental assessment of inattention, likewise with p < .05. Hyperactivity improved to where it was below the abnormal rating; however, the change was not statistically significant. Assessments used the SNAP rating scale (Swanson, 1981) and the IOWA-Conners rating scale (Atkins, 1987). Training protocol was reinforcement of 16-20 Hz, with concurrent inhibition of excess 4-7 Hz activity. Training extended over 40 sessions. The findings were replicated in a second study, which is currently in preparation for publication.
Henry A. Cartozzo reported on his thesis work at the annual meeting of the Association for Applied Psychophysiology and Biofeedback (Cartozzo, 1995), which involved a small controlled study of EEG Biofeedback for ADHD using a protocol derived from Lubar (augmentation of 12-15 Hz activity with inhibition of 4-7 Hz and 22-30 Hz, with placement at Cz). Training extended over thirty sessions. Using 8 subjects and 7 controls, Cartozzo found significant improvements in the subtests of the WISC-R intelligence test which are most closely identified with ADHD, namely Arithmetic, Digit Span, and Coding. Together with Information, which was not tracked, the above WISC-R subtests constitute the famous "ACID" test of ADHD.
The WISC-R results obtained in the Cartozzo study are shown in Table 1, with comparative data from previous studies (Tansey, 1990; Othmer, 1991). The WISC-R changes were found to be significant at the level of p < .01, whereas the control group showed no significant changes. The results also fall in line with the previous studies. How may we interpret these results? The three subtests are associated with a factor called "Freedom from Distractibility". And they are the ones most sensitive to difficulties in sequential processing. Hence, improvement in these scores can be interpreted in terms of improvement in the continuity of mental processing, in working memory, and in the ability to sustain attentional focus. These improvements would, in turn, be observed as a diminution of distractibility. A minor criticism may be advanced: The latest data suffer from the fact that the two groups were not matched in starting subtest scores. The control group had lower mean scores than the experimental group.

Figure 1. Improvements in WISC-R subtest score for three independent studies (Cartozzo, 1995; Othmer, 1991; Tansey, 1990) employing similar EEG training protocols.

The Cartozzo study also found improvements in scores on a computerized continuous performance test, the T.O.V.A. (Test of Variables of Attention) (Greenberg). The improvement in attention score was significant at the level of p < .01. The control group showed no significant movement on that test. The study also found behavioral improvements, but scores remained in the abnormal range even after the training. Furthermore, the control group improved on most behavioral measures even more than the treatment group. All changes were significant. However, they failed to confirm the expected interaction with treatment condition. On the other hand, amplitudes in the 4-7 Hz regime did decline with the training, whereas they increased in the control group. Amplitudes within the training band of 12-15 Hz did not change significantly over the course of training. The biofeedback cohort was given feedback via a PAC-Man like object which encoded the feedback signal in terms of its brightness and velocity through a maze. The control group was given the conventional PAC-Man game for the same number of sessions.
One hesitates to propose video games as a remedy for the disregulated behaviors of ADHD on the basis of the above behavioral improvements! There is a better interpretation of these findings. The Cartozzo study restricted itself to a single protocol in the study. This was unfortunate, since it is already known that different children may need different protocols (Lubar, 1991). The result is that some children improve while others do not, or may even deteriorate in behavior. This manifested itself in an increase (near-doubling) in the standard deviation of scores in the experimental group (whereas the standard deviation remained unchanged in the controls). The behavioral changes were reduced in statistical significance because of this increase in dispersion of the data.
Finally, Aubrey Fine and Larry Goldman of California Polytechnic Institute in Pomona (Cal Poly) reported preliminary results on their controlled study of ADHD at the 1994 annual meeting of the American Psychological Association in a poster presentation. This study involved two experimental groups, one getting EEG biofeedback, and a second obtaining cognitive training with a computerized tool (Captain's Log). A third group was a wait-list control. Because this was an initial study of a survey nature, the intake criteria were quite inclusive. Children were admitted who had been diagnosed not only with ADHD but also with seizure disorder, Tourette Syndrome, and depression. Most were already under medical management for these conditions. Some 80% of the group were medicated, 15% of them with more than one medication. Some of the remaining 20% of children were on summer drug holiday.
Because of the multiple objectives of the research study, assessment tools included the Wide Range Assessment of Memory and Learning, Stroop Color and Word Test, Kagan Familiar Figures Test, and the Grooved Pegboard Test. For ADHD assessment, the Conner's CPT and the Gordon Diagnostic System were employed, in addition to several parent questionnaires (Child Behavior Checklist, Home Situations Questionnaire, Child Attention Profile, Revised Conner's Questionnaire, and Social Skills Assessment).
Few of the academic skills tests revealed statistically significant improvement in either of the experimental conditions (12/51). Of course, the population had not been selected for deficiencies on those measures in the first instance, so it is difficult to judge the import of this finding. With respect to the ADHD assessments, the Conner's CPT showed improvement in omissions (inattention), commission errors (impulsivity), and response time. However, so did the control group! No statistically significant interaction with treatment condition was identified. The Gordon Diagnostic System did not yield significant change.
Parental Assessments, however, indicated some significant favorable changes. The Conner's Questionnaire yielded improvement on the impulsive-hyperactivity scale at the p < .01 level for the EEG training contingent. This finding is taken to be highly significant for several reasons. First, the behavioral improvement was identified in a population that was largely already medicated for ADHD, as stated above. Secondly, it was identified after only 20 training sessions, which is generally acknowledged to be insufficient to achieve a full resolution of ADHD with EEG training (Lubar, personal communication). Thirdly, it was identified by parents strictly on the basis of the home environment, since the first 20 training sessions were conducted over the summer months. For many of the children, the home environment is not as challenging as the school environment, so it would be harder to observe a change. And finally, a significant number of children were able to reduce their medications over the course of the EEG training. A maintenance of behavioral scores in the face of reduced medication dose should also be judged an improvement, but would not show up in the above statistics.
Part of the experimental group in the Cal Poly study was selected for an additional 20 sessions of training. The remainder of the group was composed of those who had essentially met the objectives of the training, and those who had not made significant progress in the first 20 sessions to merit continued training. The latter two groups were each a quarter of the total, so that about half (12) was selected for more training. In fact, only about 7 actually undertook the additional 20 sessions and subsequent retest. The results of testing after 40 sessions showed continued progress, as illustrated in Table 2, with four measures reaching significance at p < .05, and two reaching significance at p < .01 (hyperactive index of Conner's, and the depression scale of the child behavior checklist).

Figure 2. Effect of 20 and 40 Sessions of EEG Biofeedback on Conner's Questionnaire Scale B (Learning Problems); Scale C (Psychosomatic); Scale D (Impulsive Hyperactivity); and Scale F (Hyperactivity); and on the Child Behavior Checklist depression and delinquency scales.

The last of the three groups may be regarded as "non-responders" in the usual sense. The case can be made that the experimental data should be evaluated also on the basis of a division between "responders" and "non-responders", as is typical in drug studies [Introduction * Recent Findings * Clinical Findings * Results * Discussion

Findings from Clinical Operations
The next useful step which can be taken is to review the quantitative data derived from our actual clinical setting. In the following, T.O.V.A. data are reported which were obtained at our home office and affiliated offices for children with ADHD or ADHD-type symptoms.
The analysis was performed by David Kaiser, who had access to the data from nine independent offices employing identical protocol selection criteria for this population. The protocols consist of training to reinforce instantaneous increase in EEG amplitude in either the 12-15 Hz or the 15-18 Hz band, at either C3, Cz, or C4, as needed depending on the symptomatology. The augmentation training is accompanied by inhibition of excessive amplitudes in the 4-7 Hz and 22-30 Hz regimes.

Return to Introduction * Recent Findings * Clinical Findings * Results * Discussion

Experimental Results
Mean pre- and post-training TOVA scores are presented in Table 3 for the initial re-evaluation after 20 training sessions. A univariate Analysis of Variance (ANOVA) was used to evaluate the effect of EEG biofeedback training on four components of the T.O.V.A. The four test components were Inattention (Percent Omission), Impulsivity (Percent Commission), Response Time, and Response Variability. Because some clinicians truncated test scores at 40 points (i.e., four standard deviations below normal), all individuals who scored 40 points or below were assigned the score of 40 points. A total of 239 subjects were included in this analysis.

Figure 3. Mean standard scores for T.O.V.A. subtests before and after 20 EEG biofeedback sessions for 239 children and adults.

A significant effect of EEG biofeedback training as well as an interaction with TOVA components were observed in this data set (p < .001). After 20 EEG biofeedback sessions, subjects improved significantly in inattention scores (p < .0001); in impulsivity scores, (p < .0001); and in variability of response time (p < .0001). The mean improvement in impulsivity score is nearly one standard deviation (15 points in standard score). Response time does not show systematic improvement. This is partly understandable in that subjects increase in mean response time as they reduce in impulsivity. This dependency is documented in a regression coefficient of -.26 between impulsivity and response time scores.
The availability of a large subject pool also allows us to evaluate the effect of additional training sessions. Of the 239 subjects, 56 continued training and were retested after 40 or more EEG biofeedback sessions. Understandably, these were individuals who had made only modest progress in 20 sessions. The results are shown in Table 4. As before, a significant effect of EEG biofeedback training (p < .001) as well as an interaction with TOVA components (p < .01) were observed. Compared to the pre-training scores, after 40 EEG biofeedback sessions subjects improved significantly in inattention (p < .05); impulsivity (p < .001); and response variability (p < .001). When TOVA scores were compared between 20 sessions of training and 40 sessions of training, marginal improvement was seen in response time (p < .10); and significant improvement was observed in response variability, (p < .001).

Figure 4. Comparison of T.O.V.A. subtest scores after 20 and 40 training sessions for 56 subjects who continued training out of the 239 subjects of Figure 3.

The treatment of data on the basis of averages obscures much detail. In particular, it weights equally those who are in significant deficit with respect to a parameter, and those who may be normal. In Table 5, results for all 239 subjects are presented depending upon initial pre-treatment scores. All data with starting values above the normal score of 100 have been excluded from the table. It may aid interpretation of these results to note that in its application to titration of medication, T.O.V.A. score changes of a half standard deviation (7.5 units in standard score) are taken to be significant (Greenberg).

Figure 5. Results are presented for 312 subjects for changes in TOVA subtest scores. Each line segment represents a single subject's change from pre-training to post-training scores. The data are sorted by pre-training score. Improvement is indicated when the line segment rises above the pre-training value. Only individuals with pre-training deficits in a subtest are included in each figure.

In Figure 5, there is a systematic tendency toward improvement in all four subtests, with the most significant improvements occurring where the pre-test scores are in most severe deficit.

Return to Introduction * Recent Findings * Clinical Findings * Results * Discussion

Discussion
Taken together, the data from all of the studies support the model that EEG biofeedback training is effective in changing neurophysiological function which contributes to the symptomatology of ADHD. Improvements in IQ scores are systematically found in all of the studies which have looked for them, and such improvements are very difficult to ascribe to placebo factors. Behavioral improvements were noted in all of the studies.
The data derived from a number of clinical settings using common protocol selection criteria also appear to be quite robust in demonstrating changes in physiological function consequent to EEG biofeedback training. Since these data are not matched by controls, it remains to dispose of any placebo interpretation. First of all, the fraction of subjects favorably impacted by the training considerably exceeds that expected from typical placebo effects. The large study by Ullman and Sleator, for example, found an 18% placebo response in a medication study (Ullman and Sleator, 1986). More typically, placebo response falls in the range of 30% of the subject population.
Secondly, much placebo response is traceable to investigator bias. That is not present in this instance, since the testing is completely computer-controlled and computer-scored. Thirdly, the placebo effect can be traced to expectations on the part of the participants: Teacher questionnaires such as those used in the Ullman and Sleator study may be subject to such bias. Again, that cannot be the case in the present instance. Fourth, the placebo effect is expected to be in the direction of a favorable outcome. The data, by contrast, show that of the few who do not show favorable change, a large fraction actually worsen their scores to a degree not expected by normal drift in the measurement over time. This argues for the presence of an active agent with the potential for inducing both positive and negative change. This is inconsistent with a placebo interpretation.
The adverse changes observed in our subjects deserve some further comment. The essential clinical choice required for each subject is whether to address primarily the inattention or the impulsivity. Each of these requires its own class of protocols. If both components are present, a compromise may be required in terms of our clinical objectives. Hence, improvement in impulsivity may entail a worsening of the inattention score, and an improvement in inattention may entail an exacerbation of impulsivity. The balance may be redressed in a subsequent series of 20 training sessions when secondary symptoms are met with a different choice of protocols.
By virtue of this choice of protocols, each subject in a sense becomes his own control, since an inappropriate choice of protocols may produce adverse test results. Since a client cannot have wished for such specific and disparate outcomes with respect to impulsivity and inattention, of which he may himself be only dimly aware, a placebo explanation is ruled out. A further argument against a placebo explanation is that both Table 2 and Table 4 show continued progress when sessions are extended from 20 to 40 sessions. A placebo effect, by contrast, tends to fade over time.
EEG biofeedback is a physiologically based tool with essentially no emphasis on talk therapy. The placebo effect, whatever its nature, is psychologically mediated. It is more parsimonious to propose that a physiologically based technique actually has a physiological effect, than that the physiological effect is mediated by some psychological factor such as being mesmerized by fancy instrumentation (Barkley, 1992). General psychological factors simply lack the specificity of the tool we have at our disposal. Moreover, calling something a placebo effect does not dispense with the issue. So what if our effects are mediated psychologically to some degree? There are ultimately measurable physiological consequences, the mechanisms for which require elucidation.
When the placebo hypothesis is adduced in a test of medication efficacy, it encompasses all "non-drug" effects. However, in its application to a behavioral management technique, the placebo hypothesis is an empty hypothesis. It lacks testable predictions which would allow it to be distinguished from a postulated "real" effect. To call something a placebo has the perverse effect of barring further inquiry. In the absence of decision criteria and testability, the placebo hypothesis lacks scientific utility in application to a non-drug modality.
It remains to deal with the fact that the behavioral changes found in the controlled studies do not appear as robust as the physiological changes documented by the T.O.V.A. results from the clinical settings. The Cartozzo study suffered from a small subject group, as well as from restriction to a single protocol, as already mentioned. The problematic finding in that study was the fact that the control group improved its behavior significantly as well. This is perhaps an indication of the intrinsic volatility of behavioral variables, as well as the additional variability attributable to parental and teacher assessment. Such variability may place fundamental limitations on our ability to demonstrate systematic progress with training in small populations.
The same considerations apply to the Cal Poly study. Significant population shifts were observed in the control group with respect to a number of test variables. This limited the significance of any improvement in the experimental groups. This shortcoming can be dealt with partly by greater reliance on more physiologically based measures, such as those of the T.O.V.A., and on stable tests such as the WISC-R. The limited significance of changes, in the face of apparent high intrinsic variability of the measures chosen, can only be overcome with larger "n". One way of achieving a larger subject population is to perform a meta-analysis of the data, such as that of Table 1, where common test instruments were used in a number of studies. The total number of subjects in Table 1 comes to 54. It is our intention to solicit the cooperation of the other researchers to perform such an analysis in the future.
We suggest, in summary, that the T.O.V.A. results obtained in actual clinical settings are sufficiently cogent and robust to justify the enthusiasm for EEG training which is building among clinicians, and to justify the interest of academic researchers. Finally, the above data should cause any objective researcher to desist from asserting that clinicians may be premature in using this technique clinically. In particular, contamination by placebo factors does not invalidate the finding of significant improvement. A placebo effect is inevitably marbled through everything that a clinician does, and in fact the clinician will do everything in his power to enhance the effect!
We expect that mental health and educational professionals will be increasingly compelled to address the physiological basis of behavior, as this is elucidated in research, and that EEG biofeedback will be increasingly recognized as a useful complement to other behavioral interventions as well as to psychopharmacology. We believe that the discoveries now being made in the field of EEG biofeedback portend a watershed in the field of mental health and of education as the implications of these findings are gradually assimilated.

Return to Introduction * Recent Findings * Clinical Findings * Results * Discussion

References
Atkins, M., and Milich, R. (1987). IOWA-Conners Teacher Rating Scale. In M. Hersen & A. Bellack (Eds.), Dictionary of Behavioral Assessment, Pergamon Press, New York, pp. 273-275.
Barkley, R. A. (1992). Is EEG Biofeedback Training Effective with ADHD Children? CHADDer Box, April, p. 5-11.
Captain's Log, Sandford, J.A., 727 Twin Ridge Lane, Richmond, VA 23235 (804) 320-0105
Cartozzo, H.A., Jacobs, D., Gevirtz, R.N. (1995). EEG Biofeedback and the Remediation of ADHD symptomatology: a controlled treatment outcome study. Presented at the Annual Conference of the Association for Applied Psychophysiology and Biofeedback, Cincinnati, Ohio, March.
Fine, A. H., and Goldman, L. (1994). Innovative techniques in the treatment of ADHD: An analysis of the impact of EEG biofeedback training and a cognitive computer generated training. Paper presented at the 102nd Annual Convention of the American Psychological Association, Los Angeles, CA August 12-16.
Greenberg, L.M., Department of Psychiatry, Box 393 UMHC, 420 Delaware Street Southeast, Minneapolis, Minnesota 55455 (612) 626-5806.
Kaufman, A. & Kaufman, N. (1990). K-BIT: Kaufman Brief Intelligence Manual. Circle Pines, M.N. American Guidance Service.
Linden, M., Habib, T., Radojevic, V. (in press). A controlled study of the effects of EEG biofeedback on cognition and behavior of children with attention deficit disorders and learning disabilities. Biofeedback and Self-Regulation
Lubar, J.O., and Lubar, J.F. (1976). EEG and behavioral changes in a hyperactive child concurrent with training of the sensorimotor rhythm. Biofeedback and Self-Regulation, 1, 293-306.
Lubar, J.F. (1991). Discourse on the development of EEG diagnosis and biofeedback for attention deficit/hyperactivity disorders. Biofeedback and Self-Regulation, 16, 201-225.
Othmer, S.F, Othmer, S., and Marks, C. S., (1991) EEG Biofeedback Training for Attention Deficit Disorder, Specific Learning Disabilities, and Associated Conduct Problems, EEG Spectrum, 16100 Ventura Blvd., Suite 10, Encino, CA 91436.
Othmer, S. F., and Othmer, S., (1994). Professional Training Syllabus, EEG Spectrum, 16100 Ventura Blvd., Suite 10, Encino, CA, 91403.
Swanson, J., Nolan, W. & Pelham, W. (1981). The SNAP rating scale. Resources in Education.
Sterman, M.B. (1980). EEG biofeedback in the treatment of epilepsy: an overview circa 1980. In Leonard White and Bernard Tursky (Eds.), Clinical Biofeedback: Efficacy and Mechanisms. New York: Guilford Press.
Tansey, M.A. (1990). Righting the rhythms of reason. EEG biofeedback training as a therapeutic modality in a clinical office setting. Medical Psychotherapy, 3, 57-68.
Ullman, R.K., and Sleator, E.K., (1986). Responders, nonresponders, and placebo responders among children with Attention Deficit Disorder. Clinical Pediatrics, 25, 594-599.