The overall goal of this study was to identify by retrospective data analysis, quantifiable evidence supportive of beneficial effects of adrenal corticosteroid (prednisolone) therapy on brain, language and behavioral function of children with a history of sudden autistic regression (R-ASD). Such evidence, while retrospective, nevertheless might serve to strengthen sufficiently the existing anecdotal reports for such salutary effects [15, 24, 25], in order to potentially justify a future prospective randomized trial of steroid treatment for R-ASD. To this end a large clinical patient database of children with a diagnosis of autism spectrum disorder (ASD) was searched for pertinent target subjects who met the criteria for regressive autism, were between the ages of 3 and 5 years at the time of their first FMAER study, had an absent or distorted initial FMAER study, received steroid treatment (STAR group), and had a second FMAER study performed after steroids were discontinued. Following identification of the target group subjects, a non-steroid-treated ASD comparison group was identified. These were comparable-age children also with an initial absent or distorted FMAER study and with a second FMAER study performed between 6 and 36 months later, yet not treated with steroids (NSA group). Twenty target group children (STAR) and 24 comparison group children (NSA) were identified (Table 4). The NSA group children were followed by their clinicians in the same overall time span as the STAR sample. Additionally available data aside from the two FMAER studies included for the STAR group receptive and expressive language scores based on a clinical assessment (CLSQ) specifically developed for pediatric ASD patients who receive pharmacological treatments, as well as DSM-IV ASD total number of symptom scores. For the NSA group receptive and expressive language scores derived from neurologists’ clinical notes were calculated for the two clinical visits temporally closest to the two FMAER tests utilized. All scores were available and had been entered into the clinical database prior to study design and analysis.
The investigation’s first and most important specific goal was to compare by means of objective spectral analysis, potential changes in the 4 Hz FMAER steady-state response between the first and second study point for each of the two subject groups separately. Results showed striking differences in change between the two populations (Table 5). For the STAR group, 9 of 14 electrodes showed markedly increased second-study 4 Hz response magnitudes with maximal effect in the left inferior-posterior temporal electrode, TP9 (p ≤ 0.00037). In contrast the NSA group failed to manifest any significant change in 4 Hz spectral power at any of the 14 electrodes Thus, the steroid treatment appears associated with a significant increase in the specific FMAER stimulation elicited 4 Hz response amplitude of the superior temporal gyri (STG) in both hemispheres of the study children with regressive autism. Additionally, the FMAER response distortion present at the first study point of the STAR group was absent at the second study point (Table 1) Thus, the STAR group demonstrated both higher amplitudes and a less distortion in the FMAER after steroid treatment.
Not all electrodes manifested a significant increase in the 4 Hz response to steroids in the STAR group (Table 5). The reason for this appears to lie in the fact that the FMAER may be modeled as a single dipole within each STG. These dipoles have both a specific physical location and a specific orientation. What is seen on the scalp is the projection of the dipole source from each STG upon the overlying scalp such that the scalp response magnitude reflects the dipole orientation and location which varies slightly among subjects. By including 14 distributed scalp locations seven per hemisphere, one may conclude that failure to demonstrate any significant changes in the NSA group did not arise from a failure to record over a ‘hot’ scalp region , but from the lack of a significant response change over time.
As Figure 1 illustrates the before and after spectral plots demonstrate that the subjects receiving steroids showed not only a significant increase in magnitude of the appropriate 4 Hz FMAER response but simultaneous that distortion products in the form of other than 4 Hz response components evident in the pre-medication FFT plots were significantly reduced. In their well-known speech quantification work, Wu and Pols  note that “…in many situations the speech quality is not satisfactory, often due to factors related to the transmission of the speech signal from speaker to listener. During transmission there is a variety of factors influencing the quality of speech, such as: a limitation of the frequency range or the dynamic range, noises, echoes, and other (analog or digital) distortion components.” On the basis of the current study’s results it appears reasonable to speculate that distortion may additionally occur within the central auditory pathway of the listener as visualized by the FMAER at the level of the STG.
It is hypothesized that language regression in autism may result from the development of dysfunction in the specific STG systems that are needed to accurately and cleanly detect rapidly changing spectral information within the acoustic stream. Children with R-ASD are not deaf or ‘hard-of-hearing’ per se; they often clearly identify even very soft novel sounds. The problem appears to reside in the auditory distortion that occurs within cortex devoted to language processing. This distortion appears to be at least partly reversible with steroid therapy.
The study’s second goal was to determine whether the significant FMAER improvements might be related to steroid-based suppression of presumed-to-be minor abnormalities in the children’s EEGs as observed by visual inspection. Note that neither group demonstrated frank epileptiform transients such as spikes or spike and wave patterns. The EEG change rankings did not differ between the two groups (Table 6). Additionally the EEG changes in the STAR group included both EEG improvement and EEG worsening. These findings emphasize that there appears to be little evidence for a physiologically meaningful steroid effect on EEG. This is an important finding that runs contrary to the assumption made in Landau-Kleffner syndrome, namely that both language and FMAER improvements with steroids result from suppression of frequent spike discharges within the STG . What may be common to R-ASD and LKS children’s brains is dysfunction of the STG. However, the dysfunction need not necessarily be accompanied by spike discharges. It may, none-the-less, be ameliorated by steroids.
The third goal was to compare changes in clinical-rating-based language scores between the first and the second study. As shown in Table 7 the STAR group showed highly significant improvement in the language ratings between time study 1 and 2; the improvement was comparable for both the receptive and expressive language ratings. Moreover significantly more STAR group subjects (17/20) than NSA group subjects showed improvement (6/24 ‘better’ receptive, and 10/24 ‘better’ expressive). These data suggest that steroid treatment may be associated with improvement in language and that more subjects who receive steroid treatment may show such improvement than subjects in the non-treated group.
In order to rule out the possibility that regressive autism spontaneously improves with time estimated language change of the seven NSA group children with histories of regression was compared to that of the 17 NSA group children without such a history of regression (Table 4). Although the small population size precludes a definitive answer, there was no significant difference between the two NSA subpopulations for receptive or for expressive language ratings (Table 9). Thus, spontaneous NSA group improvement of language in the regressive autism subpopulation was not observed.
The fourth goal was to explore the possible relationships between changes in 4 Hz spectral responses at all electrodes and the language ratings change scores (Table 7) for receptive and expressive language separately. The STAR group (Table 10) showed a strong correlation between receptive language and the FMAER at the left central region (C3) and between expressive language and the FMAER in the left mid temporal region (T7). It appears clinically meaningful that STAR group language improvement was more evident in the left hemisphere as the left hemisphere is typically the dominant hemisphere for language. Given that the scalp FMAER data are highly inter-correlated and reflect activity at the level of the STG, the difference in the scalp location of maximal effect for correlations with receptive and expressive language is curious. It is possible that slightly different regions of the STG are responsible for expressive and receptive language functions resulting in subtle differences in the dipole source orientations or locations and correspondingly different in patterns of scalp projection. This is an area for future exploration. For the NSA group (Table 11) a strong correlation was found between the right parietal (P4) region and both the receptive and the expressive language scores. It is curious that the typically non-language dominant right hemisphere STG maximally correlated with NSA group language function. This may reflect relatively greater dysfunction within the left hemisphere in this group. These findings are limited by the clinical rating measures of language employed in this study. Formal language testing with standardized assessments will be required to substantiate these preliminary findings.
The fifth goal was to determine the level of behavioral improvement within the steroid-treated group. The scaled DSM-IV symptom summary scores demonstrated a highly significant improvement after steroid treatment (p ≤ 0.00001) (Table 12) Thus steroid treatment may be associated with behavioral improvement.
Finally the study’s sixth goal was to weigh the complications of steroid administration against their potential benefits (Table 13). Overall, almost all steroid-treated subjects gained weight. Working with families to counteract increased appetite and associated weight gain was a dominant clinical task in patient management. Behavior and sleep disorders were readily managed pharmacologically. The two subjects, who experienced behavioral regression after steroid taper, were successfully treated with a short period of pulse dosing. There were few serious complications although the process clearly increased the task of managing the child’s well-being and health for the parent and should only be undertaken with comprehensive, detailed understanding of the ramifications involved in such a significant medication trial. The most frequent management tasks for the parents were: (1) the constant need to monitor food intake in order to avoid weight gain which in turn might facilitate development of secondary hypertension; and (2) the difficulty in managing the child’s behavior.
It is unclear at this point how steroids might function to improve language and behavior in regressive autism. Aside from a relatively sudden clinical onset the current study fails to establish similarities between R-ASD and the Landau-Kleffner syndrome and other epileptic syndromes. Some have suggested an immune process, possibly initiated by outside factors that challenge the immune system. Only two of the 20 families in the STAR-group suggested regression onset in close relationship to inoculations (MMR, influenza vaccination). Extensive studies reported in the literature have failed to identify a clear association between autism (including R-ASD) and protective immunizations . Eighteen of the 20 current study’s R-ASD STAR group subjects provided no clinical information as to possible trigger events. Although a question of importance, the current data do not shed light on regression etiology.
As reviewed by Kurian and Korff  corticosteroids, although not necessarily a first-line treatment, have been used with some success in a number of epileptic syndromes aside from LKS. These include infantile spasms, Doose syndrome, Dravet syndrome, Rasmussen’s encephalitis, absence epilepsy, and acute ‘symptomatic’ seizures. There is, unfortunately, no agreement as to the mechanism(s) of corticosteroid actions as an anticonvulsant. However, a study by Di Chiara et al.  found in a rat model that glucocorticoids may simultaneously influence different pathways that differentially facilitate inhibitory GABA release and inhibit excitatory glutamate release, thus summing to an overall effect of neuronal suppression. Relevancy to epilepsy, LKS and R-ASD is merely speculative at this point. As reviewed by Chatham and Kimberly , corticosteroids manifest a well-documented multiplicity of cellular effects that may be relevant to treatment of immune disorders such as lupus. Again, the relevancy to R-ASD remains to be established.
Hopefully a better understanding of corticosteroid action mechanisms may lead to the development a more precisely targeted drug that has fewer secondary complications and is useful in epilepsy treatment as well as the amelioration of R-ASD symptomatology.
The current study benefits from the availability of the simple and objective FMAER test that is relevant not only for detection of response amplitude changes but also of response distortion. Nevertheless the current study has very significant limitations that preclude conclusions as to the effectiveness of steroid treatment for regressive autism. The study, however, appears to provide sufficient indication in support of the need for a prospective, randomized, double blind, placebo controlled, cross-over design trial in order to assess more rigorously the potential role of steroids in R-ASD.
Study limitations as indicated above include but are not limited to the following:
The study was retrospective.
The STAR and NSA groups were similar yet did not meet all of the same selection criteria. The STAR group was composed exclusively of children with regressive autism while the NSA group was composed mostly of children with autism without history of regression.
Aside from the FMAER different language measures were applied to STAR and NSA group. Neither of the language measures is standardized or published.
Quantification of the DSM-IV symptom list as a measure of behavior is also unpublished and not in standard usage; moreover this measure was only available for the STAR group.
Analysis by ANOVA with interaction and possibly with covariates would be preferred over paired T-test if the groups were more comparable initially.
Aside from the FMAER the language and behavior data were obtained by clinicians in collaboration with the parents rather than by independent observers.
The one year follow-up is inadequate in terms of length of time from treatment and in terms of the rigor of the language and behavior data gathered.