In the absence of revolutionary discoveries that elucidate the pathophysiology of autism and lead to accurate diagnosis in utero or infancy, treatments will focus on halting further abnormal brain development and compensating for prior aberrations. This approach is comparable to almost all neuropsychiatric disorders whose pathophysiologies are poorly understood as well as to most somatic medical disorders such as hypertension and diabetes. Conceptually, animal and human studies of sensitive periods indicate that strategies focused on overcoming early problems with brain development may be efficacious for the treatment of autism. Early intensive behavioral interventions (EIBI), which have become the standard of care in autism and appear to influence overall development and reduce the intensity of core symptomatology, are based on this strategy, (Aman 2005
; McEachin et al 1993
; Smith et al 2000
There are indications that pharmacologic manipulations are equally as capable as behavioral intervention of producing benefits if provided during sensitive periods. For instance, in the drosophila model of Fragile X syndrome (FRAX), the mGluR antagonist, 2-methyl-6-(phenylethynyl)pyridine (MPEP) can reverse associated abnormalities in both brain structure and behavior if provided during early development, but has no anatomical effects if given later (McBride et al 2005
; Yan et al 2005
). In humans with Smith-Lemli-Opitz Syndrome (SLOS), in whom autism is extremely prevalent, cholesterol supplementation prior to age five reduces the risk of autism spectrum disorders four-fold (22% versus 88%) compared to later supplementation (Tierney et al 2001
). These two disorders are intriguing because up to 25% of males with FRAX and 50% of individuals with SLOS meet diagnostic criteria for autism (Hatton et al 2006
; Sikora et al 2006
). Thus, treatments that impact experience-dependent brain organization and are provided during periods of heightened plasticity might compensate for some developmental abnormalities observed in autism.
Identifying effective treatments
The optimism related to developmental treatments must be tempered by the recognition that we do not know which pharmacologic interventions will be effective. Further, given the apparent etiologic heterogeneity of autism, it is possible that different interventions will be efficacious in different autistic subgroups. Interventions related to molecules or processes demonstrating clear developmental differences in autism as compared to typical development may be particularly effective. However, it is also possible that the developmental periods when such interventions would have been effective will have ended prior to diagnosis of autism. In that case, it will be important to evaluate treatments targeted toward molecules known to be important in the brain’s response to developmental perturbations or toward enhancing plasticity instead. The multiplicity of molecules involved in typical neurodevelopment and the complexity of their interactions suggest several potential areas in which interventions might be developed for use in autism.
In designing developmentally-based interventions, it is important to remember several key challenges. First, the intervention attempts to modulate function in an already perturbed system. The levels of almost all neurodevelopmental molecules are precisely regulated in development and some sort of disruption has already occurred. Investigators cannot assume that provision of a molecule to autistic individuals during development will have the same impact as provision of the same molecule to individuals without pre-existing perturbations in the molecule’s signaling, transport and spatial distribution, or synthesis. For instance, mice with reduced serotonin synthesis due to 1473G tryptophan hydroxylase-2 homozygosity, show typical responses to a SSRI only if provided with exogenous tryptophan (Cervo et al 2005
). Secondly, it seems essential to base intervention strategies on brain rather than peripheral findings. For instance, initial observations of hyperserotonemia led to the conclusion that brain levels of serotonin were likely to be inappropriately high. Therefore, attempts were made to reduce brain serotonin through the use of fenfluramine and L-tryptophan depletion. In both these cases, peripheral serotonin was reduced, but there was no clear benefit from treatment despite several studies (Campbell et al 1988
). Instead, irritability was common with fenfluramine and there was clear exacerbation of autistic symptoms with tryptophan depletion (McDougle et al 1996
). As interventions are proposed and developed, it will be essential to evaluate both efficacy and safety and tolerability.
Serotonergic interventions have promise
Interventions that enhance serotonergic neurotransmission during early childhood development appear to have the most immediate potential for eliciting clinically important, adaptive brain changes in children with autism(Chugani 2002
; Chugani 2005
). Serotonin plays a critical role in the development of cortical columns and experience-dependent organization. Animal work suggests that fluoxetine treatment can prevent functional brain damage from hypoxic injuries (Chang et al 2006
). Further, there is evidence of a developmental abnormality in serotonin synthesis in some young children with autism. Equally important, FDA-approved medications which are likely to enhance serotonergic neurotransmission are currently available. These agents appear to have relatively few adverse effects in human children even when exposure occurs in utero
or during early infancy (Barbey and Roose 1998
; Gentile 2005
; Isacsson et al 2005
; Levinson-Castiel et al 2006
; Malm et al 2005
; Misri et al 2006
; Moses-Kolko et al 2005
; Safer and Zito 2006
). The adverse effects that have been noted in humans are related to withdrawal syndromes, high serum levels or overdoses (Knoppert et al 2006
), and possible activation and increased suicidality. Further, the long-term impact from perinatal and/or early childhood exposure to selective serotonin inhibitors (SSRIs) is not yet known. It should be noted that three rodent studies in which developing animals were treated with high doses of SSRI for extended periods observed various late emerging side effects such presumed anxiety in adult mice (Ansorge et al 2004
). In contrast, animals treated with 67% lower doses from PND1-7 showed no late emerging behavioral effects utilizing the same assessments (Chang et al 2006
SSRIs increase availability of serotonin in the synaptic cleft. A recent review of limited data in the pediatric autism population suggests that the SSRI may have some benefits and appear safe, but definitive studies do not exist (Kolevzon et al 2006
). The most rigorous study, demonstrated clinical benefit of fluoxetine in the treatment of repetitive behaviors in children and adolescents with autism (Hollander et al 2004
). There is also case series data suggesting this SSRI may have developmental effects, particularly on language, in young children with autism (DeLong et al 2002
; DeLong et al 1998
). In response to the need to move forwards with the evaluation of promising early interventions, two centers within the NIH-funded Studies to Advance Autism Research and Treatment (STAART) network have initiated a pioneering study to examine the developmental impact of fluoxetine treatment in preschool children with autism. This developmental trial builds upon an initial feasibility study initiated by Drs. Sikich and DeLong in 1999.
Regardless, this approach is not without risk. A major impetus for treatment with an SSRI is our interpretation that there are inadequate amounts of serotonin in the brains of children with autism. However, this interpretation of Chugani’s seminal PET studies may be incorrect since the PET studies provide no information about children younger than two years. Serotonin synthesis in the central nervous system may indeed be inadequate, but may indicate an adaptation to an earlier developmental period of excessive serotonin synthesis or signaling (Whitaker-Azmitia 2005
). Subsequent use of an SSRI may impede natural compensatory changes in the developing brain, thereby worsening pathology. Further, serotonin findings in autism may merely indicate dysfunction in one or more of the neurotransmitters and factors that facilitate the development of serotonin neurons, receptors, synthesis, or release (see
). If the autism phenotype is a manifestation of an ‘upstream’ regulator, it is unclear what the effect of early SSRI treatment will be. Additionally, it is possible with serotonergic treatments, as with any other treatment provided early in development, that there may be late emerging side effects such as have been described in some animal studies. Ultimately, it will be essential to assess the balance between the potential benefits of early treatment on a devastating life long disorder and the potential risks of ineffective treatment or late-emerging adverse effects.
Other potential developmental pharmacologic interventions
Given our understanding of signaling molecule disruptions in autism, there are several potential targets in addition to serotonin for developmental interventions in autism, as summarized in . The two primary approaches would be to enhance plasticity and promote compensatory experience related changes or to remediate identified imbalances in neurotransmission. In the first case, plasticity could potentially be promoted if sensitive periods were extended or reopened as suggested by Chugani (2005)
. Reductions in GABAA
activity (e.g. by reduced GABA synthesis, increased degradation or receptor blockade) might have this effect. However, current GABA antagonists are quite toxic. Blockade of BDNF early in development might also have this effect. Another strategy might be to disrupt the extracellular matrix with an agent like tPA in order to create a more permissive environment for synaptic reorganization. Finally, one could try to augment cholinergic or dopaminergic neurotransmission in emulation of the animal studies of auditory cortical areas that demonstrated plasticity by activating the nucleus basalis or ventral tegmental area (Bao et al 2001
; Kilgard and Merzenich 1998
). Cholinesterase inhibitors or dopamine agonists might act in these ways.
Hypothetical targets for early interventions in autism.
If one takes the approach of trying to address identified disruptions in neurotransmission, evidence most strongly supports targeting the serotonin system, as discussed earlier, or the GABAergic system. The evidence of reduced GABAergic activity coupled with its critical role in experience-dependent brain organization makes it a primary target. Modulation of GABA could influence the complex excitatory:inhibitory balance which appears critical to process and synapse refinement. The availability of approved agents that enhance GABAergic neurotransmission such as valproic acid, benzodiazepines and estradiol suggest that they may be appropriate candidates for treating children with autism. Further, in a mouse model that has deficiencies in SERT,
female mice and males treated with estradiol show more normal levels of serotonin, more complex hippocampal dendrites and fewer anxiety related behaviors than untreated males or females with ovariectomy or tamoxifen (Ren-Patterson et al 2006
). In addition, if autism reflects an increased ratio of excitation to inhibition as suggested by Rubenstein and Merzenich (2003)
, benefits may also be derived from dampening excitatory neurotransmission. Agents which reduce glutamatergic activity (such as lamotrigine, topiramate or zonisamide) or enhance activity of the EAAT might be beneficial due to such actions.
If on the other hand, autism is related to a hypoglutamatergic state at NMDA receptors as proposed by Carlsson (Carlsson 1998
), highly selective NMDA partial agonists might have utility as a treatment. Unfortunately, even a regionally selective, highly specific agonist is likely to be neurotoxic. Carlsson advocates an alternative approach in the augmentation of AMPA neurotransmission through the use of ampakines. Although no ampakines have yet been approved by the FDA, a 6 month pilot study in adults with Fragile X syndrome, did not observe significant adverse effects (Berry-Kravis, personal communication). Carlsson has also suggested that the primary implication of weak NMDA tone is excessive 5HT2A
activity. Although there is no experimental support from autistic subjects for Carlsson’s hypothesis, serotonin synthesis is increased during adolescence in individuals with autism and limited use of serotonergic antagonists, such as the second generation antipsychotics, could be helpful.
Prior medication trials demonstrating some developmental effects in autism
Prior medication trials involving at least 6 children with autism that have demonstrated some benefit for core symptoms of autism are summarized in . With the exception of DeLong’s and Alcami’s studies, none of these trials focused on developmentally targeted intervention. Consequently, few young children are included. We would expect that benefits for any of these treatments might be enhanced if younger children were included. It is noteworthy that the only trials that indicate improvement in communication or social behavior are open studies. In contrast, reduction in restricted and repetitive behaviors (RRB) have been demonstrated in multiple trials, the largest of which is a trial of the second-generation antipsychotic, risperidone. It remains unclear whether this reflects the increased difficulty of assessing social and communicative behaviors or if it reflects the limitations associated with brief acute trials and the difficulties maintaining children in double-blind treatment for extended periods. More comprehensive reviews of pharmacologic treatment studies in autism are available (Buitelaar and Willemsen-Swinkels 2000
; McDougle 2005
Published clinical trials with improvement in core autism features.
How should evaluation of potential developmental treatments for autism proceed?
In order to validly assess the impact of developmentally focused interventions, it will be essential to evaluate not only acute effects but also long-term changes in core symptoms and other developmental abilities, acute and mid-range tolerability, and very late emerging adverse effects such as those described in animal studies. Because of the limitations of every system available for study, multiple approaches will be required. The three major approaches that can be utilized at this time are: 1) animal studies; 2) inclusion of children of all ages in clinical trials of agents with potential neurodevelopmental applications, and 3) more protracted trials or post-trial observation periods that allow assessment of long-term consequences of treatment.
Although our primary interest is in the initiation of rigorous, developmentally focused medication trials in autism, the need for extensive animal research in this field cannot be overstated. The use of animal models may allow us to characterize developmental windows for treatment and evaluate the appropriate duration of treatment in ways that are impossible in human studies. Further, animal models will provide the opportunity to rigorously define the relationships between treatment and changes in activity of different neuromodulators and in brain structure in ways that are not possible in humans. Such information may facilitate the design of better treatments. The use of different genetic variants (in mice) may allow us to formulate ideas about treatment specificity among subgroups of autistic individuals.
However, there are a number of prerequisites to using such models. First, it will be essential to meticulously define periods of developmental equivalence between mice, nonhuman primates and humans. It will be crucial to examine the events that are the target of interventions: 1) synaptogenesis, 2) synapse refinement, and 3) the development of the integrative pathways presumed to be impaired in autism. This task is complicated by the disparities between primate and rodent development particularly with regard to extended plateaus of synapse refinement observed only in primates. There is also extremely limited information about the time course of these events in humans. In addition, it will be essential to characterize the comparative pharmacokinetic and pharmacodynamic properties of the candidate agent in animals and children, rather than adults. Further, animal studies should use pharmacologically relevant, rather than excessive, doses of the therapeutic agents that are administered by mouth or transdermally if possible. Blood levels of the medication across species are likely to be more informative than simple mg/kg or mg/body surface area. Although CSF levels would provide the most relevant comparison, it is not realistic to measure such levels in autistic children. Further, it will be important to continue recent efforts to improve the quality of behavioral assessments in animals models (Garner et al 2006
; Nadler et al 2004
To the extent feasible, advances in lower animals should be extended to non-human primates. Nonhuman primate models allow greater control of environmental factors than human clinical trials. Further, because nonhuman primates mature more quickly than humans, late emerging adverse effects can be detected in a shorter period of time. Thus, nonhuman primate models may be useful to refine the optimal time course of promising human treatments and to define the late-emerging adverse effects. Although the costs of non-primate human studies are great, they are probably less than clinical trials in humans and may pose fewer ethical dilemmas.
Future human clinical trials
Integrating younger children into ongoing and currently planned trials is likely to yield interpretable data about developmental effects most quickly. In trials that include children with autism across the age range (e.g. 18 months to 18 years), it will be possible to examine the correlation between age and both beneficial and adverse effects of treatment. Identification of such relationships will facilitate subsequent trials that more specifically test developmental intervention hypotheses. Trials should be of sufficient size to stratify for known confounding factors such as gender, regression, language, and cognitive phenotypes (Bradford et al 2001
; Schellenberg et al 2006
). Comprehensive assessment of potential adverse effects will require active review of body systems and developmental processes rather than volunteered reports of side effects. Further, it is essential that these trials develop a mechanism for assessing late emerging adverse effects as well as potential late emerging or enduring benefits.
If an agent is being studied with explicit developmental aims, it is essential that the period of double-blind treatment be sufficiently long to identify developmental changes. Further, it will be important to improve assessments of core symptomatology, designing assessments that are sensitive to change over time. Initial attempts to do this have been undertaken (Cohen et al 2003
), but more are needed, particularly assessments that involve direct observation of the child. In addition, it will be extremely valuable to develop biologic markers of treatment response; functional magnetic resonance scans that are temporally linked to treatment may have promise in this regard. Discontinuation trials that help to define the necessary duration of treatment will be important in minimizing risks and optimizing safety. Developmentally focused, trials in children with single-gene disorders with high prevalence of autism or features of autism, such as Fragile X syndrome, Tuberous Sclerosis, and SLOS, may be particularly useful if sufficient participants can be enrolled. Such trials would have the advantage of etiologic homogeneity, but may not generalize to a majority of individuals with idiopathic autism.
Key ethical issues include: 1) balancing the desire to constrain adjunctive therapies in order to maximize power to detect meaningful drug effects with the need for adjunctive treatments; 2) the use of potent agents in children; and 3) denying potentially effective treatment to participants in the placebo-arm. These concerns are potentially heightened in young children for whom developmentally directed treatments are likely to be most salient. For instance, the potential benefits of adjunctive treatments such as EIBI are likely to be greater for very young children than for older ones because their brains are more plastic. For instance, there are repeated examples of medications having far greater toxicity in the very young, so potential risks may well be increased. Issues related to safety are heightened because very few agents are likely to be approved for use in children and pharmacokinetic testing very seldom includes the youngest children. However, it is also important to remember that there has been a tremendous increase in the use of these agents in children and adolescents with autism clinically despite the absence of any systematic safety information (Witwer and Lecavalier 2005
). These issues are likely to be heightened further if a medication’s benefits are limited to enhancing development. In that case, it may never be possible to get an indication in adults so that studies in children can proceed in the traditional manner. If phase II/III trials of such agents are to proceed, intensive and unbiased safety monitoring will be required. Additional discussion of the challenges involved in pediatric autism trials is provided in several recent reviews (Hollander et al 2004
) (Aman et al 2004
) (Anderson et al 2004a
Potential strategies to minimize the risks to participants exist. Clear discontinuation guidelines in response to clinical deterioration must be in place. Likewise, participants should be informed of their randomization status as soon as their participation is completed by someone who is not involved in study assessments or data interpretation. Further, placebo-arm participants would be guaranteed access to ‘active’ treatment either at the conclusion of their study regimen or early termination of a placebo-arm in order to insure they had access to the potential benefits of the treatment. It is acknowledged that, if the treatment is only effective during a limited period of development, these benefits may be diminished by the delay in initiating the treatment. In addition, given the clear benefits of early environmental and educational interventions, participants could be allowed access to these therapies as long as there were efforts to match their use in the treatment and control groups and they were carefully documented. Properly designed and executed RCTs provide a safe environment in which interventions can be rigorously evaluated for safety and efficacy (March et al 2004
; Sandler 2005
) without compromising the best interests of the pediatric participant or quality of the science.
The evidence that autism is a neurodevelopmental disorder which begins in utero or during the early postnatal development is extensive. Further, there is increasing awareness of very early childhood changes in autism. Intervening in autism while the brain is still plastic may provide important benefits less likely with later treatment. Indeed, the most widely accepted therapy in autism, early intensive behavioral intervention, is based on this rationale. However, it is essential to develop a broader range of therapeutic options for use during this critical developmental period. Pharmacologic interventions are particularly promising because they may be more accessible to a larger number of affected children and may be more efficacious in different subgroups, especially those with low-functioning autism. Advances in our understanding of autism and normal neurodevelopment have suggested a number of agents that may positively impact experience-dependent development in autism. Further, there are extensive and expanding networks of investigators available to develop and test the utility of promising interventions. It is essential that we undertake the translational research necessary to make early pharmacologic interventions in autism a reality.