In a large-scale review of 4120 citations and 183 articles representing 153 studies on treatment effectiveness of autistic children , the authors (Warren et al, 2011) concluded that drugs have virtually no effects other than muting 'challengin behaviors' (noncompliance, hyperactivity, self-abuse).

Risperidone and aripiprazole seem to be the only kinds that produced the effects based on parent-reports (not actual observation and frequency counts).

Yet the positives are far outweighed by the negatives. The authors concluded "Both medications also cause significant side effects, however, including marked weight gain, sedation, and risk of extrapyramidal symptoms (side effects, including muscle stiffness or tremor, that occur in individuals taking antipsychotic medications). These side effects limit use of these drugs to patients with severe impairment or risk of injury."

As to other types of therapies, "Evidence is insufficient at this time to support the use of sensory or auditory integration, insufficient for speech and language interventions, and insufficient for CAM approaches." (CAM=alternative medicine)

What about ABA (applied behavior analysis) and the UCLA/Lovaas msdel? Their review is inconclusive - promising on some, disappointing on most.
There is little research on what happen when the children are 'mainstreamed', if the improvement will sustain or continue in a more normal settings.

In fact, the sparsity of long-term follow-up likely indicates no long-term effects of these treatments. This is not surprising when one considers the paradigm in which these behaviroal programs have been conducted - labor intensive, training intensive, and situation specific (university research clinic, insitutions, etc.). Such paradigm is not practicable and if it is, cannot be sustained overtime. Many parents and teachers burn out after a while of trying these behavioral procedures.