For teams designing an autism spectrum disorder drug program, the choice of primary endpoint is the point at which a development plan either aligns with regulatory precedent or diverges from it. Because the FDA's approval record in this indication is narrow, the agency's prior decisions carry outsized weight in defining which outcome measures, rating scales, and target populations it has been willing to treat as evidence of efficacy.
The analysis below maps the endpoint architecture the FDA has actually accepted in reviewed and approved autism programs, the specific instruments behind those endpoints, and the distinction between the approved indication and the core-symptom claims that remain unapproved. It draws on approved labels, review documents, and programs that failed, to show where the regulatory line has been drawn.
Want Rhizome's help on your own question? Try it for free.
Efficacy endpoints FDA has accepted in autism spectrum disorder drug trials
The FDA-approved regulatory record for autism is narrow and specific. To date, the agency has approved drugs for a single autism-related indication, "irritability associated with autistic disorder," and never for the core symptoms of autism. That distinction drives everything about which efficacy endpoints have been accepted. The two approved programs, risperidone (Risperdal) and aripiprazole (Abilify), converged on the same primary outcome measure and the same clinician-rated global scale, and every review that has touched the core-symptom question has reiterated that no drug is approved for those domains. This article summarizes the accepted endpoints, the scales behind them, and what reviewed programs (including failures) show about what FDA will and will not treat as an acceptable outcome measure.
The two approved programs share one endpoint architecture
Both approved drugs carry the identical indication, "irritability associated with autistic disorder," in pediatric patients, and both established efficacy with the same endpoint structure 1218.
Primary endpoint: the Aberrant Behavior Checklist, Irritability subscale (ABC-I). In every pivotal trial for both drugs, the primary efficacy measure was the change from baseline to endpoint in the parent/caregiver-rated ABC-I score 14124. The ABC-I is a defined, quantitative instrument: the full Aberrant Behavior Checklist has 58 items, each scored 0 (not a problem) to 3 (severe problem), and the Irritability subscale is the sum of 15 items with a range of 0 to 45 6162. It captures the behavioral targets FDA accepted as the treatable construct: aggression toward others, deliberate self-injury, temper tantrums, and quickly changing moods, along with screaming, crying over minor annoyances, and demanding immediate satisfaction 56561.
Co-primary or key secondary endpoint: a Clinical Global Impression scale. The clinician-rated CGI supplied the second pillar. For risperidone, the Clinical Global Impression–Change (CGI-C) was a co-primary outcome in one of the pivotal studies 131618. For aripiprazole, the Clinical Global Impression–Improvement (CGI-I) scale was used alongside the ABC and is described in the clinical review as a key secondary endpoint of study CN138178 126. The CGI is a seven-point clinician judgment: CGI-S rates overall severity from absent to extremely severe, and CGI-C/CGI-I rates change from very much worse to very much improved 63. In practice, the label characterizes responders as those rated "very much improved" or "much improved" 1214.
The pairing is deliberate. A parent/caregiver-rated symptom count (ABC-I) establishes the magnitude of behavioral change, and an independent clinician global rating (CGI) confirms that the change is clinically meaningful. FDA accepted that combination as sufficient evidence of effect for the irritability indication.
Risperidone: the first approval
Risperidone was the first drug approved for irritability associated with autistic disorder 3738. Efficacy rested on placebo-controlled pediatric trials using ABC-I as the primary outcome and CGI-C as a co-primary measure 1618:
- An 8-week trial (n=101) of risperidone 0.5 to 3.5 mg/day produced significant improvement in both ABC-I and CGI-C versus placebo; 54% of risperidone patients versus 18% on placebo were "very much improved" or "much improved" at Week 8 (p<.001) 121416.
- A second 8-week trial (n=55) at 0.02 to 0.06 mg/kg/day again significantly improved ABC-I versus placebo 1214.
- A 6-week fixed-dose trial (n=96) used mean change in ABC-I from baseline to Week 6 as the primary endpoint 13.
The FDA review was explicit about what this approval did and did not represent: it stated there was "no evidence that risperidone treats the mental retardation or pervasive disruption of childhood development that are the core features of Autistic Disorder," and that the application was deliberately framed around relief of irritability-like symptoms rather than core autism features 39.
Aripiprazole: replicating the template
Aripiprazole's efficacy was established in two 8-week, placebo-controlled pediatric trials in children and adolescents aged 6 to 17, again with ABC-I as the primary endpoint and CGI-I as the accompanying measure 124:
- One 8-week trial (n=98) of flexible-dose aripiprazole 2 to 15 mg/day (mean 8.6 mg/day) significantly improved ABC-I and CGI-I versus placebo 2.
- A second 8-week trial (n=218) compared fixed doses of 5, 10, or 15 mg/day against placebo 2. The clinical review reports clinically meaningful and statistically significant improvement versus placebo emerging within 1 to 2 weeks and continuing through Week 8 on the primary measure, with a statistically significant CGI-I treatment difference for all aripiprazole groups in study CN138178 6.
The aripiprazole record confirms the template rather than extending it: same indication wording, same ABC-I primary endpoint, same CGI-I confirmation, same 8-week controlled design in the pediatric population.
The core-symptom boundary: what FDA has repeatedly said is unresolved
The most consequential lesson for endpoint strategy sits at the boundary between irritability and the core autism domains (social communication deficits and restricted/repetitive behaviors). FDA reviews draw that line sharply.
The risperidone review stated plainly that "there are no approved treatments for autism" and that off-label use had focused on irritability-like symptoms rather than core features 39. A memantine (Namenda XR) review reiterated it years later: "There are currently no medications specifically approved in the US for the treatment of any of the core domains of autism," explicitly contrasting this with risperidone and aripiprazole, which are approved only for irritability, not core symptoms 3233. Another passage in the same review states there is "currently no available Agency-approved treatment for the treatment of core symptoms of autism" 36.
The memantine program itself illustrates the endpoint problem for core symptoms. The sponsor evaluated core features across two studies using clinician-rated instruments including the CATS (measuring social interaction and communication) and CAASTS (measuring stereotyped behaviors, restricted interests, and daily function), alongside the Social Responsiveness Scale (SRS) 303134. The review concluded that the treatment effects observed were not related to memantine 30. The program is a case study in what an unsuccessful core-symptom endpoint package looks like: appropriate instruments, but no drug-attributable effect.
What the broader trial landscape reveals about acceptable measures
Registered ASD drug trials use a wider menu of scales than the two approvals, mapped to the symptom domain each program targets. This is the vocabulary a sponsor is expected to select from, and it shows how the field segments outcomes 47505860:
- Irritability / disruptive behavior: ABC-Irritability remains the anchor, targeting aggression, self-injury, and agitation; it was the primary endpoint in an arbaclofen ASD study 5860.
- Overall clinical status: CGI-S (severity) and CGI-I (improvement) serve as primary endpoints in some programs, including an L1-79 trial where CGI severity/improvement is the primary measure and a gene-therapy protocol keyed to CGI-S reduction 4750.
- Social communication (core): the Social Responsiveness Scale (SRS/SRS-2) captures reciprocal social communication, social awareness, cognition, and motivation, and the Autism Diagnostic Observation Schedule (ADOS) captures observed social and communication deficits 55585952.
- Restricted/repetitive behaviors (core): the Repetitive Behavior Scale–Revised (RBS-R), and the Children's Yale-Brown Obsessive Compulsive Scale modified for PDD (CY-BOCS-PDD) target compulsive and repetitive behavior 51565760.
- Adaptive function: the Vineland Adaptive Behavior Scales (VABS-II) measure adaptive/functional outcomes and appear as secondary endpoints across programs 475060.
- Composite core-symptom instruments: newer trials use the Autism Behavior Inventory (ABI), where the ABI Core Domain Score (Social Communication plus Restricted/Repetitive Behavior) served as a primary efficacy endpoint 57.
None of these core-symptom or composite measures has yet supported an FDA approval. Their presence in registered protocols reflects where sponsors are trying to move, not what the agency has accepted as establishing effectiveness.
What this means for endpoint strategy
For any program targeting the existing indication, the accepted path is well-defined and reproducible: an 8-week (or comparable), placebo-controlled pediatric trial with parent/caregiver-rated ABC-I as the primary endpoint and a clinician-rated CGI (CGI-I or CGI-C) as a co-primary or key secondary measure, with responder analyses framed around "much improved"/"very much improved" 121416126. That is the only endpoint architecture with a track record of approval.
For core symptoms, the regulatory precedent is a repeated statement that no treatment is approved and, in the memantine case, a specific set of core-symptom instruments (SRS, CATS, CAASTS) that were deployed but did not demonstrate a drug effect 30313234. A sponsor pursuing a core-symptom claim is therefore working in a space where the acceptable outcome measure has not been settled by an approval, and where the burden is to show a drug-attributable, clinically meaningful effect on a validated social-communication or repetitive-behavior instrument. The reviewed record defines the accepted endpoint for irritability precisely, and leaves the core-symptom endpoint an open question that a reader would do well to probe against a specific target product profile.