Mutations in distinct risk genes are now understood to contribute to autism spectrum disorder (ASD). SHANK3 is one important example; SHANK3 codes for a critical scaffolding protein in the postsynaptic density of glutamatergic synapses [1]. SHANK3 haploinsufficiency causes Phelan-McDermid syndrome (PMS) [2, 3], a common cause of ASD [4]. This study is the second of two projects examining the use of insulin-like growth factor-1 (IGF-1) as a novel treatment for PMS.Insulin-like growth factor-1 (IGF-1) is a commercially available compound that crosses the blood–brain barrier [5] and has beneficial effects on synaptic maturation and plasticity [6]. There are now several converging lines of evidence to support the use of IGF-1 in PMS based on results from a Shank3-deficient mouse model [7], neurons derived from patient derived pluripotent stem cells [8], and children with PMS [9]. In addition, evidence for the utility of IGF-1 and related compounds has been accumulating in other neurodevelopmental disorders associated with ASD, including Rett syndrome [10, 11] and Fragile X syndrome [12].

The primary aim of this study was to evaluate the safety and efficacy of IGF-1 vs. placebo in children with PMS using the Aberrant Behavior Checklist—Social Withdrawal subscale (ABC-SW) as a primary outcome measure. Our secondary aim was to explore effects on associated symptoms of ASD using measures of sensory reactivity, repetitive behaviors, and other aberrant behaviors.

Participants were enrolled in two consecutive studies, the first of which was previously published [9]. In both studies, treatment with IGF-1 or placebo was divided into two Phases (1 and 2). Participants were randomly assigned to receive either IGF-1 or placebo for 12-weeks in Phase 1 and were then switched to the other treatment condition (Phase 2) after a four-week wash-out period. The second study was completed in September, 2016 and results are presented herein, along with results combining the two studies.

The first study screened and enrolled nine children with PMS (6 females and 3 males) aged 5 to 15 years old (mean = 8.6; SD = 4.0) [9]. This second study screened 11 children and enrolled 10,one dropped out during the screening procedures. Participants were between 5 and 9 years old (mean = 6.5; standard deviation = 1.4); 6 participants were male and 4 were female. Nine of 10 participants met criteria for ASD based on clinical consensus using the Autism Diagnostic Observation Schedule, Second Edition [13], the Autism Diagnostic Interview-Revised [14], and the Diagnostic and Statistical Manual for Mental Disorders, Fifth Edition [15].

Participants were required to have pathogenic deletions or sequence variants of the SHANK3 gene for inclusion: six had terminal deletions and four had sequence variants. All participants were required to be on stable medication regimens for at least three months prior to enrollment.

Potential participants were excluded if any of the following were applicable: (1) closed epiphyses; (2) active or suspected neoplasia; (3) intracranial hypertension; (4) hepatic insufficiency; (5) renal insufficiency; (6) cardiomegaly/valvulopathy; (7) allergy to IGF-1; (8) patients with comorbid conditions deemed too medically compromised to participate.

IGF-1 is an aqueous solution for injection containing human insulin-like growth factor-1 (Increlex; Ipsen Biopharmaceuticals, Inc) produced by recombinant DNA technology. Placebo was normal saline prepared in identical bottles by the research pharmacy at the Icahn School of Medicine at Mount Sinai. Dose titration was initiated at 0.04 mg/kg twice daily by subcutaneous injection, and increased, as tolerated, every week by 0.04 mg/kg per dose to a maximum of 0.12 mg/kg twice daily. Medication was administered subcutaneously twice daily with meals and glucose monitoring was performed by parents prior to each injection and at bedtime.

Participants underwent comprehensive medical evaluations, including physical and neurological examination, routine hematology and blood chemistry, bone X-ray for bone age, electrocardiography, and echocardiography to determine eligibility for participation and repeated throughout the study to assess safety. Tolerability was monitored using a safety monitoring report form. Patients were monitored at weeks 2, 4, 6, 8, and 12 in both treatment phases. The most common side effects of IGF-1 are related to its insulin-like activity and hypoglycemic risks. Training was conducted with parents at baseline visits for drawing finger stick blood glucose levels and monitoring for signs and symptoms of hypoglycemia. Training in administering subcutaneous injections was also performed. Hypoglycemia was defined as glucose < 50 mg/dL.

The clinical outcome measures were administered at baseline and weeks 4, 8, and 12. Measures included the ABC [16], the Repetitive Behavior Scale—Revised (RBS-R; [17]), the Sensory Profile (SP; [18]), and the Clinical Global Impressions—Improvement and Severity Scales (CGI; [19]). Results from the SP and CGI were not previously reported in the first IGF-1 trial.

All statistical analyses were conducted in the statistical package R. Before testing for efficacy, we conducted analyses to test for potential bias in study design. We first performed a pre-test to check the assumption of negligible carryover effects between Phase 1 and Phase 2 of the crossover trial, as previously described [20], to ensure data collection was highly standardized across all patients and that the wash-out period was successful in removing significant carry-over effects. In brief, the sum of the values measured in the two phases was calculated for each subject and compared across the two sequence groups by means of a statistical test for independent samples. This trial used a randomized crossover design; thus, order (phase) of assessment is nested (repeated) within treatment and treatment is nested within subjects. Therefore, in terms of efficacy measurement, we applied a treatment × time interaction analysis using two-way repeated measures analysis of variance (ANOVA), which estimates the differential change in the two treatments on the outcome measures. We also specified an error term to account for individual variation. Subsequently, data derived from the first IGF-1 study (n = 9) [9], which was also a randomized crossover design, were assembled with the current study (n = 10) to generate a larger pooled data set (N = 19). Using this pooled data set, we applied the same treatment × time interaction analysis while controlling for baseline measurements and data set effects as covariates. Finally, because of the small sample size and to cast a wide net of informative treatment effects at threshold significance level, we applied an exploratory approach implementing a Mann–Whitney U-test on fold-changes observed at week 12 for each treatment group using the combined sample (N = 19).