In children with autism, is intravenous secretin more effective than placebo in improving social skills, communication, behaviour or global functioning?

In 1998, a case series (1) and television program (NBC Dateline) described children whose symptoms of autism improved remarkably after secretin infusion for a gastrointestinal procedure. These and other testimonials sparked a huge demand for off-label secretin use. Multiple Web sites started advertising secretin infusions at great costs to parents. Autism interest groups were claiming that improvements were seen in 70% of patients. Then in 1999, alarmed by the claims of several thousand recipients of secretin for autism, there was a call within academia to provide better quality evidence.

Secretin is no more effective than placebo in treating children with autism, including improving social skills, communication, behaviour or global functioning (Grade of recommendation: A, based on several randomized controlled trials [RCTs]).

A literature search of PubMed, EMBASE, MEDLINE and the Cochrane Library identified 14 RCTs (2–15) examining the efficacy of secretin versus placebo published from 1999 to 2003 using the following search terms: autism, autistic disorder and secretin. The methodological quality of these trials was generally strong as assessed using the Jadad score (16), which examines randomization, double-blinding, and description of withdrawals/dropouts (median Jadad score 4 out of 5, interquartile range 3 to 5).

Many of the studies highlighted different dimensions to the question and addressed previous criticisms: nine studies used porcine secretin (3–7,11–13,15), five examined the human synthetic form (2,8,9,14,15), nine studies used cross-over designs (3,5,7–11,14,15), and multiple doses were given in two studies (12,14). Several studies examined possible subgroup responders; patients with gastrointestinal symptoms (5–7,9,12,15), different degrees of autism severity (7,8,12,13,15), and those for whom other medical therapies were allowed (3,6,8,10–12) or excluded (15). Secretin doses used were similar to those of the original reports. None of these trials found evidence to support the use of secretin in autism. For a more in-depth review, three studies with the highest methodological quality and the largest study populations were selected.

Dunn-Geier et al (6) conducted an RCT of 95 children (ages two to seven) randomized to receive a single dose of porcine secretin (2 Cu/kg) or saline placebo. Autism diagnosis was based on a Diagnostic and Statistical manual of the American Psychiatric Association’s (17) criteria score of six or higher and on a Childhood Autism Rating Scale score (CARS) of 30 or higher (18), as well as clinician judgment. Mean CARS scores indicate these were children at the moderate to severe end of the autism spectrum. This study was methodologically strong, with adequate allocation concealment, appropriate methods to ensure double-blinding, complete follow-up for all patients, and an intention-to-treat analysis. The groups were similar at the start of the trial on important prognostic factors except that the average baseline Preschool Language Scale-3 (PLS-3) (19) score was higher in the placebo group. A sensitivity analysis, adjusting for baseline language skills, did not change the study’s conclusions. At baseline and three weeks after treatment, patients were assessed using the CARS and Autism Behavior Checklist (ABC). Receptive and expressive communication skills were measured using the PLS-3 (19).

The authors found no statistically significant differences between secretin and placebo groups in any outcomes at three weeks. Interestingly, both groups tended to show some improvements on total PLS-3, CARS and ABC scores.

Roberts et al (12) addressed criticisms of previous studies by using more than one dose and assessing gastrointestinal symptoms and differing cognitive levels. Sixty-eight children (ages two to seven years) were recruited and 94% completed the study. A diagnosis of autism was confirmed using ‘gold standard’ diagnostic measures, the Autism Diagnostic Interview – Revised (20), and Autism Diagnostic Observation Scale (ADOS) – Generic (21). Two equal groups (n=34) were allocated to receive two doses of porcine secretin (2 Cu/kg) or saline placebo given six weeks apart. Three weeks after each dose there was an assessment of autistic symptoms (ADOS, ABC, Behavior/Side Effect Rating Scale [21]), language (PLS-3, ADOS) plus a visual-spatial attention task and a gastrointestinal symptoms questionnaire designed specifically for this study. Cognitive function (Leiter [22] or Vineland [23]) was assessed at baseline and final follow-up. This study conducted subgroup analyses based on low versus high IQ, presence of gastrointestinal symptoms or history of developmental regression. The sample size had a power of 0.8 to detect a moderate effect size on the PLS-3 at a 5% level of significance.

The study found that, even with two doses of secretin, there were no statistically significant differences on any measure. Secretin did not appear more beneficial for the subgroups divided by low or high cognitive level, presence of gastrointestinal symptoms or history of regression. Both groups did show improvements on receptive and expressive, but not total language.

Levy et al (8) evaluated a single dose of human synthetic secretin (HSS, 2 Cu/kg) versus saline placebo (SP). Sixty-eight children (ages 44 to 104 months) with Autism Diagnostic Interview – Revised confirmed diagnoses of autism were recruited; 91% completed the study. Autism severity was classified using the CARS. Two groups of 31 children were randomized to receive either HSS/SP or SP/HSS with a six-week intervening washout period. Baseline patient characteristics were similar except one group had a greater than expected number of ‘severe’ classifications on the CARS to a small, but statistically significant degree. The analysis therefore included CARS severity category as a covariable. Outcome measures taken at baseline and weeks two and four postinfusion included part of the Communication and Symbolic Scale (24), Ritvo Real-Life Rating Scale (25), weekly Global Rating Scale (designed for this study) by parents and teachers and daily log of gastrointestinal symptoms.

Levy et al’s (8) study also concluded that when compared with placebo, secretin treatment is not effective in changing behaviour and communication in children on the autism spectrum.

In a second part of this study (26), parents were asked to guess whether their child received secretin or placebo. Authors thought there may be dimensions of behaviour detectable by parents, but not by standardized measures. Results were obtained on 60 of the 62 subjects, finding that parents guessed right 50% of the time. The study concluded that in a controlled setting, parents of young children with autism were unable to distinguish the short-term behavioural effects of secretin from placebo.