Naltrexone for Self-Injurious Behavior in ASD/Intellectual Disability
Opioid-excess-theory mechanism with genuinely mixed, inconclusive trial evidence — not disproven, but not resolved either — and real tension when better-evidenced antipsychotics have failed or caused unacceptable effects in a safety-critical behavior.
Three generations of T.W.'s family sit in the room today, which is its own kind of evidence about how long this problem has been managed at home before ever reaching this office: his mother, who has provided most of his direct care for nineteen years, and his grandmother, who cared for an older cousin with a strikingly similar self-injurious pattern decades ago and recognized what she was watching almost immediately when it began in T.W. as a toddler. T.W. is 19, has autism and moderate intellectual disability, and has engaged in head-striking against hard surfaces since early childhood — a behavior that has required protective helmet use for years and has, twice in the past year, produced injuries serious enough for an emergency department visit.
Two antipsychotic trials, risperidone and later aripiprazole, each produced measurable reductions in the behavior's frequency, but both were discontinated: risperidone for a forty-pound weight gain over eighteen months that his medical team judged an unacceptable cardiometabolic trajectory, and aripiprazole for akathisia severe enough that it appeared to worsen his overall agitation rather than help it. His family has done their own reading and specifically raised naltrexone, built on what his grandmother remembers being tried, without much documented success, for his cousin in the 1990s. What the family doesn't have is a clear sense of how that history should actually inform a decision today, given that the drug's evidence record itself has never fully resolved one way or the other.
T.W.'s mother describes the helmet not as a piece of medical equipment at this point but as something closer to a fixture of his identity within the family — he has worn some version of one for so long that a cousin's young children, meeting him for the first time last year, reportedly assumed it was simply part of what he looked like, the way glasses might be for someone else.
She says this without visible distress, more as an illustration of how thoroughly the family has adapted around a problem that has never actually gone away, only been made survivable. That adaptation is itself part of why today's visit carries real weight beyond the immediate clinical question: nineteen years of managing around a behavior is a long time to keep hoping the next option might be different, and the family's patience, while genuine, is not the same as the problem having gotten any smaller.
A mechanism-plausible drug with a genuinely unresolved evidence record
I want to represent the actual evidence honestly, because "mixed" undersells how genuinely split it is. Willemsen-Swinkels and colleagues' double-blind, placebo-controlled crossover trial in 33 adults with intellectual disability found naltrexone had no benefit for self-injury and actually increased stereotypic behavior relative to placebo. Symons and colleagues' later quantitative synthesis across the accumulated small trials found naltrexone did produce measurable reduction in a meaningful subset of cases, though response was far from universal, and earlier work, including Kars and colleagues, reported real attenuation of self-injury in some patients. This isn't a drug with weak evidence pointing one direction; it's a drug where controlled trials have gone both ways.
Given that split record, I'd weigh two other facts more heavily than the average trial result: T.W. has already failed two mechanistically different antipsychotics for reasons unrelated to efficacy — both actually worked on the behavior itself before being stopped for tolerability — and naltrexone's opioid-antagonist mechanism is genuinely distinct from anything he's tried, not a third pass at the same dopaminergic pathway. A mixed evidence record for a mechanistically novel option, in a patient who has run out of better-evidenced choices, reads differently than the same mixed record would for a first-line decision.
I'd agree with that framing but want it explicit in what we tell the family: we're recommending a trial because the alternative options are worse, not because we expect this to work reliably — the Willemsen-Swinkels finding of a possible worsening effect in some patients is real and needs to be watched for directly, not glossed over because we're optimistic about a different mechanism.
A structured, time-limited trial with explicit stereotypy tracking alongside the primary SIB measure addresses that directly — if his stereotypic behavior worsens the way it did for a subset of the Willemsen-Swinkels cohort, that's a specific, pre-named stopping criterion, not a surprise finding requiring a fresh judgment call under pressure. Four to six weeks, clear outcome measures, and a genuine willingness to stop if the negative-trial pattern shows up rather than his own positive-trial pattern.
Agreed: a structured six-week naltrexone trial with dual tracking of self-injury frequency and stereotypic behavior, an explicit pre-named criterion for stopping if stereotypy worsens rather than improves.
T.W.'s grandmother asked directly whether this was "the same thing" tried for her nephew decades ago. The team's honest answer — the same drug, but a genuinely different, more structured trial design than what was likely available at the time, and no guarantee of a different outcome — was offered without either overpromising or dismissing her family's own remembered experience.