Presymptomatic SMA Treatment After Newborn Screening: The Three-Copy Question
A newborn screens positive for SMA with three copies of SMN2 — a genotype whose natural history genuinely ranges from wheelchair-dependent to lifelong ambulatory. Treat every screen-positive infant immediately, or does three-copy SMA earn a different conversation?
Baby G. is nine days old, born by uncomplicated vaginal delivery to first-time parents who had never heard the term spinal muscular atrophy until a phone call two days ago told them their state's newborn screening panel had flagged their daughter as SMN1-deletion positive. She is, by every bedside measure available at nine days, a normal infant — good tone, strong cry, feeding well at breast, reflexes intact, and a CHOP-INTEND of 64 out of 64, which is exactly what a presymptomatic infant should score and therefore tells the team nothing about what is coming. That normalcy is what makes today's conversation harder rather than easier.
Confirmatory testing shows three copies of SMN2, the single number on which the rest of this turns. Two copies predicts an infantile-onset course severe enough that early treatment is close to unanimous practice; four or more predicts a course mild enough that some centers still favor watching; three sits in the genuinely contested middle, historically spanning children who lose independent ambulation before age ten and adults who reached their sixties walking, undiagnosed the whole way. What is striking about her particular chart is how closely it matches the population that generated the strongest evidence available. SPR1NT's three-copy cohort had a median birth weight of 3.4 kilograms and a median molecular diagnosis at eight days; Baby G. weighs 3.4 kilograms and was diagnosed on day seven. She is not an extrapolation from that trial. She is, almost line for line, the infant it enrolled.
And those children did well: all fifteen stood independently before twenty-four months and fourteen walked. NURTURE reported the same pattern for presymptomatic nusinersen, RAINBOWFISH for risdiplam. But SPR1NT required treatment within six postnatal weeks, which means the window in which she matches that population is about a month wide, and the trials share a structural limitation no amount of follow-up will fix: every child in them was treated. None can answer what this family is actually asking, which is what would have happened to their own daughter untreated. Three-copy natural-history cohorts gathered before these drugs existed show real bimodal spread, and no biomarker available today — not her copy number, not her maximal CHOP-INTEND, not her normal exam — sorts a nine-day-old into one arm of that distribution or the other before treatment, or before symptoms, removes the chance to find out.
In the NICU consult room, day nine of life
I don't think we should let three copies talk us out of treating now. NURTURE, SPR1NT, and RAINBOWFISH weren't run on two-copy infants alone — SPR1NT's three-copy cohort was fifteen children, and every one of them stood independently, with fourteen walking — rates that simply don't exist in any untreated natural-history cohort. Every day between today and whenever symptoms would eventually appear is a day of motor neuron loss we cannot get back afterward. 'Watchful waiting' sounds neutral, but it isn't — it's a real choice with a real cost, it's just a cost that doesn't show up on an exam yet.
I want to be honest that this argument doesn't distinguish which of the three agents to use — only that whichever one we pick, the clock should start now rather than at her first missed milestone.
You're right that the treated-cohort data are strong, and I'm not arguing against treatment as a real option here. But I don't think 'three copies' is a diagnosis that tells us enough to make immediate treatment the obvious default the way it is at two copies. Historical three-copy natural history genuinely spans children who lose ambulation before age ten and adults who are still walking in their sixties, undiagnosed the whole time. Committing every three-copy screen-positive infant to what is, for two of these three agents, effectively lifelong or one-time-irreversible therapy treats a population that we know contains people who would have done fine without it.
The counterargument that early treatment simply protects everyone equally well doesn't hold up to the trial populations themselves — SPR1NT and RAINBOWFISH both enrolled presymptomatic infants precisely because nobody could tell in advance which ones would have stayed mild. That's not evidence every three-copy infant needed treating. It's evidence the trials, like us, had no way to sort them first.
I don't think we can cleanly separate 'when' from 'which drug' the way this conversation keeps doing. If we lean toward treating now, a presymptomatic three-copy infant is arguably the single best-case scenario for onasemnogene abeparvovec specifically — she has the most motor neurons left of any patient we'll ever treat with it, and the most years ahead to benefit from a durable, one-time transgene rather than a chronic splicing modifier that commits the family to years of ongoing decisions. If instead we lean toward watching, that's a real, defensible position too, given what the geneticist just laid out — but it should be a structured watch, with a defined re-evaluation interval and named CHOP-INTEND or motor-milestone triggers for stepping in, not an open-ended 'we'll see.'
I'll say plainly that I don't think this family needs us to resolve which of those two paths is correct in the abstract today. They need a specific plan, whichever path they choose, with a specific next appointment already on the calendar.
Not agreed, and not falsely smoothed over: the neonatologist and geneticist left this meeting recommending different defaults for three-copy infants in general, a disagreement that predates this family and will outlast this visit. What was agreed for Baby G. specifically, after her parents asked directly for the team's honest uncertainty rather than a false consensus, was a structured surveillance plan rather than immediate treatment — reassessment at four weeks with CHOP-INTEND and formal motor exam, an earlier visit if her parents notice anything at home, and an explicit understanding that either drug remains available and is not disadvantaged by waiting a few weeks longer than a two-copy infant safely could.