Switching or Adding SMA Therapy After a Plateau on Nusinersen
A toddler on nusinersen since infancy has plateaued and is now losing function. Switch to risdiplam, add gene therapy on top of an SMN2 splicing modifier, or hold the current course longer — real practice with only limited safety and efficacy data behind any of the three.
D.K. is 20 months old, the only child of parents who moved across the state shortly after her diagnosis specifically to be near this clinic, and has been on nusinersen since she was six weeks old — started presymptomatically after an older cousin's diagnosis prompted family testing. For fourteen months her trajectory looked like the one presymptomatic treatment was supposed to buy: she sat independently at 8 months, pulled to stand at 13, and was cruising along furniture by 16. Those are not merely normal milestones; they are the specific outcomes NURTURE's presymptomatic three-copy children achieved and held, which is why her family understood the first year as evidence the drug was working rather than as a promise still being tested.
Over the past six months that trajectory flattened, then reversed. The clearest evidence is functional rather than numeric: she has stopped attempting to cruise at all and tires within minutes of supported standing, neither of which was true in the spring. Her CHOP-INTEND has fallen from 58 to 51 across two consecutive quarterly visits, which is consistent with that picture but is the weaker of the two signals and deserves to be read as such. CHOP-INTEND was built to avoid floor effects in profoundly weak infants; it compresses badly at the top of its range, so a child who was cruising a few months ago sits close enough to its 64-point ceiling that it registers real loss only sluggishly. A seven-point drop at her motor level almost certainly understates what her parents are describing. She turns two in four months, when the Hammersmith expanded scale becomes appropriate and will track her considerably more sensitively.
The confounders that usually explain a decline are absent: no missed doses, and an anti-drug antibody panel drawn last month came back negative. What remains is a genuine question about the drug. A sustained reversal after fourteen months of documented gains is a different clinical entity from a plateau that never showed early response, and only the second is well described in nusinersen's own SHINE extension, where some children plateau and later resume gains. Her pattern is the first, which is rarer and less well characterized in the follow-up literature than either the family or the team would like. That leaves two paths — switching to a mechanistically different agent, or adding one on top — and the honest answer about the evidence behind either, for a child in exactly her situation, is not very much.
At the six-month reassessment, after a documented decline
I want to be careful about calling this drug failure after one flat-to-declining six-month window. SHINE's own long-term cohort has documented children who plateau for a period and then resume gains — SMA motor trajectories on treatment are not a straight line, even in responders. She has no missed doses, no antibody development, and fourteen months of real, unambiguous gains behind her. I'd want to see this hold for another quarter, with tighter functional tracking in between, before treating it as evidence the drug has stopped working rather than a genuine but temporary plateau.
If this were her first assessment showing no response at all, rather than a reversal after real documented gains, I'd be having a very different conversation with you both right now.
You're right that a single plateau shouldn't trigger a switch reflexively, and I take the SHINE precedent seriously. But this isn't a plateau — it's a documented reversal after fourteen months of real gains, on two consecutive assessments, with clean adherence and a negative antibody panel ruling out the most obvious confounders. And I'd add that the instrument we're arguing over isn't the one that should be carrying this. CHOP-INTEND is ceiling-compressed for a child who was cruising; the seven-point drop understates what her parents are describing. As for where to go, JEWELFISH is the study that enrolled patients previously treated with nusinersen or onasemnogene and switched them to risdiplam — it was a safety and pharmacodynamic study rather than a functional-outcome trial, so I won't oversell it, but it established that the switch raises SMN protein in previously-treated patients, and its systemic distribution reaches non-CNS SMN-deficiency effects that nusinersen's intrathecal delivery doesn't touch. And practically: she's 20 months old with no scoliosis yet, but that's exactly the window in which switching away from repeated lumbar punctures is easiest, before spinal instrumentation or curvature makes the procedure itself a growing risk.
'Wait another quarter and watch more closely' sounds conservative, but it isn't free — if this is a genuine drug-response failure and not a temporary plateau, we'd be spending three more months of irrecoverable motor neuron loss confirming what the trend already suggests, on the theory that SHINE's plateau pattern applies to her specifically when we don't actually know that it does.
I don't think switching and continuing are the only two options on the table, and I want to name the third honestly rather than let it go unsaid because it's the least comfortable one. Onasemnogene abeparvovec works through gene replacement, a genuinely different mechanism from either splicing modifier — there's a real biological argument for additive SMN protein restoration in a child whose splicing-modifier response has stalled, rather than trading one splicing strategy for another. I want to be direct that this is the thinnest evidence base of the three paths we're discussing today — small case series, real hepatotoxicity-stacking risk on top of whatever nusinersen or risdiplam contributes, and insurance approval for combination therapy in a non-trial setting is genuinely uncertain.
I'm not proposing this as the plan for today. I'm proposing that if the switch to risdiplam doesn't produce a response within a defined window, this should already be on the table as the next conversation rather than something we discover only after that fails too.
Agreed: discontinue nusinersen and start risdiplam, with reassessment at 12 weeks rather than the usual 6-month interval, given how much motor function is already in question. Not agreed, and stated plainly rather than papered over: whether a switch that doesn't produce stabilization by 12 weeks should lead next to onasemnogene abeparvovec as an add-on, or to a longer trial of risdiplam alone first — the geneticist and pharmacologist left with genuinely different thresholds for how much additional waiting is reasonable once one switch has already failed to help.