Tofersen for SOD1-ALS: Treating on a Biomarker After the Trial's Own Symptom Endpoint Missed
A single patient, newly confirmed with a rare inherited form of ALS. The disagreement isn't about whether tofersen lowers a lab value — it does — it's about whether that's reason enough to start a drug whose own pivotal trial didn't show the thing patients actually feel.
R.M., a 52-year-old land surveyor, has spent the past six months watching his right hand fail him in ways that started as an inconvenience — buttons that no longer cooperate, a compass he keeps switching to his left hand — and became something harder to explain away once his forearm began visibly thinning. His father died of ALS at 58, a fact R.M. had filed away as bad luck rather than warning until his own weakness sent him back to the neurologist his father once saw. Genetic testing confirmed a pathogenic SOD1 variant, the same gene his father's own workup had flagged only in hindsight. R.M. is otherwise healthy — normal renal and hepatic function, no prior neurologic disease — and his ALSFRS-R functional score today sits at 41 of 48, still working, still driving, weeks into a diagnosis he has already researched obsessively online.
What he found is tofersen, and what he's asking the team to explain is why a drug he read about as approved for his exact mutation comes with an asterisk. VALOR, the drug's 28-week pivotal trial, missed its primary endpoint: ALSFRS-R decline in treated patients wasn't significantly slower than placebo's. What it did show clearly was a roughly 35% reduction in CSF SOD1 protein and a 55% drop in plasma neurofilament light — a marker of ongoing axonal injury that rises long before functional decline is measurable on any rating scale. The FDA's own advisory committee split 3 votes to 5, with one abstention, against calling that convincing evidence of the clinical benefit patients actually feel. The agency approved tofersen anyway, on the accelerated pathway, betting the neurofilament signal is reasonably likely to predict a functional benefit VALOR itself was underpowered and too short to show directly. R.M.'s neurofilament, drawn today, is already elevated — a number with a treatment attached to it, and a real argument about whether that's enough.
Weighing a biomarker approval against a symptomatic diagnosis
Start tofersen now, and follow his neurofilament level as our real-time readout of whether it's working. I know VALOR's own ALSFRS-R endpoint came back negative at 28 weeks — that's real, and I'm not going to argue it away. But this is a rapidly fatal, irreversible disease, and the trial measured a functional scale over six months in a drug whose actual job is slowing molecular injury that shows up on neurofilament light months before ALSFRS-R can register it.
VALOR still showed a roughly 35 percent drop in CSF SOD1 protein and a 55 percent drop in plasma neurofilament — the toxic gain-of-function pathway itself responding to the drug, not a soft surrogate invented to save a failed trial. And the combined analysis of VALOR with its own open-label extension — reported by Miller and colleagues in 2022 — showed a possible signal favoring the early-start cohort over the delayed-start cohort on clinical endpoints, which is the closest thing we have to a timing argument. I want to be careful here: ATLAS, the presymptomatic-carrier trial, was built to test exactly whether treating before clinical onset delays symptom emergence, but it is still ongoing and has not reported that result. I'm arguing from the early-start/delayed-start signal, not from a finding ATLAS hasn't delivered yet. R.M.'s neurofilament is elevated today, and he's early in his disease course — every month we wait for a functional trial VALOR was never long enough to deliver is a month of motor neurons we don't get back.
You're right that the biomarker signal is real and mechanistically on-target — I'm not disputing that tofersen does what it's supposed to do to SOD1 protein and neurofilament. What I'm not willing to do is treat that as equivalent to the FDA's own advisory committee voting 3 to 5, with one abstention, against convincing evidence of clinical benefit on the endpoint that actually describes how a patient functions. This drug is dosed intrathecally, every four weeks after loading, indefinitely — and VALOR's own safety data include aseptic meningitis, myelitis, and elevated CSF protein with pleocytosis in a meaningful minority of treated patients.
And I'd go further on ATLAS than you just did: it isn't only that R.M. is symptomatic rather than a presymptomatic carrier under a research protocol — it's that ATLAS has no result to extrapolate from yet. Leaning on a trial's design as though it were a trial's finding is exactly how an accelerated-approval drug accumulates a reputation its own evidence hasn't earned. The early-start signal you're actually relying on is a post-hoc comparison in an open-label extension, not a randomized result. At roughly $180,000 a year with a real procedural burden attached, I'd want more than a plausible mechanism before committing him to this indefinitely.
There's a version of this that doesn't require either of you to be fully right. R.M.'s ALSFRS-R is 41 of 48 — early, still working, still driving — closer to the presymptomatic end of the spectrum than to VALOR's more advanced enrolled patients, even though he's clinically symptomatic. Post-hoc analyses of VALOR's own extension data suggested numerically less decline in the earliest, mildest treated patients, though the trial wasn't powered to confirm that as a formal subgroup finding.
I'd start tofersen now specifically because he's early, not as a blanket answer for every SOD1-ALS diagnosis regardless of stage — and I'd set an explicit measurement plan: neurofilament and ALSFRS-R every three months, with a real conversation at six months about whether the biomarker trajectory and his functional trend are moving together or apart. If they diverge, that's a signal this bet didn't pay off for him specifically, not proof the drug does nothing at all.
Agreed within the visit: start tofersen, with the loading regimen begun this week and the first neurofilament recheck at three months. Riluzole continues unchanged.
Not agreed, and explicitly carried forward rather than resolved: whether this same decision would look the same in a later-stage, faster-progressing SOD1-ALS patient. The Clinical Pharmacologist's underlying objection — that VALOR's own primary endpoint was negative — wasn't answered today, only judged less decisive given R.M.'s early stage specifically. The three-month monitoring plan exists in part to keep that question open rather than treated as settled by today's decision.