Dyne Therapeutics released additional one-year data on September 29 from the multiple ascending dose portion of ACHIEVE, its Phase 1/2 study of z-basivarsen in myotonic dystrophy type 1, and for the first time put the results next to a matched natural history cohort rather than only against baseline and placebo. The pooled dose group, drawn from three dose cohorts and numbering 25 to 26 participants, improved on hand myotonia, on a sit-to-stand test, on walking speed, on quantitative muscle strength, and on a patient-reported index that tracks the disease's cognitive and fatigue burden. The company reported no related serious treatment-emergent adverse events, with a safety cutoff of April 20, 2026.
The timing of the release is what makes it more than an incremental update. Three weeks ago, Novartis disclosed that del-desiran, the centerpiece of its $12 billion acquisition of Avidity Biosciences, failed the primary endpoint of the Phase 3 HARBOR study in the same disease. That endpoint was video hand opening time, the same measure of muscle relaxation that anchors Dyne's entire accelerated approval strategy. When the field's largest and best-funded program misses on a measure, the measure itself goes on trial, and every other sponsor using it inherits the doubt. Dyne's one-year dataset is the most direct evidence yet about whether that doubt belongs to del-desiran or to video hand opening time.
What the twelve-month data showed
The analysis pooled participants from the 3.4 mg/kg every-four-weeks, 5.4 mg/kg every-eight-weeks, and 6.8 mg/kg every-eight-weeks cohorts of the ACHIEVE dose escalation, including participants originally randomized to placebo who were re-baselined to the visit before their first active dose. That group ran from 25 to 26 participants depending on the measure.
On video hand opening time, or vHOT, the pooled group started with a mean of 8.2 seconds and improved by 3.2 seconds at six months, which the company described as sustained improvement. The placebo group in the same portion of the trial worsened by 0.4 seconds over the same window, a relative difference of 3.6 seconds. The measure itself is worth pausing on: a patient with myotonia cannot quickly release a closed hand, so a task that takes a healthy adult a fraction of a second stretches to eight seconds. A three-second improvement is the kind of change a person can feel when opening a jar. In the expansion cohort, the protocol narrows the test further to the middle finger, and the primary endpoint is defined as the change from baseline in that measurement at six months against placebo.
The functional measures moved in the same direction at 12 months. Participants improved by 1.2 seconds on the five-times sit-to-stand test and by 0.3 seconds on the ten-meter walk or run, with the walking measure diverging from the matched natural history cohort. That cohort came from END-DM1, an ongoing natural history study, with 41 to 46 participants propensity-matched to the treated group. Muscle strength rose by 4.8 percent of predicted on both the quantitative muscle testing total score and hand grip, again diverging from the natural history comparison.
The patient-reported results were the broadest of all. The Myotonic Dystrophy Health Index total score improved by 25.2 percent, and the improvement extended into six subscales covering cognitive impairment, sleep disturbance, fatigue, communication, emotional issues, and pain. That comparison used an unmatched natural history cohort of 410 people, a different and weaker design than the propensity-matched analysis used for the physical measures, and the company labels it accordingly.
The endpoint that carries the whole strategy
Dyne's path to market runs through the registrational expansion cohort of ACHIEVE, which enrolled 71 participants and completed enrollment with the 6.8 mg/kg every-eight-weeks dose as the registrational regimen. The primary endpoint of that cohort is the change from baseline in vHOT at six months. If the cohort succeeds, Dyne intends to file for accelerated approval in the third quarter of 2027. The company disclosed that the mean baseline vHOT across the 71 participants is 8.3 seconds, a number that matters because a sicker, slower baseline gives a drug more room to show a change.
That dependency explains why HARBOR's failure re-priced Dyne by roughly 29 percent in a single session earlier this month, even though the failure belonged to a competitor. RBC cut its price target to $30 from $35, and Raymond James cut its target to $37 from $40, according to reports at the time. The reasoning was mechanical: if video hand opening time cannot detect a real drug effect, an accelerated approval built on it becomes unattainable regardless of how good the underlying therapy is.
The counter-argument has also been forming. After the World Muscle Society data were released this week, Jefferies reiterated its buy rating and its $50 target, pointing out that the baseline vHOT in the expansion cohort closely resembles the baseline in the Phase 1/2 population. Similar baselines mean similar variance and, in the analyst's reading, a better chance of reaching statistical significance on the primary endpoint. That is a point about statistical powering, not about biology, and it cuts both ways. A drug whose approval case depends on the baseline characteristics of enrolled patients matching a prior cohort is a drug whose approval case depends on luck.
The two-trial structure mirrors the two approval standards. Accelerated approval can rest on a surrogate that is reasonably likely to predict clinical benefit, which is how vHOT functions in the expansion cohort, but it obliges the sponsor to confirm the benefit afterward. HARMONIA is that confirmation, and its primary endpoint is not vHOT at all but the change from baseline in the five-times sit-to-stand test at 12 months. So the accelerated filing rides on how fast a hand releases, and the traditional filing rides on how quickly a person rises from a chair, two measures of the same underlying muscle function taken in different trials. The design gives Dyne two chances, and it also means the company has to be right about the biology twice.
The limits the company wrote into its own release
Dyne's disclosure is unusually candid about what the pooled analysis can and cannot support, and the caveats are not boilerplate.
The first is dose. Only 8 of the 26 participants in the pooled group received the registrational dose for the full 12 months analyzed, and most received a lower dose for part of the period. The results therefore describe a group that was, for much of the study, underdosed by the standards of the regimen Dyne will file on. That is an argument for optimism about the expansion cohort, since the registrational dose should perform better than a mixture diluted by lower exposures. It also means the one-year data cannot be used to predict the size of the effect the expansion cohort will show.
The second is variance. The mean improvement in vHOT of 3.2 seconds came with a standard deviation of 3.5 seconds. In plain terms, the average moved, but the spread around that average was wider than the average itself, which means some participants improved substantially while others did not improve at all. A group-level result and an individual-level benefit are different claims, and the release supports the first, not the second.
The third is comparison design. The natural history analyses are not randomized comparisons, and two of them use different cohorts with different matching. The propensity-matched cohort is the stronger of the two, and it covers the physical measures. The patient-reported outcomes lean on the unmatched cohort of 410, which cannot exclude the possibility that people who volunteer for a treatment trial differ from people who do not in ways that shape how they answer a quality-of-life questionnaire.
The fourth is the sit-to-stand endpoint, which had no natural history comparison at all because END-DM1 does not include that test. That matters because the five-times sit-to-stand test is the primary endpoint of HARMONIA, the Phase 3 trial that Dyne intends to run as the confirmatory study for traditional approval.
Safety and the disease it is aimed at
The safety update covers participants from the dose escalation through an April 20, 2026 cutoff. The company reported no serious treatment-emergent adverse events related to z-basivarsen and described the profile as favorable. For an antisense oligonucleotide engineered to reach both muscle and the central nervous system, that is the claim that matters most in a disease where patients would take the drug for decades.
The disease itself is rare, progressive, and has no approved disease-modifying treatment. DM1 affects roughly 40,000 people in the United States and 55,000 in the European Union. It is caused by a mutation that produces toxic RNA, which disrupts the processing of other genes' messages throughout the body. The result is a disease that attacks muscle, the heart, breathing, sleep, cognition, and the gut at once, and kills people early. Z-basivarsen is designed as an antibody fragment that binds the transferrin receptor to ferry an antisense payload into cells, where it is meant to reduce the toxic RNA at its source. The functional measures in the trial sample that multi-system failure at a few points: how fast a hand releases, how quickly a person rises from a chair, how far they can walk, how they describe their own fatigue and thinking.
What the first quarter of 2027 decides
Three readouts will settle the question the field is now asking. The registrational expansion cohort reports topline in the first quarter of 2027 on vHOT at six months, and it is the trial that determines whether Dyne files for accelerated approval. HARMONIA, recruiting at more than 35 sites worldwide, reports later on the sit-to-stand test, and it determines whether the drug earns traditional approval and reaches patients outside the United States. Between them sits the endpoint question that HARBOR raised and this week's data partially answered.
The strongest reading of the new results is that vHOT responds to a drug that engages its target, and that del-desiran's miss was a fact about del-desiran. The weakest reading is that a 3.2-second average with a 3.5-second standard deviation, measured in a pooled group that mostly did not receive the filing dose, cannot tell you much about what a clean randomized test will show. Both readings will be tested by the same endpoint in the same disease within six months of each other, and the answer will determine whether accelerated approval for DM1 rests on a measure that works or on one that has now failed once.
Primary sources
- Dyne Therapeutics, Exhibit 99.1 to the September 29, 2026 Form 8-K, for the pooled dose group results, the natural history comparisons, the safety update, the expansion cohort baseline, and the regulatory timelines.
- Novartis, HARBOR study update, September 2026, for the del-desiran primary endpoint failure.
- Mavengity, coverage of the HARBOR failure, September 9, 2026, for the trial design and the $12 billion Avidity acquisition context.
- ClinicalTrials.gov, ACHIEVE trial record, for the trial design and registrational dose selection.