๐ Pharma Economics
0.99 Ataxia Points Separated the Groups at the Start. The Adjusted Difference Is 0.49.
Infilbray's approval rests on one combined test, p=0.0137. The ataxia difference is 0.49 points on a 16-point scale, and the interval includes zero.
Zero point nine nine.
Patients who stayed on imiglucerase started LEAP2MONO at a mean four-item ataxia score of 3.18, patients moved to oral Infilbray started at 2.19, and 3.18 minus 2.19 is 0.99 on a scale that runs from 0 to 16, before week 52 existed at all.
Adjusted gap: 0.49.
On November 18, 2026, FDA approved Infilbray (venglustat) for neurological manifestations of Type 3 Gaucher disease in adults and in pediatric patients weighing at least 15 kilograms, and Sanofi's release that evening printed a combined p-value of 0.0137 while leaving both score differences, and both intervals, off the page.
One test, two intervals
Table 6 of the prescribing information, repeated on Sanofi's label PDF, comes from a mixed model that already includes baseline as a covariate. Least-squares changes on the four-item SARA, gait, stance, sitting, and speech, each scored 0 to 4, are โ0.56 on the pill and โ0.08 on the infusion, and subtracting those printed means gives โ0.48 against a printed difference of โ0.49, interval โ1.29 to 0.32, a one-hundredth gap that is rounding rather than a second result.
Divide 0.99 by 0.49 and the arms began 2.02 times farther apart than the adjusted gap that followed; divide 0.49 by 16 and the gap is 3.06 percent of the scale; the interval width is 0.32 minus โ1.29, which is 1.61, and 1.29 divided by 1.61 puts 80.1 percent of that width on the better side of zero, with 0.32 divided by 1.61 leaving 19.9 percent on the worse side.
Zero sits inside.
Cognition clears zero where ataxia does not: 3.69 minus โ4.63, the two RBANS least-squares changes, is the printed 8.32, interval 0.75 to 15.89, and 4.63 divided by 8.32 is 55.6 percent of that gap from the infusion falling, against 44.4 percent from the pill rising. Baselines were 78.57 versus 68.36, a 10.21-point head start, 1.23 times 8.32.
O'Brien's 1984 paper placed ordinary least squares among five ways to combine endpoints and recommended the nonparametric rank-sum for general use, and the Biometrics abstract says that rank-sum controlled size and had relatively good power in the simulations, with generalized least squares as the alternative for normal data in moderate or large samples, against a file of 21 patients and 22.
ClinicalTrials.gov lists the two score changes as separate primary outcomes and lists spleen volume, liver volume, hemoglobin, and platelet count as secondary outcomes, while Sanofi's release calls one prespecified O'Brien ordinary-least-squares test the primary endpoint and names those four systemic measures without numbers, and none of the four appears in the label section that holds Table 6.
Week-52 means are not a paired subtraction on the infusion: SARA of 2.81 minus 3.18 is โ0.37 against a printed mean change of โ0.14, and RBANS of 66.14 minus 68.36 is โ2.22 against a printed โ3.52, while pill-arm SARA, 1.71 minus 2.19, matches โ0.48, and the label never says who is missing from which cell.
| Score | Infilbray, n=21 | Imiglucerase, n=22 | Arithmetic on the printed cells |
|---|---|---|---|
| SARA baseline, mean | 2.19 | 3.18 | 3.18 โ 2.19 = 0.99. Then 0.99 / 0.49 = 2.02 times the later adjusted gap |
| SARA least-squares change | โ0.56 | โ0.08 | โ0.56 โ (โ0.08) = โ0.48, against a printed โ0.49 (โ1.29 to 0.32) |
| SARA as a fraction of 16 | 0.49 / 16 = 0.030625 | 3.06 percent of the scale. Width 1.61; 1.29/1.61 = 80.1 percent of it lies on the better side of zero | |
| RBANS baseline, mean | 78.57 | 68.36 | 78.57 โ 68.36 = 10.21, and 10.21 / 8.32 = 1.23 times the adjusted gap |
| RBANS least-squares change | +3.69 | โ4.63 | 44.4 percent of 8.32 is the pill rising; 55.6 percent is the infusion falling |
| RBANS difference | 8.32 (0.75 to 15.89) | 0.75 / 8.32 = 9.0 percent, so the floor is about a tenth of the point estimate | |
Substrate, concentration, and the infusion that stopped
From the pill arm's own baseline, plasma GL-1 fell 73.1 percent and cerebrospinal-fluid GL-1 fell 64.3 percent, while plasma lyso-GL-1 fell only 15.2 percent with a standard error of 8.9. Then 1.96 times 8.9 is 17.44, and 15.2 minus 17.44 is โ2.24, so a normal approximation around that one biomarker includes no reduction. The same step on cerebrospinal-fluid GL-1, 64.3 minus 1.96 times 7.9, equals 48.8 percent and excludes zero, bounds that are ours rather than the label's.
Steady-state peak is 150 ng/mL and the area under the curve is 2,610 ngยทh/mL, so 2,610/24 is 108.75 ng/mL. Times the printed fluid-to-plasma ratio of 0.05, that peak is 7.5 ng/mL, or 32.6 percent of the IC50 of 23 ng/mL, and the average is 5.44 ng/mL, 23.6 percent of it. Nothing says which plasma level the ratio uses, and cerebrospinal-fluid GL-1 still fell 64.3 percent anyway.
Cerezyme's indication is the non-central-nervous-system manifestations of Type 1 or Type 3, and Infilbray's is the neurological manifestations of Type 3, and the trial switched people already stable on enzyme replacement for at least three years, with a modified SARA of at least 1, from the infusion onto tablets plus a placebo infusion, or kept the infusion and added a placebo tablet. NORD says enzyme replacement does not primarily reach the brain, a 2006 assessment said Cerezyme cannot cross the barrier, and a 2026 registry analysis calls that claim largely assumptive because some reports describe neurological stabilization on the infusion, which was still the control.
Splenomegaly was recorded in 3 of 21 on the pill and in 0 of 22 who stayed on imiglucerase, 14.3 percent where the table prints 14, and depression in 2 of 21 versus none. On another venglustat formulation, not approved for this disease, depression was 28 of 362, 7.73 percent, matching the printed 7.7, against 19 of 340, and 19 divided by 340 is 5.59 percent where the sentence prints 5.7. Suicidal ideation or behavior in that pool was 4 of 362, 1.10 percent printed as 1 percent, against none on placebo.
Section 8.4 counts 19 patients aged 12 to 17 in the study and 9 of the 21 pill recipients in that band, so 19 minus 9 is 10 adolescents on the infusion if both sentences count the same people. Dosing down to 15 kilograms also uses pharmacokinetics from 13 children aged 2 to 17 in an unapproved population, and the yearly eye exam rests on rat lens findings the label calls uncertain in children.
| Labeled regimen, 60 kg adult | Units per infusion | Infusions in 52 weeks | Vials opened | WAC for the year |
|---|---|---|---|---|
| 60 units/kg every 2 weeks, top of the range | 3,600 | 26 | 9 ร 26 = 234 | 234 ร $1,751.47 = $409,843.98 |
| 2.5 units/kg three times a week, bottom of the range | 150 | 156 | 156, one vial each time | 156 ร $1,751.47 = $273,229.32 |
Sanofi's Colorado disclosure puts the January 1, 2026 wholesale acquisition cost at $1,751.47 per 400-unit vial and says that figure is not a transaction price. Sanofi's dosing page repeats 2.5 units per kilogram three times a week up to 60 units per kilogram every two weeks. BioPharma Dive reported on November 19 that Infilbray's price is undisclosed, so dividing $409,843.98 by 8.32 would price a comparator dose the trial never stated.
European Type 3 prevalence sits near 0.02 to 0.04 per 100,000 in that review, and under 1 in 100,000 in the 2006 British assessment already cited, with a larger neuronopathic share in parts of Asia, and the efficacy file is 43 people, median age 19, 19/43 or 44.2 percent aged 12 to 17, median 14 years on enzyme replacement. Gold's adult RBANS, mean 100 and standard deviation 15 for ages 20 to 89, turns 8.32 into 0.55 of a standard deviation and 0.75 into 0.05, which the label does not claim for this younger sample.
What You Can Do
Do not stop an infusion because this pill was approved. Ask whether the prescription is added or substituted, and ask for hemoglobin, platelets, liver volume, and spleen volume, because those secondary results are missing from Table 6 and splenomegaly was 3 of 21 versus 0 of 22. Ask which score is the target: ataxia's interval includes no change, and cognition's floor is 0.75, then ask the pharmacy for a price, because none is public, against an illustrated $409,843.98 Cerezyme year for a 60 kilogram adult at the top of that label. Book a child's eye exam before the first tablet and yearly after. Use contraception through 11 days after the last dose, and report psychiatric history first.
Limitations
This does not prove the pill fails, that the combined test was the wrong primary, or what any named patient should do, and microdata were not reanalyzed. Infusion-arm means do not match baseline plus the printed change, biomarker bounds are ours rather than printed intervals, and the fluid ratio is not tied to a named plasma level. Dose, pill price, and the four systemic endpoints are absent. Dollar rows assume 60 kilograms, labeled extremes, single-dose vials, and list price with no rebate. Sanofi says the open-label phase is ongoing; the registry lists completion on October 2, 2025 and October 30, 2026, status active, not recruiting.
The Strongest Case Against This Analysis
Grant the objection its full strength: a first brain drug has to beat the infusion people already take, p=0.0137 is the prespecified bar, and requiring each interval to exclude zero is stricter than that protocol. Divide 0.49 by the pill arm's baseline of 2.19 and the gap is 22.4 percent of the points that arm started with, after a study that required a modified SARA of at least 1, while cerebrospinal-fluid GL-1 fell 64.3 percent and 43 patients is what this disease yields.
Count the falling arm.
That case is right about the comparator and the scarcity, and incomplete about 8.32. Both ataxia changes are improvements, the difference is 0.49, and the interval includes zero and a worsening of 0.32. Of the cognitive gap, 4.63 points are the infusion getting worse and 3.69 are the pill getting better, so a claim of an 8.32-point gain spends the control decline. Dividing by 2.19 is not a percent the label calculated. A minimum score of 1 explains the low baselines. It does not push zero out of โ1.29 to 0.32.
The Bottom Line
Read Table 6 before the p-value if the question is whether to replace an infusion that has held liver, spleen, and blood counts for a median of 14 years. Arms started 0.99 ataxia points apart, the adjusted difference is 0.49, and that interval includes no change. Cognition's 8.32-point gap starts at a floor of 0.75, and 55.6 percent of it is the infusion falling, cerebrospinal-fluid GL-1 fell 64.3 percent, splenomegaly was 3 of 21 against 0 of 22, and no pill price is public, while a 60 kilogram adult at the top of the Cerezyme label illustrates at $409,843.98 a year.