Before the Elevidys deaths, "reduced liver uptake" was a line in a slide deck. After two kids died of acute liver failure from Sarepta's AAVrh74 gene therapy — boxed warning, non-ambulatory indication pulled, stock down 83% — it became the most important differentiator in DMD gene therapy. SLDB designed their capsid for it. RGNX didn't. And at $7 against a $16 analyst consensus, the market is pricing SLDB like it loses.

Three companies are racing through the door Elevidys opened: RGNX with RGX-202 (AAV8 capsid), SLDB with SGT-003 (modified AAV9, engineered for reduced liver tropism), and Genethon with GNT0004 (EU-focused). SLDB has the better science. RGNX has the better timeline. This memo is about whether the science matters if it arrives second.

The Data

SGT-003 Phase 1/2 (INSPIRE DUCHENNE, 41 patients dosed, Feb 23 2026 data cutoff):

Microdystrophin expression by western blot: 60% at Day 90 (n=20), 91% at Day 360 (n=3). Gene therapies usually decay. This one is building — vector copy numbers held steady (11 to 12 per nucleus), the transgene isn't being silenced. By mass spec: 52% at D90 rising to 86% at D360. Same direction, different assay, same conclusion.

The biomarker constellation is consistent: CK -37%, LDH -38%, eMHC -44% (disrupts the chronic degeneration-regeneration cycle), cardiac LVEF stabilized-to-improved. One SAE in 41 patients (Grade 3 myositis, resolved, DSMB recommended continue without interruption). Zero liver events.

The construct itself is unique. SGT-003 is the only microdystrophin — approved or investigational — with the R16/R17 binding domain that localizes nNOS to muscle. That's not branding. Nitric oxide signaling in dystrophic muscle is a real mechanism, and no one else has it.

The caveat that matters: D360 is n=3. Three biopsies. The D90 data (n=20, 60%) is statistically real but lower than Elevidys Week 12 in young ambulatory patients (93-99% adjusted). FDA explicitly says cross-assay comparisons are invalid — different sample processing, reference materials, quantitation methods. The market will compare anyway. If the expanded D360 cohort (expected mid-2026) regresses below 70%, the headline evaporates.

The Better Vehicle

Factor regression (250 daily observations) reveals something the price doesn't:

SLDB: 71% idio, α = +16.2%, XBI β = 1.96
RGNX: 57% idio, α = -42.0%, XBI β = 2.04

Both are 2x levered biotech bets. Nearly identical sector loading. But strip out XBI: SLDB generates +16% annualized residual alpha. RGNX destroys -42%. If you believe the DMD gene therapy replacement thesis, SLDB is the better vehicle by a wide margin.

One caveat the regression doesn't advertise: backward-looking α on a binary biotech after a positive data release (March 11, stock +29.5%) is partly tautological — the event was idiosyncratic and positive, so of course the residual is positive. The RGNX comparison is more defensible because both are measured over the same contaminated window. But treat the absolute +16.2% as directional, not precise.

The Problems

RGNX reports first. RGX-202 pivotal topline data arrives Q2 2026 — before SLDB's Phase 3 even generates meaningful data. RGNX's primary endpoint is a low bar (≥10% expression at Week 12). They'll probably hit it. If RGNX files a BLA mid-2026 while SLDB is still enrolling, the "Elevidys replacement" narrative lands in RGNX's lap. RGNX's data could be good news for SLDB (validates gene therapy post-Elevidys) or bad (captures the market position). We have no edge on this outcome.

14.5% short interest, 8.3 days to cover. Someone has a material bearish position on this name and we don't know why. Possible explanations: biotech sector hedging (clinical-stage shorts are common), a thesis on Phase 3 execution risk (SLDB has never completed a clinical trial), or a view that RGNX captures the market first. We haven't identified a public short report. The honest answer is this is an open gap — we can't dismiss it and we can't name it. It warrants investigation before sizing.

The Phase 3 is ex-US. IMPACT DUCHENNE sites are in Australia, Canada, EU, and UK. No US sites are activated — SLDB says they're "evaluating" US sites due to KOL demand. If FDA requires US enrollment data for accelerated approval, this is a hidden blocker. The FDA Type C meeting (Feb 2026) aligned on trial design but didn't resolve the US data question. This risk is absent from analyst coverage and could independently tank the AA pathway probability. We've opened this as a high-priority research gap.

3.5x dilution in 14 months. Shares went from 40.5M (Dec 2024) to ≈141M fully diluted (post-March 2026 placement). The $226M raised in March at $5.61/share keeps the lights on through H1 2028. But this isn't a one-time event — it's the mechanism through which the bear case plays out. If AA is denied and Phase 3 becomes a 3-year grind, they'll need another raise before data reads out. At that point, existing holders own progressively less of whatever value remains.

The Binary

The thesis runs through one gate: FDA accelerated approval pathway for SGT-003.

SLDB is seeking guidance in H1 2026 meetings. The precedent exists — Elevidys got AA in June 2023 on expression data alone. But the FDA that approved Elevidys is not the FDA that exists today. Post-deaths, post-boxed-warning, post-AAVrh74 Platform Tech revocation, the bar has moved. SLDB chose a functional primary endpoint (Time to Rise velocity at 18 months) rather than a biomarker surrogate — more conservative, arguably better science, but possibly incompatible with the accelerated approval framework which relies on surrogate endpoints "reasonably likely to predict clinical benefit."

If granted: timeline compresses to 12-18 months from Phase 3 data. Stock reprices toward $14-18.

If denied: 3+ year Phase 3, burn rate at $156M/year and accelerating, against $414M cash. That's roughly 2.7 years of runway. Phase 3 with an 18-month functional endpoint plus enrollment time could easily exceed that. FDA denial doesn't kill the science — it kills the math.

We put this at 50/50. Not hedging — genuinely uncertain. The pathway exists, but the regulatory posture has shifted and the ex-US enrollment question adds a variable the market isn't pricing.

The Market's Math vs Ours

The current price (≈$1B FD market cap) implies the market weights positive outcomes at roughly 30%. We weight them at 55%. Here's the decomposition (market probabilities reverse-engineered from current price using scenario targets):

Our EV: $9.40 (+32% from current). The gap comes from three things confirmed cross-ticker:

The Elevidys safety shift is structural. SRPT's 10-K confirms the full timeline — deaths March and June 2025, boxed warning November 2025, non-ambulatory removed. RGNX's CEO explicitly positions against it in Q4 earnings: "clamoring for something where not worry about post-treatment events." CAPR's CEO positions cell therapy as complementary, filling the cardiac gap. This isn't a news cycle — the competitive landscape permanently rearranged. SRPT -83% 1Y, CAPR +155% 1Y prices the shift at the extremes. SLDB at +48% 1Y with 128% upside to consensus sits in the middle, still catching up.

The expression trajectory is unusual. Gene therapies don't usually improve from D90 to D360. Vector copies held steady. If this holds at larger N, it's a biological advantage that cross-assay caveats can't dismiss.

POLARIS-101 liver safety is preclinical but directionally confirmed. 41 patients, zero liver events. Not proof. But a data point that weighs more today than it did before the Elevidys deaths.

What we don't have edge in: XBI sector direction (31% of variance, β=1.96), RGNX data outcome (Q2 2026, no visibility), and the FDA's actual decision on AA pathway.

The Setup

The stock ran 29.5% in 9 days on the March 11 data release. Institutional floor is $5.61 (March 9 placement, 43M shares). A pullback to the 50-day MA (≈$6.35) would deliver 48% upside to EV vs 32% at current — meaningfully better risk/reward for the same underlying thesis.

This is a binary-sizing situation. Position-level Sharpe at current price is 0.22 — not a core holding. The probability gap is real but embedded in 80% annualized idio vol.

XBI hedging matters structurally here. At 71% idio, the position sits below the 75% target. Neutralizing sector beta isolates the idio component — the factor where the informational advantage actually lives. Hedged alpha improves from 17.6% to 27%; hedged Sharpe from 0.22 to 0.34.

Three resolvers sharpen the picture: (a) pullback to $6-6.50, (b) RGNX pivotal data Q2 2026, (c) FDA AA pathway guidance mid-2026.

Scenarios

Bull (30%) → $16, 6-12 months. FDA grants AA pathway. Mid-2026 data confirms D360 expression >85% at expanded N. RGNX data comparable, not dominant. Market reprices to analyst consensus.

Base-bull (25%) → $10, 12 months. FDA positive but non-committal on AA. Data holds. Pipeline breadth (FA Phase 1b data H2 2026, CPVT first-in-class Phase 1b) provides optionality. Gradual appreciation.

Base-bear (25%) → $6, 3-6 months. FDA requires full Phase 3 completion. Expression regresses at larger N. RGNX captures narrative. Stock retraces to pre-data-release levels. Another dilutive raise becomes probable.

Bear (20%) → $3, 6-12 months. Liver safety event in SGT-003 — destroys the POLARIS-101 narrative and the entire competitive positioning thesis. Or D360 expression collapses below 50% at larger N. Forced raise at distressed levels. Accumulated deficit already at $958M.

The bear case isn't about bad science. It's about bad math: AA denial → 2.7-year runway → burns through cash before Phase 3 reads out → dilutive raise at the worst possible moment → equity value erodes to option on eventual data.

Evidence