# Tofersen and the Surrogate-Endpoint Approach in ALS: Evidence Since Approval and Gaps for Generalization

**Date:** 2026-05-04

---

## Executive Summary

Tofersen (Qalsody) received FDA accelerated approval in April 2023 for SOD1-mutated amyotrophic lateral sclerosis (ALS) based primarily on robust reduction of plasma neurofilament light chain (NfL) and cerebrospinal fluid (CSF) SOD1 protein—surrogate endpoints deemed "reasonably likely to predict clinical benefit." Notably, the Phase 3 VALOR trial did **not** meet its primary functional endpoint (ALSFRS-R at 28 weeks; p=0.097), making the approval a landmark regulatory decision with significant implications for broader ALS drug development. This briefing reviews evidence accumulated since approval and identifies critical gaps that must close before a biomarker-driven, accelerated-approval framework can be generalized beyond SOD1-ALS.

---

## (a) Functional Clinical Outcomes Since Approval

### Phase 3 VALOR Trial — Primary and Surrogate Results

The VALOR trial randomized 108 adults with confirmed SOD1-ALS to intrathecal tofersen 100 mg or placebo. At the primary analysis (Week 28):

- **Primary endpoint (ALSFRS-R):** Not met. The adjusted mean difference favored tofersen but fell short of statistical significance (two-sided p=0.097).
- **Key surrogate endpoints:** Highly significant reductions in both plasma NfL (p<0.0001) and CSF SOD1 protein (p<0.0001), confirming central target engagement.

### Open-Label Extension (OLE) and Long-Term Analyses

Following completion of the placebo-controlled period, participants entered an open-label extension in which all received tofersen. Evidence accumulated since the initial approval includes:

1. **Delayed divergence of functional trajectories.** Integrated analyses comparing the early-treatment cohort with the delayed-start (placebo-to-tofersen) cohort showed progressive separation in ALSFRS-R slope over extended follow-up, with differences becoming more discernible beyond 52 weeks. Early-treated participants exhibited nominally slower rates of functional decline.

2. **Preservation of respiratory and muscle strength measures.** Secondary analyses suggested slower deterioration in slow vital capacity (SVC) and quantitative muscle strength (hand-held dynamometry) in the early-treatment group relative to delayed-start controls.

3. **Sustained biochemical response.** Plasma NfL and CSF SOD1 reductions have remained durable with continued administration, demonstrating that pharmacodynamic suppression of the toxic protein is maintained over time.

### Limitations and Interpretive Caveats

The post-hoc and open-label nature of the long-term functional analyses limits causal inference. Factors such as expectation bias, differential attrition, and regression to the mean in the delayed-start arm complicate interpretation. The FDA’s accelerated approval framework explicitly requires ongoing confirmation of clinical benefit; at present, the confirmatory dataset from the OLE and emerging real-world registries continues to mature, but definitive, prospective evidence of long-term functional preservation remains incomplete.

### Key Takeaway

The functional signal appears to emerge slowly in SOD1-ALS, likely reflecting the biology of the disease: reduced neurofilament release indicates diminished acute axonal injury, but stabilization of functional scores may require many months to become detectable above background variability. This latency has important design implications for future trials that rely on biomarker-based efficacy signals.

---

## (b) Translation of the Surrogate-Endpoint Approach to Non-SOD1 ALS Subtypes

### NfL as a General ALS Biomarker

Plasma NfL is broadly elevated across ALS subtypes, including sporadic ALS and genetic forms linked to *C9orf72*, *TARDBP*, and *FUS* mutations. Natural-history cohorts have robustly established that baseline NfL correlates with:

- Rate of ALSFRS-R decline,
- Overall survival,
- Upper motor neuron involvement, and
- Progression from symptom onset to diagnosis.

These properties have positioned NfL as a leading prognostic biomarker and a candidate treatment-response marker across ALS.

### Why Extrapolation Is Not Straightforward

Despite its prognostic value, translating the tofersen/NfL surrogate model to non-SOD1 ALS faces substantial conceptual and empirical hurdles:

**1. Etiologic specificity versus a final common pathway.**
In SOD1-ALS, tofersen directly suppresses a defined toxic protein encoded by a single pathogenic variant. The causal chain—mutant SOD1 expression → protein aggregation and toxicity → motor neuron injury → NfL release → clinical impairment—is relatively linear and genetically homogeneous. In sporadic or other genetic ALS, NfL reflects downstream neuroaxonal damage arising from heterogeneous upstream causes. A compound that lowers NfL through anti-inflammatory, anti-apoptotic, or other generic mechanisms may not interrupt the primary driver of neurodegeneration. Consequently, NfL reduction in these contexts may not be "reasonably likely to predict clinical benefit" in the same way.

**2. Differential NfL dynamics across subtypes.**
Genetic subtypes differ in baseline NfL and trajectory. For example, *FUS*-ALS often presents with exceptionally high NfL and rapid progression in young patients; *C9orf72*-ALS tends to show higher baseline NfL and significant frontotemporal cognitive involvement compared to many SOD1 variants. The quantitative relationship between relative NfL reduction and functional preservation established in SOD1-ALS may not generalize one-to-one to these distinct biological contexts.

**3. Mixed pathology in sporadic disease.**
Many patients with clinically diagnosed sporadic ALS have unrecognized genetic variants, and the majority display TDP-43 proteinopathy of uncertain etiology. Without a single, druggable molecular target, demonstrating that a therapy’s effect on NfL maps reliably to patient-centered functional outcomes is substantially more complex.

**4. Inconsistent biomarker–clinical correlations in other trials.**
Therapeutic trials in non-SOD1 ALS (e.g., agents targeting oxidative stress, inflammation, or axonal transport) have yielded mixed or inconsistent relationships between NfL modulation and functional outcomes. These experiences underscore that NfL modulation alone is insufficient for regulatory confidence unless tightly coupled to a clear mechanistic rationale.

### Current Regulatory Posture

As of 2024–2025, no therapy for non-SOD1 ALS has received approval based predominantly on an NfL surrogate endpoint. Regulatory agencies (FDA, EMA) have indicated openness to incorporating NfL into benefit–risk assessments but have not accepted it as a standalone surrogate for accelerated approval in heterogeneous or sporadic ALS absent a compelling mechanistic link between target engagement and the biomarker response.

---

## (c) Emerging Targets in Antisense and Gene Therapy Development

Beyond SOD1, several antisense oligonucleotide (ASO) and gene-therapy strategies are in early-stage development for ALS.

### Active ASO Programs

| Target | Biological Rationale | Development Stage | Notes |
|--------|-------------------|---------------------|-------|
| **STMN2 (Stathmin-2)** | TDP-43 pathology causes a cryptic exon to be included in STMN2 pre-mRNA, leading to nonsense-mediated decay and loss of stathmin-2—a protein required for axonal regeneration and maintenance. An ASO blocking this cryptic splice event could restore stathmin-2 expression. | Phase 1/2 trials ongoing (Biogen/Ionis collaboration) | Highly promising because TDP-43 pathology is present in ~97% of ALS, including most sporadic cases; thus, applicability is potentially far broader than monogenic approaches |
| **C9orf72 (GGGGCC repeat)** | Hexanucleotide repeat expansion produces toxic RNA foci and dipeptide repeat (DPR) proteins that drive cellular toxicity. ASOs targeting the repeat-containing transcript aim to reduce production of these pathogenic species. | Phase 1 completed; development continues | Complex phenotype (ALS, FTD, or both); earlier programs encountered safety or CSF exposure challenges, but optimized chemistries remain under investigation |
| **FUS** | FUS mutations cause aggressive, often early-onset ALS. ASOs designed to lower mutant FUS transcript are under preclinical and early clinical exploration. | Preclinical/early Phase 1 | Very rare population; ultra-rapid progression complicates traditional trial designs and endpoint selection |
| **ATXN2** | Intermediate-length polyglutamine expansions in ATXN2 are a known modifier/risk factor for ALS. Reducing ATXN2 expression has shown preclinical promise in mitigating TDP-43 toxicity. | Preclinical/early clinical | Represents an indirect, upstream modulatory strategy rather than a direct monogenic driver approach |

### Gene Therapy (Non-ASO) Approaches

| Modality | Description | Stage |
|----------|-------------|-------|
| **AAV–microRNA SOD1** | Novartis (via Avexis acquisition) is exploring intrathecal delivery of adeno-associated virus vectors encoding artificial microRNAs to suppress SOD1 expression. Offers potential for one-time administration versus repeated intrathecal injections. | Preclinical/early clinical development |
| **Gene editing (CRISPR/CRISPRoff)** | Proof-of-concept studies in patient-derived iPSC models have demonstrated viability for both SOD1 and C9orf72 repeat expansion knockdown. Not yet in human ALS trials. | Discovery/preclinical |

### Special Challenge: TARDBP / TDP-43

TDP-43 is an essential nuclear RNA-binding protein involved in splicing, RNA stability, and transport. Complete elimination via ASO knockdown would likely be neurotoxic. Consequently, traditional ASO-mediated transcript reduction strategies for *TARDBP* face a narrow therapeutic index. Emerging strategies focus instead on:

- **Restoring nuclear localization** of TDP-43 and clearing cytoplasmic aggregates,
- **Targeting downstream defects** such as the STMN2 cryptic splicing event (see above), or
- **Protein-quality-control pathways** that enhance clearance of misfolded TDP-43 without reducing total protein.

This distinction is important: the STMN2 ASO program is functionally a TDP-43 pathway therapy without the risks of directly suppressing *TARDBP* expression.

---

## Gaps That Must Close Before Surrogate-Endpoint Generalization

The following gaps represent critical barriers to extending the tofersen/NfL accelerated-approval model to broader ALS populations. Closing them will require coordinated efforts across academia, industry, and regulatory science.

### Gap 1: Formal Surrogate Validation in Mechanistically Heterogeneous Contexts

Tofersen’s regulatory case rested on a **closed causal loop**: a defined toxic protein (SOD1) is lowered, neurofilament (a marker of axonal injury downstream of that toxicity) falls, and the biological plausibility of clinical benefit is strong because the intervention targets the root genetic cause. In non-SOD1 or sporadic ALS—where the upstream causal landscape is heterogeneous and often unknown—NfL reduction may reflect generic neuroprotection rather than interruption of disease-driving pathology. Before NfL can support accelerated approval for such therapies, large cohort studies and pooled trial databases must formally validate, within each mechanistic context, that therapy-induced NfL changes reliably predict functional or survival outcomes.

### Gap 2: Standardization of Assays and Clinically Meaningful Response Thresholds

Multiple platforms measure plasma NfL (Simoa, ECLIA, MSD, mass spectrometry) with differing reference ranges, sensitivities, and preanalytical requirements. For generalization, the field needs:

- **Assay harmonization** or traceability to a clinical reference standard,
- **Prospectively defined thresholds** for what constitutes a meaningful relative or absolute NfL reduction, and
- **Age-, sex-, and BMI-adjusted normative frameworks** to interpret individual patient responses.

Without these, comparing NfL data across trials—or approving drugs based on an NfL signal—is methodologically fraught.

### Gap 3: Confirmation That Surrogate-Based Approvals Translate to Patient-Centered Benefit

Under accelerated approval, confirmatory evidence of clinical benefit is required post-market. The tofersen experience remains the test case. If long-term follow-up (from the VALOR OLE and real-world registries) fails to demonstrate unequivocal functional or survival benefit with rigorous methodology, confidence in NfL surrogacy will erode—regardless of biochemical response. The ALS field’s recent experience with the withdrawal of AMX0035 (initially approved on a modest functional signal, subsequently refuted by a larger trial) illustrates how fragile regulatory and clinical trust is. A clear, prospective demonstration that an NfL-based approval translates into durable patient benefit is essential before generalizing this model.

### Gap 4: Disentangling Target Engagement from Generic Neuroprotection

A therapy in SOD1-ALS demonstrated both **target engagement** (CSF SOD1 lowering) **and** **surrogate improvement** (plasma NfL lowering). In non-SOD1 ALS, a drug might lower NfL without clear evidence that it engages the disease-causing mechanism. Regulatory science needs frameworks to stratify NfL responses into:

- **Mechanism-specific** (linked to engagement of an established pathogenic driver), and
- **Generic neuroprotective** (downstream injury reduction of unclear causal significance).

The former supports accelerated approval far more robustly than the latter. Distinguishing these categories prospectively—through multi-analyte biomarker panels, molecular target-engagement assays, or pathway-specific biomarkers—is critical.

### Gap 5: Management of Disease Heterogeneity and Subgroup-Specific Trajectories

Even within SOD1-ALS, different mutations (e.g., A4V vs. D90A) exhibit variable penetrance, age of onset, and progression rates; functional signals were detectable only with post-hoc integrated analyses across a relatively homogeneous, genetically defined population. Generalizing surrogate-endpoint reliance to sporadic ALS—where clinical heterogeneity (bulbar vs. limb onset, fast vs. slow progressors, cognitive involvement) is enormous—would require prospectively defined, biomarker-stratified trial designs. Natural-history biomarker studies must establish whether the NfL–clinical outcome relationship holds within clinically relevant subgroups (e.g., bulbar-onset, older patients, co-morbid frontotemporal dementia) before a broad surrogate-endpoint framework can ethically or scientifically be applied.

---

## Conclusions

Tofersen’s approval was a milestone that validated biomarker-driven accelerated approval in a monogenic neurodegenerative disease. However, evidence accumulated since 2023 reinforces that NfL surrogacy is **context-dependent**: it is most defensible when anchored to a known genetic etiology, measurable target engagement, and a biologically plausible causal pathway to clinical outcomes. Before this approach can be safely generalized to the much larger non-SOD1 and sporadic ALS populations, the field must close gaps in **formal surrogate validation**, **biomarker standardization**, **confirmatory clinical evidence**, **mechanistic specificity**, and **management of disease heterogeneity**. Progress in emerging ASO and gene-therapy pipelines—particularly the STMN2 program, which may apply broadly to TDP-43–positive ALS—will test these principles, but will not by themselves resolve the epistemic and regulatory challenges of surrogate-endpoint generalization.

---

## Sources and Analytical Context

This synthesis integrates publicly available knowledge from:

- The VALOR Phase 3 trial and open-label extension analyses,
- FDA and EMA review documents and advisory committee deliberations for tofersen,
- Published natural-history and biomarker studies on NfL in SOD1-, C9orf72-, TARDBP-, and FUS-related ALS,
- Pipeline disclosures from Biogen, Ionis, Novartis, and academic consortia (e.g., Target ALS, ENCALS), and
- Regulatory frameworks for surrogate endpoints in neurodegenerative diseases.

*Note: This document was prepared using established scientific literature and public regulatory records. Specific 2024–2025 trial-in-progress data not yet available in the public domain are summarized as pipeline stages based on verified public disclosures.*