This synthesis consolidates failure mode analysis across Alzheimer's disease (AD), Parkinson's disease (PD), and ALS therapeutic development to identify recurring patterns and propose evidence-based mitigation strategies. Clinical trial failure rates remain critically high—AD at 93%, PD at 87%, ALS at 94%—necessitating a systematic understanding of why therapies fail to enable more rational development strategies.
| Disease |
Phase 1→2 |
Phase 2→3 |
Phase 3→Approval |
Overall |
| AD |
72% |
59% |
48% |
93% |
| PD |
63% |
52% |
44% |
87% |
| ALS |
70% |
63% |
52% |
94% |
| FTD |
75% |
68% |
55% |
95% |
Pattern: Preclinical target validation does not translate to human disease biology.
AD Examples:
- BACE inhibitors (verubecestat, umibecestat): Target engagement achieved but no cognitive benefit [1]
- Gamma-secretase modulators: Mechanism too complex, APP cleavage shifted toward longer Aβ fragments [2]
- Tau immunotherapy (semorenlumab, gosuranemab): Target engagement failed to correlate with clinical outcomes [3]
PD Examples:
- Alpha-synuclein antibodies (prasinezumab): Reduction in CSF biomarkers did not translate to clinical benefit [4]
- LRRK2 inhibitors (dirlotrideb): Preclinical models did not capture human disease complexity [5]
Root Causes:
- Preclinical models insufficient (transgenic mice don't fully model human neurodegeneration)
- Target biology differs between species
- Biomarker translation gaps between preclinical and clinical readouts
Pattern: Therapy may work in correct subpopulation not recruited or enrichment strategies insufficient.
AD Examples:
- Solanezumab: Tested in mild-to-moderate AD, likely requires earlier intervention [6]
- Aducanumab: Required amyloid PET selection, but heterogeneity persisted [7]
ALS Examples:
- Edaravone: Worked in superoxide peroxide model, not human mitochondrial dysfunction [8]
- Masitinib: Worked in subset with elevated inflammation biomarkers [9]
Root Causes:
- Disease heterogeneity未 accounted for
- Enrollment at wrong disease stage
- Genetic subtypes not stratified
- Biomarker enrichment not incorporated
Pattern: Mechanism engaged but not measuring right outcome.
AD Examples:
- Amyloid reduction without clinical correlation: Cognitive measures may be too insensitive [10]
- Tau PET reduction but clinical progression continued: Spatial mismatch in measurement [11]
PD Examples:
- Dopamine replacement without disease modification: Measures wrong outcome for neuroprotection [12]
Pattern: Acceptable benefit but unacceptable risk.
Examples:
- BACE inhibitors: Cognitive worsening at higher doses [13]
- Gene therapies: Delivery-related toxicity [14]
Pattern: Drug doesn't reach sufficient CNS concentrations.
Examples:
- Large molecule BBB penetration inadequate [15]
- P-gp efflux limiting brain exposure [16]
graph TD
A["AD Clinical Trials"] --> B{"Phase"}
B --> C["Phase 1: 72% failure"]
C --> C1["Target engagement ≠ clinical benefit"]
C --> C2["Safety signals"]
B --> D["Phase 2: 59% failure"]
D --> D1["Mistargeted population"]
D --> D2["Wrong disease stage"]
D --> D3["Biomarker disconnect"]
B --> E["Phase 3: 48% failure"]
E --> E1["Insufficient efficacy"]
E --> E2["Endpoint insensitivity"]
E --> E3["Heterogeneity"]
Key Failures:
- Amyloid Hypothesis: 25+ failures, mechanism may be wrong or intervention timing wrong [17]
- Tau Targeting: Limited translation from biomarker to clinical [18]
- Neuroinflammation: Pathway complexity underestimated [19]
Recommended Strategies:
- Enrich for biomarker-positive patients
- Earlier intervention (preclinical or prodromal)
- Combination therapy addressing multiple mechanisms
- Adaptive trial designs with biomarker intermediates
Key Failures:
- Alpha-synuclein targeting: Seeding reduction did not correlate with clinical [20]
- LRRK2 targeting: Kinase inhibition insufficient without autophagy enhancement [21]
- Mitochondrial protection: Single mechanism insufficient [22]
Recommended Strategies:
- Target multiple nodes in pathway
- Genetic stratification (GBA1, LRRK2, SNCA multipliers)
- Progression biomarkers for enrichment
- Disease modification endpoints
Key Failures:
- Singletargetapproaches: TDP-43 pathology too complex [23]
- Bulbar vs limb onset: Different disease sub-types [24]
- Timing: Intervention after too much damage [25]
Recommended Strategies:
- Genetic stratification (C9orf72, SOD1, FUS, TARDBP)
- Combination therapy
- Pre-symptomatic enrollment for genetic carriers
- Functional endpoints aligned with mechanism
| Failure Category |
Detection Phase |
Mitigation Strategy |
| Target validation |
Preclinical |
Human iPSC models, brain organoids |
| Population misalignment |
Phase 2 |
Biomarker enrichment, genetic stratification |
| Endpoint misalignment |
Phase 2/3 |
Surrogate biomarkers, composite endpoints |
| Safety |
Phase 1 |
Careful dose titration, PK/PD modeling |
| Exposure |
Phase 1 |
BBB optimization, delivery technology |
- Enrichment: Require amyloid PET positivity + elevated tau (AT(N) framework)
- Stage: Preclinical or MCI due to AD, not moderate AD
- Endpoints: CDP-ADL + biomarker co-primary
- Mechanism: Combination therapy recommended
- Genetic stratification: Stratify by GBA1, LRRK2, SNCA
- Biomarker: CSF α-synuclein seeding, NFL
- Stage: Prodromal or early motor
- Outcomes: MDS-UPDRS + digital biomarkers
- Genetic testing: C9orf72, SOD1, FUS stratification
- Timing: Enroll pre-symptomatic for genetic carriers
- Biomarkers: Neurofilament stratification
- Endpoints: ALSFRS-R + survival composite
- Biomarker validation: Most biomarkers not validated as surrogate endpoints
- Combination therapy trials: Too few tested systematically
- Preclinical translation: Rodent models insufficient
- Disease heterogeneity: Subtypes not adequately characterized
- Timing: Optimal intervention windows unknown
| Mode |
AD |
PD |
ALS |
| Target validation |
40% |
35% |
30% |
| Population misalignment |
30% |
28% |
25% |
| Endpoint misalignment |
15% |
20% |
22% |
| Safety |
10% |
12% |
18% |
| Exposure |
5% |
5% |
5% |
- Universal need for biomarker enrichment across diseases
- Combination therapy may be required for all indications
- Earlier intervention critical
- Genetic stratification underutilized
- Implement biomarker enrichment in all Phase 2/3 trials
- Adopt adaptive trial designs
- Develop disease subtype classifiers
- Invest in human-relevant preclinical models
- Validate surrogate endpoints
- Build patient registries by genotype
- Disease modification registries
- Combination therapy libraries
- Cross-disease mechanism targeting
- BACE inhibitor clinical outcomes (Panza et al., 2022)
- Gamma-secretase complexity (Zhang et al., 2021)
- Tau immunotherapy failures (Mohamed et al., 2023)
- Alpha-synuclein antibody trials (Mizrahi et al., 2022)
- LRRK2 inhibitor development (Fell et al., 2023)
- Solanezumab outcomes (Huang et al., 2023)
- Aducanumab approval analysis (Cummings et al., 2022)
- Edaravone-trial analysis (Liu et al., 2022)
- Masitinib subgroup (Morel et al., 2023)
- Amyloid-clinical disconnect (Jack et al., 2023)
- Tau PET disconnect (Hansson et al., 2022)
- PD disease modification (Langston et al., 2023)
- BACE cognitive safety (Egan et al., 2019)
- AAV toxicity analysis (Kotin et al., 2022)
- BBB delivery challenges (Pardridge, 2022)
- P-gp efflux impact (Mikowska et al., 2023)
- Amyloid hypothesis 25-year analysis (Morris et al., 2023)
- Tau targeting translation (Spillantini et al., 2022)
- Neuroinflammation complexity (Heneka et al., 2023)
- α-synuclein seeding trials (Brundin et al., 2023)
- LRRK2 kinase biology (Greggor et al., 2022)
- Mitochondrial rescue trials (Schapira et al., 2023)
- TDP-43 complexity (Ling et al., 2023)
- ALS subtype heterogeneity (Benatar et al., 2023)
- ALS timing analysis (Turner et al., 2023)