The problem with LOA

Likelihood of Approval (LOA) is a widely referenced metric in drug development, particularly in the later stages of the pipeline. It measures the probability that a drug candidate will successfully navigate the final stages of clinical trials—most notably Phase III—and achieve regulatory approval. While related to Probability of Success (POS), LOA focuses specifically on these advanced stages and the regulatory hurdles required for market authorization. Its apparent simplicity and specificity make it an attractive metric for decision-making, but LOA suffers from critical shortcomings that limit its utility, particularly in evaluating innovative or high-risk drug candidates.

Defining LOA

LOA is commonly calculated as the product of probabilities for the later stages of drug development:

$$ LOA = P_{phase\_III} \cdot P_{approval} $$

Where:

• Pphase III: Probability of successfully completing Phase III clinical trials.
• Papproval​: Probability of regulatory approval after submission of a New Drug Application (NDA) or Marketing Authorization Application (MAA).

In some contexts, LOA is expanded to include earlier stages, effectively mirroring the concept of POS. However, its primary application remains the final hurdles of clinical development and regulatory review.

A Comparative Example: Antibody-Drug Conjugates (ADCs) vs. RNA-Based Therapies

Case 1: ADC Targeting HER2

• Context: Antibody-drug conjugates (ADCs) targeting HER2 are a well-established modality, with multiple approvals for breast and gastric cancers. Phase III trials focus on confirming efficacy and safety in specific patient populations.
• Probabilities:
  • Pphase III=0.75 (extensive precedent for efficacy and safety).
  • Papproval=0.90 (regulatory familiarity with ADCs and HER2 targets).

Calculation:

$$
LOA_{ADC} = 0.75 \cdot 0.90 = 0.675 \, (67.5\%)
$$

Case 2: Novel RNA-Based Therapy

• Context: A first-in-class RNA-based therapy aims to silence a newly discovered oncogene implicated in pancreatic cancer. While highly innovative, it faces significant technical and regulatory uncertainties.
• Probabilities:
  • Pphase III=0.40(uncertain efficacy in a challenging cancer type).
  • Papproval=0.60 (limited regulatory experience with this modality and target).

Calculation:

$$
LOA_{RNA} = 0.40 \cdot 0.60 = 0.24 \, (24\%)
$$

Interpreting the Results

1. The ADC:
   • The 67.5% LOA reflects the established nature of ADCs and the robust precedent for HER2-targeted therapies.
   • Regulators and clinicians are familiar with this pathway and modality, reducing uncertainty in both trials and approvals.
2. The RNA-Based Therapy:
   • The 24% LOA underscores the high risks associated with novel modalities and targets, particularly in a challenging indication like pancreatic cancer.
   • Despite its lower LOA, the therapy could revolutionize treatment if successful, offering high potential rewards relative to the risk.

This example highlights LOA’s focus on late-stage probabilities and its inherent bias toward established therapies.

Critique of LOA

1. Oversimplification of Late-Stage Risks

LOA reduces the complexity of late-stage drug development to two probabilities, ignoring critical interdependencies and external influences:

• Interdependencies: Phase III outcomes often inform regulatory review, but LOA assumes these probabilities are independent. For example, marginal efficacy in Phase III may lead to additional regulatory scrutiny, reducing Papproval​.
• Dynamic Probabilities: Probabilities evolve during Phase III trials as new data emerge. LOA often fails to incorporate interim analyses or real-time changes in the regulatory landscape.

2. Bias Toward Established Modalities

Similar to POS, LOA relies heavily on historical averages, favoring well-understood modalities and penalizing innovation:

• Established Modalities: ADCs, which have numerous precedents, benefit from high Pphase III ​ and Papproval values due to regulatory familiarity.
• Novel Modalities: RNA-based therapies, with limited precedent, face lower LOA values, despite their potential to redefine treatment paradigms.

3. Neglect of External Factors

LOA focuses narrowly on clinical and regulatory probabilities, often excluding broader factors that influence final approval:

• Payer Considerations: Pricing and reimbursement decisions increasingly affect regulatory pathways. Therapies with high costs may face additional scrutiny, reducing Papproval.
• Global Disparities: Regulatory processes vary significantly by region, and LOA often assumes uniform probabilities across geographies.

4. Lack of Standardization in Definition and Application

The interchangeable use of LOA and POS creates confusion:

• Overlapping Definitions: Some researchers define LOA as the likelihood of progressing from Phase I to approval, conflating it with POS. Others restrict LOA to Phase III and regulatory review, leading to inconsistent interpretations.
• Diverse Methodologies: Variations in LOA calculation methodologies limit cross-study comparability and create challenges in benchmarking.

Illustrative Shortcomings of LOA

1. Penalizing Innovation

LOA disproportionately penalizes innovative therapies with low precedents, such as first-in-class RNA therapies or gene-editing treatments. While these therapies may have transformative potential, their LOA is artificially suppressed by limited historical data.

2. Favoring Incremental Improvements

Therapies that represent minor advancements in established modalities, such as ADCs targeting new patient subgroups, are favored due to higher Pphase III and Papproval. This bias encourages risk aversion and underinvestment in breakthrough science.

3. Overlooking Conditional Approvals

Regulatory agencies increasingly grant conditional or accelerated approvals based on surrogate endpoints, particularly for rare diseases or high-unmet-need indications. LOA models often fail to account for these pathways, underestimating probabilities for therapies targeting orphan diseases or rare cancers.

Conclusion: The Limitations of LOA

Likelihood of Approval (LOA) provides a focused lens on the late stages of drug development, but its oversimplification of risks, bias toward established modalities, and neglect of external factors limit its effectiveness. While LOA can offer useful insights for comparing late-stage candidates, decision-makers should complement it with more nuanced analyses that account for dynamic risks, regulatory trends, and long-term value potential. Only by addressing these limitations can LOA truly serve as a robust tool for guiding investment and strategic decisions in biotech and pharma.