Orphan Therapies: Making Best Use of Postmarket Data

Judith C. Maro, Jeffrey S. Brown, Gerald J. Dal Pan, Lingling LiJournal of General Internal Medicine 2014 · DOI

Introduction

Orphan drugs—therapeutics designed for rare diseases—face a fundamental "small numbers" problem. Due to tiny patient populations, pre-licensure clinical trials are frequently insufficient to detect rare but serious adverse events. Consequently, the FDA often mandates postmarket registries. However, these registries typically operate as information silos; data from competing manufacturers is rarely compared, and traditional non-sequential studies wait for a fixed enrollment period before analysis, potentially delaying critical safety warnings by years.

This paper proposes a solution: linking product-specific registries through distributed electronic health data networks to enable "live" sequential statistical monitoring. By sending analysis code to the data rather than pooling patient-level records, manufacturers can collaborate on safety surveillance without compromising proprietary information.

Methodology

The researchers employed a two-pronged simulation to evaluate the efficiency of this framework:

  1. Conditional Sequential Sampling Procedure (CSSP): This allows for repeated interim looks at data as it accumulates, calculating the "sample size savings ratio" compared to traditional fixed-sample methods.
  2. Agent-Based Modeling (ABM): To translate mathematical savings into real-world time, the authors simulated a population of 3,000 patients (modeled after Homozygous Familial Hypercholesterolemia) and their adoption of new therapies over time.

These methods allowed for the calculation of an Analytic Calendar Time Savings Ratio, measuring the actual reduction in surveillance duration.

Efficiency Gains and Market Dynamics

The study found that sequential analysis consistently detects safety signals earlier than traditional methods. Interestingly, the most substantial savings in sample size occur when market share is highly imbalanced (e.g., one drug holds 90% of the market). In these scenarios, the sequential approach can reduce the required sample size by up to 30-40%.

Savings ratios for sample size and calendar time across various simulation parameters.

For "common" outcomes (defined as 1 event per 100 person-years), the analytic calendar time savings ratio ranged from 0.05 to 0.31. This suggests that researchers could reach definitive safety conclusions 5% to 31% sooner than with non-sequential methods.

The "30-Year Wall"

Despite the statistical advantages of sequential monitoring, the study highlights a sobering reality regarding extreme rarity. For "infrequent" outcomes (1 event per 1,000 person-years), many simulation scenarios failed to reach statistical significance within a practical 30-year window.

Detailed simulation metrics highlighting the '30-year wall' for rare event detection.

As shown in the table above, many results for infrequent events are labeled "Not Determined" (N/D) because the median length of surveillance exceeded 30 years. This suggests that for certain orphan drugs, traditional comparative postmarket surveillance may be functionally impossible, regardless of the statistical method used.

Conclusion and Future Directions

While a distributed, sequential network offers clear efficiency gains, the "rarity hurdle" remains significant. The authors suggest that future work should explore incorporating "prior information"—such as data from animal studies or early-phase trials—into these sequential models to further compress the time needed for a definitive signal. While the front-end costs of building such a network are high, the potential for earlier detection of life-threatening side effects provides a compelling public health argument for breaking down registry silos.

Summary updated

BibTeX
@article{maro2014orphan,
  author = {Judith C. Maro and Jeffrey S. Brown and Gerald J. Dal Pan and Lingling Li},
  title = {{Orphan Therapies: Making Best Use of Postmarket Data}},
  journal = {Journal of General Internal Medicine},
  year = {2014},
  doi = {10.1007/s11606-014-2882-1},
}
Figure