Randomized and non-randomized designs for causal inference with longitudinal data in rare disorders.

Area of Science:

• Clinical Trials
• Epidemiology
• Biostatistics

Background:

• Rare diseases affect 30 million in the US, with only 10% having treatments.
• Small populations, diverse genetics, and complex progression challenge traditional trial designs for rare diseases.

Purpose of the Study:

• To review longitudinal study designs for rare disease research.
• To draw parallels between randomized and observational controlled study designs.
• To discuss causal inference methods for rare disease therapy evaluation.

Main Methods:

• Review of longitudinal designs (crossover, N-of-1, sequential).
• Comparison with observational designs (case series, case-crossover, cohort studies).
• Discussion of causal inference techniques (matching, weighting).

Main Results:

• Self-controlled randomized crossover and N-of-1 designs align with observational case series/crossover.
• Randomized sequential designs are comparable to longitudinal cohort studies with sequential matching/weighting.
• Longitudinal designs enhance study power and reduce variability in treatment effect estimation.

Conclusions:

• Longitudinal and carefully designed observational studies are crucial for rare disease research.
• Valid causal inference can be achieved using these advanced designs and analysis methods.
• Examples from urea cycle disorder and cystic fibrosis illustrate the application of these methods.