Biometric Colloquium
S. Wied (1), D. Bodden (1,2), V. Ihl (1,2), S. Schoenen (1,2)
(1) Sigmund Freud Private University, Vienna, Austria
(2) RWTH Aachen University, Aachen, Germany
Challenges in rare disease clinical trials: Randomization, bias and statistical inference
April 23rd, 2026 at 09:00 am
Thursday 23rd of April 2026 at 9:00am
Jugendstilhörsaal der Medizinischen Universität Wien
Rektoratsgebäude, Ebene 02+03
Medical University of Vienna, 1090 Wien
Hosts: Nico Bruder
Abstract:
Clinical trials in rare diseases face distinct methodological challenges, primarily due to small sample sizes, heterogeneous patient populations, and limited understanding of the natural history course of the disease. As a result, existing evidence is sparse, and approximately 95% of patients affected by one of the nearly 7,000 identified rare diseases suffer from a lack of effective treatment options. Improving the design and analysis of clinical trials in this setting is therefore essential for enabling robust evaluation of new therapies.
Randomization is widely regarded as the gold standard for protecting against bias and ensuring the validity of statistical inference. However, randomization becomes more challenging to implement in complex trial settings, particularly in adaptive designs, as well as in studies with multiple endpoints. In
rare disease trials, where blinding is often li mited, allocation bias represents a central concern.
Allocation bias arises when future treatment assignments can be predicted based on prior allocations, potentially leading to the preferential assignment of patients with specific characteristics to treatment arms. In addition, other sources of bias, e.g. chronological bias, may emerge, particularly in complex
designs such as platform trials. We examine the impact of these biases on statistical inference across various clinical trial designs, including platform trials and multiple-endpoint settings. By employing direct measures such as the mean actual Type I error probability, mean actual power and the go-no- go criterion, we quantify these effects and provide guidance on selecting randomization procedures that minimize the trial’s susceptibility to bias.
Statistical inference in rare disease trials is further complicated by small sample sizes and patient heterogeneity. In particular, for binary outcomes, standard large-sample methods may perform poorly due to low power and unreliable asymptotic approximations. In this context, a novel exact stratified
test based on Boschloo’s test is introduced and compared with existing approaches, including the (stratified) Fisher’s exact test, Boschloo’s unconditional exact test, and the Cochran–Mantel–Haenszel test. The results demonstrate that tailored testing strategies can improve power while maintaining control of the type I error rate.
Overall, these findings highlight the need for carefully chosen randomization strategies to reduce susceptibility to bias, alongside adapted statistical methodologies to ensure valid inference in rare disease clinical trials. Such approaches are critical for generating reliable evidence and ultimately
improving treatment options for patients with rare diseases.
Funding
SW, DB, VI and SS are members of RealiseD supported by the Innovative Health Initiative Joint Undertaking (IHI JU) under grant agreement No 101165912. The JU receives support from the European Union’s Horizon Europe research and innovation programme and COCIR, EFPIA, Europa Bío, MedTech Europe, and Vaccines Europe. Views and opinions expressed are those of the author(s) only.
This publication reflects the author’s views. They do not necessarily reflect those of the Innovative Health Initiative Joint Undertaking and its members, who cannot be held responsible for them. For more information on RealiseD visit https://realised-ihi.eu/