Clinical Development optimization

Modeling Huntington Disease progression to de-risk Phase III trials

Modeling Huntington’s disease progression heterogeneity allows to de-risk Phase III trials while optimizing their probability of success by targeting rapidly progressing patients.
Huntington disease is the the most common single-gene neurological disorder to an inexorable decline in motor, cognitive & behavioral functions due to unstable expression of a trinucleotide repeat (CAG) in the huntingtin gene. Variability in nature and progression rate is known to be high. There is currently no cure for Huntington’s disease, yet a number of pharmaceutical companies are planning to launch concomitantly a drug over the next few years.


Characterize progression profiles and identify patients with higher benefit vs. Competitors (optimal positioning)

Substantiate the link between short-term efficacy and long-term benefits.

Predict the benefits of a new treatment in real life.

Methods & solution

Using a clustering algorithm and a random forest model to characterize patients endotypes of progressor profiles

A clustering algorithm was used on observational data to identify endotypes of disease progression.

In a second step, a random forest predictive model was applied to clinical and MRI data to identify prognostic factors for disease development and progression.

The identified prognostic factors were integrated into a Bayesian progression model (dynamic networks).


Modeling the causal and temporal relationship and characterizing different progression endotypes

Six different endotypes of progression speed/severity were identified and described, they were homogeneous in terms of progression speed and stage. The analysis confirmed and quantified the strong correlation established between age of onset of motor disturbance & repetition of the CAG sequence and highlighted the determining role of brain volume.

Half of the cohort showed a “classic” progression. Nevertheless, different severity profiles withing patients were also highlighted, underlining the existence of yet unidentified progression factors.

The causal and temporal relationship was modeled so to enable long term outcomes prediction given patients characteristics.


De-risked trial design and positioning

  • New drug can be positioned vs higher-benefit patients so to maximize value vs. Upcoming competitor.
  • The probability of success of upcoming phase IIb & III trials cto be significantly improved.
  • The developed model can also be used as a basis for synthetic control arms emulations.
  • The next step is to enrich the model with additional data, to increase the number of patients and include more variables in the model, in particular biomarkers (NfL, mHTT) as well as imaging data to refine the progression model.