Osteoarthritis Prognosis

Context

Patients with osteoarthritis (OA) can be roughly divided in two categories. A majority of them (70-90%) will have a mild/moderate progression of their disease, spreading over more than 10 years (this is good news for them). On the other hand, the rest of the OA patients (10-30%) will develop an aggressive disease, damaging their articulations in only a few years. Although there is no effective drug against the progression of OA yet, many pharmaceutical companies are working to develop new components in that field. Their problem is: “how can we prove the efficacy of a treatment over the duration of a clinical trial when in any case up to 90% of the patients we recruit will not evolve even without treatment?” Our client in this project is a CRO specialized in OA which has also developed kits allowing measurements of several markers monitoring the cartilage degradation.

Objective

The objective of our customer was to develop a solution based on its cartilage degradation biomarkers that could allow predicting the OA prognosis of any single patient: aggressive progression or not? The aim is to provide this tool to pharmaceutical companies wanting to prove the efficacy of an OA treatment on a limited set of patients prognosed with aggressive OA progression. On those patients, the chance of proving a significant difference between a test and a control arm would be much higher. Data have been collected by our client from more than 200 patients. Biomarkers have been measured at baseline and several later time points. Clinical information consisting in both traditional patient information (gender, age, BMI, …) and specific indexes (WOMAC) as well as Imaging information (Joint Space Width) have also been collected at several time points.

What we did

Based on the data made available to us by our customer, we:

  1. assessed the discrimination power of each candidate marker from our client in terms of OA aggressiveness / progression
  2. assessed the discrimination power of several combinations of markers (these combination are deduced by our algorithms)
  3. suggested to combine clinical information and indexes with the markers for an increased predictive power
  4. performed a pharmaco-economic study taking into account all costs of a traditional clinical trial in this field, including the costs of patients screening and non-enrollment (if predicted as non-progressor) to identify the most cost-effective trade-off in terms of sensitivity and specificity for the model
  5. designed and developed a web application making these results usable in practice

At the end of the project, although still an prototype, our solution was an OA progression system able to reduce the cost of a clinical trial in OA by 25%, all costs included, and operational in practice via a convenient web interface.

Customer’s feedback

“DNAlytics carried out this project swiftly and with tangible deliverables. We can now make our future clients more aware of how our markers can in practice have an impact on their costs and the design of their projects. A good point was to suggest the incorporation of clinical information in the models along with our own markers’ data. By making the data analysis objectives clearer, DNAlytics brought interesting inputs feeding our reflexion on our service offer and positioning.”

What if it had to be re-done?

We really care about the fact that the models we design are fit as much as possible for the real-life context in which they would be used. In this (preliminary) project, the available cohorts presented some mismatch for some factors (namely in terms of gender and BMI) with respect to a real life setting. Typically, this is also the kind of input we are able to provide when we are associated to the design of a clinical protocol.

Do you have to deal with a similar project?

Contact us!