© J Renwick Beattie. Data Scientist. Privacy Policy | Terms and Conditions | External Resource Links | Contact Form

Services Home Osentia Signal Processing Kerus Protein Ageing

Case Note Sections:


The Osentia Test - a screening tool for fracture risk

The background:

When Dr Beattie commenced working on the Osentia project it was backed by numerous small scale preliminary academic studies that demonstrated a strong connection between the proteins in human nails and their bone health. A spin out company was created to develop and commercialise the promising work, which promptly set about designing and undertaking a number of follow on clinical studies in association with a range of 3rd party collaborators. The studies showed promising trends of scientific interest, where neither proving commercially nor clinically useful.

The first task was to review the reports and data acquired from the clinical studies. These made it clear that something useful was happening in the cohort, but it was getting lost in the midst of noise.

The schematic to the left shows that the variability involved in repeating measurements exceeds the average difference between healthy and fracture. What does this mean?

It is common place for one measurement to predict healthy and a repeat measurement to predict high fracture risk

Schematic showing the variability in repeated measurement of a nail compared with the average difference between healthy and fracture. The inherited methodology had resampling error 1.2 x the size of the group separation.

Sampling variability and healthy-fracture contrast

Such a state of affairs is clearly unacceptable for a health screening tool. A strategic plan was drawn up to:

  1. Identify the sources of error within the methodology
  2. Assess what measures were required to control variability
  3. Quantify the tolerance limits of the methodology

The strategy used the results of historical data analysis combined with existing documentation to identify 4 key suspects within 1 month. For each of these error bottlenecks a protocol was drawn up detailing what experiments needed to be carried out and how the different sources of variation could be isolated and controlled.

As part of the strategy the business case was reviewed to understand the required level of regulatory adherence, the demands of the end use case and the standards to which the test must ultimately perform. These then provided a benchmark to assess the outcome of the protocols against.