Bcal Diagnostics Limited provided additional information to supplement the announcement dated 7 February 2022, following analysis of its scientific data carried out independently by its Australian team and by a highly regarded team of international diagnostic experts. BCAL Diagnostics provided scientific data on Cohorts 1, 2, 3, and 4, to be independently reviewed and retrospectively analysed in a strategic collaboration with BSC-Medical from the United States, involving Dr Szilard Voros, and Dr Aruna Bansal. Both experts have a proven track record in developing IVDs.

BSC analysed the data in a blinded fashion, without any access to prior in-house results. In this formal analysis conducted pursuant to a pre-specified Statistical Analysis Plan, candidate signatures were developed for five discovery sets. Overall, twelve lipids selected, with two lipids selected multiple times: lipid 1 and lipid 2. Each candidate signature contained one or both of these lipids, and each candidate signature was supported by Area Under the Curve (AUC) > 70% in at least one other dataset.

Two specific signatures developed for cohort 4 and combined cohorts 2+3+4 showed strong performance in the validations sets. To clarify, the cohort 4 signature was well supported by cohorts 2 and 3 with impressive AUC of 78% and 85% respectively. The cohort 2+3+4 signature was well supported by the 2+3+4 validation set with AUC of 83%.

Acknowledging that the patients overlapped, these two signatures were very similar in composition, had only three lipids each, and two of the lipids were consistent, with a third lipid for each signature from one class of lipids. The supportive lipids were highly correlated. The findings demonstrated strikingly internally consistent results across all datasets.

An overall accuracy of 77% was achieved in independent validation. The results are especially exciting because when testing a signature on a never seen before data set, a considerable drop in performance (15-20% based on empirical testing) from training to testing sets is frequently observed while the algorithm is being developed and fine-tuned, which was not the case here for cohort 4 and the combined cohort 2+3+4 results. This lends itself to the strength of the identified signatures.

An in-house comparison of the 12 lipids identified by Dr Bansal to BCAL's results showed a 50% overlap despite the use of very different approaches. Furthermore, the 2 lipids in the most promising signatures identified by Bansal are part of BCAL's 18 lipid panel and 2 of the 12 are part of BCAL's locally optimised 6- lipid signature. This is a significant outcome given that the 12 lipids were identified out of 400+ candidates and half of them were mutually identified.

In summary, the outcome of the analysis by the two teams is that a considerably reduced number of markers, compared with previous information, can be used to distinguish between blood samples of breast cancer patients and normal control samples. Such a reduction in the number of markers to be examined for each test considerably improves its commercial feasibility and attractiveness, reducing the time, cost and difficulty of analysis of each sample.