PD-L1 assay comparability study published

Staburo supported the evaluation of the first multicentric analytical comparability study of PD-L1 expression in urothelial bladder cancer using four clinically developed immunohistochemistry assays. Statistical analyses included the investigation of inter-assay differences using ANOVA and intra-class correlations and the concordance for clinically relevant cut-offs based on Fleiss‘ kappa.

The results suggest that all four currently clinically relevant assays are analytically similar for the evaluation of PD-L1-stained immune cells and three (SP263, 22C3, and 28-8) for PD-L1-stained tumor cells.

The paper, which was recently published in Virchows Archiv, can be downloaded here (Staburo’s Statistics contribution is mentioned in the Acknowledgments): https://rd.springer.com/article/10.1007%2Fs00428-019-02610-z

R Shiny App for Standardized Detection of Outliers in Bioassay Development

The availability of reliable and sensitive assays is an important building block in the production and quality control of biological products. In the development of such assays, unusual or erroneous measurement results need to be detected, but adjudication by human operators is often inconsistent or prone to bias.

To help overcome this challenge, a Shiny app for the statistical detection of such outliers is being developed at Staburo: the Optimal Tool for Tracking of Outliers (OTTO). Close collaboration with the end users throughout the development process ensures that the statistical outlier detection results are in good agreement with subject matter expert evaluations. In particular, a mixed method approach was chosen which significantly reduced false outlier detection rates, compared to conventional statistical methods.

Thanks to the capabilities of the R Shiny technology, the user interface is intuitive and well suited for everyday use in the lab. Features of the app include application of outlier tests, automatic curve refitting after removal of outliers, and report generation for documentation purposes.

New employee @ Staburo

We are very happy to welcome Dr. Yesilda Balavarca in our team. Yesilda joins Staburo with her work experience from Deutsches Krebsforschungszentrum (DKFZ) Heidelberg and her excellent education background from University of Hasselt and University of Göttingen. She supports our clients in the areas of biostatistics and statistical programming as a Senior Biostatistician. We are looking forward to a great collaboration!

Training @ Staburo: Response Evaluation Criteria in Solid Tumors (RECIST)

At Staburo, we are working on a lot of oncology projects. Dealing with data from this particular therapeutic area does not only require good statistical knowledge, but also to understand how the status of the patients themselves is evaluated and which are the medical objectives behind that.

More precisely, this training intended to describe how the patients with solid tumors are evaluated in oncology trials. This occurs with the commonly used RECIST criteria, from which the latest version (1.1) was presented. First, CT or MRI scans are performed for each patient at baseline to describe the existing lesions. At several timepoints of the trial, images are performed again to evaluate these already detected tumor lesions and to detect potentially new lesions. Measurements of lesion sizes are performed on these images, which then lead to a response evaluation: “Complete response”, “partial response”, “stable disease”, “progressive disease” (PD), or “not evaluable”. This response can trigger the (dis-)continuation of patients in the trial (mostly in case of PD) but is also useful to evaluate efficacy of the treatments tested.

Details of the RECIST criteria were presented such as definitions of measurable and non-measurable lesions, target and non-target lesions, measurement rules (including Sum of Longest Diameters), as well as rules for assessing a response to a patient.

The next step of the analysis is the derivation of endpoints with the help of the response assessment. These endpoints were described in the presentation: they are mostly binary endpoints such as Objective Response, and time-to-event endpoints such as Progression-Free Survival. The usual methods of analysis were as well part of the presentation.

Finally, other types of criteria in oncology were shortly introduced: RECIST applies to solid tumors, but not to blood cancers for example where specific criteria are necessary.