Raising standards in the PDX field

Patient-derived cancer xenografts (PDXs) are costly and time-consuming to develop. Hence, data sharing and collaboration has been essential to the growth of this research field but is it still limited by a lack of central guidance and suitable infrastructures. The EurOPDX Research Infrastructure (EDIReX EU project, see Session 3 talk by Enzo Medico) will aim to improve preclinical and translational cancer research and promote innovation in oncology by integrating a European PDX repository and facilitating access to this essential resource for researchers from Europe and worldwide.

A key aspect of the collaborative efforts in this area is the sharing of PDXs between developers and researchers. To maximise the utility of these models, it is essential that they are established in a traceable and quality-controlled manner, and that all necessary secondary information is provided, thus allowing each PDX to be fully exploited. Moreover, in order to extend research efforts beyond a single lab or institute, interoperability and the establishment of common standards are also essential, not only for biobanking and quality control aspects, but also for the performance of in vivo drug efficacy studies. Finally, a major bottleneck in the process of model sharing is the different health status of the animal facilities, and the specific microbiological status of PDX models. The availability of clear information about the latter is of crucial importance, as it can directly influence experimental parameters and impose a risk to the personnel.

In the frame of the EurOPDX Research Infrastructure set-up, we will present our current efforts to devise common operating procedures and standards across the central nodes of the project, and ensure that all models offered for access have been quality-controlled, health-monitored, and are presented with all the required secondary data. Therefore, in order to facilitate the transfer of PDXs to users, an accompanying fact booklet has been developed. This “PDX Passport” summarises all available health screening results, ethical considerations and minimal information, as defined by the PDX minimal information (PDX-MI) guidelines* and in collaboration with various partners across the EurOPDX research infrastructure project. Here, we will present a first draft of the PDX passport. We aim to extend the use of the PDX passport beyond the confines of the EurOPDX research infrastructure and consortium, eventually reaching consensus at an international level, thus promoting collaborative efforts in the field of PDX research as a whole. Through this initiative, this biological resource can be widely distributed to foster scientific excellence, guarantee interoperability, and comply with ethical and legal requirements.

Ioannis Zavrakidis, Michael Hauptmann, Katarzyna Jozwiak

The Netherlands Cancer Institute

Introduction

In typical PDX experiments, tumor cells are injected into mice and the volume of the growing tumor is measured every 2-3 days until the mouse dies or is sacrificed. When the tumor reaches a certain volume, mice are randomized into treatment and control groups, and the rate of tumor growth is to be compared between groups. Often, average tumor size is compared at subjectively selected time points using a t-test or ANOVA, but this does not take into account the correlation among repeated measurements.

Material and Methods

We evaluate the performance of mixed effects models to test differences in tumor growth between groups of mice. Mixed models can accommodate random effects, i.e., factors whose impact on the outcome might differ per mouse, and the correlation among repeated measurements from the same mouse. These methods use all data from an experiment and avoid arbitrary selection of a time point. Statistically, an interaction term between time of volume measurement and treatment group is included to evaluate whether tumor growth differs between groups. The small-sample properties of these models for this particular application are not known. We therefore use data from two real experiments as well as simulated data to evaluate the operational characteristics.

Results and Discussion

Mixed effects models performed well under all scenarios. Random effects seemed not to be necessary since all mice start at the same baseline of the pre-defined tumor size. However, accounting for the correlation structure in the residuals was important.

Conclusion

Mixed effects models are the state-of-the-art method for experiments comparing tumor growth rates and are preferable over cross-sectional analyses using a T-rest or ANOVA. We show how sample sizes can be calculated for such experiments.