A risk proportionate approach to GCP training might say that, for clinical samples which are not involved in safety or efficacy parameters and which are fully anonymised, GCP training would only need to cover the data quality and data integrity aspects.