Day 7 - September 15, 2021
Erika Cheung described Edison quality-control failures, data-point removal, proficiency-testing disparities, and her internal and external reporting. Cross-examination addressed her limited role, Theranos validation records, compressed timelines, and differences between written procedures and laboratory practice.
Full day summary
2. Erika Cheung — Direct/Cross (Continued)
Erika Cheung finished direct testimony about Theranos laboratory concerns before the defense began cross-examination on her qualifications, assay validation, quality control, and laboratory procedures.
Highlights
Erika Cheung — Directtestimony highlightCheung explained that VIP demonstration samples were divided among three teams and processed using Edison devices and third-party analyzers; Edison testing could be dropped when delays or problems arose.Erika Cheung — Directtestimony highlightCheung testified that Theranos frequently removed two of six generated data points without a defined outlier protocol, which could make QC or validation data appear better.Erika Cheung — Direct“SO IT -- THERE WAS NO STANDARD PROTOCOL AT THERANOS FOR WHAT OUTLIER DELETION WAS. IT WAS SOMETHING THAT WAS CONDUCTED QUITE FREQUENTLY, AND IN NORMAL SENSE THIS WOULD BE CONSIDERED CHERRY PICKING, WHERE YOU'RE ESSENTIALLY PICKING THE BEST DATA POINTS TO MAKE IT APPEAR THAT THE QUALITY OF THE DATA IS BETTER THAN IT IS.”— Erika CheungCheung characterized Theranos's outlier-removal practice and explained why she believed it overstated data quality.Erika Cheung — Direct“IT HAD VERY POOR PERFORMANCE. WE WERE HAVING QC FAILURES AT A MAXIMUM EVERY THREE DAYS, POSSIBLY MORE.”— Erika CheungCheung gave a memorable assessment of the frequency and quality of Edison QC performance.Erika Cheung — Directtestimony highlightComparative proficiency-testing data showed disparities between Edison and predicate-device results, including inconsistent Edison reruns; Cheung testified that Edison-generated proficiency results had not been provided to regulators during her employment.Erika Cheung — Direct“AT MY TIME AT THERANOS WE DID NOT PROVIDE REGULATORS WITH THE RESULTS GENERATED FROM THE THERANOS EDISON DEVICES SIMPLY FROM THE PREDICATE METHOD. SO THIS INFORMATION WAS NOT PROVIDED TO REGULATORS.”— Erika CheungCheung testified from her involvement in proficiency testing about which results were provided to regulators during her employment.Erika Cheung — Directtestimony highlightMarch 2014 data showed a 25.6 percent overall Edison QC-run failure rate, which Cheung described as typical rather than an unusually bad month; she contrasted that with rare QC failures on third-party devices.Erika Cheung — Directevidence eventThe court admitted Tyler Shultz's email to Elizabeth Holmes solely to show notice, and later admitted the law-firm letter delivered to Cheung; Cheung said she subsequently contacted CMS about the laboratory concerns.Erika Cheung — Crosstestimony highlightCheung described her lab-associate position as involving basic test operations and acknowledged that she did not know several laboratory classifications, regulatory qualifications, or assay-specific accuracy and precision standards.Erika Cheung — CrossadmissionCheung explained the extensive R&D and clinical-validation workflow but said Theranos compressed work that would usually take about three months into periods ranging from three days to a week and a half; she generated data while others performed the analysis and incorporated it into reports.Erika Cheung — Cross“BUT AT THERANOS WE WERE RUNNING THEM IN A SHORTENED TIMEFRAME, SO WE WOULD JUST TRY AND GET THEM DONE AS SOON AS POSSIBLE. SO IT WOULD BE LIKE THREE DAYS TO A WEEK AND A HALF.”— Erika CheungAlthough the defense emphasized the amount of validation work, Cheung testified that Theranos compressed a process she said would usually take about three months.Erika Cheung — Crosstestimony highlightCheung agreed that QC samples were not patient samples and that patient testing followed successful QC, but she also testified that actual laboratory practices sometimes differed from the written SOP and that frequent reader failures led to added recalibration steps requiring substantial time.