Presenting at the DIA’s 21st Annual Workshop in Japan for Clinical Data Management, I had the opportunity to compare academic studies in the US vs those in Japan as well as the differences in data management practices between academic and industry studies.
I researched with interest a paper published recently comparing investigator-initiated trials (ITTs) between Japan (data extracted from Japanese registries) and other countries (data extracted from clinicaltrials.gov). It was initiated because, despite the new ability to submit ITTs for investigational new drug (IND) applications in Japan as of 2003, the number of new ITTs for this purpose in Japan remained low (56 trials over 8 years), especially when compared with other countries. Comparisons were conducted for academic/government-sponsored trials and for industry-sponsored trials; the academic/government-sponsored trials in all countries tended to be early-phase with relatively small sample sizes.
This seemed relevant for a presentation for this conference, in Japan, about how academic studies and industry studies take different approaches to clinical data management. This could be explained by the study characteristics in each domain. Academic studies tend to be smaller in terms of both budget and scale and aim to answer a specific research question, compared with industry studies, which typically have a large budget and scale and aim to bring a product to market.
As a result, academic studies are more likely to continue to use paper-based data collection and manual transcription of data into the electronic data collection (EDC) system, in which the forms have been created by the academic center itself. Those forms are usually designed by the build team, less frequently with standards. Vendor-contracted EDCs are preferred by industry studies, containing forms that were created using CDISC standards by the EDC vendor or CRO. Herein lies the first big difference – academic research centers often do not have data standards: the same type of data can be entered and stored differently study by study, limiting any potential data sharing across studies or across the academic industry. The National Institutes of Health (NIH) does require investigators to have data sharing plans for larger studies, but the prevalent smaller studies, which could benefit from data pooling, are excluded from these requirements. In contrast, industry tend to use big data repositories based on industry standards, which facilitate data sharing and pooling.
These data collection processes also have trickle down effects to data cleaning activities. Not only are manually transcribed data likely to require data cleaning, due to human error and delays in data access for monitoring purposes, but academic centers are more likely to use tools such as Excel and SAS for their data cleaning activities. Data cleaning by data managers according to a prespecified Data Management Plan (DMP), as recommended by the Society for Clinical Data Management (SCDM), is more often used by industry.
However, even across the industry, manual processes to track clean patient data continue to be more prevalent than automated processes that provide system-generated output for review – based on a 2017 survey that we conducted. We know through our own experience at Bioclinica that programmatic-based analyses allow the team to concentrate only on the tasks that are needed for patients to be clean, rather than aiming for 100% SDV. This not only increases the velocity at which the patients are determined to be “clean” but also expedites database lock. Our advanced analytics in our Clean Patient Optics service provide instant visibility into these remaining tasks.
Despite the ongoing disparities in practice, we are seeing some convergence in the data management practices between academic centers and industry, in part driven by the intersection of real-world evidence (RWE)-based trials, outcomes-focused trials, and clinical trials. There is increased use of standard operating procedures (SOPs) and evidence of documented processes, and investigators are increasingly being required to have data sharing plans. As part of any study, a DMP should be designed with the intent of sharing and mining collected data as part of larger pooled datasets, to help with decisions regarding the safety and efficacy of treatments.
Of course, there are tradeoffs to be considered when selecting data capture systems for academic studies, including the price model, ability to handle patient surveys, and use of cloud-based solutions. In addition, the presence of core functionality such as queries, CRF library considerations, and the ability to work in SDTM, Excel and SAS are other considerations.
A good place to start in understanding clinical data management is to review the SCDM Good Clinical Data Management Practices (GCDMP). And, if you’re interested in learning more about our EDC and Clean Patient Optics offerings, you can contact me at David.Kiger@bioclinica.com.