People ask what does Atorus Research do?
Clinical Trials are experiments.
They form a hypothesis, build an experiment to test said hypothesis, and the test results may or may not bring confirmation, inconclusiveness, or review more questions. The community is competitive in racing to bring new products to market or expand labeling of currently approved ones. With approval, results must be shared to the extent that the labelling can guide the consumer as to what it can be used for and any potential hazards that may be associated with the product.
The protocol is the document stating the hypothesis and experimental methods. It has both a clinical and statistical hypotheses and design.
In the past 30+ years, protocols have been harmonized, and data has been standardized through CDISC with the FDA expecting SDTM as a default data vehicle. Data collection has jumped from double-data entry of paper CRFs to the advent of EDC, and now the emergence of EHR for quicker data accessibility. The Prescription Drug User Fee Act (PDUFA) brought change to the agency in accelerating the review cycle of submissions. All these innovations assist in the competitiveness of drug development, bringing new therapies quicker to the agency and faster review, leading, hopefully, to a successful product to help the patients who need them.
What hasn’t really changed?
Statistical appendices and graphics for FDA submissions have not changed.
SAS has had a lock on the market due to fear and false belief in the user community that their in-house developed macros are sacrosanct, proprietary, and should not be shared, and that they bring a competitive advantage to the statistical community.
Why do we have so many companies with legacy macros that are so complicated and unable to be modified without fear of breaking. Why are there CROs who will not share their code with customers when producing tables, listings, and graphs? Does this help or hinder the industry? Remember, these outputs are produced quickly at the end of the study to describe the results. When only a few know the intricacies of the code, there will be bottlenecks and delays.
This blockade must end.
Open source languages, like R and Python, bring all the statistical expertise of a community the size of tens of thousands of people. It is widely used and taught in academia, and many reviewers in both industry and the FDA use it as a QC tool for clinical review. Individuals write the code then share it with the community who ARE STILL THE OWNERS OF THE CODE. The code can be reviewed and critiqued by all, and the number of people who can learn and use a given package dwarfs what any single organization can hope to achieve. The code deemed proprietary to an organization for competitive reasons can remain so but still tap into outside packages for efficiencies.
The FDA stated in 2015 that it is not married to any one statistical software, https://www.fda.gov/media/109552/download; rather, that the statistical analyses should be reliable and “fully documented in the submission, including version and build identification.” This is a sound, scientific request.
Atorus Research has made a commitment to open source. We are spending millions on the development of code to be shared across industry. It is our belief that as people use our validated packages and provide feedback, together we can accelerate the commercial, regulatory, and academic communities review and analysis of information in a format agreeable to all. This disruption is necessary for the sake of the patients. The practice of considering SAS code proprietary used to generate a demographic table, which is part of an ICH mandated output, has lived long past its usefulness. We are not stating SAS is bad; it is a fantastic language. Ask yourself, does industry really need 10,000 different versions of the same code?