Over the next few posts, I discuss the data review process for clinical trials and highlight some new features for JMP Clinical 4.1 that streamline this monumental endeavor. Ideally, the data from a clinical trial should be examined by as many eyes as possible – including data and protocol managers, monitors, clinicians, regulatory associates and statisticians. Each individual brings a unique skill set that is important for understanding the safety of the patients under study, whether the protocol is being adhered to appropriately, or whether insufficiencies in the data can affect the analysis at the end of the trial. It’s important to note that even small, relatively straightforward trials can generate large amounts of data. So, in general, the data review effort for any trial can be very time-consuming.
During my days in the pharmaceutical industry, the biostatistics and programming group would regularly supply listings of the study data so that individuals in other departments could understand specific questions about safety or protocol compliance. Inevitably, as these reviews continued, the question from our clinical colleagues became: “Is it possible to just provide the new data so I don’t have to review what I’ve already seen?”
Of course, if you’ve ever had to program any of these tables or listings, this simple question may not be as straightforward as it initially sounds. It gets even more difficult when there is interest in seeing previously available values that change during the course of the study (well, what was this value before?). And to further complicate things, we found that what people really wanted was the new stuff AND the previous data as well since in order to truly appreciate the patient’s current status, you need to understand his or her prior condition. In other words, they wanted the ability to subset or filter to new data at will. In a paper or static (say, a PDF of a table) world, however, this had the potential to generate twice as many analyses or require flags on each individual piece of data in a listing.
JMP Clinical 4.1 has several enhancements to solve these classic data review dilemmas. When a study is added or updated to a new data snapshot, JMP Clinical collects a host of new information about each data record:
Is this record new? If not, was there a change in this record since it was previously available?
For which variables did changes occur, and what were the previous values?
Are there duplications of any records in my data?
All these questions are now easily answered.
Figure 1 shows output from the new Domain Viewer Analytical Process (AP). Records are labeled as previously available with no changes (green), new (yellow) or previously available with a modification (red). For modified records, cells are highlighted to show which variables experienced a change between the current and previous snapshot of the data (here, the SDTM variable Subject Reference End Date/Time). The user can select these rows and click View Record-Level Notes to see what changes occurred for this variable (or any others, if applicable). These notes are generated by JMP Clinical when the study data is updated, and the user is made aware that notes are available for a given record based on the row marker (an asterisk). Further, the Domain Viewer AP enables the user to create and save their own record-level notes, show previously-available records that were dropped from the current snapshot, as well as highlight any records that may be duplicates.
Figure 2 shows output from the AE Distribution AP. In many JMP Clinical analyses, subsetting to new data is as easy as selecting the “New” AE Review Flag from the JMP Data Filter. Further, all clinical APs allow the user to create and review notes at the AP- or subject-level. Any user-defined notes are date-time stamped and saved to a central location for review at any time.
The ability to understand data changes at the click of a mouse, create and review notes from any AP or to subset analyses to new records on the fly allows for truly efficient data reviews for clinical trials. A clear understanding of last-minute changes also provides a degree of comfort when locking the study database. In these crucial instances, it is imperative to understand any final changes that make their way into the final study data sets. Overlooking an unexpected last-minute change may result in unlocking the trial database, lost time and other unpleasantness.
Next week, we discuss what makes this all possible -- it’s all about the KEYS!