Data Mining as a Path to Better Trial Design
The prospects of Big Data are tantalizing for the life sciences industry, but there’s still much left to do with small data.
Significant inefficiencies remain in the pharma industry, and pharma companies can look closely at available data to create value – starting with how to determine the feasibility of a new clinical trial.
Robert Califf, the FDA commissioner nominee, has spoken in recent months about using available data to improve the clinical trials process. “Improving the quality and efficiency of clinical trials will leverage improvements in many other aspects of the healthcare system. The only way we can get there is to use integrated information,” Califf said at a Tufts Center for Drug Development event in May. [Disclosure: I was a speaker at the event.]
The pharma industry is one of the few industries where the level of risk increases as a company gets closer to a product launch. We need new trial designs to correct that. The pharma industry is phenomenal at statistics, but it doesn’t always include that analysis upstream in the clinical trial protocols where it should be.
When sponsors design clinical trial protocols, they look for data that can be used to calculate a trial’s feasibility. Insurance data or electronic health record data can often help assess the number of patients that meet the trial’s eligibility criteria and where the highest concentrations of patients are located.
But many sponsors overlook the huge amount of internal data available that can be used to determine the likelihood of success in study feasibility. Sponsors can examine enrollment rates for trials with similar inclusion/exclusion criteria; dropout rates in studies with specific clinical procedures; the most common reasons for protocol amendments; and how common a particular procedure is used per therapeutic area.
However, this data can be buried in documents and spreadsheets that cannot be searched. Sponsors need tools in place to take advantage of internal data as well as industry benchmark data.
Additionally, it’s vital to be able to identify previous studies with a similar set of design characteristics to the current study. This requires the adoption of a structured approach to capture core protocol design information such as eligibility criteria and clinical endpoints.
Ideally, this protocol design information is then paired with operational data. Enrollment rates, dropout rates, the number of protocol amendments and the reasons for the amendments – all of these operational data points can provide more powerful insight when they can be linked to to protocols with clinical trials with comparable objectives.
If a sponsor company intends to use data mining to improve its protocol design, it is critical that a master data record across all trials. A master data record can help sponsors identify clinical procedures from trial to trial and connect information about a procedure’s cost, difficulty and its burden to the patient between trials. This can be done by connecting procedures to a common code such as the American Medical Association Current Procedural Terminology code.
By linking procedures to previous studies, sponsors can identify procedures that have an impact on a site or a patient’s willingness to participate, especially if it is a complex or invasive procedure. The same data can be used to determine if a procedure is particularly unusual in a specific clinical trial phase and indication.
The Tufts Center for the Study of Drug Development estimates that approximately 25% of procedures in a typical trial are not associated with primary or secondary clinical endpoints. While tertiary and additional endpoints can provide insight, unnecessary procedures increase the complexity of a study, make it harder for a clinical trial site to execute, and potentially makes the study less appealing for patients.
Sponsors can use structured data to analyze the cost and complexity of each objective to ensure that procedures can support the most important goals of a study. As Dr. Califf says, the pharma industry can improve the quality of efficiency of clinical trials, but it requires integrated information to get there.
This article originally appeared in our MedidataVoice series on Forbes.