FDA Goes In-Depth on Using EHR and Claims Data in Drug and Biologic Trials
The FDA has issued its most detailed guidance yet on using electronic health records (EHR) and medical claims data in drug and biologic clinical trials, publishing an extensive 35-page draft guidance for sponsors and researchers.
The guidance applies to all study designs that look to use EHR or medical claims data to help show safety or effectiveness and first touches on general considerations, including advice on selecting data sources to maximize data accuracy and completeness. All studies intending to use and submit such data need to have their protocols and statistical analysis plans submitted prior to initiation, the guidance explains.
Protocols also need to identify all data sources the study plans to use. A description should be included in the protocol that explains how well the selected data sources have historically captured study elements (such as inclusion/exclusion criteria, exposures, outcomes and key covariates) and how the data can be validated for a particular research task, the agency noted.
In addition, all of the essential elements of study design, analysis, conduct and reporting should be predefined, and for each element, the protocol and final study report should describe how they were determined from the selected real-world data source, including applicable validation studies, the guidance says. The guidance includes a specific section on study design elements that goes into further detail.
“I think real-world evidence has been accumulating, and so now because it’s coming from many sources, [the FDA is] trying to create a structure, that there are common rules, if you will, for how data are aggregated, interpreted and how research questions are posed to use it,” Sandra Smith, senior vice president of clinical solutions and strategic partnerships at WCG, told CenterWatch Weekly.
For data sources, the guidance stresses the importance of understanding potential limitations inherent in existing electronic healthcare data. Examples to consider include:
- Medical claims data’s purpose being to support payment for care. Claims may not accurately reflect a particular disease or a patient may have a particular disease or condition that isn’t reflected in claims data;
- EHR data being generated for use in clinical care, billing and auditing practice quality measures. Data recorded in an EHR system depend on each healthcare system’s practices for patient care and the clinical practices of its providers. Additionally, data collection is limited to the data captured within an EHR system or network and might not represent comprehensive care; and
- Limited ability to add modules to EHR systems to extensively collect additional patient data during routine care means EHR-based data collection may not be comprehensive.
“Part of what’s … fueling this is there’s just a lot more data aggregation, so you do have medical records that are being used now, not only by hospitals but by private practices, by homecare agencies, skilled nursing facilities, so the healthcare industry as a whole has pretty much digitized the patient data that they are collecting,” Smith said.
“We also have more mobile devices where we can gather patient-reported outcomes. The wearables, the biosensors, all of that is just gathering large amounts of health-related data, and so now, this data is being looked at as to how they can better design and conduct clinical trials and answer questions that before may not have been able to be answered,” Smith added.
Because of the differences in healthcare systems, patient characteristics and medical practice globally, data may not be as relevant for certain studies. With that in mind, the agency recommends providing reasoning for selecting particular data sources to address specific hypotheses and a description of prescribing and use practices in the specific healthcare system the data is being drawn from, if it’s available.
In addition, background information should be provided about the healthcare system, including any specified method of diagnosis and preferred treatments for the disease being studied, as well as the degree to which it collects and validates the information in the proposed data sources.
“In certain types of care, and I’ll use oncology as an example, many organizations have adopted clinical pathways which give a roadmap for the care of patients with certain types of cancers. If an organization is following a certain clinical pathway, there may be some differences in how one organization is treating than another,” Smith said. “A lot of those pathways are based on scientific evidence from previous studies, but these guidances are digging more deeply to make sure that when data is being used from multiple sources, that it can be harmonized appropriately.”
The guidance also notes that EHR and medical claims data are only collected during routine care; thus, the information a study needs to answer certain questions may not be provided by EHR and medical claims data sources. Sponsors should be able to show that each of their data sources will provide enough detail and completeness to fully capture the study populations, exposures, key covariates, outcomes of interest and other parameters, the guidance says.
When using EHR and medical claims data sources, it’s important to document patients’ continuity of coverage, as they often switch health plans throughout their lives, such as when changing jobs. The validity of results from trials using such data depends partly on the documentation of patients moving in and out of health plans and healthcare organizations. The guidance explains that this enables studies to define enrollment periods (when data are available on the patient of interest) and disenrollment periods (when data aren’t available). Accordingly, definitions of enrollment or continuous coverage should be developed and documented in the trial protocol, the FDA said.
The guidance also cautions trials that intend to use drug coverage and medical care data from patients with certain privacy conditions, including sexually transmitted infections, substance abuse and mental health conditions. Obtaining comprehensive amounts of this data can be difficult and failing to do so can lead to incomplete or incorrect information. Additionally, the agency pointed out that there are certain populations that sign up more often for experimental clinical trials, such as certain cancer patients or patients who get their medicine from pharma company assistance programs. In these cases, electronic healthcare data sources may not give a complete picture of their health data. If these issues are relevant to a study, the protocol should lay out how they will be addressed.
Additionally, if uncaptured prescriptions are relevant exposures to the trial, the guidance cautions those using medical claims data that there could be dispensed prescriptions not associated with insurance claims. Uncaptured prescriptions may consist of low-cost generics, drugs gotten through discount programs, drug company samples dispensed by healthcare providers, drugs sold over the internet or out-of-pocket purchases, the agency said.
It also includes a general discussion on capturing data and a section on things to consider when examining the quality of data over the course of its lifecycle, which may vary depending on the type of data and setting but generally involves multiple phases. In addition, it lists considerations to make for data quality control procedures and quality assurance.
“This really has been … building, so now, with even more detailed guidance … [the agency is] trying to provide more guidance on defining the relevance of those data sources, so if studies were being designed to use these data sources, objectively defining why those data sources are being chosen, having a little bit more background on healthcare systems that this data is coming from,” Smith said.
Read the full guidance here: https://bit.ly/3onSyBO.