By Kristen Booth
Late last month, the FDA published a guidance document in which the Agency clarifies the specific criteria that should be considered when deciding whether it is appropriate to leverage existing clinical data to support pediatric claims, and if so, to what extent. The guidance, entitled “Leveraging Existing Clinical Data for Extrapolation to Pediatric Uses of Medical Devices,” is the finalized version of FDA’s May 6, 2015 draft guidance. The final document was released on June 21, 2016, but does not take effect until September 19, 2016, giving manufacturers nearly three months to understand and comply with the new regulations and guidelines.
Borrowing Strength from Adult Data
“Extrapolation enables a sponsor to leverage adult data to support demonstration of a reasonable assurance of effectiveness and possibly the safety of a medical device for pediatric use.” In its guidance, FDA states that sponsors can use the quantitative information that is provided by the existing adult data in one of two ways. Extrapolated data can be used:
- As a substitute for pediatric data
- To supplement pediatric data within a statistical model
The statistical model that combines both the adult and pediatric data has the potential to strengthen the scientific evidence that is available to show effectiveness and/or safety among the pediatric population. This is known as “borrowing strength.” The FDA notes that the exact model used to borrow strength can vary from case to case, but regardless of the model used, the degree of leveraging depends on the similarity between borrowed data and any pediatric data that will be collected. Furthermore, “the extent of borrowing may also be moderated by clinical judgments that are not inherently implied by the statistical modeling. This may include consideration of the quality of the data, the particulars of the populations and the studies, and whether such data are intended to demonstrate either safety or effectiveness (or both).”
Full vs. Partial Extrapolation
According to statistic modeling, existing clinical data can be leveraged either fully or partially to support the safety and/or effectiveness of a medical device in a pediatric population. With full extrapolation, existing data is used as a complete substitute “for prospective pediatric clinical data in support of a determination of a reasonable assurance of effectiveness or of safety for a pediatric device. No prospective pediatric clinical data are anticipated for the endpoint being fully extrapolated.” Because of the number of potential differences between adult and pediatric patients, it is expected that full extrapolation will rarely be used to demonstrate safety.
In addition, manufacturers can also use partial extrapolation, which is when existing adult data and pediatric data sources (or prospective pediatric clinical data) are combined (using a statistical model) to demonstrate a claim of safety and/or effectiveness. The guidance document states that “the construction of such a statistical model is anticipated to require the availability of measured variables that will help connect the adult outcomes to the pediatric outcomes. If necessary variables are not available in the data sources, partial extrapolation may not be appropriate. If the model is determined to be appropriate, then the inferences obtained from it may be used to support a pediatric indication.”
Extrapolation for Effectiveness vs. Safety
At times, existing clinical data can be used to support claims of safety, effectiveness, or both in medical devices. However, oftentimes the study endpoints used to demonstrate effectiveness are typically different from those used to demonstrate safety. Because of this, FDA recommends that manufacturers consider whether the data will be extrapolated for safety or effectiveness (or both) independently.
Furthermore, due to the physiological differences between adults and children that may have an impact on device safety, FDA anticipates that extrapolation for safety will be rarer than extrapolation for effectiveness.
Pediatric Extrapolation Decision Process
In the guidance document, FDA states that the appropriateness of extrapolation largely depends on the following factors:
- “The similarity of the existing adult response data and/or population characteristics to the intended pediatric sub-population.”
- “The quality of the adult data in terms of study design, data collection, and measurement.”
- “Whether extrapolated data may be used to fairly and responsibly decide whether there is a reasonable assurance of the safety and effectiveness (or probable benefit, for HDEs) of a medical device (i.e., constitute valid scientific evidence).”
In situations where similarity and quality are both determined to be high, there is a great level of certainty that the existing data can be appropriately considered for extrapolation to the intended pediatric subpopulation. Conversely, in circumstances where neither similarity nor quality are determined to be high, it may not be appropriate to use the existing adult data for extrapolation purposes.
When determining if it is appropriate to extrapolate existing data and, if so, whether extrapolation should be full or partial, FDA recommends using a decision tree. In the guidance, FDA specifically states that the decision tree is intended to act only as an aid in making the final judgement. As such, “a conclusion from the decision tree that extrapolated data may be used does not necessarily mean that these data will support an approval decision for the PMA, de novo, or HDE application.”
In the guidance, FDA provides a sample of the decision tree that sponsors should use as well as a list of the questions to consider in using the decision tree. For additional details, view the full guidance document.
Factors that Could Limit Extrapolation
There are a number of general factors that can aid in determining whether, and to what extent, extrapolation is necessary and appropriate.
Factors that may prevent the extrapolation of adult data include, but are not limited to:
- “There is little knowledge of the disease or condition in pediatrics.
- The device is not FDA-approved or -cleared for adults.
- Endpoints cannot be directly borrowed.
- Statistical models cannot account for differences.
- Human factors and growth can affect safety in pediatric patients (these factors don’t exist in adults).
- Appropriate labeling cannot be written for the pediatric population or subpopulation(s) targeted.
- The practice of medicine has changed since the device was initially approved to such an extent that historical data would likely be different than prospectively collected data.
- Appropriate risk mitigation cannot be assured.”
“Factors that may limit extrapolation to a partial extent and thus require conducting a prospective study of pediatric patients include, but are not limited to, the following:
- The age difference between the pediatric (sub)population and the available adult data is too great, making it difficult to infer similarity in risk or effectiveness. In such cases, it may be more appropriate to extrapolate to a pediatric age that is closer to the mean age of the adult population. For example, it might be more appropriate to extrapolate young adult data to an adolescent indication than to a neonate indication.
- Other supportive pediatric data are outdated and may not properly represent current treatment trends and practices.
- There are important differences between the adult and pediatric (sub)population(s) such that the adult data cannot substitute for data from a prospective pediatric study to fairly and responsibly conclude that there is reasonable assurance of the safety and effectiveness of the device in the pediatric population.”
Uncertainty in Extrapolating Data
Extrapolation, whether full or partial, adds uncertainty into FDA’s assessment of the device’s effectiveness and safety. The extent of this uncertainty is dependent upon the differences between the two populations, as well as the quality of the data. When making benefit-risk determinations, FDA takes this uncertainty into consideration.
As such, FDA will only permit extrapolated data to be used when it can be done in a way that supports reasonable, scientifically sound conclusions about medical device effectiveness and safety based on valid scientific evidence.
Statistical Methodology for Extrapolation
When it is appropriate to extrapolate adult data, sponsors have various options for how to do so. “Available options could depend on whether a prospective study of pediatric patients is needed and feasible, and/or whether sufficiently robust pediatric data can be obtained in other ways, such as from prior studies run by the sponsor, studies in the literature, or pediatric registries.”
FDA provides fairly explicit details regarding the options for extrapolating data. For additional information on these options and the statistical methodology involved, view the full guidance document.