3 Key Questions When Developing the Integrated Summary of Safety (ISS)
January 7, 2014
Rob Woolson is a senior biostatistician with 12 years’ experience in the analysis of complex data. He has conducted statistical analyses in all phases of drug development (Phase I through IV, including NDAs) and has led SDTM/ADaM dataset conversion projects in multiple therapeutic areas. He has held a leadership role in six CDISC-compliant FDA submissions, having guided the creation of ISS/ISE statistical analysis plans; integrated analysis dataset design and production; integrated display design and production; and submission-related documentation development.
A new drug application (NDA) covers information about a product from inception through clinical trials. The integrated summary of safety (ISS) is a section of the NDA that provides comprehensive safety information collected throughout the development program. The goal of the ISS is to characterize the overall safety profile of the drug and to identify risks that should be included on the product label. This article discusses three key questions to address as a part of your ISS analysis plan.
(1) What are the safety parameters of interest?
Safety parameters of interest typically include those specified in FDA guidance, those that are a priori special interest or concern for the program or compound, and those identified during data review. Some examples of safety parameters are exposure, concomitant medications, deaths, adverse experiences (occurrence, relatedness, severity, seriousness, duration, timing, etc.), laboratory measures, and vital signs. A good test of whether one has selected the right parameters is to ask whether the summary of the estimates of these parameters sufficiently describes the overall drug safety profile.
(2) How does one present and analyze data to convey key safety messages and describe the overall safety profile?
Once safety parameters have been selected, one must decide the best way to present them. Some guiding principles we’ve come up with are:
- The presentation should help a reviewer get at the true value of the parameter of interest.
- The parameter should be presented with a high degree of confidence (maximum precision, minimum bias).
- The presentation should permit the reviewer to make meaningful comparisons between active and placebo. An absolute means little unless it is compared to something.
There are a number of ways we can characterize these parameters including proportions (i.e., a crude rate), incidence rate (per unit time), total incidence rate (events per unit time which may be useful where there are multiple events per subject and different exposure times), time to event, and change from baseline. One challenge frequently encountered in presenting safety data is that different trials have collected data differently which lead to additional complexity in how data should be presented in the ISS. For example, one may have to deal with several different follow up times. The solutions selected must convince a reviewer that they are provided an unbiased presentation of the data. There are several methods that can be used for making between group comparisons. These methods include difference of proportions, ratio of proportions (mentioned in FDA guidance), difference in rates, ratio of rates, hazard ratio, survival curves, and difference in means.
(3) Should safety data from all studies be pooled and, if so, how?
One of the first things to consider in characterizing the information is if and how data will be pooled. There is a regulatory obligation to present all safety data; however, there is no requirement that data from all studies be pooled.
It is important that whatever pooling strategy is taken, one is prepared to justify it to FDA reviewers and/or an advisory committee.
- A reason to pool data is that one may be able to provide more precise and more reliable estimates of safety parameters. Also, pooled data may allow conclusions to be drawn (e.g., comorbidities) that wouldn’t be seen by looking at the studies individually.
- A reason not to pool data is that studies and the populations in those studies may be so different that it is difficult to make sensible comparisons. Additionally, analyses of pooled data can be time consuming and expensive when compared to summarizing and presenting the analyses that were performed as part of the studies. (Going too far down that path, however, may lead reviewers to believe the analyses are insufficient.)
In most cases, it will make sense to pool at least some data. Some general principles for creating a pooling strategy:
- Combine the data in the most valid way(s)
- The safety message should drive the pooling strategy
- Summaries should produce transparent results
- No masking of safety signals
- Estimates produced should be as unbiased as possible
Some factors to consider when looking at individual protocols to determine whether it makes sense to pool data from those protocols are:
- Design similarity
- Doses studied
- Controlled/uncontrolled/choice of control
Keep in mind that differences in these factors introduce variation to the safety parameters of interest.
In conclusion, it is important that the finished ISS tells FDA reviewers a clear and focused story about the drug’s safety profile. Doing this effectively is necessary to move through the approval process efficiently.