The text below provides responses to CBER’s September 16, 2009 comments on the draft concept protocol (submitted to the BLA on August 13, 2009) for a phase IV epidemiological study to assess the incidence of autoimmune disease following administration of Cervarix. For ease of review, CBER’s comments are presented in bold text followed by GSK’s responses.
1) Please include “reactive arthritis” among pre-specified endpoints.
“Reactive arthritis” will be included in the aggregate primary and secondary endpoints, as well as the exploratory endpoints.
2) Section 7, related to adverse events and serious adverse events, will require further clarification.
The objective of this study is to evaluate the incidence of autoimmune disease following administration of Cervarix. Section 7 describes the proposed reporting of only adverse events that are not included in the objectives and are considered serious. These serious adverse events (SAEs) will be reported only when:
1. they are verified in the medical chart AND
2. they are recorded in the medical record as related to the vaccine;
The reason for applying these criteria is because of the database nature of the study.
SAEs will be reported to GSK within 24 hours of notification to the principal investigator. GSK will then make the assessment of expectedness and will proceed with any expedited report submissions to health authorities, including CBER. GSK will expedite reports of such events to VAERS within 15 calendar days following chart review which includes confirmation of diagnosis and causality assessment.
3) As an exploratory analysis, please conduct SaTScan (spatial and temporal scan statistics) or similar statistical analysis to identify onset interval clusters of all primary and secondary endpoints specified above within the 12 month follow-up period.
GSK agrees to use a "scan statistics" method to detect possible temporal clusters of outcomes. This will be an exploratory analysis by applying this approach to the two co-primary endpoints and the secondary endpoints but not to each individual auto-immune disease because of the risk of a false safety signal due to multiplicity.
4) Please define time windows, after vaccination, specific to each adverse event.
By “adverse event”, GSK has assumed that CBER is referring to the individual events included in the proposed endpoints. Our primary and secondary objectives consider two aggregate AID endpoints; the only instance were we will be assessing individual autoimmune events is in the exploratory objective. For the aggregate AID endpoints, as advised by AID experts, we have considered a 12-month follow-up period after the first dose of Cervarix, regardless of the number of doses received. We will apply criteria in the study protocol for varying dosing schedules that frequently occurs with routine vaccine use.
For the exploratory objective we can develop a time window for the most probable occurrence of AID, if it is available in the literature or in our databases.
5) Please perform primary analyses of serious adverse events occurring with onset times which are specific to the adverse event, in addition to the proposed 12 month follow-up.
GSK requests clarification of this comment. Is CBER requesting analysis of serious autoimmune events that are included in the proposed endpoints?
If so, we have developed aggregate endpoints as primary and secondary objectives to address the issues of multiple statistical tests. Individual autoimmune events will be assessed only in the exploratory objective. Is CBER proposing a primary analysis of each of the 37 AIDs each with a specific time window? We can develop a time window for the most probable occurrence of AID as mentioned in response to Question 4, but we are not clear on how to address the multiple testing issue.
6) Please implement additional self-control design methods in the MCO-base study for all primary and secondary endpoints specified above.
We did not consider self-control methods as our main analyses for the following reasons:
a. The aggregate nature of the endpoints. GSK is not aware that this method has been used for such an aggregated endpoint.
b. The chronic nature of the outcome. The self-control analysis is not suited for chronic diseases, which is the case for most AIDs.
c. It is unclear which period could be used as control period. Since it is very probable that occurrence of an AID will affect the decision to vaccinate, the period before vaccination cannot be used. The only alternative control period is the time after the risk period, which is more than a year after the first dose. The acceptability of this period as completely risk free can be challenged, however.
Given the limitations and uncertainties and to reduce the risks associated to multiple testing, GSK prefers not to include such testing in the analytical plan.