FDA Statistical Review and Evaluation
Document
for the Vaccines and Related
Biological Products Advisory Committee (VRBPAC)
February 20 – 21, 2008
Rotarix, Live
Attenuated Human Rotavirus (HRV) Vaccine, Oral (GSK)
Indication:
prevention of rotavirus gastroenteritis in infants and children caused by the
serotypes G1, and non-G1 types (including G2, G3, G4, and G9).
Jingyee Kou, Ph.D.
FDA/CBER/OBE
GSK – Rotarix
Contents
- Overview
- Safety: Intussusception cases
- Efficacy
- Reviewer’s Overall Conclusion
1. Overview
In the application for licensure, the applicant GSK has submitted information from several clinical trials. This statistical briefing document presents only the results from study Rota-023, the pivotal trial that the applicant conducted to demonstrate safety with respect to intussusception and efficacy against gastroenteritis caused by rotavirus. The goal of this document is to provide information to the Vaccine and Related Biological Products Advisory Committee; however, it is not the final statistical review to the FDA.
- Safety: Intussusception Cases
The first rotavirus vaccine licensed, RotaShield (Wyeth), has been associated with an increased risk of intussusception (IS). A case-control study conducted by CDC confirmed that the risk of IS appeared to be increased among recipients of RotaShield during the 3- to 14-day period after the first dose and during the 3- to 7-day period after the second dose. Therefore, it is crucial for future rotavirus vaccines to demonstrate safety with regard to IS.
Because of the RotaShield experience, GSK’s trial design considered IS as the main safety endpoint for the clinical trial Rota-023. One of the two co-primary objectives in Rota-023 concerned the intussusception issue:
• In all subjects (N = 60,000), to determine the safety of GSK Biologicals’ human rotavirus vaccine (HRV) with respect to definite IS within 31 days (Day 0 to Day 30) after each HRV vaccine dose.
This objective was reached if:
- the upper limit of the two-sided 95% Confidence Interval (CI) of the Risk Difference for the percentage of subjects reporting definite IS within 31 days (Day 0 to Day 30) after any dose was below 6/10,000, a limit based on the study sample size and the anticipated IS incidence rate, and
- there was no statistically significant increase in the percentage of subjects reporting definite IS within 31 days (Day 0 to Day 30) after any dose (the lower limit of the two-sided 95% CI of the Risk Difference had to be below 0).
Study design
This study was designed
as a randomized, double-blind, placebo-controlled, multi-country and
multi-centre study conducted in 12 countries (11 countries in Latin America and
Graphic presentations of study design are presented for subjects followed only for safety from Dose 1 until Visit 3 (40,000 subjects planned) and for subjects followed for safety and efficacy from Dose 1 until Visit 3 and beyond (20,000 subjects planned, Subset A). [From ‘rota-023-report-body.pdf’ submitted by the applicant.]
Sample size for safety evaluation
In the original protocol, the primary safety objective was defined as: “With an assumed background rate of 3 IS cases per 100,000 in the placebo group during the observation period, 60,000 subjects will allow exclusion of an IS attributable risk greater than 2 : 10,000 vaccinees (observed attributable risk ≤ 1 : 10,000 vaccinees, upper limit of 90% CI ≤ 2 : 10,000 vaccinees, at least 80% power).”
However, while the trial was ongoing, it was determined that the background rate was much higher than 3/100,000. A different study estimated the background rate to be about 51/100,000. Consequently, the primary objective for safety was revised to its final form.
The applicant provided
the following statements concerning the changes in of the primary objective: [rota-023-report-body.pdf]
“As of 18 May 2004, a total of 14 IS cases were
observed within 31 days post vaccination period. This led to an overall IS
incidence rate between 2 and 4/10 000, which far exceeded the anticipated
definite IS incidence rate of 0.3/10 000 subjects that was expected to occur in
the Placebo group in this same time window. This higher incidence of IS could
reasonably be attributed to geographical differences and/or the active
surveillance for IS in the study.
The higher IS incidence was further substantiated by a
concurrent, prospective, multicenter epidemiological study conducted in the
same 11 Latin American countries as those participating in study 023. Study
epi-204 assessed the incidence of IS through active surveillance in children
less than 2 years of age and not vaccinated with HRV. An interim analysis of
the epidemiological study showed that most IS cases occurred before one year of
age. IS hospitalization was uncommon before two months of age, but increased
from three months and peaked at five months of age. Preliminary calculation of
background incidence rates in children < than 1 year suggested an overall
incidence of 51/100 000, with a range among countries [Study Report 99910/204;
Breuer, 2004].
Due to the higher overall IS incidence (study remained
blinded) the width of the CI of the Risk Difference had become so large that,
under identical IS incidences in both study arms (HRV vaccine minus Placebo),
the upper limit of the 90% CI exceeded the initially specified 2/10 000 limit.
Therefore the original criterion for meeting the co-primary safety objective
was no longer appropriate.
For this reason, the primary safety objective was
revised (see Section 5.8.1, amendment 3) so that a vaccine with an identical IS
incidence as placebo would meet the objective:
• The upper limit of the two-sided 95% confidence
interval of the Risk Difference for definite IS occurring within 31 days post
vaccination should be below 6/10 000, a limit based on the study sample size
and the anticipated IS incidence rate.
• There should be no statistically significant
increase in the incidence of definite IS occurring within 31 days post
vaccination (the lower limit of the two-sided 95% CI of the Risk Difference
should be below 0).”
Results from Applicant
A total of 63,225 infants
(31,673 in vaccine group and 31,552 in placebo group) were enrolled and
vaccinated in 11 countries in Latin America and
Rotarix is a rotavirus vaccine to be administered in two doses. The applicant has provided the following results for definite IS diagnosed within 31 days (Day 0 to Day 30) after any dose. There are 6 cases in the vaccine group and 7 cases in the placebo group within the 31 days after either dose.
[rota-023-report-body.pdf, Table 18]
The applicant concluded that since the upper limit of the 95% CI for the relative difference is < 6/10,000, the revised primary objective for safety has been demonstrated.
Reviewer’s comments and analysis
1. Study
Rota-023 was performed outside the
2. The study was designed with the assumption of a background rate of IS in placebo group of 3/100,000. However, because another study obtained an estimate of 51/100,000, together with the number of accumulating IS cases observed during the trial of Rota-023, the primary objective was revised during the conduct of the study. Since changing the primary objective while the trial is ongoing could potentially compromise the integrity of the study, and CBER did not concur with this change during the study, CBER is currently in the process of obtaining more detailed information from the applicant to ensure that proper procedure was followed.
3. The applicant presented the definite IS cases within 31 days (Day 0 to Day 30) by the diagnostic date, not the start date of the symptoms. However, there was one case in the vaccine arm that for which the symptoms started on Day 29 but was not diagnosed until Day 31, and, hence, was excluded from the reporting period. The following table, created by the reviewer, displays the IS cases by the onset day of the symptoms for all individuals during the 31-day window.
|
Days since last dose when an IS case occurred |
Ratio of the number of cases |
Risk Difference (95% CI) Per 10,000 |
Relative Risk (95% CI) Per 10,000 |
||
|
Rotarix ( N=31673) |
Placebo (N=31552) |
||||
|
Dose 1 (day 0 – day 30) |
18 |
16, 22 |
1 : 2 |
|
|
|
Dose 2 (day 0 – day 30) |
3, 3, 16, 17, 25, 29 |
6, 9, 18, 24, 28 |
6 : 5 |
|
|
|
Any Dose |
|
|
7 : 7 |
- 0.008 (-2.63, 2.61) |
0.996 (0.36, 2.72) |
From the results obtained by the reviewer, the upper limit of the 95% CI for risk difference is 2.61, which is still below the revised criterion of 6/10,000. Hence, the revised primary objective was achieved.
CBER considers the relative risk as a measure for assessing adverse events for preventive vaccine. The rationale for this preference is that the risk difference, which may be a useful metric for public health policymakers (e.g., determining how many new hospital beds are needed), would tend to minimize the risk of uncommon adverse events associated with vaccination. Since preventive vaccines will potentially be given to many millions of healthy individuals, it is important not to minimize any potential risk.
The reviewer calculated the upper limit of the 95% CI for the relative risk to be 2.72, which may be considered acceptable.
4. The following table, created by the reviewer, displays the onset day of the symptoms that lead to all IS cases after any dose at all times during the follow-up period. There is no apparent pattern for when the IS cases occurred after each dose.
|
|
Days since last dose when an IS case occurred |
Ratio of the number of cases |
|
|
|
Rotarix (N = 31673) |
Placebo (N = 31552) |
|
|
Dose 1 (day 0 - day 30) |
18 |
16, 22 |
1 : 2 |
|
Dose 1 (day 31 +) |
53 |
41, 51, 68, 74, 81, 224 |
1 : 6 |
|
Dose 2 (day 0 - day 30) |
3, 3, 16, 17, 25, 29 |
6, 9, 18, 24, 28 |
6 : 5 |
|
Dose 2 (day 31 +) |
56, 68, 86, 144, 231 |
35, 46, 50, 106, 126, 127, 222 |
5 : 7 |
|
|
|
|
13 : 20 |
5. Since the risk of IS appeared to be increased among recipients of RotaShield during the 3- to 14-day period after the first dose and during the 3- to 7-day period after the second dose, the reviewer created the following table, displaying the days of IS cases for the periods of 3-7 days and 3-14 days
|
|
Days since last dose when an IS case occurred |
Ratio of number of IS cases (Rotarix : Placebo) |
|
|
Rotarix ( N = 31673) |
Placebo (N = 31552) |
||
|
Dose 2 (Day 3 - Day 7) |
3, 3 |
6 |
2 : 1 |
|
Dose 2 (Day 3 - Day 14) |
3, 3 |
6, 9 |
2 : 2 |
There were no IS cases post dose 1. Although there were 2 cases in the vaccine arm versus 1 case in the placebo arm for the period of 3-7 days after the second dose, the period of 3-14 revealed 2 cases in each arm. Due to the small number of cases within these periods of time, one cannot rule out that they occurred on these days by chance alone.
3. Efficacy
The primary objective for efficacy in study Rota-023 was defined by the applicant as the following:
“• In the efficacy subset (N = 20 000), to determine
if two doses of GSK Biologicals’ HRV vaccine can prevent severe RV GE caused by
the circulating wild-type RV strains during the period starting from 2 weeks
after Dose 2 until one year of age.
Assuming a 1.5% incidence of severe RV GE in the
placebo group during the observation period, and a 70% vaccine efficacy, the
sample size of 20 000 subjects had at least 80% power to detect a lower limit
of the 95% CI for the vaccine efficacy above 50%.
Severe GE: An episode of diarrhea with or without vomiting that
required hospitalization and/or re-hydration therapy (equivalent to WHO plan B
or C) in a medical facility.
Severe RV GE: An episode of severe gastroenteritis occurring at
least two weeks after the full vaccination course in which rotavirus other than
vaccine strain was identified in a stool sample collected during the episode of
severe gastroenteritis.
• Occurrence of severe RV GE caused by the wild RV
strains during the period starting from 2 weeks after Dose 2 until one year of
age.”
Definition of vaccine efficacy
The applicant provided the following concerning the definition and analysis of vaccine efficacy:
“The vaccine efficacy was calculated using the
formula: 1 – RR = 1 – (ARV/ARU), where RR = relative risk = ARV/ARU
ARU = disease attack rate in unvaccinated population
(estimated from the Placebo group) = number of subjects reporting at least one
severe RV GE episode / total number of subjects in the placebo (control) group.
ARV = disease attack rate in vaccinated group = nv/Nv
= number of subjects reporting at least one severe RV GE episode / total number
of subjects in the HRV vaccine group.
Two-sided Fisher’s exact test (significance level of a
= 0.05) was used to compare these percentages between HRV and Placebo groups.”
Results from the Applicant
Vaccine efficacy analysis was performed on the According To Protocol (ATP) cohort which included all subjects from the ATP safety cohort and who received 2 doses of either the investigational vaccine or the placebo, had follow-up beyond 2 weeks after Dose 2 through the end of the first efficacy follow-up period, and had no vaccine strain in stool samples collected between the day of Dose 1 administration and 2 weeks after Dose 2 was administered.
There were 17,867 subjects (9,009 in the investigational vaccine group and 8,858 in the Placebo group) included in the ATP efficacy cohort.
The following table showing efficacy results was submitted by the applicant.
[rota-023-year-1-report-body.pdf, Table 12]
Since the lower bound of the 95% CI for vaccine efficacy is above 50%, the applicant concluded that the primary efficacy objective was reached.
The applicant also used the Cox proportional-hazard model to estimate vaccine efficacy against severe RV GE caused by the circulating wild-type as 84.8% (95% CI: 72.0%; 91.7%). The applicant submitted the following table as an amendment at CBER’s request.
Table 1 Percentage of subjects reporting severe RV GE episodes and efficacy of the vaccine from 2 weeks after Dose 2 up to Visit 4, by COX - ATP cohort for efficacy
|
|
|
|
|
n/T |
Vaccine Efficacy |
|
||||
|
|
|
|
T |
|
95%CI |
|
95%CI |
|
||
|
Group |
N |
n |
(year) |
value |
LL |
UL |
% |
LL |
UL |
P-value |
|
Severe RV GE of any wild gtype |
||||||||||
|
HRV |
9009 |
12 |
5914.1 |
0.002 |
0.001 |
0.004 |
84.8 |
72.0 |
91.7 |
<0.001 |
|
Placebo |
8858 |
77 |
5777.1 |
0.013 |
0.011 |
0.017 |
|
|
|
|
|
G1 wild type |
||||||||||
|
HRV |
9009 |
3 |
5916.4 |
0.001 |
0.000 |
0.002 |
91.8 |
73.5 |
97.5 |
<0.001 |
|
Placebo |
8858 |
36 |
5788.6 |
0.006 |
0.004 |
0.009 |
|
|
|
|
|
Pooled Non G1 (G2, G3, G4, G9) |
||||||||||
|
HRV |
9009 |
10 |
5914.7 |
0.002 |
0.001 |
0.003 |
75.5 |
51.0 |
87.6 |
<0.001 |
|
Placebo |
8858 |
40 |
5792.3 |
0.007 |
0.005 |
0.009 |
|
|
|
|
Notes:
N = number of subjects included in each group
n = number of subjects reporting at least one specified severe RV GE
episode in each group
T= sum of follow-up period expressed in year censored at the first occurrence
of the specified severe RV GE episode, in each group
n/T= person-year rate of the specified severe RV GE in each group
95% CI,LL,UL = Lower and upper limits of the 95% confidence interval
P-value from Cox regression model to test H0 = {VE=0%} (Y = Time to Event)
Reviewer’s comment
1. For the definition of the vaccine efficacy in this study, CBER considers it is more appropriate to use the “time-to-first-episode” analysis than using the number of subjects who had at least one episode among the subjects enrolled in each arm. The rationale for this preference is that the time-to-event approach accounts for differential follow-up of subjects, while the latter approach does not. Therefore, CBER is inclined to place more importance on the Cox proportional-hazards model results. The result of this analysis was submitted to CBER as an amendment. The reviewer has verified the efficacy estimate and the 95% confidence intervals for any wild-type, as described in the primary objective.
4. Reviewer’s Overall Conclusion
- Study Rota-023 was not conducted under US/FDA IND regulation. Therefore, none of the protocols and amendments were concurred upon by CBER before or during the trial.
- By the reviewer’s relative risk calculation based on the onset day of the IS symptoms, the applicant has reached the revised primary safety objective. However, CBER is still evaluating the process by which the applicant revised the primary safety objective.
- The efficacy results based on the Cox proportional-hazard model were confirmed by the reviewer, who considers them to be a more appropriate measure for the efficacy estimate due to large variations in the follow-up times of the subjects.