Statistical Review - Cervarix
|Date||August 2, 2009|
|From||Lev A. Sirota
Vaccines Evaluation Branch
Acting Team Leader, VBT
A. Dale Horne
|Subject||BLA 125259 (GSK.)
Cervarix. Human Papillomavirus Vaccine, AS04 Adjuvant-Adsorbed c.c. Chronological File
Executive Summary Statement and Conclusion
The applicant addressed all major concerns related to assay validation statistics. I do not see any major bioassay-related statistical issues preventing STN 125259 from being approved by the agency.
This review covers statistical methodology in assay validation for Human Papillomavirus Vaccine, AS04 Adjuvant-Adsorbed. It is concentrated on answers presented by the applicant to CBER’s questions. The BLA amendments containing related answers are listed in the above subject heading.
My comments are in italics.
Results and Comments to CBER
Serological assays validation
Original data on Serological assay validation are presented in module 3 of the STN 125259. The applicant addresses CBER questions related to assays validation in amendments listed in the subject heading to this review.
The following CBER’s Questions were addressed in the BLA 125259/0.21 Amendment.
Please clarify the definition of ----(b)(4)---- of the dose-response curve used in your assay validation (Section 6.1.2, Page 18). Additionally, please provide a detailed explanation and your justification of the acceptance criterion for ----(b)(4)---- of the dose-response curves.
The applicant clarified that they consider dose-response curves to be ------------(b)(4)------------- --------------------------------------------------------. The applicant also noted that in addition to this pre-established criterion, other acceptance criteria for ----(b)(4)--- were investigated. The applicant provided illustration of ----(b)(4)--- in terms of both doses-response curves for ELISA HPV-18 (comparison of the standard and -----(b)(4)-----), and tabulated parameters demonstrating that the sigmoids are visually undistinguishable and very close parametrically.
You state that inter-dilution CV for all L1 VLP antigen content values range between --(b)(4)-- (Section 6.2, Page 18). We note that in Table 10 (Page 18), the CV for theoretical starting concentration of ---(b)(4)--- is stated to be -(b)(4)- (column 3; 10/03/04). This value is outside the --(b)(4)-- interval. Please comment.
The applicant commented that the inter-dilution CV L1 VLP antigen content referred to in Table 10 is indeed ranging from ---(b)(4)--- . The original value of -(b)(4)- was a typographic error.
Please justify using CV -(b)(4)- for determining whether the starting concentration is suitable in your analysis described in Section 6.2 (Page 18)
The applicant commented that a CV of -(b)(4)- for determining an antigen concentration by ELISA is commonly used in ELISAs developed by GSK Biologicals. The results are commonly considered valid, as long as the antigen content is determined from at least 3 serial dilution points with an inter-dilution CV -(b)(4)- . In the context of question 3, the criterion of an inter-dilution CV of -(b)(4)- was also applied to determine whether an HPV- 18 L1 VLP antigen concentration evaluated with the HPV- 18 L1 VLP ELISA was accurate. This criterion was applied for all types of test materials for which the HPV-18 L1 VLP antigen content needed to be determined. For instance, the inter-dilution CV of -(b)(4)- was applied during the evaluation of the range of suitable so-called starting concentrations of HPV-18 L1 VLP antigen to be used as the first dilution of test material in the HPV-18 L1 VLP ELISA. The applicant also explained that the evidence for using an inter-dilution CV -(b)(4)- for determining HPV-18 L1 VLP antigen concentration when using the HPV-18 L1 VLP ELISA comes from evaluating the impact of the inter-dilution CV on the antigen content determined for the Internal Control of the
HPV-18 L1 VLP ELISA. Based on analysis of -(b)(4)- tests, they concluded that only when assay CV exceeds -(b)(4)- does the proportion of rejected tests increase dramatically.
Regarding evaluation of the range, we have the following comments:
(a) We note that the ratio of calculated starting concentration/theoretical starting concentration ------------(b)(4)------------- from -----(b)(4)------ as the theoretical starting concentration decreases from --------------(b)(4)--------------- (data in last column of Table 10 on Page 18). These data show systematic deviation from a proportional relationship between the calculated and theoretical values in this range. However, you conclude that the data presented in Table 10 "demonstrate that there is a proportional relationship between the calculated and theoretical values for starting concentrations ranging from ---(b)(4)-- ----------------- .” Please comment.
(b) If you consider the deviation from proportionality to be acceptable, please provide your justification.
The applicant agreed that there is a trend for ---(b)(4)--- ratios (recoveries) of the calculated versus theoretical starting concentrations. They further comment that the range of theoretical antigen concentrations ( -----(b)(4)----- ) evaluated in this analysis brackets by about -(b)(4)- fold the amount of antigen engaged in the HPV-18 L1 VLP ELISA assay. According to the HPV-18 L1 VLP ELISA Standard operating procedure the standard material, the internal control and the test samples are -------------------------------------(b)(4)-------------------------------------- . It follows from Table 10 in m3.2.S.4.3 (HPV-18) of the initial BLA that the use of a ---(b)(4)--- starting concentration lies in between the ----------(b)(4)---------- theoretical starting concentrations tested during the evaluation of the range of a suitable starting concentration for HPV-18 L1 VLP antigen. The recoveries observed for the theoretical starting concentrations of ------(b)(4)------ -------------------------------- , respectively. The -(b)(4)- difference observed between -----(b)(4)----- is in line with the intermediate precision determined during analytical method validation (Section 6.2 in m3.2.S.4.3 (HPV18) of the initial BLA). Based on the fact that the standard material, internal control and test samples are all diluted to ----(b)(4)--- as their starting concentration for the HPV-18 L1 VLP ELISA and that the recovery rates for this starting concentration are in line with the intermediate precision of the assay, the Sponsor confirms that the ELISA assay allows for accurate quantification of HPV-18 L1 VLP antigen.
For analysis of reproducibility (Section 6.3.3, Page 20), please calculate the 95% confidence interval for the inter-laboratory difference.
The applicant presented the calculated 95% confidence intervals for the antigenic content determined for two different batches of ------(b)(4)---- as requested. The presented data demonstrates that 95% CI observed for different labs are comparable.
Please clarify whether your assay validation included assessment of operator-to-operator and day-to-day variability. If these parameters related to your assessment of precision were evaluated, please provide the data. If these analyses were not performed, please explain why you do not consider assessment of these parameters to be relevant for this assay.
The applicant assesses Inter-assay variation or Intermediate precision of the HPV-18 L1 VLP ELISA by evaluating the variability of the HPV-18 L1 VLP antigenic content or antigenic activity measured on -(b)(4)- different samples from -(b)(4)- different HPV-18 L1 VLP -(b)(4)- tested in the same laboratory on different days or by different operators. The applicant presented data demonstrating a coefficient of variation of -(b)(4)- for both the antigen content or antigenic activities for both purified bulks. The sponsor performed a n additional analysis to evaluate the individual contribution of the operator-to-operator and day-to-day variability to the total variability of the HPV-18 L1 VLP ELISA assay as requested. The day-to-day variability was shown to be around -(b)(4)- for antigenic activity and antigen content. Operator-to-operator variability was around -(b)(4)- as for antigenic activity so for antigen content.
We have the following additional comment regarding your assessment of antigenic activity of the HPV-16 L1 VLP by ELISA (m3.2.S.4.3, Section 6):
For your analysis of reproducibility (Section 6.3.3, Page 18), please calculate the 95% confidence interval for the inter-laboratory differences.
As requested, the applicant calculated and presented the 95% confidence Intervals (CI) for the difference between laboratories in antigenic content determined for -(b)(4)- different batches of ------(b)(4)------ . The data shows that the 95% CI observed for the independent determinations of HPV-16 L1 VLP antigen content performed at the Analytical-R&D laboratory (ARD) and compared to the antigen content determinations performed at the QC-New Projects laboratory (QC-NP) are comparable for each purified bulk batch evaluated.
The following CBER’s Question was addressed in BLA 125259/0.32 Amendment.
CBER notes that you propose an infinite upper limit for the assay range for the measurement of antibody in human serum samples. For example, in Module 220.127.116.11
(p.7 of 52), VALDOC: HPV16EIA2PCV01, you state that there is no upper limit to the analytical range of the assay. According to the ICH definition of the range, suitable level of precision, accuracy, and linearity should be demonstrated within the range. Please use this criterion to establish appropriate upper limits for the ranges of the assays.
The applicant acknowledges that precision, accuracy, and linearity should be demonstrated within the range of titers including high titers. The applicant performed additional experiments to establish appropriate upper limits for the analytical range of the HPV-16, -18, ---(b)(4)--- ELISA as requested. The upper limit for each HPV ELISA is provided below:
HPV-16 ELISA: ------(b)(4)------
HPV-18 ELISA: ------(b)(4)------
The following CBER’s Questions were addressed in BLA 125259/0.31 Amendment.
We note a high level of operator variability, -(b)(4)- , for all serological assay validations submitted (VALDOC: HPV31EIA: Tables 14 and 15, p. 27 of 53). Operator variability is the largest component and exceeds -(b)(4)- . Please comment
The applicant acknowledges the fact that the operator variability is an important component of the assay variability. In routine conditions, the operator variability for each HPV ELISA is controlled. They explained that each new operator is certified for the test prior to assaying study samples using the following acceptance criteria:
- At least -(b)(4)- serum samples representative of the titer range likely to be encountered in clinical trials must be tested.
- The geometric mean ratio of serum titers compared to a reference operator must be within --(b)(4)-- .
- There must be no more than -(b)(4)- of samples that show a titer two-fold higher or two-fold lower compared to the reference titer
Each tested ELISA plate needs to be within the Quality Control specification limits.
We note that there are several samples that demonstrated unusually high variability (VALDOC: HPV18EIA2PCV01; p. 18 of 52). For example, CV% for day+operator for sample 4 is above -(b)(4)- and CV% for day+operator+reagents is exceeding -(b)(4)- for 6 samples out of -(b)(4)- reaching almost -(b)(4)- for sample 2. Please comment
The applicant acknowledges that 6 out of -(b)(4)- sera used to evaluate the variability of the
HPV-18 ELISA assay have shown an unusual high variability. However, they explain that four out of the 6 samples (samples 1, 2, 3 and 4) for which the CV% day +operator + reagent is exceeding -(b)(4)- had titers lower than -(b)(4)- . In routine testing conditions, ELISA titers within the grey zone, ranging from ----------(b)(4)--------- , must be confirmed by another test giving a titer that must be included within a two fold ratio of the initial titer. This increases the level of confidence for results in the titer range close to the assay cut-off ( -(b)(4)- ). Two out of the -(b)(4)- samples (samples 7 and 12) for which the CV% day + operator +reagent was -(b)(4)- and -(b)(4)- respectively, had titers higher than -(b)(4)- . The applicant considers that a high level of variability on 2 samples out of 12 is acceptable as the total CV measure is acceptable ( -(b)(4)- percent).
Validation reports pertaining to MPL
Process validation and/or evaluation section summarizes the process validation and/or evaluation studies that were conducted to characterize the MPL manufacturing process. All process control variables are categorized as critical or non-critical parameters according to the sponsor’s global process validation policy. The applicant used standard statistical process control methods in this section and presented statistical process control charts to support conclusions. I have no statistical comments regarding this part.
Other minor comments
On page 115, Figure 60 in 3.2.A.3 of the original submission, the applicant presented data of ---------(b)(4)--------- of S.Minnesota R595 by ------(b)(4)------ extrapolated far outside the area of measured data.
From the statistical point of view, this is not a recommended practice. Nevertheless, it may be acceptable if the ----(b)(4)--- process of S. Minnesota R595 by -(b)(4)- is well understood.