Statistical Issues Related to

1
Statistical Issues Related to Pneumococcal
Vaccine Evaluation Brian D. Plikaytis Centers
for Disease Control and Prevention
2
Statistical Issues Related to Pneumococcal
Vaccine Evaluation

• Discuss non-inferiority in the context
• of T-cell independent vaccines

• Extend these points to T-cell dependent vaccines

• Use of multiple correlate endpoints
• to predict protection

• Briefly summarize WHO meeting, June, 2003

3
T-cell Independent Vaccine
Licensed vaccine
Antibody Level
New vaccine
threshold or reference
Time
4
Endpoints for Vaccine Comparison

• Antibody response (GMC)

• Opsonophagocytic Titer (GMT)

• Fold-increase in titer ( fold-responders)

• Proportion above threshold or reference level

• Additional assays / endpoints?

5
Hypothesis Testing
6
T-cell Independent Vaccine
Antibody Concentrations or Titers
Licensed vaccine
Antibody Level
New vaccine
threshold or reference
Lic. Vac New Vac
Time
7
T-cell Independent Vaccine
Antibody Concentrations or Titers
Licensed vaccine
Antibody Level
New vaccine
threshold or reference
Lic. Vac New Vac
Time
8
T-cell Independent Vaccine
Antibody Concentrations or Titers
Licensed vaccine
Antibody Level
New vaccine
threshold or reference
Lic. Vac New Vac
Time
9
T-cell Independent Vaccine
Antibody Concentrations or Titers
Licensed vaccine
Antibody Level
New vaccine
threshold or reference
Lic. Vac New Vac
Equivalence boundary for testing non-inferiority
(?)
?
Time
10
T-cell Independent Vaccine
Antibody Concentrations or Titers
Licensed vaccine
Antibody Level
New vaccine
threshold or reference
Lic. Vac New Vac
Equivalence boundary for testing non-inferiority
(?)
?
Time
11
T-cell Independent Vaccine
Proportion Above Threshold
Licensed vaccine
Antibody Level
New vaccine
threshold or reference
Licensed vaccine
Proportion Threshold
New vaccine
threshold or reference
Time
12
T-cell Independent Vaccine
Proportion Above Threshold
Licensed vaccine
Proportion Threshold
New vaccine
threshold or reference
Diff in Prop. Threshold
Time
13
T-cell Independent Vaccine
Proportion Above Threshold
Licensed vaccine
Proportion Threshold
New vaccine
threshold or reference
Equivalence boundary for testing non-inferiority
(?)
Diff in Prop. Threshold
? for diff. in prop. ? thresh. 0.10
?
Time
14
Is Antibody Response an Accurate Correlate for
T-cell Dependent Vaccines?
Licensed vaccine
Antibody Level
New vaccine
threshold or reference
Natural exposure or vaccine boost
Priming Dose Series
15
Predicting Protection
Given that we can measure more than one
correlate of protection .
How do we utilize them to predict protection
more effectively than each test individually?
16
Predicting Protection
Sn95 Sp93
Unprotected
Protected
Frequency
Antibody Concentration
17
Predicting Protection
Sn90 Sp90
Unprotected
Protected
Frequency
8
16
32
64
128
256
512
1024
4
Titer
18
Predicting Protection
Antibody Concentration
r 0.64 r2 0.41
8
16
32
64
128
256
512
1024
4
Titer
19
Predictive Model Building

• Multiple approaches
• logistic discriminant analysis, cluster analysis,
  other exploratory techniques (CART, etc.)
• Logistic Discriminant Analysis
• Model relates a series of assay endpoints to
  protection
• Construct a discriminant function from results

IS b1x1 b2x2 bpxp
IS immunologic score xi correlate data bi
model coefficients p number of variables in
the model
20
Immune Response vs. Protection Model Building
1
(Protected)
Protected
Unprotected
0
(Unprotected)
Antibody Concentration
21
Predicting Protection
Antibody Concentration
r 0.64 r2 0.41
8
16
32
64
128
256
512
1024
4
Titer
22
Predicting Protection
Sn97 Sp97
Antibody Concentration
8
16
32
64
128
256
512
1024
4
Titer
23
Predicting Protection
Sn95 Sp93
Unprotected
Protected
Frequency
Antibody Concentration
24
Predicting Protection
Sn97 Sp97
Unprotected
Protected
Frequency
Immunologic Score
25
Limitations

• Protected/unprotected status must be determined
• using criteria independent of the analysis
• variables

• If more than two correlates exist, one of them
  may
• serve to define protected/unprotected status

• For this to work, multiple lab tests will need
  to be
• run on all specimens (assay multiplexing?)

• For highly efficacious vaccines, pool of
  unprotecteds
• will be small

• In general, cannot perform these analyses on
  data
• acquired from vaccine evaluation studies.

26
RCD Curves for Vax and Control Groups at 4 Weeks
Post Dose 3
100
80
60
Vaccine
gt Ab Conc
40
Control
20
0
0.01
0.1
1
10
100
Ab Concentration
Jódar, L., et al., Vaccine, 213265-3272, 2003.
27
Schematic RFD Curves for Vax and Control Groups
at 4 Weeks Post Dose 3
Vaccine
Control
Relative Frequency
Ab Concentration
28
Conclusions / Questions

• Study endpoints need to be carefully selected
• to fully optimize group comparisons

• T-cell dependent vaccines complicate using
• circulating antibody levels as a sole
  correlate
• for protection (develop new correlates?)

• Two or more correlates may be utilized to better
• predict vaccine protection (aid in
  licensure?)

29
(No Transcript)
30
Serological Criteria for the Evaluation and
Licensure of New Pneumo. Conj. Vaccine
Formulations for Use in Infants

• IgG Ab measured by ELISA from sera collected
• 4 weeks after 3 dose primary series

• Single threshold or reference level of 0.35 µg/ml

• Threshold not necessarily predictive of
  protection
• in an individual

• Threshold defined using ELISA without 22F
• absorption

• ELISA should be calibrated against reference
• assay (standardization)

31
Serological Criteria for the Evaluation and
Licensure of New Pneumo. Conj. Vaccine
Formulations for Use in Infants

• Alternative assays should be bridged to derive
• equivalent threshold concentration

• Proportion of responders above threshold
• should be used to determine noninferiority

• Head-to head comparison with registered
• vaccine preferred

• Noninferiority for each serotype in the
  registered
• vaccine desirable but not required.

• New serotypes compared to aggregate response
• of serotypes in registered vaccine

32
Serological Criteria for the Evaluation and
Licensure of New Pneumo. Conj. Vaccine
Formulations for Use in Infants

• Demonstrate functional activity of antibody
  using
• opsonophagocytic assay

• Opsono. assay should be comparable to
• reference assay (standardization)

• Evidence of memory should be demonstrated
• using a booster dose of pneumo. PS vaccine

• Avidity of antibodies also a useful marker for
• immunological memory

33
Further Information
Plikaytis, B.D., Carlone, G.M., Statistical
Considerations for Vaccine Immunogenicity Trials,
in Carbohydrate Based Vaccines, Chris Jones and
Neil Ravenscroft, editors, Blackwell Science,
Ltd., Oxford, submitted.
Sponsors
Centers for Disease Control and Prevention
World Health Organization