Title: TRAINING WORKSHOP ON PHARMACEUTICAL QUALITY, GOOD MANUFACTURING PRACTICE
1TRAINING WORKSHOP ON PHARMACEUTICAL QUALITY, GOOD
MANUFACTURING PRACTICE BIOEQUIVALENCE
- Statistical Considerations for Bioequivalence
Studies - Presented by
- John Gordon, Ph.D.
- Consultant to WHO
- e-mail john_gordon_at_hc-sc.gc.ca
2Introduction
- Performance will never be identical
- Two formulations
- Two batches of the same formulation?
- Two tablets within a batch?
- Purpose of bioequivalence (BE)
- Demonstrate that performance is not
significantly different - Same therapeutic effect
- What constitutes a significant difference?
3Introduction cont.
- Agencies must define a standard consisting of the
following - Bioavailability metrics
- One or more acceptance criteria for each metric
- Number and type of metrics may vary
- Dependent on drug formulation
4Metrics for BE studies
- Concentration vs. time profiles
- Area under the curve (AUC)
- Maximal concentration (Cmax)
- Time to Cmax (Tmax)
- Statistical measures of BE metrics
- Mean
- Variance
5Logarithmic Transformations
- Distribution of BE metrics
- Skewed to the right
- Consistent with lognormal distribution
- Proportionate effects
6Example
- What would be the expected drop in AUC if a
patient received 20 less drug? - Subject 1
- Original AUC 100 units
- 20 drop 20 units
- Subject 2
- Original AUC 1000 units
- 20 drop 200 units
7Example cont.
- Log transformation
- Absolute intrasubject differences become
independent of patients AUC - Log(80) log(100) log(800) log(1000)
- Log transformation for concentration dependent
measures - Accepted by regulatory agencies
8Analysis of Variance
- ANOVA
- Most common technique of analysis and estimation
- Lognormal distribution
- Raw data must be log transformed
- Comparison of means and variances of transformed
data - Geometric mean
- Results reported in original scale
9ANOVAHypothesis Testing
- Null hypothesis test
- No formulation difference
- Convey little detail
- Statistically significant difference
- Clinically significant?
- Imprecise estimates (high variability)
- No statistically significant difference
10Confidence Intervals (CI)
- Inference from study to wider world
- Range of values within which we can have a chosen
confidence that the population value will be
found - Study findings expressed in scale of original
data measurement
11Confidence Intervals cont.
- Width of CI indication of (im)precision of sample
estimates - Width partially dependent on
- Sample size
- Variability of characteristic being measured
- Between subjects
- Within subjects
- Measurement error
- Other error
12Confidence Intervals cont.
- Degree of confidence required
- More confidence wider interval
- In other words, width of CI dependent on
- Standard error (SE)
- Standard deviation, sample size
- Degree of confidence required
13Confidence Intervals cont.
- Statistical analysis of pharmacokinetic measures
- Confidence intervals
- Two one-sided tests
14Typical BEAssessment Criteria
- 90 confidence interval
- Ratio of geometric means
- Acceptance criteria 80 125
- Log transformed AUCT Cmax
15Statistical Approaches for BE
- Average bioequivalence
- Population bioequivalence
- Individual bioequivalence
16Statistical approaches cont.
- Average BE
- Conventional method
- Compares only population averages
- Does not compare products variances
- Does not assess subject x formulation interaction
17Statistical approaches cont.
- Population and individual BE
- Include comparisons of means and variances
- Population BE
- Assesses total variability of the measure in the
population - Individual BE
- Assesses within subject variability
- Assesses subject x formulation interaction
18Design Considerations
- Non-replicated designs
- Most common
- Crossover designs
- Two-formulation, two-period, two-sequence,
crossover design - Average or population BE approaches
- Parallel designs
19Design Considerations
- Replicated designs
- Can be used for all approaches
- Critical for individual BE approach
- Suggested replicated design
- Two-formulation, four-period, two-sequence
- T R T R
- R T R T
20Statistical effects in model
- Sequence effect
- Subject (SEQ) effect
- Formulation effect
- Period effect
- Carryover effect
- Residual
21Outliers
- Statistical outliers
- Valid clinical/physiological justification
- Re-testing?
22Add-on designs
- All studies should be powered appropriately
- If study fails the standard
- Reformulate
- Undertake larger study
- Add-on study
- Consistency testing
- Group-sequential designs
- Penalty for peeking at results