SPADA at AOAC Annual Meeting 2023
B IOLOGICAL T HREAT A GENT M ETHOD V ALIDATION G UIDELINES
AOAC O FFICIAL M ETHODS OF A NALYSIS (2012)
Appendix I, p. 20
Table D1. Example data summary table: Detection of Bacillus anthracis spores on nitrocellulose aerosol collection fi lters by PCR a Organism Concentration N x POD C 95% LCL 95% UCL B. anthracis Ames 20,000 spores/ fi lter 96 95 0.9896 0.9433 1.0000 B. cereus E33L 200,000 spores/ fi lter 96 1 0.0104 0.000 0.0567 a N = number of replicates tested; x = number of positive outcomes; and POD C = POD for the con fi rmed candidate method.
ANNEX D Summary Data Table for Qualitative Method: Method Developer and Independent Validation Studies Prepare one table per matrix and indicate laboratory or intended use conditions ( see Table D1). ANNEX E Calculation of CPOD and dCPOD Values from Qualitative Method Collaborative Study Data Calculations are done in four steps. Step 1 .—The overall fractional response (mean POD = CPOD) for the method is calculated from the pooled POD i responses of the individual collaborators ( i = 1, 2, ..., C). Then the standard deviation of an individual x value, called the repeatability standard deviation s r , is estimated. Next a 95% con fi dence interval on s r is obtained. Step 2 .—The standard deviation s C of the collaborator variance component due to differences in detection among collaborators is estimated. s C is called the “collaborator effect” standard deviation. A 95% con fi dence interval for s C is obtained. Step 3 .—Both s C and s r are used to estimate a 95% con fi dence interval for the expected value of CPOD. Also estimated is the reproducibility standard deviation s R , which is the standard deviation of measurement of a single x from a single collaborator, including both the collaborator–collaborator and repeatability error sources (15). A 95% con fi dence interval is estimated for the expected value of s R . Step 4 .—A t statistic based on the χ 2 distribution or Fisher randomization test of homogeneity of POD across laboratories is performed to determine if the observed collaborator effect is detectably greater than zero. CPOD Report the CPOD estimates with 95% con fi dence intervals for the candidate method and, if included, the presumptive and con fi rmed results. CPOD estimates are determined as for the single-laboratory case, but are based on the composite data across collaborators.
Repeatability Standard Deviation (s r ) Estimate the repeatability standard deviation:
2
C
2 1 1 n s 1 i i i C i n
i x x i
n
i
2
s
r
N C
i
1
2
s
s
r
r
where s 2 i is the variance of the results from collaborator i , x i is the number of positive detections from collaborator i , n i is the number of observations from collaborator i , N is the total number of data, and C is the number of collaborators. Collaborator Effect Standard Deviation (s C ) Estimate the collaborator effect standard deviation:
2
s
2 s max 0, s C
2
r
POD
n
where
2
n
1
i
N
n
N C
1
is the weighted average number of replicates per collaborator and POD CPOD s 1 i C is the observed variance in POD values across collaborators. The degrees of freedom, df, for s POD are calculated as follows: 2 2 POD
2 2
s r
2
s L
( )
L N
df
2 2
s r
2 2
s L
( )
N
i x x n N
L
1, 2,3,..., i
C
CPOD
1 Reproducibility Standard Deviation (s R ) Calculate the reproducibility standard deviation: 2 2 2 R s s s r C If the collaborator effect s C is near zero, then r R s s CPOD 1 CPOD L N L
i
where N = Σ n i is the total number of data, x = Σ x i is the total number of positive detections, and C is the number of collaborators. The CPOD is the proportion detected as positive in the entire set of data across collaborators for the concentration level for the particular method. CPOD is distinguished from POD in that it includes between-collaborator variation as well as within- collaborator variation.
© 2012 AOAC INTERNATIONAL
Made with FlippingBook - Online Brochure Maker