Instruction Manual

19
Each pool was tested and the percent of negative results for these pools was calculated. Of the 103 4-sample pools
in the 1-3% prevalence group, 99.0% (102/103, 95%CI 94.9-99.8%) were negative, and 1/103 was inconclusive. Of
the 107 4-sample pools in the 3-6% prevalence group, 99.1% (106/107, 95%CI 94.9-99.8%) were negative, and 1/107
was invalid. Of the 37 4-sample pools in the 6-10% prevalence group, 100% (37/37, 95%CI 90.6-100%) were negative.
Overall in the study, 99.2% (245/247, 95% CI 97.1-99.8%) of 4-sample pools with 4 negative samples were negative.
Two pools would have had to be subsequently deconvoluted, with each sample being tested individually.
Efficiency with Pooled Negative Specimens (n = 247)
Positivity
Rate
Group
n
Results of 4-sample Pools
% Negative Percent
Agreement *
Negative
Inconclusive
Positive
Invalid
1-3%
103
102
1
0
0
99.0% (102/103)
95%CI 94.9-99.8%
3-6%
107
106
0
0
1
99.1% (106/107)
95%CI 94.9-99.8%
6-10%
37
37
0
0
0
100% (37/37)
95%CI 90.6-100%
Total
247
245
1
0
1
99.2% (245/247)
95% CI 97.1-99.8%
* Since any pool that is not negative (i.e., positive, inconclusive and invalid) is re-tested as an
individual sample, the parameter NPA affects the efficiency of 4-sample pooling
18) Pooling Validation – In Silico Sensitivity in Population with Positivity Rate 1%-10%
Quest Diagnostics conducted an in silico analysis to evaluate the effect of 4-sample pooling on the clinical sensitivity
of the SARS-CoV-2 assay. This analysis was conducted using Passing-Bablok regression analyses from the “
Pooling
Validation /
Sensitivity for Pools with One Positive Sample and Three Negative Samples (n = 101)” data to calculate
the Ct shift resulting from the dilution effect of 4-sample pools (1 positive sample combined with 3 negative samples).
In the regression analysis, the X-axis displayed individual Ct values for positive samples and the Y-axis displayed Ct
values for the corresponding pools with one positive sample and 3 negative samples. This analysis was conducted in
three populations with different positivity rates: 1-3% (n=820), 3-6% (n=1,113) and 6-10% (n=1,158). The de-identified
data were selected from sequentially tested positives based on the Quest Diagnostics SARS-2-CoV RT-PCR
molecular assay. The regression analysis was used to calculate an interval of Ct values [X*, 40] where individual
samples with Ct values within this interval would have negative results in 4-sample pools (1 positive and 3 negative)
due to dilution effects. For each population, the percent of individual samples with Ct values ranging from [X*, 40] was
calculated. The X* values for the N1 target in the three populations were 37.0 (1-3%), 38.3 (3-6%) and 37.65 (6-10%).
The X* values for the N3 target in the three populations were 37.45 (1-3%), 38.7 (3-6%) and 38.1 (6-10%).
Of the 820 samples in the 1-3% prevalence group, 100% (820/820, 95% CI 99.5-100%) of the samples would not have
negative results in 4-sample pools: 97.3% were positive (798/820, 95% CI 96.0-98.3%), 2.7% were inconclusive
(22/820, 95% CI 1.7-4.0%), and none were negative.
Of the 1,113 samples in the 3-6% prevalence group, 100% (1,113/1,113, 95% CI 99.7-100%) of the samples would
not have negative results in 4-sample pools: 99.4% were positive (1,106/1,113, 99.4-99.8%), 0.6% were inconclusive
(7/1,113, 95% CI 0.3-1.3%), and none were negative.
Of the 1,158 samples in the 6-10% prevalence group, 100% (1,158/1,158, 95% CI 99.7-100%) of the samples would
not have negative results in 4-sample pools: 98.5% were positive (1,141/1,158, 95% CI 97.7-99.1%), 1.5% were
inconclusive (17/1,158, 95% CI 0.9-2.3%), and none were negative.
Overall in the study, none of the 3,091 samples, if pooled, would have been incorrectly determined to be negative
(0/3,091, 95%CI 0.0-0.12%).