Early identification and isolation of SARS-CoV-2 infected individuals remains a key strategy to interrupt community transmission. In this pilot study, we successfully implemented hypercube pooling for frequent testing of a professional sports team. We show that this procedure reliably detects a single positive sample in a group of 81, provided the starting sample has a Ct ≤ 32. Whilst there is a need for further evaluation, particularly with samples with a wider range of Ct values and pools with multiple positive samples (for different viral prevalence scenarios), these results suggest that the pooling strategy is indeed a promising approach for cost-efficient RT-PCR testing.
Other studies have demonstrated successful implementation of pooled testing using groups of up to 5 and 8 samples each.6 To our knowledge, this is the first study that has implemented the SARS-CoV-2 hypercube-based pooled testing strategy, using group sizes as large as 81.
We selected for analysis samples with Ct values that are typical for the population under study. For comparison, we also selected samples with Ct values that are one standard deviation away from the mean Ct value. We used historical data from the population of samples tested routinely in KZN to determine the distribution of Ct values present at various percentile ranges (Supplementary Information Fig. 1). The asymmetrical (left-skewed) distribution of the data may be attributed to the population of samples that was used to generate the curve i.e. symptomatic people with active disease (and therefore lower Ct scores) presenting at health facilities.
The retention of the sensitivity of the pooled test procedure as the hypercube dimension size (and consequently the dilution) increases is an obvious concern.
Our results demonstrated no loss of assay sensitivity for samples with an initial (undiluted) Ct value ≤ 32 (Fig. 2). Samples with higher Ct values would typically have a lower viral load, which may be deemed as clinically and / or epidemiologically insignificant, since the infection is not likely to be contagious (a cut-off Ct > 30 can be associated with non-infectious samples).10
Propagating SARS-CoV-2 from clinical samples can also be used as a valuable proxy for infectiousness. There are, however, conflicting reports on the cut-off Ct at which the virus is not cultivable which can range anywhere from 30 to > 35.11,12
In any event, high Ct values are expected to be associated with low infectivity;13 the maximum viral load occurs during the onset of symptoms and presents the highest risk of transmission.14 In respiratory samples, the viral load is highest during the initial stage of infection (patients in the early stages of infection usually have Ct values of 20–30 or less12), and reaches a peak in the second week followed by lower viral loads.15 The pooling strategy we have implemented is therefore most beneficial when applied to low-prevalence, asymptomatic or pre-symptomatic people who are on the viral load incline. In addition, the effects of the loss of assay sensitivity can be mitigated by implementing more frequent testing – a solution which could be more easily afforded given the cost savings of the pooled testing strategy.
There have also been reports of inherent inaccuracy due to higher false negative rates associated with the pooled testing method.16 Our method, however, includes many consistency checks. For example, detection of a single positive slice pool with the other slice pools negative is indicative of a testing mistake.8 This is especially beneficial when compared to methods such as binary testing methods17 which rely on repeated testing of the positive sample in subsequent rounds of testing, which can incorrectly terminate the identification of the positive where there are false negative results. Another cause of false negative results is sequence variation at primer or probe binding sites on the viral RNA.18 In South Africa, a new SARS-CoV-2 lineage (501Y.V2) characterised by various mutations in the spike (S) gene has been reported.3 However, there was no apparent loss of sensitivity for the S-gene in the RT-PCR assay used in this study.
We have developed and successfully implemented a pilot SARS-CoV-2 pooled testing strategy for a prominent South African rugby team. The successful implementation of pooled testing requires that 3 important criteria are met viz. (i) efficiency, (ii) sensitivity and (ii) operational feasibility6. Our study was successful in satisfying each of these criteria by (i) achieving a 40-fold reduction in test usage (and therefore cost) compared to individual testing. The cost reduction is most impressive when all individuals are negative, because 27 or 81 negative results can be achieved using just one qPCR reaction, in a 3-d or 4-d hypercube, respectively. This was the case with the screening of the South African Rugby team at the KRISP laboratory, where, during the first 8 weeks, all of the samples were tested using pooled tests of 27 or 81 samples, and only in the 9th week of consecutive tests were an additional round of slice testing required as the prevalence rose; (ii) no significant loss of sensitivity for samples with appreciable viral loads (Ct ≤ 32) was demonstrated and (iii) the practical nature of our pooling workflow, i.e., the implementation of liquid-handling robots for automated pooling and software applications for the automated identification of the positive sample.
There are more conventional methods for high-throughput diagnostic testing which are available. For example, SARS-CoV-2 lateral flow antigen (LFA) tests are point of care rapid tests that can be used for large-scale screening. Although LFAs may be appealing because they are cheap, do not require a laboratory with specialized equipment or personnel and can provide results in 15–30 minutes, they do present some limitations; the sensitivity of this type of technology can have considerable variation (average sensitivity of 56.2% (95% CI 29.5 to 79.8%),20 thereby decreasing its utility in screening certain populations (such as health-care workers and other front-line personnel). We have performed a detailed cost comparison of the hypercube-based pooled testing strategy compared to the LFAs. We estimate that, at prevalences below 0.43%, it costs over 6 times more to achieve reliable detection of SARS-CoV-2 infectious individuals by using LFAs compared to hypercube testing (assuming a detection probability of 99.9%) (Supplementary Information - Cost comparison of hypercube-based pooled tests vs. lateral flow antigen tests).
The pooled testing strategy described in this study, together with the downstream automation has led to a significant reduction in cost, kit usage and turnaround time at our laboratory. Our application of hypercube pooling to frequent testing of a professional sports team can potentially be extended to other low prevalence, asymptomatic population groups. Further evaluation of this approach in different epidemic situations (with variable prevalence) is required.