RR\ID Evidence Scale rating by reviewer:
Potentially informative. The main claims made are not strongly justified by the methods and data, but may yield some insight. The results and conclusions of the study may resemble those from the hypothetical ideal study, but there is substantial room for doubt. Decision-makers should consider this evidence only with a thorough understanding of its weaknesses, alongside other evidence and theory. Decision-makers should not consider this actionable, unless the weaknesses are clearly understood and there is other theory and evidence to further support it.
***************************************
Review: The paper describes a novel pooling method based on Hadamard matrices - which allow decoding by pattern matching and claim it is well suited for low-prevalence screening in cases where a large number of samples are being tested. Under specific assumptions the method can provide decoding using a single round of testing, but this is not the case when the positivity rate increases even slightly.
The method decodes tests using a binary test result - positive/negative, and in essence ignores the quantitative results provided by current qRT-PCR methods that are widely used in clinical settings. As shown in previous studies, this information is useful for decoding and can allow to also identify individuals with low viral load, which may either be recovering, or at the initial stages of their infection.
Another claim made in the paper is that other pooling algorithms require computational resources that are "expensive". Previous work by Shental et. al, and later followup work by Zissmanov et al. (Communications Medicine 2024 https://doi.org/10.1038/s43856-024-00531-w) reported a clinically validated combinatorial pooling method that was used to test over 800K samples using software that was installed on standard laboratory PC computers and provided real--time decoding of results automatically. These studies demonstrate that single-stage methods can be effectively used and their computational requirements are not ones that cannot be readily supported within real-world lab settings.
Another important consideration in real-world testing settings is that positive samples may sometimes cluster together due to sample collection strategies. - e.g. an outbreak in an elderly home, or samples from a family with several infected individuals. Therefore, even if the prevalence is low, a pooling method needs to be robust to batch variations that may occur.
Dorfman pooling, which has been widely used for diagnostic screening since World War I, is very easy to implement, and in fact does not require any computational power at all. When the prevalence is very low, pool sizes can be selected to be larger thereby further reducing the number of tests required.
Other testing strategies used during the recent pandemic included collecting both pooled samples and individual samples from schools and nursing homes. Pools were "generated" at the sampling sites directly, reducing the need for laboratory based pooling which requires liquid dispensing robots. Pools were first tested and positive pools were then deconvoluted from the individual samples.
Creative Commons Attribution 4.0 International License (CC-BY 4.0)
