Metagenomic sequencing as a pathogen-agnostic clinical diagnostic tool for infectious diseases: a systematic review and meta-analysis of diagnostic test accuracy studies.
Govender KN., Street TL., Sanderson ND., Eyre DW.
Background: Metagenomic sequencing is frequently claimed to have the potential to revolutionise microbiology through rapid species identification and antimicrobial resistance (AMR) prediction. We assess progress towards this.Methods: We perform a systematic review and meta-analysis of all published literature on culture-independent metagenomic sequencing for pathogen-agnostic infectious disease diagnostics to August 12, 2020. Methodologic bias and applicability were assessed using QUADAS-2. (PROSPERO CRD42020163777)Results: A total of 2023 clinical samples from 13/21 eligible diagnostic test accuracy studies were included in the meta-analysis. Reference standards were culture, molecular testing, clinical decision or a composite measure. Sensitivity and specificity in the most widely investigated sample types were 90%(78-96%) and 86%(45-98%) for blood, 75%(95%CI, 54-89%) and 96%(72-100%) for CSF, and 84%(79-88%) and 67%(38-87%) for orthopaedic samples respectively. We identified limited use of controls, especially negative controls which were used in only 62%(13/21) studies. AMR prediction and comparison to phenotypic results was undertaken in four studies: categorical agreement was 88%(80%-97%), very major and major error rates were 24%(8-40%) and 5%(0-12%) respectively. Better human DNA depletion methods are required: a median 91%(IQR 82-98%)[range 76-98%] of sequences were classified as human. The median(IQR)[range] time from sample to result was 29(24-94)[4-144] hours. The reported consumables cost per sample ranged from $130-$685.Conclusions: There is scope for improving the quality of reporting in clinical metagenomic studies. Although our results are limited by the heterogeneity displayed, our results reflect a promising outlook for clinical metagenomics. Methodological improvements, and convergence around protocols and best practises may improve performance in future.