Microbiome analysis of over 2000 NHS Bowel Cancer Screening Programme (NHSBCSP) samples shows the potential to improve screening accuracy

Background CRC is the fourth global cause of cancer deaths. As CRC is associated with an altered faecal microbiome, there is the potential for faecal microbiome analysis to improve CRC screening. This has been demonstrated by a number of studies, however it has never been tested using samples collected routinely by a national CRC screening programme.MethodsBetween 2016-2019 the largest of the five NHS Bowel Cancer Screening Programme (NHSBCSP) hubs, the Southern Hub, prospectively collected and anonymised processed card samples with recorded outcomes: blood-negative (n=491); CRC (n=434); adenoma (n=668); colonoscopy-normal (n=303); and non-neoplastic (n=380). Extracted DNA underwent V4 16SrRNA NGS, with QIIME2, LEfSe and Random Forest analysis.Findings The microbiome was successfully analysed from screening samples. LEfSe analysis confirmed CRC-microbiome associations described in the existing literature, but interestingly showed that these were affected by choice of control group (blood-negative or colonoscopy-normal). Random Forest models based on relative abundance of genera distinguished: CRC from blood-negative samples with AUC: 0.896 (0.869-0.923); CRC from colonoscopy-normal samples with AUC: 0.77 (0.72-0.815); neoplasm from blood-negative samples with AUC: 0.825 (0.788-0.856); and neoplasm from colonoscopy-normal samples with AUC: 0.736 (0.688-0.781). Models remained robust when restricted to the top ten-fifteen taxa. CRC-enriched taxa included Fusobacterium, Peptostreptococcus, Parvimonas, Gemella and Porphyromonas. The microbiome of samples stored at ambient temperature up to 23 months was stable. Random Forest models based on relative abundance of genera distinguished: CRC from blood-negative samples with AUC(validation): 0.896 (0.869-0.923); CRC from colonoscopy-normal samples with AUC(validation): 0.77 (0.72-0.815); neoplasm from blood-negative samples with AUC(validation): 0.825 (0.788-0.856); and neoplasm from colonoscopy-normal samples with AUC(validation): 0.736 (0.688-0.781). All of the models performed significantly better than models generated for comparison which used age and gender alone.InterpretationThis study demonstrates that microbiome analysis can be performed on samples collected routinely by a national CRC screening programme to improve accuracy. The models required as few as ten taxa, which raises the potential of an inexpensive qPCR-based test. This could reduce the number of unnecessary colonoscopies in countries which use faecal occult blood test screening.