Urinary Volatile Organic Compounds as Non-Invasive Biomarkers to Distinguish Hepatocellular Carcinoma from Cirrhosis: A Multicentre Diagnostic Accuracy Study.

Hepatocellular carcinoma (HCC) typically arises on a background of cirrhosis and carries high mortality, in large part because current surveillance tools-ultrasound and serum AFP-have suboptimal sensitivity for early-stage disease. We investigated whether urinary volatile organic compounds (VOCs) could provide a more accurate non-invasive approach. Across four UK centres, 256 participants were recruited: 72 with confirmed HCC and 184 with cirrhosis but no malignancy. Complete urine VOC profiles were obtained using an eight-sensor electronic nose in 190 participants (58 HCC, 132 cirrhosis). Seven of the eight sensors showed statistically significant differences between groups, all reaching p < 0.001 (S1-S4, S6-S8), while one (S5) did not reach significance (p = 0.682). Principal components analysis demonstrated clear separation between HCC and cirrhosis populations, with the first two components accounting for 70.7% of total variance. A radial basis function neural network, trained on 20% of the data and tested on an independent second replicate, achieved sensitivity of 98.28% (95% CI: 90.76% to 99.96%), specificity of 96.97% (95% CI: 92.42% to 99.17%), PPV of 93.4% (95% CI: 84.44% to 97.40%) and NPV of 99.2% (95% CI: 94.83% to 99.89%). The area under the ROC curve was 0.99 (95% CI: 0.959-1.000). Performance was robust in a mixed treated/untreated cohort, with 50% of HCC patients untreated at sampling; two treated cases fell within the cirrhosis cluster, indicating that VOC profiling may be most sensitive for active, untreated disease. What's New? Urinary VOC profiling demonstrates improved diagnostic accuracy for distinguishing HCC from cirrhosis compared to conventional screening of serum alpha-fetoprotein combined with ultrasound imaging. This non-invasive approach offers substantial clinical potential for HCC screening in high-risk cirrhotic populations at lower cost but will need validation and economic evaluation in larger independent cohorts.