Breath volatile organic compound testing in colorectal cancer: diagnostic trade-offs and clinical implications

We read with interest the study by Ji et al.,1 which developed a breath panel comprising 10 volatile organic compounds (VOCs) that discriminated colorectal cancer (CRC) from controls and other gastrointestinal conditions, reporting that the addition of four serum biomarkers increased the area under the curve from 0.834 to 0.881. However, a reappraisal of the test-set metrics shows that this improvement was primarily driven by greater specificity, while sensitivity declined markedly (recall, from 0.82 to 0.61).1 When expressed as likelihood ratios, the VOC-only model yielded an estimated positive likelihood ratio (LR+) of about 5.5 and a negative likelihood ratio (LR−) of about 0.21, whereas the combined VOC-serum model achieved a slightly higher LR+ near 6 but a weaker LR− around 0.4. This trade-off improves rule-in performance but reduces the ability to rule out disease, an important limitation for CRC screening frameworks that prioritize sensitivity and tolerate false positives followed by diagnostic colonoscopy.2