Letter: Methodological Gaps Undermine Conclusions on Electronic Nose Use for Colorectal Cancer Detection

Abstract

We write to express our interest in the study by van Riswijk et al. [1], which evaluated the diagnostic performance of an electronic nose (e-nose) for colorectal cancer (CRC) in faecal immunochemical test (FIT)-positive patients. While the multicentre design and focus on external validation are commendable, the study's conclusions may be compromised by methodological and interpretative gaps that warrant discussion.

A limitation is the absence of validated biomarker-linkage. The authors assumed that exhaled volatile organic compounds (VOCs) correlate with CRC but did not establish the specificity of these VOCs as tumour-derived biomarkers. Without mechanistic data on how CRC-specific metabolism or microbiota produce detectable VOCs in breath, the association between e-nose signals and disease remains observational. Prior studies [2] have emphasised the need to validate VOC-tumour causality through stool or tissue correlation, which was absent here. The lack of molecular-pathological validation weakens the scientific foundation for e-nose utility.

Additionally, the study did not adequately control for confounding variables. Key factors such as smoking, alcohol use and pre-existing conditions (e.g., haemorrhoids in controls) can influence VOC profiles, yet multivariate analyses to isolate their effects were absent. Moreover, the distribution of CRC cases across 21 devices introduced unquantified ‘device-to-device’ variability, described in sensor research [3] as a major confounder. These issues raise doubt about whether the observed low AUC (0.542) reflects true biological noise or technical/epidemiological biases.

Finally, the static assessment of e-nose performance diminished its clinical relevance. By relying on single-timepoint breath tests, the study overlooked the dynamic nature of VOC profiles across tumour stages (e.g., adenoma progression). Longitudinal data or repeated testing—critical for evaluating screening tools—were absent, making it difficult to determine if the technology's limitations are inherent or due to sampling timing. This contrasts with microbiome-based CRC studies [4], which demonstrate improved accuracy through longitudinal modelling.

In conclusion, while the study has provided valuable real-world data, its failure to address biomarker specificity, control confounding variables and assess temporal dynamics undermines the robustness of its conclusions. To strengthen the evidence for or against e-nose technology, future research should prioritise mechanistic validation of VOC-tumour links, implement rigorous statistical adjustment for confounders and incorporate longitudinal sampling. These points are essential for advancing non-invasive CRC screening methodologies.

Ding Shi: conceptualization, methodology, investigation, validation, formal analysis, funding acquisition, project administration, resources, data curation, supervision, software, visualization, writing – original draft, writing – review and editing. Buyuan Dong: software, writing – original draft, writing – review and editing, conceptualization, data curation, formal analysis, visualization, methodology, investigation, supervision, project administration, validation, resources.

The authors declare no conflicts of interest.

This article is linked to van Riswijk et al. paper. To view this article, visit https://doi.org/10.1111/apt.70207 and https://doi.org/10.1111/apt.70243.