eNose technologies in the detection of cancer: a systematic review and meta-analysis
Abstract
Background
In 2004, researchers in the United Kingdom were able to train six dogs to distinguish between urine samples from bladder cancer patients and healthy controls using smell, achieving a detection rate of 41% (95% confidence interval [CI]: 23-58%), far above the 14% expected by chance. After numerous subsequent studies validated that the odor of breath and urine samples could be used by dogs to detect cancer, researchers pivoted to electronic noses (eNoses), sensor-based systems that mimic the sense of smell using arrays of chemical detectors. In this study, we review the potential efficacy of eNoses in the detection of selected cancer types in human biological samples.
Methods
We identified and performed a meta-analysis on 37 studies of eNose technology, comprising 1 365 cancer patients and 2 249 control subjects. We calculated the pooled sensitivity and specificity stratified by cancer type, sample type, and sensor type. Meta-regressions were conducted on these variables as well as the number of sensors used in the sensor array.
Results
All six cancer types analyzed—breast, colorectal, gastric, lung, ovarian, and prostate—achieved pooled sensitivities and specificities above 70%, with most around 85%. The overall pooled sensitivity was 85.9% (95% CI: 82.3-88.9%) and specificity was 83.6% (95% CI: 78.6-87.7%). Meta-regression revealed that the number of sensors in the sensor arrays, up to 15 sensors, was predictive of sensitivity with PFDR < 0.001.
Conclusion
This analysis found that eNoses constitute a promising tool in the early detection of cancer. However, more research is necessary before it can be introduced into clinical settings.