Abstract
Colorectal cancer (CRC) is a significant public health burden, with New Zealand exhibiting some of the highest incidence rates worldwide and poorer mortality rates and survival outcomes compared to Australia. The slow progression of CRC offers an opportunity for early detection through screening; however, current diagnostic tests have limitations in cost, acceptability, accuracy, and/or invasiveness, highlighting the need for a screening test better suited to New Zealand’s population needs.
Volatile organic compounds (VOCs) in breath have shown promise in discriminating advanced colorectal neoplasia (CRC and advanced adenoma) from healthy individuals with high accuracy. However, most studies have been conducted under optimal research conditions, with few in clinical settings. Additionally, previous studies only assessed outcomes from breath stored for less than one day, whereas implementing a breath test in New Zealand screening setting would require longer storage, particularly for rural communities. This need for sample storage also allows for potential centralisation of analysis.
To address these real-world requirements, we adopted a practical sampling, storage, transport, and analysis system tailored to New Zealand. Breath VOCs were stored in easily posted thermal desorption (TD) tubes, shipped via air freight, and analysed using selected ion flow tube mass spectrometry (SIFT-MS) - a fast, automated technology able to process hundreds of samples daily. We selected 24 candidate VOCs, prioritising aldehydes and carboxylic acids based on recent research.
In Chapter 3 we assessed technical baselines, finding residual VOCs in ‘clean’ tubes, with ethanol being particularly high. Refrigerated tube storage produced less variable VOC outcomes than room temperature storage. We found that aldehyde levels in breath samples decreased with storage time. Hand sanitiser use significantly increased ethanol and butanol levels, which could cause invalid results. We also noted some high levels of technical variability in some breath samples for some VOCs, and that our results were not comparable with results from other research using alternative protocols.
In Chapter 4, assessing natural variation in healthy breath, we found no fasting/fed differences in samples taken on the same day, although individual variance in VOC concentrations were high. Acetone was the only VOC that showed significant day-to-day variation when fasted.
In Chapter 5, evaluating 21 remaining candidate VOCs in a cohort of patients symptomatic for CRC (n=139), 3 VOCs showed significant differences between advanced neoplasia and controls, based on area under the curve (AUC) values. Butanoic acid and propanoic acid levels increased with disease (AUC 0.67 and 0.65 respectively) whereas acrolein levels decreased with disease (AUC 0.65). An acrolein/propanoic acid panel provided an AUC of 0.68. We also showed that levels of carcinoembryonic antigen (CEA) did not significantly correlate with VOC levels in a small subset of CRC patients.
While our breath test detected advanced neoplasia with moderate accuracy, sensitivities and specificities were too low for clinical utility compared to FIT in symptomatic cohorts. However, improving VOC capture and storage and assessing all aldehydes and carboxylic acids in samples could increase accuracy. Further research exploring this non-invasive screening approach is recommended, potentially allowing opportunistic primary care sampling.