Volatile organic compounds (VOCs) are generated during pathologic processes giving unique characteristics to breath. Breath analysis has thus become an important method to diagnose and monitor a variety of diseases. There is increasing data to support a critical role for the intestinal microbiome in graft-versus-host disease (GVHD). Various microorganisms are known to produce characteristic volatile metabolites, and may alter VOCs expelled in breath. We hypothesized that patients with gastrointestinal (GI) GVHD will have a unique breath signature compared to transplant recipients without GVHD and performed a pilot study of breath analysis in HCT recipients.

Exhaled breath samples were obtained from 19 patients with clinical grade 2-4 acute GI GVHD and 10 patients with no evidence of GVHD by day 100; and analyzed using mass spectrometry technology. Baseline characteristics and VOCs were compared between patients with and without GI GVHD. 22 VOCs were included in stepwise linear discriminant analysis to identify the subset that could best discriminate between GI GVHD and no GVHD.

Median age at transplant was 58 years (range 13-71). The majority of patients (93%) were White, and 66% male. 55% of patients underwent myeloablative conditioning and 45% reduced intensity, with 52% receiving peripheral blood stem cells, 41% bone marrow, and 7% cord blood grafts. 45% patients had a matched unrelated donor, 34% matched sibling donor, 14% haploidentical, and 7% cord blood transplants. GVHD prophylaxis primarily consisted of a calcinuerin inhibitor in combination with mycophenolate mofetil or methotrexate. Among recipients who developed GI GVHD, median time to onset of GVHD was 43 days from transplant. 9 (47%) patients had grade 2, 8 (42%) grade 3, and 2 (11%) grade 4 GVHD. 26% had only upper GI GVHD, 21% lower GI GVHD and 53% had both upper and lower GI manifestations.

Stepwise canonical discriminant analysis identified 5 VOCs that distinguished between patients with and without GI GVHD: 2-propanolol, acetaldehyde, dimethylsulfide, isoprene, and 1-decene (Wilks' Lambda 0.43, F statistic 6.08, P=0.001). The model correctly classified 89% (17/19) and 90% (9/10) of patients with and without GI GVHD, respectively (Figure 1). We found no significant correlation between VOCs and GVHD severity. However, 2 VOCs, pentane and ammonia, distinguished patients who achieved a complete response (CR) to therapy by day 28 and those who did not (Wilks' Lambda 0.57, F statistic 6.05, P=0.011). The model correctly classified 90% (9/10) patients with CR and 78% (7/9) of patients with less than CR (Figure 2). An additional VOC, acetone, distinguished patients with at least a partial response versus no response. Patients with higher levels of acetone were less likely to have any response, odds ratio 0.22, 95% confidence interval 0.06-0.82, P=0.024.

Breath analysis is a feasible and promising non-invasive method to diagnose and monitor acute GI GVHD. Further study of serial breath analysis and the gut microbiome in a larger cohort are needed to validate these findings.

Disclosures

Gerds: CTI BioPharma: Consultancy; Incyte: Consultancy. Majhail: Sanofi: Honoraria; Anthem, Inc.: Consultancy.

Author notes

*

Asterisk with author names denotes non-ASH members.

Sign in via your Institution