Prospective Cohort Study of Non-COVID Infection in Patients with AML/MDS Undergoing Treatment: Implications for Practice

Infection remains a major cause of mortality and morbidity in patients with acute myeloid leukaemia (AML) and high-risk myelodysplasia (MDS). The recent advent of venetoclax based therapies have altered the degree of immunosuppression, increasing treatment options for more frail patients. The marked heterogeneity in the frequency and severity of infections between all patients with AML is poorly understood. To improve our ability to stratify infection risk in patients undergoing treatment, we prospectively collected clinical information alongside microbiome and transcriptome data, from patients with AML/MDS undergoing chemotherapy as part of the PACE (Patients with AML and COVID-19 Epidemiology) study.

Methods: Between May 2020 and April 2021, we prospectively collected information on treatment, and disease alongside rates and severity of all infections in patients with AML undergoing chemotherapy. We have previously reported in this cohort of patients, the risk of SARS-CoV2 infection and their response to vaccination. We collected blood, stool and sputum from 65 participants to explore biomarkers that could be used to understand how patients respond to serious infections, in conjunction with treatments.

Results: 230 patients with AML or high risk MDS undergoing chemotherapy were recruited from 30 sites. The median age was 65 (range:16-86), of whom 128/230 (55.7%) were male. 202/230 (87.8%) of the cohort had a diagnosis of AML, 36 (17.8%) of whom had prior MDS. 197/230 (85.7%) of patients were newly diagnosed, with the remainder having relapsed disease. 126/230 (54.8%) patients were intensively treated (INT), 49 (21.3%) patients received venetoclax based therapy (VEN), and 51 (22.2%) non-intensive treatment (NON-INT).

160 patients reported to have at least one non-COVID related infection whilst on study, 107/126 (84.9%) INT patients had at least one significant infection. In this group, 389 infections were recorded, occurring at a mean rate of 0.5 infection per month (range: 0.0-15.2). In contrast, 26/49 (53.1%) of VEN patients at a rate of 0.2 infections per month (0-1.8) and 26/47 (55.3%) NON-INT patients had one significant infection, occurring at a mean rate of 0.3 infections per month (0-4.2). 78 (48.8%) patients had >/=3 infections. In keeping with an increased rate of infection, of these 78 patients 66 (52% of INT patients) were intensively treated, whilst only 5 (10%) had VEN, and 7 (15%) had NON-INT.

We next examined the pattern of infections in the different treatment groups. The most common infections overall, were systemic/blood infections (260/472; 55.1%). However, this differed according to treatment class. In the INT group, 96/107 (89.7%) of patients with at least one infection, had experienced a systemic/blood infection. In contrast, in the NON-INT and VEN group only 14/26 (53.8%) and 10/26 (38.5%) respectively had a systemic/blood infection. In these latter two groups of patients, localised infections were comparable in frequency (NON-INT 12/26 (46.2%); VEN 15/26 (57.7%)). Length of hospital stay differed according to treatment group. 426/126 (number of episodes of hospitalisation with >/=1 infection /patients) for INT, 122/47 for NON-INT and 76/49 for VEN were reported.

In our correlative sampling study in this cohort, 16S microbiome was highly abnormal in AML/MDS patients: alpha diversity in sputum were significantly lower compared to healthy controls. Beta diversity revealed clear differentiation vs healthy via weighted or unweighted UniFrac distance on principal component analysis (p=0.001 for both). At times of infection, quantitative PCR revealed a significant increase in serum bacterial DNA and transcriptomic analysis revealed multiple differentially expressed host genes whilst stool microbiome was altered significantly compared to baseline or monthly time points (e.g. lower abundance of Proteobacteria, p=0.006).

Summary: Our prospective real-word data on infection epidemiology supports a high burden of significant infection. There are highly different patterns of infection (both source and severity) between INT patients and others, but a more similar pattern between VEN and NON-INT. This has been complemented by the results of our multiomic profiling of patients suggesting significant microbiome and immune dysfunction, following infections. Our study has implications on informing novel anti-microbial strategies in these groups of patients.