Antibiotic prophylaxis and the gastrointestinal resistome in paediatric patients with acute lymphoblastic leukaemia: a cohort study with metagenomic sequencing analysis

Abstract

Background: Although antibiotic prophylaxis with levofloxacin can reduce the risk of serious infection in immunocompromised patients, the potential contribution of prophylaxis to antibiotic resistance is a major drawback. We aimed to identify the effects of levofloxacin prophylaxis, given to paediatric patients with acute lymphoblastic leukaemia to prevent infections during induction chemotherapy, on antibiotic resistance in gastrointestinal microbiota after completion of induction and consolidation therapy.

Methods: This prospective, single-centre (St Jude Children's Research Hospital, Memphis, TN, USA) cohort study included children (≤18 years) receiving therapy for newly diagnosed acute lymphoblastic leukaemia and who received either primary levofloxacin prophylaxis or no antibacterial prophylaxis (aside from Pneumocystis jirovecii prophylaxis with trimethoprim-sulfamethoxazole) and provided at least two stool samples, including one after completion of induction therapy. We used metagenomic sequencing to identify bacterial genes that confer resistance to fluoroquinolones, trimethoprim-sulfamethoxazole, or other antibiotics, and to identify point mutations in bacterial topoisomerases (gyrA, parC) that confer resistance to fluoroquinolones. We then used generalised linear mixed models to compare the prevalence and relative abundance of antibiotic resistance gene groups after completion of induction and consolidation therapy between participants who had received levofloxacin and those who received no prophylaxis.

Findings: Between Feb 1, 2012, and April 30, 2016, 118 stool samples (32 baseline, 49 after induction, and 37 after consolidation) were collected from 49 evaluable participants; of these participants, 31 (63%) received levofloxacin prophylaxis during induction therapy and 18 (37%) received no antibacterial prophylaxis. Over the course of induction therapy, there was an overall increase in the relative abundance of trimethoprim-sulfamethoxazole resistance genes (estimated mean fold change 5·9, 95% CI 3·6-9·6; p<0·0001), which was not modified by levofloxacin prophylaxis (p=0·46). By contrast, the prevalence of topoisomerase point mutations increased over the course of induction therapy in levofloxacin recipients (mean prevalence 10·4% [95% CI 3·2-25·4] after induction therapy vs 3·7% [0·2-22·5] at baseline) but not other participants (0% vs 0%; p<0·0001). There was no significant difference between prophylaxis groups with respect to changes in aminoglycoside, β-lactam, vancomycin, or multidrug resistance genes after completion of induction or consolidation therapy.

Interpretation: Analysing the gastrointestinal resistome can provide insights into the effects of antibiotics on the risk of antibiotic-resistant infections. In this study, antibiotic prophylaxis with trimethoprim-sulfamethoxazole or levofloxacin during induction therapy for acute lymphoblastic leukaemia appeared to increase the short-term and medium-term risk of colonisation with bacteria resistant to these antibiotics, but not to other drugs. More research is needed to determine the longer-term effects of antibacterial prophylaxis on colonisation with antibiotic-resistant bacteria.

Funding: Children's Infection Defense Center at St Jude Children's Research Hospital, American Lebanese Syrian Associated Charities, and National Institutes of Health.

Figure 1:. Changes in prevalence and relative abundance of trimethoprim-sulfamethoxazole antibiotic resistance genes during induction therapy for acute lymphoblastic leukaemia

(A) Prevalence of trimethoprim-sulfamethoxazole resistance genes estimated from generalised linear mixed models, with and without prophylaxis group effect and accounting for sequence depth. The significance of the change in prevalence of trimethoprim-sulfamethoxazole resistance genes during induction therapy was not evaluated because the overall prevalence was too high. (B) Estimated changes in relative abundance of trimethoprim-sulfamethoxazole antibiotic resistance genes, estimated from generalised linear mixed models and expressed as the fold change in RPKM. RPKM=reads per kilobase gene length per million bacterial reads.

Figure 2:. Changes in prevalence and relative abundance of fluoroquinolone antibiotic resistance genes and topoisomerase point mutations during induction therapy for acute lymphoblastic leukaemia

(A) Prevalence of all quinolone antibiotic resistance genes and resistance mutations and of the topoisomerase point mutation subset (mutations in the gyrA and parC, which confer resistance to fluoroquinolones), estimated from generalised linear mixed models, with and without prophylaxis group effect and accounting for sequence depth. (B) Estimated changes in relative abundance of quinolone antibiotic resistance genes, primarily topoisomerase point mutations, estimated as the fold change in RPKM from generalised linear mixed models. RPKM=reads per kilobase gene length per million bacterial reads.

Figure 3:. Changes in prevalence and relative abundance of selected antibiotic resistance gene classes during induction therapy for acute lymphoblastic leukaemia

(A) Prevalence of aminoglycoside, β-lactam, multidrug, and vancomycin resistance genes estimated from generalised linear mixed models. (B) Estimated changes in relative abundance of aminoglycoside, β-lactam, multidrug, and vancomycin resistance genes estimated from generalised linear mixed models. RPKM=reads per kilobase per million bacterial reads.