The impact of different antimicrobial exposures on the gut microbiome in the ARMORD observational study
- Oxford University Hospitals NHS Foundation Trust, United Kingdom
- Nuffield Department of Medicine, University of Oxford, United Kingdom
- Anthony Nolan Research Institute, Royal Free Hospital, Hampstead, United Kingdom
- NIHR Health Protection Research Unit in Healthcare Associated Infections and Antimicrobial Resistance, University of Oxford, United Kingdom
- Brighton and Sussex Medical School, United Kingdom
- University Hospitals Sussex NHS Foundation Trust, United Kingdom
Figures
Figure 1
Independent effect of specific antimicrobial exposures on Shannon diversity in (A) cross-sectional and (B) longitudinal analyses.
Multivariable estimates are in black, univariable (unadjusted) estimates in grey. Error bars represent 95% confidence intervals. Numbers by exposure represent sample size (n). Non-antimicrobial covariates are not shown but were included in the model and can be found in Source data 1. Results are not plotted for four antimicrobials (n=6–12) that had standard errors >3 and did not differ significantly from zero. Estimates represent the impact of prolonged use, when exposure ≈ 1 (approximately 42 days, see Appendix 1—figure 2).
Figure 2
Independent effects of specific antimicrobial exposures on relative abundance of selected taxa, in (A) cross-sectional and (B) longitudinal multivariable analyses.
Error bars represent 95% confidence intervals. Numbers by each exposure represent sample size (n). Non-antimicrobial covariates are not shown but were included in the model and can be found in Source data 1. Estimates represent the impact of prolonged use, when exposure ≈ 1 (approximately 42 days, see Appendix 1—figure 2).
Figure 3
Independent effects of specific antimicrobial exposures on relative abundance of selected AMR genes, in (A) cross-sectional and (B) longitudinal multivariable analyses.
Error bars represent 95% confidence intervals. Numbers by each exposure represent sample size (n). Non-antimicrobial covariates are not shown but were included in the model and can be found in Source data 1. Estimates represent the impact of prolonged use, when exposure ≈ 1 (approximately 42 days, see Appendix 1—figure 2).
Appendix 1—figure 1
Depiction of antimicrobial exposure model.
Modelled exposure for each antimicrobial is equal to the sum of shaded areas. The microbiome disruption half-life (lambda) in this example is 6 days.
Appendix 1—figure 2
Modelled exposure to antimicrobial courses of varying duration.
Modelled exposure to a single course of varying duration, starting at day 0. The microbiome disruption half-life (lambda) in this example is 6 days.
Appendix 1—figure 3
ARMORD study CONSORT diagram.
HCT = Haematopoietic cell transplant.
Appendix 1—figure 4
Independent effects of exposure to different antimicrobial classes on Shannon diversity in (A) cross-sectional and (B) longitudinal analysis.
Multivariable estimates are in black, univariable (unadjusted) estimates in grey. Error bars represent 95% confidence intervals. Numbers by each exposure represent sample size (n). Non-antimicrobial covariates are not plotted here but were included in the model. ‘Narrow’ beta-lactams are penicillin, amoxicillin, flucloxacillin, and first-generation cephalosporins; all others are defined as ‘broad’. Estimates represent the impact of prolonged use, when exposure ≈ 1 (approximately 42 days, see Appendix 1—figure 2).
Appendix 1—figure 5
Independent effects of exposure to different antimicrobial classes on relative abundance of selected taxa in (A) cross-sectional and (B) longitudinal analysis.
Error bars represent 95% confidence intervals. Numbers by each exposure represent sample size (n). Non-antimicrobial covariates are not shown but were included in the model. Antimicrobial categories are the same as Appendix 1—figure 4. Estimates represent the impact of prolonged use, when exposure ≈ 1 (approximately 42 days, see Appendix 1—figure 2).
Appendix 1—figure 6
Independent effects of exposure to different antimicrobial classes on relative abundance of selected AMR genes in (A) cross-sectional and (B) longitudinal analysis.
Error bars represent 95% confidence intervals. Numbers by each exposure represent sample size (n). Non-antimicrobial covariates are not shown but were included in the model. Antimicrobial categories are the same as Appendix 1—figure 4. Estimates represent the impact of prolonged use, when exposure ≈ 1 (approximately 42 days, see Appendix 1—figure 2).
Tables
Table 1
Characteristics of participants in cross-sectional analysis.
| Healthy volunteers (n=33) | General medical patients (n=91) | HCT patients (n=101) | All participants(n=225) | |
|---|---|---|---|---|
| Age, years (median, IQR) | 37 (31–49) | 76 (67–83) | 58 (50-66) | 64 (50–73) |
| Sex (n, %) | ||||
| Male | 7 (21%) | 53 (58%) | 60 (59%) | 120 (53%) |
| Female | 26 (79%) | 38 (42%) | 41 (41%) | 105 (47%) |
| Recent antibiotic use (n, %) | ||||
| Receiving antibiotics at time of sampling | 0 (0%) | 55 (60%) | 42 (42%) | 97 (43%) |
| Use in past month (but not at time of sampling) | 3 (9%) | 26 (29%) | 22 (22%) | 51 (24%) |
| Use in past year (but not in past month) | 4 (12%) | 6 (7%) | 33 (33%) | 43 (19%) |
| No antibiotics in past year | 26 (79%) | 4 (4%) | 4 (4%) | 34 (15%) |
| Max Charlson index in past year (median, IQR)* | 0 (0–0) | 4 (0–13) | 0 (0–8) | 0 (0–8) |
| Maximum values in past 14 days (median, IQR) | ||||
| NEWS2* | 0 (0–0) | 5 (2-8) | 3 (2-4) | 3 (1-5) |
| C-reactive protein† | 0.2 (0.2–0.2) | 63 (22–163) | 10 (3–69) | 21 (2–81) |
| White cell count‡ | 7.5 (7.5–7.5) | 11.5 (8.4–14.4) | 7.6 (5.8–10.9) | 8.4 (7.4–12.4) |
| Days of chemotherapy at time of sampling (median, IQR) | 0 (0–0) | 0 (0–0) | 3.0 (1.4–7.2) | 0 (0–2.8) |
-
*
Imputed as 0 if no observations recorded, see Methods. NEWS2 is National Early Warning Score 2.
-
†
Imputed as 0.2 if no result recorded.
-
‡
Imputed as 7.5 if no result recorded.
Appendix 1—table 1
Identification of best-fit antimicrobial exposure half-life.
The microbiome disruption half-life with the lowest AIC (i.e. best fit) was used for subsequent analyses.
| Model half-life (days) | R2 | Adjusted R2 | Akaike Information Criterion (AIC) |
|---|---|---|---|
| 1 | 0.4078 | 0.3162 | 477.3 |
| 2 | 0.4331 | 0.3351 | 473.47 |
| 3 | 0.4583 | 0.358 | 467.24 |
| 4 | 0.4676 | 0.369 | 463.35 |
| 5 | 0.4747 | 0.3774 | 460.33 |
| 6 | 0.505 | 0.4102 | 448.95 |
| 7 | 0.5048 | 0.41 | 449.05 |
| 8 | 0.5035 | 0.4084 | 449.65 |
| 9 | 0.5015 | 0.406 | 450.55 |
| 10 | 0.4993 | 0.4035 | 451.51 |
| 14 | 0.4925 | 0.3921 | 456.57 |
Appendix 1—table 2
Relative abundance of major taxa in baseline samples.
| Taxon | Median relative abundance (IQR) % |
|---|---|
| Enterococcus | 0.13 (0.056–1.3) |
| Enterobacteriaceae | 1.4 (0.055–7.7) |
| Bacteroidetes | 38 (18–61) |
| Clostridia | 22 (10–40) |
| Actinobacteria | 3.3 (0.70–9.1) |
| All taxa above | 91 (85–96) |
Additional files
-
Supplementary file 1
ARMORD case report form.
- https://cdn.elifesciences.org/articles/97751/elife-97751-supp1-v1.docx
-
MDAR checklist
- https://cdn.elifesciences.org/articles/97751/elife-97751-mdarchecklist1-v1.docx
-
Source code 1
R project allowing reproduction of the analyses in this paper.
- https://cdn.elifesciences.org/articles/97751/elife-97751-code1-v1.zip
-
Source data 1
Deidentified study data and outputs used to plot Figures 1—3 and Appendix 1—figures 4–6.
- https://cdn.elifesciences.org/articles/97751/elife-97751-data1-v1.zip
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
The impact of different antimicrobial exposures on the gut microbiome in the ARMORD observational study
eLife 13:RP97751.
https://doi.org/10.7554/eLife.97751.3
