Effect of prophylactic antibiotics on days of antibiotic spectrum coverage in comatose post-cardiac arrest patients: a secondary analysis of PROTECT

Abstract

Background

Antibiotic resistance is a threat to public health driven in part by widespread antibiotic administration. Days of antibiotic spectrum coverage (DASC) is a novel metric to quantify both duration and breadth of antibiotic exposure that has not previously been used as an endpoint in a clinical trial. We calculated DASC using data from the Ceftriaxone to Prevent Pneumonia and Inflammation after Cardiac Arrest (PROTECT) trial to determine the association of ceftriaxone prophylaxis with DASC and with the acquisition of antibiotic resistance genes (ARGs).

Methods

PROTECT randomized out-of-hospital cardiac arrest subjects to ceftriaxone or placebo for 3 days. ARGs were measured from rectal swabs collected at Days 0, 3 and 7 post randomization. DASC was calculated for each subject and compared using a two-sided Mann–Whitney U-test. Correlations between DASC and new ARGs, antibiotic-free days (AFD) and days of therapy (DOT) were tested using Kendall’s tau-alpha.

Results

PROTECT enrolled 52 subjects, 26 per treatment group, and treatment groups were similar at baseline. Median DASC scores were lower in the ceftriaxone group (19.5; IQR: 0, 43) compared with placebo (53; IQR: 16, 81). We found no correlation between DASC and new ARGs at either timepoint, or between DASC and AFD. DASC was correlated with DOT.

Conclusions

DASC post intervention was lower in the ceftriaxone group, representing less antibiotic exposure following the intervention. There was no correlation between new ARGs and DASC. Further study is needed to understand the relationship between antibiotic prophylaxis, subsequent antibiotic exposure and resistome changes in the critically ill.

Introduction

Antibiotic resistance is partly a consequence of widespread or untargeted antibiotic administration, common practices when treating critically ill patients.^1,2 The burden of antibiotic resistance is steadily increasing globally and is forecast to kill up to 10 million persons annually by 2050.^1–3 Antibiotic stewardship programmes (ASPs) monitor antibiotic use and are endorsed by the CDC as part of their National Action Plan for Combating Antibiotic Resistant Bacteria.^4,5 ASPs can help to reduce antibiotic resistance and resistant infections through escalation/de-escalation efforts and through antibiotic allocation monitoring efforts for quality improvement and benchmarking.⁶

There are two established metrics recommended for monitoring antibiotic use. The CDC recommends days of therapy (DOT) to quantify antibiotic exposure, which is calculated by counting the number of days an antibiotic is given and taking the aggregate sum if more than one agent is used.⁶ The WHO endorses DDD as an alternative to DOT. DDD is the average maintenance dose per day for a drug being used for its main indication in adults and based on the anatomical therapeutic chemical classification.⁷ Importantly, neither DOT nor DDD consider the spectrum of activity of an antibiotic.⁸

The antibiotic spectrum index was developed in 2017⁹ and assigns a single point for each organism against which an antibiotic is active; however, it was limited in its scope of included antibiotics and organisms.⁸ An extension of the antibiotic spectrum index is the days of antibiotic spectrum coverage (DASC) score, which includes more bacteria and selected resistance patterns (WT without acquired resistance versus commonly isolated with acquired resistance) thereby creating the antibiotic spectrum coverage score, which is then multiplied by duration of antibiotic administration to generate the final DASC score.⁸ DASC incorporates 16 bacteria and 77 antibiotics from 23 classes.⁸ DASC is correlated with DOT, but no clinical trial has examined the effect of interventions on DASC.^8,10–16

Additionally, clinical trials often report antibiotic-free days (AFD) to compare antibiotic regimen lengths or measure changes in antibiotic administration following an intervention, but this metric provides no granularity to describe antibiotic spectrum, which is important when considering the effects of an antibiotic strategy.^17–19 There is no single best metric to describe antibiotic use in clinical trials. One recent trial in brain-injured critically ill patients, the Ceftriaxone to Prevent Early Ventilator-Associated Pneumonia in Patients with Acute Brain Injury (PROPHY-VAP) trial, reported AFD as a metric to describe reduction in post-intervention antibiotic administration.²⁰ Conversely, another recent trial in a similar patient population, the Prevention of Early Ventilator-Associated Pneumonia after Cardiac Arrest (ANTHARTIC) trial, reported the relative percentage of days of antibiotics while admitted in the ICU.²¹

We performed a secondary analysis of the Ceftriaxone to Prevent Pneumonia and Inflammation After Cardiac Arrest (PROTECT; NCT04999592) trial to calculate DASC, DOT and AFD in subjects randomized to ceftriaxone or placebo.^22,23 We hypothesized that DASC after the study drug period would be lower in subjects receiving ceftriaxone compared with placebo. We also hypothesized that DASC would be positively correlated with the presence of newly acquired antibiotic resistance genes (ARGs).

Materials and methods

Parent trial design

PROTECT was a randomized, double-blinded, placebo-controlled pilot trial of antibiotic prophylaxis with ceftriaxone 2 g IV every 12 h for 3 days or placebo in mechanically ventilated, comatose, out-of-hospital (OHCA) subjects treated with targeted temperature management. The trial design and protocol were previously published.²² The trial was approved by the local Institutional Review Board and monitored by an independent Data Safety Monitoring Board. Trial enrollment occurred using traditional consent from a legally authorized representative since subjects were comatose at time of enrollment. Enrollment occurred via Exception from Informed Consent when no legally authorized representative could be contacted within 30 min of initial assessment by the study team. Once a legally authorized representative was able to be contacted, they were asked for informed consent. If subjects regained consciousness, they were asked for informed consent.^22,23 The primary outcome was adjudicator-confirmed early-onset pneumonia ≤4 days after intubation, and the secondary outcomes included open-label antibiotic exposure and ARG acquisition.

Antibiotic resistance genes

ARGs were measured from material recovered from rectal swabs collected prior to study drug initiation (Day 0), 3 days post randomization (after study drug completion), and 7 days post randomization. No minimum amount of detectable faecal material was specified to be collected. ARGs at each post-randomization timepoint were considered new if they were not present at the previous timepoint. Total nucleic acids were extracted from faecal material recovered from rectal swabs using ZymoBIOMICS DNA/RNA plus extraction kit (Zymo Research Corporation). Sequencing libraries were prepared using the Kapa BioSystems HyperPlus Kit (KR1145, v3.16).

Sequencing was performed on a NovaSeq 6000 with an SP flow cell (paired-end 250 bp reads). Data were demultiplexed using bcl2fastq v2.20.0.422. FASTQ files were examined for quality using FastQC v0.11.5²⁴ and multiqc v1.11²⁵ with default settings. Estimation of duplication rates, adapter sequence trimming and read merging were completed using fastp v0.23.2²⁶ with default settings and the flags ‘--include_unmerged’, ‘--merge’, ‘-l 50’ and ‘-g’ to remove poly-G tails produced by the NovaSeq 6000 platform. Human reads were identified by mapping the data against the human reference genome (GRCh38) using Burrows-Wheeler Aligner-MEM v0.7.17²⁷ and filtering the mapped reads using samtools v1.19.2²⁸ (bit flags -f 4 -F262, -f 8 -F 260, and -f 12 and -F 256). Metagenomic assembly was completed from the unmapped reads using the SPAdes pipeline v3.15.5²⁹ with the metagenomic flag and default settings. QUAST v5.2.0³⁰ was used to characterize the contiguity of the genome assemblies. Per sequence coverage statistics were calculated using Burrows-Wheeler Aligner-MEM v0.7.17 and samtools v1.19.2 with default settings.

BLAST v2.14.0³¹ was used to compare sequences against the nucleotide database. Blobtools v1.1.1³² was then used to taxonomically classify each sequence using the ‘bestsum’ taxonomic rule and the NCBI taxonomy database.³³ Any human contigs were removed from the assemblies based on taxonomic assignments. PROKKA v1.14.0³⁴ was used for coding sequence annotation. ARGs were characterized using AMRFinderPlus³⁵ using both the genome assemblies (FASTA) and the annotated genes from PROKKA (FAA). Sequencing quality metrics are available in File S1 (available as Supplementary data at JAC-AMR Online).

Days of antibiotic spectrum coverage

DASC was calculated from subjects randomized to ceftriaxone or placebo using two methods: (i) after the study drug period, and (ii) including the 3 day study drug period. The antibiotic spectrum coverage score was determined for each subject using antibiotics received during the entire hospitalization and then multiplied by the duration of administration to calculate DASC. We adhered to the rubric used in the DASC derivation paper by Kakiuchi et al.⁸ to determine the antibiotic spectrum coverage score for each antibiotic. Subjects receiving ceftriaxone were assigned a DASC score of 6 for each day of study drug, totalling as high as 18 if all three days of study drug were received. Open-label antibiotic administration was at the discretion of bedside treatment teams and consistent with clinical practice guidelines for pneumonia in mechanically ventilated patients.³⁶

Antibiotic-free days and days of therapy

AFD in the first 28 days after randomization were calculated for each subject using previously published methods.³⁷ For subjects randomized to ceftriaxone, AFD were calculated using the day of randomization as Day 0 and included all doses of the study drug. For subjects randomized to placebo, AFD were calculated both after the study drug was completed and including the study drug days, which were without antibiotics. For subjects who died, AFD were counted until the date of death. For subjects who survived and were discharged from the hospital before Day 28, it was assumed they did not receive an antibiotic prescription before the end of the 28 day observation period. DOT was determined by counting the number of days an antibiotic was given and using the aggregate sum if more than one agent was used.⁶

Statistical analysis

Continuous variables were reported using mean and standard deviation (SD), or median and interquartile range (IQR), and categorical variables were reported using numbers with percentages. DASC scores were not normally distributed; therefore, we compared DASC scores between subjects randomized to ceftriaxone or placebo using the two-sided Mann–Whitney U-test. We used Kendall’s tau-alpha (tau-a) to evaluate the correlations between DASC and new ARGs, DASC and AFD, and DASC and DOT. To determine significance, we calculated the Somers’ D estimates with 95% CIs and P values for the Mann–Whitney U-test and Kendall’s tau-a using the somersd statistical package in Stata.^38,39 To report CIs for correlation, we applied Greiner’s rho transformation of Kendall’s tau-a.^38,39 The parent trial was powered to detect an absolute reduction of 25% in the incidence of early-onset pneumonia, from 55% to 30%, providing 80% power for a two-sided relative risk assessment with alpha = 0.05 and a total sample size of 120 subjects; however, the trial was terminated early with 52 subjects enrolled, due to slow enrollment during the COVID-19 pandemic leading to an expiration of funding. All statistical testing was performed using Stata Version 18.5 SE.

Results

From August 2021 to January 2024, 411 subjects were screened, and 53 subjects were enrolled; one subject withdrew consent, leaving 52 subjects for analysis (26 ceftriaxone subjects and 26 placebo subjects; Figure S1). The results of PROTECT have been published previously.²³ Briefly, the treatment groups were similar at baseline, though there was a numerically lower rate of shockable rhythm and higher rate of witnessed arrest with bystander cardiopulmonary resuscitation in subjects randomized to ceftriaxone (Table 1). Mean age was 60 (SD: 52, 67) and 85% were male in the ceftriaxone group, whereas the mean age was 59 (SD: 52, 65) and 92% were male in the placebo group. Early-onset pneumonia within 4 days of intubation was diagnosed in 10 (38%) subjects randomized to ceftriaxone compared with 18 (69%) randomized to placebo. There were no non-pulmonary infections or bacteraemia discovered.²³ Three days of study drug were completed in 16 (62%) subjects randomized to ceftriaxone and in 17 (65%) subjects randomized to placebo. Open-label antibiotics were prescribed by bedside treatment teams to 14 (54%) subjects receiving ceftriaxone and to 22 (85%) subjects receiving placebo (Tables S1 and S2). Broad-spectrum antibiotics were prescribed to 10 (38%) subjects receiving ceftriaxone and to 20 (77%) subjects receiving placebo (Table S2). Unique ARGs isolated in each treatment group are described in Tables S3 and S4, and specific ARGs of high public health relevance (i.e. ESBL, vancomycin, methicillin/penicillin) are listed in Tables S5–S7.

Table 1.

Demographics and baseline characteristics of participants in PROTECT

	Ceftriaxone n = 26	Placebo n = 26
Age, y	60 (SD: 52, 67)	59 (SD: 52, 65)
Sex, male	22 (85)	24 (92)
Race, n (%)
White	25 (96)	25 (96)
Black/African American	1 (4)	0 (0)
Other/multiple races	0 (0)	1 (4)
Ethnicity, n (%)
Non-Hispanic/Latino	26 (100)	25 (96)
Unknown	0 (0)	1 (4)
Weight, kg	87 (IQR: 75, 100)	89 (IQR: 77, 106)
Medical history, n (%)
Coronary disease	9 (35)	3 (12)
Myocardial infarction	7 (27)	3 (12)
Chronic obstructive pulmonary disease	7 (27)	3 (12)
Heart failure	7 (27)	3 (12)
Hypertension	7 (27)	12 (46)
Kidney disease	4 (15)	2 (8)
Out-of-hospital cardiac arrest
Witnessed, yes	20 (77)	13 (50)
Bystander CPR, yes	22 (85)	16 (62)
Time to ROSC, min	23 (IQR: 13, 31)	27 (IQR: 21, 36)
Initial heart rhythm, n (%)
Shockable	11 (42)	15 (58)
Non-shockable	13 (50)	11 (42)
Unknown	2 (8)	0 (0)
Defibrillation, yes	15 (58)	16 (62)
Hospital admission
Glasgow Coma Scale, motor	1 (IQR: 1, 4)	1 (IQR: 1, 1)
SOFA score	11 (IQR: 7, 14)	11 (IQR: 9, 13)
APACHE IV score	71 (IQR: 54, 99)	102 (IQR: 70, 112)
Shock on admission, yes	9 (35)	13 (50)
Lactate, mmol/L	4.0 (IQR: 2.1, 7.9)	5.4 (IQR: 3.2, 9.2)
Targeted temperature management
Target temperature 33°C	21 (81)	22 (85)
Target temperature 36°C	5 (19)	4 (15)

Single values in parentheses represent percentages, dual values in parentheses represent SD (Standard Deviation) or IQR (Interquartile range) as indicated. CPR, cardiopulmonary resuscitation; ROSC, return of spontaneous circulation.

Median DASC score after the study drug period was 19.5 (IQR: 0, 43) for subjects randomized to ceftriaxone compared with 53 (IQR: 16, 81) for subjects randomized to placebo (Somers’ D = −0.37; asymmetric 95% CI: −0.62 to −0.06; P = 0.01). When including the study drug period, median DASC score was 31.5 (IQR: 18, 56) for subjects randomized to ceftriaxone compared with 53 (IQR: 16, 81) for subjects randomized to placebo (Table 2; Somers’ D = −0.11; asymmetric 95% CI: −0.42 to 0.22; P = 0.52).

Table 2.

Antibiotic exposure and antibiotic resistance gene outcomes in PROTECT

Antibiotic exposure outcomes	Ceftriaxone n = 26	Placebo n = 26	Somers’ D coefficient (95% CI)	P value
DASC score post study drug intervention period, median [IQR]	19.5 [0, 43]	53 [16, 81]	−0.37 (−0.62 to −0.06)	0.01
DASC score including study drug intervention period, median [IQR]	31.5 [18, 56]	53 [16, 81]	−0.11 (−0.42 to 0.22)	0.52
Antibiotic-free days, excluding study drug period, median [IQR]	21 [2, 26]	2 [1, 18]	0.49 (0.16 to 0.72)	<0.01
Days of therapy, excluding study drug period, median [IQR]	3 [0, 7]	6 [2, 11]	−0.33 (−0.59 to −0.01)	0.03

Antibiotic resistance genes (ARGs)a	Ceftriaxonea	Placeboa	Somers’ D coefficient (95% CI)	P value
New ARGs at Day 3 following randomization, median [IQR]	2 [0, 5]	2 [0, 12]	−0.10 (−0.46 to 0.29)	0.63
New ARGs to frequently used antibiotics at Day 3 following randomization, median [IQR]	0 [0, 1]	1 [0, 5]	−0.20 (−0.52 to 0.18)	0.29
New ARGs at Day 7 following randomization, median [IQR]	1 [0, 8]	3 [0, 7]	−0.05 (−0.51 to 0.44)	0.87
New ARGs to frequently used antibiotics at Day 7 following randomization, median [IQR]	0 [0, 2]	1 [0, 4]	−0.21 (−0.62 to 0.31)	0.42

Correlation of DASC with other antibiotic exposure metrics	Greiner’s rho (95% CI)	P value
DASC with DOT	0.95 (0.91 to 0.97)	<0.01
DASC with AFD	−0.15 (−0.43 to 0.15)	0.32
DASC with new ARGs at Day 3	0.06 (−0.33 to 0.43)	0.78
DASC with new ARGs at Day 7	0.27 (−0.29 to 0.68)	0.32

AFD, antibiotic-free days; ARG, antibiotic resistance gene; DASC, days of antibiotic spectrum coverage; DOT, days of therapy.

^aThere were 36 subjects in total with usable ARG data. All included subjects had Day 0 ARG data; however, some subjects had Day 3 and Day 7 ARG data, whereas others had either Day 3 or Day 7 ARG data. There were only 34 subjects (19 in the placebo group and 15 in the ceftriaxone group) with Day 3 ARG data. There were 21 subjects (10 in the placebo group and 11 in the ceftriaxone group) with Day 7 ARG data. ARGs are assessed at the subject level, and one subject can have multiple ARGs; however, statistical testing is performed at the group level.

There were 34 subjects, 15 randomized to ceftriaxone and 19 randomized to placebo, with Day 3 ARG data (Figure S2; Tables S3 and S4). ARGs were determined for each subject and compared statistically at the group level. There were 2 (IQR: 0, 6) new ARGs at 3 days post randomization in subjects randomized to ceftriaxone compared with 2 (IQR: 0, 12) new ARGs in subjects randomized to placebo (Table 2; Somers’ D = −0.10; asymmetric 95% CI: −0.46 to 0.29; P = 0.63). We then considered new ARGs only to antibiotics frequently used in the ICU, defined as β-lactams, macrolides, fluoroquinolones, sulfamethoxazole/trimethoprim and vancomycin (Tables S3 and S4). There were 0 (IQR: 0, 1) new ARGs to frequently used antibiotics in subjects randomized to ceftriaxone at Day 3, compared with 1 (IQR: 0, 5) new ARG in subjects randomized to placebo (Table 2; Somers’ D = −0.20; asymmetric 95% CI: −0.52 to 0.18; P = 0.29).

There were 21 subjects, 11 randomized to ceftriaxone and 10 randomized to placebo, with Day 7 ARG data. There was 1 (IQR: 0, 8) new ARG at 7 days post randomization in subjects randomized to ceftriaxone compared with 3 (IQR: 0, 7) new ARGs in subjects randomized to placebo (Table 2; Somers’ D = −0.05; asymmetric 95% CI: −0.51 to 0.44; P = 0.86). There were 0 (IQR: 0, 2) new ARGs to frequently used antibiotics at Day 7 in subjects randomized to ceftriaxone, compared with 1 (IQR: 0, 4) new ARG in subjects randomized to placebo (Table 2; Somers’ D = −0.21; asymmetric 95% CI: −0.64 to 0.31;, P = 0.42).

There was no correlation between DASC and new ARGs at 3 days post randomization (Greiner’s rho = 0.06; asymmetric 95% CI: −0.33 to 0.43; P = 0.78) or between DASC and new ARGs at 7 days post randomization (Greiner’s rho = 0.27; asymmetric 95% CI: −0.29 to 0.68; P = 0.32). No correlation was noted at Day 3 or Day 7 for new ARGs to frequently used antibiotics either (Table 2).

The median number of AFD was 21 days (IQR: 2, 26) for subjects randomized to ceftriaxone compared with 2 days (IQR: 1, 18) for subjects randomized to placebo after the study drug period (Somers’ D = 0.49; asymmetric 95% CI: 0.16–0.72; P < 0.01). Median DOT was 3 (IQR: 0, 7) for subjects randomized to ceftriaxone compared with 6 (IQR: 2, 11) for subjects randomized to placebo when the study drug period was not included (Somer’s D = −0.33; asymmetric 95% CI: −0.59 to −0.01; P = 0.03). AFD were not correlated with DASC (Greiner’s rho = −0.15; asymmetric 95% CI: −0.43 to 0.15; P = 0.32), but DOT was correlated with DASC (Greiner’s rho = 0.95; asymmetric 95% CI: 0.91–0.97; P < 0.01).

Discussion

Widespread and untargeted antibiotic exposure is a global driver of antibiotic resistance and a major risk for critically ill patients. Quantification of antibiotic exposure is a clinically meaningful outcome often included in trials that test different antibiotic strategies; DASC is a novel metric because it measures both duration of treatment and spectrum of coverage. To our knowledge, no prospective, randomized trial has included an outcome that quantifies antibiotic exposure by considering duration of therapy and spectrum of coverage. In this secondary analysis of data from PROTECT, DASC following the study drug period was lower in patients randomized to ceftriaxone compared with placebo. Despite this, there was no correlation between DASC and presence of new ARGs evaluated at 3 and 7 days post randomization. AFD were lower in subjects randomized to ceftriaxone compared with placebo, but there was no correlation between DASC and AFD. DOT was similarly lower in subjects randomized to ceftriaxone compared with placebo, and DOT was correlated with DASC. Due to the small sample size of PROTECT, our analysis was underpowered to detect statistical differences in any of the tested outcomes.

The most appropriate way to report antibiotic exposure after the administration of antibiotic prophylaxis in critically ill brain-injured patients is unknown. PROTECT reported a 36% reduction in subsequent open-label antibiotic use in OHCA subjects randomized to prophylactic ceftriaxone.²³ The ANTHARTIC trial, in which subjects were randomized to antibiotic prophylaxis with amoxicillin/clavulanate or placebo, reported a 54% reduction in percentage of days of subsequent open-label antibiotic use outside of the trial and while still in the ICU in the intervention group compared with placebo.²¹ The PROPHY-VAP trial reported a 40% increase in AFD in subjects randomized to ceftriaxone compared with placebo (Table 3); notably, this trial excluded OHCA patients, but the critically ill brain-injured cohort is not dissimilar to the PROTECT and ANTHARTIC cohorts.²⁰ Interestingly, each of these trials incorporated a different definition for antibiotic exposure, and none used DOT or DASC. In PROTECT, the number of AFD following the study drug period was higher for subjects randomized to ceftriaxone, and DOT mirrored this trend. AFD were not correlated with DASC; however, DOT was highly correlated with DASC, which is consistent with prior studies.^8,10–16 Despite the different ways in which antibiotic use outcomes were measured, there was an overall trend in PROTECT, ANTHARTIC and PROPHY-VAP supporting antibiotic prophylaxis as a strategy to reduce subsequent open-label antibiotic use broadly in critically ill brain-injured patients.

Table 3.

Comparison of antibiotic exposure reporting in trials of antibiotic prophylaxis in brain-injured patients

	PROTECT	ANTHARTIC	PROPHY-VAP
Intervention	CRO 2 g IV q12h for 3 days	Amoxicillin/clavulanate 1 g/200 mg IV for 2 days	CRO 2 g IV once
Percent open-label antibiotic use	54% versus 85%	N/A	N/A
Percent reduction in open-label antibiotic use	36%	N/A	N/A
Open-label antibiotic reduction RR (95% CI)	0.64 (0.39–0.90)	N/A	N/A
Percent days of antibiotics in ICU	N/A	23% versus 50%	N/A
Percent reduction in days of open-label antibiotics prescribing in ICU	N/A	54%	N/A
Antibiotic-free days	21 versus 2	N/A	21 versus 15
Percent increase in antibiotic-free days	91%	N/A	29%
Days of therapy	3 versus 6	N/A	N/A

CRO, ceftriaxone; RR, relative risk.

Antibiotic prophylaxis must be balanced against the risk of increasing antibiotic resistance through exposure and selective pressure on the resistome.^40–45 Antibiotic prophylaxis is recommended for specific surgical, oncological and ICU populations, as well as high-risk populations with recurrent infections, but routine use in the general population is not advised.^46–48 It is possible that early-onset pneumonia after acute brain injury is common enough that the risks of antibiotic prophylaxis are outweighed by the benefits. This may not be true for other indications where infection is less frequent or the exposure duration is longer, such as those with spontaneous bacterial peritonitis or bite wounds.³⁴

The resistome was not evaluated in ANTHARTIC or PROPHY-VAP, whereas PROTECT provided only a crude overview of resistome changes.^20,21,23 Specifically, PROTECT measured the presence of ARG isolates using metagenomic sequencing, but it did not assess gene expression or identify the organism associated with the ARG isolate.²³ Since ANTHARTIC, PROHPY-VAP and PROTECT consistently demonstrated reductions in post-prophylaxis pneumonia and subsequent antibiotic use, we reasoned that antibiotic administration (quantified by DASC) would be correlated with the presence of newly acquired ARGs in PROTECT. We did not find a correlation between DASC and newly acquired ARGs at either 3 days or 7 days after randomization in either treatment group. However, the ARG assessment was limited by lack of subjects and power; only 34 of 52 (65%) subjects had usable rectal swabs at 3 days post randomization, and only 21 of 52 (40%) subjects had usable rectal swabs at 7 days post randomization.

Whether a short-duration antibiotic prophylaxis strategy can induce minimal changes in the resistome by reducing downstream infection and broad-spectrum antibiotic administration (i.e. increased selection pressure on the resistome) remains an important and yet unanswered question. The consistent direction of the DASC and ARG analyses suggests that antibiotic prophylaxis in this patient population may actually bolster antibiotic stewardship initiatives by reducing subsequent antibiotic exposure and antibiotic resistance while also reducing clinical disease. A future adequately powered study, with accompanying detailed resistome taxonomic analysis as well as transcriptomic analysis to describe expression of ARGs, is needed to provide greater insight into the implications of antibiotic prophylaxis and its impact on antibiotic resistance in critically ill patients. Incorporating an endpoint, such as the DASC score, to measure antibiotic exposure is a logical next step for interventional trials that test different antibiotic strategies.

Limitations

These results lack the statistical power to recommend for or against antibiotic prophylaxis as a strategy to reduce subsequent antibiotic exposure or ARG acquisition in the resistome. In addition to lack of statistical power, the metagenomic analyses are limited by a lack of accompanying identification of the source organism and transcriptomic data. Future studies should assess ARG presence and expression as well as identify the organism associated with the isolate. The ARG assessment may also have been hindered by lack of usable rectal swabs due to an absence or dearth of faecal material collected on the swabs. The protocol did not specify a minimum amount of faecal material to be recovered at each available timepoint. Refined sample collection techniques to maximize the amount of faecal material collected for sequencing analysis could have increased the number of subjects with ARG data. Future studies should ensure robust collection of sample specimens to maximize usable material for metagenomic analyses. PROTECT was a pragmatic study, the protocol did not specify collection of screening cultures; however, sputum, blood and urine cultures were routinely collected as standard of care and no non-pulmonary infections or bacteraemia were noted.²³ Carbapenem-resistant Enterobacterales screening is not routinely performed at the clinical site, nor did we collect data on Candida species; these are additional limitations of the study.

Conclusions

The DASC score is a comprehensive metric that measures antibiotic exposure, and it can be feasibly applied to interventional clinical trials investigating antibiotic prophylaxis strategies. Future studies can test whether interventions that change the DASC score alter the acquisition of new ARGs.

Funding

This work was supported by the National Institute of General Medical Sciences through a Center of Biomedical Research Excellence in Acute Care Research and Rural Disparities award (5P20GM139745-04).

Transparency declarations

None to declare. The authors have no real or perceived financial conflicts of interest to disclose. The funder had no decision-making role in the research. The authors did not use artificial intelligence or a professional medical writer to prepare this manuscript.

Author contributions

Concept and design: D.J.G., K.M.B., A.J.W., R.R.R., T.L.M., D.B.Seder, D.J.D.; acquisition of data: D.J.G., K.M.B., R.R.R., T.L.M., D.B.Seder; analysis and interpretation of data: D.J.G., K.M.B., A.J.W., R.R.R., T.L.M., C.W.C., D.B.Seder; drafting of the manuscript: D.J.G., K.M.B., A.J.W., R.R.R., D.B.Seder; critical revision of the manuscript for important intellectual content: all authors.

Data availability

Additional trial data are available upon reasonable request to the corresponding author.

Supplementary data

File S1, Figures S1 and S2, and Tables S1 to S7 are available as Supplementary data at JAC-AMR Online.