Antibiotic duration for common bacterial infections—a systematic review

Abstract

Background

Reducing antibiotic duration is a key stewardship intervention to mitigate antimicrobial resistance (AMR). We examined current evidence informing antibiotic duration for common bacterial infections to identify any gaps in terms of settings, patient populations and infectious conditions. Trial methodologies were assessed to identify areas for improvement.

Methods

MEDLINE and Embase were searched up to July 2024 for randomized trials comparing antibiotic durations in hospital and community settings (PROSPERO 2021, CRD42021276209). A narrative synthesis of the results was performed with a review on the major guidelines published by IDSA, NICE, WHO and other international societies to assess the impact of these trials on practice guidance.

Results

Out of 315 studies, 85% concluded equivalence or non-inferiority of shorter courses. Adult bacterial sinusitis, community-acquired pneumonia, female cystitis/pyelonephritis, uncomplicated cellulitis and intra-abdominal infection with adequate source control and perioperative prophylaxis had robust evidence supporting shorter durations. Few trials studied severe infections, such as bloodstream infections and ventilator-associated pneumonia. Twenty-three (7%) of the trials were conducted in intensive care settings and only 43 trials (14%) enrolled patients from low-to-middle- or low-income countries. Only 15% of studies were at low risk for bias.

Conclusions

Reducing antibiotic duration likely remains an important strategy for antibiotic stewardship, and an area of active research. While shorter antibiotic courses may be suitable for many bacterial infections, more evidence is needed for severe infections and in low- and middle-income settings.

Introduction

Human consumption of antibiotics is a major driver of antimicrobial resistance (AMR),¹ and reducing unnecessary antibiotic use is a key intervention for reducing AMR. Increasing calls to view antibiotics as a non-renewable resource has given rise to systematic and evidence-based antibiotic stewardship programmes, which are strongly advocated by the WHO and international infectious disease authorities.²

Reducing antibiotic duration is one of the most commonly implemented antibiotic stewardship strategies reported in the literature,³ and the one that is deemed to be safest and most acceptable by practising clinicians.⁴ This includes shortening courses for established bacterial infections, preventing bacterial infections and rapid discontinuation of antibiotic prescriptions when bacterial infections have been excluded.

The conventional principle for antibiotic duration was to treat beyond clinical improvement in order to prevent both relapse of infection and development of antibiotic resistance. The message of always completing antibiotic courses to prevent the development of resistance has remained widespread until recent years, promoted by the WHO, international health authorities, national health campaigns and school curricula.^4,5

However, this led to a parallel concern of excess antibiotic use. The spotlight on this and the emergence of widespread AMR in the late 1990s sparked numerous randomized controlled trials to shorten antibiotic duration. Aside from directly comparing durations, inflammatory markers such as serum C-reactive protein and procalcitonin, and individualized expert opinion such as infectious disease specialist consultations have also been used in trials to inform duration decisions.⁶

This systematic review focuses on all randomized controlled trials evaluating definitive antibiotic duration for common bacterial infections and perioperative prophylaxis in humans and aims to: (i) review current evidence for shorter-course antibiotic durations; (ii) identify gaps in evidence for antibiotic durations in terms of settings, patient populations and infectious conditions; and (iii) appraise trial methodologies and identify areas for improvement.

Methods

Data sources and searches

This systematic review was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines (Supplementary material, available as Supplementary data at JAC-AMR Online). Randomized controlled trials published in the English language and indexed in MEDLINE or Embase were sought.⁷ Unpublished studies and preprints were excluded. This article includes surveillance through to 31 July 2024. This systematic review was registered on PROSPERO (reference: PROSPERO 2021, CRD42021276209).

Study selection

A Boolean search strategy with search terms pertaining to antibiotic treatment, duration, bacterial infections, human health, and randomized controlled trials was adopted (Table S2). Additional relevant titles found in other systematic reviews were also included.

The specific inclusion criteria were: (i) patients from either hospital or community settings who received antibiotics for the prevention or treatment of bacterial infections; (ii) random assignment to trial arms with varying duration of antibiotic treatment (examples of the specific interventions used to guide antibiotic duration include, but are not limited to, the use of inflammatory markers, expert opinion or electronic prescription systems); and (iii) clinical outcomes (e.g. cure, relapse, mortality). The exclusion criteria were: (i) patients with non-infectious conditions, fungal or viral infections, or TB; (ii) intervention involving no antibiotic courses (e.g. comparison between antibiotic courses and a surgical procedure); and (iii) feasibility or pilot studies.

Y.M. and W.C.T. reviewed each abstract for potential full-text review independently. One investigator then reviewed each full-text article for inclusion, and a second verified the decision to exclude. Disagreements were resolved through consensus, with B.S.C. making the final decision.

Data extraction and quality assessment

Full texts for all studies included in the systematic review were retrieved for data extraction. Data extracted from all included studies were: age group; type of bacterial infection; treatment durations; blinding; how duration was determined; and healthcare settings. To provide a more detailed assessment of current methodologies and aspects of trial design and conduct affecting risk of bias, additional data were extracted from trials conducted since 2006 (chosen arbitrarily). These additional data included: study hypothesis and sample size calculation; how the bacterial infection was identified; antibiotic choice; randomization process; follow-up period; non-adherence and how this was monitored; analysis methods; and types of outcomes reported. Randomized trials published from 2006 onwards, and which achieved the target enrolment number, were further assessed according to the key domains outlined in the revised tool for assessing risk of bias in randomized trials (RoB 2).⁸

Y.M. and W.C.T. reviewed the shortlisted articles and extracted the underlying data for the systematic review independently. Both authors verified the extracted data. Conflicts during the screening and data extraction processes were resolved through discussion, with B.S.C. making the final decision.

Data synthesis and analysis

Quantitative synthesis was not possible owing to heterogeneity in types of bacterial infections, study design, patient populations, comparisons and analytic methods. Instead, we performed a narrative synthesis of the results.

In addition, we reviewed major guidelines published by IDSA, NICE, WHO and other national/international societies to assess the impact of the randomized trials on practice guidance.

Results

Overview of antibiotic duration randomized trials

The initial electronic database search produced 4036 unique records published from 1969 to 31 July 2024 (Figure 1). Of these, 315 fulfilled the inclusion criteria. The most frequently studied bacterial infections were upper (52/315; 17%) and lower (66/315; 21%) respiratory tract infections and genitourinary infections (52/315; 17%). Perioperative prophylaxis (52/315; 17%) was also extensively studied (Figure 2). Eighty-five percent (267/315) of the trials concluded that there was either no statistical difference, equivalence or non-inferiority between the short and long antibiotic courses when clinical outcomes were compared.

Most antibiotic treatment duration trials prior to 2004 exclusively tested various arbitrary durations (Figure 3). Biomarkers such as procalcitonin and C-reactive protein were increasingly studied in the 2000s but declined from the mid-2010s. Another type of intervention is the use of stopping rules incorporating individual clinical response.

In terms of patient characteristics, the most frequently represented groups consisted of adults from either the hospital general ward or the community in upper- and upper-middle-income countries (Figure 3). Only 23 (7%) of the trials were conducted in intensive care settings, and 43 trials (14%) enrolled patients from low- and middle- or low-income countries.

Evidence from antibiotic treatment duration randomized trials for specific bacterial infections

Evidence for antibiotic treatment durations for common bacterial infections are discussed in detail below (Figure 4). Specific recommendations from major guidelines are included in Table 1.

Respiratory tract infections

Acute group A Streptococcus (GAS) pharyngotonsillitis

Twenty-eight trials studied antibiotic duration for GAS pharyngotonsillitis. Important outcomes in these trials included both symptom resolution and microbiological eradication to prevent immunological sequelae such as rheumatic fever. A short course of 5–7 days phenoxymethylpenicillin was found to have a lower microbiological cure rate compared with 10 days of treatment [number needed to treat 13 (range 8–37)] in a single systematic review and meta-analysis published in 2008.⁹ Hence, IDSA recommended 10 days of penicillin or amoxicillin,¹⁰ while WHO and NICE recommended 5–10 days depending on the risk of GAS as the cause for pharyngotonsillitis.^11,12

In contrast to most studies involving phenoxymethylpenicillin, which used 10 day courses, shorter courses of 5 days of macrolides (e.g. azithromycin, erythromycin) and late-generation cephalosporins (e.g. cefpodoxime, cefdinir) were studied.^13,14 However, shorter-course cephalosporin studies lacked strict entry criteria, did not assess adherence to antibiotics, and did not differentiate the serotypes or genotypes between infection episodes that failed treatment versus those that were deemed to be newly acquired.¹⁰ In addition, macrolides and late-generation cephalosporins are broader in antimicrobial spectrum than penicillins and more costly. Therefore, 5 days of macrolides is recommended as second-line treatments, and late-generation cephalosporins were not recommended across major guidelines.^10–12

There remain important evidence gaps, especially for evaluation of long-term clinical complications from reduced antibiotic treatment duration. No trial evaluated the effect of an alternative antibiotic to penicillin for preventing acute rheumatic fever. In addition, few trials were performed in low- and middle-income settings, where the prevalence of rheumatic fever is high.¹⁵

Otitis media in children

Thirteen trials studied antibiotic duration in otitis media amongst children. The conventional antibiotic treatment duration adopted in the trials was 10 days, which was compared with shorter durations of 5–7 days. These randomized trials showed varied results.^16–19 Generally, in older children more than 2–5 years old, whose host immunity tends to be mature and the Eustachian tube spontaneously drains, shorter courses or no antibiotics might be needed.^20–22 In a landmark placebo-controlled trial conducted in the USA, children younger than 2 years old were randomized between 10 and 5 days of amoxicillin/clavulanate.¹⁶ Children in the shorter-course arm had more clinical failure (34% versus 16%, difference 17%, 95% CI 9%–25%). In addition, prolonged antibiotics might be indicated for those with recurrent or severe infections.

The trials were heterogeneous in terms of diagnostic criteria, antibiotic choices (which may have different bioavailability) and definitions of symptom resolution.²³ High rates of spontaneous resolution in many of these trials might have suggested that antibiotics might not have been required. Evidence gaps exist for older children older than 2 years, and in low- and middle-income countries, where prevalence of antibiotic resistance is higher.

Acute bacterial sinusitis

Ten trials studied antibiotic duration in acute bacterial sinusitis, performed exclusively in adults. A meta-analysis showed that there was no difference in cure or improvement between shorter-course (3–7 days) and longer-course (6–10 days) antibiotics in adults, and is associated with fewer antibiotic side effects.²⁴ In children, IDSA recommended 10–14 days of antibiotics based on expert opinion as there was no randomized trial evidence,^25,26 while NICE and WHO recommended 5 days only in those who are systematically unwell, with symptoms and signs of more serious infection or at high risk of complications.^12,27

The trials used heterogeneous interventions, which included various adjunctive therapies for symptom relief. Enrolled patients were heterogeneous, due to varying symptom duration prior to antibiotic treatments, non-specific diagnostic criteria and lack of microbiological testing. These may have led to patients with viral sinusitis being enrolled into the trials.

Community-acquired pneumonia

There were 38 randomized trials that studied antibiotic duration in community-acquired pneumonia. Shorter-course antibiotics (5 days) were consistently shown to be non-inferior to longer durations, especially in mild to moderate infections in outpatient settings in children and adults.^28–32 This result included all empirical antibiotic options for community-acquired pneumonia, i.e. macrolides, quinolones and β-lactams.³³ All major guidelines recommended to prolong antibiotics in patients with delayed clinical response and severe disease, e.g. pulmonary or extrapulmonary complications, inadequate procalcitonin response or antibiotic-resistant organisms.^{12,32,34–36} However, no randomized trials were performed that defined antibiotic treatment duration for patients who might require longer-course antibiotics.

A wide variety of antibiotics with different bioavailability were compared within and between these trials. None of the trials compared the same antibiotic at the same daily dosage in the outpatient setting.³⁷ Though there were studies that supported 3–5 day antibiotic courses for treatment in children with mild community-acquired pneumonia,^38–40 lack of chest imaging and microbiological testing might have allowed patients with viral pneumonia or upper respiratory tract infections to be enrolled. There were no randomized controlled trials for radiographically confirmed pneumonia in children.^30,41,42

Ventilator-associated pneumonia

There were six trials that studied antibiotic duration for ventilator-associated pneumonia, which all showed non-inferiority of shorter-course antibiotics (7–8 days) against 14 days.^43–48 Hence, major guidelines recommended 7–8 days for patients with no pulmonary or extrapulmonary complications.^49,50 While a higher rate of recurrence with shorter-course antibiotics was observed in patients with pneumonia caused by non-fermenting Gram-negative rods, such as Pseudomonas aeruginosa, a meta-analysis showed no increased risk of 60 day mortality or recurrence in this group and those associated with carbapenem resistance.^47,48,51,52 Antibiotic duration could be further reduced to individual response to culture-directed antibiotics, as demonstrated by the REGARD-VAP (individualized, short-course antibiotic treatment versus usual long-course treatment for ventilator-associated pneumonia), specifically stable haemodynamics and resolution of fever.⁵²

There were no specific diagnostic criteria used to define ventilator-associated pneumonia, which might have led to patients with no pneumonia being enrolled into the trials. Procalcitonin has been used in randomized trials to define antibiotic durations but it is yet to be routinely recommended by major guidelines.^49,50 A meta-analysis of randomized trials performed in 2018 showed that using procalcitonin to stop antibiotics in ventilator-associated pneumonia reduced treatment duration from 13 to 11 days.⁵³

Bacterial exacerbation of COPD

There were 18 trials that studied antibiotic duration in infective exacerbations of COPD. These trials found that shorter-course antibiotics (less than 6 days) were comparable to longer courses (7 days or more of the same antibiotic) in terms of symptom resolution, and were associated with fewer antibiotic side effects.⁵⁴ Shorter-course antibiotics were comparable to longer courses when evaluated using outcomes at 1, 2 and 3 weeks, and in both inpatient and outpatient settings.⁵⁴

These trials adopted highly heterogeneous diagnostic criteria for COPD, e.g. smoking exposure, airflow obstruction and exacerbation episodes were not all accompanied by microbiological confirmation. This may have led to patients with bronchiectasis and chronic asthma being enrolled into the trials.

Genitourinary tract infections

Simple cystitis in women

There were 22 trials that studied antibiotic duration in female simple cystitis. Shorter-course antibiotics, e.g. 5 days of nitrofurantoin or 3 days of trimethoprim/sulfamethoxazole, were supported by numerous randomized trials across non-pregnant women of different age groups. In pregnant women, the consensus across major guidelines was for 5–7 days,^12,55–57 in view of potentially severe complications in this group, e.g. upper tract involvement and negative consequences for the neonate, and higher chances of microbiological eradication with a longer antibiotic course.⁵⁶ None of the antibiotic-duration randomized trials for urinary tract infection enrolled pregnant women.

Increasing prevalence of AMR (e.g. trimethoprim/sulfamethoxazole resistance, ESBL production) among common bacteria that cause urinary tract infections were important considerations in treatment guideline recommendations.^12,55

Complicated urinary tract infections (in males, catheter-associated, upper urinary tract involvement)

There were 25 trials that addressed antibiotic duration in complicated urinary tract infections. Urinary tract infections in males are considered to be at higher risk of complications than in females due to the potential development of outlet obstructions, upper tract involvement and progression to bloodstream infections.⁵⁸ One randomized trial concluded that 7 days of ciprofloxacin or trimethoprim/sulfamethoxazole was non-inferior to 14 days (symptom resolution 93% in the 7 day group versus 90% in the 14 day group, difference 3%, lower limit of one-sided 97.5% CI 5%).⁵⁹ However, enrolled patients were afebrile and did not have microbiological confirmation from urine samples.⁵⁹ Another recent study compared 7 days or continued ofloxacin for 14 days in men with febrile urinary tract infections. This double-blinded, placebo-controlled study failed to demonstrate non-inferiority of the shorter-course treatment [56% in the 7 day group versus 78% in the 14 day group met treatment success, risk difference −21.9 (95% CI −33.3 to −10.1)].⁶⁰ This trial was underpowered as treatment failure was lower than expected.

As for catheter-associated urinary tract infections, one randomized trial found 5 days of levofloxacin was non-inferior to 10 days of ciprofloxacin in complicated urinary tract infections (72% of the total patients enrolled, number of catheter-associated urinary tract infections not reported) and acute pyelonephritis (29% of the total patients enrolled) with a non-inferiority margin of 15% (91% in the 5 day levofloxacin group versus 87% in the 10 day ciprofloxacin group, 95% CI −10% to 1%).⁶¹ In the second trial, which studied exclusively catheter-associated urinary tract infections in patients with spinal cord injury, 5 day antibiotic with catheter exchange was compared with 10 day antibiotic with catheter retention.⁶² All patients achieved clinical cure at the end of therapy, but more patients in the shorter-course group had recurrent urinary tract infections than in the long-course group (32% versus 11%, P = 0.043). Catheter change prior to antibiotic administration, which may reduce the need for prolonged antibiotics, was studied in a single randomized trial involving 54 older adults in a long-stay care facility.⁶³ This study found a significant difference in cure or improvement, favouring catheter change at 72 h (93% versus 41%), 28 days (89% versus 59%), but not at 7 days.⁶³ Given this limited, low-quality evidence for antibiotic treatment of catheter-associated urinary tract infections, the guidelines recommended 7–14 days depending on clinical response and antibiotic choice, based on expert opinion.^64,65

For females with pyelonephritis, 7 days of ciprofloxacin was found to be superior to 14 days of trimethoprim/sulfamethoxazole in terms of bacteriological and clinical cure rates.⁶⁶ As there were no randomized trials for shorter-course β-lactams or trimethoprim/sulfamethoxazole, a course of 7–14 days was recommended in the guidelines.^12,56,64,65

No trials evaluated antibiotic duration in microbiologically confirmed male urinary tract infections. The role of catheter exchange or removal remains unclear in catheter-associated urinary tract infections. Acute complicated urinary tract infections included in the trials to date were highly heterogeneous due to the presence of nephrolithiasis, tumours or implants, various sites of infection within the urinary tract and infection severity.

Bone and joint infections

Adult osteomyelitis/septic arthritis and prosthetic joint infection

Eleven trials evaluated antibiotic duration for osteomyelitis, septic arthritis or prosthetic joint infection in adults. Four studies focused on prosthetic joint infections, four on diabetic foot osteomyelitis, two on native joint septic arthritis and one on vertebral osteomyelitis. Interventions varied in terms of surgical debridement and one- or two-staged exchange arthroplasty. In the largest trial, which included patients with microbiologically confirmed prosthetic joint infection, 6 weeks was not non-inferior to 12 weeks of antibiotic therapy (persistent infection 18% in the 6 week group versus 9% in the 12 week group, risk difference 9%, 95% CI 2%–16%) within 2 years follow-up. Outcomes in the shorter-course group were worse, and mainly accounted for by those patients who had implant retention (31% versus 15%).⁶⁷ Due to limited, low-quality evidence, there was lack of consensus in the guidelines.^12,68–74 and expert opinion was to prolong antibiotic treatment for up to 12 weeks if implants were retained.⁷⁵ In addition, the general consensus for retained infected implants was life-long antibiotic suppression therapy.⁷⁶

In diabetic foot osteomyelitis, 3 weeks of antibiotic therapy post-debridement was shown to be non-inferior to 6 weeks in terms of clinical remission with a 25% margin (84% in the 3 week group versus 73% in the 6 week group) at 2 month follow-up.⁷⁷ Two trials found antibiotics could be stopped early in those with mild infection (no need for surgical debridement) and without severe peripheral arterial disease.^78,79 In one, exact antibiotic durations administered were not reported.⁷⁸ In the second, 6 weeks was comparable to 12 weeks of antibiotics in terms of remission (12/20 versus 14/20, respectively, P = 0.50). For vertebral osteomyelitis, 6 weeks was non-inferior to 12 weeks of antibiotics for clinical cure (160/176; 91% in the 6 week group versus 159/175; 91% in the 12 week group, difference 0.05%, 95% CI −6.2% to 6.3%).⁸⁰

A single randomized trial studied antibiotic duration in native joint septic arthritis post-surgical drainage, where the majority (99/154; 64%) involved hand and wrist joints and the commonest organism was MSSA and Streptococcus spp.⁸¹ Infection recurred in one patient in the 2 week arm and two in the 4 week arm (97% cure rate) within a median 2 years of follow-up.

Bone and joint infections are highly heterogeneous in terms of joints involved, infection severity and complications, time from symptom onset to surgical debridement, and bacterial pathogens. This made defining inclusion and exclusion criteria challenging. Interventions applied in these trials were also varied, e.g. single versus two-stage arthroplasty, removal versus retention of implants, requirement for surgical debridement, and different antibiotic regimens with variable bioavailability.⁸²

Childhood osteomyelitis/septic arthritis

Two trials studied antibiotic duration for osteomyelitis or septic arthritis in children. Both were conducted in Finland. The primary outcomes were clinical cure and recurrence. Patients with acute osteomyelitis caused primarily by MSSA were given clindamycin or a first-generation cephalosporin for 10 days versus 30 days (first trial) or 20 days versus 30 days (second trial) after an initial IV treatment of 2–4 days in both groups.⁸³ Antibiotics were discontinued with clinical symptom resolution and serum C-reactive protein normalization. Both studies showed comparable outcomes in shorter- versus longer-course groups.

Due to limited randomized trial evidence, there is no current consensus on antibiotic duration for childhood osteomyelitis or septic arthritis.^72–74 The major guidelines generally recommended a short course of IV antibiotic treatment followed by an oral antimicrobial therapy for 2–3 weeks in septic arthritis and 3 weeks in osteomyelitis.⁸⁴ However, for MRSA or Panton–Valentine leucocidin-producing Staphylococcus aureus, 4–6 weeks of treatment are recommended.^72,83

Most trials that evaluated antibiotic duration for childhood bone and joint infections were underpowered.⁸⁵ These trial results may have limited generalizability as the studies mainly enrolled mild cases with minimal complications, with infections predominantly caused by bacterial pathogens that responded quickly to antibiotic therapy e.g. MSSA.⁸⁶

Bloodstream infections

Five trials evaluated antibiotic duration in Gram-negative bloodstream infections. Three trials were performed in adults, and two in neonates. In adult uncomplicated Gram-negative bloodstream infections, 7 days of treatment was non-inferior to 14 days of antibiotic treatment in terms of clinical failure using a 10% non-inferiority margin.^87,88 The interpretation of one trial was limited by very low primary outcome rate (5% 30 day clinical failure) compared with the 10% non-inferiority margin.⁸⁸ As for the other trial, its results could only be applied to Enterobacterales bloodstream infections (over 90% of patients enrolled), and in patients whose infection sources were controlled.⁸⁷ In neonates, 10 days was comparable to 14 days if the infant clinically improved and C-reactive protein normalized.^89,90 However, the findings might not be generalizable as both studies were conducted in single centres in India and sample sizes were small (202 patients in total). There was no randomized trial evidence to guide antibiotic duration for MDR Gram-negative bloodstream infections or for immunocompromised patients.

There was one randomized trial that studied antibiotic duration in S. aureus bloodstream infection. In this trial, the majority of the patients enrolled had CoNS bloodstream infections.⁹¹ The current guidelines recommended a conservative approach of at least 2 weeks of IV antibiotics, with extension to more than 4 weeks if clinical or microbiological response was delayed, in the presence of prothesis or secondary sites of dissemination.⁹²

In bloodstream infections complicated by infective endocarditis, international guidelines typically recommended antibiotic duration of up to 6 weeks. This was established empirically with no randomized controlled trial data.^93–96

Non-tuberculous meningitis

Seven trials evaluated antibiotic duration for non-tuberculous meningitis in neonates and children. Antibiotic durations vary according to bacterial pathogen, and should be determined according to clinical response and microbiological clearance (Table 1). Although a 2009 meta-analysis found that 4–7 days was similar to 7–14 days in terms of clinical cure rate, it also highlighted important methodological issues including lack of blinding, small sample sizes, relatively short follow-up and, most importantly, high variability in the causative pathogens.⁹⁷ None of the trials evaluated antibiotic duration in adult meningitis.

Cellulitis and skin abscesses

There were 10 trials that studied antibiotic duration for cellulitis and skin abscess. In uncomplicated cellulitis, four trials showed that 5–6 days of oral antibiotics was adequate, if clinical response was observed at the end of treatment.^98–101 Single doses of dalbavancin,¹⁰² azithromycin¹⁰³ or oritavancin were also non-inferior to longer courses, including for skin and soft tissue infections caused by MRSA.¹⁰⁴ However, in these trials, definition of clinical response was poorly defined and was at the discretion of the treating physician.¹⁰⁵ There were no trials that compared antibiotic duration for complicated soft tissue infections, e.g. recurrent cellulitis or abscesses.

Intra-abdominal infections

Nine trials studied antibiotic duration in intraabdominal infections. Generally, about 4 days was comparable to longer courses given adequate surgical control of infection source.^106,107 The landmark trial that established this compared a 4 day course versus a maximum 10 day course, which found comparable surgical site infection, recurrent intra-abdominal infection, or death (56/257; 22% in the 4 day group versus 58/260; 22%, difference −0.5%, 95% CI −7.0 to 8.0, P = 0.92).¹⁰⁷

There were no trials that compared antibiotic duration for intra-abdominal infections with inadequate source control, and deep organ abscesses. However, adequacy of source control, infection sites, bacterial pathogens and antibiotic resistance patterns are highly heterogeneous and challenging to define.

Perioperative prophylaxis

Perioperative prophylaxis is the most well-studied indication, with 52 antibiotic duration trials. Generally, a single dose or repeated dosing for less than 24 h was found to be adequate, across different surgical procedures. This was associated with comparable efficacy at preventing surgical site infections and fewer antibiotic side effects.^108–114 For some procedures in a clean surgical field without prosthesis insertion, such as head and neck operations and orthopaedic procedures, IDSA did not recommend any antibiotic prophylaxis.^{12,71,115–117}

Evidence gaps remain in defining redosing frequencies, especially in certain specialized types of surgery, e.g. cardiothoracic surgery, where surgery may be complicated and prolonged.¹¹⁵

Quality of trial design, conduct and analysis

Bias assessment

Bias assessment with the RoB 2 tool was performed for 80 trials (25%), which were published from 2016 onwards and were adequately powered. Twenty-seven (34%), 41 (51%) and 12 (15%) trials were assessed to be at high risk, with some concerns, and low risk of bias, respectively (Supplementary material). The studies classified as having the highest risk of bias were those that were unblinded, had a high degree of non-adherence or crossovers (Figure S1) and those where the analysis failed to account adequately for non-adherence. Ninety-five (30%) of the trials were double-blinded.

Out of 89 non-inferiority or equivalence trials published since 2006, 62 trials reported on the number of adherent participants. The majority of these trials had non-adherence (90%; 56/62), with a median of 11% non-adherent participants (IQR 5%–19%). Only 17 trials reported both the prespecified and actual treatment durations observed in the respective randomization arms; the difference between the actual and the intended treatment durations stated in the protocol ranged from −3 to 6 days (mean of 0 days).

Most of the trials published since 2006 reported a clinical outcome (97%; 143/147). The overall mortality reported in the trials was low (median 0%, IQR 0%–2%). Sixty-seven trials (46%; 67/147) obtained follow-up samples from the participants to check for clearance or newly acquired antibiotic-resistant bacteria at the sites of infection. Only 17 trials (12%) took surveillance cultures during follow-up to assess for emergence of resistant bacteria colonization, most commonly in the upper respiratory, urinary or gastrointestinal tracts. Additionally, 43% (63/147) of the trials reported economic outcomes, including lengths of stay, direct costs (of antibiotics) or indirect costs (consumables and manpower to deliver antibiotics).

Non-inferiority margin

Of the 86 non-inferiority trials, 76 reported a non-inferiority margin (88%). The mean non-inferiority margin was 10%, ranging from 0.35% to 25% (Figure S2).

Discussion

In this systematic review of antibiotic duration trials published to date, most trials (87%; 274/315) successfully concluded that shorter antibiotic courses had been comparable or non-inferior to longer courses. These trials mostly studied infections associated with low mortality, such as respiratory tract and genitourinary infections, and perioperative prophylaxis. Patients from low- and middle-income countries, critically ill patients and children were under-represented. Using arbitrarily determined duration was the most common trial intervention throughout the years. Non-adherence to allocated antibiotic duration was prevalent and involved a median proportion of 11% in each trial. Few trials took surveillance cultures to assess for resistant bacteria colonization with various antibiotic durations.

Some infections had high-quality evidence to shorten antibiotic duration, such as community-acquired pneumonia and uncomplicated urinary tract infections. However, there are also conditions in which practice guidelines have continued to support long treatment duration despite numerous randomized trials showing non-inferiority of a short course in terms of clinical outcomes. These include infections that require longer follow-up (e.g. immunological sequelae of GAS infections), those in low- and middle-income countries where antimicrobial resistance to first-line empirical antibiotics is prevalent (e.g. otitis media in children), and highly heterogeneous infections for which trials with broad inclusion criteria were not informative but where, if inclusion criteria are too specific, participant accruement becomes challenging (e.g. bacterial meningitis, prosthetic joint infections). There remain few randomized trials defining antibiotic duration for relatively common conditions such as Gram-negative and S. aureus bloodstream infections and ventilator-associated pneumonia caused by non-fermenting Gram-negative bacilli.

There is limited evidence to show the effectiveness of shortening antibiotic duration in reducing AMR. The effect of antibiotic duration on the development and spread of AMR is expected to vary due to multiple pathogens, host and environmental factors. The penalties of prolonged antibiotic courses are the side effects and associated costs for both treated individuals and, through increased selection for resistance, the population. Direct side effects of antibiotics for the individuals include allergies, kidney and liver injuries, and opportunistic infections such as Clostridioides difficile colitis. Though mostly mild, these side effects are not infrequent and have been reported in up to 20% of hospitalized patients who received antibiotics.^118,119 While there are sound theoretical reasons for expecting reduced antibiotic duration to often lead to reduced AMR, surprisingly little empirical evidence has been collected to explore this relationship. As noted in this review, most antibiotic duration trials investigated the short-term clinical outcomes of reduced antibiotic duration on individual patients, but only nine followed up treated patients to determine subsequent AMR colonization. This is echoed by earlier reviews of studies attempting to quantify the impact of stewardship strategies, which found that microbiology outcomes are seldom reported, especially those from longer-term follow-up.³

Non-inferiority was the most commonly adopted design for antibiotic duration trials. Non-inferiority trials compare the short with the standard treatment duration against a margin that the investigators are willing to sacrifice in terms of treatment effects, to reap benefits such as reduced resistance, cost savings or reduced side effects. The non-inferiority margin is usually defined with prior knowledge of the standard-of-care antibiotic duration’s efficacy compared with a placebo, and consensus from subject-matter experts.¹²⁰ However, prior placebo-controlled trials to inform the choice of the non-inferiority margin are frequently lacking, and the efficacy of standard-of-care antibiotic duration may be increasing with time due to improvements in general healthcare delivery.¹²¹ This systematic review showed that 86% of the antibiotic duration non-inferiority trials published in the last 18 years reported a low event rate of less than 10%. A lower-than-expected event rate can lead to insufficient power, favouring non-inferiority.¹²²

Conclusions

Reducing antibiotic duration is likely to remain an important strategy for antibiotic stewardship, and an area of active research. Innovative and robust trial designs have been proposed to define duration–response relationships, rather than relying on arbitrary durations compared with non-inferiority hypotheses.¹²³ Larger randomized trials will allow for meaningful conclusions for patient subsets with varying host and pathogen characteristics. More evidence to define treatment duration is needed for severe bacterial infections and in low- and middle-income settings.

Acknowledgements

Y.M. and B.S.C. conceptualized the study. Y.M. and W.C.T. performed the systematic review independently and cross-checked data extracted. Y.M. drafted the first version of this manuscript. All authors reviewed and edited the manuscript. We acknowledge Professor Sarah Walker and Professor Marc Bonten for their helpful insights in refining this manuscript.

Funding

This work was supported by the Wellcome Trust [106698/Z/14/Z to Mahidol Oxford Tropical Medicine Research Programme], Singapore National Medical Research Council [NMRC/Fellowship/0051/2017 to M.Y.] and the UK Medical Research Council [MR/V028456/1 to B.S.C.]. For the purpose of Open Access, the author has applied a CC BY public copyright licence to any Author Accepted Manuscript version arising from this submission. The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Transparency declarations

None to declare.

Data availability

Data extracted from all included antibiotic treatment duration randomized controlled trials are available at https://github.com/moyinNUHS/abxdur-review.

Supplementary data

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

References