Effect of Oral Prednisolone on Symptom Duration and Severity in Nonasthmatic Adults With Acute Lower Respiratory Tract Infection: A Randomized Clinical Trial

Key Points

Question  Does a moderate dose of oral corticosteroid reduce the duration or severity of acute lower respiratory tract infection in adults without asthma presenting to primary care?

Findings  In this randomized trial of 401 adults with symptoms of acute lower respiratory tract infection, treatment with oral prednisolone, 40 mg/d for 5 days, compared with placebo did not significantly reduce the median duration of moderately bad or worse cough (5 days in each group) or the mean severity of symptoms between days 2 and 4 (1.99 vs 2.16 points out of 6).

Meaning  These findings do not support the use of oral steroids for the treatment of acute lower respiratory tract infection in the absence of asthma.

Importance  Acute lower respiratory tract infection is common and often treated inappropriately in primary care with antibiotics. Corticosteroids are increasingly used but without sufficient evidence.

Objective  To assess the effects of oral corticosteroids for acute lower respiratory tract infection in adults without asthma.

Design, Setting, and Participants  Multicenter, placebo-controlled, randomized trial (July 2013 to final follow-up October 2014) conducted in 54 family practices in England among 401 adults with acute cough and at least 1 lower respiratory tract symptom not requiring immediate antibiotic treatment and with no history of chronic pulmonary disease or use of asthma medication in the past 5 years.

Interventions  Two 20-mg prednisolone tablets (n = 199) or matched placebo (n = 202) once daily for 5 days.

Main Outcomes and Measures  The primary outcomes were duration of moderately bad or worse cough (0 to 28 days; minimal clinically important difference, 3.79 days) and mean severity of symptoms on days 2 to 4 (scored from 0 [not affected] to 6 [as bad as it could be]; minimal clinically important difference, 1.66 units). Secondary outcomes were duration and severity of acute lower respiratory tract infection symptoms, duration of abnormal peak flow, antibiotic use, and adverse events.

Results  Among 401 randomized patients, 2 withdrew immediately after randomization, and 1 duplicate patient was identified. Among the 398 patients with baseline data (mean age, 47 [SD, 16.0] years; 63% women; 17% smokers; 77% phlegm; 70% shortness of breath; 47% wheezing; 46% chest pain; 42% abnormal peak flow), 334 (84%) provided cough duration and 369 (93%) symptom severity data. Median cough duration was 5 days (interquartile range [IQR], 3-8 days) in the prednisolone group and 5 days (IQR, 3-10 days) in the placebo group (adjusted hazard ratio, 1.11; 95% CI, 0.89-1.39; P = .36 at an α = .05). Mean symptom severity was 1.99 points in the prednisolone group and 2.16 points in the placebo group (adjusted difference, −0.20; 95% CI, −0.40 to 0.00; P = .05 at an α = .001). No significant treatment effects were observed for duration or severity of other acute lower respiratory tract infection symptoms, duration of abnormal peak flow, antibiotic use, or nonserious adverse events. There were no serious adverse events.

Conclusions and Relevance  Oral corticosteroids should not be used for acute lower respiratory tract infection symptoms in adults without asthma because they do not reduce symptom duration or severity.

Trial Registration  ISRCTN.com Identifier: ISRCTN57309858

Quiz Ref IDAcute lower respiratory tract infection, defined as an acute cough with at least 1 of the symptoms of sputum, chest pain, shortness of breath, and wheeze,¹ is one of the most common conditions managed in primary care internationally. In 2009-2011, an estimated 65% to 75%^2,3 of patients were prescribed antibiotics despite good evidence that they do not reduce symptom duration or severity⁴ and guidelines to the contrary.¹ Annual antibiotic prescribing costs are estimated at US $726 million in the United States⁵ and US $300 million for consultations and antibiotics in the United Kingdom.⁶

Antimicrobial resistance is one of the greatest challenges to modern public health.⁷ Primary care is responsible for 80% of health service antibiotic prescribing,^2,8 with a high proportion regarded as unnecessary² and contributing to antimicrobial resistance.⁹ Both US¹⁰ and UK¹¹ national antimicrobial resistance action plans recommend finding alternatives to antibiotics, but none is currently proven for acute lower respiratory tract infection in adults.

Quiz Ref IDSymptoms of acute lower respiratory tract infection are similar to those of exacerbated asthma.¹² Bronchial epithelial changes are similar in people with and without asthma during a respiratory tract infection, with both groups showing reductions in forced expiratory volume and airways inflammation,¹² and prolonged acute lower respiratory tract infection symptoms are thought to be due to bronchial hyperresponsiveness.¹³ Oral and inhaled corticosteroids are highly effective for acute asthma, but US, British, and European guidelines do not provide guidance on whether corticosteroids should be used for acute lower respiratory tract infection. Despite this, US and European clinicians are increasingly using oral and inhaled steroids, with 1 US study¹⁴ reporting oral prednisolone use in 15% of adults without asthma with acute lower respiratory tract infection.

A previous systematic review¹⁵ found insufficient evidence regarding the role of inhaled corticosteroids and found no oral corticosteroid studies for acute lower respiratory tract infection. The aim of the current study was to investigate the effects of a moderate dose of oral corticosteroids in adults without asthma presenting to primary care with acute lower respiratory tract infection.

Ethical approval for this study was granted by the Central Bristol Research Ethics Committee (12/SW/0180) and all patients gave written informed consent. The Oral Steroids for Acute Cough (OSAC) trial was a multicenter, placebo-controlled, individually randomized study conducted between July 2013 and October 2014. Quiz Ref IDFamily physicians and nurses (recruiting clinicians) were trained in study procedures by 4 centers at the Universities of Bristol, Southampton, Nottingham, and Oxford. They were asked to assess eligibility in consecutive patients: age 18 years or older and presenting for an acute (≤28 days) cough as the main symptom with at least 1 lower respiratory tract symptom (phlegm, chest pain, wheezing, or shortness of breath) in the previous 24 hours. Patients were excluded if they were clinically suspected to have or their medical records showed evidence of chronic pulmonary disease; had received any asthma medication in the past 5 years; met National Institute for Health and Clinical Excellence criteria for severe infection/complications¹; required same-day hospital admission; or required same day antibiotics (see Supplement 1 for full list). Participants were recruited on the day of or the day following presentation. Following consent, demographic and clinical data were collected, including self-reported ethnicity using UK-approved¹⁶ categories, to assess sample representativeness.

The treatment allocation schedule was computer generated by a statistician independent of the trial team. Randomization to prednisolone or placebo in a 1:1 ratio used a variable block size (4, 6, 8, and 10) and was stratified by center. Allocated medication was added to numbered participant packs by pharmacists independent of the team. All packs were identical and centers distributed 4 packs at a time to family practices. Following eligibility confirmation, participants were given the next pack.

Participant packs contained either ten 20-mg oral prednisolone tablets (Galen Pharma GmbH) or placebo tablets matched on dimension, appearance, and taste (Piramal Healthcare Ltd). Participants were asked to take 2 tablets once daily for 5 days, starting on the day of consultation, if possible before starting any antibiotics (if receiving a “delayed” prescription). The dose and duration of prednisolone was selected to reflect the dose and duration known to be effective for acute asthma.¹⁷ Participants, recruiting clinicians, and the trial team were masked to treatment allocation until data analyses were complete.

Participants were invited to report (using web or paper versions) the presence and severity of symptoms using a validated¹⁸ diary shown to be sensitive to change.^4,19 Symptoms were measured using a scale from 0 (no problem) to 3 (moderately bad) and up to 6 (as bad as it could be). All symptoms were measured daily, with twice-daily peak expiratory flow, for 28 days or until symptom resolution. Cough was measured for a further 28 days in case of late treatment effects. A research nurse telephoned participants weekly to support symptom diary completion. Participants were given £5 (US $6.60) shopping vouchers at 14 and 28 days. Medical notes were reviewed at 3 months for new diagnoses of asthma, chronic obstructive pulmonary disease, whooping cough, and lung cancer.

Two primary outcomes were assessed. The first was duration of moderately bad or worse cough, defined as the number of days from randomization to the last day with a score of at least 3 points prior to at least 2 consecutive days with a score of less than 3, up to a maximum of 28 days. This was regarded as the more important of the 2 primary outcomes because cough was the main presenting symptom of the illness, and it included measures of both duration and severity. The second primary outcome was the mean severity score (range, 0-6) of the 6 main symptoms (cough, phlegm, shortness of breath, sleep disturbance, feeling generally unwell, and activity disturbance) on days 2 to 4; the mean score was calculated across the symptoms for each day and then an overall mean was calculated, with a maximum value of 6.

Secondary outcomes specified a priori were total duration and severity of each symptom up to 28 days (cough, phlegm, shortness of breath, wheeze, blocked/runny nose, chest pain, fever, muscle aching, headache, sleep disturbance, feeling generally unwell, activity disturbance), duration of moderately bad or worse and any cough up to 56 days, duration of abnormal peak flow, antibiotic use, adverse events, reconsultation with evidence of illness deterioration, patient satisfaction with treatment, and intention to use the same treatment if it were to be available in the future (more detail about the derivation of these outcomes is provided in Supplement 1). Quality of life, National Health Service treatment, and investigation costs are not reported in this article.

Prespecified potential treatment effect modifiers were age; prior cough duration; presence of wheeze; antibiotic use; β-agonist use; smoking status; history of hay fever, asthma, or eczema; and new diagnoses (at 3 months) of asthma, chronic obstructive pulmonary disease, whooping cough, or lung cancer. Baseline impression of severity of illness was added as a post hoc subgroup analysis because the investigators determined it was important to differentiate between participants with severe vs mild symptoms.

The distributions of both primary outcomes were expected to be positively skewed; hence, sample size calculations were based on the log-normal distribution. The mean durations of moderately bad or worse cough and symptoms severity score (days 2-4) were estimated to be 5.8 (SD, 4.1) days and 2.3 (SD, 1.1) points, respectively.¹⁹ This corresponds to 1.56 (SD, 0.64) log days (or a geometric mean of 4.74 days) for cough duration and 0.73 (SD, 0.45) points on the log scale (or a geometric mean of 2.08) for severity of symptoms. Because there were no previous studies of oral steroids to inform the minimum clinically important difference in both outcomes, the investigative team considered the balance of potential benefits and adverse effects and reached a minimum clinically important difference consensus of 20%, corresponding to a geometric mean in the active treatment group of 3.79 days (mean, 1.33 log days) for duration of cough and 1.66 points (mean, 0.51 log points) for severity of symptoms. Allowing for 20% attrition, 218 participants needed to be randomized per group to retain 174 at follow-up and achieve 90% power with a 2-sided α = .05 for primary outcome one. A final achieved sample size of 174 participants per group would provide 89% power to detect a 20% reduction in severity of symptoms, with an adjusted 2-sided α = .001 to reflect the second primary outcome status (Supplement 1).

A prespecified analysis plan was approved by the trial steering and data monitoring committees and the study protocol was published before data collection had finished (Supplement 2). Changes that were made to the statistical analysis plan after the analysis was begun are also described in Supplement 2. All analyses were performed in Stata software, version 13.1.²⁰

The primary comparative analyses considered patients in the groups to which they were randomized, without imputation for missing outcome data. These analyses were adjusted for center (Bristol, Nottingham, Oxford, and Southampton) and the relevant baseline measure (prior cough duration [1-28 days] for duration of moderately bad or worse cough and patient-reported illness severity in last 24 hours for severity of symptoms [0 = completely well; 10 = extremely unwell]). Time-to-event methods were used to analyze the duration of moderately bad or worse cough. Semiparametric Cox proportional hazard models were used (to enable comparison with previous studies) and the assumption of proportional hazards checked by visual inspection of log-log survival curves and calculation of Schoenfeld residuals.²¹ Hazard ratios were reported comparing the instantaneous rate of resolution of cough between the prednisolone and placebo groups, with 95% confidence intervals and P values. To assist interpretation against the minimum clinically important difference of a 20% reduction in time to resolution, for which hazard ratios are unhelpful, parametric Weibull accelerated failure time models were used to present cough duration treatment effects as time ratios. Such models can be formulated as proportional hazards or accelerated failure time models; hence, hazard ratios were also produced from the Weibull models to ensure comparability with the Cox models.

To further aid interpretation, we calculated absolute measures of effect for the primary outcome of duration of moderately bad or worse cough. There is no single absolute measure of treatment effect for time-to-event data because it varies over the duration of follow-up; it can, however, be calculated at a specific time point. Of particular clinical interest is day 7, because this is a time in the illness trajectory when clinicians and patients want to know about expected benefits, and when steroids should have affected symptoms if effective. Survival curves were produced from the Cox regressions at given values of center (Bristol) and duration of prior cough (median value). Predicted survival probabilities at day 7 in the prednisolone and placebo groups were obtained and an absolute risk difference estimated as the survival probability in the prednisolone group minus that in the placebo group. Ninety-five percent confidence intervals were obtained using the method proposed by Altman and Andersen.²²

Mean severity score from days 2 to 4 was considered in linear regression models. Models considered mean severity score and log mean severity score, and distributional checks of residuals were undertaken to determine the most appropriate model. Differences between the prednisolone and placebo groups are reported with 95% confidence intervals and P values.

For both primary outcomes, secondary analyses additionally adjusted for factors demonstrating imbalance at baseline (difference >5% for binary and >0.5 SD for continuous outcomes) and for smoking because it is known to be prognostically important.¹⁵

Analyses of secondary outcomes used regression models as appropriate. Consideration of potential effect modifiers used formal tests of interaction. We calculated absolute measures of effect for time-to-event secondary outcomes as described above for the primary outcome. Absolute risk differences were also obtained (with 95% confidence intervals) for the binary secondary outcomes of antibiotic use (up to 7 and 28 days), patient satisfaction, and intention to use the same treatment if it were to be available. Sensitivity analyses considered multiple imputation of missing data (using a 2-fold fully conditional specification algorithm),^23,24 treatment adherence, day of recruitment, and inclusion of those with no moderately bad or worse cough at baseline (post hoc) (see Supplement 1 for details).

Fifty-eight family physicians and 50 practice nurses based in 54 family practices assessed 525 patients for suitability, of whom 401 were eligible, consented, and were randomized, 199 to prednisolone and 202 to placebo (Figure 1), equating a mean patient recruitment rate of 0.5 patients per month per practice. Two placebo group patients requested complete withdrawal immediately after randomization, and a duplicate patient was subsequently identified in the prednisolone group (the participant remained in the group to which he/she was first allocated), leaving a sample of 398. Enrollment was stopped when the required number of participants was achieved. At baseline, participants had a mean age of 47.4 (SD, 16.0) years; 37% were men; 3.5% had diabetes; 17% were currently smoking; 5% had received asthma medication more than 5 years previously; 77% reported phlegm, 46% chest pain, 47% wheezing, and 70% shortness of breath; and 42% had abnormal (defined as <80% expected) peak flow. Baseline characteristics were similar between the groups with respect to deprivation, smoking status, weight, height, and clinical characteristics of the acute lower respiratory tract infection, although compared with placebo, the prednisolone group was slightly more likely to be male, be older (and hence retired), and have received an influenza vaccine in the last 12 months (Table 1).

Symptom diaries were returned by 374 participants (94%; 192 in the prednisolone group and 182 in the placebo group). For duration of moderately bad or worse cough, data were available for 334 participants (84%), with 40 reporting an initial cough severity of less than 3 points (that is, not moderately bad or worse) and 24 lost to follow-up. For severity of symptoms, follow-up data were available in 370 (93%). However, 1 participant in the prednisolone group had no baseline measure of illness severity and this participant’s data could not be used in the adjusted analysis. Patients who withdrew or were lost to follow-up were younger (median, 30 years vs 49 years), were less likely to be white (85% vs 97%), were more likely to be employed (86% vs 69%), and had a higher English Index of Multiple Deprivation score (median score, 18 vs 11).

The median duration of moderately bad or worse cough was 5 days (interquartile range, 3-8 days) in the prednisolone group and 5 days (interquartile range, 3-10 days) in the placebo group (Table 2). Kaplan-Meier survival curves were similar for both groups (Figure 2). Visual inspection of the log-log survival curves and calculation of the Schoenfeld residuals (P = .52) provided no evidence against proportional hazards. Comparing prednisolone with placebo, the Cox model adjusting for center and baseline cough duration resulted in a hazard ratio of 1.11 (95% CI, 0.89-1.39; P = .36 with α = .05). The hazard ratio represents the instantaneous risk of resolution from moderately bad or worse cough in the prednisolone group compared with placebo; a hazard ratio greater than 1 demonstrates a beneficial effect of prednisolone. The Weibull accelerated failure time model ratio was 0.91 (95% CI, 0.76-1.10), indicating that the time to resolution was reduced by 9% (0.45 days) with prednisolone compared with placebo (P = .34); the lower limit of the 95% CI did not exclude the 20% a priori minimum clinically important difference. Further (secondary analysis) adjustment for factors demonstrating possible imbalance at baseline (age, sex, and influenza vaccine in last 12 months) and smoking had no effect on the models (Table 2). The difference between the prednisolone group and the placebo group, expressed as the absolute difference in percentage with unresolved moderately bad or worse cough at day 7, was −3.61 (95% CI, −10.64 to 4.23) (Table 3); this can be interpreted as 3.61% fewer participants in the prednisolone group who still had an unresolved moderately bad or worse cough at the end of day 7.

Mean symptoms severity scores (and residuals) were normally distributed. The mean symptoms severity scores were 1.99 (SD, 0.99) and 2.16 (SD, 1.09) points for the prednisolone and placebo groups, respectively. Adjusting for center and baseline illness severity, the mean symptoms severity difference was 0.20 point (95% CI, −0.40 to 0.00 point; P = .05 with a priori α = .001) between prednisolone and placebo (Table 2). With a mean symptoms severity score of 2.16 in the placebo group, a difference of 0.20 equates to a relative reduction of 9.3%. The lower limit of the 95% CI of this reduction was 18.5% and excluded the 20% a priori minimum clinically important difference. Additional adjustment for factors demonstrating imbalance at baseline and smoking marginally attenuated the difference in means and reduced the strength of evidence against the null hypothesis (Table 2).

None of the sensitivity analyses had any effect on the primary comparisons, including those with no moderately bad or worse cough at baseline, multiple imputation of missing data, per-protocol analysis, and adjusting for day of recruitment (eTable 1 in Supplement 1).

There were no significant effects on any symptom duration or peak flow up to 28 days or cough duration up to 56 days (Table 4). No significant effects were observed for antibiotic use; patient satisfaction or intention to use the same treatment if it were to be available in the future; nonserious adverse events (Table 4); expected, unexpected, or cough-related adverse events; or reconsultations (eTable 2 in Supplement 1). The nature of the adverse events was similar between the groups (eTable 3 in Supplement 1), no new urinary or visual symptoms were reported, and none of the patients reporting fatigue, thirst, or dry throat (eTable 3 in Supplement 1) had diabetes. There were no serious adverse events. Four participants (3 in the prednisolone group and 1 in the placebo group presented to the emergency department but were not hospitalized.

All 95% confidence intervals for the interaction effects included values consistent with no significant subgroup effect (eTable 4 in Supplement 1).

Quiz Ref IDIn this randomized trial of 401 adults, 5 days of moderate-dose oral prednisolone did not reduce the duration of moderately bad or worse cough, or the severity of symptoms between days 2 and 4 adults without asthma who presented to primary care with acute lower respiratory tract infection. No effects were observed for duration and severity of any acute lower respiratory tract infection symptom, duration of abnormal peak flow, antibiotic use, or adverse events, including worsening of glycemic control in patients with diabetes.

This study has several strengths. It was an adequately powered, multicenter, fully masked, randomized trial with low rates of missing baseline and follow-up data. The design was pragmatic, using eligibility criteria easily reproduced in routine clinical practice and clinically relevant, validated¹⁸ outcomes. The final sample included participants with high rates of self-reported sputum production and wheeze and was generalizable to adults without asthma presenting to primary care with acute lower respiratory tract infection in whom an immediate antibiotic is not necessary. With 398 participants, this trial more than doubles the number of patients recruited to primary care trials of corticosteroids for acute lower respiratory tract infection¹⁵ and, to our knowledge, is the first to investigate the effects of oral rather than inhaled steroids. The trial also contributes to a growing body of evidence suggesting that systemic and topical corticosteroids have a limited role in the treatment of common infections and their postinfectious complications in primary care.^26,27 This contrasts with an increasing number of studies suggesting that corticosteroids are effective for secondary-care patients with community-acquired pneumonia,²⁸ croup,²⁹ acute sinusitis,³⁰ and severe sore throat.³¹

This study also has several limitations. First, the low patient recruitment rate suggests that patients may have been selectively invited to participate, affecting the generalizability of the final sample. However, the rate was higher than a similar previous trial,¹⁹ not all practices were active throughout the recruitment period, and the characteristics of the final sample appears representative of primary care adult patients with acute lower respiratory tract infection. Second, there was a higher than expected number of participants with zero duration of moderately bad or worse cough, although a sensitivity analysis including these participants did not influence the results. Third, other baseline biomarkers (eg, inflammatory, microbiological, spirometric, or radiographic) were not measured, and it is possible that patients with more severe, inflammatory, eosinophilic,^32,33 or microbiological (eg, rhinovirus)³⁴ etiology entered the trial or could have differentially benefited. However, the study used readily recognized, pragmatic entry criteria facilitating replication in routine clinical practice. Fourth, study eligibility criteria might have included some patients with chronic or postinfectious cough rather than acute lower respiratory tract infection. However, 100% of participants had evidence of active lower respiratory tract involvement (sputum, shortness of breath, wheeze, or chest pain) and more than 75% had a preconsultation cough duration of less than 21 days. Fifth, the study used a patient-reported outcome rather than an objective primary outcome measure (such as digitally measured cough severity). This was chosen because it was considered the strongest option in the presence of a fully masked intervention, it closely reflected patient priorities, and it allowed comparison with other trials.^4,19 Sixth, the lack of effects and a similar between-group pattern of adverse events could reflect poor adherence. However, this is unlikely because standard methods³⁵ were used to establish similar and high levels of adherence to both prednisolone and placebo, and adverse events were similar to another trial in which a similar dosage of prednisolone was proven effective.³⁵

Quiz Ref IDThis trial suggests that oral corticosteroids should not be used in adult primary care patients without asthma or chronic obstructive pulmonary disease who do not require treatment with an immediate antibiotic. Further research is needed to establish effectiveness in primary care patients with more severe infections, such as those with elevated C-reactive protein levels or requiring immediate antibiotic treatment, and larger studies or meta-analysis are needed to address effects in subgroups, such as those with longer preconsultation illness and nonsmokers.¹⁵

Among adults without asthma who developed acute lower respiratory tract infection, the use of oral prednisolone for 5 days did not reduce symptom duration or severity. These findings do not support oral steroids for treatment of acute lower respiratory tract infection in the absence of asthma.

Corresponding Author: Alastair D. Hay, FRCGP, Centre for Academic Primary Care, NIHR School for Primary Care Research, Population Health Sciences, Bristol Medical School, University of Bristol, Canynge Hall, 39 Whatley Rd, Clifton, Bristol BS8 2PS, England (alastair.hay@bristol.ac.uk).

Accepted for Publication: July 21, 2017.

Author Contributions: Dr Brookes had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Hay, Little, Harnden, Thompson, Wang, Kendrick, Brookes, May, Carroll, El-Gohary, Moore.

Acquisition, analysis, or interpretation of data: Harnden, Wang, Kendrick, Orton, Brookes, Young, May, Hollinghurst, Carroll, Downing, Timmins, Lafond, Moore.

Drafting of the manuscript: Hay, Kendrick, Orton, Brookes, Young, Downing, Timmins, Lafond, El-Gohary, Moore.

Critical revision of the manuscript for important intellectual content: Hay, Little, Harnden, Thompson, Wang, Kendrick, Orton, Brookes, Young, May, Hollinghurst, Carroll, Lafond, Moore.

Statistical analysis: Brookes, Young, May.

Obtained funding: Hay, Little, Harnden, Wang, Kendrick, Orton, Moore.

Administrative, technical, or material support: Thompson, Orton, Downing, Timmins, Lafond, Moore.

Supervision: Hay, Little, Harnden, Wang, Kendrick, Brookes, May, Hollinghurst, Lafond.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Thompson reports that he has received funding from Alere Inc to conduct research on C-reactive protein point-of-care tests, has received funding from Roche Molecular Diagnostics for consultancy work, and is a cofounder of Phoresa Inc, which is developing point-of-care tests for primary care. No other disclosures were reported.

Funding/Support: This article presents independent research funded by the National Institute for Health Research (NIHR) School for Primary Care Research (grant reference 117a). Dr Hay is funded by an NIHR Research Professorship (NIHR-RP-02-12-012). The study sponsor was the University of Bristol.

Role of the Funder/Sponsor: Neither the funder nor the sponsor had no involvement in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.

Disclaimer: The views expressed herein are those of the authors and not necessarily those of the NIHR, the National Health Service, or the UK Department of Health.

Additional Contributions: We thank the participants, the recruiting primary care sites, the NIHR Clinical Research Network, and all members of the OSAC team. We also thank the members of the trial steering committee (who provided independent supervision on behalf of the funder and sponsor) and the data monitoring committee (who oversaw safety), the Nottingham University Hospitals NHS Trust pharmacy, and the University Hospitals Bristol NHS Foundation Trust. We thank Mark Ebell, MD, University of Georgia, who (without compensation) conducted a secondary data analysis to estimate the use of oral prednisolone for acute lower respiratory tract infection using administrative data from the southeastern United States.