Prednisolone Versus Dexamethasone for Croup: a Randomized Controlled Trial

We conducted a prospective, double-blind, randomized controlled trial at 2 urban EDs: Princess Margaret Hospital for Children (a tertiary pediatric center) and Joondalup Health Campus (an urban district hospital). Hospital research ethics committee approval was obtained at both centers, and the study was registered with the Australian New Zealand Clinical Trials Registry (identifier ACTRN12609000290291). De-identified individual participant data will not be made available.

Patients were a convenience sample of children presenting to either of the EDs with croup during the study period (March 2009–July 2012). The disease state was defined as a clinical diagnosis of croup (laryngotracheitis) by the ED doctor after a history and physical examination. If there was any diagnostic uncertainty, clinicians could refer to a Guidance for Doctors information sheet (Supplemental Information), in which a croup diagnosis is defined as a hoarse voice or barking cough and stridor (with or without increased work of breathing) directly observed or elicited in the history. Inclusion criteria were age >6 months, contactable by telephone, and English-speaking caregivers. A maximum weight of 20 kg was imposed to limit the maximum possible dexamethasone dose to 12 mg (adult dose). Exclusion criteria included known prednisolone or dexamethasone allergy, immunosuppressive disease or treatment, steroid therapy or enrollment in the same study within the previous 14 days, and a high clinical suspicion of an alternative diagnosis, with specific prompts to include bacterial tracheitis, inhaled foreign body, retropharyngeal abscess, epiglottitis, angioedema, vascular ring, and subglottic stenosis. Signed consent was obtained from caregivers by the treating doctor after the doctor provided a standardized information sheet (Croup Study: Information for Parents, see Supplemental Information).

Patients were randomly assigned to 1 of 3 interventions: dexamethasone (0.6 mg/kg; standard “control” treatment), low-dose dexamethasone (0.15 mg/kg), or prednisolone (1 mg/kg). The randomization list was computer-generated at www.randomization.com by using randomly permuted block sizes in the ratio of 4 patients from each group, with block randomization by center. All medications were prepared, randomly assigned, and labeled by a clinical trials pharmacist at Princess Margaret Hospital. To preserve the influence of palatability, no masking agents were used; however, individual doses of the trial medication were identically packaged, dispensed via a dispensing rack in order of randomization, and dosed at 0.3 mL/kg by the nurse after enrollment and initial severity scoring. Both staff (administering and assessing treatments) and patients were therefore blinded to the treatment allocation.

The study had 2 primary outcome measures: (1) an objective and validated^3,19  measure of croup severity, the Westley Croup Score (WCS),²⁰  and (2) re-attendance for follow-up of ongoing symptoms.

The WCS is a clinical croup score, with a range of 0 to 17 points, that is based on stridor, retractions, air entry, cyanosis, and level of consciousness (Supplemental Fig 5). It is possible to have a clinical diagnosis of croup that is based on a history of a barking cough and yet have a WCS of 0, with no stridor or retractions and normal air entry. The WCS was assessed at baseline, at 1 hour after treatment, hourly up to 6 hours, and again at 12 hours for patients not yet discharged from the hospital; patients were discharged when clinically appropriate (Croup Oral Steroid Study: Guidance for Doctors, see Supplemental Information). The second primary outcome was unscheduled medical re-attendance (for any reason) during the 7 days after treatment. This information was obtained via a telephone call to caregivers within 4 weeks after discharge. If the caregiver was deemed as not contactable (after several failed contact attempts), re-attendance to the ED with a diagnosis of croup was searched for electronically.

Secondary outcomes included total hospital stay, ED length of stay, vomiting, use of nebulized epinephrine (adrenaline), endotracheal intubation, need for additional steroid doses, and need for admission to an inpatient ward, emergency short-stay unit, or ICU.

The study was designed to demonstrate the noninferiority²¹  of low-dose dexamethasone and prednisolone relative to the standard full dose of dexamethasone for the 2 primary outcomes: (1) WCS (from baseline) at 1 hour and subsequent hourly intervals after administration and (2) the rate of unscheduled medical re-attendance in the 7 days after administration. The analysis was conducted per protocol via intention to treat; noninferiority was specified as the upper bound of the 2-sided 95% confidence interval (CI) for the reduction in WCS (for the intervention group [low-dose dexamethasone or prednisolone] relative to the standard treatment group [dexamethasone]) not exceeding 0.5. Whereas authors of some previous studies have used a WCS of 1 as a clinically important difference,²²  we selected 0.5 to increase the study’s discriminatory ability²¹  among milder croup cases (scores between 0 and 2), which typically account for the majority of patients.³  Under this design, it was determined that 437 participants would be required, per arm, to provide 90% power to demonstrate noninferiority (by using a 2-sided t test; common SD of 1.8; α = .05). The expected re-attendance rate for croup at our institution is 4% to 6%⁴ ; this sample size would provide 85% power to detect an absolute difference in re-attendance rate of 5% (based on 10% vs 5%; α = .05).

Descriptive statistics are presented as mean and SD, median, or count and percentage, as appropriate. A per-protocol analysis was used to employ linear regression, adjusted for baseline levels, of the postadministration hourly assessment of the WCS to calculate 95% CIs for the difference between groups. The χ² test was used to compare the frequency of dichotomous and categorical variables; a 1-way analysis of variance or Student’s t test was used to compare parametric variables, either between all groups or between each intervention group and the control group, respectively. Secondary outcomes were analyzed by using the same techniques; because of a moderate right skew, length of stay was analyzed after a log transformation, and the ratio of geometric means between groups is reported. For patients whose WCS was not recorded at an hourly assessment, if, and only if, their baseline WCS was 0 and they were discharged within the 60 minutes preceding that assessment, their score was assumed to be 0 for that assessment. Post hoc consistency analyses included additional modeling approaches to validate the per-protocol analysis. This included dichotomizing patients according to their status of recovered (either discharged or with a WCS of 0 at the hourly clinical review [assuming it was not 0 at baseline]) versus not recovered (still within the hospital) or improved (either discharged or with a WCS lower at the hourly clinical review than at baseline) versus not improved (still within the hospital) and using logistic regression to generate odds ratios (ORs) and 95% CIs for each treatment relative to the control group. The WCS at the hourly clinical review was also analyzed unmodified via ordinal logistic regression to generate ORs with 95% CIs. Models were adjusted for age and study center and, when appropriate, baseline WCS. Because of the higher than anticipated number of patients with multiple attendances within the study, post hoc analyses were conducted to calculate the intraclass correlation coefficient and to replicate the per-protocol analysis with clustering on subject identification. All data manipulation and analysis were conducted in R.²³ 

There was no patient and public involvement in the study design or implementation. No patients or their representatives were asked to help interpret or disseminate the results.

A total of 1252 patient attendances underwent random assignment into the trial. After exclusions (Fig 1), 1231 patients entered the analysis set; 410 were assigned to dexamethasone (0.6 mg/kg), 410 were assigned to low-dose dexamethasone (0.15 mg/kg), and 411 assigned to prednisolone (1 mg/kg). These 1231 attendances included 105 repeat enrollments: 48 patients enrolling twice and a further 3 patients enrolling 3 times each, all outside the 14-day exclusion period. Twenty-eight patients were enrolled despite meeting ≥1 exclusion criterion; 19 children weighing >20 kg, 4 children with laryngomalacia, 4 children with steroid use in the 14 preceding days, and 3 children <6 months of age all were included in the analysis on an intention-to-treat basis. The distribution of repeat enrollments and patients meeting an exclusion criterion was relatively balanced across treatment groups (dexamethasone/low-dose dexamethasone/prednisolone distributed at 1:1.48:1.14 and 1:1.29:1.71, respectively). The mean age and weight of patients was 30 months and 14.0 kg, respectively; 38% of patients were girls, and these characteristics, in addition to croup severity at presentation (baseline WCS), did not differ between groups (Table 1).

The percentage of patients available for the 1-hour assessment was similar between groups at 88.3%, 88.3%, and 89.1% for dexamethasone, low-dose dexamethasone, and prednisolone, respectively. There was no statistically significant difference between the 3 groups for the WCS at the 1-hour assessment, with the adjusted difference in scores at 1 hour (relative to the dexamethasone group) being 0.03 (95% CI −0.09 to 0.15) for low-dose dexamethasone and 0.05 (95% CI −0.07 to 0.17) for prednisolone (Table 2, Fig 2). The upper limits of these CIs fall within the prespecified noninferiority margin of 0.5.

Re-attendance rates (Table 2) were modest at 17.8% (dexamethasone), 19.5% (low-dose dexamethasone), and 21.7% (prednisolone), and similarly, ED re-attendance rates were low at 5.9% (dexamethasone), 8.8% (low-dose dexamethasone), and 7.5% (prednisolone), with no statistical difference between treatment groups (Table 3).

For the low-dose dexamethasone group relative to the dexamethasone group, the WCS was 0.11 higher at 2 hours and 0.23 higher at 3 hours (Fig 2); although the difference was significant at the 3-hour assessment (P = .04), the upper limit of the 95% CI (0.45) was within the noninferiority margin (Table 4, Fig 3). These estimates were 0.04 at 2 hours and 0.04 at 3 hours for the prednisolone group, with the upper limit of the 95% CIs not crossing 0.5. These results must be interpreted with caution given the lower numbers.

Consistency analyses for the odds of recovery (WCS of 0 or discharged) and improvement (lower WCS) are shown in Table 4, with no statistically significant differences demonstrated.

The median length of stay across patients was 124 minutes; this did not differ significantly between the study groups (P = .23), as reflected in the number of patients for which hourly assessment data were available (87.4%, 12.4%, 4.2% for hours 1, 2, and 3, respectively). The ratio of geometric means for total length of stay, relative to dexamethasone, was 0.99 (95% CI 0.92 to 1.07) and 1.04 (95% CI 0.97 to 1.12) for low-dose dexamethasone and prednisolone, respectively (Fig 4). There was no difference (P = .63) between groups in the percentage of participants whose total length of stay exceeded 4 hours (8.8%, 7.1%, and 8.5% for dexamethasone, low-dose dexamethasone, and prednisolone, respectively). The intraclass correlation coefficient was lower for the recovery at 1 hour (0.17) but higher for length of stay (0.52); despite this, incorporation of clustering only saw changes to the SEs in the third and fourth decimal place; therefore, per-protocol analyses are presented.

There were no differences between treatment groups in the need for nebulized epinephrine (2.2%–3%) or in incidence of vomiting (up to 4%) after treatment (Table 3). A repeat dose of epinephrine was given to 0%, 1.2%, and 1.0% of participants in the dexamethasone, low-dose dexamethasone, and prednisolone groups, respectively. No study participants required intubation, and none were admitted to intensive care during their hospitalization. One or more additional steroid doses were given to 11.3%, 15.1%, and 18.9% of participants in the dexamethasone, low-dose dexamethasone, and prednisolone groups, respectively (P = .04).

Adverse events were reported in only 4 patients; 1 child assigned to dexamethasone had a 30-second febrile convulsion ∼30 minutes after dosing, which was not attributed to the medication by the treating clinicians; 1 child assigned to prednisolone developed insomnia (dose at ∼5:00 pm and awake until 3:00 am); 1 child assigned to low-dose dexamethasone was transferred back to the ED from the ED short-stay unit and treated with nebulized epinephrine for possible stridor; and 1 patient assigned to low-dose dexamethasone developed hyperactivity 30 minutes after the dose.

The ToPDoG (Trial of Prednisolone/Dexamethasone Oral Glucocorticoid) study, is, to the best of our knowledge, the largest croup randomized controlled trial published to date. Our findings confirm the clinical experience of safety²⁴  and efficacy^1,8,9,25  of oral steroids for croup. We studied 2 different but complementary primary outcome measures: an objective measure of acute severity and improvement (the WCS) and also a real-world, clinically relevant outcome, re-attendance rate, which has implications for patient and family satisfaction as well as use of resources in hospitals and the wider community.

Dexamethasone is generally not available outside of the hospital environment, so there is a distinct advantage of being able to use prednisolone to treat croup in the community setting. We chose not to use masking agents in the preparation of trial medications because palatability issues affect the real-world utility of these medications, especially in pediatric populations.

A number of limitations were applicable to our study. The study population was a convenience sample from 2 institutions, and ∼1 in 7 patients with croup were enrolled. Our power calculation (based on hypothesis testing) predicted that 1311 patients were required; we only enrolled 1252 subjects. However, the CIs in our data suggest that our sample was large enough to answer the clinical questions posed.

We were unable to record the number of participants screened for inclusion, and the number of patients who were excluded or who declined consent because data collection sheets were only retained for those who met enrollment criteria. Because of limited resources and challenging logistics (general population, ED sample), follow-up was not as robust as intended, with only ∼70% of families contactable by phone. For the remaining 30%, we had to search ED attendance records for re-attenders diagnosed with croup; we were not able to assess all re-attendances in the study group; therefore, we may have missed those who re-attended ED with a different diagnosis, and we were unable to determine the rate of general practitioner (GP) re-attendance in this group.

Our results indicate that it is acceptable to use any of the 3 commonly used oral steroid regimes to treat croup in children. The vast majority (92%) of patients were successfully treated and discharged within 2 hours, improving from an average WCS of ∼1.5 to ∼0.5 over the first hour after treatment, with no differences between the 3 groups. The “ceiling effect” proposed by Geelhoed and Macdonald,¹²  whereby steroid doses higher than a certain threshold would not have any additional benefit, seems to be applicable in these patients, in line with several decades of experience with using low-dose dexamethasone for croup at our institution.⁴  Our study revealed ED re-attendance rates similar to those from other studies,^7,8,16–18  with no differences between groups for either GP or ED croup re-attendance.

When comparing the groups at 2- and 3-hour clinical reviews, there appears to be a progressive divergence of the WCS for low-dose dexamethasone compared with dexamethasone. Although this difference reached statistical significance (P = .042), the effect size was moderate (0.23), and even with the relatively reduced sample size available for the 3-hour clinical review, the upper limit of the CI lies within the predefined noninferiority margin of a 0.5 difference in the WCS. This result is broadly consistent with noninferiority,²¹  although it is suggestive of a worse outcome for low-dose dexamethasone. One possible explanation would be that the steroid ceiling is at a dose higher than 0.15 mg/kg for a minority of patients.

Duration of treatment has been raised by some authors^18,26  who suggest that treatment with prednisolone should constitute multiple doses (3 days in the study by Garbutt et al¹⁸ ) to cover the expected duration of the illness because prednisolone has a shorter clinical duration of action.¹⁵  Our study was not designed to test different durations of treatment, but it did reveal that patients treated with a single dose of prednisolone were statistically more likely (P = .02) to receive additional doses of the steroid than those treated with dexamethasone.

One suggestion for further study relates to the apparent weakening performance for low-dose dexamethasone (0.15 mg/kg) at the 3-hour assessment. This effect may be due to a small number of patients who do not respond to oral steroid treatment within 1 to 2 hours, constituting a treatment-resistant cohort; <4% of our patients were still in the ED at the 3-hour mark. These nonresponders may have different responses to steroid treatment, or they may require higher doses to effectively treat their croup.

Oral steroids are an effective treatment of croup, and the type of steroid seems to have no clinically significant impact on efficacy, both acutely and during the week after treatment. Children treated with prednisolone initially are more likely to require additional doses to cover the duration of the illness.

We gratefully acknowledge Sharon O’Brien (research assistant) who conducted the telephone follow-up of enrolled patients. Dr Gareth Kameron helped with early study administration, including the study drug dispensing mechanism, telephone follow-up of patients, and initial data collation. Dr Dami Denbali helped with the telephone follow-up. Trial pharmacists Margaret Shave and Thanh Tan assisted greatly with medications management. We also thank ED doctors and nurses at Princess Margaret Hospital and Joondalup Health Campus for consenting and enrolling patients during their busy shifts.

CI

confidence interval

ED

emergency department

GP

general practitioner

OR

odds ratio

WCS

Westley Croup Score