Of 3257 potentially relevant records identified, 26 were relevant to sore throat, tonsillitis, or pharyngitis (fig 1). A further 17 studies were excluded because they examined postoperative or postintubation sore throat (13 studies), included inpatients (one), did not have a placebo group (one), or were duplicate publications (two). Of the nine that met our inclusion criteria, one was excluded for not describing the method of randomisation.21
Fig 1 Flowchart of search results
The eight studies included 743 patients (369 children, 374 adults): 348 (47%) had exudative sore throat, and 330 (44%) were positive for group A β-haemolytic streptococcus. Patients were recruited from emergency department and general practice settings in four countries: United States (five studies), Canada (one), Israel (one), and Turkey (one) (table 1). Corticosteroids used included betamethasone 2 ml (estimated dose 8 mg, one study), dexamethasone (up to 10 mg, six studies; or prednisone 60 mg, one study): doses were reasonably comparable for their potency. Corticosteroids were administered either intramuscularly (three studies), orally (four), or both (one). Six trials used one dose of corticosteroids, and two trials prescribed more than one dose of corticosteroids to a subgroup of participants.w1 w2 In the eight included studies, methodological quality was high with a low risk of bias (for example, all trials reported adequate allocation concealment and clear methods of randomisation). Table 2 reports the specific elements of methodological quality in the selected studies.
Table 1 Characteristics of trials included in meta-analysis (see web appendix for references)
Table 2 Methodological quality of included studies (see web appendix for references)
Outcome measures included complete pain resolution at 24 hours (four studies) and 48 hours (three studies), mean time to onset of pain relief (five), mean time to complete resolution of pain relief (six), reduction in visual analogue scale pain score (five), number of days missed from school or work (three), and recurrence rates (four). All eight trials prescribed antibiotics to both intervention and placebo groups and allowed simple analgesia. In four trials, analgesia use was recorded, which was not significantly different between placebo and corticosteroid groups. Two trials restricted analgesia to paracetamol for 24 hours or 72 hours, recording no difference in usew6 and not reporting usew7 respectively. Four trials reported outcomes separately for patients positive and negative for bacterial pathogens.w3-w6
Complete resolution of pain at 24 or 48 hours
In a pooled analysis of four trialsw1 w2 w6 w7 patients treated with corticosteroids were three times more likely to have complete resolution of pain at 24 hours (relative risk 3.2, 95% confidence interval 2.0 to 5.1, P<0.001, I2=44%) (fig 2). The number needed to treat was 3.7 (2.8 to 5.9). Significant effects were recorded in adult patients only (relative risk 4.3, 2.3 to 8.1, P<0.001)w1 w6 w7 and in those receiving oral corticosteroids only (2.6, 1.6 to 4.3, P<0.001).w1 w2 w6 Data were insufficient to undertake further subgroup analysis.
Fig 2 Effect of corticosteroids on number of patients experiencing complete pain relief at 24 and 48 hours. See web appendix for references
In three trialsw1 w2 w7 corticosteroids also increased the likelihood of complete resolution of pain at 48 hours (relative risk 1.7, 95%CI 1.3 to 2.1, P<0.001), number needed to treat was 3.3 (2.4 to 5.6) (fig 2). Results were similar in trials with adult patients only (1.8, 1.3 to 2.3, P<0.001)w1 w7 and in those receiving oral corticosteroids only (1.6, 1.2 to 2.1, P=0.004).w1 w2
Mean time to onset of pain relief
Six trials reported the mean time to onset of pain relief, which occurred at an average of 6.3 hours earlier with corticosteroids than without (95% CI 9.3 to 3.4, P<0.001) (fig 3).w3-w8 The wide variation in individual response times caused high heterogeneity (I2=73%). A sensitivity analysis, which excluded each trial in turn, demonstrated a range of weighted mean difference of 5.1 to 7.2 hours, but no loss of significance. The majority of the heterogeneity arose from the trial by Tasar et al, which showed the largest benefit of corticosteroids with small standard deviations.w7 Removal of this trial from the meta-analysis gave a mean time to onset of pain relief 5.1 hours earlier in patients given corticosteroids.
Fig 3 Effect of corticosteroids on mean time to onset of pain relief in hours. See web appendix for references
In patients with an exudative sore throat, corticosteroids also reduced the mean time to onset of pain relief (weighted mean difference 6.2 hours, 8.4 to 4.0). Similarly, we recorded a reduction in mean time to pain relief in sore throat that was bacterial pathogen positive (5.3, 8.0 to −2.6) and in trials selecting for severe sore throat (7.2, 10.1 to 4.3). All three categories of sore throat (exudative, bacterial pathogen positive, and severe) were significant (P<0.001) with no heterogeneity (I2=0) (fig 4).
Fig 4 Effect of corticosteroids on mean time to onset of pain relief analysed by subgroup using meta-regression. PO=oral delivery. IM=intramuscular delivery
The direction of effect for mean time to onset of pain relief was similar in trials with adults only, in trials with intramuscular and oral routes of steroid administration, and in trials in which severe sore throat was not selected. The effect on children only, those with less than 50% exudative sore throat, and non-bacterial pathogen positive only did not reach significance .We did not find significant changes in mean time to onset of pain relief in trials with children only, trials with less than 50% exudative sore throat, and in the subgroup of patients with sore throat not positive for bacterial pathogens. Meta-regression analysis revealed no significant differences across all subgroups (fig 4).
Figure 5 shows the change in visual analogue scale at baseline to 72 hours. The data from Bulloch et al’s trial display the lowest baseline visual analogue scale score (that is, less severe) and the least response at 24 and 48 hours.w3 This finding is reflected in the non-significant effect (and the smallest effect) observed in the mean time to onset of pain relief for this trial.
Fig 5 Mean pain score on visual analogue scale at baseline and after corticosteroids or placebo. See web appendix for references
Time to complete resolution of symptoms
Five trials assessed the mean time to complete resolution of pain.w3-w5 w7 w8 High heterogeneity prevented pooling: three studies showed a benefit of corticosteroids,w5 w7 w8 and two showed non-significant effects in opposing directions. Time to complete resolution ranged from 15 to 45 hours in the corticosteroid group and 35 to 54 hours in the placebo group.
Adverse events, relapse rates, and days missed from school or work
Only one trialw5 of 125 participants reported adverse events: five patients (three steroid, two placebo) were hospitalised for fluid rehydration, and three patients developed peritonsillar abscess (one steroid, two placebo). Three studies reported no significant differences in days missed from school or work.w1 w2 w4 Four trials reported no difference in the incidence of recurrent symptomsw1-w4 (measured between 5 days and 1 month after treatment), whereas one trial found significantly increased recurrence in the placebo group.w6