Summary of main results

In participants with sore throat, oral or intramuscular corticosteroids in addition to antibiotics significantly increased the proportion of participants experiencing complete resolution of pain at 24 and 48 hours, although the effect was moderate. Five people need to be treated with corticosteroids to prevent one person continuing to experience pain at 24 hours. Corticosteroids decreased the mean time to onset of pain relief by 6 hours and the mean time to complete resolution of symptoms by 11.6 hours. However, both analyses were associated with high heterogeneity. Corticosteroids reduced the subjective rating of pain at 24 hours by an additional 10.6%, again associated with high heterogeneity. Subgroup analyses revealed a greater effect size for intramuscular versus oral corticosteroids in both the proportion of participants achieving complete pain resolution at 24 hours and the mean time to onset of pain relief, although tests for subgroup effects did not reach significance. The effect on mean time to onset of pain relief was greater in those participants with severe, exudative, and bacterial pathogen‐positive sore throat. A small number of trials reported adverse events, recurrence/relapse rates, or days missed from work or school; we found no difference in the likelihood of these outcomes between participants receiving corticosteroids and those receiving placebo. Only two trials reported funding sources.

Overall completeness and applicability of evidence

We included nine trials that were conducted in general practice and emergency department settings. Most trials were performed in the USA, and only two in Europe. Additional trial data may therefore be warranted in other populations, including those from lower‐ and middle‐income countries, who are likely to differ in consultation patterns and clinical practice (including antibiotic prescribing guidelines), before the results can be generalised to other populations.

Only two trials included children, and the results of these trials were inconsistent, therefore we were unable to draw reliable conclusions about the clinical benefit of corticosteroids in this population. Children commonly present with sore throat, and it will be important to await results from future trials in these cohorts to develop a clearer understanding of whether corticosteroids also provide a benefit in this population.

The most important limitation to applicability of the evidence was that in eight of the nine included trials, participants received antibiotics in addition to corticosteroid or placebo treatment. Only one trial did not give immediate antibiotics, excluding those with a sore throat severe enough to require urgent antibiotics from the study, and asking clinicians to offer delayed antibiotics where they felt it to be necessary.

Quality of the evidence

Whilst we included only randomised, double‐blind, placebo‐controlled trials, in a number of cases the trials did not specify their methods of randomisation or allocation concealment. The included trials reported a variety of outcome measures, in some cases with inadequate reporting, no standard deviations, or use of graphical representation only. Significant heterogeneity was evident in a number of our analyses. This heterogeneity, and the relatively small number of participants in some analyses, resulted in a GRADE evidence certainty assessment of moderate for most of our analyses; the evidence for two analyses (complete resolution of pain at 24 hours and complete resolution of pain at 48 hours) was assessed as of high certainty.

The outcome measures of mean time to onset of pain relief and to complete resolution of pain were limited by recall bias because they relied on participants’ subjective recall and recording. The mean time could also be skewed by a few participants who had sore throat pain for especially long or short periods. A median time may have been more appropriate, although data were insufficient to calculate this. The included studies were also underpowered to detect rare adverse effects of corticosteroid therapy, as well as relapse rates. Only two studies included missed days from work or school, making it difficult to draw robust conclusions from this analysis.

Finally, the limited number of trials meant that we were unable to assess publication bias using a funnel plot, although we attempted to address this issue by using citation searching.

Potential biases in the review process

We included one new study for this update (Hayward 2017). Hayward 2017 included authors involved in this Cochrane Review. To limit bias, Hayward 2017 was appraised and data extracted by two review authors who were not authors of the study. This review update was drafted by an author who was not a published author for Hayward 2017. As outlined in the Differences between protocol and review section, the primary outcomes were changed to reflect the available evidence and to focus on the most relevant symptom of sore throat (pain). Future updates will focus on these outcomes, but if other outcomes are routinely reported in future studies, such as painful swallowing, these could be included.

Agreements and disagreements with other studies or reviews

Four systematic reviews addressing the question of corticosteroids in sore throat have been published in addition to this Cochrane Review (Korb 2010; Mullarkey 2011; Sadeghirad 2017; Wing 2010).

This review update included a recent double‐blind, placebo‐controlled randomised trial by Hayward 2017. Inclusion of this trial greatly increased the numbers of participants incorporated into the analysis. Inclusion of the study reduced the effect size of likelihood of complete resolution of pain at 24 hours, with the effect size maintained at 48 hours. Similarly, the effect size of mean time to onset of pain relief was reduced, whilst mean time to complete resolution of pain was maintained. This is potentially explained by the exclusion criteria in Hayward 2017. Hayward 2017 sought to examine the effects of corticosteroids in the absence of antibiotics, and participants were not routinely given antibiotics. Those participants deemed to require immediate antibiotics at baseline were excluded from the study, which will have excluded those with the most severe sore throats. Although there was no significant difference in outcomes between those participants offered delayed antibiotic prescriptions and those who were not, it is possible that participants with more severe sore throats may derive a greater benefit from steroids early on in an infection, which was not captured in this review. The majority of participants included in this review were adults, whereas studies carried out in children have shown a significant benefit with corticosteroids (Bulloch 2003; Olympia 2005). Also, patients in the Hayward 2017 study were recruited from a general practitioner setting and not an emergency department setting, which could also account for a difference in baseline severity of symptoms. Finally, reporting of time to relief/resolution of pain differed across the studies, with Hayward 2017 contacting participants directly at 24 and 48 hours in addition to participants using a self‐reported symptoms diary.

The most recent systematic review, Sadeghirad 2017, used a similar search strategy to this update and included all trials included in our review as well as a trial by Ahn 2003. We had previously identified this trial as providing poor‐certainty evidence, requiring estimation of standard deviations and scoring zero on the Jadad scale (Jadad 1996). Personal communication with the trial authors revealed that this was not a double‐blind randomised controlled trial. No placebo was given, and participants were not randomised but rather treated consecutively until equal numbers were achieved in both treatment groups. Consequently, we excluded Ahn 2003 from our review. Although Sadeghirad 2017 identified this bias, the study was included in the analysis, therefore likely accounting for the shorter mean time to onset of pain relief found in participants treated with corticosteroids of 4.8 hours (95% CI −1.9 to −7.8). Overall, participants taking corticosteroids were twice as likely as placebo participants to experience pain relief at 24 hours (RR 2.2, 95% CI 1.2 to 4.3) and 1.5 times as likely to experience pain resolution at 48 hours (RR 1.5, 95% CI 1.3 to 1.8). These results were in line with our findings.

Korb 2010 used a similar search strategy to our review but did not search the Database of Abstracts of Reviews of Effects (DARE), the NHS Economic Evaluation Database (NHS EED), or the World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP). Korb 2010 included all trials included in our review with the exception of Hayward 2017, but did not document excluded trials. Although Korb 2010 did not perform a meta‐analysis, similar conclusions to our review were drawn regarding a small but significant beneficial effect of corticosteroids using a descriptive analysis.

Wing 2010 performed an extensive search of the literature, including conference proceedings. In addition to eight of the nine trials included in our review (the exception was Hayward 2017), Wing 2010 also included two trials with low‐certainty evidence (Ahn 2003; Hahn 1951). Hahn 1951 achieved a Jadad score of zero and did not contribute any data to the meta‐analyses. We excluded Hahn 1951 from our review because it was performed in hospitalised participants. However, results reported by Wing 2010, including the trials with low‐certainty evidence, are also in line with our review. Wing 2010 reported mean time to clinically meaningful pain relief as 4.5 hours (MD −4.54, 95% CI −7.19 to −1.89; I² = 81%; P < 0.001), compared to 6 hours in our review, but failed to address the high heterogeneity in this analysis. Wing 2010 also presented a meta‐analysis of absolute VAS scores at 24 hours and demonstrated an MD of −0.9 on a 0‐to‐10 VAS (MD −0.9, 95% CI −1.5 to −0.3; I² = 74%, P = 0.003). This is a difficult finding to interpret because it did not take account of the baseline VAS score, which was variable amongst studies. The trial authors argued that a reduction of 1.3 points represents a clinically meaningful change, and hence this was not a significant result. However, this a priori level of significance has not been extensively validated in the field. In our analysis, which accounted for baseline scores by assessing change from baseline, we found that the reduction from baseline was greater in the corticosteroid group by 11%, or 1.1 cm.

Mullarkey 2011 was a single‐author systematic review that limited the search to adult participants presenting to emergency departments. Mullarkey 2011 included five of the studies included in our review and one systematic review (Hayward 2009), but no additional studies. No meta‐analysis was attempted. A descriptive analysis was supportive of the benefit of corticosteroids as an adjunct in the management of acute pharyngitis.