Table 2.
GRADE summary of findings for corticosteroids (intervention) versus no corticosteroids (control) in patients with sore throat
| Outcome and timeframe | Study results (95% CI) and measurements | Absolute effect estimates | Quality of evidence | Summary | ||
|---|---|---|---|---|---|---|
| No corticosteroids | Corticosteroids | Difference (95% CI) | ||||
| Complete resolution of pain at 24 hours | Relative risk: 2.24 (1.17 to 4.29). 1049 patients in 5 studies | 100/1000 | 224/1000 | 124 more (17 more to 329 more | Moderate (inconsistency and imprecision)* † ‡ | Corticosteroids probably increase chance of complete resolution of pain at 24 hours |
| Complete resolution of pain at 48 hours | Relative risk: 1.48 (1.26 to 1.75). 1076 patients in 4 studies | 425/1000 | 629/1000 | 204 more (111 more to 319 more) | High‡ | Corticosteroids increase chance of complete resolution of pain at 48 hours |
| Recurrence/relapse of symptoms | Relative risk: 0.52 (0.16 to 1.73). 372 patients in 3 studies | 65/1000 | 34/1000 | 31 fewer (55 fewer to 47 more) | Moderate (serious imprecision)‡ § ¶ | Corticosteroids probably have no important effect on chance that symptoms recur |
| Antibiotics prescription | Relative risk: 0.83 (0.61 to 1.13). 342 patients in 1 study. Follow-up 28 days | 564/1000 | 468/1000 | 96 fewer (220 fewer to 73 more) | Low (very serious imprecision)** | Corticosteroids might decrease chance of taking antibiotics in patients given prescription with instructions to take antibiotic if unimproved or worse |
| Mean time to onset of pain relief (hours) | 907 patients in 8 studies | 12.3 hours | 7.4 hours | 4.8 fewer (7.8 fewer to 1.9 fewer) | Moderate (inconsistency and imprecision)‡ †† ‡‡ §§ | Corticosteroids probably shorten the time until pain starts to improve. |
| Mean time to complete resolution of pain (hours) | 720 patients in 6 studies | 44.0 hours | 33.0 hours | 11.1 fewer (21.8 fewer to 0.4 fewer) | Low (serious imprecision and inconsistency)‡ †† ‡‡ ¶¶ | Corticosteroids might shorten duration of pain |
| Pain reduction 24 hours | Scale: high better. 1247 patients in 8 studies | Mean 3.3 hours | Mean 4.6 hours | 1.3 higher (0.7 higher to 1.9 higher) | Moderate (inconsistency and imprecision)‡ †† ‡‡ *** | Corticosteroids probably reduce severity of pain at 24 hours |
| Duration of bad/non-tolerable symptoms | — | — | — | 0 (0 to 0) | — | No studies provided information on this outcome |
| Days missed from work or school | 181 patients in 2 studies. Follow-up to 14 days | Two trials reported days missed from work/school. In Kiderman et al, 22/40 (55%) in steroids group and 27/39 (69%) in placebo group took time off work (relative risk 0.79, 95% CI 0.56 to 1.13). Marvez-Valls et al reported average time patients in each arm missed from work/school: average 0.4 (SD 1.4) days in intervention group adults v and 0.7 (SD 1.4) days in placebo group adults; mean difference 0.30 days, −0.28 to 0.88) | Moderate (serious imprecision and some concerns of risk of bias)††† ‡‡‡ | Corticosteroids probably have no important effect on days missed from work or school | ||
| Serious adverse events | 808 patients in 3 studies. Follow-up to 10 days | Few adverse effects reported in trials, mostly disease related complications, and occurred with similar frequency in intervention and control groups (see table 3) | Moderate§§§ | Corticosteroids probably do not increase risk of adverse events | ||
*Considerable heterogeneity (I2=69%). Not rated down because clinical inconsistency was deemed not important as all results of included studies have similar clinical implication.
†Limits of confidence interval suggest small benefit in one extreme and benefit important to patients in other. Because imprecision is linked to inconsistency, certainty of evidence rated down by only one level.
‡Publication bias not tested because of small number of studies.
§Not rated down for risk of bias as one of three trials judged to be at high risk of bias from missing participant data.
¶Confidence interval suggests that corticosteroids increase chance of recurrence of symptoms in one extreme but decrease this chance in other extreme.
**Confidence interval suggest that corticosteroids could largely reduce chance of taking antibiotics in one extreme but could slightly increase this chance in other extreme.
††Not rated down for risk of bias as equal number of trials judged to be at high and low risk of bias, but P value for test of interaction showed no difference between two estimates.
‡‡Large unexplained clinical and statistical inconsistency.
§§Confidence interval suggests small benefit in one extreme and benefit that some patients might consider important in other extreme. As this imprecision was result of inconsistency, certainty of evidence rated down by only one level.
¶¶Confidence interval suggests trivial benefit in one extreme and benefit that would be considered important by most patients in other extreme.
***Confidence interval suggests small benefit in one extreme and benefit important to patients in other. As this imprecision was related to inconsistency, rated down by only one level.
†††One study was at high risk of bias from concerns with regards to allocate concealment.
‡‡‡Studies showed that corticosteroids could increase days missed from school or work in one extreme but decrease them in other extreme.
§§§High risk of bias studies showed similar results as low risk of bias studies; however, high risk of selective outcome reporting was possible.