Summary of findings for the main comparison. Glucocorticosteroids for people with alcoholic hepatitis.

Glucocorticosteroids for people with alcoholic hepatitis
Patient or population: participants with alcoholic hepatitis at high risk of mortality and morbidity Settings: hospitals and clinics Intervention: glucocorticosteroids Comparison: placebo or no intervention
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Placebo or no intervention	Glucocorticosteroids
All‐cause mortality: up to 3 months' follow‐up after randomisation	299 per 1000	278 per 1000 (210 to 344)	RR 0.90 (0.70 to 1.15)	1861 (15 RCTs)	⊕⊝⊝⊝ Very lowa	We downgraded for inconsistency because of selection bias in the trials: trials either included or excluded people with gastrointestinal haemorrhage, active peptic ulcer disease, pancreatitis, renal failure, bacterial infections. The OIS was 7870 participants.
Health‐related quality of life: up to 3 months (measured with European Quality of Life – 5 Dimensions – 3 Levels (EQ‐5D‐3L)b scale)	The mean value was 0.592	The mean value was 0.553 (0.502 to 0.604)	MD –0.04 (–0.11 to 0.03)	377 (1 RCT)	⊕⊕⊝⊝ Lowc	—
Serious adverse events during treatment	362 per 1000	381 per 1000 (398 to 467)	RR 1.05 (0.85 to 1.29)	1861 (15 RCTs)	⊕⊝⊝⊝ Very lowd	The OIS was 4197 participants.
Liver‐related mortality: up to 3 months' follow‐up after randomisation	299 per 1000	267 per 1000 (207 to 341)	RR 0.89 (0.69 to 1.14)	1861 (15 RCTs)	⊕⊝⊝⊝ Very lowe	The OIS was 7987 participants.
Participants with any complication: up to 3 months following randomisation	444 per 1000	462 per 1000 (382 to 564)	RR 1.04 (0.86 to 1.27)	1861 (15 RCTs)	⊕⊝⊝⊝ Very lowf	The OIS was 5980 participants.
Participants with non‐serious adverse events: up to 3 months' follow‐up after randomisation	52 per 1000	104 per 1000 (38 to 285)	RR 1.99 (0.72 to 5.48)	160 (4 RCTs)	⊕⊝⊝⊝ Very lowg	The OIS was 2698 participants.
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; MD: mean difference; OIS: optimal information size; RCT: randomised controlled trial; RR: risk ratio.
GRADE Working Group: certainty of evidence grades High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

^aDowngraded three levels: one level due to within‐study risk of bias (high overall risk of bias in all the trials but one (Thursz 2015); one level due to inconsistency of the data (there is wide variation in the effect estimates across studies; there is little overlap of confidence intervals associated with the effect estimates; presence of moderate heterogeneity: I² = 45%; heterogeneity could be explained with selection bias); one level due to imprecision (the OIS was not reached).
 ^bEQ‐5D‐5L: a self‐report, multiple‐choice questionnaire that provides a simple descriptive profile and a single index value for health status. The EQ‐5D‐5L essentially consists of two pages: the EQ‐5D descriptive system (on page 2) and the EQ visual analogue scale (EQ VAS) (on page 3). The descriptive system comprises the following five dimensions: mobility, self‐care, usual activities, pain/discomfort, and anxiety/depression. Each dimension has five levels: no problems; slight problems; moderate problems; severe problems; and extreme problems. The EQ VAS records the respondent's self‐rated health on a vertical, visual analogue scale. A summary index with a maximum score of 1 can be derived from these five dimensions by conversion with a table of scores. The maximum score of 1 indicates the best health state, by contrast with the scores of individual questions, where higher scores indicate more severe or frequent problems. Utility values for perfect health and death are 1 and 0, respectively. In addition, there is a visual analogue scale to indicate the general health status with 100 indicating the best health status.
 ^cDowngraded two levels: one level due to within‐study risk of bias (high overall risk of bias in the trial); one level due to imprecision of effect estimates (fewer than 400 participants).
 ^dDowngraded three levels: one level due to within‐study risk of bias (high overall risk of bias in all the trials but one); one level due to inconsistency of the data (there was wide variation in the effect estimates across studies; there was little overlap of confidence intervals associated with the effect estimates; presence of moderate heterogeneity: I² = 36%; heterogeneity could be explained with selection bias); one level due to imprecision (the OIS was not reached).
 ^eDowngraded three levels: one level due to within‐study risk of bias (high overall risk of bias in all the trials but one); one level due to inconsistency of the data (there is wide variation in the effect estimates across studies; there is little overlap of confidence intervals associated with the effect estimates; presence of moderate heterogeneity: I² = 46%; heterogeneity could be explained with selection bias); one level due to imprecision (the OIS was not reached).
 ^fDowngraded three levels: one level due to within‐study risk of bias (high overall risk of bias in all the trials but one); one level due to inconsistency of the data (there was wide variation in the effect estimates across studies; there was little overlap of confidence intervals associated with the effect estimates; presence of moderate heterogeneity: I² = 41%; heterogeneity could be explained with selection bias); one level due to imprecision (the OIS was not reached).
 ^gDowngraded four levels: one level due to within‐study risk of bias (high overall risk of bias in all the trials but one); one level due to inconsistency of the data (there is little overlap of confidence intervals associated with the effect estimates); one level due to publication bias (only four trials with a small number of participants reported on non‐serious adverse events); one level due to imprecision (the OIS was not reached).