Twenty-nine studies were included in the review. Twenty-three studies satisfied more than five of 11 quality criteria, although selection bias (particularly differences in baseline pain status between study groups) was apparent, largely in studies evaluating pain.
Pain: Compared with control, epidural steroid injections showed a favourable effect for pain at six months of follow-up (WMD -0.41, 95% CI -0.66 to -0.16; 10 comparisons). At 12 months follow-up the difference was no longer statistically significant (nine comparisons). There was no substantial heterogeneity. The results at both time points were not statistically significant when the analysis was adjusted for baseline pain score.
Disability: There were no statistically significant differences at six months (eight comparisons) or 12 months (nine comparisons) between epidural steroid injections and placebo/other procedures. Substantial heterogeneity was observed (Ι²=70.6% for six-month comparisons and Ι²=89.6% for 12-month follow-up comparisons). Adjustments for baseline score did not alter the main result.
Subsequent surgery: There was no statistically significant decrease in numbers of patients treated with epidural steroid injections (delivered by any approach) who subsequently underwent surgery compared with placebo/other procedures (19 comparisons). There was no evidence of heterogeneity. After baseline adjustments were made, subgroup analyses did not reveal any statistically significant impacts.
Publication bias was a possibility in studies that evaluated disability. Additional results were reported in the paper.