Few evidence-based treatment guidelines for tendinopathy exist. We undertook a systematic review of randomised trials to establish clinical efficacy and risk of adverse events for treatment by injection. We searched eight databases without language, publication, or date restrictions. We included randomised trials assessing efficacy of one or more peritendinous injections with placebo or non-surgical interventions for tendinopathy, scoring more than 50% on the modified physiotherapy evidence database scale. We undertook meta-analyses with a random-effects model, and estimated relative risk and standardised mean differences (SMDs). The primary outcome of clinical efficacy was protocol-defined pain score in the short term (4 weeks, range 0-12), intermediate term (26 weeks, 13-26), or long term (52 weeks, ≥52). Adverse events were also reported. 3824 trials were identified and 41 met inclusion criteria, providing data for 2672 participants. We showed consistent findings between many high-quality randomised controlled trials that corticosteroid injections reduced pain in the short term compared with other interventions, but this effect was reversed at intermediate and long terms. For example, in pooled analysis of treatment for lateral epicondylalgia, corticosteroid injection had a large effect (defined as SMD>0·8) on reduction of pain compared with no intervention in the short term (SMD 1·44, 95% CI 1·17-1·71, p<0·0001), but no intervention was favoured at intermediate term (-0·40, -0·67 to -0·14, p<0·003) and long term (-0·31, -0·61 to -0·01, p=0·05). Short-term efficacy of corticosteroid injections for rotator-cuff tendinopathy is not clear. Of 991 participants who received corticosteroid injections in studies that reported adverse events, only one (0·1%) had a serious adverse event (tendon rupture). By comparison with placebo, reductions in pain were reported after injections of sodium hyaluronate (short [3·91, 3·54-4·28, p<0·0001], intermediate [2·89, 2·58-3·20, p<0·0001], and long [3·91, 3·55-4·28, p<0·0001] terms), botulinum toxin (short term [1·23, 0·67-1·78, p<0·0001]), and prolotherapy (intermediate term [2·62, 1·36-3·88, p<0·0001]) for treatment of lateral epicondylalgia. Lauromacrogol (polidocanol), aprotinin, and platelet-rich plasma were not more efficacious than was placebo for Achilles tendinopathy, while prolotherapy was not more effective than was eccentric exercise. Despite the effectiveness of corticosteroid injections in the short term, non-corticosteroid injections might be of benefit for long-term treatment of lateral epicondylalgia. However, response to injection should not be generalised because of variation in effect between sites of tendinopathy. None. Copyright © 2010 Elsevier Ltd. All rights reserved.