The need for randomised treatment studies in neglect research.

PURPOSE The aim was to review the methodological quality of trials to evaluate rehabilitation for spatial neglect and to determine the overall effectiveness of interventions METHODS A systematic literature review and meta-analysis were conducted of trials completed by 2005. Trials identified were independently assessed for methodological quality by two reviewers. Outcomes were analysed as the standardised mean difference and 95% confidence intervals with random effects models. RESULTS 25 trials of neglect rehabilitation were identified, 12 randomised controlled trials and 13 controlled clinical trials. The methodological quality was generally poor with only 4 trials achieving an A rating, i.e. low risk of selection bias. The immediate effect of cognitive rehabilitation on disability was small, 0.26 [-0.16, 0.67] and neither this nor the persisting effect 0.61 [-0.42, 1.63] was statistically significant. The most frequently used standardised neglect test (number of single letters correctly cancelled) favoured the experimental group 0.58 [0.10, 1.05] but was not significant. When cancellation errors were measured there was a small immediate effect favouring the experimental group, of borderline statistical significance, -0.65 [-1.28, -0.01] p=0.05, and a significant persisting effect -0.76 [-1.39, -0.13] p=0.02. Cognitive rehabilitation also significantly improved immediate (p=001) and persisting (p=0.02) line bisection performance but these findings are based on only four and one study respectively. CONCLUSIONS The quality of trials identified was poor. Analysis of randomised controlled trials showed some evidence of an effect of intervention on measures of impairment. There was no evidence to support the effects of intervention on measures of disability. Further trials must use methods that reduce bias, have adequate statistical power, and include valid disability outcome measures.