Unilateral neglect (UN) is a heterogeneous perceptual disorder that often follows stroke, especially after right hemisphere lesion. Its most typical feature is failure to report or respond to stimuli presented from the contralateral space, including visual, somatosensory, auditory, and kinesthetic sources. Sufferers may even fail to perceive their own body parts (Mesulam, 1999). The reported incidence varies from 10 to 82% following right- and from 15 to 65% following left-hemisphere stroke (Plummer et al., 2003). Subject selection criteria, lesion site, the nature and timing of the assessment, and lack of agreement on assessment methods are all responsible for the variability in these reported rates (Stone et al., 1991; Azouvi et al., 2002). UN has a significant negative impact associated with functional recovery at home discharge (Jehkonen et al., 2006; Mutai et al., 2012).

Different treatment approaches and assessment tools have been developed to evaluate and address UN. The most recent literature shows that rehabilitation can be classified under two types of behavioral approaches: recruiting the hemiplegic limbs to reduce spatial preference for the ipsilesional space, or improving awareness of the contralesional space to promote patients’ attention (Pierce and Buxbaum, 2002; Paci et al., 2010). More than 18 methods using these general approaches have been put into practice (Luauté et al., 2006) with varying results based on a large number of outcome measures. Although the reported quality is moderate for most of the RCTs in neglect rehabilitation (Paci et al., 2010), some interventions appear to be more promising. Comments have also been made that the effects of treatment are often task-specific or transient and cannot be generalized to daily functioning (Pierce and Buxbaum, 2002; Bowen et al., 2007). Due to a lack of evidence, it is also hard to report which approach is the optimal recommendation for clinical practice (Luauté et al., 2006), and interestingly, professional therapists rarely use these scientifically proven interventions (Petzold et al., 2012).

Many RCTs have employed “pencil-and-paper” tasks, including line bisection, cancelation tasks, copying, and drawing, as treatment outcomes for UN. One of the commonest tests, and one that has been used extensively as an outcome measure for UN, is the Behavioral Inattention Test (BIT) (Bowen et al., 1999, 2007). This is a criterion-referenced test for UN or visual inattention in patients suffering from stroke or brain injuries, comprising two parts: the conventional and the behavioral subtests (Halligan et al., 1991). The conventional subtests include six traditional paper-and-pencil tasks: line crossing, letter cancelation, star cancelation, figure copying, line bisection, and representative drawing. The behavioral subtests consist of nine simulated daily living tasks: picture scanning, telephone dialing, menu reading, article reading, telling and setting the time, coin sorting, address and sentence copying, map navigation, and card sorting. Both parts can be used separately in clinical for impairment and function level assessments, and it has been recommended as a good predictor of functional performance in daily living with good construct and predictive validity (Hartmanmaeir and Katz, 1995).

The aim of this study was to develop a systematic review to assess the effectiveness of rehabilitation for UN as measured by the BIT and to evaluate the effects of the interventions reported in the RCTs using a meta-analysis.

Figure 1 illustrates the selection process. The initial search yielded 201 citations from January 1997 through June 2012. After removing duplicates, 153 citations remained. Based on the title and abstract of the articles, 32 potentially relevant articles were selected. After careful evaluation by the reviewers, we identified 25 clinical trials (Wiart et al., 1997; Robertson et al., 2002; Harvey et al., 2003; Pizzamiglio et al., 2004; Katz et al., 2005; Fong et al., 2007; Nys et al., 2008; Schroder et al., 2008; Ertekin et al., 2009; Luukkainen-Markkula et al., 2009; Polanowska et al., 2009; Serino et al., 2009; Song et al., 2009; Tsang et al., 2009; Saevarsson et al., 2010; Turton et al., 2010; Ferreira et al., 2011; Kamada et al., 2011; Kim et al., 2011; Làdavas et al., 2011; Mizuno et al., 2011; Welfringer et al., 2011; Gorgoraptis et al., 2012; Ianes et al., 2012; Koch et al., 2012) to be included in the final assessment. Of these, 12 articles were included in our final review (Robertson et al., 2002; Harvey et al., 2003; Fong et al., 2007; Nys et al., 2008; Luukkainen-Markkula et al., 2009; Serino et al., 2009; Tsang et al., 2009; Turton et al., 2010; Ferreira et al., 2011; Làdavas et al., 2011; Mizuno et al., 2011; Koch et al., 2012) with the others excluded because the BIT was not used as the primary outcome measure.

The quality of all 12 RCTs was fair to good (Table 1). Four (33.3%) were identified as of fair quality as their scores were below six in the scale. Two studies (Mizuno et al., 2011; Koch et al., 2012) used double-blind designs whereas others were mostly single-blind.

Our systematic review indicates that there is modest evidence for the use of PA to reduce UN in stroke, with immediate and long-lasting effects, and eye patching as shown by BIT-C scores for immediate effects. Other studies obtained positive effects from the use of visual scanning training (Ferreira et al., 2011), visuomotor feedback (Harvey et al., 2003), and TBS (Koch et al., 2012). Since Koch et al. (2012) only report the BIT (Total) and not the BIT-C and BIT-B subscale scores, it is impossible to draw any conclusion that rTMS is better than PA in improving the performance of tasks in the BIT-C and the BIT-B for neglect patients as no comparison could be done.

According to this review, PA is inclined to exhibit the highest ES for immediate effects, but this was not statistically significant as the 95% CI crossed over the zero point. The possible neural mechanism underlying the therapeutic effect of PA is that it reduces spatial neglect by enhancing the recruitment of intact brain areas responsible for visuo-spatial output through short-term sensori-motor plasticity pathways (Rossetti et al., 1998; Luauté et al., 2006). Although this technique has produced some improvement in a wide range of neglect symptoms, especially visual (Shiraishi et al., 2010; Mizuno et al., 2011; Rusconi and Carelli, 2012), some contradictory results have also been reported (Ferber et al., 2003; Rousseaux et al., 2006). The inconsistent results are probably due to the lack of comparability of treatment apparatus, treatment duration, the tasks used to assess PA effects, and post-stroke duration. Similar to PA, hemiplegic half-field eye patching is another compensational intervention for neglect which works by blocking the ipsilesional visual field. The initial study by Tsang et al. (2009) demonstrates a significant result with an ES of 0.71 immediately after intervention. More good-quality RCTs are needed to assess its long-lasting effects on UN.

Transcranial magnetic stimulation is a safe and non-invasive procedure to detect or modulate brain activity by passing a strong brief electrical current through an insulated wired coil placed on the skull which generates a transient magnetic field in the brain (Hummel and Cohen, 2006). TBS is a kind of rTMS using a lower stimulation intensity and a shorter time of stimulation to induce long-lasting effects in the cortex (Cárdenas-Morales et al., 2010) which demonstrates a relatively high ES as measured by the BIT total scores discussed in this review. TMS has become a popular method to stimulate the human brain, with rTMS attracting particular interest for its therapeutic potential to modify cortical excitability (Funke and Benali, 2011), which sheds light on the use of the inter-hemispheric rivalry model in explaining the recovery after neglect disorder in stroke patients. According to the literature, rTMS induces and repairs the inter-hemispheric imbalance (a neglect-like behavior) in the left or right posterior parietal cortex in healthy humans (Kinsbourne, 1977, 1994; Oliveri et al., 2001; Rounis et al., 2007). Based on this model, some studies have explored whether the use of inhibitory rTMS over the contralesional hemisphere to reduce the pathological hyperactivity of either hemisphere may be useful in promoting recovery from neglect after stroke with promising results (Oliveri et al., 2001; Brighina et al., 2003; Shindo et al., 2006; Koch et al., 2008; Nyffeler et al., 2009; Song et al., 2009). Compared to traditional standard cognitive intervention, rTMS can accelerate clinical recovery (Oliveri et al., 2001; Shindo et al., 2006; Song et al., 2009; Paik and Paik, 2010). It seems that patients more severely affected at baseline also benefited more from this intervention. However, the small sample size of the TBS study makes it impossible to draw any conclusion based on robust evidence. There may be a publication bias whereby large studies will report small ES whereas small studies will report large ES.

This review cannot determine the best time to commence neglect rehabilitation interventions, because most participants in the studies included here were recruited in either the subacute or chronic phases. Only two studies implemented rehabilitation within 1 month of stroke (Fong et al., 2007; Nys et al., 2008). As most of the spontaneous recovery after stroke happens in the first month (Kerkhoff and Schenk, 2012), further research is necessary to determine the effects of early but specific intervention for UN compared to conventional rehabilitation in order to avoid the confounding effect of spontaneous recovery. Neglect is the best single predictor of long-term functional impairment and poor rehabilitation outcome in the early stage (Jehkonen et al., 2001; Nys et al., 2005). One study (He et al., 2007) based on neuroimaging shows that 2 weeks after stroke, the normally functional connectivity between the left and right dorsal parietal cortex was disrupted, with the degree of breakdown correlated with the severity of left spatial neglect. It is therefore reasonable that patients should start a neglect intervention as soon as possible in the acute stage, in order to avoid non-use of the hemiplegic limbs, by increasing multisensory inputs or stimulation to the ipsilateral brain regions, and thus slowing down the secondary changes in the brain related to neglect. For further research, we also recommend adequate follow-up to maximize the benefits and monitor the persistence of the effect of neglect rehabilitation interventions.

The results of this review confirm that PA appears to be the most common and effective rehabilitation intervention for UN, and that rTMS might be a promising approach for future treatment. As shown by the insignificant long-lasting effects, rehabilitation interventions often had a transient impact and could not be generalized across time to an improvement in daily functioning. All studies faced the same weakness of low power with smaller samples and a limitation in the blindness design. More rigorous studies of various interventions should be done before coming to a firm conclusion.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.