Summary of main results

The primary aim of this review was to investigate the effectiveness of group CCT for improving mobility after stroke. For our primary outcome measure of gait capacity, we found CCT to be superior to other interventions for improving the distance walked on the 6mWT. The minimal clinically‐meaningful improvement on the 6mWT has been estimated at 34.4 m for people later after stroke (Eng 2004) and 61 m for people early after stroke (Perera 2006). Thus, we can be confident that the mean improvement found in the meta‐analysis of 60.86 m represents a real and applicable clinical change. The positive finding for the 6mWT is of functional relevance as it has been shown to be a stronger predictor of the community walking ability than measures of walking speed (Fulk 2010; Mudge 2009b; Rand 2009), which may overestimate community ambulatory ability (Taylor 2006). Furthermore, the 6mWT has been shown to correlate significantly with quality of life after stroke (Muren 2008). We also confirmed that the positive effects were present for people both early and late after stroke suggesting potential for improvement does not necessarily decline. A further positive feature of the primary outcome analyses was that heterogeneity was low. However, we downgraded the GRADE designation to 'moderate' because of the potential for risk of bias in some included studies.

We also found a small favourable effect of CCT in regards to improvements in walking speed, with the magnitude of the between‐group difference (0.15 m per second) being greater than the estimated smallest worthwhile effect of 0.06 m per second (Perera 2006). Perera 2006 suggest that a difference of greater than 0.14 m per second represents a substantial meaningful change for people after stroke. Thus, we can be confident that our results represent real clinical change. Our results suggest that CCT as an intervention has a positive effect on improving independence in walking, with the odds of being fully independent in walking (Functional Ambulation Scale score of 5) after the intervention being significantly greater for intervention participants compared with people allocated to the control intervention.

The evidence for the effectiveness of CCT in improving walking ability after stroke can be considered robust as it is consistently positive across a range of clinic‐based walking measures (6mWT, walking speed, Functional Ambulation Classification) and self‐reported physical function (Stroke Impact Scale ‐ Physical, Rivermead Mobiity Index). The intervention across all studies included a strong emphasis on continuous walking practice. Therefore, the positive results are in line with evidence for intensity and task‐specificity of training, that is to say 'what is trained is what is gained' (Verbeek 2014).

There is some evidence that improvements in walking capacity and ability gained through CCT may also translate into behaviour change. In this updated review, two studies included measures of daily step counts, measured using either a pedometer (Dean 2012) or an ankle‐worn accelerometer (Mudge 2009a). Both trials found that participants who received CCT increased their daily step count to a significantly greater degree than control participants. This is important, as lack of adequate physical activity is linked to increased all‐cause morbidity and mortality (Lollgen 2009) and cardiovascular disease‐specific morbidity and mortality (Thompson 2003), as well as increased risk of stroke (Feigin 2014; McDonnell 2013).

Importantly, CCT may also be an effective method of training for improving cardiorespiratory fitness for people later after stroke. Many studies (Marsden 2013; Marsden 2016; Smith 2012) have reported fitness levels of stroke survivors at less than the minimum requirement for activities of daily living in older adults, that is, 15 mL/kg/min to 18 mL/kg/min (Shephard 2009). An improvement in fitness in the order of magnitude found in our meta‐analysis (2.8 mL/kg/min) is similar to that conferred by exercise interventions with an aerobic component (Marsden 2013). This amount is clinically important as it can improve the exercise reserve of stroke survivors (Ivey 2006) and has the potential to reduce the risk of death (Kodama 2009).

The effectiveness of CCT for improving postural control is less clear. We found significant between‐group differences in favour of CCT for the Activities‐specific Balance Confidence scale and the TUG that exceeded the minimal detectable difference on these measures (Flansbjer 2005; Salbach 2006). However, between‐group differences were non‐significant for the step test, and too small to be clinically worthwhile on the Berg Balance Scale: MD of 1.36 points compared with minimal detectable change of 6.9 points early after stroke (Stevenson 2001), and 4.13 points later after stroke (Flansbjer 2012).

There is some suggestion that providing CCT to people receiving in‐hospital rehabilitation after stroke may reduce length of hospital stay with a mean between‐group difference of 16.35 days. However, the heterogeneity in the study results was higher (I² = 51%) and the difference just failed to reach significance using random effects in the analysis. There are many factors that influence length of hospital stay. A recent individual patient meta‐analysis was conducted where data were pooled from two large multicentre trials investigating the effect of additional weekend therapy for people with stroke. The meta‐analysis identified a range of factors that significantly contributed to length of rehabilitation hospital stay, including age and degree of disability at admission (English 2016). Interestingly, this paper also reported considerable variability in length of stay between individual hospital sites, highlighting the complexity of factors that influence how long people with stroke spend in inpatient rehabilitation. However, a secondary analysis of data from the CIRCIT trial (English 2015) found that when controlling for other influencing factors, receiving CCT as the sole method of physiotherapy service delivery (as compared to usual care physiotherapy) was an independent predictor of a shorter length of stay, in the order of ‐11.6 days (95% CI ‐21.3 to ‐1.9, P = 0.019) (Abstracts Asian Pacific Stroke Congress p6). Reducing length of stay has the potential for significant savings to the healthcare system, but we currently lack high‐quality economic data to establish the cost effectiveness of such an approach.

With regards to adverse events, there were more falls (albeit not statistically different) reported among participants receiving CCT compared with other interventions. Any intervention aimed at improving mobility and balance after stroke carries an inherent risk of causing falls because it is necessary for participants to undertake activities at the limits of their abilities for the interventions to be effective. The greater falls rate in the intervention group is perhaps not surprising considering that the control group was either undertaking interventions that did not expose the participants to an increased risk of falls; for example, seated upper extremity exercise programmes (Blennerhassett 2004; Dean 2000; Pang 2005), stretching (Marigold 2005; Moore 2015), education (Holmgren 2010; Mudge 2009a), or had significantly less risk exposure because they spent significantly less time engaged in physical therapy sessions (English 2015). Nevertheless, it would be pertinent for future studies to more closely examine the link between CCT and falls in therapy.

Carer burden was reported as increased in two studies comparing CCT plus education against no intervention (Harrington 2010; Marsden 2010). It is unknown how the burden was generated and whether it was simply because of receipt of an intervention per se ‐ this requires clarification in future studies.

Based on the results of the two available trials, there is currently no evidence for superior effectiveness of CCT when combined with education. Similarly, there is insufficient evidence for the relative effectiveness of CCT delivered at higher versus lower intensity (based on heart rate targets).

Overall completeness and applicability of evidence

The content of the intervention provided was similar across all studies with many of the same exercises and activities included (see Characteristics of included studies and Table 1). The majority (11) of the trials were conducted with participants later after stroke (more than one year), compared with earlier after stroke (less than one year, six trials) and whilst our subgroup analyses failed to show a significant difference in effect between the two time frames, there was a larger improvement noted in the later group for several measures. The influence of time alone on recovery after stroke remains largely unknown, although it has been estimated to account for between 16% and 42% of improvements in function in the first six to 10 weeks after stroke (Kwakkel 2004). This may mask any potential benefits of CCT over and above usual care in studies conducted with people earlier after stroke.

There were insufficient data available to examine the impact of CCT on sensorimotor impairment after stroke. No studies included measures of movement kinematics or stroke recovery biomarkers such as imaging. Therefore, we cannot determine the degree to which improvements in mobility measures are related to recovery of motor function, specifically 'true neurological recovery' (Levin 2009) versus compensation and overcoming deconditioning.

This updated review included four trials with sample sizes greater than 100 participants. When we pooled data from only these trials, the magnitude of effect for CCT was smaller, but remained statistically significant for the 6mWT. Smaller trials tend to over‐estimate treatment effects (Pereira 2012). The implications of population heterogeneity across large and small trials need to be considered. Furthermore, our 6mWT results were upheld after a sensitivity analysis for trials with low versus unclear/high risk of bias.

The ability to pool data across trials was somewhat limited by the diversity in outcome measures used. Across the 17 included trials, a total of 62 different outcome measures were used. Lack of commonality in outcome measures is a major issue hampering the progress of stroke rehabilitation and recovery research. An analysis of 38 trials in the Virtual International Stroke Archives in 2012 found at least 44 reported outcome measures, with age being the only common metric across trials (Ali 2013). A group of international experts is currently working on addressing this issue with consensus statements being produced as a result (Bernhardt 2016).

Quality of the evidence

The trials were of varying levels of assessed risk of bias. Most commonly, failure to report one or two domains led to a greater overall risk and it remains to be seen if standards of trial conduct and reporting improve in the future. We cannot differentiate between failure to report versus failure to control the risk and this is a potential source of bias in the review process. Hence we downgraded all GRADE determinations as a result of this uncertainty. Three studies achieved 'low' risk ratings in all six domains, confirming that stroke rehabilitation studies can be conducted and reported in an acceptable manner.

Potential biases in the review process

Potential biases in the review process need to be considered in that the three review authors are stroke rehabilitation trialists and take a pragmatic stand on trial design. For example, we did not assess trials as having a risk of bias where the therapist or the participants were not blinded, as we did not consider this possible in these kinds of clinical trials (other than to maintain the participant naive as to which arm of the trial was of interest to the researchers). The definition of CCT was relatively prescriptive and it may be that studies using an alternate circuit format were not included. For example, Kim 2016b compared group CCT with individualised CCT ‐ however their definition of individualised CCT met the criteria for this review's group CCT, thus not offering a useful comparator. It is important that studies such as this are considered in future, as they may help ascertain which aspects of CCT are effective.

Agreements and disagreements with other studies or reviews

This updated review strengthens the findings of our previous review and the systematic review of Wevers 2009, that CCT is an effective intervention for improving walking ability in people after stroke. The updated findings highlight that the benefit of CCT is reported regardless of time after stroke. This update also provides new evidence that CCT may be an effective method of improving cardiorespiratory fitness and increased daily physical activity (step counts).