Response to Letter by Manigandan and Charles
Manigandan and Charles make 3 criticisms of our randomized trial of splinting the wrist after stroke.1 In our opinion those criticisms do not survive close scrutiny. We respond to each in turn.
The Splints May Have Been Applied too Late or for an Insufficient Duration to Be Effective
The literature on mechanisms of contracture (which we and others have reviewed)2–4 provides little guidance as to when and how splinting should be applied. It is not implausible that, if splinting had beneficial effects, the magnitude of those effects might depend on when the splinting was first applied, but we see no reason to expect qualitatively different effects of splinting applied 4 weeks after stroke (the mean in our study) or sooner after stroke. Our data support this view. The interaction between effects of splinting (0=no splint, 1=splint) and time since stroke (in weeks) on wrist range of motion at the 4-week follow-up was 0.8° (95% CI, −3.8 to 5.6), suggesting that time had little bearing on the efficacy of hand splinting in this population.
Our interest was in the effectiveness of the widespread practice of splinting the wrist soon after stroke (not long-term splinting), so we chose to implement a 4-week splinting program. The splints were applied for an average of 10 hours each night. It would not be possible to wear splints for longer periods each day without also wearing the splints during the daytime. We chose not to apply splints during the daytime because we were concerned that doing so would prevent attempts at active use of the arm and hand. Thus, the splints were applied for the longest possible duration that was consistent with our aim of investigating splinting soon after stroke and which, in our opinion, would not enforce disuse. It is unlikely that shorter durations (smaller doses) of splinting would produce greater effects than the splinting protocol we instituted.
The Lateral Photographs May Be Associated With Methodological Error or Bias
In our trial, participants were randomly allocated to groups, measurements were blind to allocation, and estimates of effects of splinting were based on differences between groups. Under these circumstances estimates of effects of splinting cannot be biased by systematic errors in the outcome measures. Random measurement error would contribute to uncertainty in estimates of effects of splinting. That is, random measurement errors make confidence intervals of estimates of effects wider. The confidence intervals of estimates of effects of splinting in our trial were sufficiently narrow to rule out clinically worthwhile effects of stretching, indicating that random measurement errors do not preclude meaningful interpretation of our results.
Participants in the Splint Groups Experienced Improvements in Secondary Outcomes
Within-group improvements provide very poor estimates of the effect of an intervention because they do not control for natural recovery, statistical regression or other potential sources of serious bias.5 Estimates based on between-group comparisons, which provide a much higher degree of experimental control, showed that the effects of splinting on secondary outcomes were, at best, small. Only 1 of 20 comparisons between splint and control groups was statistically significant. The most plausible interpretation is that this was a type I statistical error.6
We conclude by clarifying a misunderstanding apparent in the comments made by Manigandan and Charles. We concluded, on the basis of our trial, that the routine practice of hand splinting to prevent muscle contracture during acute rehabilitation after stroke should be discontinued. We did not draw any conclusions about the effects of long-term splinting because the effects of long-term splinting were not evaluated in our study.
Lannin NA, Cusick A, McCluskey A, Herbert RD. Effects of splinting on wrist contracture after stroke: a randomized controlled trial. Stroke. 2007; 38: 111–116.
Herbert R. The passive mechanical properties of muscle and their adaptation to altered patterns of use. Aust J Physiother. 1988; 34: 141–149.
Herbert RD, Jamtvedt G, Mead J, Hagen KB. Practical Evidence-BasedPhysiotherapy. Oxford: Elsevier; 2005.