Optimising the methods of evaluation of the effectiveness of botulinum toxin treatment of post-stroke muscle spasticity.

T he effectiveness of botulinum toxin type A (BtxA) in reducing poststroke muscle hypertonia and its complications has been demonstrated in adult patients in at least eight large randomised controlled trials (RCTs). The clinical benefits were reported when diverse treatment protocols and different dose schedules and injection techniques were used for the management of the same clinical indication. Nonetheless, some doubt as to the value of this treatment remain. What is the explanation for this apparent paradox? The contradiction may be explained by the inadequate design of these studies and the inherent difficulties with the interpretation of the results of RCTs of a drug treatment that is carried out in the context of rehabilitation for a person with chronic neurological disability. The WHO classification of functioning, disability, and health provides a useful framework for understanding the complex relation between the statistical significance and the clinical significance of a given treatment and helps to explain the complexity of rehabilitation research. A statistically significant change at the level of impairment (as a result of the intervention) may not alter the subject’s functional abilities or potential for social participation. The outcome of treatment with BtxA is clinically significant only if it results in the full or partial attainment of functional goals that are meaningful to the recipient of this treatment. An example of such outcomes is the achievement of safe walking and prevention of falls in a patient with dynamic foot equinus after BtxA injections into the calf muscles. Successful treatment reduces muscle tone in the ankle plantar flexors and allows correct placement of the foot in stance (which ensures stability) and adequate foot clearance of the ground in the swing phase of the gait cycle (which prevents tripping over). The major RCTs on the effectiveness of BtxA in spasticity that have been published to date suffer from methodological problems, especially in relation to the choice of the outcome measures and the data analysis and interpretation. The Ashworth scale was often chosen as the primary outcome measure, although the value of this scale as a measure of spasticity is doubtful. Furthermore, the Ashworth scale measures change at the level of impairment. It does not inform the observer about the effect of treatment on functional abilities. Another drawback of the Ashworth scale is that the measurement technique is not standardised. The scale also does not reliably distinguish spasticity from a fixed contracture (which is not amenable to medical treatment). It is clear from the above thesis that the Ashworth scale is not a suitable outcome measure of treatment effectiveness, although it may be used as a screening tool for inclusion for RCTs. Similarly, measurements of the joint range of motion on passive muscle stretch have little or no value in the assessment of the outcomes of treatment with BtxA. Another frequently encountered weakness of the hitherto published RCTs is the inappropriate statistical analysis of the data. For example, the scores derived from the Ashworth scale (which is an ordinal scale) were often analysed with parametric statistics, as if they were the result of an interval level outcome measure. The treatment of muscle spasticity is usually undertaken as part of a holistic rehabilitation programme that seeks to reduce the subject’s disability and to promote their social participation (that is, reduce handicap). Under these circumstances the overall rehabilitation care is likely to be more important in producing functional change than a single specific intervention, such as BtxA injections. Consequently, inadequately designed RCTs may be misleading when used to examine whether treatment with BtxA improves motor function. This is because antispasticity treatment is directed at the level of impairment and reversal or improvement in the impairment—that is, the muscle hypertonia—does not necessarily translate into better functional abilities. Furthermore, the measurement of the treatment outcomes is confounded by the effects of other therapeutic interventions, including physiotherapy, the use of orthotic devices, and so on. Other important confounding factors are the variability in the treatment goals between subjects and the influences of the patient’s personal attributes, personal rehabilitation objectives, physical environment, and social context. The standardised outcome measures used in RCTs do not address the variability in the personal experience or treatment objectives of individual patients which are important determinants of the clinical outcome. It is considered a major advantage of RCTs that the data generated are easily amenable to statistical analysis and statistical significance testing. However, often there is no direct relation between statistical and clinical significance and this is particularly relevant for rehabilitation research. This is because statistical significance does not measure the size or importance of the treatment effect that is deemed ‘‘significant.’’ It only excludes the probability of it occurring by chance. Therefore, a statistically significant result of a given intervention—for example, the reduction of muscle spasticity with BtxA—may not correspond to or reflect a meaningful functional gain to the patient. A statistically significant reduction in muscle tone may even be detrimental in some patients although it may be useful in others with an identical diagnosis or impairment. RCTs are more suited to the study of the effectiveness of treatment at the level of impairment. However, intervention at this level is seldom useful in the context of complex neurological disability except occasionally when the successful treatment of the impairment has a direct linear relation with the desired functional outcome. In addition, the results of RCTs demonstrate the general trends of the study variable in a given study population. The averaged group data of RCTs obscure the poor response to treatment in individual patients. As poor response to treatment may reflect inadequate control for the confounding factors described above, efforts to minimise this effect are necessary before randomisation for group trials. The EDITORIAL 665