More randomised controlled studies in speech and language therapy

Randomised controlled trials remain the most widely accepted way of evaluating new treatments. Clinical services such as speech and language therapy, however, have been particularly reluctant to produce randomised controlled trials as evidence of efficacy of treatment.¹,² An evidence base is emerging for the efficacy of a number of speech and language therapy interventions, especially in dysphasia, stammering, laryngectomy, and dysphonia.³ Most interventions, however, have been evaluated by uncontrolled before and after comparisons. One of the first randomised controlled trials in speech and language therapy to evaluate voice therapy in dysphonia appears in this issue.⁴ This trial shows that it is possible to design and carry out randomised controlled trials to examine complex behavioural interventions.

Randomised controlled studies are difficult to apply in some areas⁵; the limitations of such trials in general medicine, surgery, and behavioural therapies are well documented.²,⁶,⁷ Researchers in speech and language therapy have been reluctant to use randomised controlled designs because of the complexity and individuality of human communication behaviour, the consequent difficulties in standardising the content and delivery of specific intervention programmes, and the problem of measuring outcomes of a multifaceted phenomenon. The methods of randomised controlled trials are not appropriate in many areas of current clinical research—for example, exploratory research into factors related to health that influence speech and language behaviour.⁸ Also, it may not be possible for studies of behavioural treatments to meet all the criteria typically used to define the control group in a randomised study or to use blind or placebo treatment. Study of the individual, psychosocial, and cultural components of a speech or language disorder requires good qualitative research to establish promising interventions.

Despite these difficulties, important randomised controlled trials in speech and language therapy have been done in recent years.¹ In the trial described in this issue the application of a randomised controlled design was successful because the authors were able to evaluate interventions with well defined treatment objectives.⁴ The authors wisely confined the study to a number of well established techniques in voice therapy but did not restrict the treatment variable to an unrealistic degree. Voice therapy for dysphonia consists of a series of techniques that aim to correct maladaptive and inappropriate vocal behaviours.⁸,⁹ The selection and emphasis of particular behavioural techniques are based on the specific nature of a person's observed vocal pathophysiology and the contributing psychological, occupational, and social influences.¹⁰ These individual treatment programmes constitute a complex intervention and therefore require careful handling in a randomised controlled trial.¹¹

The evaluation of a complex intervention often requires multidimensional assessment because of the absence of a single standard outcome measure that can encompass the complexity of the disorder. A package of outcome measures to reflect the nature of voice disorders was used in this trial. The human voice is a complex phenomenon that involves anatomical structures, physiological mechanisms, acoustic output—the voice we hear—and associated psychosocial factors that are unique to each person. Hence a variety of assessments is used to measure various aspects of voice production before and after treatment. However, regardless of the phenomenon being evaluated, it is difficult to achieve statistically significant results across a number of related measures, especially in studies with small numbers of participants.¹² Alternatively, a strong correlation between different measures is convincing evidence of change, even if some results in themselves are not significant.¹³

Clinical trials of complex behavioural treatments such as voice therapy are most meaningful when the trials combine measures giving insights into the impact of specific underlying structures or mechanisms (impairment) with measures that reflect disability or handicap. In the evaluation of most behavioural therapies, assessment of disability (limitation of performance) and handicap (loss of social function) often prove more valuable than assessment of impairment (an abnormality in physical or mental functioning). Consequently, reliable questionnaires on disability or handicap in specific diseases have become valuable outcome measures.¹⁴ This point is illustrated in the study under discussion, where a measure of impairment (a laryngoscopic judgment) was not sufficiently refined or sensitive to reflect change in outcome.⁴ However, end points relating to disability and handicap that were judged by the patients differed significantly between the treatment groups.

The use of a randomised controlled design does not prevent further qualitative analysis of “the active ingredients” of the therapy.² Careful documentation of individual differences in a treatment programme is a vital component of evidence based practice. Similarly, tracking the factors that are suspected to contribute to individual patients' disorders (for example, vocal demands or hydration levels) allows retrospective analysis (correlation), so that more patient specific schemes for predicting response to treatment can be developed.

In studies of complex human behaviour it can be difficult to represent accurately the nature of particular treatment programmes and to determine a satisfactory set of outcome measures. To be successful these studies need a pragmatic approach whereby the effectiveness rather than efficacy of the treatment is measured. To produce high quality scientific results in this context is acknowledged as being difficult.²,⁷,¹³ The benefits may, however, be great, because only studies of effectiveness can contribute to a meaningful evidence base and bring about a change in clinical practice.⁷

Papers p 658