BoTULS: a multicentre randomised controlled trial to evaluate the clinical effectiveness and cost-effectiveness of treating upper limb spasticity due to stroke with botulinum toxin type A

Article history

The research reported in this issue of the journal was commissioned by the HTA programme as project number 02/41/06. The contractual start date was in February 2005. The draft report began editorial review in May 2009 and was accepted for publication in January 2010. As the funder, by devising a commissioning brief, the HTA programme specified the research question and study design. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the referees for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Ipsen Ltd provided the botulinum toxin type A used by the study free of charge. They also provided sponsorship for launch meetings at study sites. The design, analysis and reporting of the study were undertaken independently of Ipsen Ltd. LS, HR, CP, FvW, GF, PS and NS have no competing interests. MB and LG use botulinum toxin regularly in clinical practice. They have received sponsorship from Ipsen Ltd to attend and teach at conferences and meetings, but have no personal financial interest in botulinum toxin or any related product.

Copyright statement

© 2010 Queen’s Printer and Controller of HMSO. This journal is a member of and subscribes to the principles of the Committee on Publication Ethics (COPE) (http://www.publicationethics.org/). This journal may be freely reproduced for the purposes of private research and study and may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NETSCC, Health Technology Assessment, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2010 Queen’s Printer and Controller of HMSO

Overview of stroke

Stroke is a major cause of death and disability in the UK. In England, over 900,000 people are living with the consequences of stroke, 300,000 of them are moderately or severely disabled. 1 The direct cost of stroke to the National Health Service (NHS) is £2.8B per annum, although the overall cost to the economy is much higher (£7B per annum) once informal care costs and lost productivity are taken into account. 1

Upper limb problems following stroke

Upper limb impairments such as muscle weakness, spasticity, poor co-ordination and sensory disturbance are common after stroke. These impairments alone, or in combination, can result in a range of functional limitations. Between 50% and 70% of stroke patients have long-term upper limb functional limitations2–4 and many feel that insufficient attention is paid to upper limb rehabilitation. 5 In contrast, about 80% of stroke survivors are able to walk again. 6

Spasticity

Spasticity is traditionally defined as ‘a motor disorder characterised by a velocity-dependent increase in tonic stretch reflexes (muscle tone) with exaggerated tendon jerks, resulting from hyperexcitability of the stretch reflex, as one component of the upper motor neurone syndrome. 7 A recent definition, which is broader and perhaps more clinically relevant, is ‘disordered sensori-motor control, resulting from an upper motor neurone lesion, presenting as intermittent or sustained involuntary activation of muscles’. 8 However, put simply, spasticity is involuntary overactivity of muscles as a result of damage to the brain or spinal cord.

Spasticity may cause reduced function, pain and deformity, and in the longer term may lead to the development of contracture. 9 Patients with upper limb spasticity can develop abnormal limb posturing such as the classic adducted internally rotated shoulder, flexed elbow, flexed wrist and clenched fist10 (Figure 1). These positions can make washing of the axilla, elbow crease and hand difficult, leading to hygiene problems, which in turn can lead to skin breakdown, infections and pressure sores. 10 Dressing can also be a challenge. Activities such as opening the hand for washing and putting an arm down a sleeve may need the assistance of a carer or the unaffected limb if the patient cannot carry out the task voluntarily with the affected limb. Tasks completed by another person or the unaffected limb are commonly described as ‘passive’ functional tasks. 11

FIGURE 1.

Examples of severe upper limb spasticity. From Brashear and Mayer entitled ‘Spasticity and other forms of muscle overactivity in the upper motor neuron syndrome’. 14 Reproduced with permission from WeMove; New York, NY: 2009.

Although the relationship between spasticity and motor performance remains unclear,12,13 upper limb spasticity is thought to lead to reduced ‘active’ function as overactive muscles around the shoulder and elbow limit reaching activities and spastic finger flexors impair potential finger extension. 10

Upper limb spasticity after stroke is readily recognised clinically, but studies of the prevalence of the condition are lacking. The largest prospective cohort study to date (n = 106) found that 31% of patients had upper limb spasticity at 12 months. 15 A further study (n = 95) found that 20% of stroke patients had upper limb spasticity 5 days after stroke and 18% had upper limb spasticity at 3 months. 16 Despite the lack of prevalence or prospective cohort studies, upper limb spasticity after stroke is an important clinical issue and identification and treatment of spasticity is a key component of stroke rehabilitation. 17

Because of the complex multifaceted definition, measurement of spasticity is a challenge and no tool covers all aspects of the definition. 18 Clinicians and researchers often measure resistance to passive movement (muscle tone) using a clinical assessment scale, e.g. the Modified Ashworth Scale. 19 Resistance to passive movement (muscle tone) in addition to measuring some components of spasticity such as hyperactive tonic stretch reflexes also measures muscle biomechanical/viscoelastic changes. 20 Biomechanical measures have been developed to measure resistance to passive movement21 and the stretch reflex can be quantified using neurophysiological techniques,22 but neither have been widely used in clinical practice or trials.

Botulinum toxin

Botulinum toxins are proteins produced by the bacterial species Clostridium botulinum. Seven serotypes of toxin, labelled A–G, are produced from different strains of C. botulinum. 23,24 All serotypes of toxin can cause botulism, which is a life-threatening condition involving symmetrical flaccid paralysis, autonomic dysfunction and respiratory compromise. 25

The clinical syndrome of botulism was first accurately described in 1820 when Justinus Kerner published his observations of ‘sausage poisoning’. 26 Kerner correctly hypothesised that the syndrome was caused by a biological poison that interrupted nerve conduction. Although he was unable to isolate the toxin, he did suggest that it may be possible to use it therapeutically. 26

Botulinum toxins act by blocking the release of acetylcholine from cholinergic nerves leading to blockage of transmission at the neuromuscular junction and paralysis, and blockage of cholinergic autonomic nerves with resulting autonomic disturbance. 27 When given by intramuscular injection, botulinum toxin causes local and temporary paresis. It has been suggested that botulinum toxin may also block transmission of sensory neurotransmitters providing an analgesic effect independent of muscle relaxation. 27–29 The effects of botulinum toxins are not permanent. Each serotype has a different length of activity with that of botulinum toxin type A being the longest – lasting for 3–4 months. 30

Botulinum toxin was first used clinically in 1977 when it was given by local injection to treat a patient with squint due to overactive ocular muscles. 31 Since then, botulinum toxin has been used to treat various conditions including dystonia, tremor, spasticity, achalasia, migraine, incontinence and sweating. Three preparations of botulinum toxin type A [Dysport^® (Ipsen Ltd), Botox^® (Allergan Inc.) and Xeomin^® (MerzPharmaceuticals)] and one preparation of botulinum toxin type B [Neurobloc^® (Eisai Ltd)] are currently available in the UK. 32 Only Dysport and Botox currently have a licence for treating spasticity. The potency of each product is different and doses are not interchangeable. 32,33

Adverse effects following injection with botulinum toxin are generally mild and transient. 9,34 Local reactions such as erythema, rash and oedema have been reported at the injection site. Migration of toxin into adjacent tissues can lead to weakening of surrounding muscles and autonomic effects; for example, injections into the neck can result in dysphagia and dry mouth. Systemic effects such as fatigue, malaise and flu-like symptoms are reported and systemic transport of toxin to tissues distant to those injected can also occur. Systemic transport of toxin has given cases of botulism-like illness35 or cases of muscle weakness distant from the injection site,32 but these events are very rare.

Review of the evidence for treating upper limb spasticity due to stroke with botulinum toxin

Sixteen randomised controlled trials (RCTs)36–51 and five systematic reviews52–56 evaluating the clinical effectiveness of botulinum toxin as a treatment for upper limb spasticity after stroke have been published. Eight trials44–51 and four systematic reviews53–56 were published following the start of this study in 2005. Details of our literature search strategy, methodological appraisal of the papers, overview of the studies, and a summary of the systematic reviews can be found in Appendix 1.

Trials have reported that treating upper limb spasticity due to stroke with botulinum toxin results in a measurable reduction in resistance to passive movement (muscle tone), which is evident by 1–2 weeks post-treatment. The treatment effect usually lasts for 3–4 months. Although trials varied in the dose and type of botulinum toxin used, the magnitude of initial change in muscle tone/spasticity was approximately a one-point decrease on the Modified Ashworth Scale, which reflects a clinically significant improvement. 36,38–42,44–46,48,49,51

The main benefits of spasticity reduction appear to be in terms of improved global patient/physician ratings36,38,41,42,44,51 and itemised passive functional tasks, notably hand hygiene. 37,40,42 Only one trial reported an improvement in active upper limb function. 45

Botulinum toxin has also been shown to reduce shoulder pain associated with spasticity,46,47 but its role in preventing or treating other types of upper limb pain associated with spasticity is unclear. Only one trial found that upper limb pain was reduced in those who received botulinum toxin compared with those in the control group. 45 Trials reported no unexpected adverse events, however, the event reporting system was often unclear. No trial considered the cost-effectiveness of treatment.

As the treatment effect of botulinum toxin lasts for 3–4 months, injections may need to be repeated to offer sustained benefit. There is limited evidence to support the continued use of repeated botulinum toxin injections for spasticity reduction. Two RCTs considered the impact of repeat injections demonstrating a decrease in resistance to passive movement following a second injection. 44,51 One of the trials also demonstrated sustained improvement in global ratings and patient-selected goals. 51 Six uncontrolled studies have examined the effects of repeated injections. 57–62 These studies are summarised in Appendix 1, Table 36. Repeat injections decreased muscle tone by a similar amount after each injection and improved passive functional scores by a similar amount to the initial injection. 58–60,62 Two studies measured active function58,61 and one found improvement. 61

Botulinum toxin and upper limb rehabilitation

The use of botulinum toxin to treat spasticity forms only one part of upper limb rehabilitation following stroke. Guidelines for the use of botulinum toxin in the treatment of spasticity recommend that it should be used in combination with a rehabilitation programme to achieve optimal beneficial effect. 9,17,63 It is recommended that the rehabilitation programme should consist of 2–8 weeks of physical and/or occupational therapy. 17

Limitations of previous studies

Previous trials varied in methodological quality, size, type of patients included, muscles treated with botulinum toxin, dose of botulinum toxin delivered and outcome measures used. It was often unclear how randomisation was undertaken, whether blinding was robust and follow-up complete. A number of trials developed outcome measures specifically for their studies which were not assessed for validity or reliability and which focused on passive benefits rather than active function. Participants were significantly younger (average age 52–66 years) than typical stroke patients (the average age of incident stroke is 74 years64). Studies were mostly undertaken in specialist rehabilitation centres and were small, lacking statistical power.

Studies published to date have not attempted to standardise upper limb therapy and the amount and content of the therapy received was usually poorly described. It was also often unclear what concurrent medication or additional antispasticity treatments patients received.

Justification for the current study

Botulinum toxin is increasingly used to treat upper limb spasticity due to stroke. Although botulinum toxin does reduce muscle tone and facilitates activities such as hand hygiene and dressing, the impact of this treatment upon upper limb active function is unclear. In clinical practice, botulinum toxin injections may be repeated every 3–4 months, but the effectiveness of repeat injections has not been adequately studied. It is also important that botulinum toxin is evaluated as part of a rehabilitation programme which is clearly described. In addition, cost-effectiveness of treatment is not fully established.

Multidisciplinary care on a stroke unit is currently the gold standard for stroke care regardless of age or stroke severity. 65 Evaluations of botulinum toxin should recruit participants from stroke services rather than tertiary referral centres to avoid selection bias and ensure results are applicable to routine care.

The Botulinum Toxin for the Upper Limb after Stroke (BoTULS) trial was designed to evaluate the clinical effectiveness and cost-effectiveness of botulinum toxin type A plus an upper limb therapy programme for the treatment of post-stroke upper limb spasticity.

Study design

The BoTULS trial was a multicentre open-label parallel-group RCT comparing the clinical effectiveness and cost-effectiveness of botulinum toxin type A plus an upper limb therapy programme with the upper limb therapy programme alone for the treatment of upper limb spasticity due to stroke in adults.

Primary objective

1. To compare the upper limb function of participants with spasticity due to stroke who receive botulinum toxin type A injection(s) to the upper arm and/or forearm flexors/hand/shoulder girdle plus a 4-week evidence-based upper limb therapy programme (intervention group) with participants who receive the upper limb therapy programme alone (control group) 1 month after study entry. Upper limb function was assessed using the Action Research Arm Test (ARAT). 66

Secondary objectives

1. To compare the upper limb function, impairment and activity limitation of participants with spasticity due to stroke who receive botulinum toxin type A injection(s) to the upper arm and/or forearm flexors/hand/shoulder girdle plus a 4-week evidence-based upper limb therapy programme (intervention group) with participants who receive the upper limb therapy programme (control group) 1, 3 and 12 months after study entry. Upper limb function, impairment and activity limitation was assessed by: ARAT,66 Nine-Hole Peg Test,67 basic upper limb functional activity questions used in previous upper limb spasticity studies,37,39,41 Modified Ashworth Scale,19 Motricity Index,68 grip strength69 and Barthel Activities of Daily Living (ADL) Index. 70

2. To compare attainment of participant-selected upper limb goals, upper limb pain, and stroke-related quality of life/participation restriction between intervention and control groups at 1, 3 and 12 months. The following measures were used: attainment of participant-selected upper limb goals (1 month only) – Canadian Occupational Performance Measure (COPM);71 upper limb pain – numerical rating scales;72 stroke-related quality of life/participation restriction – Stroke Impact Scale,73 European Quality of Life-5 Dimensions (EQ-5D) measure of health-related quality of life74 and Oxford Handicap Scale. 75

3. To seek the experience and views of participants about treatment.

4. To compare the health-care and social services resources used by control and intervention groups during 3 months following study entry.

5. To report adverse events and compare the use of other antispasticity treatments between intervention and control groups.

6. To investigate the influence of severity of initial upper limb function and time since stroke upon the efficacy of the intervention.

Summary of design of randomised controlled trial

Figure 2 summarises the study method.

FIGURE 2.

Study method.

A list of all case record forms used in the study is shown in Appendix 2.

Setting

The study involved a collaborative network of 12 stroke services in the north of England. Expertise in the management of spasticity and use of botulinum toxin was provided by the regional spasticity clinic based at the International Centre for Neurorehabilitation, Newcastle upon Tyne, UK. This model, i.e. stroke units with close links to a specialist spasticity service, enabled stroke patients to access specialist care (both in terms of stroke and spasticity management) and could be replicated in other settings.

Case ascertainment

Potential participants were identified from a number of sources in each study centre which were components of locally organised stroke services (stroke unit, outpatients, day hospital and community rehabilitation teams). They were given an information leaflet and had an opportunity to discuss the study with a member of the clinical team. Training was given to clinical teams about the project and research governance. A member of the research team then arranged to see the participant to discuss the study. Consent was sought at the screening visit.

Some potential participants were not in contact with rehabilitation or stroke services. Local community stroke clubs and day centres were given information about the study and individuals were able contact a member of the study team directly.

Inclusion criteria

Adults with a stroke more than 1 month previously who had moderate/severe spasticity and reduced upper limb function who fulfilled all of following criteria were eligible:

• age over 18 years

• at least 1 month since stroke

• upper limb spasticity [Modified Ashworth Scale19 > 2 at the elbow and/or spasticity at the hand, wrist or shoulder (there is no validated measure of spasticity at these sites)]

• reduced upper limb function (ARAT66 score 0–56)

• able to comply with the requirements of the protocol and upper limb therapy programme

• informed consent given by participant or legal representative.

Exclusion criteria

• Significant speech or cognitive impairment which impeded ability to perform the ARAT66 assessment.

• Other significant upper limb impairment, e.g. fracture or frozen shoulder within 6 months, severe arthritis, amputation.

• Evidence of fixed contracture.

• Pregnancy or lactating.

• Female at risk of pregnancy and not willing to take adequate precautions against pregnancy for the duration of the study.

• Other diagnosis likely to interfere with rehabilitation or outcome assessments, e.g. registered blind, malignancy.

• Other diagnosis which may contribute to upper limb spasticity, e.g. multiple sclerosis, cerebral palsy.

• Contraindications to intramuscular injection.

• Religious objections to blood products [botulinum toxin type A (Dysport) contains human albumin].

• Contraindications to botulinum toxin type A, which include bleeding disorders, myasthenia gravis and concurrent use of aminoglycosides.

• Use of botulinum toxin to the upper limb in the previous 3 months.

• Known allergy or hypersensitivity to any of the test compounds.

• Previous enrolment in this study.

Screening assessment

Having sought consent, the screening assessment was completed by a study therapist or clinical research associate. The assessment consisted of demographic details, review of medical history and medication, handedness, Abbreviated Mental Test Score,76 Sheffield Aphasia Screening Test,77 prestroke limitations (Oxford Handicap Scale),75 time since stroke, stroke type and subtype,78 self-reported current neurological impairment and activity limitation (Barthel ADL Index70), quality of life (EQ-5D74), assessment of spasticity (Modified Ashworth Scale19) and measurement of upper limb function (ARAT66). Details of antispasticity treatment and concomitant medications were recorded.

Baseline assessment

The baseline visit was undertaken within 2 weeks of the screening visit by a study therapist or clinical research associate. The inclusion/exclusion criteria were reviewed to ensure that the participant was still eligible. Participants underwent a clinical assessment and were asked to complete the following assessments: Modified Ashworth Scale,19 Motricity Index,68 grip strength,69 ARAT,66 Nine-Hole Peg Test,67 upper limb functional activity questions,37,39,41 Stroke Impact Scale73 and upper limb pain. 72 Female participants with child-bearing potential (i.e. those who were not either surgically sterile or at least 1 year post last menstrual period) had to have a negative urine pregnancy test to be included in the study. Such participants agreed to use adequate contraception throughout the study if they were randomised to receive botulinum toxin type A. Participants were randomised once the baseline assessment was completed.

Randomisation

Randomisation was by a central independent web-based randomisation service from the Clinical Trials Unit, Newcastle University, Newcastle upon Tyne, UK. Participants were stratified according to research site and level of upper limb function (ARAT 0–3, ARAT 4–28, ARAT 29–56), and randomised to intervention or control in a 1 : 1 ratio using permuted block sequences.

Botulinum toxin

Participants in the intervention group received botulinum toxin type A (Dysport). Dysport is available as a white lyophilised powder for reconstitution containing 500 units of C. botulinum type A toxin–haemagglutinin complex together with 125 µg of a 20% albumin solution and 2.5 mg lactose in a clear glass vial.

The range of muscles and dosages injected were as described in ‘The management of adults with spasticity using botulinum toxin: a guide to clinical practice’. 9 The maximum dose of botulinum toxin type A (Dysport) that could be administered at any one time point was 1000 units. All injectors were clinicians trained in the assessment and injection of botulinum toxin in the context of upper limb spasticity.

The use of aminoglycosides was prohibited during the study because they enhance the effects of botulinum toxin, thereby increasing the risk of toxicity. Clinicians were advised to use muscle relaxants with caution because the effects of botulinum toxin are enhanced by non-depolarising muscle relaxants. The international normalised ratio of participants taking warfarin was checked before injection. Information about concomitant drug use was given in the patient information sheet and in letters to consultants and general practitioners.

If further treatment was necessary at 3, 6 or 9 months, further injections were provided to those in the intervention group. At each visit a letter was sent to the participant’s stroke physician, general practitioner and physiotherapist. At the 12-month review, participants in both the intervention and control groups who required botulinum toxin were referred to the spasticity clinic.

If during the course of the trial the study therapist decided that a participant in the control group had an unacceptable degree of symptomatic spasticity, further management was discussed with the stroke physician, physiotherapist, occupational therapist and/or a member of the local or regional spasticity team and the participant could then be referred to a local spasticity service for botulinum toxin.

The upper limb rehabilitation programme

Guidelines highlight that it is important that botulinum toxin is not used in isolation but as part of a comprehensive rehabilitation programme. 9,17,63,79,80 Focal reductions in upper limb spasticity from any pharmacological intervention are unlikely to translate into sustained improvements in function or patient-selected rehabilitation goals without a targeted therapy programme.

The upper limb therapy programme was based upon available research evidence from the stroke rehabilitation and skill acquisition literature as well as clinical practice80–98 and consisted of two menus. Participants with ARAT 0–3 received menu 1, which was designed specifically for participants with no active upper limb function. Menu 1 aimed at improving and maintaining range of movement, encouraging active assisted upper limb movement in the context of functional activities, along with hand hygiene and positioning88–93. Menu 2 was for participants with some retained active upper limb movement (ARAT 4–56) and was piloted in a previous study. 99 Following stretching of soft tissues affected by spasticity, this menu specifically concentrated on task-orientated practise aimed at patient-centred goals. Upper limb goals were measured by the COPM. 71 Each menu standardised the category of tasks, the number and order of repetitions as well as the amount of feedback for each session, but within these parameters the therapist was able to tailor the specifics of each activity to the ability of the patient. Manuals and training programmes were developed for both upper limb therapy menus and all therapists were trained in the delivery of the programme.

The upper limb therapy programme was provided by study therapists and each participant received 1 hour per day, twice a week for 4 weeks, in addition to their other rehabilitation needs. The study therapist could transfer participants between menu 1 and menu 2 according to their clinical opinion. Participants were given a written exercise programme, which was based on the content of the face-to-face sessions with the therapist, to carry out by themselves or with a carer (following training) on the weekdays on which they were not attending therapy.

If the participant was already receiving rehabilitation, then the upper limb therapy programme was delivered in that setting, e.g. stroke unit, outpatients, day hospital or home. In each case, the study therapist liaised closely with the rehabilitation team to ensure the participant’s needs were addressed and therapy was well co-ordinated. At the end of the 4-week intervention period participants were given advice by the study therapist regarding maintenance of upper limb function.

Participants were reviewed by the research team every 3 months. If further therapy was required, this was provided by a study therapist. Those in the intervention group could also receive further botulinum toxin type A injections. Participants in both the intervention and control groups who had symptomatic spasticity at the 12-month follow-up appointment were referred to the spasticity clinic.

Participants who made a good recovery before completing the 4-week upper limb therapy programme could be discharged from the programme provided that they had achieved a maximum score on the ARAT66 and achieved their upper limb goals.

Outcome assessments

Outcomes were measured 1 month (+/– 3 days), 3 months (+/– 5 days) and 12 months (+/– 5 days) after study randomisation.

Each outcome assessment consisted of two stages – stage 1 outcome assessment was a self completion postal questionnaire (Barthel ADL Index,70 Oxford Handicap Scale,75 Stroke Impact Scale,73 EQ-5D,74 upper limb functional activity questions37,39,41 and resource utilisation) which was sent to participants 1 week before stage 2. Participants were asked to bring the completed proforma to their stage 2 appointment.

Stage 2 outcome assessments consisted of assessment of upper limb impairment and function (Modified Ashworth Scale,19 Motricity Index,68 grip strength,69 ARAT,66 Nine-Hole Peg Test67 and upper limb pain72) and face-to-face interview seeking participant’s experience and views of the study treatment. Information was sought about side effects, use of other antispasticity treatment and any change in the participant’s concomitant medications. Any new adverse events or changes in existing adverse events that had occurred since the previous visit were sought. The stage 1 questionnaire was checked for completeness. The stage 2 assessment was completed by a study therapist or clinical research associate.

Blinding

Outcome assessments were undertaken by an assessor who was blinded to the randomisation group. Participants and the study therapists who provided the upper limb therapy programme were not blind to the randomisation group. To enable blinding to be achieved, study therapists undertook screening and baseline assessments and provided the upper limb therapy programme in one research centre and undertook outcome assessments in adjacent centres. As it was possible for assessors to become unblinded, at each outcome assessment an evaluation of blinding was performed.

Participant withdrawal criteria

No specific withdrawal criteria were defined for the study. If a participant discontinued the study prematurely (i.e. before completion of the protocol), the primary reason for discontinuation was recorded when given. In all cases the investigator ensured that the participant received medical follow-up as necessary. Withdrawn participants were not replaced.

Study completion/early termination visit

Study completion was the last outcome visit. If a participant discontinued from the study prematurely, every effort was made to perform an early termination visit consisting of all outcome assessments. At the participant’s last study visit details of their completion of the study/withdrawal from the study were recorded. Female participants of child-bearing potential (i.e. those who were not either surgically sterile or at least 1 year post last menstrual period) in the intervention group were asked to undertake a urine pregnancy test.

Safety evaluation

Side effects of botulinum toxin type A are generally mild and transient. Local muscle weakness may occur as a result of toxin spread to nearby muscles. Five per cent experience flu-like symptoms 1 week to 10 days after injection. Pain at the injection site and a dry mouth can occur. Transient dysphagia has been reported. Anaphylaxis rarely occurs. Excessive doses may produce distant and profound neuromuscular paralysis. Respiratory support may be required where excessive doses cause paralysis of respiratory muscles.

The safety of botulinum toxin type A in the treatment of participants with upper limb spasticity post stroke was evaluated by examining the occurrence of all adverse and serious adverse events as defined by the Medicines for Human Use (Clinical Trials) Regulations, 2004. 100 Follow-up of each adverse event continued until the event or its sequelae resolved or stabilised at a level that was acceptable to the investigator.

Study schedule

Table 1 summarises the study schedule.

Resource utilisation and economic evaluation

This is discussed in Chapter 4.

Data completeness

Data quality checks were performed regularly throughout the study. Missing data and data queries were discussed with the local site teams and collected/resolved as possible.

Statistical analysis

Analyses were undertaken on an ‘intention-to-treat’ basis; participants were analysed in the group to which they were randomised. Data were exported from the study microsoft access database to spss for analysis. All available data were analysed, missing data were not imputed.

The primary end point was the ARAT score at 1 month. For each participant it was determined if there had been a significant improvement in function based on the change in ARAT score from baseline. It is suggested that the minimal clinically important difference for the ARAT is 10% of its range (six points);101 however, we estimated that a smaller treatment effect would be clinically beneficial in those with poor initial upper limb function (ARAT 0–3) compared with those with some retained active function (ARAT 4–56).

A successful outcome was defined as:

• a change of three or more points on the ARAT scale for a participant whose baseline ARAT score was between 0 and 3

• a change of six or more points on the ARAT scale for a participant whose baseline ARAT score was between 4 and 51

• a final ARAT score of 57 for a participant whose baseline ARAT score was 52–56.

The proportion of ‘successes’ in each group was compared using Fisher’s exact test and an interval estimate of the effect of the intervention in the form of an approximate 95% confidence interval (95% CI) for the relative risk was calculated.

Secondary outcomes providing binary data were compared using Fisher’s exact test (or chi-squared test if unable to compute exact form). Secondary outcomes providing ordinal or continuous data were compared using the Mann–Whitney U-test (exact form where possible). Two-tailed p-values are reported. All secondary outcomes were analysed using scale score at follow-up and change in scale score from baseline to follow-up. Change in score was believed to be the key analysis and is presented in the Results section. Scale score at follow-up is presented in Appendix 3.

Although the Mann–Whitney U-test gives an indication of statistical significance it does not provide any information about the magnitude of difference between the groups or whether the difference is clinically important. For some outcomes, the presentation of median values was not helpful in determining clinical importance because of the presence of skewness (changes in the tail of the distribution may not result in any change in the median score). To address this we used resampling methods (bootstrapping) to generate an interval estimate of the effect of the intervention on the group means for each outcome (changes in the tails of the distribution will affect the mean score). The 2.5 and 97.5 percentiles of the bootstrap distribution based on 10,000 replications are reported. This analysis was not prespecified in the study protocol because it was not anticipated to be necessary before viewing the results of planned non-parametric tests.

To enable comparison with previous studies, the basic upper limb functional activity questions were also analysed by comparing the proportion of participants in each randomisation group who had improved by one or more points on the scale from baseline.

As a secondary analysis, logistic regression modelling was used to estimate the effect of the intervention on the primary outcome (ARAT ‘success’) adjusting for randomisation strata (research site and baseline upper limb function).

There were two prespecified subgroup analyses. Response to treatment was compared for:

• participants who had a stroke ≤ 1 year ago and those who had a stroke > 1 year ago

• participants with no initial active upper limb function (with a baseline ARAT score of 0–3) and participants with some initial upper limb function (baseline ARAT 4–56).

Subgroup analysis of the primary outcome was undertaken using logistic regression procedures by adding a subgroup by treatment interaction term to a model that already included the main effects. For secondary outcomes, subgroup analysis was only undertaken where the difference between treatment groups in the main analysis had been statistically significant. For these secondary outcomes, resampling procedures were used to estimate the difference between the treatment effects in the two subgroups.

A power calculation was performed before the start of the study using prognosis based methodology. 102 A clinically important treatment effect was defined as a difference in good outcomes between intervention and control groups of 15% where a good outcome was defined as listed above for each ARAT group; it was expected to see 20% of the control group achieve good outcomes and 35% of the intervention group achieve good outcomes. Using Fleiss’s method for a binary outcome103 and inflating the sample size by 10% to allow for attrition, we needed to recruit a total sample of 332 participants to give 80% power to detect a 15% difference in good outcomes assuming a two-tailed test and a significance level of 5%. The study aimed to recruit 50% of the sample from the ARAT 0–3 group and 50% from the ARAT 4–56 group.

Ethical arrangements and research governance

Multicentre Research Ethics Committee approval was granted. For each individual centre, a site-specific approval was obtained from the appropriate local research ethics committee. Research and development approval was obtained from each participating Trust. The trial was commenced subsequent to the UK Medicines for Human Use (Clinical Trials) Regulations, 2004100 and was one of the first investigator-led trials of an investigational medical product to be undertaken in the UK following the introduction of this legislation. Regulatory approval was granted by the Medicines and Healthcare Products Regulatory Agency and the trial was conducted in accordance with the legislation, the International Conference on Harmonisation–Good Clinical Practice (ICH-GCP),104 and the Research Governance Framework for Health and Social Care. 105

A Trial Steering Committee (TSC) and Data Monitoring and Ethics Committee (DMEC) following Medical Research Council guidelines were established. 106 The TSC comprised an independent chairperson, two independent researchers (all of whom had expertise in rehabilitation research and clinical trials), a consumer representative and three members of the study team.

The DMEC was chaired by a clinical academic with expertise in RCTs of complex interventions in the field of stroke. A biostatistician with expertise in multicentre trials, a researcher experienced in running rehabilitation trials and the study statistician were also members.

An informal interim analysis was performed at the request of the DMEC following recruitment of 50% of the required sample size. Only the DMEC had access to these data. They were able to recommend discontinuation of the study if significant ethical or safety concerns arose, or if there was clear evidence of benefit (clinical or statistical).

The study was monitored for compliance with ICH-GCP by an independent monitor from the Newcastle Clinical Trials Unit.

Amendments to the study following commencement

Study recruitment

Between July 2005 and March 2008, 333 participants were recruited to the BoTULS trial. One hundred and seventy were randomised to the intervention group and 163 to the control group. Monthly and cumulative recruitment are shown in Figure 3.

FIGURE 3.

BoTULS trial recruitment: (a) monthly recruitment, and (b) cumulative recruitment.

Two hundred and eight (62%) participants were randomised before July 2007 and entered the trial for 12 month follow-up. The remaining 125 (38%) participants were followed for 3 months. Details of recruitment in each study site are given in Appendix 4.

Study attrition

Figure 4 shows participant flow through the trial. There were nine deaths during the study period and 12 participants withdrew. Reasons for attrition are described in Table 2.

FIGURE 4.

BoTULS participant flow chart.

One participant withdrew consent shortly after randomisation and asked for their data to be excluded from the analyses. Baseline data are therefore presented for 332 participants. The 1-month assessment was completed by 155/163 (95%) participants in the control group and 167/170 (98%) participants in the intervention group. The 3-month assessment was completed by 151/163 (93%) participants in the control group and 163/170 (96%) in the intervention group.

Of the 332 participants randomised into the study, 208 (63%) were enrolled for 12-month follow-up. The 12-month assessment was completed by 92/103 (89%) participants in the control group and 97/105 (92%) in the intervention group. Outcome data were missing at all time points for one study participant only.

Study population

Randomisation groups were well matched at baseline with regard to demography, stroke characteristics and comorbidity (Table 3). The median age of participants was 67 years [interquartile range (IQR) 59–74], 225 (67.8%) were male and the majority were living at home. One hundred and eighty-one participants (54.5%) were randomised within 1 year of stroke.

Randomisation groups were well matched for the distribution, severity and current treatment of upper limb spasticity (Table 4). The majority of participants had spasticity present throughout the upper limb and the median score on the Modified Ashworth Scale at the elbow was two. Previous treatment with botulinum toxin was not an exclusion criterion provided that it was given more than 3 months before study entry and potential participants were prepared to temporarily relinquish further upper limb injection(s) should they be randomised to the control group. Botulinum toxin treatment had been previously received by 27 (16.7%) of the control group compared with 21 (12.4%) of the intervention group. Randomisation groups were also well matched for other measures of upper limb impairment (Table 4).

One hundred and eighty-four participants (55.4%) had no active upper limb function (ARAT 0–3) and 148 (44.6%) had some retained active function (ARAT 4–56) at randomisation. The median initial ARAT in both intervention and control groups was three (Table 5). Most participants had no or poor dexterity and were unable to complete the Nine-Hole Peg Test. Participants experienced moderate difficulty with upper limb functional activities such as putting their arm through a sleeve or opening the hand to clean the palm and the majority were unable to use cutlery as a result of their stroke. The control group had a lower level of participation restriction when assessed using the Oxford Handicap Scale, but it is unlikely that this possible imbalance at baseline is clinically important or had an impact upon outcome assessments.

The median rating for pain was moderate in both groups. The median pain score was 5/10 in both the intervention and control groups (Table 6).

Primary outcome

Improved upper limb function (predefined treatment success on the ARAT) at 1 month was achieved by 30/154 (19.5%) participants in the control group and 42/167 (25.1%) in the intervention group. This difference was not significant (p = 0.232). The relative risk of having a ‘successful treatment’ in the intervention group compared with the control group was 1.3 (95% CI 0.9 to 2.0).

Secondary outcomes

Trial treatments

Botulinum toxin

Participants in the intervention group received botulinum toxin type A injections to the upper limb immediately following study entry, plus repeat injections at 3 , 6 and 9 months if clinically indicated and they remained in the study for 12-month follow-up.

An initial set of injections was received by 164/170 (96.5%) intervention group participants. Six intervention group participants did not receive the planned initial treatment; three withdrew from the study before treatment, two became unwell following randomisation and a decision was taken not use botulinum toxin (although they did receive upper limb therapy) and one participant had insufficient hypertonicity when seen in the study injection clinic. The latter participant was believed to have variable tone, having been assessed as having increased tone at the screening visit. She remained in the intervention group for the purpose of analysis.

At 3, 6 and 9 months, further injections were received by 71/105 (67.7%), 64/105 (61.0%) and 54/105 (51.4%) intervention group participants, respectively. Muscles treated and botulinum toxin type A doses used are shown in Table 17. Muscles treated were also grouped as limb areas (Table 18). The median (IQR) total botulinum toxin type A doses per participant at initial treatment, 3 months, 6 months and 9 months were 200 units (100–300), 300 units (150–400), 300 units (150–450) and 300 units (188–450), respectively.

TABLE 17 - Intervention group botulinum toxin type A treatment

Muscle	Participants injected n (%)	Dose (units) median (IQR)
Flexor digitorum superficialis
Initial	90 (54.9) (n = 164)	100 (50 to 100)
3 months	50 (70.4) (n = 71)	100 (100 to 100)
6 months	46 (71.9) (n = 64)	100 (100 to 100)
9 months	39 (72.2) (n = 54)	100 (100 to 100)
Flexor digitorum profundus
Initial	63 (38.4) (n = 164)	100 (50 to 100)
3 months	37 (52.1) (n = 71)	100 (100 to 100)
6 months	39 (60.9) (n = 64)	100 (100 to 120)
9 months	35 (64.8) (n = 54)	100 (100 to 100)
Flexor pollicis longus
Initial	6 (3.7) (n = 164)	100 (72.5 to 112.5)
3 months	5 (7.0) (n = 71)	80 (50 to 100)
6 months	7 (10.9) (n = 64)	50 (50 to 100)
9 months	7 (13.0) (n = 54)	50 (50 to 80)
Forearm flexors
Initial	17 (10.4) (n = 164)	200 (200 to 300)
3 months	7 (9.9) (n = 71)	300 (100 to 300)
6 months	4 (6.3) (n = 64)	200 (100 to 300)
9 months	0 (0.0) (n = 54)	0 (0 to 0)
Flexor carpis ulnaris
Initial	57 (34.8) (n = 164)	100 (50 to 100)
3 months	29 (40.8) (n = 71)	100 (100 to 100)
6 months	31 (48.4) (n = 64)	100 (100 to 100)
9 months	29 (53.7) (n = 54)	100 (100 to 100)
Flexor carpis radialis
Initial	10 (6.1) (n = 164)	50 (28.8 to 100)
3 months	3 (4.2) (n = 71)	100 (100 to 100)
6 months	1 (1.6) (n = 64)	100 (100 to 100)
9 months	1 (1.9) (n = 54)	100 (100 to 100)
Biceps brachii
Initial	125 (76.2) (n = 164)	100 (50 to 100)
3 months	55 (77.5) (n = 71)	100 (100 to 100)
6 months	47 (73.4) (n = 64)	100 (100 to 150)
9 months	41 (75.9) (n = 54)	100 (100 to 175)
Brachioradialis
Initial	25 (15.2) (n = 164)	100 (50 to 100)
3 months	13 (18.3) (n = 71)	100 (100 to 100)
6 months	8 (12.5) (n = 64)	100 (100 to 100)
9 months	5 (9.3) (n = 54)	100 (100 to 150)
Pronator teres
Initial	2 (1.2) (n = 164)	100 (100 to 100)
3 months	0 (0.0) (n = 71)	0 (0 to 0)
6 months	0 (0.0) (n = 64)	0 (0 to 0)
9 months	0 (0.0) (n = 54)	0 (0 to 0)
Pectoralis major
Initial	9 (5.5) (n = 164)	100 (50 to 100)
3 months	5 (7.0) (n = 71)	100 (50 to 100)
6 months	1 (1.6) (n = 64)	200 (200 to 200)
9 months	2 (3.7) (n = 54)	150 (100 to 200)

TABLE 18 - Intervention group botulinum toxin type A treatment grouped as limb area

FCR, flexor carpis radialis; FCU, flexor carpis ulnaris; FDP, flexor digitorum profundus; FDS, flexor digitorum superficialis; FPL, flexor pollicis longis.

Where there were three observations, the 25th percentile was calculated as the mean of the first and second value listed in order and 75th percentile as the mean of the second and third values listed in order.

Limb area	Participants injected n (%)	Dose (units) median (IQR)
Hand only (FDS and/or FDP and/or FPL)
Initial	19 (11.6) (n = 164)	100 (50 to 300)
3 months	9 (12.7) (n = 71)	150 (100 to 250)
6 months	8 (12.5) (n = 64)	100 (100 to 188)
9 months	5 (9.3) (n = 54)	200 (100 to 200)
Wrist only (FCU and/or FCR)
Initial	3 (1.8) (n = 164)	50 (50 to 125)a
3 months	1 (1.4) (n = 71)	10 (100 to 100)
6 months	3 (4.7) (n = 64)	100 (100 to 100)
9 months	2 (3.7) (n = 54)	100 (100 to 100)
Elbow only (Biceps and/or Brachioradialis)
Initial	37 (22.6) (n = 164)	100 (50 to 100)
3 months	10 (14.1) (n = 71)	100 (50 to 163)
6 months	6 (9.4) (n = 64)	100 (88 to 113)
9 months	6 (11.1) (n = 54)	100 (88 to 225)
Shoulder only (Pectoralis major)
Initial	2 (1.2) (n = 164)	75 (50 to 100)
3 months	0 (0.0) (n = 71)	–
6 months	0 (0.0) (n = 64)	–
9 months	1 (1.9) (n = 54)	100 (100 to 100)
Hand and wrist
Initial	13 (7.9) (n = 164)	300 (175 to 300)
3 months	6 (8.5) (n = 71)	300 (275 to 325)
6 months	6 (9.4) (n = 64)	300 (300 to 338)
9 months	5 (9.3) (n = 54)	400 (300 to 425)
Hand and elbow
Initial	45 (27.4) (n = 164)	300 (150 to 400)
3 months	21 (29.6) (n = 71)	300 (300 to 500)
6 months	19 (29.7) (n = 64)	400 (200 to 500)
9 months	13 (24.1) (n = 54)	300 (250 to 425)
Wrist and elbow
Initial;	7 (4.3) (n = 164)	100 (100 to 300)
3 month	2 (2.8) (n = 71)	150 (100 to 200)
6 month	3 (4.7) (n = 64)	150 (125 to 175)a
9 month	5 (9.3) (n = 54)	150 (125 to 200)
Hand and wrist and elbow
Initial	29 (17.7) (n = 164)	400 (250 to 400)
3 months	17 (23.9) (n = 71)	400 (400 to 500)
6 months	18 (28.1) (n = 64)	450 (400 to 613)
9 months	16 (29.6) (n = 54)	450 (400 to 575)
Hand and wrist and elbow and shoulder
Initial	4 (2.4) (n = 164)	500 (363 to 600)
3 months	2 (2.8) (n = 71)	700 (600 to 800)
6 months	0 (0.0) (n = 64)	–
9 months	0 (0.0) (n = 54)	–
Other
Initial	5 (3.0) (n = 164)	200 (150 to 300)
3 months	3 (4.2) (n = 71)	150 (150 to 250)a
6 months	1 (1.6) (n = 64)	700 (700 to 700)
9 months	1 (1.9) (n = 54)	850 (850 to 850)

The trial protocol allowed for participants in the control group to be considered for treatment with botulinum toxin injections if they had an ‘unacceptable degree of symptomatic spasticity’. In addition, it was possible for trial participants to be referred to routine spasticity/botulinum toxin services by local health-care providers outside the study.

In the initial 3-month study period botulinum toxin was received by 4/163 (2.5%) control group participants. Only one participant in the control group received treatment before the study 1-month outcome assessment. Eight (7.8%) control group participants commenced botulinum toxin treatment after 3 months. Figure 5 shows the use of botulinum toxin in both intervention and control groups during the study period. All participants treated with botulinum toxin in the control group remained in the control group for the purposes of analyses.

FIGURE 5.

Summary of botulinum toxin treatment in both randomisation groups.

At the end of the study period (3 or 12 months) participants in both groups whom research therapists felt would benefit from botulinum toxin treatment were referred to local spasticity services. Following the last outcome assessment 87/170 (51.2%) in the intervention group and 67/163 (41.1%) in the control group were referred to a spasticity service for botulinum toxin.

Trial safety evaluation

For clinical trials using investigational medicinal products, trial safety is evaluated by examining the occurrence of all adverse events as defined by the Medicines for Human Use (Clinical Trials) Regulations. 100 All serious adverse events were assessed for causality and expectedness, and were immediately reported to the trial sponsor. Systems were in place for expedited reporting of suspected unexpected serious adverse reactions to the appropriate bodies. Annual safety reports were submitted to the Medicines and Healthcare Products Regulatory Agency and the Multicentre Research Ethics Committee as required.

Safety data were analysed according to treatment received. Serious adverse events occurred in 34/156 (21.7%) participants who had not received botulinum toxin type A during the study and in 36/176 (20.5%) participants who had received botulinum toxin type A treatment. However, for two participants who received botulinum toxin type A during the study, the serious adverse events occurred before receipt of injections so they were included in the ‘no botulinum toxin’ group for the purposes of the analyses.

Fifty serious adverse events were reported from the ‘no botulinum toxin’ group and 52 from the ‘received botulinum toxin’ group. These events were summarised and categorised blinded to treatment received (Table 21). The most commonly reported serious adverse events were neurological events, including further strokes and seizures. Several musculoskeletal and respiratory events were also reported from both groups. There were no statistically significant differences between the groups for any serious adverse event type.

All serious adverse events were assessed for their relationship to the study drug at the time of reporting. Only one event was reported as potentially related to botulinum toxin type A. This was an event recorded as dysphagia of unclear cause. As this is a known adverse reaction to botulinum toxin type A it was reported as a suspected serious adverse reaction. However, when further information became available about this event, it was assessed as unrelated to the study drug. No suspected unexpected serious adverse reactions were reported

Adverse events occurred in 70/156 (44.9%) participants who had not received botulinum toxin type A during the study and 90/176 (51.1%) participants who had received botulinum toxin type A treatment. However, for the participants who received botulinum toxin type A during the study, six had events that occurred before receipt of toxin and three had events that occurred both before and after botulinum toxin injection(s). For the purposes of the analyses those that had events before they received botulinum toxin type A were included in the ‘no botulinum’ group. Participants that had events both before and after botulinum toxin type A were included in the ‘received botulinum toxin’ group, which was a pragmatic decision as analyses were performed in exclusive categories.

One hundred and fourteen adverse events were reported from the ‘no botulinum toxin’ group and 147 from the ‘received botulinum toxin’ group. Adverse event descriptions were summarised and categorised blinded to treatment received (Table 22). It was not intended to calculate relative risk for any of the individual adverse events (only the categories), but on examining the data it appeared that there were several more ‘chest infections’ and ‘general malaise/flu-like/cold symptoms’ in the ‘received botulinum toxin’ group. A relative risk estimate was therefore calculated for these. For chest infection, the botulinum toxin relative risk was 1.4 (95% CI 0.5 to 3.5) and for general malaise/flu-like/cold it was 7.6 (95% CI 1.8 to 32.3). General malaise/flu-like/cold symptoms are recognised side effects of botulinum toxin type A.

As with serious adverse events, all adverse events were assessed for their relationship to botulinum toxin type A at the time of reporting. In addition, adverse events were also assessed for their relationship to study upper limb therapy. Twenty-eight events were reported as possibly or probably related to botulinum toxin type A and 16 events were reported as possibly or probably related to study upper limb therapy (Table 23). All other events were assessed as not related to study botulinum toxin or study therapy.

Secondary analysis of primary outcome

A logistic regression model was used to adjust the primary outcome (predefined treatment success on the ARAT at 1 month) for randomisation strata factors (Table 24). Adjustment for research centre and baseline upper limb function (ARAT 0–3, 4–28, 29–56) had very little impact on the magnitude of the estimated effect of botulinum toxin; the relative odds of a ‘treatment success’ changed from 1.39 to 1.41.

Preplanned subgroup analyses

The effect of time since stroke and severity of initial upper limb function upon the primary outcome and statistically significant secondary outcomes was addressed.

Summary

Table 27 summarises the main results of the RCT.

Assessment of costs

Participants’ use of resources was categorised under four general headings: (1) upper limb therapy sessions with or without botulinum toxin type A; (2) other antispasticity medication; (3) management of adverse events attributable to botulinum toxin type A and/or upper limb therapy requiring a hospital contact; and (4) other health-care and social services resource use. A breakdown of the individual items of resource use under these headings and their unit costs are presented in Table 28.

With respect to therapy sessions and botulinum toxin type A, data were collected in the study case record forms on the number of therapy sessions each participant received and the number of 500-unit vials of botulinum toxin type A they used. Each therapy session was staffed by one therapist and lasted for 1 hour. Unit cost data were obtained from Curtis 2007109 and the British National Formulary (BNF). 110

Data on other antispasticity medication were collected in the study case record forms. The forms recorded which drugs the participants were on at baseline, 1 and 3 months. For most participants, the length of time they spent taking a particular drug was straightforward to calculate insofar as if no changes to medications were noted, they were assumed to be taking the drug until such time as a change was noted. Problems arose when, for example, a participant was not taking a drug at baseline, but was noted as taking a drug at a later follow-up time point. Since the precise time at which the participant began taking the drug was not recorded, it was assumed that such participants began taking the drug at the mid-point between baseline and the time at which it was first noted they were taking the drug, i.e. 2 weeks for the 1-month follow-up, and 2 months for the 3-month follow-up. In the few cases where follow-up data were missing, it was assumed that participants remained on the drugs they were taking at baseline.

Another problem regarding antispasticity medication was failure to record the drug dosage that participants were taking. To deal with this, it was assumed that participants were taking the standard dose stipulated in the BNF 2007. 110 If a standard dose was not stated, then it was assumed that participants were taking the maximum recommended dose.

Data on adverse events attributable to botulinum toxin type A and/or therapy which led to a hospital contact were obtained from the study adverse event monitoring forms and from participant responses to specific resource use questions included in the participant assessment questionnaires. A distinction was made between outpatient attendances and events requiring an inpatient stay. Although there were cases of participants encountering hospital services as a consequence of their initial stroke, clinical review showed that none of the hospital contacts were attributable to either therapy or botulinum toxin type A. As a result, hospital resource use due to adverse events did not form part of the cost-effectiveness analysis.

Data on other health-care and social services resource use were obtained from the participant assessment questionnaires, which were administered at baseline, 1 and 3 months. These included questions on participants’ use of health-care and social services, such as day hospitals and day centres, and health-care and social services professional contacts, such as general practitioners and social workers. Unit cost data were mainly obtained from Curtis 2007. 109 Where unit cost data were not available, assumptions were made (see notes accompanying Table 28).

The resource use questions asked participants to report resource use for the 1-month period prior to completion of the questionnaire. This means that participant-reported resource use data were comprehensive for months 1 and 3, but that extrapolations had to be made for month 2. The extrapolation method adopted in the base-case analysis was to assume resource use in month 2 was the same as that in month 3. The impact on the results of alternatively assuming that resource use in month 2 is the same as in month 1 is investigated in the sensitivity analysis.

Assessment of outcome

Participant health-related quality of life was assessed using the EQ-5D,111 which was included in the participant assessment questionnaires. Differences between the randomisation groups at follow-up with respect to EQ-5D scores were investigated using multiple regression analysis of covariance. The average follow-up score of the EQ-5D (the mean of the 1-, 3- and 12-month assessments) was estimated and included as the dependent variable in a linear regression model. Covariates in the model were the ‘baseline EQ-5D score’ and the ‘randomisation group’ (coded 0 for therapy alone and 1 for botulinum toxin type A plus therapy). The regression coefficient estimate for randomisation group represents the difference in mean EQ-5D follow-up scores between the therapy alone and the botulinum toxin plus therapy groups after adjustment for baseline EQ-5D values.

Participant responses to the EQ-5D questionnaire were converted to health-state utility values using the UK tariff values. 112 These values were then multiplied by duration in each health state to estimate quality-adjusted life-years (QALYs). QALYs were estimated using an area under the curve (AUC) approach. 113 To illustrate this approach, consider an individual whose baseline and 3-month quality of life weights are 0.6 and 0.8, respectively. When located on a two-dimensional plane where the y-axis corresponds to the quality of life weight and the x-axis corresponds to time in years, the AUC that joins these two points defines a trapezium, the area of which is equal to the number of QALYs enjoyed by the patient in the 3 months since randomisation. The area of a trapezium with a base width of 3 months (one-quarter of a year) and whose sides are defined as a and b is equal to ½ × ¼ (a + b). In this example, a = 0.6 and b = 0.8. Hence, the AUC corresponds to 0.175 QALYs.

To estimate QALYs, participant responses at baseline and 3 months were used to map out the AUC. An alternative approach would be to use the baseline, 1-month and 3-month points. Although this latter approach provides a more precise estimate of QALYs, the inclusion of the 1-month values increases missing values and consequently decreases the number of participants on which the estimation is based. Therefore, the first approach was used in the base-case analysis, with the possible impacts on the results of the second approach being investigated in the sensitivity analysis.

Assessment of cost-effectiveness

To assess the relative cost-effectiveness of botulinum toxin type A plus upper limb therapy relative to upper limb therapy alone, data on cost and outcome were brought together to estimate an incremental cost-effectiveness ratio (ICER). Specifically, the incremental cost per QALY gained of botulinum toxin type A plus therapy relative to therapy alone was estimated.

The ICER for botulinum toxin type A plus therapy can be located on the cost-effectiveness plane (Figure 6), which is a two-dimensional space in which the origin represents the comparator intervention – in this case therapy alone. The x-axis represents the average difference in effectiveness per participant between botulinum toxin type A plus therapy and therapy alone, while the y-axis represents the average difference in cost per participant between the interventions. The four quadrants are conventionally referred to as points on the compass, namely north-west (NW), north-east (NE), south-west (SW) and south-east (SE). The ICER can be plotted as a point on this plane, with the slope of the line from the origin to the ICER representing the value of the ICER. Treatments with ICERs located in the NW quadrant (more costly, less effective) are said to be dominated by the comparator treatment, whereas treatments with ICERs located in the SE quadrant (less costly, more effective) are said to dominate the comparator treatment. In practice, most new treatments locate in the NE quadrant where increased effectiveness is achieved at increased cost. In this instance the decision to adopt the new treatment will depend upon whether the ICER lies below the acceptable ceiling ratio of the decision-maker. If the decision-maker’s willingness to pay for a unit of effectiveness (λ) is greater than the ICER, then on efficiency grounds the treatment should be recommended for adoption.

FIGURE 6.

The cost-effectiveness plane.

The point estimate of the ICER is subject to uncertainty and it is therefore important that this uncertainty is taken into account. Because of the problems associated with estimating CIs for ratio statistics, the approach of non-parametric bootstrapping is adopted to represent the uncertainty surrounding the ICER estimate. 114 A cost-effectiveness acceptability curve (CEAC), which summarises the evidence in support of botulinum toxin type A plus therapy being cost-effective for a range of values of λ, is also presented. The probabilistic interpretation of this curve should be from a Bayesian perspective. In effect, the CEAC provides information to decision-makers on the level of uncertainty associated with a potential decision to recommend the use of a new or additional intervention. For example, a 0.82 probability of an intervention being cost-effective at a ceiling ratio of £20,000 per QALY implies an error probability (i.e. the probability of making a wrong decision) of 0.18 (1 – 0.82). In making a decision regarding the potential recommendation of a new intervention, the decision-maker must weigh up these probabilities against one another. Alternatively, instead of deciding whether or not to recommend the new intervention on the basis of the currently available evidence, the decision-maker may demand an expected value of perfect information analysis to compare the expected cost of the uncertainty with the value of conducting further research to reduce the uncertainty (see Claxton et al. 115 for more details on expected value of perfect information analysis).

In addition to addressing the uncertainty surrounding the point estimate of the ICER, sensitivity analysis was undertaken to investigate the impact on the results of making alternative assumptions and varying key parameters. As part of the sensitivity analysis, to take into account missing data, an additional set of analyses were carried out using the technique of multiple imputation in which missing data were imputed using the norm package116. Five data sets were imputed using age, sex, place of residence, Barthel ADL score and time between stroke and randomisation as explanatory variables. A point estimate of the ICER and the accompanying CEAC for botulinum toxin type A plus therapy were estimated.

Results of the outcome assessments

Table 29 shows the mean EQ-5D scores over time for the botulinum toxin type A plus therapy and therapy alone groups.

Regression analysis of covariance indicated that there was no significant difference between the groups with respect to mean follow-up EQ-5D scores after adjusting for baseline values.

Results of the cost-effectiveness analysis

The base-case analysis was based on 283 participants who provided EQ-5D responses at baseline and 3 months, of whom 150 were in the intervention group and 133 were in the control group.

There was no significant difference in the mean number of upper limb therapy sessions received by participants in the botulinum toxin type A plus therapy and therapy alone groups (7.64 versus 7.56, respectively; p = 0.46). With respect to use of botulinum toxin type A, the average number of vials used by each patient in the intervention group was 1.01. The numbers of participants taking other antispasticity drugs were 38 in the botulinum toxin type A plus therapy group and 31 in the therapy alone group, with the difference between the groups not being significant (χ² = 0.157; p = 0.692).

A breakdown of other health-care and social services resource use among participants is presented in Table 30.

Chi-squared tests of differences in the proportion of participants in each randomisation group reporting a contact reveal significantly more practice nurse and social worker contacts among participants in the botulinum toxin type A plus therapy group (χ² = 5.115; p = 0.024 and χ² = 4.586; p = 0.032, respectively) and significantly more continence advisor contacts among participants in the therapy alone group (χ² = 4.319; p = 0.038).

The only significant difference in the mean number of contacts among participants reporting a contact is with respect to day hospital contacts, with participants in the therapy alone group having significantly more contacts on average (9.4 versus 3.1, respectively; 95% CI of the difference, 2.42 to 10.19).

Table 31 shows the contribution of botulinum toxin type A costs, upper limb therapy costs, other antispasticity medication costs, and other health-care and social services costs to the overall mean cost per participant.

The overall mean cost per participant was higher in the botulinum toxin type A plus therapy group, although the difference was not significant. There were also no significant differences between the groups with respect to upper limb therapy costs, antispasticity medication costs and other health-care and social services costs. The biggest contributor to total costs for both groups was the cost of other health-care and social services contacts, accounting for 81% in the therapy alone group and 77% in the botulinum toxin type A plus therapy group.

Table 32 shows the point estimate of the ICER of botulinum toxin type A plus therapy relative to therapy alone.

Botulinum toxin type A plus therapy was associated with an incremental cost of £374 and an incremental QALY gain of 0.004, compared with therapy alone. When combined, these data gave an ICER for botulinum toxin type A plus therapy of £93,500 per QALY gained.

Bootstrapping the point estimate of the ICER for botulinum toxin type A plus therapy resulted in 27% of the replications being located in the NE quadrant of the cost-effectiveness plane (more costly, more effective), 21% being located in the SE quadrant (less costly, more effective), and 7% being located in the SW quadrant (less costly, less effective). The largest proportion of the replications (45%) is located in the NW quadrant, where botulinum toxin type A plus upper limb therapy is more costly and less effective, and therefore dominated by, upper limb therapy alone.

Figure 7 shows the CEAC for botulinum toxin type A plus therapy relative to therapy alone. The probabilities that botulinum toxin type A plus therapy is cost-effective at ceiling ratios of £10,000, £20,000, £50,000 and £100,000 per QALY are 0.29, 0.36, 0.41 and 0.42, respectively.

FIGURE 7.

Cost-effectiveness acceptability curve for botulinum toxin type A plus therapy relative to therapy alone.

Sensitivity analysis

The impact on the results of the following sensitivity analyses were explored:

• making an alternative extrapolation assumption for participant reported resource use

• rerunning the analysis using data from participants with complete EQ-5D data at baseline and at 1 and 3 months

• allowing the cost of botulinum toxin type A to fall to zero

• a best–worst QALY analysis in which the impact of alternative assumptions regarding the timing of health-state changes is explored

• rerunning the analysis following multiple imputation of missing data.

The impact on the results of assuming that participant-reported resource use in month 2 is the same as that in month 1 was minimal. Under this assumption the incremental cost of botulinum toxin type A plus therapy increased by £14 to £388, which resulted in the point estimate of the ICER increasing from £93,500 per QALY gained to £97,000 per QALY gained.

The results of the remaining sensitivity analyses, along with those of the base-case analysis, are summarised in Table 33.

Complete EQ-5D data

Rerunning the analysis to include only those participants for whom complete EQ-5D data were available (i.e. responses at baseline and at 1 and 3 months) had little impact on the results. The numbers of participants included were 248, of whom 116 were in the botulinum toxin type A plus therapy group and 132 were in the therapy alone group. The average cost per patient and QALYs enjoyed for therapy alone were £1773 and 0.079 QALYs, respectively. The corresponding figures for botulinum toxin type A plus therapy were £2255 and 0.086 QALYs, respectively. Hence, compared to the base-case analysis, botulinum toxin type A plus therapy was associated with a higher incremental cost (£482 versus £374) and a bigger incremental QALY gain (0.007 versus 0.004), compared with therapy alone. The resultant ICER from combining these data was lower than that in the base-case analysis (£68,857 versus £93,500 per QALY gained, respectively).

Bootstrapping the point estimate of the ICER for botulinum toxin type A plus therapy resulted in 27%, 20%, 7% and 46% of the replications being located in the NE, SE, SW and NW quadrants of the cost-effectiveness plane, respectively.

Figure 8 shows the CEAC for botulinum toxin type A plus therapy relative to therapy alone for participants with complete EQ-5D data at 3 months. The probabilities that botulinum toxin type A plus therapy is cost-effective at ceiling ratios of £10,000, £20,000, £50,000 and £100,000 per QALY are 0.29, 0.34, 0.40 and 0.43, respectively.

FIGURE 8.

Cost-effectiveness acceptability curve for botulinum toxin type A plus therapy relative to therapy alone for participants with complete EQ-5D data at 3 months.

Base-case data and zero cost for botulinum toxin

The impact on the results of allowing the cost of botulinum toxin type A to fall was investigated by considering the extreme assumption that the cost of botulinum toxin type A was zero. Under this assumption the average cost per patient and QALYs enjoyed for therapy alone were £1793 and 0.081 QALYs, respectively. The corresponding figures for botulinum toxin type A plus therapy were £2016 and 0.085 QALYs, respectively. Hence, despite the extreme assumption of a zero cost of botulinum toxin type A, botulinum toxin type A plus therapy still had a positive ICER compared with therapy alone, with the ICER having fallen from a base-case value of £93,500 to £55,750.

Bootstrapping the new point estimate of the ICER for botulinum toxin type A plus therapy resulted in 26%, 22%, 8% and 44% of the replications being located in the NE, SE, SW and NW quadrants of the cost-effectiveness plane, respectively.

Figure 9 shows the CEAC for botulinum toxin type A plus therapy relative to therapy alone assuming a zero cost for botulinum toxin type A. The probabilities that botulinum toxin type A plus therapy is cost-effective at ceiling ratios of £10,000, £20,000, £50,000 and £100,000 per QALY are 0.31, 0.39, 0.42 and 0.43, respectively.

FIGURE 9.

Cost-effectiveness acceptability curve for botulinum toxin type A plus therapy relative to therapy alone assuming a zero cost for botulinum toxin type A.

Base-case data and best–worst QALY assumptions

The AUC approach to estimating QALYs described above assumes that the rate of change in health status between any two points (in this case, EQ-5D tariff values) is linear. For example, if the baseline value is 0.5 and the 3 months value is 0.8, then it is assumed that after 1 month, the health-state value is 0.6, and after 2 months it is 0.7. This is the method most commonly employed in AUC analyses in the literature. 113 However, many other assumptions could be made, each of which may have an impact on the results. In light of the relatively high ICERs associated with botulinum toxin type A plus therapy, it was decided to investigate the impact on the results of the timing of the health-state changes favouring the use of botulinum toxin type A. Specifically, it was assumed in the botulinum toxin type A plus therapy group that participants moved into the health-state value reported at 3 months almost immediately (approximately 3 days or 0.1 of a month) after baseline. This can be regarded as a best-case outcome scenario for botulinum toxin type A. On the flip side, a worst-case outcome scenario for the therapy alone group was defined whereby it was assumed that participants in this group remained in their baseline reported health state for 2.9 months, at which point they moved into the health state value reported at 3 months. The results of these assumptions were that the average number of QALYs enjoyed by botulinum toxin type A plus therapy participants increased from 0.085 in the base case to 0.087, while the average number of QALYs enjoyed by participants in the therapy alone group remained at 0.081. Combining the incremental QALY gain of 0.006 with the incremental cost of £374 gives a best–worst QALY scenario ICER for botulinum toxin type A plus therapy of £62,333.

Bootstrapping the point estimate of the ICER for botulinum toxin type A plus therapy resulted in 30%, 19%, 9% and 42% of the replications being located in the NE, SE, SW and NW quadrants of the cost-effectiveness plane, respectively.

Figure 10 shows the CEAC for botulinum toxin type A plus therapy relative to therapy alone for the best–worst QALY scenario. The probabilities that botulinum toxin type A plus therapy is cost-effective at ceiling ratios of £10,000, £20,000, £50,000 and £100,000 per QALY are 0.29, 0.35, 0.43 and 0.45, respectively.

FIGURE 10.

Cost-effectiveness acceptability curve for botulinum toxin type A plus therapy relative to therapy alone for the best-worst QALY scenario.

Missing data imputed using multiple imputation

Imputing missing data using multiple imputation resulted in estimates of average cost per patient and QALYs enjoyed for therapy alone of £1807 and 0.077 QALYs, respectively. The corresponding figures for botulinum toxin type A plus therapy were £2237 and 0.082 QALYs, respectively. Compared with the base-case analysis, therefore, botulinum toxin type A plus therapy was associated with a higher incremental cost (£430 versus £374) and a bigger incremental QALY gain (0.005 versus 0.004), compared with therapy alone. The resultant ICER from combining these data was lower than that in the base-case analysis (£86,000 versus £93,500 per QALY gained, respectively).

Bootstrapping the point estimate of the ICER for botulinum toxin type A plus therapy over the five imputed data sets resulted in 28%, 20%, 11% and 41% of the replications being located in the NE, SE, SW and NW quadrants of the cost-effectiveness plane, respectively.

Figure 11 shows the CEAC for botulinum toxin type A plus therapy relative to therapy alone when missing data have been imputed. The probabilities that botulinum toxin type A plus therapy is cost-effective at ceiling ratios of £10,000, £20,000, £50,000 and £100,000 per QALY are 0.34, 0.39, 0.42 and 0.46, respectively.

FIGURE 11.

Cost-effectiveness acceptability curve for botulinum toxin type A plus therapy relative to therapy alone following multiple imputation of missing data.

Key findings

Methodological issues

Implications for clinical practice

National clinical guidelines for the management of spasticity in adults using botulinum toxin were updated in January 2009. 63 Our study has followed the guidelines by ensuring that botulinum toxin was not used in isolation, but was part of a multidisciplinary approach to the management of spasticity.

Our results will help to inform clinicians regarding the outcomes that may be improved by treating upper limb spasticity due to stroke with botulinum toxin type A. Management of spasticity should include clearly agreed goals for treatment and for our study the primary goal was improvement in upper limb function. The guidelines suggest that active functional gain is an appropriate goal in some cases. However, our primary analysis and preplanned subgroup analyses did not find any significant clinical improvement in active upper limb function. Other goals suggested in the guidelines include improvement of basic upper limb tasks and pain reduction, and our results support these choices. Although repeated treatment is commonly used in clinical practice and suggested in the guidelines, our study was not able to demonstrate sustained improvement for all outcomes.

Guidelines suggest that botulinum toxin ‘has the potential to reduce the overall cost of ongoing care in people with severe spasticity through the prevention of contracture and deformity, and improved ease of care and handling’. Botulinum toxin type A was not a cost-effective treatment for the patients included in our study, but we acknowledge that we did not examine the longer-term consequences.

Implications for research

We have demonstrated that it is possible to undertake an investigator-led multicentre study of an investigational medicinal product and standardised therapy programme in the UK. Only a small number of multicentre stroke rehabilitation studies have been published to date, although since the Stroke Research Network was established in 2005 numbers are increasing.

Further research is needed to increase our understanding of the natural history and clinical impact of spasticity following stroke, and to explain the relationship between spasticity and functional limitation. Studies are needed to improve the measurement of spasticity and to develop valid measures for all upper limb joints.

Further work is required to establish the optimum dosage and pattern of injections of botulinum toxin type A to treat upper limb spasticity due to stroke and to define the efficacy of repeat injections. The timing of outcome measures should relate to the time when botulinum toxin type A is likely to be optimally effective and measure outcome in the longer term.

There is a need for patients, clinicians and the research community to clearly define important clinical outcomes for patients with upper limb impairment. Further work is needed to develop robust person-centred outcome measures that can be used in national or international multicentre studies.

A national register to record the clinical details of patients who receive treatment with botulinum toxin, their goals and outcomes would be helpful to assist the development of RCTs to identify which patients benefit from this treatment.

Contribution of authors

Dr Lisa Shaw (Clinical Research Associate) was responsible for day-to-day management of the study, delivery of botulinum toxin type A injections and drafting the report. Professor Helen Rodgers (Professor of Stroke Care and chief investigator) designed the RCT and was responsible for the overall management of the study and drafting the report. Dr Christopher Price (Consultant Stroke Physician and Clinical Senior Lecturer) designed the RCT and undertook the initial literature review. Dr Frederike van Wijck (Senior Lecturer, Physiotherapy) contributed to the design of the RCT, designed the upper limb therapy programme, and trained the research therapists. Dr Phil Shackley (Senior Lecturer, Health Economics) designed and conducted the economic analyses. Dr Nick Steen (Principle Research Associate, statistics) was responsible for the trial statistics. Professor Michael Barnes (Professor of Neurological Rehabilitation) contributed to the design of the RCT and provided expertise in spasticity management and the use of botulinum toxin. Professor Gary Ford (Professor of Clinical Pharmacology) contributed to the design of the RCT. Dr Laura Graham (Consultant in Rehabilitation Medicine) provided expertise in spasticity management and the use of botulinum toxin, and was involved in delivery of botulinum toxin type A injections. All authors have contributed to analyses, interpretation of results and drafts of the report.

Disclaimers

The views expressed in this publication are those of the authors and not necessarily those of the HTA programme or the Department of Health.