Participants

Patients aged 18–64 years with a diagnosis of MS according to the McDonald criteria,11 who had had stable disease for the previous 6 months and troublesome and ongoing muscle stiffness for at least 3 months before enrolment (as shown by a current disability score of at least 4 on an 11 point category rating scale (CRS)), were enrolled.

All physiotherapy regimens or medications likely to affect spasticity were adjusted, where necessary, before study entry and not altered in the 30 days before study start. Patients with active sources of infection or taking immunomodulatory drugs that might affect spasticity (eg, β-interferon, but not azathioprine) were excluded. Other exclusion criteria included fixed tendon contractures, severe cognitive impairment, history of psychosis, major illness, pregnancy and cannabis use in the 30 days before study start.

The study was approved by the Scotland A Research Ethics Committee and is registered at http://www.clinicaltrials.gov (NCT00552604).

Randomisation and masking

Participants were evenly allocated to CE or placebo by means of an interactive voice response system. Computer generated permuted block randomisation was used, stratified by centre, ambulatory status (able to walk or not) and concurrent use of antispasticity medication (yes or no). Active treatment was an extract from Cannabis sativa L (extraction medium ethanol 96%) in soft gelatine capsules, standardised on cannabidiol (range 0.8–1.8 mg) and containing 2.5 mg Δ⁹– tetrahydrocannabinol (THC) as the main cannabinoid (Cannador; Society for Clinical Research, Berlin, Germany). Matched placebo capsules contained the same partial glyceride vehicle (Imwitor 742; Hüls AG, Marl, Germany). The study coordinating team, all investigators and participants were blinded to treatment allocation throughout.

Study conduct

The study consisted of a pretreatment screening period of 1–2 weeks, a 2 week dose titration phase and a 10 week maintenance phase. Total treatment duration was 12 weeks. Participants were assessed at 2, 4, 8 and 12 weeks after the start of treatment.

The starting dose was 2.5 mg Δ⁹THC (one capsule) twice daily. Subsequent doses were individually titrated upwards by 5 mg Δ⁹THC/day every 3 days for up to 12 days to optimise the ratio between therapeutic effect and side effects, and to achieve a maximum daily dose. The maximum allowable total daily dose was 25 mg Δ⁹THC. Participants used diaries to record capsules taken daily.

In the event of intolerable side effects, the daily dose was reduced by one capsule until the side effect(s) resolved. After resolution, one rechallenge with a further dose escalation was required. If the side effect(s) returned, the dose was reduced again, with no further rechallenge allowed.

The primary outcome measure was an 11 point CRS to evaluate perceived change in muscle stiffness after 12 weeks of treatment compared with the premedication phase. At the final visit, participants answered the following question on a symptom questionnaire: “Compared with before the study started, my muscle stiffness over the last week has been…” providing a rating on an 11 point numerical CRS where 0=very much better, 5=no difference and 10=very much worse. Categories 0–3 of the rating scale were classified as ‘relief of muscle stiffness’—that is, as a clinically relevant response.

Secondary outcome measures included further equivalent CRSs measuring perceived relief from body pain, muscle spasms and sleep disturbance compared with pretreatment (at 4, 8 and 12 weeks). Absolute amount of muscle stiffness, body pain, muscle spasms and sleep disturbance (at screening, baseline and 4, 8 and 12 weeks) was also asked for by similar 11 point rating scales stating “Over the last week, the (symptom) I have had is …” providing a rating in categories between ‘no (symptom)’ and ‘extreme (symptom)’. Further patient reported secondary outcome measures included validated disease specific multi-item rating scales measuring aspects of spasticity in MS (MS Spasticity Scale (MSSS-88)),12 the physical and psychological impact of MS (MS Impact Scale (MSIS-29))13 and walking ability (MS Walking Scale (MSWS-12))14 (at screening, baseline, 4 and 12 weeks).

Investigators evaluated severity of disability (Expanded Disability Status Scale (EDSS))15 and all safety parameters at screening and at week 12. Safety was evaluated using laboratory parameters, vital signs and adverse events (AEs), in accordance with good clinical practice.

Calculation of sample size

In the CAMS study, 27% of placebo participants and 42% of CE participants showed relief from muscle stiffness, as defined in the present study. Assuming a two sided test of hypothesis with a power of 80%, a 5% level of significance and an active treatment to placebo ratio of 1:1, about 170 evaluable participants per group would be required to detect a significant improvement in relief rate compared with placebo, according to the sample size approximation of Casagrande et al for Fisher’s exact test.16 With an assumed dropout rate of 15%, it was planned to enrol a total of 400 patients.

An unblinded adaptive interim analysis was planned by means of ADDPLAN software V.4 and performed at an α error level of α=0.0026 after the first 200 patients had completed the study, applying the inverse normal procedure of Wassmer with stopping boundaries of O’Brien and Fleming.17–19 On the basis of this analysis, the Independent Data Monitoring Board recommended reducing participant numbers from 400 to 300 as this sample size was sufficient to maintain a conditional power of 95%. The post-interim results confirmed the validity of this approach.

Statistical analysis

All decisions regarding primary outcome data were finalised by a blind data review panel before unblinding. SAS/STAT software (V.8.02) was used, based on a pre-agreed analysis plan.20 All primary and sensitivity analyses were carried out on the full analysis set (FAS) (ie, all patients enrolled), had at least screening/baseline measurements and had taken at least one dose of study medication. For the efficacy analysis, missing data were replaced by the last observation carried forward approach. In addition, as a sensitivity assessment of the primary efficacy parameter, worst case and multiple imputation substitutions were performed.

The primary analysis of efficacy compared the proportion of patients with relief from muscle stiffness (0–3 on the CRS) in the CE and placebo arms. A logistic regression model was used with fixed factors treatment (binary), ambulatory status at screening (binary), use of spasticity medication (binary) and geographical region (binary dummy coded). A one sided test for superiority of CE over placebo was performed with a prespecified α error level of α=0.024 for the final analysis, incorporating a global error level of α=0.025 combined from the interim and final analyses.

Secondary efficacy endpoints were analysed in the same way as the primary endpoint for numerical rate of change scales or non-parametrically for amount of symptoms in CRS and MS specific questionnaires. The MSSS-88 has eight sections measuring three spasticity related symptoms, effect of spasticity on three aspects of physical function and two aspects of psychosocial function. Analysis included the factors treatment, ambulatory status and antispasticity medication (all binary) using Lehmann’s non-parametric aligned rank test.

The MSIS-29 has two scales measuring the self-reported physical impact (20 items) and psychological impact (nine items) of MS, each with five response options. The MSWS-12 is a self-rated measure of walking ability. This questionnaire was not administered in patients unable to walk and the analysis therefore included only the factor treatment (binary), evaluated using Wilcoxon’s rank sum test.

The MSSS-88, MISIS-29 and MSWS-12 were also analysed using Rasch measurement methods.21–23 This modern psychometric analytic method has the advantage of deriving linear measurements from ordinal patient responses to items, and enabling legitimate analysis at the individual person level.

EDSS data at screening and after 12 weeks of treatment were listed and summarised.

The Cochran–Mantel–Haenszel–χ² test was used to test for differences between the proportion of patients with AEs in the CE and placebo groups stratified for separate visits. Laboratory parameters were compared using Fisher’s exact test.

All secondary and safety parameters where considered statistically significant with a (local) two sided error level of α=0.05.