Multiple Sclerosis Journal - October 2017 - 1545

LE van den Akker, H Beckerman et al.
participation are also reported. The patients' treatment
adherence was determined by assessing the number of
sessions each participant attended.
Power analysis
The sample size calculation was based on a clinically
relevant between-group difference of ⩾8 points on
the CIS20r fatigue scale.22 In total, 90 patients (45 per
treatment group) were needed to detect this clinically
relevant difference with an standard deviation (SD) of
12.7,23 a power of 80%, an alpha of 0.05, and a maximum attrition rate of 20%.
Statistical analysis
Statistical analyses were pre-specified and performed
with SPSS for Windows statistical software package
(Version 22; SPSS, Inc., Chicago, IL). Primary analyses were performed according to the intention-to-treat
principle and consisted of all participants (analyzed as
randomized).
We used mixed linear models to study the effectiveness of CBT. Mixed-model analysis takes into account
the dependency of the repeated observations within
the patient. In total, three models were analyzed: (1) a
crude model, only adjusted for the baseline value of
the particular outcome; (2) a model adjusted for
center; scores of patients who belong to the same
center are correlated,23 therefore randomization was
stratified per center; (3) a model adjusted for covariates that were likely to have a prognostic influence to
improve precision.24 The covariates were determined
a priori: gender, disease severity (EDSS), general
self-efficacy (General Self-Efficacy Scale (GSES)),
anxiety (HADS, subscale anxiety), and comorbidities
(Cumulative Illness Rating Scale (CIRS)) at baseline.
For all models, both an overall effect of CBT and the
effects of CBT at the different measurement points
were estimated. For the latter, time and interaction
between group and time were added to the model.
Time was thereby treated as a categorical variable
represented by dummy variables.
Furthermore, we calculated the absolute risk reduction (ARR), the number needed to treat (NNT), and
the 95% confidence interval (CI) for patients who
showed an improvement of ⩾8 points on the CIS20r
fatigue scale and for patients who scored <35 on the
post-treatment measurement (T16). The NNT is the
average number of patients who need to be treated for
one patient to benefit in comparison to the control
group (in this case an improvement of ⩾8 points or a
score <35 points).25
journals.sagepub.com/home/msj

Results
Participants
The flow chart is displayed in Figure 1. Between
November 2011 and July 2014, 91 patients were randomized, of whom 44 were allocated to CBT and 47
to the MS nurse. Baseline demographic and disease
characteristics are displayed in Table 1.
Three patients (7%) in the CBT study group and eight
(17%) in the control study group dropped out during
the study. In total, 28 (64%) patients in the CBT study
group completed ⩾10 sessions, with a median of 10.5
sessions (interquartile (ICQ1-3): 8.8-11.0). In the control study group, 37 (79%) completed all three MS
nurse consultations, with a median 3 of sessions
(ICQ1-3: 3-3). The demographics of the drop-outs are
available in Supplementary Appendix 3.
During the treatment period, one serious adverse
event (MS relapse confirmed by a neurologist, followed by inpatient treatment) was reported in the
CBT study group, and two (one MS relapse and one
gall bladder surgery) were reported in the control
study group. During the long-term follow-up period,
three serious adverse events (two MS relapses and
one surgery) were reported in the CBT study group
and one (MS relapse) was reported in the control
study group. These events were reported to and judged
by the Medical Ethics Committee not to be directly
associated with the interventions.
Effect on fatigue and societal participation
The mean scores (SD) of the fatigue and participation
outcomes at baseline and follow-up are presented in
Table 2.
The between-group effects of the second model (i.e.
adjusted for center) of CBT on fatigue, societal participation, and secondary fatigue outcomes are displayed in Table 3. The results of the other models are
provided in Supplementary Appendix 4.
Post-treatment CIS20r fatigue scores showed a positive treatment effect for CBT compared to the MS
nurse intervention (β = −6.7 (95% CI = −10.7; −2.7))
(Figure 2) that gradually wore off during the subsequent 8 months (T52: β = 0.5 (95% CI = −3.5; 4.5)).
IPA domain scores did not show between-group differences at any time point, except for a significant
between-group effect on the IPA work and education
domain at T26 (β = −0.3 (95% CI = −0.7; −0.0)). In
addition to the CIS20r fatigue results, the secondary
fatigue measures showed comparable scoring patterns
1545
https://journals.sagepub.com/home/msj
Table of Contents for the Digital Edition of Multiple Sclerosis Journal - October 2017
