STATISTICS

talk, health promotion and evaluation of readiness for change. Each patient received a printed copy of his or her physical activity prescription. Additionally, patients were encouraged, for the duration of the 14 week intervention period, to perform at least 30 minutes of physical activity at-moderate-intensity at least three days per week. They were given a training dairy to record their exercises and other alternative physical activities e.g. walking, running and garden work during their spare time. Throughout the period, the patients were guided by the physiotherapist to increase their overall activity and general strength. An example of exercise is shown in Figure 5. During the 14 weeks of follow-up the patients were allowed and

encouraged to contact the physiotherapist as many times as they needed.

Figure 5. General physical activity for the physical activity group.

4.5.1 Study I

The study groups at different time points, specifically 6 weeks, 3, 6 and 12 months, and finally 2 years were compared using analysis of covariance adjusted for baseline score, sex, and age.

The effects of surgery next to potential complimentary effects of the post-operative routine at all time points were assessed with repeated-measure analysis of covariance.

Effect sizes of the surgical and post-operative interventions were assessed with Cohen´s d (Cohen, 1992). Cohen´s guidelines for interpretation of effect size are as follow: 0.20 = small effect, 0.50 = medium effect, and 0.80 = large effect.

Analyses of statistical differences between groups over time were conducted on an intention- to-treat principle for the continuous and discrete data. Thus, all patients, regardless of their loss to follow-up, drop out, or non-compliance, remained in the analysis of the group to which they had been randomly assigned. The analysis was supplemented with a sensitivity- Table 3.Statistical methods applied in the different studies.

Statistics applied Study I Study II Study III Study IV Descriptive statistics

Mean, standard deviation ● ● ● ●

Median, inter quartile range ● ●

Frequency (n), percentage (%) ● ● ●

Statistical methods

Effect size ●

Repeated measure analysis of

covariance ●

Cohen´s d ●

Mann-Whitney U test

(Mann-Whitney) ●

Wilcoxon signed ranked test ● ●

Pearson´s Chi-Square x² ●

Categorical regression (CATREG) ●

Friedman´s ANOVA ●

Repeated measures analysis

(ANOVA) ●

Principal component analysis

(PCA) ●

of-missing-data imputation and was considered satisfactory after comparing results of the per protocol analysis of data exclusively from patients with completed data sets.

4.5.2 Study II

To coordinate the offset of the time between the EMG muscle activation recordings and the NME test start and stop times, the following calculations were used (Figure 6):

1. Total EMG recorded time – total NME test clock time = Start time difference.

2. Start time differences + 2 s = EMG recorded data time withdrawn from the start for analyses.

3. NME test clock time stop – 1 s = EMG recorded data stop time used in analyses.

Figure 6. Measuring time between the EMG recordings and test start and stop times.

Linear regression analysis was used to determine the EMG MDF slope during the prone and supine NME tests.

The difference between the two independent groups in Study II was tested with an

Independent t test; or else the Mann Whitney U test was used for continuous or discrete data, respectively, when comparing groups.

Friedman ̓s test or Wilcoxon ̓s rank sign test was used for ordinal data and non-normally-distributed data. Wilcoxon ̓s rank signed tests were used for within-group comparison of right and left side MF EMG.

Start

End

Start

EMG End

EMG Start

Test

Clock time

End Test 1sec

Imbalance between left-and-right-side MDF EMG slopes was calculated as a ratio for every recorded second (division of right-side value/left-side value). The mean of all the transformed ratios was used to represent the imbalance behavior in the muscle groups.

A categorical regression (Van der Kooij, 2004) method was used to analyze nonlinear relations between dependent variable (NME test time), mean of the bilateral right-and-left-side slope of each muscle group during the NME test, NDI, pain (VAS), TSK and self-rated fatigue at the end of the NME test for the CR group.

4.5.3 Study III

Present neck pain intensity (VAS) and self-perceived fatigue (Borg CR-10) were treated as ordinal data, and non-parametric statistical methods were used. For ordinal data, between-groups differences were tested with the Mann Whitney U test, using the differences in scores between baseline and each follow-up occasion (Study III).

A one-way repeated measure analysis of variance (ANOVA) tests, the same entities that take part in all conditions of an experiment (Field, 2012). The one-way ANOVA was used due to the study ̓ s within-subjects design, which involved repeated measures on the same

participants (with several observations over time). To correct for multiple comparison, a Bonferroni post-hoc test was applied.

A four-way repeated measures ANOVA was used to evaluate the changes in the slope of the MDF for the extensor muscles during the NME test in extension, with group (interventions), muscles (SCap, UT, MT), side (ipsilateral, contralateral) and time (baseline, 14 weeks, one year) as factors. Additionally, a three-way repeated ANOVA was used to evaluate changes in the slope of the MDF for the SCM muscle during flexion contraction, with group

(interventions), side (ipsilateral, contralateral) and time (baseline, 14 weeks, one year) as factors.

The EMG amplitude (ARV) of the SCM and SCap muscles was expressed as a percentage change relative to the initial epoch.

Significant differences revealed by ANOVA were followed using post-hoc Student Newman-Keuls (SNK) tests pair-wise for comparing flexion and extension.

4.5.4 Study IV

Principal component analysis (PCA) with oblique rotation was used in study IV. PCA identifies groups or clusters of variables, and the aim was to understand the structure of the set of variables. Oblique rotation was chosen as it was assumed that the components would be correlated. The PCA was iterated until variables fulfilled the criteria for inclusion. The

Kaiser-Meyer-Olkin measure of sampling adequacy (KMO) was used to ascertain whether the sample was appropriate for a factor analysis. In Study IV, three of the most

well-established criteria were used: oblique rotation, the Scree-plot, and the percentage of non-redundant residuals values over 0.05. The KMO measure of sampling adequacy and Bartlett ̓s test of sphericity were used as measure of appropriateness of the PCA. The component loadings were as follows: over 0.71 = excellent; 0.63 - 0.70 = very good; 0.55 - 0.62 = good;

0.45 - 0.54 = fair; and below 0.32 = poor (Tabachnik BG, 2013).

4.5.5 Missing values analysis

Missing data in the questionnaires were dealt with in the following way: for missing items less than 30%, an imputation value was calculated; that is, the mean value of the non-missing item. Questionnaires with more than 30% missing data were excluded from the analysis.

4.6 ETHICAL APPROVAL

The projects were approved by the Stockholm Regional Ethical Review Board (Dnr: 01-396, 2009/1756-31/4, 2011/692-32), and were carried out in compliance with the Helsinki

Declaration.

5 RESULTS

This section presents the main results of the present work. Detailed results for each study are given in the publications and manuscript at the end of the thesis.