Version 13

The multifactorial effects of variables were studied. The Mann-Whitney U tests were used for nonparametric analyses of the differences between pairs of groups, and Kruskal-Wallis tests followed by Mann-Whitney U tests were used to examine the differences between the three groups. Pearson correlation analysis was performed between JOA recovery ratio and ages, duration of symptoms, cervical alignment, cervical stenosis (Pavlov's ratio), maximal spinal canal stenosis, the space available for the spinal cord (SAC), preoperative JOA score, SI length, or postoperative JOA score. Spearman correlation analysis was performed between JOA recovery ratio and gender, OPLL type, or severity of preoperative signal intensity (SI). The correlations between the variables were determined with regression analyses using MedCalc version 13.1 (MedCalc, Mariakerke, Belgium). p values <0.05 were considered significant.

Data were analysed and graphs were created with SPSS version 22 (IBM SPSS Statistics, Armonk, NY, USA) and MedCalc version 13.1 (MedCalc Software, Mariakerke, Belgium).
We constructed Bland–Altman plots to visualise agreement between the two measurement techniques [17 , 18 (link)]. After the inclusion of 18 cases, the Bootstrap resampling method [17 ] was used to estimate the number of participants needed in the study. We concluded that the reliability of the 95% limits of agreement would not improve by including >30 participants in the study. One-sample t-tests were used to assess if the mean differences between data obtained with the two methods differed significantly from zero, indicating the presence of a consistent bias. To evaluate the existence of proportional bias, i.e. that the methods did not agree equally through the range of measurements, the difference between the methods was regressed on the average of the two methods. To evaluate the effect of the facemask, we compared the means of the different breathing parameters from the VSP recording with and without facemask using paired sample t-tests.

Data were expressed as numbers/proportions, mean ± SD or median (interquartile range), as appropriate. Categorical variables were compared using χ² testing. Between-group comparisons were made using either the independent-samples t test or the Wilcoxon test as appropriate.
Cox proportional hazard analysis was performed to identify predictors of NRS progression/regression on a per-patient basis. Baseline variables that were considered clinically relevant (statin equivalent dose) or showed a univariate relationship with the outcome at p < 0.10 were entered into the multivariable analysis. The Kaplan-Meier survival method was used to compare progression rates according to the existence of independent predictors, using the log-rank test.
Intra-observer variability was determined by repeated analysis of 93 segments by the same observer at > 4 weeks time points, blinded to the previous results. The κ test result was 0.90 (95% CI: 0.79–1.00), showing very good agreement [13 (link)].
P-values < 0.05 were considered to be statistically significant. Calculations were performed with MedCalc Version 13.1.2.0 (MedCalc Software, Mariakerke, Belgium).

Statistical analyses were performed using SPSS ver. 17.0 (SPSS Inc., Chicago, IL, USA). Continuous data were expressed as mean and standard deviation and median and interquartile range, as appropriate. Categorical variables were shown as absolute number and percentage. Difference between distances walked during 6MWT and ISWT was verified by the Wilcoxon signed-rank test. Correlations were evaluated by Pearson or Spearman correlation test. The significance level was set at 5%. The agreement analysis between distances walked during submaximal field tests was verified by the Bland–Altman graph and the differences values were plotted against their mean. A ROC curve was constructed to verify the accuracy of 6MWT and ISWT in identifying those patients with peak VO₂ below 20 mL/kg/min functional impairment, defined as values according to Webber classification (Weber et al., 1982 (link)). The statistical difference between the curves was determined by chi-square using the software MedCalc version 13.1.2.0 (MedCalc Software, Ostend, Belgium).