4.8. Statistical analysis

In studies I, II and IV, the data were recorded and calculated using the SOLO statistical software. In study I, the distributions of SF-ICTP and SF-PIIINP by Larsen’s grade were visualized with box plots. In the comparison of means, statistical significance was evaluated by analysis of variance. Statistical significance was evaluated by the Mann-Whitney test when comparing medians. The SF:S ratios were calculated for each patient, and the mean values with confidence intervals were reported as Larsen’s grade values. The correlations between the levels of the different markers of collagen metabolism and disease activity were calculated using Spearman’s rank correlation coefficient. When the significance of the differences between two parameters was assessed by dividing them into two groups, Pearson’s χ2-test or Fisher’s exact test were used as appropriate. In study II, Fisher’s exact probability test was used to evaluate statistical significance when comparing the proportions of patients with or without a need for total joint replacement surgery. In study IV, medians were used to describe the average value of markers of collagen metabolism. Statistical significance was evaluated by the Kruskal-Wallis or Mann-Whitney test when comparing medians. The χ²-test was used when comparing the proportions of patients who needed one or more corticosteroid injections by stable or deteriorated Larsen’s grade.

In study III, the data were recorded and analyzed on a microcomputer using the SPSS statistical software. The significances of the differences between the measurements were determined by Wilcoxon’s test for paired data. When the significance of the differences between two parameters was assessed by dividing them into two groups, crosstabs and Pearson’s χ²-test or Fisher’s exact test were used as appropriate. The correlations were calculated using Spearman’s rank correlation coefficient. The confidence intervals of Spearman’s rank correlation coefficients were calculated with the Confidence Interval Analysis version 1.0 (CIA) in study III.

In study V, the BMDP statistical software was used for statistical analysis. The significance of the differences in Larsen’s score values between the groups formed according to the baseline laboratory values was determined by the Mann-Whitney rank-sum test with Bonferroni correction. Score 20 was used as the cut-off point for the change in Larsen’s score from baseline to 36 months. The age- and sex-adjusted associations of different abnormal laboratory tests or their combinations with radiologic progression were also expressed as odds ratios (OR) with 95% confidence intervals (CI) calculated by logistic-regression analysis.

For all the tests in all the studies, p-values < 0.05 were considered significant.