Statistical methods
Descriptive statistics were calculated as means, standard deviations
(SD), and 95% confidence intervals (CI). Categorical variables were
summarized with relative frequency distribution. All continuous data
were normally distributed and so summarized with mean and standard
deviation (SD). A chi-squared test (or Fisher’s exact test if any
expected count was lower than 5) was used to compare categorical values
between the two groups. An unpaired t-test was performed to compare
continuous variables.
Differences in RAND-36 domains between the two treatment groups were
tested with the non-parametric Mann–Whitney U test. The magnitudes of
group differences were estimated by calculating the effect size (ES;
Cohen’s d ). ES makes it possible to interpret the importance of a
group difference and facilitates comparison across different measures.
ES was calculated as the mean difference between groups divided by the
pooled SD20, and was judged according to the standard
criteria proposed by Cohen: trivial (0.0 to <0.2), small (0.2
to <0.5), medium (0.5 to <0.8), and large
(≥0.8).
In addition, propensity score
matching (PSM) was used to estimate the average treatment effects
between the two groups, using 1:1 nearest neighbour matching based on
the propensity scores. The matched sample size was 96 (48:48). PSM was
used as a sensitivity analysis to assess the robustness of the primary
analysis results. Propensity scores of patients treated with no-touch or
conventional SVG were estimated using a logistic regression model with
age, sex, smoking, hypertension, diabetes mellitus, and creatinine level
as predictors21.
The statistical analyses were performed in version 27.0 of SPSS (IBM,
Armonk, NY, USA) and version 16.1 of Stata (StataCorp, College Station,
TX, USA).