2.4 Statistical analysis

The number of patients required per group was calculated after assuming a two‐sided hypothesis to be rejected if the p‐value was below 5% and with a power of 80%. Primary outcome was mean peri‐implant bone level change, measured per implant, and a mean difference of 0.5 mm (standard deviation 0.8 mm) was chosen as a meaningful level of difference to be detected. Compensating for a withdrawal rate of 20% resulted in a sample size of 100 patients. Each study center could enroll patients up to a maximum of 33 participants.

The study protocol pre‐determined the statistical tests.

When testing changes over time a non‐parametric statistical approach was applied because of the nature of the data that may not be normally distributed. Wilcoxon signed rank test (exact) was used for continuous data to test if the changes over time within treatment group, for example, marginal bone levels and probing pocket depth, were equal to zero. McNemar's test was used for categorical data, for example, bleeding on probing and plaque. Fisher's exact test was used for categorical data, for example, BOP and implant survival, testing the hypothesis that the percent of BOP and the survival rate are equal in the two treatment groups.