In addition, periodontal examination, in the form of probing depth (PD) and bleeding on probing (BoP), was performed at pre-screening and at the 12-month follow-up.

Descriptive summary statistics were computed for all parameters, and quantitative parameters were described using mean, standard deviation, median, quartiles, minimum, and maximum. For qualitative variables, absolute and relative frequencies were given. All descriptions were done separately for treatment groups and visits.

The hypothesis was that the change in crestal bone level between baseline and 6 months would be non-inferior for early loading compared to conventional loading. Non-inferiority was defined as a clinically relevant difference of up to 0.3 mm because in the same type of study by Bornstein et al. [35], most implants demonstrated 0.0 to 0.3 mm after 3 years of observation period, and it did not reach statistical significance. The null hypothesis (H0) was therefore that the mean crestal bone loss is >0.3 mm higher with early loading compared to conventional loading. The hypothesis was tested by calculating whether the difference of crestal bone loss was within 0.3 mm between conventional loading and early loading at 6 months (p < 0.05).

There were three types of data sets used: PPS, FAS, and SAS. The SAS included all patients who got implant treatment, including patients who did not meet LC1 and evacuated before randomization. The FAS included all patients who received an implant and who had at least one post-randomization measurement, irrespective of any premature termination or major protocol violations; this set therefore includes the PPS. The PPS includes all patients who completed the study with no major protocol violations.

Based on a two-group one-sided t test with significance level of 0.025, a sample size of 29 patients per group was calculated to have 80 % power to reject the null hypothesis, assuming an expected difference in means of 0, a common standard deviation of 0.4, and a non-inferiority lower limit of 0.3 mm. A subject drop-out rate of 20 % was assumed, giving a sample size of 37 patients per group (total of 74 patients). Calculations were made using nQuery Advisor 6.01.