Statistical analysis

The variables addressed to evaluate the in vitro lysis capacity of the supernatants from the different strains were analysed using a split-plot model as measurements have an equal variance at all the times and dilutions, and pairs of measurements from the same strain were equality correlated (the Huynh–Feldt condition [37]). The model included the ST classification (ST96 and ST121), the dilution of supernatants (10%, 25%, 50% and 100%), repetition day (1, 2 and 3) and their interactions as fixed effects, as well as the strain (ST) as a random effect.

The data from the flow cytometry and immunohistochemistry studies were analysed using a general linear model (SAS/STAT version 9.2, SAS Institute, North Carolina, USA). The flow cytometry data did not demonstrate normal distribution, and log₁₀ transformation was applied prior to the analysis. The statistical model included only lesion type as the fixed effect. The least square means comparison was made by a Student’s t test. The effect of the control level of the abscess (compact, loose, non-encapsulated) and the spreading level of the lesion (unifocal, multifocal, lobular pattern) on the immunohistochemistry parameters were determined using contrasts as the difference between their various levels (by testing their significant difference from 0; P < 0.05). The effect of the control level of the abscess, the spreading level of the lesion and the MLST (levels: 96, 121) on lymphocyte populations was determined using contrasts as the rate between their different levels (by testing their significant difference from 1; P < 0.05). To test the relation between the immunohistochemistry parameters in lesions and the lymphocyte populations detected by flow cytometry, Pearson’s correlation coefficients (ρ) were calculated.