GCF sampling GCF was sampled 1 week after the clinical examination in order not to alter the nature

of the GCF. Two non-contiguous diseased sites (i.e. PD and CAL > 5 mm, BoP and no furcation involvement) and two noncontiguous healthy sites (i.e. PD and CAL < 4 mm without BoP and/or MB) were randomly chosen per patient for GCF sampling. After removal of the supragingival biofilm with sterile cotton pellets, the sites were isolated with cotton rolls and gently dried with an air syringe to eliminate the possibility of contamination with saliva. GCF was collected by inserting standard paper strips (Periopaper; Oraflow Inc., Smithtown, NY, USA) of approximately 2 mm into the sulcus/pocket for 20 s. Strips that were visually contaminated with blood were discarded. The GCF volume was measured in a calibrated device (Periotron 8000; Proflow Inc., Amityville, NY, USA) and the readings were converted to an actual volume (ll) by reference to a standard curve. Strips from the four selected sites were immediately placed into separate microcentrifuge tubes and stored at 80C for subsequent assays. Multiplex bead immunoassay The tubes were vortexed for 15 s and centrifuged for 5 min at 1500 g to elute. Samples were analysed using a multi-analyte method by means of a 14-multiplex fluorescent bead-based immunoassay for 14 cyto/chemokines [eotaxin, G-CSF, GM-CSF, interferon (IFN)-c, IL-1b, IL-2, IL-6, IL-7, IL-8, IL-10, IL-12, protein chemoattractant monocyte (MCP)-1, MIP-1a and tumour necrosis factor (TNF)-a] using commercially available kits (Milliplex Human Cytokine/Chemokine; EMD Millipore, Billerica, MA, USA) and a plate reader (Luminex 100TM; EMD Millipore), according to the manufacturer's recommendations. The amount of protein in each sample was extrapolated from standards using an appropriate software (Beadview; EMD Millipore). The

minimum detectable concentration for eotaxin, G-CSF, GM-CSF, IFNc, IL-1b, IL-2, IL-6, IL-7, IL-8, IL-10, IL-12, MCP-1, MIP-1a and TNF-a are 1.2, 0.5, 9.5, 0.1, 0.4, 0.3, 0.3, 1.8, 0.2, 0.3, 10.5, 0.9, 3.5 and 0.1 pg/ml, respectively. The results were reported as concentrations of cyto/chemokines per volume of GCF (pg/ll). Statistical analysis As no precise information regarding the clinical significance of the differences in the levels of the studied mediators between diabetic and nondiabetic groups exists, no formal sample size calculation could be performed for this study. The number of sites for GCF sampling was based on a previous study that evaluated the levels of the same cyto/chemokines in GCF (Tymkiw et al. 2011). Statistical analyses were performed using commercial software (SigmaPlot 11.0; Systat Software, San Jose, CA, USA). Data were first examined for normality by the Shapiro Wilk test and the data that did not achieve normality were analysed using non-parametric methods. The percentage of sites with visible PI, MB, BoP and SUP, mean PD and CAL and levels of HbA1c and FPG were computed for each subject and then averaged across diabetic and non-diabetic groups. GCF volume and the concentration of cyto/chemokine were evaluated in two subsets: healthy and diseased sites. Subsequently, all data were averaged in the diabetic and nondiabetic groups. The differences in age, clinical and glycemic parameters between groups were compared using the Unpaired Student's t-test. The MannWhitney U-test was used to compare the cyto/cytokine levels between diabetic and non-diabetic groups. The v2 test was used to detect differences in the frequencies of gender between groups. Eight models of multivariable logistic regression (MLR) analysis were undertaken to estimate the association between DM and GCF concentrations of relevant chemokines (eotaxin, IL-8 and MIP-1a), innate response-related cytokines

(GM-CSF, IL-6 and TNF-a), T helper-(Th) related cytokines (IFN-c, IL-10, IL-12) and T cell-growth factor cytokines (IL-2 and IL-7) in healthy and diseased sites. For each model, the dependent variable was the presence of DM and, an odds ratio (OR) with a 95% confidence limit was calculated. After using random effects models, the ORs were estimated as follows: OR > 1 nondiabetic subjects had higher levels of cyto/chemokine than the dependent group (i.e. diabetic subjects); OR < 1 non-diabetic subjects had lower levels of cyto/chemokine than the dependent group. The level of significance was set at 5%. Adjustments were made for multiple comparisons when the levels of the 14 cyto/ chemokines were evaluated. An overall p of 0.05 = 1 (1 k)14 was computed, where k was the desired individual p value. Therefore, for cyto/chemokine comparisons, p < 0.0035 was considered statistically significant at p < 0.05. Results Clinical The study was conducted between July 2010 and December 2011. Table 1 presents the mean clinical parameters and the demographic characteristics of the studied population. The only clinical parameter that differed significantly between diabetic and non-diabetic groups was the mean PD. Diabetic subjects showed a higher mean full-mouth PD than non-diabetic subjects (p < 0.05). As expected, the nondiabetic group presented lower levels of HbA1c and FPG than the diabetic group (p < 0.05) (Table 1). The duration of DM reported by the subjects was 6.4 0.9 years. Regarding DM regimens, 18 subjects reported that they were under diet control and using oral hypoglycemic agents, six reported to be under diet control and in use of insulin whereas two subjects reported to be under diet control only. There were no differences between groups for GCF volume, PD and CAL in healthy and diseased sampled sites (p > 0.05).