We also excluded the studies if OSA always find useful information was not diagnosed by measured by polysomnography. For example Kohler 2009 was excluded since OSA was defined by Oxygen Desaturation Index instead of AHI. Some studies that fulfilled our inclusion criteria had to be excluded because values of inflammatory markers were exponentially larger than the values for the same inflammatory marker in all other studies. Tamaki et al. [60], for example, measured the production of TNF-�� by monocytes before and after treatment with CPAP, and the values, when converted, were 1000 times greater than the other studies measuring TNF-�� levels. Intercellular adhesion molecule (ICAM) and interleukin-8 (IL-8) were not included in this meta-analysis because there were not enough studies available to performed meta- analysis.

Moreover oxyhemoglobin desaturation data was not included for the same reason. For studies in which no numerical data accompanied the graphical data, the authors were contacted in order to obtain the data. Authors of one study produced two independent papers that included the same CPAP and inflammatory markers data, so we only included the data from Schiza et al. 2010 [61] and not from Mermigkis et al. 2011 [62]. Statistical analyses were done using RevMan software version 5. Pooled mean difference was calculated using a random effects model for all outcomes due to the high level of heterogeneity present. Heterogeneity was assessed by calculating the Cochrane Q statistic. I2 statistics were also calculated to help quantify the amount of heterogeneity.

An I2 of the following percentages represents different levels of heterogeneity: 25-49% low, 50-74% moderate, and 75-100% high. Measurement units of inflammatory markers we used in the meta-analysis were mg/dl for CRP and pg/ml for IL-6 and TNF-��. If values of any of these markers were not reported in the same standard measurement unit we used, the values were converted to the appropriate unit. Primary principal measures were differences in means of inflammatory markers before and after CPAP treatment. Results A total of 3835 studies were reviewed for inclusion with 23 studies pooled for analysis. The quality of evidence was low (3B-individual case�Ccontrol study) for all 23 studies. A total of 14 studies with 771 patients were pooled for CRP; 9 studies with 209 patients were pooled for TNF-��; and 8 studies with 165 patients were pooled for IL-6 (Figure 1).

The studies measuring key serum inflammatory markers are outlined in Tables 2, ,33 and and44. Figure 1 Study selection methodology. Table 2 Selected studies measuring serum CRP before and after CPAP Table 3 Selected GSK-3 studies measuring serum TNF-�� before and after CPAP Table 4 Selected studies measuring serum IL-6 before and after CPAP C-reactive protein With respect to CRP, study level means ranged from 0.18 to 0.85 mg/dl before CPAP treatment and 0.10 to 0.72 mg/dl after CPAP treatment. Mean differences, at a study level, ranged from ?0.05 to 0.50.