Estimating the association between systemic Interleukin-6 and mortality in the dialysis population. Re-analysis of the global fluid study, systematic review and meta-analysis

Background Systemic inflammation, measured as circulating Interleukin-6 (IL-6) levels, is associated with cardiovascular and all-cause mortality in chronic kidney disease. However, this has not been convincingly demonstrated in a systematic review or a meta-analysis in the dialysis population. We provide such evidence, including a re-analysis of the GLOBAL Fluid Study. Methods Mortality in the GLOBAL fluid study was re-analysed using Cox proportional hazards regression with IL-6 levels as a covariate using a continuous non-logarithmic scale. Literature searches of the association of IL-6 levels with mortality were conducted on MEDLINE, EMBASE, PyschINFO and CENTRAL. All studies were assessed for risk of bias using the QUIPS tool. To calculate a pooled effect size, studies were grouped by use of IL-6 scale and included in the meta-analysis if IL-6 was analysed as a continuous linear covariate, either per unit or per 10 pg/ml, in both unadjusted or adjusted for other patient characteristics (e.g. age, comorbidity) models. Funnel plot was used to identify potential publication bias. Results Of 1886 citations identified from the electronic search, 60 were included in the qualitative analyses, and 12 had sufficient information to proceed to meta-analysis after full paper screening. Random effects meta-analysis of 11 articles yielded a pooled hazard ratio (HR) per pg/ml of 1.03, (95% CI 1.01, 1.03), I-2= 81%. When the analysis was confined to seven articles reporting a non-adjusted HR the result was similar: 1.03, per pg/ml (95% CI: 1.03, 1.06), I-2 =92%. Most of the heterogeneity could be attributed to three of the included studies. Publication bias could not be determined due to the limited number of studies. Conclusion This systematic review confirms the adverse association between systemic IL-6 levels and survival in people treated with dialysis. The heterogeneity that we observed may reflect differences in study case mix.