Does CytoSorb Pose Unique Challenges to Pooled Estimates in Meta-Analysis?*

Systematic review and meta-analysis has transformed the way we interpret and apply evidence-based medicine (1). While numerous iterations of the famed evidence pyramid have been proposed, some variation of systematic review and meta-analysis consistently lay atop in the hierarchy, with the assumption that these studies provide the highest validity and applicability. However, more contemporary versions of the evidence pyramid emphasize that systematic review and meta-analysis is a lens through which evidence is appraised, and certainty in the evidence depends on the quality of the primary studies (2). In this issue of Critical Care Medicine, Heymann et al (3) conduct a systematic review and meta-analysis that serves as a case in point. Heymann et al (3) analyzed 17 randomized trials including 855 critically ill patients with hyperinflammatory syndromes assessing the inflammatory marker removal efficacy of the CytoSorb hemoadsorption device. Random-effects meta-analysis demonstrated no significant differences in interleukin (IL)-6, C-reactive protein, or procalcitonin at 1 day following treatment with the device, as well as after 2, 3, and 5 days. No significant differences were demonstrated in other inflammatory markers including IL-1β, IL-8, IL-10, and tumor necrosis factor-α, although the number of studies with available data for quantitative pooling in these was limited. Certainty of evidence for all outcomes was judged to be low or very low, which means that future research will likely yield different estimates of effect. This is despite trial sequential analysis which indicated additional trials would not alter the null findings for IL-6 reduction. We do question the assumptions used in this trial sequential analysis that showed an optimal information size of 46, which is very small. Most meta-analyses of continuous outcomes yield imprecise estimates when their total sample size is below 800 (4). We note that the meta-analysis of IL-6 at day 1 included only 441 patients, and other meta-analyses in this systematic review had even smaller size. Thus, there is high uncertainty, and the evidence is far from being conclusive. Another factor that adds to this uncertainty and makes the interpretation of these results more challenging relates to the heterogeneity of treatment effect caused by variability in the baseline values of the continuous variable. Does the magnitude of change in IL-6 depend on the baseline IL-6 level, in which case meta-analysis assumptions may be violated? CytoSorb is a form of extracorporeal blood purification therapy that consists of a column composed of hemocompatible porous polymer-based adsorber beads with a total surface area in excess of 45,000 m2 (5). While the construct of the device supports adsorption of molecules with medium to large molecular size and those with hydrophobic properties, removal appears to be nonspecific and includes a very wide range of substances (6). Concentration-dependent removal has been claimed by the manufacturer (5), as well as in research publications (6), although empirical evidence appears to be limited. In pigs exposed to supratherapeutic dosages of antimicrobials, substance clearance was greatest in the first 30 minutes of extracorporeal hemoperfusion with CytoSorb compared with a sham extracorporeal circuit, for most antimicrobials assessed (7). Interestingly, while substance removal continued over the next several hours, the calculated clearance gradually declined. This suggests that in a concentration-dependent removal model, the efficiency of the device may depend on the magnitude of substance exposure and declines over time as exposure is reduced (Fig. 1). Although such an effect was observed in a nonhuman experiment with small drug molecules, it is unclear if concentration-dependent removal of inflammatory markers exists in critically ill patients. In the review by Heymann et al (3), the baseline IL-6 concentrations were only reported in 11 of 17 studies, and ranged from untraceable to 2,299 pg/mL (3). Interestingly, the baseline IL-6 values in six of the 11 studies that reported these were within the range of IL-6 values that have been observed in healthy persons (8). Whether IL-6 removal efficiency by CytoSorb in such patients is similar to those with higher baseline IL-6 values cannot be concluded.Figure 1.: Conceptual schema of the hypothesized concentration-dependent removal phenomenon.Should such a phenomenon exist with CytoSorb, wherein the efficiency of the device depends on the magnitude of baseline substance concentrations and declines over time, meta-analysis estimates may not be reliable without a meta-regression that accounts for baseline IL-6 values. In conclusion, due to high certainty and inconclusive evidence about surrogate outcomes, along with a signal of increased mortality, routine use of the CytoSorb device in critically ill patients cannot be recommended. The device should be studied further in trials that evaluate patient-important outcomes (9) and not surrogate inflammatory markers.