Dialysis Dose and Flux as Potential Effect Modifiers of Sphingolipid–Mortality Associations

We read with great interest the report by Lidgard et al., published in JASN, examining circulating sphingolipids and subsequent all-cause and cardiovascular death in patients with kidney failure receiving maintenance hemodialysis.1 We commend the authors for leveraging the HEMO biorepository to address this understudied pathway.2 A chain-length gradient emerged: Long–acyl-chain sphingolipids (e.g., ceramide-16:0) were associated with higher risk, whereas very long–acyl-chain species (e.g., ceramide-22:0) were associated with lower risk. Similar patterns have been reported in community cohorts and are biologically plausible given the diverse signaling roles of ceramides and related sphingolipids.3,4 In a population where conventional lipid-lowering strategies have not improved outcomes, these observations may help refine risk stratification and highlight alternative lipid pathways. Because the analyses were performed in baseline sera from the HEMO trial—a 2×2 factorial randomized study of dialysis dose and dialyzer flux—one additional question seems readily answerable within the existing dataset: Are these sphingolipid–mortality associations consistent across the randomized arms?2 The authors appropriately adjusted for treatment assignment. Still, adjustment does not address effect modification. The randomized design offers a rare chance to test whether delivered dose or membrane flux changes the prognostic meaning of a given sphingolipid profile.2 We would welcome the authors' perspective on a focused interaction analysis. Specifically, it would be helpful to test prespecified interaction terms between selected sphingolipids (for example, ceramide-16:0 and ceramide-22:0, or the long– versus very long–acyl-chain summaries highlighted in the report) and each randomized factor (standard versus high dose; low- versus high-flux membrane) for both all-cause and adjudicated cardiovascular death.2 If precision is limited for the full panel, restricting analyses to a small a priori subset could keep the question sharp and reduce multiple-testing concerns.3 For interpretability, stratified hazard ratios by randomized arm—paired with absolute event rates across sphingolipid quartiles within each arm—could help readers gauge whether any heterogeneity is clinically meaningful. Reporting P values for interaction and confidence intervals for stratum-specific estimates would further support interpretation. Clarifying consistency across dose and flux arms could strengthen the translational interpretation. Uniform associations would support sphingolipids as robust risk markers independent of modifiable dialysis parameters, whereas heterogeneity might suggest treatment-responsive pathways and motivate biomarker-enriched trials.4 Conversely, null interactions would reassure readers that the associations are generalizable across common dialysis prescriptions. Either result would be informative for clinicians and investigators considering how broadly to apply sphingolipid-based risk assessment. We thank the authors for their contribution.