Comment on: Neutrophil-to-lymphocyte ratio as a biomarker for disease onset and mortality risk in systemic sclerosis: a real-world national cohort study

Dear Editor, We read with great interest the study by Keret et al. [1] published in Rheumatology, which investigated neutrophil-to-lymphocyte ratio (NLR) as a biomarker for disease onset and mortality risk in systemic sclerosis (SSc). This large national cohort provides valuable insights into NLR’s prognostic utility. However, we wish to raise methodological concerns regarding the variable selection and the influence of inflammatory biomarkers. First, regarding variable selection in multivariable analysis, the study used Cox regression adjusted for age, sex, interstitial lung disease (ILD) and pulmonary hypertension (PAH) to demonstrate NLR’s independent association with mortality. While these covariates are clinically relevant, the absence of univariate analysis as a screening step limits transparency in variable selection. Established guidelines recommend preliminary univariate analyses to identify significant variables for inclusion in multivariate models, reducing the risk of overfitting or omitted confounders [2, 3]. Without this, the model’s robustness may be questioned. Second, although the authors excluded patients with prednisone use to mitigate corticosteroid-induced confounding, other immunosuppressants (e.g. mycophenolate mofetil, rituximab) were not excluded from baseline NLR assessment. As the staff work in the clinical laboratory, medications can significantly alter neutrophil and lymphocyte counts. The study obtained NLR at diagnosis, but treatment data were recorded as ‘ever used’, leaving uncertainty about whether baseline measurements were influenced by ongoing therapies. This could introduce bias, as NLR values might reflect treatment effects rather than pure disease biology. For instance, the higher MMF use in patients with NLR ≥4 (19% vs 10%) suggests potential confounding.