Association between hemoglobin, albumin, lymphocyte, and platelet (HALP) score and rheumatoid arthritis: A cross-sectional study from NHANES 1999–2018

The hemoglobin, albumin, lymphocyte, and platelet (HALP) score is a potential composite biomarker that integrates nutritional and immune-inflammatory profiles. However, its association with rheumatoid arthritis (RA) prevalence remains undefined. In this large-scale cross-sectional study, we investigated the relationship between HALP and RA, explored the cross-sectional association between HALP score and RA prevalence, and evaluated HALP as a concurrent composite biomarker reflecting immune-inflammatory and nutritional status in RA. Our analysis included 34,079 participants from the National Health and Nutrition Examination Survey 1999–2018 cycles. We calculated the HALP score using the following formula: Hemoglobin (g/L) × albumin (g/L) × lymphocyte count (×10 ⁹ /L)/platelet count (×10 ⁹ /L). Weighted multivariable logistic regression models, adjusted for demographic and clinical covariates, were used to assess the HALP-RA association, with restricted cubic splines evaluating potential nonlinearity. We performed comprehensive sensitivity analyses and subgroup examinations to test result robustness. For cross-sectional discriminative modeling, we used a 3-stage variable selection process: Univariate screening, least absolute shrinkage and selection operator regression with 10-fold cross-validation for dimension reduction, and multivariable logistic regression. Model performance was assessed to evaluate how well HALP and clinical variables distinguish participants with prevalent RA from those without. In this large national study, the HALP score was inversely and nonlinearly associated with RA, consistent across subgroups. HALP score is inversely and nonlinearly associated with prevalent RA in a large US population. Given the cross-sectional design, HALP likely represents a consequence of established RA rather than a predictive risk factor. These findings are hypothesis-generating and require validation in clinically confirmed RA cohorts. Current evidence does not support using HALP for RA case identification or prediction. In this cross-sectional study of 34,079 participants, including 2166 unweighted RA cases (unweighted percentage 6.36%, weighted prevalence 4.85%), higher ln(HALP) levels were inversely associated with RA (adjusted odds ratio, 0.74; 95% confidence interval, 0.64–0.85; P  < .001), with a nonlinear inverse association ( P for nonlinearity < .001). Subgroup analyses showed no significant effect modification (all P for interaction > .05). A HALP-integrated cross-sectional discriminative model yielded an area under the curve of 0.747, which was mainly driven by conventional clinical variables. The incremental value of HALP was not evaluated.