# Oxidative balance score predicts chronic kidney disease risk in overweight adults: a NHANES-based machine learning study

## TL;DR

A new score based on diet and lifestyle factors can predict chronic kidney disease risk in overweight adults, with machine learning models showing strong performance.

## Contribution

The oxidative balance score (OBS) is shown to predict CKD risk in overweight adults using machine learning and NHANES data.

## Key findings

- Higher oxidative balance score was inversely associated with CKD risk in overweight adults.
- Machine learning model GLMBoost achieved an AUC of 0.833 in predicting CKD.
- Age, LDL-C, and SBP were key predictors, with OBS components like physical activity and magnesium also contributing.

## Abstract

Oxidative stress plays a pivotal role in the pathogenesis of chronic kidney disease (CKD), particularly in overweight and obese populations where adipose tissue dysfunction exacerbates systemic inflammation and metabolic derangements. The oxidative balance score (OBS) is a composite index that integrates dietary antioxidants and pro-oxidant exposures, offering a quantifiable surrogate of oxidative burden. However, its utility in CKD prediction among overweight adults remains unclear.

We analyzed data from 28,377 overweight or obese participants in ten NHANES cycles (1999–2018). OBS was calculated based on 16 dietary components and 4 lifestyle factors. CKD was defined using KDIGO guidelines. Survey-weighted logistic regression models were used to assess the association between OBS and CKD, with multivariable adjustment. Restricted cubic spline regression examined dose–response patterns, and subgroup analyses evaluated effect modifiers. Additionally, 14 machine learning algorithms were trained and validated using SMOTE-balanced data and five-fold cross-validation. Model interpretability was enhanced through SHapley Additive exPlanations (SHAP) analysis.

A higher OBS was inversely associated with CKD risk (fully adjusted OR per unit increase, 0.975; 95% CI, 0.969–0.981; p < 0.0001), with a significant linear dose–response relationship. This protective association was attenuated in morbid obesity (BMI ≥ 40 kg/m2; Pinteraction < 0.001), a finding driven by the abrogation of the dietary score’s effect, while the lifestyle score remained protective in this subgroup. Among 14 machine learning models, GLMBoost was the top performer, achieving an Area Under the Curve (AUC) of 0.833 on the independent test set. SHAP analysis identified age, LDL-C, and SBP as primary predictors, but also revealed the significant protective contributions of OBS components—most notably physical activity and magnesium—and showed that age critically modifies the effects of both clinical and lifestyle factors.

Higher OBS was associated with lower CKD risk in overweight and obese adults. This may support the role of oxidative balance in kidney health and its potential for early prevention strategies.