The National Center for Health Statistics (NCHS) performed the National Health and Nutrition Examination Survey (NHANES), a population-based cross-sectional survey, to gather information about Americans’ health and nutritional status. It is conducted on a two-year cycle using complex multi-stage stratified probability sampling; hence, the study samples are representative.

All NHANES study protocols were approved by the NCHS Research Ethics Review Board, and each survey participant provided a signed informed permission. The public can access all comprehensive NHANES study designs and data at www.cdc.gov/nchs/nhanes/. The present investigation adhered to the cross-sectional research reporting criteria and guidelines of the Strengthening Norms for Reporting Observational Studies in Epidemiology (STROBE).

This study is based on 4 NHANES cycles from 2003 to 2010 because only those cycles contains complete PSA data.

In this analysis, participants with complete PSA and AGR data were included. The following exclusion criteria were applied to the subjects in this study: (1) age < 40 years, (2) lack of complete data on PSA and SII, and (3) factors affecting PSA (drug use such as 5-ARI, prostate hyperplasia, prostate infections and inflammation, undergoing prostate biopsy within 1 week, urological surgeries within 1 month and prostate cancer, etc.).

AGR is calculated using the formula: AGR = serum albumin/serum globulin. Serum specimens were processed, stored under proper refrigeration (2–8 °C), and shipped to Collaborative Laboratory Services for testing and analyses using the Beckman Synchron LX20 and Beckman UniCel DxC800® Synchron. In our analyses, the AGR served as the primary variable for investigation.

The total PSA was measured using the Hybritech PSA methods for recording serum total PSA concentrations (ng/mL) and the Beckman access, Department of Laboratory Medicine Immunology. Serum total PSA data was used as outcome variables in our analyses.

Covariates that may influence the relationship between AGR and PSA were also included in this study, including demographic characteristics: age, ethnicity, education level, body mass index and alcohol use; laboratory indices: blood urea nitrogen (mmol/L), cholesterol (mg/dL), glucose, serum (mg/dL), lactate dehydrogenase (U/L), total bilirubin (mg/dL), triglycerides (mmol/L), serum uric acid (mg/dL), serum creatinine (mg/dL), alachlor aminotransferase (U/L) and alanine aminotransferase (U/L); history of chronic diseases: hypertension (yes/no), diabetes mellitus (yes/no/critical), coronary artery disease (yes/no), angina pectoris (yes/no) and history of neoplastic diseases (yes/no);

This study additionally included covariates that might have an impact on the link between AGR and PSA. Percentages are used to indicate categorical data, while means and standard deviations are used to express continuous variables. Using student’s t-tests for continuous variables or chi-square tests for categorical data, the differences among the AGR (quartiles) groups were evaluated. Multiple regression models were used to explore the relationship between AGR and PSA based on 3 different models. For Model 1, no adjustment was made for covariates. Model 2 was adjusted for age and race. Model 3 was adjusted for age, race, education, body mass index, drinking status, urea nitrogen, glutamic oxaloacetic transaminase, glutamic alanine aminotransferase, cholesterol, lactate dehydrogenase, total bilirubin, triglycerides, serum uric acid, serum creatinine, hypertension, diabetes mellitus, coronary artery disease, angina pectoris and history of neoplasia. Building on Model 3, subgroup studies on the relationship between ARG and PSA were carried out, utilizing age, BMI, diabetes and hypertension as stratification factors. In addition, an interaction test was added to this study as a way to test the heterogeneity of the associations among subgroups. The nonlinear connection between the ARG and PSA level was evaluated using the generalized additive model (GAM) regression and smoothed curve fitting (penalized spline approach). Finally, based on Model 3, the nonlinear relationship between AGR and PSA was further validated using a two-stage linear regression model for threshold effect analyses. p < 0.05 was considered statistically significant. R version 4.3 and Empower software version 2.0 were used.