Cui · Frontiers in nutrition 2022 · repeated cross-sectional study · n=71,685

Prevalence, trend, and predictor analyses of vitamin D deficiency in the US population, 2001-2018.

Cited 116 times in the scientific literature.

Level 4 - case-series / case-control

Repeated cross-sectional survey analysis (NHANES 2001-2018)

PubMed 36263304 · doi:10.3389/fnut.2022.965376 · record verified 2026-08-30

What was done

The authors analyzed NHANES cross-sectional data from 2001 to 2018 to evaluate prevalence, secular trends, and predictors of vitamin D deficiency (VDD) in the US population. Serum 25(OH)D concentrations were classified as severe VDD (<25 nmol/L), moderate VDD (25-50 nmol/L), insufficiency (50-75 nmol/L), and sufficiency (>75 nmol/L). Weighted multivariate linear regression models were used to identify predictors, adjusting for the complex survey design using Mobile Examination Center (MEC) weights.

What was found

Among 71,685 participants, weighted prevalence was 2.6% for severe VDD, 22.0% for moderate VDD, 40.9% for insufficiency, and 34.5% for sufficiency. Severe and moderate VDD were higher among women, non-Hispanic Black Americans, individuals aged 20-29 years, and during winter. From 2001 to 2018, slight linear decreases occurred in moderate VDD (coefficient = -0.847; P = 0.009) and insufficiency (coefficient = -0.810; P = 0.014), while sufficiency increased (coefficient = 1.693; P = 0.004). Severe VDD showed no trend change (coefficient = -0.037; P = 0.698). Age, sex, ethnicity, season, sun-protective behaviors, lower BMI, lower socioeconomic status, alcohol use, and lower milk intake predicted severe VDD.

Why it matters

This study provides updated national prevalence estimates through 2018, demonstrating slight population-wide improvements in sufficiency alongside persistent disparities and stagnant rates of severe deficiency among high-risk demographic groups.

Limits

The repeated cross-sectional design cannot establish causality or individual longitudinal trajectories. Serum 25(OH)D was measured at single timepoints, and the abstract does not provide exact effect sizes or confidence intervals for the identified predictors.

Cited by