Urinary Albumin-to-Creatinine Ratio Across NHANES Cycles

Overview

Urinary albumin-to-creatinine ratio (UACR) is a key marker of chronic kidney disease (CKD) and cardiovascular risk. NHANES has measured urinary albumin (URXUMA) across all ten continuous cycles (1999–2018), making it well suited for pooled multi-cycle survival analyses.

This vignette illustrates two practical challenges that arise when assembling UACR from NHANES:

File disambiguation: urinary creatinine (URXUCR) appears in multiple laboratory files per cycle. In early cycles it was measured in both the albumin/creatinine file and several environmental-exposure files (phthalates, heavy metals). nhanes_variable_map() may find it in the wrong file; we show how to use keep_vars to pin it to the correct source.
Method comparability: NCHS changed the urinary albumin assay and analyser platform between cycles. Before pooling, analysts should consult the NHANES laboratory methods documents for each cycle and apply any published bridging equations.

Step 1 — Locate the variables

library(nhanesR)

# Confirm URXUMA is present across all ten cycles
nhanes_search_variables("albumin", component = "Laboratory")

# Variable map for urinary albumin: maps cleanly to ALB_CR_* files
map_uma <- nhanes_variable_map("URXUMA")
map_uma

# nhanes_variable_map("URXUCR") resolves to the wrong file in early cycles:
# URXUCR appears in phthalates / heavy-metals files (e.g. PHPYPA 1999-2000)
# as a dilution-adjustment variable, and the catalog finds those first.
nhanes_variable_map("URXUCR")

The early-cycle rows point to files such as PHPYPA (1999–2000) and SSNO3P_B (2001–2002). These are not the source we want.

The solution is simpler than it first appears: nhanes_download_analyte() downloads the complete file for each cycle, not just the named variable. The albumin/creatinine files (LAB16, L16_B, L16_C, ALB_CR_*) contain both URXUMA and URXUCR together, so downloading via URXUMA gives us both variables in a single pass.

Step 2 — Download and stack

# Restrict to cycles with public-use mortality linkage (1999-2018)
cycles <- nhanes_cycles()[nhanes_cycles()$has_lmf_public, "cycle"]

# nhanes_download_analyte resolves the changing file name automatically.
# Each returned data frame contains the full ALB_CR file — including
# URXUMA (urinary albumin) AND URXUCR (urinary creatinine).
alb_cr_list <- nhanes_download_analyte("URXUMA", cycles)
alb_cr      <- nhanes_stack(alb_cr_list)

Step 3 — Calculate UACR

NHANES reports:

Variable	Units
`URXUMA`	ug/mL (= mg/L)
`URXUCR`	mg/dL

The conversion to mg/g is:

$\text{UACR (mg/g)} = \frac{\text{URXUMA (mg/L)}}{\text{URXUCR (mg/dL)} \times 0.01 \text{ (g/dL per mg/dL)}} = \frac{100 \times \text{URXUMA}}{\text{URXUCR}}$

alb_cr$UACR <- with(alb_cr, 100 * URXUMA / URXUCR)

# Clinical thresholds (KDIGO 2012)
alb_cr$UACR_cat <- cut(
  alb_cr$UACR,
  breaks = c(0, 30, 300, Inf),
  labels = c("Normal-mildly increased (<30)",
             "Moderately increased (30-300)",
             "Severely increased (>300)"),
  right  = FALSE
)

Step 4 — Method comparability and bridging

Known method changes

NCHS changed the urinary albumin assay platform and reagents in [VERIFY: exact cycle transition]. Analysts pooling data across this transition should check the NHANES laboratory methods documentation for each cycle and apply published bridging equations before pooling if values are not directly comparable.

# The NHANES laboratory methods document for each cycle is linked from the
# online data browser.  For urinary albumin (ALB_CR files), look for entries
# under "Albumin, Urine" describing the analyser make/model and reagent kit.
# Key things to record for each cycle:
#   - Analyser (e.g. Roche Hitachi 917, Beckman UniCel DxC 800, ...)
#   - Reagent/method (immunoturbidimetric, immunonephelometric, ...)
#   - Calibrator traceability (JCTLM-listed reference material?)
#   - LOD (ug/mL) -- differs across cycles

Bridging equations for serum creatinine

As a reference example of how NCHS handles bridging, serum creatinine (LBXSCR) was transitioned from the Jaffe (alkaline picrate) method to the enzymatic/IDMS-traceable method around the 2007–2008 cycle. NCHS published regression-based bridging equations to convert Jaffe values to the enzymatic scale, enabling pooling across the transition. See the NHANES laboratory methods documentation and the NKDEP guidance for the specific equations.

Urinary creatinine (URXUCR) may require analogous treatment if method changes are identified in the laboratory methods documents.

Step 5 — Survival analysis with survey weights

Once UACR is assembled and any bridging applied, the workflow follows the standard nhanesR mortality linkage pipeline.

library(survival)
library(survey)

# Link mortality
alb_cr_mort <- nhanes_mortality_link(alb_cr)
alb_cr_surv <- nhanes_survival_prep(
  alb_cr_mort,
  origin     = "exam",
  time_unit  = "years",
  weight_var = "WTMEC2YR"
)

# Survey design
options(survey.lonely.psu = "adjust")
svy <- svydesign(
  ids     = ~SDMVPSU,
  strata  = ~SDMVSTRA,
  weights = ~survey_weight,
  nest    = TRUE,
  data    = alb_cr_surv
)

# Unadjusted rate by UACR category
svyby(~event, ~UACR_cat, svy, svymean, na.rm = TRUE)

Notes and caveats

The 2017–2020 pre-pandemic file (P_ALB_CR) should be used instead of the 2017–2018 file in pooled analyses to avoid double-counting participants. nhanes_variable_map() will warn about this overlap.
Spot urine samples (used here) are subject to substantial within-person variability. Single-void UACR is standard in epidemiological cohorts but regression dilution bias should be considered in exposure-response analyses.
UACR below the limit of detection (URXUMA < LOD) should be handled using standard methods (e.g. LOD/sqrt(2) substitution). The LOD varies by cycle and is documented in the NHANES laboratory methods pages.