Abstract
BACKGROUND: RBC folate (RBF) is an indicator of folate status and risk of neural-tube defects. It is calculated from whole blood folate (WBF), serum folate (SFOL), and hematocrit (Hct). SFOL and/or Hct are sometimes unavailable; hemoglobin (Hb) is generally available in surveys. OBJECTIVES: We assessed the ability of different RBF approximations to generate population data in women aged 12-49 y. METHODS: Using SFOL, RBF, Hct, Hb, and mean corpuscular Hb content (MCHC) from prefortification (1988-1994) and postfortification (1999-2006, 2007-2010) NHANES we applied 6 approaches: #1) assume SFOL = 0; #2) impute SFOL (population median); #3) impute Hct (population median); #4) estimate Hct (Hb/MCHC); #5) assume SFOL = 0 and estimate Hct; and #6) predict SFOL (from WBF) and estimate Hct. For each approach, we calculated the paired percentage difference to the "true" RBF and estimated various statistics. RESULTS: For 2007-2010 (unweighted data), the median relative difference from "true" RBF was lowest for approaches #2 (-0.74%), #4 (-0.96%), and #6 (-1.15%), intermediate for #3 (-3.36%), and highest for #5 (4.96%) and #1 (5.78%). The 95% agreement limits were smallest for approach #1 (2.33%, 13.0%) and largest for #3 (-20.8%, 11.3%). Approach #2 showed concentration-dependence (negative compared with positive differences at low compared with high RBF). Using weighted data, we found similar patterns across approaches for mean relative differences by demographic subgroup for all 3 time periods. CONCLUSIONS: We obtained the best agreement between estimated and "true" RBF when we predicted SFOL using a regression equation obtained from a subset of samples (approach #6). Alternatively, the consistent overestimation of RBF when assuming SFOL = 0 (∼6%) could be addressed by adjusting the data (approach #5). Similar observations for pre- and postfortification periods suggest applicability to low and high folate status situations, but should be confirmed elsewhere. To estimate RBF, at least WBF and Hb are needed.