Beyond fixed serum vitamin B12 thresholds: A multivariable risk approach to reflex methylmalonic acid testing (NHANES 2011-2014)
Related Experiment Videos
1Department of Clinical Laboratory, The First People's Hospital of Kunming, Kunming, Yunnan, China.
Background:
Reflex methylmalonic acid (MMA) testing is usually guided by fixed B12 thresholds, which may fail when kidney function is impaired.
Methods:
We analysed 10,174 NHANES 2011-2014 adults with complex-survey weighting. Four prespecified B12 cutoffs were compared with multivariable logistic models for elevated MMA (≥271 nmol/L) using decision curve analysis.
Results:
Weighted prevalence of elevated MMA was 8.9%. The 221 pmol/L cutoff had sensitivity 33.4% (9.4% at eGFR <45). Discrimination rose across a comparator ladder: dichotomised B12 0.623, continuous B12 0.690, B12 + eGFR 0.805, full model 0.819. ΔAUC over B12 alone was 0.128 (JKn 95% CI 0.100-0.157) overall and 0.052 (JKn 95% CI 0.032-0.071) with eGFR ≥60; post hoc temporal validation preserved discrimination (validation AUC 0.817). Grey-zone-refitted models retained discrimination (optimism-corrected AUC 0.71 for B12 + eGFR); frozen full-sample models were modest (AUC 0.651 two-variable, 0.669 full) and require local recalibration. At a 5% risk threshold the full model detected 83.2% of cases versus 33.4% for the cutoff-7.6 additional tests per additional case. Continuous-scale kidney-function interaction was sensitive to the test denominator (p = 0.028-0.096).
Conclusions:
Single B12 thresholds had limited sensitivity and no kidney-stratified threshold system was supported or validated. A multivariable model-using only B12 and eGFR-showed better discrimination and apparent net benefit than fixed cutoffs. Post-hoc grey-zone refits showed gains, but the frozen full-sample model showed poor grey-zone calibration; external validation is required before clinical use.
