Background:
Vitamin B12 deficiency is associated with variations in red blood cell and white blood cell morphologies. These changes are also reflected in automated red cell measures. Are such changes of sufficient strength to influence clinical decision‐making about suspected deficiency states? Prior studies and case series disagree about the utility of hematologic indices in predicting B12 deficiency. This study sought a clinical prediction rule for low B12 levels using red ceil indices.
Methods:
A single institution's computerized laboratory records were searched from December 1, 2002, through November 30, 2007, for adults with a serum B12 measurement. Those who had not had a complete blood count performed simultaneously or within the previous 48 hours were excluded. For each case, the following values were noted: serum B12 level, hemoglobin value (HGB), mean corpuscular volume (MCV), and red cell distribution width (RDW). A low B12 level was defined as below 200 pg/mL Receiver operating characteristic (ROC) curves were constructed to assess sensitivity and specificity over the range of predictor values. The area under the ROC curve was used to quantify discrimination of the various predictors. An area under the curve of 0.5 suggests no ability to discriminate, whereas an area under the curve of 0.7 is considered fair, 0.8 good, and 0.9 excellent. HGB, MCV, and RDW were tested individually for their predictive ability, and logistic regression was used to combine the effect of HGB, MCV, and RDW to predict low B12 level.
Results:
Data were available for 50,405 patients, of whom 1352 (2.7%) had low B12 levels. The areas under the respective ROC curves for low B12 level versus HGB, low B12 level versus MCV, and low B12 level versus RDW were .61, .54, and .57, respectively. The logistic regression model to infer low B12 from a combination of all 3 indices was significant at P < 0.001. However, discrimination of the model was poor, with an area under the curve of only 0.62. The combined model had positive and negative likelihood ratios only as high as 1.6 and as low as 0.6 for 95% of the patient population.
Conclusions:
Red cell indices singly or in combination lack sufficient sensitivity and specificity to either rule in or rule out B12 deficiency. Although low B12 levels had statistically significant associations with changes in HGB level, MCV, and RDW, none of the relationships were strong enough to permit creation of a usable clinical prediction rule. Patients suspected of vitamin B12 deficiency should have a direct assay of B12 performed regardless of the values of hematologic indices.
Author Disclosure:
G. Buran MD, none.