Failure of Assay to Identify Low Cobalamin Concentrations

We wish to report a serious problem in the ADVIA:Centaur® chemiluminescence assay for cobalamin (Bayer Diagnostics; formerly the ACS:Centaur® assay, Chiron Diagnostics). The problem is urgent for two reasons: (a) our findings suggest that many cobalamin-deficient patients are being missed; and (b) the assay is used by increasing numbers of laboratories, which some time ago abandoned radioisotopic methods and were recently faced with the withdrawal of widely used assays by Abbott Laboratories. Several months after our hospital’s clinical laboratory introduced the Centaur assay, we noted that the number of subnormal results being reported had declined from ∼15–20 to 4–5 per month. When we reassayed two serum specimens that had given subnormal cobalamin results with the Abbott IMx® method earlier and had been stored at −20 °C, the results obtained with the Centaur method were within reference values. As a result of these observations, we selected 33 sera from our research collection of specimens that had been assayed with our radioisotope dilution assay (RIDA) and stored at −20 °C. The methodologic details of the RIDA, which uses pure intrinsic factor as the binding protein, have been published (1), with a reference interval of 140–750 pmol/L established in 332 subjects (2). Twenty-two of the 33 sera that we selected for testing had low cobalamin concentrations by RIDA and were from well-characterized patients with cobalamin deficiency. As established by clinical and metabolic evaluation by one of us (R.C.), pernicious anemia or ileal malabsorption (diagnosed by Schilling tests and, where appropriate, by serum anti-intrinsic factor antibody and serum gastrin testing) was responsible for the deficiency in 13 of the 22 cases; all 13 patients had documented megaloblastic anemia and/or neurologic symptoms. The other 9 cobalamin-deficient sera were from patients with mild preclinical deficiency [see Carmel (3) for background], which had been documented by metabolic tests of cobalamin status, such as the deoxyuridine suppression test and methylmalonic acid and homocysteine assays. The remaining 11 of the 33 sera had cobalamin concentrations within the reference interval by RIDA and came from patients with normal metabolic test results. All 33 specimens were assayed in a blinded fashion with the Centaur method in the clinical laboratory. Table 1 shows that the results were, in all but two cases, higher than the RIDA results; the median Centaur-to-RIDA ratio was 1.77 (range, 0.66–94.0). Much more importantly, the Centaur method failed to give low results in 16 of the 22 cobalamin-deficient sera (73% failure rate). We also showed that the poor performance of the Centaur method was not limited to our clinical laboratory. We gave 12 serum aliquots (8 from cobalamin-deficient patients) to a representative of Bayer Diagnostics for assay in a laboratory of his choosing. The results were very similar, and perhaps worse. Falsely normal values were obtained with the Centaur method in seven of the eight deficient sera (169–292 pmol/L, with one outlying value of >1500, vs the 68–124 pmol/L we had obtained by RIDA). Finally, we largely confirmed the previous RIDA results with another method and demonstrated that the poor performance of the Centaur method was not attributable to storage effects on the sera. We reassayed in a blinded fashion the 21 available samples (16 of which were from cobalamin-deficient patients) with the Quantaphase II® radioassay (Bio-Rad). Table 1 shows that the Bio-Rad assay failed to identify deficiency in only 2 of the 16 cobalamin-deficient sera (13%) compared with the 73% failure rate with the Centaur assay. For several reasons, the poor performance of the Centaur assay does not appear to be soluble simply by shifting the reference range. The degree of error varied greatly. Some discrepancies were very large, whereas others were relatively small. In addition, perusal of the College of American Pathologists ligand survey results suggests that the Centaur assay performs adequately in the normal reference range but consistently overestimates concentrations in the low range. For example, one survey from 1999 showed that the mean (± SD) result from responding laboratories for the low-cobalamin specimen (K-11) was 130.7 ± 17.2 ng/L with the Centaur method compared with 71.1 ± 7.9 ng/L for the Bio-Rad Quantaphase II method and 83.7 ± 15.1 ng/L with the IMx method. In contrast, the three methods agreed very closely in their results for two ligand specimens with cobalamin concentrations within the reference range (K-12 and K-13) in the same survey. This suggests that the inaccuracy of the Centaur assay may be confined to the low end of cobalamin values, which is exactly the critical area for clinicians. The misidentification of 73% of cobalamin-deficient sera as normal by the Centaur method is alarming, given the serious clinical consequences that can arise from failure to identify and treat cobalamin deficiency promptly. We do not yet know the reason for the Centaur assay malfunction. However, it is worth noting in general that published documentation and analysis of the new cobalamin assays, especially since the introduction of chemiluminescence assays, has been meager. Comparison of cobalamin results obtained with three assays. Results below the reference interval are shown in bold. Reference range, 140–750 pmol/L. Manufacturer’s reference range, 156–672 pmol/L. Cobalamin deficiency, <151 pmol/L (manufacturer’s data). Comparison of cobalamin results obtained with three assays. Results below the reference interval are shown in bold. Reference range, 140–750 pmol/L. Manufacturer’s reference range, 156–672 pmol/L. Cobalamin deficiency, <151 pmol/L (manufacturer’s data).