Tracing a roadmap for vitamin B₁₂ testing using the health technology assessment approach

Laboratory workload

The current workload in our laboratory is ~5,000 B12 tests per year. Most requests are related to the evaluation of patients carrying myelogenous leukemia (22%) or admitted to Departments of Internal Medicine (20%) and Infectious Disease (13%). The number of requests coming from the Divisions of Neurology (8%), Nephrology and Dialysis (8%), and Gastroenterology (4%) is lower. Finally, ~9% of total requests regarded blood donors and 8% pregnant women under standard surveillance. By considering the reference interval of B12 assay in use by our laboratory (Roche Modular EVO system: 190–665 ng/L), we may report the percentage of test results for marker concentrations as follows: <190 ng/L (6.8%), 190–665 ng/L (74.7%), 666–2000 ng/L (17.4%) and >2000 ng/L (0.1%). As 92.2% of tested subjects exhibited physiologic or elevated B12 concentrations, it can be assumed that the test is mostly used to screen for vitamin deficiency or in subjects undergoing cobalamin supplementation.

Searching for recommendations

Bone marrow failure and demyelinating nervous system disease (causing brain atrophy and consequent dementia [20]) are the main overt clinical conditions associated with B12 deficiency [21]. Pernicious anemia (PA) is currently considered as the main responsible factor for B12 malabsorption and deficiency (20%–50% of B12 deficiency cases in adults) [22]. Referring to Northern Europe, where the disease has been well documented, the estimated prevalence of PA is ~2% in subjects ≥60 years and nearly 0.1% in the general population [23].

In the elderly, the frequency of B12 deficiency is relevant (>20% of total cases) [24], with a further significant increase in those patients with cardiovascular disease [25]. Recent epidemiological literature has reported that low B12 concentrations and/or mild increases of homocysteine (HCY) levels represent strong risk factors for neurodegenerative and cardiocerebrovascular diseases [21]. However, most meta-analyses of randomised controlled trials show that there is no evidence to support the use of B12 supplements for prevention of cardiocerebrovascular diseases [21, 26, 27]. Recent statistics have revealed that the primary etiology of B12 deficiency in the elderly is food-cobalamin malabsorption (syndrome of non-dissociation of B12 from carrier proteins associated to hypochlorhydria) [23]. In the overall population, the main causes of mild to severe malabsorption have been ascribed to various gastrointestinal diseases and to consumption/abuse of drugs, such as nitrous oxide, gastric acid inhibitors or metformin [21]. Furthermore, an increasing number of cases due to poor B12 intake have been detected in vegetarians, strict vegans, and newborns of vegan women [21]. Other physiologic (pregnancy) or pathologic (renal failure) conditions are cautionary screened and monitored overtime for cobalamin deficiency to be possibly treated as potentially life threatening or highly invalidating [21, 28]. B12 deficiency in pregnant women can affect the neurodevelopment in infants, whereas in renal failure this is reported to stifle treatment effect [29, 30]. As detailed below, in these patients a great caution should be taken in interpreting marker results, thereby the treatment is generally initiated independently of the results [21, 22, 30–32]. Due to potentially severe consequences of B12 deficiency and in agreement with results of large surveys, an increase of B12 and functional indicators [HCY and methylmalonic acid (MMA)] testing along with oral/intramuscular cyanocobalamin supplementation has been recorded in several countries [12].

Although these premises suggest the need for guidance on the interpretation of cobalamin status biomarkers, there is currently no agreement on recommendations for definition, diagnosis, treatment, and monitoring of B12 deficiency. For instance, no guidelines have been released by the American Society of Hematology, even though general practitioners often request B12 measurement in the evaluation and treatment of anemia, pancytopenia and myelopathy [21, 33].

A diagnostic algorithm for PA, including B12 determination as first-line test, was proposed in 1995 [34]. The evaluation of MMA, HCY, and anti-intrinsic factor antibodies (anti-IF) were considered mandatory in those patients with B12 concentrations both <200 ng/L decisional threshold and higher up to 300 ng/L. However, if clinical data are suggestive for PA, all the tests included in the algorithm had to be performed independently of B12 results [34]. More recently, different strategies have been suggested for PA diagnosis. Andres et al. [24, 35] recommended anti-IF evaluation (with a predicated 100% diagnostic specificity) only in individuals with serum B12 concentrations <150 pmol/L (203 ng/L) on two separate occasions or, alternatively, with serum B12 concentrations <150 pmol/L and HCY concentrations >13 µmol/L or MMA concentrations >0.4 µmol/L (in absence of renal failure and folate or vitamin B₆ deficiency). A more cautionary diagnostic approach with respect to cobalamin deficiency is suggested by the American Academy of Neurology (AAN), recommending the evaluation of serum B12, with MMA, and with or without HCY, for all patients suspected for polyneuropathy [36]. Although the poor evidence level of the recommendation, AAN supports the determination of all markers for screening purposes, accounting for a disease prevalence of ~2.4% in the overall population, rising to 8% in subjects >55 years [36]. The document warns on the wide discrepancies between B12 and metabolite results, implying that the additional determination of MMA and HCY can significantly increase the detection of cobalamin deficiency [36]. Similarly, the US Institute of Medicine Food and Nutrition Board (IM-FNB), remarking that a diagnosis of cobalamin deficiency may be only performed on a biochemical basis, suggested to measure MMA and HCY as the most sensitive markers [37], as B12 results may be within the reference interval in patients with reported PA or undergone previous gastrectomy [38]. In addition to these pragmatic indications, the IM-FNB document warned on the poor agreement between MMA and HCY results, by reporting limitations and interfering conditions [37]. Importantly, in monitoring cobalamin supplementation IM-FNB just recommends to evaluate the hematological response as index of therapeutic success, indicating erythrocyte/reticulocyte counts, blood hemoglobin concentrations, and hematocrit values as accurate measurements [37]. The National Health and Nutrition Examination Survey (NHANES) roundtable has recently recommended monitoring B12 status in the general population for the joint measurement of serum B12 [or holotranscobalamin (holoTC)] and of one functional measure of B12 inadequacy (preferably MMA) [12]. Finally, the American Dietetic Association and Dietitians of Canada, updating the nutritional guidelines for vegetarians, did not report any suggestion about diagnosis of cobalamin deficiency, but only recommendations for B12 intake [39].

We have previously stated that a deficit of B12 may result in severe illness among patients with renal failure and in newborns of pregnant women with the condition; however, within these clinical frameworks, guidelines only report an ancillary role for B12 testing. The Society of Obstetricians and Gynaecologists of Canada has released clear statements for B12 status evaluation and intake in pregnancy [40]. However, laboratory measurements were not required prior to initiating vitamin supplementation. In patients with chronic kidney disease, recommendations concern only the B12 supplementation, although deleterious effects of its deficiency are clearly reported. In particular, in patients with end-stage renal disease, B12 deficiency affects the maintenance of hemodialysis and increases erythropoietin stimulating agent (ESA) resistance. In fact, the effect of ESA treatment is significantly lowered by B12 deficiency and this problem is further amplified by the high costs of ESA [41]. Although MMA is reported to be highly influenced by renal failure, this biomarker with a 0.8 μmol/L threshold is currently included in the diagnostic criteria of ongoing studies. Saifan et al. have reported in this population a prevalence of B12 deficiency of 58%, according to MMA results [30]. It is quite difficult to establish how much this prevalence can be overestimated despite the adoption of an adjusted cut-off for MMA. It is noteworthy that these patients have a higher predisposition to B12 deficiency due to the constrained poor nutritional intake and to the fact that B12, as a middle size molecule, is easily cleared by modern high flux dialyzers [30, 41].

The functional B12 deficiency (defined as impairments in transport, processing and delivery of B12 to the physiological targets) should be shortly mentioned in this context, as the measurement of total serum B12 is not a reliable marker of this condition, whereas at present holoTC and MMA seem to provide a better index [42]. Overall, functional B12 deficiency is an increasing problem in the general population as its prevalence as a sub-clinical condition is higher than previously expected [42]. In infants, the functional deficiency, whether excluding a maternal undiagnosed disease, is rare and mainly due to inborn errors affecting cobalamin absorption (inherited intrinsic factor deficiency, Imerslund-Gräsbeck syndrome), transport (transcobalamin deficiency), cellular uptake, and intracellular metabolism (cblA-cblG) [43]. Twelve different inborn errors have been identified resulting in severe hyperhomocysteinemia and life-threatening MMA aciduria and metabolic ketoacidosis [43].

From systematic reviews to expert opinions

The Assessing Service and Technology Use to Enhance Health (ASTUTE Health) study group has recently published a SR and meta-analysis of the diagnostic performance of serum B12 testing [9]. Positive likelihood ratios of 2.7 and 3.3 (pooled 95% CI: 0.92 to 12.10) were reported when using either MMA or clinical criteria as diagnostic reference, respectively. Consequently, B12 concentrations below the decisional threshold were associated with a poor diagnostic value for vitamin deficiency. Additionally, negative likelihood ratios of 0.59 and 0.34 (pooled 95% CI: 0.13 to 0.89) were obtained with either diagnostic reference, also indicating the poor capability of B12 concentrations results above the diagnostic threshold to exclude cobalamin deficiency. As highlighted by the authors, the lack of a “gold standard” method for diagnosis of cobalamin deficiency prevented an accurate evaluation of the diagnostic accuracy of B12 testing. Most selected studies reported the measurement of MMA as the diagnostic reference method, associated or not with HCY, whereas the remaining studies adopted other criteria, such as hematological indices of deficiency, clinical evidence of symptom resolution, hematological response to therapy, or normalization of MMA/HCY concentrations. Under this condition, meta-analyzing pooled data is generally not recommended [14]. Furthermore, additional sources of heterogeneity across selected studies affected accuracy of pooled likelihood ratio estimates. For instance, B12 was assayed as serum cyanocobalamin or, alternatively, as holoTC by using a plethora of analytical methods (immunoassays, chromatographic and microbiological assays). Moreover, investigated cohorts included individuals under different conditions and/or from different clinical settings: healthy subjects, patients with clinical diagnosis or suspected for the vitamin deficiency. As a consequence, the prevalence of the disease (pre-test probability) was quite different among populations and pooling data may result in significantly biased estimations of the diagnostic performance. Finally, in addition to these methodological drawbacks, the strength of the computed evidence was further decreased by the poor quality of reporting in primary studies, not meeting Quality Assessment of Diagnostic Accuracy Studies (QUADAS) criteria [44].

Although results from this SR may aid in sharing unsolved issues and pitfalls existing around biomarkers of cobalamin deficiency, they should not be exploited to decide on discontinuing B12 testing in the clinical laboratory. We should also remind the experience of those laboratories replacing B12 measurement with HCY and MMA, as being attracted by the preliminary positive results on their higher sensitivities. Later findings on the limited diagnostic value of these metabolites suggested reinstating B12 testing. Currently, this is fostered by the potentially higher diagnostic accuracy of holoTC measurements, estimated by testing some populations at risk (e.g., vegetarians) [45, 46], in the face of the weak specificity of both HCY and MMA measurements [47, 48].

The ASTUTE group has shown that the conduction of a SR of primary studies is not feasible and useful in the framework of serum cobalamin tests, but, according to the HTA model, the identification of snowball references from papers identified by experts in the field may aid in building local guidelines. In particular, expert opinions gain a relevant role when discussing: (i) requirements for marker diagnostic accuracy in various specific settings; (ii) the optimal threshold value; and (iii) ethical issues related to the marker measurement (testing populations carrying life-threatening conditions) [18]. Experts currently speculate on the use of diagnostic algorithms including ≥2 biomarkers of B12 deficiency in combination with clinical data [35], despite that, increasing the number of markers has not resulted in any diagnostic improvement and, theoretically, contrasts with cost-effectiveness purposes. Some authors have recommended the co-detection of MMA and serum B12 (or holoTC) as an optimal diagnostic approach for cobalamin deficiency: the simultaneous detection of abnormal results for both markers (i.e., B12 decrease and MMA increase, according to selected thresholds) theoretically identifies those patients requiring cobalamin supplementation [22]. This approach maximizes the diagnostic specificity, and thus, the rule-in capability of the test. In addition, it meets the agreement of epidemiologists speculating on the higher relevance to identify an overt B12 deficiency than an early subclinical condition [subclinical cobalamin deficiency (SCCD)] [22]. SCCD represents the prevalent type of B12 deficiency reported in epidemiologic survey defined according to the isolated finding of abnormal concentrations of single markers not associated to clinical expression of the disease. Physicians are rather confused on the management of SCCD patients, generally monitored overtime by biomarkers and sometimes cautiously treated. Theoretically, SCCD, whether excluding possible interferences on biomarker determinations, appears associated with transient and minimally malabsorptive deficiency, generally progressing very slowly or not at all [22].

In the setting of primary care, other experts have suggested the interpretation of cobalamin test results in terms of “risk for deficiency” (low/medium/high susceptibility to cobalamin deficiency) [33]. Accordingly, in patients with hematological abnormalities or neurological symptoms, a B12 concentration <100 ng/L, is associated with a specificity of 90% [49], and may indicate a probable cobalamin deficiency. A possible or unlikely deficiency may be associated with marker results, respectively, below and above 300 ng/L. In patients with B12 concentrations >400 ng/L the vitamin deficiency may be excluded. Therefore, only patients with B12 <100 ng/L must be immediately supplemented, whereas patients with concentrations up to 300 ng/L are treated whether MMA and HCY results increase over the threshold [33].

More recently, experts have worked to release local guidelines and/or indications for checking B12 status [21, 28, 50]. However, controversial recommendations on test ordering have emerged; for some experts, B12 testing may apply to a broad range of clinical conditions (hematological and neurological diseases, pregnancy, glossitis, malabsorption, metformin therapy, and in dialysis patients) [28, 50], whereas for others only the detection of B12 metabolites should be considered [21]. Discordant approaches result also in monitoring B12 replacement. Stabler recommended to only perform the clinical evaluation of symptoms correction at 6 months, or blood counts at 2 months [21]. Others recommend a minimum re-testing interval of 6 months for B12 measurement [28]. The recently released Association for Clinical Biochemistry and Laboratory Medicine (ACB) recommendations on minimum re-testing intervals state that repeating the B12 measurement is useless [51].

Diagnostic performance and clinical requirements

A large body of literature has extensively discussed the methodological issues threatening the clinical performance of markers of cobalamin deficiency [12, 21, 22, 50, 51, 52]. MMA has been recognized as the most sensitive and specific marker, being considered as the reference for evaluating the diagnostic performance of different cobalamin assays. However, its diagnostic accuracy is strictly dependent on the clinical setting and in most cases MMA is reported to perform only marginally better than serum B12 [53]. In addition, impaired renal function markedly affects MMA (and HCY) concentrations, while influencing at lower extent serum B12 concentrations. This is not negligible when considering the relevance of testing patients receiving dialysis [54, 55]. Other interfering conditions are reported to spuriously increase B12 concentrations, as well as metabolites, thus lowering the diagnostic specificity. Individual genetic variations, various drugs used to treat gastrointestinal disease (e.g., proton pump inhibitors), liver disease and conditions implying variations in plasma cobalamin-binding proteins are well recognized causes of B12 elevation or decrease [21, 22]. Consequently, spuriously low cobalamin concentrations are detectable in pregnancy [32] as well as in hypothyroidism [56], whereas dramatic elevations may characterize chronic myelogenous leukemia and some tumors due to low cobalamin consumption [32]. Paradoxically, most confounders characterize those subjects (hemodialysis patients, pregnant women) at higher risk for B12 deficiency, for whom this condition may be most severe and its detection assumes an ethical significance.

Concerning clinical sensitivity, heterogeneous and contrasting data are available [22, 53]. Overall, Carmel’s report of diagnostic sensitivities >95% for B12 and MMA [22] seems too optimistic in face of the amount of data showing the low capability of tests to exclude the vitamin deficiency. With regard to serum B12, normal concentrations have been detected in patients with overt deficiency (PA or gastrointestinal resection) [38, 57, 58] and recognized not to exclude cobalamin-responsive hematological or neurological disorders [59]. In addition, B12 aberrant results have been pointed out in cases of PA, when the measurement was performed by immunoassays using IF as assay-binding protein [60]. However, the optimization of the initial step inactivating IF-blocking antibodies overcame the analytical interference. Data from NHANES have further reinforced the evidence on the low sensitivity of B12 testing: 3%–5% of patients with overt clinical deficiency had B12 concentrations within the reference interval and any attempt to shift the threshold level did not improve the diagnostic sensitivity [61]. There is also a “biological” explanation for the low sensitivity of B12 measurement in overt clinical conditions: the misleading evaluation of unchanged serum B12 concentrations can be due to cyanocobalamin liver storage, which can satisfy the daily metabolic requirement for a long time (up to 5 years), and also when absorption mechanisms are impaired [62, 63]. Actually, in absence of a specific marker of B12 absorption, the early deficiency appears to be undetectable [64].

The poor specificity of cobalamin metabolites (both MMA and HCY) exhibited in some critical settings (renal failure), together with the limited availability of MMA assay due to the cumbersome measurement procedure, have encouraged biochemists to invest on more reliable B12 laboratory evaluations [48]. The main challenge now is to show whether the replacement of total serum B12 (including cobalamin bound both to transcobalamin and haptocorrin) with holoTC (known as bioavailable fraction, representing 5%–20% of total vitamin) measurement may provide a “pure” marker of vitamin absorption and increase the diagnostic performance [65]. The possibility to directly measure the B12 fraction potentially entering the cells (holoTC) has reinforced the idea for an early detection of impaired mechanisms of intake and preliminary results on the highest sensitivity of holoTC have increased the interest on the detection of this form [64]. However, the lack of definitive data on holoTC as specific index of B12 absorption and on potential interfering conditions have refrained the theoretical attractiveness for this test [66, 67].

The relevance of claiming traceability of B12 results

The difficulty to evaluate the diagnostic accuracy of B12 testing according to standardized criteria increases the need to assure at least the analytical equivalence of marker results by different assays through the implementation of their traceability to internationally recognized reference materials (RMs) [12, 68]. Tracing back the calibration of commercial assays to suitable RMs enables to harmonize patient results and standardize the recommended decision limits [69].

In 1992, the RM from the National Institute for Biological Standards and Controls (NIBSC) coded 81/563 was first proposed [70] and adopted by the National Center for Environmental Health (NCEH), Centers for Disease Control and Prevention, in providing data by the Bio-Rad Quantaphase (QP) II radioimmunoassay, the assay used by NHANES up to 2006 [12]. When 81/563 was discontinued because of positivity to virological markers, NCEH started to use the new World Health Organization (WHO) RM coded 03/178, with a B12 value assigned by a consensus approach [70], replacing QP with the Roche E-170 electrochemiluminescent immunoassay. In a crossover study, NCEH re-analyzed NHANES samples previously measured by QP to obtain data on comparability, agreement and bias with Roche E-170 assay [12]. Although the correlation coefficient between differently calibrated assays resulted 0.98 over a range of B12 concentrations of 61–1490 ng/L, the Roche procedure was found to overestimate B12 concentrations when compared to QP (median, 593 vs. 528 ng/L), with a mean relative bias of 10.7% (95% limits of agreement: −5.6% to 27.7%) [12]. This could be expected as no continuity between the two RMs and their commutability was actually unproven. To solve the problem of trueness of B12 results, the National Institute of Standards and Technology (NIST) is developing a new RM (SRM 3951), consisting of fresh-frozen pooled human serum at three different concentrations, value-assigned with liquid chromatography coupled with tandem mass spectrometry, intended for direct calibration of commercial assays [71]. Considering its characteristics, the new material is expected to be commutable with native human sera. Given this resource available, manufacturers may finally establish traceable method calibrations and laboratories implement these standardized (“zero-biased”) assays in clinical setting. Nevertheless, the B12 standardization process remains complex for several reasons: (i) existence of multiple forms of B12 that may differently characterize serum samples, foods, RMs and proficiency testing (PT) materials [72]; (ii) B12 tight binding to proteins in the circulation; and (iii) low B12 levels present in serum (ng/L).

All the knowledge reported about the clinical validity of B12 testing and about the decision-making criteria used by physicians is based on assay-dependent marker results obtained by using non-harmonized methods. In this situation, through correlation studies appears to be mandatory to assess the degree of comparability, if any, among different B12 assays. Several authors have reported poor agreement and comparability of results among marketed automated immunoassays (IF-binding assays), the predominant methods employed in clinical laboratories [57, 58, 61, 73]. Data warned on the wide discrepancies in the classification of patients when using dichotomized results obtained by these immunoassays according to the selected cut-off. Potential reasons for this disagreement have been ascribed to: (i) sample pre-treatment steps; (ii) binding characteristics of the IF preparation used for ligand binding; and (iii) assay calibration using pharmaceutical vitamin preparations and not serum-based commutable materials [73]. Furthermore, the lack of sound evidence on comparability of B12 measurements across time unavoidably jeopardizes time-trend analyses, the major design component of NHANES [61]. Differences in B12 concentrations have been detected from one survey period to the next and even within a single survey period when measurement procedures have changed [61]. Therefore, it can be difficult to establish if differences in B12 concentration trends over time may be due to analytical artifacts or reflect true population changes in nutrient status, even if time-trend analyses were adjusted for differences in measurement procedures [12, 61, 74, 75].

Pursuing optimal diagnostic thresholds: a fallacious appeal?

Authoritative sources had stated that distinguishing subjects with normal and cobalamin-deficient conditions by a threshold level of serum B12 concentrations is unreliable and clinically misleading [76, 77]. In spite of this, a wide range of diagnostic thresholds for B12 have been reported and recommended in the literature. Most studies have adopted a 200 ng/L threshold, assuming a good comparability between assays [22]. Other authors have suggested shifting the cut-off from 200 to 350 ng/L in order to improve the diagnostic sensitivity. Statistical models have been proposed to derive B12 cut-off for deficiency based on the inverse and biphasic relationship between plasma concentrations of vitamin and functional indicators. Accordingly, Selhub et al. [78] estimated cut-offs of 203 ng/L and 406 ng/L when considering MMA and HCY concentrations, respectively. However, these estimates were shown to be deceptive because of unreliability of the statistical model [61].

The poor comparability of results from current B12 assays, shown by head-to-head comparisons and reinforced by PT surveys, has further complicated the attempts of harmonizing the diagnostic cut-off. Manufacturers generally provide reference intervals for healthy subjects, trusting the estimate of the decision cut-point to each laboratory. The selection of threshold level depends indeed on disease prevalence and may vary considerably among different clinical settings. More often, laboratories adopt the best threshold level estimated by maximizing the diagnostic sensitivity of the test, as clinicians generally tend to prioritize the rule-out capability of the test, by accepting to treat those subjects showing low B12 concentrations, but without overt vitamin-deficiency, as no adverse effects have been associated with an excess intake from supplementation.

Any effort to optimize the diagnostic thresholds for serum B12 might be impaired by the biological behavior of the marker, whether the within-person variation (CVw) results so low as to imply a low index of individuality (II), or the ratio between intra- and inter-individual (CVg) variation [79]. In this case, only monitoring serial individual changes is effective to classify patients, whereas a dichotomized transversal interpretation of a single result, according to a defined cut-off, can be misleading. In the framework of B12 biological variation, three studies have reported quite comparable data [80–82]. A cross sectional study from 1999 to 2002 NHANES estimated an average CVg of 43.6% and a CVw of 14.6%, resulting in an II of 0.34 [80]. Using the same design, the previous NHANES sub-study (1991–1994 survey) had reported a CVg of 41.6%, a CVw of 13.4%, and an II of 0.32 [81]. A third study designed to evaluate B12 seasonal variations on one single cohort estimated a CVg of 69% and a CVw of 15% (II=0.22) [82]. In agreement with the clinical evidence, the low II value (~0.30) resulted in a poor utility of reference intervals for B12 interpretation [79]. To reinforce this evidence, one can estimate the theoretical number of blood samples that should be collected to ensure that the mean marker result is within ±10% of the individual’s homeostatic set point, according to the statistics: 1.96² (analytical CV²+CVw²)/100 [83]. By using an average estimated CVw of 14.3% and a mean analytical CV of 2.9% (from our 2012 internal quality control), it can be estimated that each individual should undergo to approximately 8 marker measurements to achieve a sufficiently accurate estimate of his own B12 set point, which is clearly not feasible in clinical practice.