Mbs reviews vitamin b12 testing report february 2014 table of contents section Page

5.5 Quality of testing according to testing platform

5.5.1 Diagnostic performance of the serum vitamin B12 assay

The Ontario rapid review identified one systematic review and meta-analysis performed by Willis et al (2011)⁽⁸⁹⁾ on the diagnostic accuracy of the serum tests for assessing vitamin B12 (or cobalamin) across patient subgroups. They searched the literature from 1990 to November 2009 and identified 54 studies for inclusion. Of these, 12 were classified as case-control studies, while 42 were classified as cohort studies. The cohort studies were further subdivided into18 studies enrolling patients with suspected vitamin B12 deficiency (suspicion based on previous investigation, e.g. serum vitamin B12, MMA, homocysteine, clinical signs, etc.), a single study enrolling patients with suspected normal vitamin B12 levels (based on previous investigation, e.g. serum vitamin B12, MMA, homocysteine, clinical signs, etc.) and 23 studies enrolling patients with unknown vitamin B12 status. In this systematic review, the existing evidence base for serum vitamin B12 tests rated poorly against the criteria of a widely used and validated quality assessment tool (QUADAS)⁶. Studies that employed a rigorous reference standard method, such as that reported by Matchar et al. (1994)⁽⁹⁰⁾, were rated as positive against the QUADAS criterion relating to reference standards. The absence of a gold standard resulted in many studies being rated as unclear (rather than poor quality) on this criterion.
The included studies compared serum vitamin B12 to a reference standard (all reference standards were employed). They reported that there was no consistent reference standard used to measure the accuracy of the serum vitamin B12 test.⁷ They also reported a wide range of variability for sensitivity and specificity across the studies. The review found low levels of test sensitivity and, to a lesser extent, specificity across clinical indications and practice settings. For sensitivity, the range was 13% to 75%, and for specificity the range was 45% to 100%.
Test sensitivity was low when MMA and Hcy were employed as separate reference standards [MMA 0.52 (95% CI: 0.39, 0.65) and Hcy 0.40 (95% CI: 0.27, 0.54)]. Specificity estimates were considerably higher than those estimated for sensitivity [MMA 0.81 (95% CI: 0.70, 0.89) and Hcy 0.84 (95% CI: 0.73, 0.90)]. The authors of the systematic review attributed the wide ranges to the inconsistent use of a reference standard.⁽⁸⁹⁾
The systematic review by Willis et al. (2011)⁽⁸⁹⁾ included studies that assessed the diagnostic accuracy of serum B12 test using holoTC (four studies) as a reference standard⁽⁵¹⁾. HoloTC binds only a small portion of the total plasma vitamin B12 (20–30%) but is responsible for delivery of vitamin B12 to cells and is considered to be the functionally active fraction of vitamin B12⁽⁴⁴⁾ and has been advocated as having comparable diagnostic performance to serum vitamin B12 assays^{(46, 51, 52, 91)}. The meta-analysis performed by Willis et al. (2011) reported that serum B12 test sensitivity was higher when holoTC was used as a referent as compared to when MMA and Hcy were employed as referants (holoTC 0.70, 95% CI: 0.55, 0.82). Specificity was 0.79 (95% CI: 0.68, 0.86) when holoTC was used as the referent.⁽⁸⁹⁾
Assay methods for serum vitamin B12 analyses varied across studies; however, similar values for sensitivity and specificity were observed for both MMA and Hcy analyses. Radioassays were the most commonly reported methods for quantifying serum vitamin B12 (eight studies for MMA analysis, 11 studies for Hcy analysis), and demonstrated low levels of sensitivity in pooled analyses [MMA 0.57 (95% CI: 0.37, 0.74), Hcy 0.40 (95% CI: 0.24, 0.59)]. Under both reference standards, studies employing chemiluminescence (three studies for MMA analysis, one study for Hcy analysis), which is the most recent technology innovation for quantifying serum vitamin B12, demonstrated significantly lower sensitivity [MMA 0.36 (95% CI: 0.33, 0.38), and Hcy 0.13 (95% CI not reported)] compared to studies using radioimmunoassays (three studies for MMA analysis, two studies for Hcy analysis) [MMA 0.75 (95% CI: 0.66, 0.83), Hcy 0.70 (95% CI: 0.53, 0.83)]. This may indicate that the transition from older assay methods to newer technologies was not associated with improved diagnostic accuracy. However, these results need to be interpreted in light of the limitations of the existing evidence base, particularly the absence of an internationally accepted ‘gold standard’ for diagnosing conditions amenable to vitamin B12 supplementation, existing reference standard imperfections, and evolving meta-analytical methods for pooling data across diagnostic accuracy studies.⁽⁸⁹⁾
Results of the meta-analysis performed by Willis et al. (2011)⁽⁸⁹⁾ showed that positive likelihood ratios⁸ (PLR) close to one (suggesting limited value in differentiating between disease presence or absence) were observed for serum vitamin B12 across all subgroups. Among studies employing MMA as the referent, the PLR ranged from 1.23 (95% CI: 0.87, 1.70) in patients with previous vitamin B12 deficiency, to 3.70 (95% CI: 2.66, 5.13). Similar results were found in analyses of Hcy as the referent, with the PLR ranging from 0.90 (95% CI: 0.54, 1.27) to 3.68 (95% CI: 2.77, 4.89) in cohort studies enrolling patients with unknown vitamin B12 status. Negative likelihood ratios⁹ (NLRs) ranged from 0.95 (95% CI: 0.82, 1.09) in studies restricted to neuropsychiatric patients and Hcy as the reference standard, to 0.34 (95% CI: 0.22, 0.46) for studies using radioimmunoassay and using MMA as the reference standard, or those employing a clinical reference standard [0.34, (95% CI: 0.13, 0.89)]. The overall PLR and NLR were 2.72 (95% CI: 1.95, 3.81) and 0.59 (95% CI: 0.49, 0.72), respectively, in studies employing MMA as the referent. The overall PLR and NLR were 2.40 (95% CI: 1.78, 3.23) and 0.72 (95% CI: 0.62, 0.84), respectively in studies employing Hcy as the referent. In studies employing a clinical reference standard (e.g. response to therapy), PLR was 3.33 (95% CI: 0.92, 12.10) and NLR 0.34 (95% CI: 0.13, 0.89).
In summary, the 2011 review by Willis et al. was unable to demonstrate that diagnosis of conditions potentially amenable to vitamin B12 supplementation on the basis of a single serum vitamin B12 measurement is an accurate method for identifying vitamin B12 deficient patients. Across clinical indications and practice settings, the authors of the review found low levels of test sensitivity and, to a lesser extent, specificity. While age associated changes in vitamin B12 levels have been reported, the available data did not demonstrate improved sensitivity or specificity of vitamin B12 tests by age group. The transition from older assay methods to newer technologies (e.g. chemiluminescence) was not associated with improved diagnostic accuracy. The authors of the review stated that whilst these results may be suggestive of poor diagnostic accuracy for serum vitamin B12 assay, they need to be interpreted in light of the limitations of the existing evidence base, particularly the absence of an internationally accepted gold standard for diagnosing conditions amenable to vitamin B12 supplementation, existing reference standard imperfections, and evolving meta-analytical methods for pooling data across diagnostic accuracy studies.⁽⁹²⁾
The authors of the review acknowledged that not all variation in serum vitamin B12 test performance seen in the results is attributed to serum vitamin B12 assay deficiency. Therefore, it is wrong to conclude that serum vitamin B12 assays have no role in the diagnostic management of patients with diseases potentially amenable to vitamin B12 supplementation. Attempts have been made to outline diagnostic algorithms using a range of available tests including serum vitamin B12, Hcy and MMA. In combination with individualised patient management, such algorithms may be able to better tailor the use of serum vitamin B12 assays alongside clinical examination. Moreover, the role of serum vitamin B12 in medical decision making and the utility of results to influence therapeutic actions, requires greater clarification to more precisely define the place of this test in the management of patients with potential vitamin B12 deficiency.⁽⁹²⁾

5.5.2 Diagnostic performance of the holoTC assay

Numerous other studies (shown in Table 5.1) have compared holoTC’s performance with that of total vitamin B12 for identification of patients with vitamin B12 deficiency.^{(46, 51-53, 93-98)} In a longitudinal cohort study of 2,403 randomly selected older people, Clarke et al. (2007)⁽⁹⁷⁾ reported slightly superior diagnostic performance of holoTC compared to serum vitamin B12 measures (area under the curve (AUC) 0.85 versus 0.76, p<0.001). Similarly, in a multicentre study of 360 blood samples collected by five Dutch hospitals, Heil et al. (2012)⁽⁵³⁾ demonstrated a greater AUC for holoTC than for vitamin B12 in detecting vitamin B12 deficiency characterised by three predefined cut-off levels of MMA (serum MMA >0.32 µmol/L (i.e. 90th percentile); MMA >0.45 µmol/L (i.e. 97.5th percentile); and MMA >0.77 µmol/L (i.e. 99th percentile)). Applying a cut-off value of MMA >0.45 µmol/L resulted in an AUC of 0.70 (95% CI: 0.61, 0.79) for vitamin B12 and AUC of 0.78 (95% CI: 0.69, 0.87) for holoTC (p = 0.06). In addition, a cut-off value of 32 pmol/L of holoTC resulted in the highest sensitivity (83%) with acceptable specificity (60%) in detecting MMA concentrations above 0.45 µmol/L. However, the combination of vitamin B12 and holoTC in this study did not improve the diagnostic accuracy at this cut-off level. A third study by Herrmann and Obeid (2013)⁽⁹⁸⁾ also evaluated the diagnostic accuracy of holoTC assay compared to serum vitamin B12 on 1,359 serum samples. The authors of this study demonstrated that holoTC showed a higher AUC curve compared to serum vitamin B12 for detecting MMA levels > 300 nM. The AUC results of the other studies are summarised in Table 5.1.
Table 5.1: Comparison of holotranscobalamin (holoTC) and total serum vitamin B12 for diagnosis of vitamin B12 deficiency

Author	Total number of subjects (total number of subjects with vitamin B12 deficiency)	Age (years)	Limits for MMA (tHcy) (µmol/L)	AUC for holoTC	AUC for vitamin B12
Herrmann et al. (2003)^(46)	204 (68)	21-73	>0.27	0.88	0.84
Lloyd-Wright et al. (2003)^(96)	172 (36)	>18	>0.75	0.87	0.86
Hvas and Nexo (2005)^(52)	806 (24)	>18	>0.75	0.90	0.85
Miller et al. (2006)^(51)	1,789(116)	≥60	>0.35	0.83	0.82
Clarke et al. (2007)^(97)	2,403 (129)	>65	>0.75	0.85	0.76
Obeid and Herrmann (2007)^(94)	759 (174)	8-92	>0.30	0.71	0.60
Schrempf et al. (2011)^(93)	1,279 (71)	18-98	>0.40	0.66	0.72
Valente et al. (2011)^(95)	700 (not reported)	63-97	>0.36	0.90	0.80
Heil et al. (2012)^(53)	360†(47)	≥18	>0.45	0.78	0.70

† Blood samples
All but one of the studies in Table 5.1 showed that holoTC’s performance is better than that of vitamin B12, or comparable to serum vitamin B12^{(51, 96)}, independent of the cut-off MMA value they used to classify patients’ vitamin B12 status. The study by Schrempf et al. (2011)⁽⁹³⁾ was the only study that reported that the holoTC assay did not show superior diagnostic accuracy compared to total serum vitamin B12 assay for the detection of vitamin B12 deficiency in a cohort of subjects with neuropsychiatric conditions. The AUCs were not significantly different for holoTC compared to vitamin B12 in all subjects (AUC: 0.66 [95% CI: 0.51, 0.82]; p = 0.04 vs. 0.72 [95% CI: 0.65, 0.78], p < 0.0001) (Table 5.1).
However, the primary limitation of most of these studies is that comprehensive clinical diagnostic criteria were not used to definitively categorise individuals as vitamin B12 deficient or vitamin B12 adequate. The authors of these studies acknowledge that they had no access to haematologic or neurologic assessments. Consequently, they could only categorise persons as having likely vitamin B12 deficiency based on elevated MMA and homocysteine concentrations, while accounting for potential confounding by renal dysfunction (all the studies shown in Table 5.1 involved patients with normal kidney function).
Another limitation is that ‘falsely’ elevated holoTC has been noted in patients with common genetic polymorphisms, impaired renal function and liver disease.^{(99, 100)} Like other biological markers of vitamin B12 status (MMA and Hcy), the influence of factors other than vitamin B12 levels on holoTC status has considerable effects on its utility, not only as a diagnostic tool but as a useful reference standard for evaluating serum vitamin B12 performance.⁽⁹⁹⁾
In summary, there is evidence which indicates that holoTC has a comparable or better diagnostic accuracy to that of total serum vitamin B12. There is insufficient evidence to establish holoTC testing as an alternative to either total serum vitamin B12 or levels of MMA or homocysteine in the diagnosis of vitamin B12 deficiency.