Diagnosis of multiple sclerosis (MS) is facilitated by analyzing biochemical properties of cerebrospinal fluid (CSF). Oligoclonal bands (OCBs) and immunoglobulin G (IgG) index are well-established markers for evaluating patients suspected of having MS. Myelin basic protein (MBP) is also ordered frequently, but its usefulness remains questionable. OCB, IgG index, and MBP were measured in 16,690 consecutive CSF samples. Samples were divided into 2 groups based on MS status known (n = 71) or unknown (n = 16,118). Medical charts of the MS status known group were reviewed to determine their MS status. OCBs have a stronger association to IgG index results than does MBP. Importantly, MBP does not add a statistically significant increase in diagnostic sensitivity or specificity when used in combination with OCB and/or IgG index. The data indicate that MBP is an unnecessary and overused test.
Myelin basic protein
Sensitivity and specificity
Multiple sclerosis (MS) is a disease that results in the demyelination of brain and spinal cord white matter.1 The exact etiology of MS is unknown, but a body of research supports the idea that MS is an immunologically mediated disease. The diagnosis of MS requires the objective demonstration of disseminated central nervous system (CNS) lesions in time and space.2–4 Formal guidelines, known as the McDonald criteria, have been developed for the diagnosis of MS.3,4 The McDonald criteria integrate clinical and magnetic resonance imaging (MRI) data. Biochemical analysis of cerebrospinal fluid (CSF) is no longer part of these criteria,5 but it is a component of the differential diagnosis.6 Paraclinical features are diagnostically nonspecific but can help to rule out alternate causes for the clinical features of MS,3,4 thereby enhancing the specificity of the McDonald criteria, but not the sensitivity.
Biochemical analysis of CSF provides evidence of the inflammatory response in the CNS. Two well-established CSF markers are used for the diagnosis of MS: oligoclonal bands (OCB) and immunoglobulin G (IgG) index. Both OCB and IgG index measure the intrathecal production of IgG and provide a sensitive but nonspecific indicator of CNS inflammation. OCBs are immunoglobulins generated by plasma cells specific to the CNS compartment.7 OCB appear early in the onset of MS and remain detectable throughout the course of disease, even during and after treatment. Qualitative detection of OCB with isoelectric focusing (IEF) electrophoresis has the highest diagnostic sensitivity (>95%) for MS.8 The IgG index is a calculated ratio of the CSF IgG and CSF albumin concentrations relative to their concentrations in serum.7 The IgG index is an indicator of the IgG concentration in the CSF relative to the IgG concentration in the serum. An elevated IgG index indicates increased intrathecal synthesis of IgG. IgG index was previously shown to have a sensitivity of approximately 60% and a specificity of approximately 90% for MS detection.9,10 Both IgG and OCB can be used to support a diagnosis of MS as outlined in the original McDonald criteria3,4; however, both tests are nonspecific and are mainly used to rule out CNS inflammation, including MS.
Myelin basic protein (MBP), an important component of the myelin sheath, was proposed as a marker for MS diagnosis more than 30 years ago.11,12 MBP accounts for about one third of total CNS myelin protein, and CNS MBP concentrations rise in response to neuronal damage. Thus, similar to OCB and IgG index, MBP provides a nonspecific marker of CNS inflammation. Despite the fact that its use was suggested decades ago, very few studies have evaluated the diagnostic accuracy of CSF MBP as a marker for MS. The studies that have evaluated MBP have shown considerable variation in specificity (70%–90%),13–19 leading to considerable uncertainty regarding its use in the diagnosis of MS. In general, further studies are needed to obtain better estimates of the diagnostic accuracy of MBP.7 More specifically, it is uncertain whether MBP provides additional information that is unavailable from the well-characterized OCB and IgG index tests. Thus, diagnostic performance of MBP and its incremental value for an MS diagnosis is uncertain.
As a large national reference laboratory, we process high volumes of tests for CNS inflammation. Although MBP is not included in the McDonald criteria and is not as well characterized as OCB or IgG index, we have observed that MBP is a frequently ordered test. In fact, we find that MBP is ordered independently, yet almost as frequently as OCB. This is surprising because there is no evidence to support the usefulness of MBP, nor is it recommended or even mentioned, in the guidelines for MS diagnosis.3,4 Thus, there is concern that MBP is a highly ordered but noncontributory test. The objective of this study is to ascertain how often MBP is ordered and its perceived usefulness, ultimately evaluating if MBP is an unnecessary and overused test. To that end, we present data showing the incremental diagnostic yield of MBP in patients in whom OCB and IgG index have been measured.
Materials and Methods
This was a retrospective cross-sectional study. Cases were identified by performance of the index test (MBP) and 2 comparator tests (OCB and IgG index). All CSF specimens analyzed at ARUP Laboratories (Salt Lake City, UT) for OCB, IgG index, and MBP between February 1, 2009, and November 20, 2010, were included. Specimens without results for all 3 tests were excluded. All included specimens were stratified into 1 of 2 groups: MS status known and MS status unknown. The risk of adverse events related to these tests is low and well characterized. Flow diagrams for case assignment are presented in Figure 1 and Figure 2.
As a reference laboratory, we receive many samples for testing without clinical data. The MS status unknown group consisted of samples for which clinical data were not available. The MS status known group consisted of patients evaluated at University of Utah Health Care system for which clinical data were available through medical chart review. The medical charts of these patients were evaluated for the diagnosis of MS using clinical history, radiology reports, and discharge summaries. Patients were considered to have MS if the diagnosis was explicitly listed in their chart. The interval between CSF analysis and the reference test (clinical diagnosis) was generally 1 month or less. Patients in the MS status known group were excluded if their diagnosis was not clearly defined. The MS status unknown group was composed of samples that were sent from outside the University of Utah Health Care system. As such, no clinical outcome data were available. It is conceivable that many of these samples may well have come from patients with MS. The CSF test results of these samples were used to evaluate correlations between MBP and OCB and MBP and IgG index. Approval for this study was obtained from the University of Utah’s institutional review board.
OCB testing was performed using the HYDRAGEL 9 CSF isofocusing kit (Sebia, Norcross, GA) in accordance with the manufacturer’s instructions. Briefly, total IgG in the CSF and serum were adjusted to the same concentration and then electrophoresed in parallel using IEF electrophoresis on an agarose gel. OCB were revealed by the addition of enzyme-labeled anti-IgG antiserum. The presence of 2 or more IgG bands in the CSF, but not in the corresponding serum specimen, was considered positive.
The IgG index was calculated using the following formula:
Albumin and IgG in both serum and CSF specimens were quantified using the BN II Nephelometer (Siemens Healthcare Diagnostics, Deerfield, IL). IgG index values greater than 0.66 were considered positive (reference interval, 0.28–0.66). The coefficient of variation for each of the measurements used in the calculation is between 4.5% and 6.0%.
Flowchart for selection of multiple sclerosis (MS) status known population. IgG, immunoglobulin G; MBP, myelin basic protein; OCB, oligoclonal bands.
MBP concentrations were quantified using a commercially available MBP enzyme-linked immunosorbent assay (ELISA) (Beckman Coulter, Brea, CA). Specimens were incubated in a 96-well microtiter plate coated with mouse monoclonal antihuman MBP antibody. Following incubation and wash steps, a second mouse monoclonal antihuman MBP antibody labeled with horseradish peroxidase was added. After washing, a tetramethylbenzadine substrate was added. Absorbance was measured at 450 nm with background correction at 600 nm and was directly proportional to the concentration of MBP. Each patient specimen was performed in duplicate. MBP values greater than 1.1 ng/mL were considered positive (reference interval, 0.0–1.1 ng/mL). The assay has a coefficient of variation of 13.8% and 9.7% at 0.9 and 2.3 ng/mL, respectively.
Chart review was used as the reference standard. Clinicians were not blinded to the results of the index test (MBP) or the comparators (IgG index, OCB) so it is presumed that the results of these tests were used to obtain a diagnosis. Diagnoses were generally made by neurologists. The experience level of the neurologists was unknown, as was the interrater reliability of diagnosis.
Two-by-two matrices were created to assess the strength of association between binary-coded variables (positive vs negative) using odds ratios (ORs). In some cases, it was impossible to calculate the OR (eg, because the number of false-positive results was zero), and the χ2 test was used as an alternate measure of association. Multivariate ORs were calculated using logistic regression. Diagnostic sensitivity and specificity and positive and negative predictive values were calculated from the 2 × 2 tables using standard formulas. Receiver operating characteristic (ROC) curves were determined for variables with numeric values (IgG index and MBP) and the areas under the ROC curve (AUROC) were compared using the χ2 statistic. Medians were compared using the Mann-Whitney rank-sum test. Uncertainty in accuracy statistics were calculated using 95% confidence intervals. All calculations were performed using Stata 11.0 software (Stata Corporation, College Station, TX).
Between February 1, 2009, and November 20, 2010, 16,690 CSF samples were submitted from across the United States to ARUP Laboratories for IgG index, OCB, and/or MBP measurements (Figure 1). Of these, 501 specimens were excluded from further analysis because they lacked results for IgG index (n = 278), OCB (n = 79), MBP (n = 72), OCB and IgG index (n = 67), OCB and MBP (n = 2), or all 3 tests (n = 3). The main reasons for missing test results were insufficient specimen quantity, a concentration of CSF IgG below the limit of detection (making it impossible to calculate the IgG index), and failure to submit a paired serum sample for OCB interpretation. The remaining 16,189 specimens were divided into the MS status known group (n = 71) or the MS status unknown group (n = 16,118). Eight patients in the MS status known group were excluded because the diagnosis could not be determined from chart review. The demographic characteristics and CSF test results of the MS status known and MS status unknown groups are shown in Table 1.
Correlations of CSF Markers in the MS Status Unknown Population
We analyzed both univariate and multivariate associations of IgG and MPB with OCB. In a univariate analysis, we found a statistically significant association (OR = 2.2, 95% confidence interval [CI] = 2.0–2.3, P < .001) between MBP and OCB Table 2. We also found a statistically significant association between IgG and OCB (OR = 39.0, 95% CI = 35.1–43.3, P < .001). A multivariate logistic model was also used to compare the relative strength of association of MBP and IgG index to OCB. The multivariate logistic model was:
Thus, the association between OCB and IgG index (OR = 37.1, 95% CI = 33.4–41.2) is much stronger than the association between OCB and MBP (OR = 1.5, 95% CI = 1.3–1.6). The ORs obtained for the association of MPB and IgG with OCB in the univariate analysis (Table 2) and the multivariate analysis (Equation 1) are similar.
The relationship between MBP and OCB and IgG index and OCB is shown in Figure 3. Although the binary-coded diagnostic categories (positive vs negative) show a statistically significant association, the distribution of MBP values shows considerable overlap between OCB-negative and OCB-positive cases. Figure 3 also shows that the IgG index is more likely to be positive in OCB-positive cases, but that the association is not perfect.
Correlations of CSF Markers in the MS Status Known Population
The overall pattern of univariate correlations in the MS status known patients Table 3 was similar to the pattern found in the MS status unknown patients. A univariate analysis found statistically significant associations between MBP and OCB (OR = 3.5, 95% CI = 1.1–10.5, P < .03) and between IgG index and OCB (OR = 9.8, 95% CI = 2.5–37.7, P < .001) (Table 3). A multivariate logistic model found the following relationship:
Thus, the association between IgG index and OCB is stronger (OR = 7.5, 95% CI = 1.8–31.0) than the association between MBP and OCB (OR = 2.0, 95% CI = 0.6–7.0) as was seen in the MS status unknown population. Once again, the ORs obtained in the univariate analysis (Table 3) are similar to the ORs obtained in the multivariate analysis (Equation 2). Further, in the multivariate analysis, the confidence interval for MBP includes an OR of 1.0 which indicates that there is no statistically significant association between OCB and MBP after adjusting for IgG index.
Diagnostic Performance of CSF Markers in the MS Status Known Population
The diagnostic performance of each of the CSF markers was determined using the MS status known population Table 4. The OR for MBP and IgG index were 23.3 (95% CI = 2.6–184.8) and 5.4 (95% CI = 1.4–20.9), respectively. The OR for OCB could not be calculated because there were false-negative values of 0, and for that reason, the χ2 distribution was used as an alternate measure of association.
OCB showed the strongest association with MS (χ2 = 31.9, P < .001) followed by MBP (χ2 = 13.4, P < .001) and IgG index (χ2 = 6.6, P < .02).
The sensitivity of MBP, IgG index, and OCB was 92 (95% CI = 73–100), 50 (95% CI = 17–83), and 100% (95% CI = unable to calculate due to 0 value), respectively. The specificity of MBP, IgG index, and OCB was 67 (95% CI = 53–80), 84 (95% CI = 74–95), and 84% (95% CI = 74–95), respectively. The ROC curves for the CSF markers with quantitative measures (IgG index and MBP) were determined Figure 4 (Table 4). The AUROC for the IgG index and MBP were 0.78 (95% CI = 0.61–0.94) and 0.82 (95% CI = 0.72–0.92), respectively. The difference between the AUROCs for MBP and IgG index was not statistically significant (P = .63).
Evaluation of Decision Rules
We evaluated the diagnostic accuracy of every combination of OCB, IgG, and MBP Figure 5. The data show that no decision rule using MBP dominates any decision rule based only on OCB and/or IgG index. That is, there is no decision rule that provides a statistically significant increase in diagnostic sensitivity or specificity than could be achieved using rules based on combinations of OCB and IgG index.
MS is diagnosed based on clinical presentation with supporting objective evidence from neurologic testing. The first criterion of the McDonald diagnostic criteria is “no better explanation,” often requiring the use of paraclinical testing (MRI and blood or CSF testing) to help rule out other mimicking conditions, as well as to look for patterns characteristic of MS. Biochemical markers in CSF (OCB, IgG index, and MBP) are used to help distinguish patients with MS from those without MS.7,8 OCB are generally accepted to be the most useful of the CSF markers, and clinical data support the usefulness of OCB and IgG index.8,20 In contrast, there is very little evidence to support the usefulness of MBP. Most studies that have assessed the accuracy of these CSF markers for the diagnosis of MS were completed before IEF electrophoresis with immunodetection of IgG was commercially available to detect OCB and/or used a radioimmunoassay rather than ELISA to quantify MBP.12,19,21–24
Several concerns may complicate the use of MBP for the diagnosis of MS. First, MBP is a component of peripheral nerves and therefore it is not specific to the CNS. More importantly, the literature shows that MBP has multiple proteolytic cleavage products with differing immunoreactivities.25,26 It cannot be assumed that exactly the same fragments are present in all patients with MS. Quantification of MBP, as with all protein immunoassays, relies on the use of a standard curve generated from an intact MBP or peptide reference standards. Because the patient’s specimen consists of a complex mixture of MBP fragments, and this mixture is subject to interindividual variability, the measured concentrations will not necessarily correlate to the true amount of MBP. Subsequently, MBP data are difficult to interpret.
Receiver operator characteristic curves for immunoglobulin G (IgG) index and myelin basic protein (MBP) for diagnosis of multiple sclerosis (n = 63). The area under the curve for the IgG index was 0.78 (95% confidence interval [CI], 0.61–0.94) and for MBP was 0.82 (95% CI, 0.72–0.92).
Despite the aforementioned concerns about the difficulties of using MBP as a diagnostic marker, MBP accounts for about one third of total CNS myelin protein and has therefore been an attractive marker for demyelination.12 Early studies on the clinical usefulness of MS gave discrepant results14,15,19,27; however, later studies found a strong association between MBP and MS.13,16 The consistency seen in the later studies that used immunoassays to quantify MBP was most likely because of improvement in analytical methods. Early MBP research revealed that CSF MBP epitopes were variable, that the large positive charge of MBP could lead to nonspecific binding of the protein, and that the presence of MBP autoantibodies in CSF could interfere with immunometric assays.11,28–32 The recognition of these limitations led to the design of MBP immunoassays that minimize variability because of these biochemical properties.13,16 Further, the first assays developed to quantify MBP were radioimmunoassays, which, in the case of MBP, are known to be less sensitive and precise compared with the ELISA.14,16 Using such ELISAs, 2 recent studies showed that the accuracy of MBP is comparable to that of OCB and IgG index. One study reported the diagnostic sensitivity and specificity of CSF MBP to be 83.7% and 78.3%, respectively.13 A second study showed that the concentration of MBP increased by orders of magnitude during acute episodes of MS.15 Although these studies show that MBP is associated with MS, MBP remains poorly characterized compared with OCB and IgG index. Unfortunately, neither of these studies investigated the diagnostic performance of MBP relative to OCB and IgG index.
Comparison of diagnostic accuracy of marker combinations. Each policy is a decision rule based on the outcome of cerebrospinal fluid parameters. The policies are divided into 2 categories: those that do not include MBP (policies 1–4) and those that do (policies 5–13). The squares indicate the estimated sensitivity and specificity and the lines indicate the 95% confidence interval for the estimates. The dashed vertical line represents combined average of all policies. IgG, immunoglobulin G; MBP, myelin basic protein; and OCB, oligoclonal bands.
OCB was previously shown to be a highly sensitive marker of MS.9 IgG index results were reported to correlate well with OCB, but in general have a lower sensitivity for MS.8 Our results, based on a large cross-sectional sample (n = 16,189), indicate that both IgG index and MBP are significantly associated with OCB; however, the association of IgG index with OCB is much stronger than the association of MBP with OCB. This pattern was seen in both the MS status known and MS status unknown populations. Our data also show that there is considerable overlap in the distributions of MBP values in the OCB-positive and OCB-negative groups (Figure 3) and that MBP is weakly associated with the most sensitive MS marker, OCB. This is likely because MBP is relatively nonspecific and there is a much greater tendency to have raised MBP in conditions that affect the CNS other than MS.33
A comparison of the MS status known and unknown populations showed no significant differences between the percentage of OCB-positive and IgG index–positive patients. However, there were significant differences between the MS status known and unknown populations with respect to age, sex, and the proportion of MBP-positive patients (Table 1). The overall clinical significance of the differences between the 2 groups is unclear. However, it is clear that the patients in the MS status known group probably had more complicated conditions, because the median age of those patients was considerably higher (46 years) than the mean age of MS onset (30 years).34 In addition, despite a higher prevalence of MS in women than in men (2–3:1), our MS status known population showed a ratio of 1:1. This may be a function of more stringent ordering practices for MBP at the University of Utah compared with other facilities.
The data obtained by chart review allowed us to analyze the diagnostic performance of various combinations of markers for diagnosis of MS. In univariate analysis, OCB was the most sensitive and specific CSF marker for MS. MBP has comparable sensitivity and IgG index has comparable specificity (Table 4). Because these markers are most often used to rule out CNS inflammation, sensitivity is the most critical parameter for diagnostic performance. Sensitivity and specificity can be varied by using combinations of markers and decision rules. We compared the diagnostic performance of every possible decision rule using MBP with OCB and/or IgG index against every possible decision rule using OCB and/or IgG index. We found that no MBP-based rule had sensitivity or specificity that was statistically greater than any decision rule based on OCB/IgG index alone (Figure 5) Table 5 and Table 6. We also found the overall accuracy of MBP was not statistically different from the IgG index as indicated by ROC analysis (Figure 5). Although the data from the MS status known population suggests that MBP would be equivalent as a second test to complement IgG index, the sample size in the MS status known population is relatively small and the estimates are considerably uncertain. The study was not designed to assess noninferiority of MBP. Also, data from the MS status known population show that IgG index is highly associated with OCB (the gold standard) whereas MBP is only weakly associated with OCB. Considering the evidence from both the MS status known and unknown populations, our study suggests that MBP provides no incremental diagnostic information relative to OCB and IgG index.
The strengths of this study are the large sample size used for the correlation analysis and the use of the best available analytic techniques. The study has several limitations. First, we used an imperfect reference standard. Physicians often do not record the exact criteria used to arrive at a diagnosis so the chart diagnosis may not reflect the McDonald criteria. Also, the interrater reliability of the McDonald criteria is unknown. Interpretation of neurologic signs and MRI scans can be subjective, and therefore physicians may differ in their diagnoses. Thus, there is potential for classification bias. Second, clinicians were not blinded to index test results and so there is a potential for review bias (MBP, IgG) and incorporation bias (OCB). Third, the reference standard was inconclusive in 8 of 71 cases. We conducted a sensitivity analysis to determine the potential effect of the inconclusive cases and found that although the absolute values of sensitivity and specificity changed (as expected), the same pattern of results were obtained if the inconclusive cases were all counted as having MS or not having MS Table 7 and Table 8. Fourth, our sample size of diagnosed cases was small and, as a consequence, the confidence intervals associated with our accuracy statistics are quite wide. Finally, this is an observational study and the indication for testing was not known.
Future studies could help to elucidate the value of MBP in several areas. These include larger sample size (or a meta-analysis of past studies), better defined reference test criteria, more objective patient selection criteria that are more clearly tied to the condition of interest (eg, evaluation for MS), and longitudinal studies of markers rather than cross-sectional studies.
The use of MBP for the diagnosis of MS is not currently supported by a strong base of evidence. The test is not widely offered and, to our knowledge, only 2 of the 4 largest national reference laboratories in the United States offer quantitative MBP testing. However, despite this lack of evidence, we find that it is a frequently ordered laboratory test. Further, the test volume for OCB, which includes the IgG index, is roughly equivalent to the volume of MBP, suggesting that the 2 are almost always ordered in tandem. Although there is a need for further study of the diagnostic performance of MBP, we question whether this test should be routinely ordered for clinical investigation of MS. As a reference laboratory, we are generally unaware of the clinical indication for the test and we do not know how clinicians are using MBP results. Regardless, our data suggest that MBP is a test with little clinical usefulness.
We thank Dave Davis for his computational assistance and Chia Ni Lin, PhD, for her helpful translations.
Cerebrospinal fluid oligoclonal bands in the diagnosis of multiple sclerosis: isoelectric focusing with IgG immunoblotting compared with high-resolution agarose gel electrophoresis and cerebrospinal fluid IgG index. Am J Clin Pathol. 2003;120:672–675.
. Oligoclonal immunoglobulins and immunoreactive myelin basic protein in the cerebrospinal fluid of patients with multiple sclerosis and other neurological diseases. Acta Neurol (Napoli). 1985;7:311–314.
. A study of immunoglobulin G in the cerebrospinal fluid of 1007 patients with suspected neurological disease using isoelectric focusing and the Log IgG-Index: a comparison and diagnostic applications. Brain. 1990;113(Pt 5):1269–1289.
. Diagnostic value of paraclinical tests in multiple sclerosis: relative sensitivities and specificities for reclassification according to the Poser committee criteria. J Neurol Neurosurg Psychiatry. 1995;59:152–159.
A prospective study on the predictive value of CSF oligoclonal bands and MRI in acute isolated neurological syndromes for subsequent progression to multiple sclerosis. J Neurol Neurosurg Psychiatry. 1996;60:572–575.
Dina N.Greene, Robert L.Schmidt, Andrew R.Wilson, Mark S.Freedman, David G.GrenacheAm J Clin Pathol(2012)138 (2):
262-272DOI: http://dx.doi.org/10.1309/AJCPCYCH96QYPHJMFirst published online: 1 August 2012 (11 pages)