Performance Evaluation of Serum Free Light Chain Analysis: Nephelometry vs Turbidimetry, Monoclonal vs Polyclonal Reagents

Abstract

Objectives:

Free light chain (FLC) measurement gained a lot of interest for diagnostic workup of monoclonal gammopathy.

Methods:

We evaluated the performance of turbidimetric polyclonal Freelite (The Binding Site, Birmingham, UK) assays on Cobas 6000 (Roche Diagnostics, Rotkreuz, Switzerland) and nephelometric monoclonal N Latex (Siemens Healthcare Diagnostics, Marburg, Germany) assays on BN ProSpec (Dade Behring, Deerfield, IL) vs established nephelometric Freelite assays on BN ProSpec.

Results:

Analytical performance was acceptable. Method comparison (n = 118) showed significant proportional FLC differences for N Latex assays. However, good correlation and clinical concordance were shown. Recovery study in the low concentration range demonstrated consistent over- and underrecovery for Freelite reagents, hampering future research on prognostic value of suppressed noninvolved FLC. Antigen excess detection was successful for κ FLC in three-fourths of cases with Freelite reagents and in all cases with N Latex reagents. However, the latter resulted in underestimated κ FLC concentrations.

Conclusions:

FLC analysis requires continuous awareness of analytical limitations. Monitoring of disease response requires FLC analysis on the same platform using the same reagents.

Materials and Methods

Monoclonal immunoglobulins or M-proteins are typically associated with clonal plasma cell disorders. Detection and quantification of an M-protein is considered the first serologic tumor marker for these disorders. M-proteins have a broad diversity of structure and concentration, making them challenging tumor markers for diagnosis and follow-up of monoclonal gammopathies. In addition, some clonal plasma cell disorders present without a clear M-spike on serum protein electrophoresis, the gold standard for M-protein detection. This subgroup includes light chain multiple myeloma (MM) and nonsecretory multiple myeloma, representing respectively 18% to 20% and 5% to 7% of all patients with MM.¹

Given the great diversity of clonal plasma cell disorders, multiple assays are necessary to effectively diagnose and monitor the disease. An initial diagnostic screening protocol comprises serum protein electrophoresis, serum immunofixation, and serum free light chain (FLC) determination. The latter has made urine protein electrophoresis redundant in the initial workup for monoclonal gammopathies.² Apart from its diagnostic value, baseline FLC values also have a prognostic value in newly diagnosed clonal plasma cell disorders.^3,4 Patients initially diagnosed with a monoclonal gammopathy of undetermined significance (MGUS) or smoldering MM were shown to have an 80% risk of progression to MM within 2 years if baseline FLC values result in an FLC ratio greater than 100.⁵ Therefore, the International Myeloma Working Group has updated its diagnostic criteria for MM to start prompt therapy in this high-risk asymptomatic population. Three validated biomarkers were added to conventional myeloma-defining CRAB (hypercalcemia, renal failure [ie, increased serum creatinine concentration], anemia, and lytic bone lesions) criteria, including more than 60% of plasma cells in bone marrow, more than one focal lesion on magnetic resonance imaging, and the presence of an involved/noninvolved FLC ratio greater than 100.⁶ A third application of serum FLC determination is the assessment of type of response. To establish a stringent complete response in a patient who has already achieved a complete response, a normal FLC ratio is required.^7,8

Currently, most frequently used FLC assays are latex-enhanced nephelometric or turbidimetric immunoassays using either monoclonal (N Latex reagents; Siemens Healthcare Diagnostics, Marburg, Germany) or polyclonal reagents (Freelite reagents; The Binding Site, Birmingham, UK). Depending on reagent and instrument being used, one should be aware of related analytical limitations. Vercammen et al⁹ described the occurrence of sample dilution anomalies associated with Freelite assays on BN II (Siemens Healthcare Diagnostics) analyzers, especially for FLC κ determination. Systematically performing a laboratory-determined dilution protocol next to the manufacturer’s dilution protocol is essential to avoid dilution errors. Another important issue is failure of antigen excess detection resulting in spuriously low FLC results in patients with high serum FLC amounts.¹⁰ To tackle this problem, Siemens developed an assay for BN II and BN ProSpec (Dade Behring) analyzers using monoclonal captation antibodies (N Latex assay) with a built-in antigen excess detection system.¹¹ An automatic kinetic antigen excess detection system is also available on Cobas 6000 (Roche Diagnostics, Rotkreuz, Switzerland) platforms using polyclonal Freelite reagents.

We evaluated the analytical performance of turbidimetric Freelite FLC assays on the Cobas 6000 c501 platform and the nephelometric monoclonal N Latex FLC assays on BN ProSpec and compared them with established nephelometric polycolonal Freelite assays on BN ProSpec. Clinical performance was investigated using Cohen’s κ agreement coefficients.

Results

Clinical Samples

From February to August 2015, 118 serum samples were collected from 112 different patients who went to the AZ Delta hospital Roeselare for initial screening or follow-up of a monoclonal gammopathy. All samples were analyzed for serum FLC on three different platforms: BN ProSpec (Dade Behring) with Freelite reagents (The Binding Site), BN ProSpec with N Latex reagents (Siemens Healthcare Diagnostics) and Cobas 6000 c501 (Roche Diagnostics) with Freelite reagents. Upon arrival in the laboratory, samples were aliquoted and frozen at –80°C until FLC assays were performed. Clinical diagnoses matching submitted serum samples were retrieved from the patients’ medical file. An overview of different diagnoses and median FLC values with corresponding interquartile range per condition are shown in ❚Table 1❚. Our study was performed with full respect for individuals’ right to confidentiality and in accordance with procedures supervised by local authorities responsible for ethical research.

Quantification of FLC

For clinical practice, serum FLC concentrations were assessed on BN ProSpec with polyclonal Freelite reagents. Data obtained on this platform were compared with determination on BN ProSpec with monoclonal N Latex reagents and on Cobas 6000 with polyclonal Freelite reagents. Each assay was carried out according to the manufacturer’s instructions.

Using Freelite reagents on the BN ProSpec platform, an initial dilution of 1:100 for both κ and λ FLC was used. If applicable, samples were automatically further diluted (1:400, 1:2,000, 1:8,000 or 1:32,000) or increased in volume (1:5, 1:20) until a quantitative result was obtained. Given the lack of built-in check for antigen excess, The Binding Site advises to adjust the abovementioned standard dilution protocol if (1) the measured FLC concentration or calculated κ/λ ratio is outside the reference values, (2) the sample is from a patient who has previously demonstrated antigen excess, or (3) clinical or other laboratory findings do not correlate with the obtained FLC results. If one of these conditions occurs, independent measurements at both a 1:400 and 1:2,000 dilution are advised. In case of a fourfold or more difference between results, which is suggestive of antigen excess, it is recommended to report the final result.

Analytical Performance

Discussion

Method Comparison

Obtained performance characteristics were compared with different quality goals derived from biological variation as assessed by Hansen et al,¹² who calculated intraindividual biological variations (CV_i) of 4.3% and 7.0% and interindividual biological variations (CV_g) of 21.0% and 30.0%, for κ and λ FLC, respectively. Optimum, desirable, and minimum specifications for total imprecision (I) were calculated as follows: I_optimum = 0.25 × CV_i, I_desirable = 0.50 × CV_i, and I_minimum = 0.75 × CV_i. Specifications for bias (B) were calculated as follows: B_optimum = 0.125 × √(CV_i² + CV_g²), B_desirable = 0.250 × √(CV_i² + CV_g²), and B_minimum = 0.375 × √(CV_i² + CV_g²). Optimum, desirable, and minimum specifications for total allowable error (TEa) were calculated combining respective values for imprecision and bias through formula TEa = B + 1.65 × I.

For method comparison, Spearman’s rank correlation and Passing-Bablok regression analysis were performed. Clinical concordance between samples, analyzed on the different platforms, was evaluated using Cohen’s κ agreement coefficients for both κ and λ FLC and κ /λ FLC ratio. Samples were divided into below-, within-, and above-reference range measurements.

To evaluate the analytical performance in the low concentration range for κ and λ FLC, a 9-point serial dilution was measured on all three platforms. Several routine serum samples were collected to generate a pooled sample with κ and λ FLC values within the normal concentration range. Sample dilutions were prepared by mixing the pooled sample with physiological saline. Two replicates of each dilution were measured on the three different platforms. Measured values were plotted against expected values, and regression analysis was performed (CLSI EP6-A).

Antigen excess detection was evaluated through testing of samples (n = 4) with known antigen excess on all three reagent-platform combinations. The executing laboratory technician was blinded for presumed antigen excess. Samples were tested with standard dilutions and automatically suggested dilutions, as well as with forced dilutions, to obtain a consistent test result.

Statistical analysis was performed using Medcalc (software version 12.2.1, Mariakerke, Belgium). Statistical difference between methods was analyzed with the Wilcoxon paired sample t test or normal paired sample t test if data were normally distributed. P values less than .05 were considered statistically significant. Method comparison was done using Spearman’s rank correlation, Passing-Bablok regression analysis, and calculation of Cohen’s κ agreement coefficients.

Performance data are summarized in ❚Table 2❚. For the Freelite assay on the Cobas 6000, coefficients for total imprecision (CV_t) ranged from 2.5% to 4.1% for κ FLC and from 1.8% to 3.8% for λ FLC. Imprecision for the BN ProSpec N Latex assay was comparable, showing CV_t from 2.0% to 4.1% for κ FLC and 1.8% to 2.8% for FLC λ analysis. For λ FLC, a desirable criterion of CV_t less than 3.5% based on biological variation¹² was met for both assays. Desirable specification was more stringent (CV_t <2.2%) for κ FLC and not fulfilled for most analyte levels of both assays. Even minimum specification of CV_t less than 3.2% was not met for all κ FLC levels of both assays.

The assay on the Cobas 6000 has excellent accuracy. Biases for low and high QC levels of 0.9% and 1% for κ FLC and 0.4% and 1.4% for λ FLC were assessed. Optimum specifications based on biological variation¹² were met for all analyte levels. Obtained biases for BN ProSpec N Latex assays were higher, as the low QC level presented with a bias of 8.1% and 7.5% for κ FLC and λ FLC, respectively, thereby not meeting a desirable specification for bias. The high QC level showed a bias of 5.6% for κ FLC and 2.7% for λ FLC.

Recovery Experiment

These performance data resulted in a total error (TE) of 5.2% to 6.4% for the κ FLC and 4.4% to 4.7% for the λ FLC Freelite assay on the Cobas 6000. Desirable specifications for total allowable error (TEa) were met for both κ FLC and λ FLC, and for λ FLC, even the optimum criterion of a TEa less than 6.7% was fulfilled. For the N Latex assay on the BN ProSpec, a TE of 12.4% to 12.7% for κ FLC and 7.3% to 10.5% for λ FLC was calculated, thereby meeting the minimum specification (TEa <13.4%) for κ FLC and a desirable specification (TEa <13.5%) for λ FLC.

Antigen Excess Detection Evaluation

Method comparison results are shown in ❚Figure 1❚. Freelite assays on the Cobas 6000 showed excellent correlation with established Freelite assays on the BN ProSpec, with respective Spearman’s coefficients of rank correlation (r_s) (95% confidence interval [CI]) of 0.97 (0.96-0.98), 0.98 (0.97-0.98), and 0.96 (0.94-0.97) for κ FLC, λ FLC, and κ/λ FLC ratio. Passing-Bablok regression analysis indicated a significant positive proportional bias for both κ and λ FLC assays on the Cobas 6000 compared with the Freelite assays on the BN ProSpec, which, however, did not result in a significant bias for the κ/λ FLC ratio between both platforms.

Good correlation was demonstrated also for N Latex reagents with corresponding Freelite assays on the BN ProSpec, with r_s (95% CI) of 0.94 (0.91-0.96), 0.93 (0.90-0.95), and 0.94 (0.91-0.96) for κ FLC, λ FLC, and κ/λ FLC ratio, respectively. Passing-Bablok regression analysis showed a negative proportional bias (95% CI) of 26% (20.5%-31.6%) for κ FLC and a positive proportional bias of 34% (16.1%-63.8%) for λ FLC when comparing N Latex assays with Freelite assays on the same platform. This resulted in a pronounced negative proportional bias of 70% (66.2%-76.6%) for the κ/λ FLC ratio.

In addition, we observed that for the Cobas 6000–Freelite and the BN ProSpec–N Latex platforms, correlation with the BN ProSpec using Freelite reagents was weaker for low λ FLC concentrations (<10 mg/L).

Conclusion

Statistical Analysis

Perfect recovery was observed for the BN ProSpec using N Latex reagents both for FLC κ and λ. In contrast, using polyclonal Freelite reagents on the Cobas 6000 as well as the BN ProSpec, we observed an increasing overrecovery with decreasing FLC values for both κ and λ applying the initial standard dilution. When switching to increased volume measurements, an underrecovery was observed. This over- and underrecovery was most prominent on the Cobas 6000 platform for FLC λ ❚Figure 3❚. Over- and underrecovery on the BN ProSpec with Freelite reagents had no impact on clinical interpretation of FLC results; all expected and measured values were clinically concordant. However, on the Cobas 6000, a borderline decreased expected value of 2.95 mg/L (reference values: 3.3-19.4 mg/L) for FLC κ was measured as a normal value of 4.40 mg/L, and a decreased expected value of 5.28 mg/L (reference values: 8.3-27.0 mg/L) for FLC λ was measured as a borderline normal value of 8.40 mg/L. Still, clinical interpretation of the κ/λ ratio based on expected and measured FLC values was concordant for all dilutions on all three platforms tested.

Analytical Performance

Results are listed in ❚Table 3❚. All four samples had κ antigen excess. FLC κ results measured with the Freelite assay on the BN ProSpec were all falsely low as measured with standard dilution. No flagging “> test” was produced. Forced dilutions gave rise to final fourfold to 10-fold higher κ values.

For the Freelite assay on the Cobas 6000, three samples were flagged with antigen excess warning “> Kin” for κ and were automatically diluted. Further additional dilutions confirmed the antigen excess, and final fourfold to 10-fold higher results were produced. One sample was not flagged for antigen excess and on the contrary was flagged with “< test,” and an automated rerun with increased volume (1:1) generated a spuriously low result of less than 0.80 mg/L. Forced dilutions, however, generated the antigen excess flagging “> Kin” for the 1:50 dilution and a final result at 1:500 of 1,370 mg/L.

The N Latex FLC κ assay on the BN ProSpec detected all four samples as antigen excess and reported an increased result for FLC κ without manual intervention (Table 3). However, the final κ result was underestimated as forced dilutions gave rise to much higher κ values, in line with the other platforms (taking into account the supra mentioned negative proportional bias for κ compared with Freelite reagents).

Method Comparison

Use of serum FLC immunoassays for screening and monitoring of patients with plasma cell dyscrasias is well established in clinical practice.¹³ FLC assays have detection limits more than 50-fold lower than electrophoresis and 20-fold lower than immunofixation electrophoresis, adding a level of certainty to the current battery of laboratory tests.² The ability to detect abnormal amounts of FLCs and an abnormal FLC κ/λ ratio is dependent on the characteristics of the platform being used. In the present study, we evaluated the performance of three different reagent (monoclonal vs polyclonal) and platform (nephelometer vs turbidimeter) combinations for FLC analysis: the Freelite assay on the BN ProSpec, the N Latex assay on the BN ProSpec, and the Freelite assay on the Cobas 6000 c501. Serum samples (n = 118) of different patients (n = 112) with clinical suspicion of a monoclonal gammopathy were included for method comparison. Analytical performance was assessed according to the CLSI EP5-A2 protocol. Antigen excess detection was evaluated through analysis of four samples with known antigen excess.

Studies on biological variation of serum FLC have been performed.^12,14,15 In a recent study by Hansen et al,¹² the assessed within-subject biological variations (CV_i) were small, with a CV_i of 4.3% for FLC κ and 7.0% for FLC λ, resulting in stringent analytical performance goals. For FLC κ analysis, even minimum imprecision specifications were not always met for all analyte levels of both the Freelite and N Latex assays regarding the very low CV_i. On the other hand, Katzmann et al¹⁶ assessed biological variation of serum FLC in patients with a monoclonal gammopathy over a 5-year period and reported a CV_i of 28%. This high CV_i is an overestimation of the actual biological variation due to the contribution of an inherent between-reagent lot variation when using Freelite reagents. In the present study, we demonstrated acceptable total imprecision data for the FLC assay on the BN ProSpec using N Latex reagents and on the Cobas 6000 using Freelite reagents (Table 2). The bias we reported for the N Latex assay is considerably higher compared with the bias associated with the Freelite assay on the Cobas 6000. However, we have to take into account that every new reagent lot for the Freelite assay on the Cobas 6000 goes with a new lot QC material explaining the excellent accuracy. In contrast, a specific lot of N Latex QC material can be combined with multiple N Latex reagent lots. Furthermore, polyclonal Freelite reagents are described as having considerable lot-to-lot variation (8%-45% for FLC measurement and 17%-32% for the calculated κ/λ ratio), which may lead to misinterpretation when monitoring a patient for a long period.⁹ A change in reagent lot without a change in the patient’s disease status can be associated with a 30% difference in the calculated κ/λ ratio. Monoclonal reagents, like N Latex reagents, have the theoretical advantage of less variation between reagent lots, which has been confirmed by Pretorius et al.¹⁷ In view of the reported interlot CV for Freelite reagents, the analytical performance should be monitored over a more extended period to determine if monoclonal antibodies are indeed associated with superior analytical performance.⁹ When using an FLC assay in clinical routine practice, it is advisable to monitor lot-to-lot variation by means of a pooled serum-based control sample, which is measured at the same frequency as the commercially available QC material.

Polyclonal reagents have been associated in several studies with a broader recognition profile for serum FLC.^9,18-20 Monoclonal reagents have a more limited epitope specificity and therefore are considered not equally able to detect a broad range of FLC epitopes compared with a polyclonal reagent. In the current study, the N Latex reagent failed to detect a κ FLC clone in a serum sample of a patient diagnosed with IgM κ MM that was recognized by using Freelite reagents on both the BN ProSpec and Cobas 6000. This could be due to the inability to detect a specific epitope by the monoclonal N Latex reagent. However, full recognition of all FLC epitopes by the currently available immunoassays remains a concern.

References

Method comparison between the BN ProSpec and Cobas 6000 using Freelite reagents demonstrated very good correlation for the FLC κ, λ, and κ/λ ratio. The Cobas 6000 measured slightly higher for both κ and λ FLC, resulting in no significant difference for the calculated κ/λ ratio. Weakest correlation was observed at the low end of the FLC λ concentration range. Method comparison between the BN ProSpec using Freelite reagents and BN ProSpec using N Latex reagents also showed a good correlation. The N Latex assay measured lower for κ FLC and higher for λ FLC compared with the Freelite assay on the BN ProSpec, which resulted in a pronounced negative proportional bias of 70% for the FLC κ/λ ratio. The assessment of higher λ FLC concentrations with N Latex reagents was confirmed in different studies while data concerning κ FLC measurement are not unambiguous.^11,17,20,21 Weakest correlation was again observed in the low FLC λ concentration range. Considering the absence of a gold standard, it is not possible to denote the most accurate method. Based on the performed method comparison, we can conclude that results obtained on different platforms are not readily interchangeable, and long-term follow-up of patients should be performed on the same platform using the same reagents.

The performance of the different platforms in the low-end concentration range was highlighted by means of a recovery experiment. The BN ProSpec using N Latex reagents showed a superior recovery compared with both assays using Freelite reagents. Still, all three platforms were capable of measuring low FLC values with sufficient accuracy. All platforms have built-in detection protocols automatically switching to increased volume measurements, leading to acceptable recovery in the low range regarding clinical use for normal to high FLC concentrations. Moreover, quantitative differences on the reported values had no implications on the clinical interpretation of the calculated κ/λ ratio. In addition, until now, no clinical significance has been formulated on the extent of FLC suppression. Current guidelines for measurable response criteria for monoclonal gammopathies state that the concentration of the involved FLC has to be more than 100 mg/L in combination with a κ/λ ratio outside the reference values.^6,22 Therefore, again, we can conclude that all platforms are performing sufficiently well in the low concentration ranges. On the other hand, preliminary work of Katzmann et al²³ indicated that an uninvolved heavy and light chain pair suppression (eg, suppression of immunoglobulin G [IgG] λ below the lower level of normal in a patient with a monoclonal IgG κ gammopathy) is an independent risk factor for progression to MM in patients with (MGUS). Our findings with consistent over- and underrecovery for polyclonal FLC reagents in the low range hamper future research considering the prognostic value of suppression of the noninvolved FLC. Research with N Latex reagents could in part resolve this problem, but all current guidelines for monitoring patients with monoclonal gammopathies are based on research performed with polyclonal Freelite reagents.

Author notes

The performance of the different platforms in the low-end concentration range was highlighted by means of a recovery experiment. The BN ProSpec using N Latex reagents showed a superior recovery compared with both assays using Freelite reagents. Still, all three platforms were capable of measuring low FLC values with sufficient accuracy. All platforms have built-in detection protocols automatically switching to increased volume measurements, leading to acceptable recovery in the low range regarding clinical use for normal to high FLC concentrations. Moreover, quantitative differences on the reported values had no implications on the clinical interpretation of the calculated κ/λ ratio. In addition, until now, no clinical significance has been formulated on the extent of FLC suppression. Current guidelines for measurable response criteria for monoclonal gammopathies state that the concentration of the involved FLC has to be more than 100 mg/L in combination with a κ/λ ratio outside the reference values.^6,22 Therefore, again, we can conclude that all platforms are performing sufficiently well in the low concentration ranges. On the other hand, preliminary work of Katzmann et al²³ indicated that an uninvolved heavy and light chain pair suppression (eg, suppression of immunoglobulin G [IgG] λ below the lower level of normal in a patient with a monoclonal IgG κ gammopathy) is an independent risk factor for progression to MM in patients with (MGUS). Our findings with consistent over- and underrecovery for polyclonal FLC reagents in the low range hamper future research considering the prognostic value of suppression of the noninvolved FLC. Research with N Latex reagents could in part resolve this problem, but all current guidelines for monitoring patients with monoclonal gammopathies are based on research performed with polyclonal Freelite reagents.