Scandinavian Journal of Clinical & Laboratory Investigation, 2014; 74: 575–581
ORIGINAL ARTICLE
Automated alarm to detect antigen excess in serum free immunoglobulin light chain kappa and lambda assays
PETTER URDAL1,3, ERIK K. AMUNDSEN1, KARIN TOSKA1 & OLAV KLINGENBERG2 1Department
of Medical Biochemistry, Oslo University Hospital, Ullevål and 2Rikshospitalet, Oslo, and 3Institute of Clinical Medicine, University of Oslo, Oslo, Norway
Abstract Background. Antigen excess causing a falsely low concentration result may occur when measuring serum free immunoglobulin light chains (SFLC). Automated antigen excess detection methods are available only with some analyzers. We have now developed and verified such a method. Methods. Residuals of sera with known SFLC-κ and -λ concentrations were analyzed using Binding Site reagents and methods adapted to the Roche Cobas® c.501 analyzer. Results. We analyzed 117 sera for SFLC-κ and -λ and examined how the absorbance increased with time during the 7 minutes of reaction (absorbance reading points 12–70). From this an antigen excess alarm factor (ratio of absorbance increases between reading points 68–60 and 20–12, multiplied by 100) was defined. Upon our request, Roche added to our two SFLC assays a program which calculated this antigen excess alarm factor and triggered an alarm when the factor was below a defined value. We verified this antigen excess alarm function by analyzing serum from 325 persons of whom 143 were multiple myeloma patients. All samples with a known concentration above 30 mg/L triggered either an antigen excess alarm, an ‘above test’ alarm or both. Also, all samples above 200 mg/L (SFLC-λ) and 300 mg/L (SFLC-κ) triggered the antigen excess alarm and all but one triggered the above test alarm. Conclusions. The antigen excess alarm function presented here detected all known antigen excess samples at no increased time of analysis, a reduced workload and reduced reagent cost. Key Words: Multiple myeloma, turbidimetry, nephelometry, antigen excess, free light chains
Introduction In a serum sent to us for analysis we measured immunoglobulin free light chains kappa (SFLC-κ) and lambda (SFLC-λ) to be 14 and 9 mg/L. The correct result for SFLC-κ was however later shown to be 1320 mg/L. Antigen excess occurs when antibody of a defined limited capacity is mixed with too high amounts of antigen. Antigen excess may occur with turbidimetric or nephelometric immunological assays for SFLC-κ and SFLC-λ [1–6] as well as with other proteins and different assay types [7–9], causing a false and mostly low concentration result. To detect antigen excess, the two main manufacturers of SFLC reagents and methods – Binding Site and Siemens Diagnostics – have included antigen excess alarm functions in their SFLC assays adapted to each of their analyzers
SPAplus and BN ProSpec instruments [10,11]. Binding Site has also included this function in their assay adapted to Cobas Integra® 800 [10] but not to other analyzers from Roche Diagnostics or from other companies. Laboratories using Binding Site reagents adapted to other analyzers are instead advised to dilute and reanalyze the serum when the SFLC result and/or the SFLC-κ/ SFLC-λ ratio is outside its reference range, in patients with previous antigen excess or if other clinical or laboratory findings indicate that the SFLC result might be falsely low [10]. These rules are work- and reagent-consuming. In addition, the rules would not have prevented us from reporting the wrong SFLC-κ result described above. In the present study we document an automated alarm function which detects a high SFLC concentration also in the
Correspondence: Petter Urdal, Department of Medical Biochemistry, Oslo University Hospital, Ullevål, Postboks 4956 Nydalen, 0424 Oslo, Norway. Fax: ⫹ 47 22118189. E-mail: [email protected] (Received 22 May 2013 ; accepted 25 March 2014 ) ISSN 0036-5513 print/ISSN 1502-7686 online © 2014 Informa Healthcare DOI: 10.3109/00365513.2014.915426
576
P. Urdal et al.
presence of antigen excess. This alarm function is based on Binding Site reagents and a Roche Cobas® c.501 analyzer, and uses absorbances already obtained when measuring SFLC-κ or -λ. It is similar but not identical to the antigen excess alarm functions of Binding Site, while Siemens Diagnostics uses a prereaction protocol [11].
Materials and methods Sera for use in the present study were selected among those sent to the department of medical biochemistry for analysis of SFLC-κ and SFLC-λ. During two time periods (December 2011–March 2012 and March–December 2012), two sets of sera (117 and 325 sera from 117 and 325 individuals) were collected. The first set was used for developing an antigen excess alarm function and the second set for verifying this alarm function. In principle, all sera above 200 mg/L in SFLC-κ or -λ were included, but some sera high in SFLC-κ, SFLC-λ and creatinine were not included in order to avoid a large dominance of sera from patients with renal disease. Sera below 200 mg/L in both SFLC-κ and -λ were included randomly. The sera were either stored at ⫹ 4°C for at most 7 days or stored at ⫺ 70°C for at most 6 months before being further analyzed as described below. The sera were collected at Oslo University Hospital, Ullevål and Rikshospitalet, after measurement of SFLC-κ and SFLC-λ concentrations by use of Binding Site (‘FreeLite’) reagents and methods, either adapted to a Roche Cobas® c.501 analyzer (Ullevål) or to a Siemens BN ProSpec® nephelometer (Rikshospitalet). These methods, used prior to inclusion, were run strictly as recommended by
Binding Site. They included an initial within-analyzer dilution of sample and automatic reanalysis at a fivefold lower sample fraction if the result of the first analysis was above 30 mg/L. If there was an ‘above test’ alarm reported together with the result of the automatic reanalysis, manual dilution and analysis of the diluted serum was performed. Samples in which either SFLC-κ or -λ were below 5 mg/L were also subjected to manual dilution and analysis. With one exception, i.e. SFLC-κ of the patient described in the Introduction, the SFLC concentrations measured before inclusion are the ones used in the Figures and Tables of the Results section. The patient mentioned in the Introduction who was included as a normal patient, was later found by the antigen excess alarm function to have a high SFLC-κ concentration, dilution and reanalysis found this to be 1320 mg/L. After inclusion we analyzed once more the 117 ⫹ 325 samples for SFLC-κ and SFLC-λ concentrations using the same reagents and the same method adapted to Roche Cobas® c.501. However, manual dilution and reanalysis, described above, was not performed since this had already been done prior to inclusion and was not necessary for development and verification of the antigen excess alarm function. This alarm function is described in more detail in the Results section. The project was approved by the local ethics committee.
Results Figure 1 shows the increasing absorbance from reading point 10 of eight samples analyzed for SFLC-κ. In sera with SFLC-κ concentration below 150 mg/L the absorbance increased almost linearly with time whereas in the sera well above this concentration the
Figure 1. Absorbance increase with time during analysis for SFLC-κ. The Roche Cobas® c.501 analyzer reads the absorbance 70 times with intervals of 8 seconds. After adding reagent 1 and sample, the reaction starts when the antibody-containing reagent 2 is added between reading points 10 and 11. The SFLC-κ concentrations (mg/L) of the eight sera examined are given on the right hand side. The 1320 mg/L sample is the one described in the Introduction.
Antigen excess alarm in SFLC assays early absorbance increase, between reading points 12 and 20, was higher and the late increase, between points 60 and 68, more moderate. In 117 sera we examined more closely how the early and late absorbance changes were related to the SFLC-κ and SFLC-λ concentration. The early absorbance change was continuously increasing with increasing SFLC-κ concentration (Figure 2a). The late absorbance change increased with increasing SFLC-κ concentrations up to approximately 100 mg/L but decreased for increasing SFLC-κ above 200 mg/L (Figure 2b). From all 117 SFLC-κ and -λ absorbance curves we calculated an antigen excess alarm factor (late absorbance change divided by early absorbance change multiplied by 100). The SFLC-κ alarm factor value and its relationship to the SFLC-κ concentration is shown in Figure 2c. A cut-off at 75 for the factor would discriminate most samples below 100 mg/L from all samples above 200 mg/L. Similarly, with SFLC-λ an antigen excess alarm factor of 50 would discriminate most samples below 100 mg/L from all samples above 250 mg/L (Figure 3). Upon our request, Roche added to the SFLC-κ and -λ assays a program which calculated the antigen excess alarm factor defined above and triggered an alarm when the factor was below 75 (SFLC-κ) or 50 (SFLC-λ). This alarm initiated an automatic rerun at a five-fold lower sample fraction. The alarm factor was printed as part of the result report.
577
The combined use of the above test alarm and the antigen excess alarm function was examined by analyzing the 325 samples of the second sample set, of which 143 were from multiple myeloma patients, on 27 separate days during 8 months. The sample described in the Introduction, considered at inclusion to have a SFLC-κ concentration of 14 mg/L, triggered the antigen excess alarm but not the above test alarm. Of the remaining 324 samples, none below and all above 30 mg/L in SFLC triggered the above test alarm. As for the antigen excess alarm, of the 324 samples none with concentrations in the range 15–100 mg/L triggered this alarm, whereas a small majority of the samples in the range 100–300 mg/L (SFLC-κ) and 100–200 mg/L (SFLC-λ) did (Figures 4 and 5). Above 300 mg/L (SFLC-κ) and 200 mg/L (SFLC-λ) all but two samples, one analyzed for SFLC-κ and the other for SFLC-λ, triggered the antigen excess alarm (Figures 4 and 5). If the antigen excess alarm function is modified as suggested below, these two samples would have also triggered an antigen excess alarm. A ‘false’ antigen excess alarm appeared with several samples below 11 mg/L, more frequently with the lambda (n ⫽ 12, Figure 5) than with the kappa assay (n ⫽ 2, Figure 4). The results of Figures 2a and 3a suggests that these alarms could all have been avoided if the antigen excess function had included a threshold value for the early absorbance change, the antigen excess factor being calculated
Figure 2. Early absorbance change (a), late absorbance change (b) and alarm factor value (c): Their relationship to SFLC-κ concentration. For 117 sera analyzed for SFLC-κ the early and late absorbance changes were estimated as indicated in Figure 1. The alarm factor value was defined as ratio between late and early absorbance change multiplied by 100.
578
P. Urdal et al.
Figure 3. Early absorbance change (a), late absorbance change (b) and alarm factor value (c): Their relationship to SFLC-λ concentration. For 117 sera analyzed for SFLC-λ the early and late absorbance changes and the alarm factor value were estimated as described in the Figure 2 legend.
only for sera above this threshold. By modifying the alarm function by (i) introducing early absorbance change threshold values of 0.030 for κ (cfr Figure 2a) and 0.060 for λ (cfr Figure 3a), and (ii) increasing the antigen excess alarm cut-off values to 100 (κ) and to 60 (λ) there would be no false alarms (Figures 4 and 5). The assay ranges of SFLC-κ and SFLC-λ as used with the Roche Cobas® c.501 analyzer were
approximately 5–60 mg/L and 5–75 mg/L, but Binding Site recommends rerun for samples above the upper reference limit, we use above 30 mg/L, to reduce the risk for undetected antigen excess. With seven of the 325 samples the first result was reported as a value below 80 mg/L (Table I, column 4) though their real SFLC concentration was much higher (Table I, column 3). These samples with antigen excess were the ones closest not to be detected if only
Figure 4. Verification of the antigen alarm function: The antigen excess factor of 325 samples analyzed for SFLC-κ. Shown as open or closed circles dependent upon whether the initial absorbance change was below or above 0.030, respectively.
Antigen excess alarm in SFLC assays
579
Figure 5. Verification of the antigen alarm function: The antigen excess factor of 325 samples analyzed for SFLC-λ. Shown as open or closed circles dependent upon whether the initial absorbance change was below or above 0.060, respectively.
an above test alarm was used. The antigen excess factors calculated with these samples are summarized in Table I, column 5. The five sera with high SFLC-κ concentrations (samples 1–5 in Table I) all showed very low antigen excess alarm factor values. The two high SFLC-λ sera showed moderately low antigen excess alarm factor values, but still below the alarm cut-off value of 60. Sample 3 in Table I is identical to sample 1320 mg/L in Figure 1 and is the sample described in the Introduction. On each day of analysis of the verification period a two-level quality control was run. During this period three different reagent lots were used both with the kappa and the lambda assays. This neither clearly affected the measured concentrations of the controls nor their antigen excess factor values. The quality control results for the whole verification period are summarized in Table II. The antigen excess factor values of SFLC-λ obtained in the development and the verification Table I. Selected* samples with antigen excess.
SFLC 1 2 3 4 5 6 7
κ κ κ κ κ λ λ
Concentration (mg/L)
Result (mg/L) reported after the first automatic analysis
Antigen excess factor
53400 3220 1320 14390 725 2200 6070
46.5 65.3 14.0 77.8 69.0 63.3 76.5
13 11 5 3 37 51 36
SFLC, Serum free immunoglobulin light chains. *Selected among the 325 verification samples as samples above 750 mg/L in SFLC-κ or -λ and where the result of the first analysis, before any rerun, was reported as below 80 mg/L.
time periods were largely similar. Thus, for samples with concentrations of SFLC-λ of 20–60 mg/L collected during the development time period the mean ⫾ standard deviation of the antigen excess factors (77.9 ⫾ 5.7, n ⫽ 41) was close to that of the verification period (80.2 ⫾ 6.6, n ⫽ 75). For samples with SFLC-κ concentrations of 20–60 mg/L the antigen excess factor values of the development time period (185.1 ⫾ 23.1, n ⫽ 88) differed only moderately from those of the verification period (195.0 ⫾ 23.7, n ⫽ 83).
Discussion Our antigen excess alarm is, in principle, triggered by any serum high in SFLC irrespective of whether antigen excess is also present. It thus supplements the ‘above test’ alarm which might not always give an alarm when there is antigen excess. Both alarms suggest reanalysis at a lower sample fraction, which when performed will reveal a high SFLC concentration if present.
Table II. Quality control results of the verification period. Concentration (mg/L)
Antigen excess factor
Control*
No
Mean
SD
Mean
SD
κ κ λ λ
27 22 27 26
14.6 29.3 27.0 54.9
0.72 2.09 0.97 1.28
233 256 89.7 85.3
21 16 5.3 3.9
low high low high
*The control sera used were those included in the Freelite human kappa (or lambda) free kit for use with Roche Cobas® c systems. SD, Standard deviation.
580
P. Urdal et al.
The most reliable way to detect antigen excess is by analyzing all sera at two dilutions. In the present study all sera above 30 mg/L, and those below 5 mg/L, were analyzed at two or more dilutions prior to inclusion whereas most sera in the SFLC concentration range 5–30 mg/L were analyzed at only one dilution. Our antigen excess alarm functioned reliably. All sera above 30 mg/L triggered one or both of the alarms when analyzed without manual predilution. The antigen excess alarm was not released by any serum in the concentration range 15–100 mg/L and by all sera above 300 mg/L. If in use as part of the routine assay, this reliable absence of antigen excess alarm when SFLC was above 30 but below 100 mg/L would have prevented many unnecessary manual dilutions and reanalysis. And the results shown in Figures 4 and 5 and in Table I do suggest that the antigen excess alarm has a quality for detecting high SFLC in the presence of antigen excess well above that of the ‘above test’ alarm. For sera with SFLC concentration in the range 5–30 mg/L, as measured prior to inclusion, we cannot exclude antigen excess present in some sera but missed by both types of alarm. As reported, we did find one sample included with initially reported normal SFLC-κ and SFLC-λ concentrations. When reanalyzed as part of the study, it triggered the antigen excess alarm and after dilution of sample the correct SFLC-κ concentration was found to be 1320 mg/L. Despite this case, we believe, from scarcity of reports and from our own previous experience, that antigen excess leading to a high SFLC concentration being reported as below 30 mg/L, is rare. Still, it is vital that we detect such cases of antigen excess. If we were to document that our antigen excess alarm effectively detected these probably rare cases, we would have to include a very large number of sera, and to analyze all at two dilutions. We found this to be outside the capacity of the present study, but still something that ought to be done before the antigen excess alarm may be taken into common use. We obtained many antigen excess alarms in sera below 15 mg/L in either SFLC-λ or -κ concentration. It would, however, be possible to avoid the unwanted alarms simply by restricting the antigen excess alarm function to samples where the early absorbance change is above a defined value. Such a restriction would also allow the use of a higher alarm cut-off value for the antigen excess factor, which might be of importance especially with the SFLC-λ assay where some samples with high concentrations have only moderately reduced antigen excess factor values. One might expect that a change in reagent lot may affect the antigen excess factor values measured. With the Binding Site antigen excess application the discriminatory value of the factor is established with each lot of reagents [12]. Our experience did not suggest lot adjustment to be necessary; the antigen
excess factor values of normal samples were only moderately affected by change in reagent lot. A similar antigen excess control is used by Roche to avoid antigen excess when measuring serum ferritin. However, the antigen excess control of SFLC may, one might argue, function less well given that the light chains of different myeloma patients each are monoclonal. They might differ in how they are recognized by the antibodies of the reagent. Figure 2a does not suggest a high degree of such variability for the early absorbance change, whereas Figures 4 and 5 do indicate some variability in alarm factor values. Still, excluding sera below approximately 15 mg/L by introducing an early absorbance change threshold (cfr Figures 2a, 3a, 4 and 5), the remaining sera below 100 mg/L in SFLC all showed high alarm factor values, clearly separable from the lower alarm factor values of sera above 200–300 mg/L in SFLC. Today, laboratories using Binding Site reagents adapted to other analyzers are advised to reanalyze at a lower sample fraction all samples in which the initial SFLC value is above the upper reference limit, usually above 20–30 mg/L [10,13]. Such samples are quite common among those received for SFLC analysis. For laboratories, it will be a benefit to gain access to SFLC methods that include an automated antigen excess detection. The antigen excess detection described here detected all antigen excess samples that we knew of at no increased time of analysis. If taken into use we expect it will give a safe way of detecting antigen excess, a reduced workload, and a reduced reagent cost.
Acknowledgements We thank Mohammed Jasim and Tore Øie, Roche Norway, for applying the automated antigen excess detection methods to the Roche Cobas® c.501 analyzer. We thank the medical technologists at the Department of Medical Biochemistry, Oslo University Hospital, Ullevål and Rikshospitalet, for their kind help during the project. Declaration of interest: The authors report no conflict of interest. The authors alone are responsible for the content and writing of the paper.
References [1] Murata K, Clark RJ, Lockington KS, Tostrud LJ, Greipp PR, Katzmann JA. Sharply increased serum free light-chain concentrations after treatment for multiple myeloma. Clin Chem 2010;56:16–20. [2] Daval S, Tridon A, Mazeron N, Ristori JM, Evrard B. Risk of antigen excess in serum free light chain measurements. Clin Chem 2007;53:1985–6. [3] Vercammen M, Meirlaen P, Broodtaerts L, Broek IV, Bossuyt X. Effect of sample dilution on serum free light
Antigen excess alarm in SFLC assays
[4]
[5]
[6]
[7]
chain concentration by immunonephelometric assay. Clin Chim Acta 2011;412:1798–804. McCudden CR, Voorhees PM, Hammet-Stabler CA. A case of hook effect in the serum free light chain assay using the Olympus AU400e. Clin Biochem 2009:42: 121–4. Bosmann M, Kössler J, Stolz H, Walter U, Knop S, Steigerwald U. Detection of serum free light chains: the problem with antigen excess. Clin Chem Lab Med 2010;48:1419–22. Murng SHK, Follows L, Whitfield P, Snowden JA, Swallow K, Green K, Sargur R, Egner W. Defining the impact of individual sample variability on routine immunoassay of serum free light chains (sFLC) in multiple myeloma. Clin Exp Immunol 2013;171:201–9. Papik K, Molnar B, Fedorcsak P, Schaefer R, Lang F, Sreter L, Feher J, Tulassay Z. Automated prozone effect detection in ferritin homogeneous immunoassays using neural network classifiers. Clin Chem Lab Med 1999; 37:471–6.
581
[8] Al-Mahdili HA, Jones GR. High-dose hook effect in six automated human chorionic gonadotrophin assays. Ann Clin Biochem 2010;47:383–5. [9] Akamatsu S, Tsukazaki H, Inoue K, Nishio Y. Advanced prostate cancer with extremely low prostate-specific antigen value at diagnosis: an example of high dose hook effect. Int J Urol 2006;13:1025–7. [10] Jenner E, Levoguer A, Evans J, Harding S. Serum free light chain immunoassays: a guide to antigen excess detection. Clin Chim Acta 2012;413:949. [11] Te Velthuis H, Knop I, Stam P, van den Broek M, Bos HK, Hol S, Teunissen E, Fischedick KS, Althaus H, Schmidt B, Wagner C, Melsert R. N Latex FLC – new monoclonal highperformance assays for the determination of free light chain kappa and lambda. Clin Chem Lab Med 2011;49:1323–32. [12] Product data sheet for human kappa free kit (for use on the SPAplus). Insert code: SIN 113DS, version 12 May 2011. [13] Product data sheet for human kappa free kit (for use on the Roche Cobas® systems). Insert code: SIN 129Dk, version 1 June 2011.
Copyright of Scandinavian Journal of Clinical & Laboratory Investigation is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
ORIGINAL ARTICLE
Automated alarm to detect antigen excess in serum free immunoglobulin light chain kappa and lambda assays
PETTER URDAL1,3, ERIK K. AMUNDSEN1, KARIN TOSKA1 & OLAV KLINGENBERG2 1Department
of Medical Biochemistry, Oslo University Hospital, Ullevål and 2Rikshospitalet, Oslo, and 3Institute of Clinical Medicine, University of Oslo, Oslo, Norway
Abstract Background. Antigen excess causing a falsely low concentration result may occur when measuring serum free immunoglobulin light chains (SFLC). Automated antigen excess detection methods are available only with some analyzers. We have now developed and verified such a method. Methods. Residuals of sera with known SFLC-κ and -λ concentrations were analyzed using Binding Site reagents and methods adapted to the Roche Cobas® c.501 analyzer. Results. We analyzed 117 sera for SFLC-κ and -λ and examined how the absorbance increased with time during the 7 minutes of reaction (absorbance reading points 12–70). From this an antigen excess alarm factor (ratio of absorbance increases between reading points 68–60 and 20–12, multiplied by 100) was defined. Upon our request, Roche added to our two SFLC assays a program which calculated this antigen excess alarm factor and triggered an alarm when the factor was below a defined value. We verified this antigen excess alarm function by analyzing serum from 325 persons of whom 143 were multiple myeloma patients. All samples with a known concentration above 30 mg/L triggered either an antigen excess alarm, an ‘above test’ alarm or both. Also, all samples above 200 mg/L (SFLC-λ) and 300 mg/L (SFLC-κ) triggered the antigen excess alarm and all but one triggered the above test alarm. Conclusions. The antigen excess alarm function presented here detected all known antigen excess samples at no increased time of analysis, a reduced workload and reduced reagent cost. Key Words: Multiple myeloma, turbidimetry, nephelometry, antigen excess, free light chains
Introduction In a serum sent to us for analysis we measured immunoglobulin free light chains kappa (SFLC-κ) and lambda (SFLC-λ) to be 14 and 9 mg/L. The correct result for SFLC-κ was however later shown to be 1320 mg/L. Antigen excess occurs when antibody of a defined limited capacity is mixed with too high amounts of antigen. Antigen excess may occur with turbidimetric or nephelometric immunological assays for SFLC-κ and SFLC-λ [1–6] as well as with other proteins and different assay types [7–9], causing a false and mostly low concentration result. To detect antigen excess, the two main manufacturers of SFLC reagents and methods – Binding Site and Siemens Diagnostics – have included antigen excess alarm functions in their SFLC assays adapted to each of their analyzers
SPAplus and BN ProSpec instruments [10,11]. Binding Site has also included this function in their assay adapted to Cobas Integra® 800 [10] but not to other analyzers from Roche Diagnostics or from other companies. Laboratories using Binding Site reagents adapted to other analyzers are instead advised to dilute and reanalyze the serum when the SFLC result and/or the SFLC-κ/ SFLC-λ ratio is outside its reference range, in patients with previous antigen excess or if other clinical or laboratory findings indicate that the SFLC result might be falsely low [10]. These rules are work- and reagent-consuming. In addition, the rules would not have prevented us from reporting the wrong SFLC-κ result described above. In the present study we document an automated alarm function which detects a high SFLC concentration also in the
Correspondence: Petter Urdal, Department of Medical Biochemistry, Oslo University Hospital, Ullevål, Postboks 4956 Nydalen, 0424 Oslo, Norway. Fax: ⫹ 47 22118189. E-mail: [email protected] (Received 22 May 2013 ; accepted 25 March 2014 ) ISSN 0036-5513 print/ISSN 1502-7686 online © 2014 Informa Healthcare DOI: 10.3109/00365513.2014.915426
576
P. Urdal et al.
presence of antigen excess. This alarm function is based on Binding Site reagents and a Roche Cobas® c.501 analyzer, and uses absorbances already obtained when measuring SFLC-κ or -λ. It is similar but not identical to the antigen excess alarm functions of Binding Site, while Siemens Diagnostics uses a prereaction protocol [11].
Materials and methods Sera for use in the present study were selected among those sent to the department of medical biochemistry for analysis of SFLC-κ and SFLC-λ. During two time periods (December 2011–March 2012 and March–December 2012), two sets of sera (117 and 325 sera from 117 and 325 individuals) were collected. The first set was used for developing an antigen excess alarm function and the second set for verifying this alarm function. In principle, all sera above 200 mg/L in SFLC-κ or -λ were included, but some sera high in SFLC-κ, SFLC-λ and creatinine were not included in order to avoid a large dominance of sera from patients with renal disease. Sera below 200 mg/L in both SFLC-κ and -λ were included randomly. The sera were either stored at ⫹ 4°C for at most 7 days or stored at ⫺ 70°C for at most 6 months before being further analyzed as described below. The sera were collected at Oslo University Hospital, Ullevål and Rikshospitalet, after measurement of SFLC-κ and SFLC-λ concentrations by use of Binding Site (‘FreeLite’) reagents and methods, either adapted to a Roche Cobas® c.501 analyzer (Ullevål) or to a Siemens BN ProSpec® nephelometer (Rikshospitalet). These methods, used prior to inclusion, were run strictly as recommended by
Binding Site. They included an initial within-analyzer dilution of sample and automatic reanalysis at a fivefold lower sample fraction if the result of the first analysis was above 30 mg/L. If there was an ‘above test’ alarm reported together with the result of the automatic reanalysis, manual dilution and analysis of the diluted serum was performed. Samples in which either SFLC-κ or -λ were below 5 mg/L were also subjected to manual dilution and analysis. With one exception, i.e. SFLC-κ of the patient described in the Introduction, the SFLC concentrations measured before inclusion are the ones used in the Figures and Tables of the Results section. The patient mentioned in the Introduction who was included as a normal patient, was later found by the antigen excess alarm function to have a high SFLC-κ concentration, dilution and reanalysis found this to be 1320 mg/L. After inclusion we analyzed once more the 117 ⫹ 325 samples for SFLC-κ and SFLC-λ concentrations using the same reagents and the same method adapted to Roche Cobas® c.501. However, manual dilution and reanalysis, described above, was not performed since this had already been done prior to inclusion and was not necessary for development and verification of the antigen excess alarm function. This alarm function is described in more detail in the Results section. The project was approved by the local ethics committee.
Results Figure 1 shows the increasing absorbance from reading point 10 of eight samples analyzed for SFLC-κ. In sera with SFLC-κ concentration below 150 mg/L the absorbance increased almost linearly with time whereas in the sera well above this concentration the
Figure 1. Absorbance increase with time during analysis for SFLC-κ. The Roche Cobas® c.501 analyzer reads the absorbance 70 times with intervals of 8 seconds. After adding reagent 1 and sample, the reaction starts when the antibody-containing reagent 2 is added between reading points 10 and 11. The SFLC-κ concentrations (mg/L) of the eight sera examined are given on the right hand side. The 1320 mg/L sample is the one described in the Introduction.
Antigen excess alarm in SFLC assays early absorbance increase, between reading points 12 and 20, was higher and the late increase, between points 60 and 68, more moderate. In 117 sera we examined more closely how the early and late absorbance changes were related to the SFLC-κ and SFLC-λ concentration. The early absorbance change was continuously increasing with increasing SFLC-κ concentration (Figure 2a). The late absorbance change increased with increasing SFLC-κ concentrations up to approximately 100 mg/L but decreased for increasing SFLC-κ above 200 mg/L (Figure 2b). From all 117 SFLC-κ and -λ absorbance curves we calculated an antigen excess alarm factor (late absorbance change divided by early absorbance change multiplied by 100). The SFLC-κ alarm factor value and its relationship to the SFLC-κ concentration is shown in Figure 2c. A cut-off at 75 for the factor would discriminate most samples below 100 mg/L from all samples above 200 mg/L. Similarly, with SFLC-λ an antigen excess alarm factor of 50 would discriminate most samples below 100 mg/L from all samples above 250 mg/L (Figure 3). Upon our request, Roche added to the SFLC-κ and -λ assays a program which calculated the antigen excess alarm factor defined above and triggered an alarm when the factor was below 75 (SFLC-κ) or 50 (SFLC-λ). This alarm initiated an automatic rerun at a five-fold lower sample fraction. The alarm factor was printed as part of the result report.
577
The combined use of the above test alarm and the antigen excess alarm function was examined by analyzing the 325 samples of the second sample set, of which 143 were from multiple myeloma patients, on 27 separate days during 8 months. The sample described in the Introduction, considered at inclusion to have a SFLC-κ concentration of 14 mg/L, triggered the antigen excess alarm but not the above test alarm. Of the remaining 324 samples, none below and all above 30 mg/L in SFLC triggered the above test alarm. As for the antigen excess alarm, of the 324 samples none with concentrations in the range 15–100 mg/L triggered this alarm, whereas a small majority of the samples in the range 100–300 mg/L (SFLC-κ) and 100–200 mg/L (SFLC-λ) did (Figures 4 and 5). Above 300 mg/L (SFLC-κ) and 200 mg/L (SFLC-λ) all but two samples, one analyzed for SFLC-κ and the other for SFLC-λ, triggered the antigen excess alarm (Figures 4 and 5). If the antigen excess alarm function is modified as suggested below, these two samples would have also triggered an antigen excess alarm. A ‘false’ antigen excess alarm appeared with several samples below 11 mg/L, more frequently with the lambda (n ⫽ 12, Figure 5) than with the kappa assay (n ⫽ 2, Figure 4). The results of Figures 2a and 3a suggests that these alarms could all have been avoided if the antigen excess function had included a threshold value for the early absorbance change, the antigen excess factor being calculated
Figure 2. Early absorbance change (a), late absorbance change (b) and alarm factor value (c): Their relationship to SFLC-κ concentration. For 117 sera analyzed for SFLC-κ the early and late absorbance changes were estimated as indicated in Figure 1. The alarm factor value was defined as ratio between late and early absorbance change multiplied by 100.
578
P. Urdal et al.
Figure 3. Early absorbance change (a), late absorbance change (b) and alarm factor value (c): Their relationship to SFLC-λ concentration. For 117 sera analyzed for SFLC-λ the early and late absorbance changes and the alarm factor value were estimated as described in the Figure 2 legend.
only for sera above this threshold. By modifying the alarm function by (i) introducing early absorbance change threshold values of 0.030 for κ (cfr Figure 2a) and 0.060 for λ (cfr Figure 3a), and (ii) increasing the antigen excess alarm cut-off values to 100 (κ) and to 60 (λ) there would be no false alarms (Figures 4 and 5). The assay ranges of SFLC-κ and SFLC-λ as used with the Roche Cobas® c.501 analyzer were
approximately 5–60 mg/L and 5–75 mg/L, but Binding Site recommends rerun for samples above the upper reference limit, we use above 30 mg/L, to reduce the risk for undetected antigen excess. With seven of the 325 samples the first result was reported as a value below 80 mg/L (Table I, column 4) though their real SFLC concentration was much higher (Table I, column 3). These samples with antigen excess were the ones closest not to be detected if only
Figure 4. Verification of the antigen alarm function: The antigen excess factor of 325 samples analyzed for SFLC-κ. Shown as open or closed circles dependent upon whether the initial absorbance change was below or above 0.030, respectively.
Antigen excess alarm in SFLC assays
579
Figure 5. Verification of the antigen alarm function: The antigen excess factor of 325 samples analyzed for SFLC-λ. Shown as open or closed circles dependent upon whether the initial absorbance change was below or above 0.060, respectively.
an above test alarm was used. The antigen excess factors calculated with these samples are summarized in Table I, column 5. The five sera with high SFLC-κ concentrations (samples 1–5 in Table I) all showed very low antigen excess alarm factor values. The two high SFLC-λ sera showed moderately low antigen excess alarm factor values, but still below the alarm cut-off value of 60. Sample 3 in Table I is identical to sample 1320 mg/L in Figure 1 and is the sample described in the Introduction. On each day of analysis of the verification period a two-level quality control was run. During this period three different reagent lots were used both with the kappa and the lambda assays. This neither clearly affected the measured concentrations of the controls nor their antigen excess factor values. The quality control results for the whole verification period are summarized in Table II. The antigen excess factor values of SFLC-λ obtained in the development and the verification Table I. Selected* samples with antigen excess.
SFLC 1 2 3 4 5 6 7
κ κ κ κ κ λ λ
Concentration (mg/L)
Result (mg/L) reported after the first automatic analysis
Antigen excess factor
53400 3220 1320 14390 725 2200 6070
46.5 65.3 14.0 77.8 69.0 63.3 76.5
13 11 5 3 37 51 36
SFLC, Serum free immunoglobulin light chains. *Selected among the 325 verification samples as samples above 750 mg/L in SFLC-κ or -λ and where the result of the first analysis, before any rerun, was reported as below 80 mg/L.
time periods were largely similar. Thus, for samples with concentrations of SFLC-λ of 20–60 mg/L collected during the development time period the mean ⫾ standard deviation of the antigen excess factors (77.9 ⫾ 5.7, n ⫽ 41) was close to that of the verification period (80.2 ⫾ 6.6, n ⫽ 75). For samples with SFLC-κ concentrations of 20–60 mg/L the antigen excess factor values of the development time period (185.1 ⫾ 23.1, n ⫽ 88) differed only moderately from those of the verification period (195.0 ⫾ 23.7, n ⫽ 83).
Discussion Our antigen excess alarm is, in principle, triggered by any serum high in SFLC irrespective of whether antigen excess is also present. It thus supplements the ‘above test’ alarm which might not always give an alarm when there is antigen excess. Both alarms suggest reanalysis at a lower sample fraction, which when performed will reveal a high SFLC concentration if present.
Table II. Quality control results of the verification period. Concentration (mg/L)
Antigen excess factor
Control*
No
Mean
SD
Mean
SD
κ κ λ λ
27 22 27 26
14.6 29.3 27.0 54.9
0.72 2.09 0.97 1.28
233 256 89.7 85.3
21 16 5.3 3.9
low high low high
*The control sera used were those included in the Freelite human kappa (or lambda) free kit for use with Roche Cobas® c systems. SD, Standard deviation.
580
P. Urdal et al.
The most reliable way to detect antigen excess is by analyzing all sera at two dilutions. In the present study all sera above 30 mg/L, and those below 5 mg/L, were analyzed at two or more dilutions prior to inclusion whereas most sera in the SFLC concentration range 5–30 mg/L were analyzed at only one dilution. Our antigen excess alarm functioned reliably. All sera above 30 mg/L triggered one or both of the alarms when analyzed without manual predilution. The antigen excess alarm was not released by any serum in the concentration range 15–100 mg/L and by all sera above 300 mg/L. If in use as part of the routine assay, this reliable absence of antigen excess alarm when SFLC was above 30 but below 100 mg/L would have prevented many unnecessary manual dilutions and reanalysis. And the results shown in Figures 4 and 5 and in Table I do suggest that the antigen excess alarm has a quality for detecting high SFLC in the presence of antigen excess well above that of the ‘above test’ alarm. For sera with SFLC concentration in the range 5–30 mg/L, as measured prior to inclusion, we cannot exclude antigen excess present in some sera but missed by both types of alarm. As reported, we did find one sample included with initially reported normal SFLC-κ and SFLC-λ concentrations. When reanalyzed as part of the study, it triggered the antigen excess alarm and after dilution of sample the correct SFLC-κ concentration was found to be 1320 mg/L. Despite this case, we believe, from scarcity of reports and from our own previous experience, that antigen excess leading to a high SFLC concentration being reported as below 30 mg/L, is rare. Still, it is vital that we detect such cases of antigen excess. If we were to document that our antigen excess alarm effectively detected these probably rare cases, we would have to include a very large number of sera, and to analyze all at two dilutions. We found this to be outside the capacity of the present study, but still something that ought to be done before the antigen excess alarm may be taken into common use. We obtained many antigen excess alarms in sera below 15 mg/L in either SFLC-λ or -κ concentration. It would, however, be possible to avoid the unwanted alarms simply by restricting the antigen excess alarm function to samples where the early absorbance change is above a defined value. Such a restriction would also allow the use of a higher alarm cut-off value for the antigen excess factor, which might be of importance especially with the SFLC-λ assay where some samples with high concentrations have only moderately reduced antigen excess factor values. One might expect that a change in reagent lot may affect the antigen excess factor values measured. With the Binding Site antigen excess application the discriminatory value of the factor is established with each lot of reagents [12]. Our experience did not suggest lot adjustment to be necessary; the antigen
excess factor values of normal samples were only moderately affected by change in reagent lot. A similar antigen excess control is used by Roche to avoid antigen excess when measuring serum ferritin. However, the antigen excess control of SFLC may, one might argue, function less well given that the light chains of different myeloma patients each are monoclonal. They might differ in how they are recognized by the antibodies of the reagent. Figure 2a does not suggest a high degree of such variability for the early absorbance change, whereas Figures 4 and 5 do indicate some variability in alarm factor values. Still, excluding sera below approximately 15 mg/L by introducing an early absorbance change threshold (cfr Figures 2a, 3a, 4 and 5), the remaining sera below 100 mg/L in SFLC all showed high alarm factor values, clearly separable from the lower alarm factor values of sera above 200–300 mg/L in SFLC. Today, laboratories using Binding Site reagents adapted to other analyzers are advised to reanalyze at a lower sample fraction all samples in which the initial SFLC value is above the upper reference limit, usually above 20–30 mg/L [10,13]. Such samples are quite common among those received for SFLC analysis. For laboratories, it will be a benefit to gain access to SFLC methods that include an automated antigen excess detection. The antigen excess detection described here detected all antigen excess samples that we knew of at no increased time of analysis. If taken into use we expect it will give a safe way of detecting antigen excess, a reduced workload, and a reduced reagent cost.
Acknowledgements We thank Mohammed Jasim and Tore Øie, Roche Norway, for applying the automated antigen excess detection methods to the Roche Cobas® c.501 analyzer. We thank the medical technologists at the Department of Medical Biochemistry, Oslo University Hospital, Ullevål and Rikshospitalet, for their kind help during the project. Declaration of interest: The authors report no conflict of interest. The authors alone are responsible for the content and writing of the paper.
References [1] Murata K, Clark RJ, Lockington KS, Tostrud LJ, Greipp PR, Katzmann JA. Sharply increased serum free light-chain concentrations after treatment for multiple myeloma. Clin Chem 2010;56:16–20. [2] Daval S, Tridon A, Mazeron N, Ristori JM, Evrard B. Risk of antigen excess in serum free light chain measurements. Clin Chem 2007;53:1985–6. [3] Vercammen M, Meirlaen P, Broodtaerts L, Broek IV, Bossuyt X. Effect of sample dilution on serum free light
Antigen excess alarm in SFLC assays
[4]
[5]
[6]
[7]
chain concentration by immunonephelometric assay. Clin Chim Acta 2011;412:1798–804. McCudden CR, Voorhees PM, Hammet-Stabler CA. A case of hook effect in the serum free light chain assay using the Olympus AU400e. Clin Biochem 2009:42: 121–4. Bosmann M, Kössler J, Stolz H, Walter U, Knop S, Steigerwald U. Detection of serum free light chains: the problem with antigen excess. Clin Chem Lab Med 2010;48:1419–22. Murng SHK, Follows L, Whitfield P, Snowden JA, Swallow K, Green K, Sargur R, Egner W. Defining the impact of individual sample variability on routine immunoassay of serum free light chains (sFLC) in multiple myeloma. Clin Exp Immunol 2013;171:201–9. Papik K, Molnar B, Fedorcsak P, Schaefer R, Lang F, Sreter L, Feher J, Tulassay Z. Automated prozone effect detection in ferritin homogeneous immunoassays using neural network classifiers. Clin Chem Lab Med 1999; 37:471–6.
581
[8] Al-Mahdili HA, Jones GR. High-dose hook effect in six automated human chorionic gonadotrophin assays. Ann Clin Biochem 2010;47:383–5. [9] Akamatsu S, Tsukazaki H, Inoue K, Nishio Y. Advanced prostate cancer with extremely low prostate-specific antigen value at diagnosis: an example of high dose hook effect. Int J Urol 2006;13:1025–7. [10] Jenner E, Levoguer A, Evans J, Harding S. Serum free light chain immunoassays: a guide to antigen excess detection. Clin Chim Acta 2012;413:949. [11] Te Velthuis H, Knop I, Stam P, van den Broek M, Bos HK, Hol S, Teunissen E, Fischedick KS, Althaus H, Schmidt B, Wagner C, Melsert R. N Latex FLC – new monoclonal highperformance assays for the determination of free light chain kappa and lambda. Clin Chem Lab Med 2011;49:1323–32. [12] Product data sheet for human kappa free kit (for use on the SPAplus). Insert code: SIN 113DS, version 12 May 2011. [13] Product data sheet for human kappa free kit (for use on the Roche Cobas® systems). Insert code: SIN 129Dk, version 1 June 2011.
Copyright of Scandinavian Journal of Clinical & Laboratory Investigation is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
