Printer Friendly

The importance of commutability of reference materials used as calibrators: the example of ceruloplasmin.

BACKGROUND: Different methods for ceruloplasmin tend to give different results in external quality assessment schemes. During the production of the certified reference material ERM-DA470k/IFCC discrepant measurement results were also found for ceruloplasmin measured with different methods, and consequently the protein could not be certified in the material.

METHODS: We performed a commutability study with 30 serum samples and the reference materials ERMDA470, ERM-DA470k/IFCC, and ERM-DA472/ IFCC, using 6 different methods. Data were analyzed according to the CLSI Guideline C53-A to assess whether the reference materials had the same behavior as the serum samples with respect to measurement results obtained with combinations of the methods used.

RESULTS: Measurement results from different methods showed a good linear correlation for the serum samples. ERM-DA470 showed marked noncommutability for certain combinations of methods. ERMDA470k/IFCC and ERM-DA472/IFCC were commutable for more combinations of methods. The lack of commutability of ERM-DA470 for certain combinations of methods correlates with results from the UK National External Quality Assessment Service showing discrepancies between results from these methods. For serum stored in the presence of sodium azide the results from different methods are essentially equivalent.

CONCLUSIONS: Ceruloplasmin in ERM-DA470 is a fully documented example of a situation in which, due to lack of commutability, the use of a common material for calibration did not lead to harmonization .

Ceruloplasmin is an enzyme synthesized in the liver containing 6 copper ions in its structure (1), and it is the major copper-carrying protein in the blood (2). Ceruloplasmin exhibits copper-dependent oxidase activity, which is associated with possible oxidation of [Fe.sup.2+] (ferrous iron) into [Fe.sup.3+] (ferric iron). Decreased concentrations of ceruloplasmin may indicate aceruloplasminemia, Menkes disease, or Wilson disease.

Since 1993 the majority of in vitro diagnostic measurement results for ceruloplasmin have been traceable to the certified value in the reference material ERM-DA470, which was certified for 15 serum proteins, including ceruloplasmin (3). The use of ERM-DA470 has led to a reduction of between-assay variation for the majority of the certified proteins (4). For ceruloplasmin, discrepancies between results from different methods continued to be observed by UK National External Quality Assessment Service (NEQAS)6 (5). A comparison of results obtained with a nephelometric (Beckman Immage) and a turbidimetric (Roche Cobas c 501) method showed both proportional and constant bias between the methods, despite the fact that both were traceable to the property value in ERM-DA470 (6).

In 2008, the Institute for Reference Materials and Measurements (IRMM) released ERM-DA470k/IFCC as a replacement for ERM-DA470. ERM-DA470 was used for the calibration of all measurements performed for the value assignment of ERM-DA470k/IFCC. For ceruloplasmin there were considerable discrepancies between results from different methods falling into 2 main groups. As a consequence no certified value could be assigned for this protein (7, 8).

A dedicated study was performed to assess whether these discrepant results were due to lack of commutability of ERM-DA470. Commutability of a material is a prerequisite if the common use of the material for calibration is to lead to equivalence of measurement results obtained with different methods. Commutability has been defined in the CLSI Guideline C53-A as a property of a reference material, demonstrated by the equivalence of the mathematical relationships among the results of different measurement procedures for a reference material and for representative samples of the type intended to be measured (9). The Guideline C53-A describes approaches for the qualification of commutable reference materials. Here we present results of the investigation of the commutability of (candidate) reference materials for ceruloplasmin.

Materials and Methods

The study involved the measurement of serum samples and (candidate) reference materials with 8 different methods (Table 1). Data were analyzed to assess whether the reference materials had the same properties as the serum samples with respect to their response in combinations of the methods used.

SAMPLES MEASURED

Each laboratory received 30 individual serum samples, consisting of 0.5-mL aliquots from serum donations used to produce the certified reference material ERM-DA470k/IFCC (10). They were selected to cover the measurement interval for ceruloplasmin, but as they were obtained from healthy people there were no samples with very low values. The samples had been stored at -70[degrees]C for 15 months. Laboratories also received the following (candidate) reference materials: 1 vial of ERM-DA470, 3 vials of the lyophilized ERM-DA470k/ IFCC (certified values for 12 proteins), and 3 vials of the liquid frozen ERM-DA472/IFCC (certified value for C-reactive protein) (11). ERM-DA470k/IFCC and ERM-DA472/IFCC were prepared from the same pool of processed serum, but the first was lyophilized and the second liquid frozen at -70 [degrees]C.

The serum samples and reference materials were dispatched to the participating laboratories on dry ice, together with detailed protocols and reporting sheets. Lyophilized materials were reconstituted according to the protocol in the certificate, the day before the measurements. Serum samples were thawed on the day of the measurements.

The commutability of the serum samples that had been stored frozen was verified by comparing the results for these samples with the results of 10 fresh patient samples, measured both undiluted and at a 50% (volume fraction) dilution. The fresh patient samples were anonymized routine samples that had been stored at 4 [degrees]C for less than a week. These fresh serum samples were measured next to the 30 frozen serum samples and the reference materials, with a BN ProSpec (duplicate measurements) and an Immage (triplicate measurements).

For External Quality Assessment Service (EQAS) samples units of venous blood were collected into "dry" bags (without anticoagulant) from healthy donors attending the local UK National Blood Service donor center and allowed to coagulate. The blood was then centrifuged and the serum collected, frozen, and stored as individual donations at -40[degrees]C for 1-4 weeks. These serum donations were thawed, pooled (3-20 donations per pool), and mixed, and any stabilizing agent (sodium azide, Bronidox, or none) added, and then the pool was divided into aliquots. Aliquots were stored at 4[degrees]C and distributed to participants by post at ambient temperature within 5 days of pool preparation (fresh samples), or frozen and distributed after storage at -40[degrees]C for 4-16 months (aged samples).

MEASUREMENT PROCEDURE

The methods used are summarized in Table 1. Laboratories reconstituted the samples and performed the measurements according to a detailed protocol, and reported data in reporting sheets. All samples were measured in triplicate. First they were all measured in one analytical sequence in a first order. Then they were remeasured in the same order, and finally in reverse order. Control samples were measured at the beginning and at the end of each sequence. Measurement results for the reconstituted lyophilized reference materials were normalized taking into consideration the exact mass of the water used to reconstitute the materials.

DATA ANALYSIS

Method comparison. Results from different methods were compared to each other on the basis of the mean values for the serum samples, using Passing-Bablok regression as implemented in the software program Analyse-It. The Pearson correlation coefficient was also calculated for each method comparison.

Commutability. We assessed commutability according to procedures described in CLSI Guideline C53-A (9) using Excel and R (12). The analysis is based on the pair-wise comparison of sets of results using Deming regression, the calculation of a 95% prediction interval, and the conclusion on the commutability of the reference material on the basis of the position of its values with respect to the prediction interval.

EQAS data. Data from distributions of pools of serum distributed by UK NEQAS over the years 2001-2008 were examined. The pools were stored frozen at -70[degrees]C either without preservative, with sodium azide (1 g/L), or with Bronidox (1 g/L). The same pools were distributed 2 or 3 times, with intervals of 4-16 months between the first and last distribution. The methods were grouped by platforms as Roche Integra, Dade (BN ProSpec and BNII), Immage, Olympus, and Abbott. The groupings were those used by UK NEQAS at the time of the measurements. Values reported were trimmed means for each method, expressed relative to the all-laboratory trimmed mean (ALTM).

Results

METHOD COMPARISONS

The mean CVs for the measurement results of the serum samples obtained when using the different methods were between 1.7% and 6.1%. The Pearson correlation coefficients were >0.98 for all pair-wise method comparisons. This indicates a very good correlation of the results of different methods. In contrast the slopes for the method comparisons varied from 0.56 to 1.42 (Table 2). These values show that the measurement of the same serum sample provides very discrepant results with the different methods.

COMMUTABILITY OF ERM-D470, ERM-DA470K/IFCC, AND ERM-DA472

Laboratories measured a set of 30 individual serum samples together with ERM-DA472/IFCC, reconstituted ERM-DA40, and reconstituted ERM-DA40k/ IFCC in a single analytical run.

The suitability of the frozen samples used for the commutability study was verified in one laboratory in which both the BN ProSpec and the Beckmann Immage methods were implemented. The slopes of the measurement results of BN ProSpec with respect to Immage were 0.699 for the frozen samples and 0.705 for the fresh samples; the difference between the slopes was not statistically significant.

The commutability of the reference materials was evaluated by different approaches. The results obtained using Deming regression and comparison of the means of the results for the reference material with the prediction interval are shown in Fig. 1. Table 2 summarizes the results of the commutability study for ERM-DA470. It is evident that ERM-DA470 shows considerable lack of commutability for certain combinations of methods. On the other hand it is commutable for all combinations of Roche Integra, Roche Hitachi, Dako Hitachi, and AU2700. ERM-DA470k/IFCC is commutable for most combinations, except for Dako Hitachi combined with Abbott Architect, Immage, or BN ProSpec I.

In some cases the 2 BN ProSpec datasets gave different results because the reference material is close to the limit of the prediction interval and will be inside or outside the prediction interval depending on the width of the prediction interval, which is a function of the repeatability obtained in that particular laboratory.

RESULTS FROM UK NEQAS

UK NEQAS regularly distributes aliquots of pooled patient EQAS samples to laboratories. The results had shown a lack of equivalence between results from different methods. In Fig. 2 a more detailed analysis of results from a large number of proficiency test distributions is shown. In particular, we analyzed results from samples that have been distributed both as fresh samples and after different periods of storage. The following trends can be observed.

For fresh EQAS samples there was a systematic bias between results obtained with different methods. Measurements with Beckman Immage gave the highest ceruloplasmin values and measurements, with Roche and Olympus methods giving the lowest values (Fig. 2A). As an example the UK NEQAS results from March 2008 showed the following B-scores (bias with respect to the ALTM): Beckman Immage, 15.5; Dade Behring nephelometry, 5.1; Abbott, 2.3; Olympus, -5.3; and Roche Integra, -14.2.

For EQAS samples that were stored after the addition of sodium azide and then redistributed, the discrepancies between results from different methods are considerably reduced. For fresh EQAS samples the between-method CV was between 9% and 25%, and for EQAS samples containing sodium azide the between-method CV wasbetween 1% and 6% (Fig. 2, B and C). For aged EQAS samples the results from different methods in fact agreed very well.

EQAS samples stored in the absence of a stabilizing agent showed variable ageing effects.

EQAS samples stored in the presence of Bronidox showed very comparable between-method CVs before and after storage (Fig. 2C). The means per method did not change significantly after different storage times (Fig. 2D).

Discussion

Measurement results for ceruloplasmin evaluated in this study are claimed to be traceable to the ceruloplasmin value in ERM-DA470 because this material was used in the calibration process. The results from the commutability study presented here show that the ERM-DA470 is not commutable for ceruloplasmin measurements for a substantial number of combinations of methods. It is commutable for subgroups of methods, e.g., for all combinations of Roche Integra, Roche Hitachi, Dako Hitachi, and Olympus AU2700 and for combinations of the Abbott Architect with BN ProSpec and Immage.

The deviation of results for ERM-DA470 from the correlation line for serum samples is largest for the combination of Immage and Dako Hitachi (Fig. 1C). This is also the combination of methods for which results for EQAS samples are most discrepant, with Immage giving high values and Dako Hitachi giving low values.

In the results from UK NEQAS, methods using Olympus and Roche Integra platforms have a negative bias with respect to the means. Methods using Immage, Dade Behring, and Abbott platforms have a positive bias. This clustering is the same as found in the commutability studies. In Fig. 3 results from UK NEQAS for a material distributed when fresh and after storage for 5 and 11 months in the presence of Bronidox are compared to results from the commutability study for 3 individual serum samples. It is clear that the trend is the same for both types of material. ERM-DA470 is commutable for combinations of Olympus and Roche, which both have a negative bias, or for combinations of Abbott and BN ProSpec (from Dade Behring) and Immage, which both have a positive bias.

The same clustering was also found during the value assignment of ERM-DA470k/IFCC (7, 8). ERMDA470k/IFCC was produced to replace ERM-DA470. The intention was to certify the concentration of ceruloplasmin in ERM-DA470k/IFCC. Value assignment measurements were carried out by 13 laboratories, using BN ProSpec, BNII, Olympus AU640, Immage, Abbott, and LX-2200 platforms with the reagents from the respective companies, and Hitachi platforms with Roche and Dako reagents. These measurements were all calibrated with ERM-DA470. With the Immage and Abbott methods high values were found for ceruloplasmin in ERM-DA470k/IFCC, whereas Olympus, Roche, and Dako Hitachi methods gave low values. The new reference material ERM-DA470k/IFCC was found to behave more like a fresh patient sample, and the bias between the values of the different methods was comparable to that found in UK-NEQAS. The commutability study also showed that ERM-DA470k/IFCC behaves more like a typical serum sample. ERM-DA470k/IFCC and ERM-DA472/IFCC are both commutable for more combinations of methods (see Table 1 in the Data Supplement that accompanies the online version of this report at http:// www.clinchem.org/content/vol59/issue9). However, the results from value assignment could not be used for certification because ERM-DA470, which was used for calibration, was not commutable. It is also possible that ERM-DA470k/IFCC and ERM-DA472/IFCC become less commutable upon storage, as both contain sodium azide. Therefore ceruloplasmin is not yet certified in ERM-DA470k/IFCC.

One could speculate on the causes of the lack of commutability of "aged" serum materials (serum materials stored frozen after addition of sodium azide), whether as EQAS samples or as reference materials. It was proposed that an alteration in ceruloplasmin structure upon storage leads to changes in antigen-antibody binding characteristics (5). However, further investigations are necessary to fully understand the molecular changes caused by the ageing process.

In conclusion, the reference material ERM-DA470 behaves like an aged patient sample with respect to the measurement of ceruloplasmin. When the aged ERM-DA470 was the basis of the traceability for results of fresh EQAS samples, the between- method variation was high. When aged ERM-DA470 was used to provide traceability for aged EQAS samples the between-method variation was very low. The fact that it was used by manufacturers to transfer values to in-house and kit calibrators then resulted in between-method bias for fresh serum samples. When aged samples, whose properties in terms of ceruloplasmin analysis resemble ERM-DA470, were distributed by EQAS the between-method variation was very low. This indicates that the value transfer chain, reference material--in-house calibrator--kit calibrator--patient sample, was reasonably well under control by the manufacturers. This is also shown by the fact that for all methods the ceruloplasmin values found for ERM-DA470 are close to the certified value, 0.205 g/L with expanded uncertainty = 0.011 g/L. Therefore the between-method variation can to a large extent be explained by the lack of commutability of ERM-DA470.

When ERM-DA470 was produced, all processing and value assignment procedures were extensively tested and validated. The reference material had led to a major improvement in the standardization of many serum proteins (4). However, proteins in serum are very complex, and minor modifications in either the protein of interest or the matrix can have a great impact on their measurement by homogeneous immunoassay. These modifications are often not detected by single-method approaches. The verification of the commutability of new reference materials and EQAS samples is necessary to prevent the introduction of method-dependent bias in measurement results. However, many reference materials used on a regular basis have not been characterized in terms of commutability. Therefore external quality assessment schemes may uncover between-method variations which then have to be investigated regarding the commutability of reference materials. However, this requires that the EQAS samples themselves are commutable. The present work and the publication from Beetham and colleagues (5) show that EQAS samples may lack commutability even when their processing is limited to pooling, addition of sodium azide, and freezing.

The example of ceruloplasmin clearly shows that the use of a common calibrant that is not commutable will not result in full equivalence of results obtained with different methods.

Author Contributions: All authors confirmed they have contributed to the intellectual content of this paper and have met the following 3 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; and (c) final approval of the published article.

Authors' Disclosures or Potential Conflicts of Interest: Upon manuscript submission, all authors completed the author disclosure form. Disclosures and/or potential conflicts of interest:

Employment or Leadership: None declared.

Consultant or Advisory Role: T. Keller, IRMM.

Stock Ownership: None declared.

Honoraria: None declared.

Research Funding: None declared.

Expert Testimony: None declared.

Patents: None declared.

Role of Sponsor: No sponsor was declared.

Acknowledgments: We thank the laboratories that participated in the commutability study, and in particular Suzanne Adam (Roche Professional Diagnostics Ltd), Pia Lundblad (Dako), Matt McCusker (Beckman Coulter Biomedical Ltd.), Fabio Rota and Lucini Ruggero (Sentinel Diagnostics), and Wiebke Schreiber (Siemens Healthcare Products GmbH).

References

(1.) Holmberg CG, Laurell CB. Investigations in serum copper II: isolation of the copper-containing protein and a description of some of its properties. Acta Chem Scand 1948; 2:550-6.

(2.) Twomey PJ, Viljoen A, House IM, Reynolds TM, Wierzbicki AS. Relationship between serum copper, ceruloplasmin, and non-ceruloplasmin-bound copper in routine clinical practice. Clin Chem 2005; 51:1558-9.

(3.) Baudner S, Bienvenu J, Blirup-Jensen S, Carlstrom A, Johnson AM, Milford Ward A, et al. The certification of a matrix reference material for imunochemical measurement of 14 human serum proteins CRM470. Report EUR 15423 EN and 16882 N. Luxembourg: Office for Official Publications of the European Communities; 1993.

(4.) Merlini G, Blirup-Jensen S, Johnson AM, Sheldon J, Zegers I; IFCC Committee on Plasma Proteins (C-PP). Standardizing plasma protein measurements worldwide: a challenging enterprise. Clin Chem Lab Med 2010; 48:1567-75.

(5.) Beetham R, White P, Riches P, Bullock D, MacKenzie F; UK NEQAS Protein Assays Specialist Advisory Group. Use of CRM 470/RPPHS has not achieved true consensus for ceruloplasmin measurement. Clin Chem 2002; 48:2293-4.

(6.) Infusino I, Valente C, Dolci A, Panteghini M. Standardization of ceruloplasmin measurements is still an issue despite the availability of a common reference material. Anal Bioanal Chem 2009; 397:521-5.

(7.) Zegers I, Keller T, Schreiber W, Sheldon J, Albertini R, Merlini G, et al. Characterisation of the new serum protein reference material ERMDA470k/IFCC: value assignment by immunoassay. Clin Chem 2010; 56:1880-8.

(8.) Schimmel H, Zegers I, Emons H. Standardization of protein biomarker measurements: is it feasible? Scand J Clin Lab Invest Suppl 2010; 242:2733.

(9.) Vesper HW, Emons H, Gnezda M, Jain CP, Miller WG, Rej R, et al. Characterization and qualification of commutable reference materials for laboratory medicine; approved guideline. Wayne (PA): CLSI; 2010. CLSI document C53-A; vol. 30, no. 12.

(10.) Zegers I, Schreiber W, Linstead S, Lammers M, McCusker M, Munoz A, et al. Development and preparation of a new serum protein reference material: feasibility studies and processing. Clin Chem Lab Med 2010:805-13.

(11.) Zegers I, Charout-Got J, Rzychon M, Trapmann S, Emons H, Schimmel H, et al. Certification of C-reactive protein in reference material ERMDA472/IFCC. Luxembourg: Office for Official Publications of the European Communities; 2009. Report EUR 23756 EN.

(12.) Team RC. R: a language and environment for statistical computing. Vienna, Austria, 2012. http://www.r-project.org/ (Accessed April 2013).

Ingrid Zegers, [1] * Robert Beetham, [2] Thomas Keller, [3] Joanna Sheldon, [4] David Bullock, [5] Finlay Mackenzie, [5] Stefanie Trapmann, [1] Hendrik Emons, [1] and Heinz Schimmel [1]

[1] Institute for Reference Materials and Measurements (IRMM), Joint Research Centre, European Commission, Geel, Belgium; [2] Department of Clinical Bio chemistry, Frenchay Hospital, Bristol, UK; [3] Acomed Statistik, Leipzig, Germany; [4] Protein Reference Unit, St. Georges Hospital, London, UK; 5 UK NEQAS, Queen Elizabeth Medical Centre, Birmingham, UK.

* Address correspondence to this author at: European Commission, Joint Research Centre, Institute for Reference Materials and Measurements (IRMM),

Retieseweg 111, 2440 Geel, Belgium. Fax +32-14-571-548; e-mail ingrid. zegers@ec.europa.eu.

Received December 19, 2012; accepted April 22, 2013.

Previously published online at DOI: 10.1373/clinchem.2012.201954

[6] Nonstandard abbreviations: NEQAS, National External Quality Assessment Service; IRMM, Institute for Reference Materials and Measurements; EQAS, External Quality Assessment Service; ALTM, all-laboratory trimmed mean.

Table 1. Methods (combinations of platform and reagents) used in the
commutability study.

Platform                                    Reagents

Abbott Architect c16000    Abbott Ceruloplasmin REF 6K9101 Lot 70185
BN ProSpec                 Siemens reagents (N antiserum to human
                             ceruloplasmin)
BN ProSpec                 Dade Behring anti-ceruloplasmin reagent
                             lot: 155289, Dade Behring standard SL
                             lot:083678
Beckman Immage             Beckman calibrant 2 lot:704575, Beckman
                             anticeruloplasmin reagent lot:706567
Hitachi 917                Tina-quant C4-2, Id. 11 875 051, lot 696
                             008: C4, ACN 032
Roche Integra              Ceruloplasmin, Id. 20 764 663, lot 696825:
                             CERU3, Test-Id. 0-666
Hitachi 917                DAKO (antibody: Q0121 lot 00012259
                             ceruloplasmin)
Olympus systems analyzer   Olympus OSR6164 # 5576
  AU2700 OA28

Platform                                  Laboratory

Abbott Architect c16000    Sentinel Diagnostics, Milan, IT
BN ProSpec                 Siemens Healthcare Diagnostic Products
                             GmbH, Marburg, DE
BN ProSpec                 St. Georges Hospital, London, UK

Beckman Immage             St. Georges Hospital, London, UK

Hitachi 917                Roche Professional Diagnostics, Penzberg,
                             DE
Roche Integra              Roche Professional Diagnostics, Penzberg,
                             DE
Hitachi 917                DAKO, Glostrup, DK

Olympus systems analyzer   Olympus Clare (now Beckman Coulter
  AU2700 OA28                Biomedical), Clare, IE

Platform                       Code

Abbott Architect c16000    Abbott
BN ProSpec                 ProSpec I

BN ProSpec                 ProSpec II

Beckman Immage             Immage

Hitachi 917                Roche Hitachi

Roche Integra              Roche Integra

Hitachi 917                DAKO Hitachi

Olympus systems analyzer   AU2700
  AU2700 OA28

Table 2. Comparison of results for serum samples and commutability of
reference materials. (a)

                                    Commutability (b)

                                    Abbott   ProSpec   ProSpec   Immage
                                                I        II

Slope of the        Abbott                   1         1         1
linear regression
                    ProSpec I       1.00               1         0
                    ProSpec II      0.93     0.93                0
                    Immage          1.33     1.34      1.42
                    Roche Hitachi   0.99     0.99      1.06      0.73
                    Roche Integra   0.94     0.95      1.02      0.70
                    DAKO Hitachi    0.75     0.75      0.80      0.56
                    AU2700          0.92     0.92      0.98      0.68

                                    Commutability (b)

                                     Roche     Roche     Dako
                                    Hitachi   Integra   Hitachi

Slope of the        Abbott          0         0         0
linear regression
                    ProSpec I       0         1         0
                    ProSpec II      0         1         0
                    Immage          0         0         0
                    Roche Hitachi             1         1
                    Roche Integra   0.96                1
                    DAKO Hitachi    0.77      0.80
                    AU2700          0.93      0.96      1.21

                                    AU2700

Slope of the        Abbott          0
linear regression
                    ProSpec I       0
                    ProSpec II      0
                    Immage          0
                    Roche Hitachi   1
                    Roche Integra   1
                    DAKO Hitachi    1
                    AU2700

(a) The upper-right part of the table contains results from the
evaluation of the commutability of ERM-DA470. The lower-left half of
the table contains the slope of Passing--Bablok regression.

(b) 1 indicates commutability, 0 indicates noncommutability.
COPYRIGHT 2013 American Association for Clinical Chemistry, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2013 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Zegers, Ingrid; Beetham, Robert; Keller, Thomas; Sheldon, Joanna; Bullock, David; Mackenzie, Finlay;
Publication:Clinical Chemistry
Article Type:Report
Date:Sep 1, 2013
Words:4025
Previous Article:Gestational diabetes mellitus.
Next Article:Processed B-type natriuretic peptide is a biomarker of postinterventional restenosis in ischemic heart disease.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters