- Email: [email protected]
Report of a Collaborative Study to Assess the Suitability of a Reference Reagent for Antibodies to Hepatitis E Virus
Biologicals (2002) 30, 43–48 doi:10.1006/biol.2001.0315, available online at http://www.idealibrary.com on
Report of a Collaborative Study to Assess the Suitability of a Reference Reagent for Antibodies to Hepatitis E Virus 1
Morag Ferguson1*, Dawn Walker1, Eric Mast2 and Howard Fields2 National Institute for Biological Standards and Control, Blanche Lane, South Mimms, Potters Bar, Herts, EN6 3QG, U.K.; 2Hepatitis Branch, Centers for Disease Control and Prevention, Atlanta, Georgia, 30333, U.S.A.
Abstract. Several commercial and ‘‘in-house’’ assays have been developed for the detection of antibodies to hepatitis E virus, a major causative agent of enterically transmitted non-A non-B hepatitis. As these kits contain a variety of synthetic peptides or recombinant proteins, greater standardisation is required. A collaborative study was therefore carried out to assess the suitability of a freeze dried preparation designated 95/584 to serve as a reference reagent for hepatitis E virus serum IgG. Preparation 95/584, which is a serum from a previously infected individual, was assayed along with four coded samples, one of which D, was a coded duplicate of 95/584, and three individual sera, coded A, B and C. These preparations were sent to seven laboratories in five countries who tested them in eight different enzyme immunoassays. In most laboratories the coded duplicate gave a mean potency of within 20% of the candidate reference reagent despite the wide range of assays used. However, the potencies of the coded samples which were from different individuals gave somewhat variable potencies relative to the candidate reference reagent. This is not surprising as each sample will have varying proportions of antibodies against individual viral proteins and result in the variation in results observed. Nevertheless, this material will be of use in the standardisation of diagnostic tests for use in sero-prevalence studies and for assessing immunity. Preparation 95/584 was found to be suitable to serve as a reference reagent for hepatitis E serum IgG and has been established as an interim Reference Reagent for Human anti-hepatitis E serum. Each ampoule contains 50 Units per ampoule.
Introduction
detection ranged from 1 in 5 to 1 in 160. Greater standardisation is therefore required so that the sensitivity of di#erent assays and the results obtained in di#erent laboratories can be compared. There will also be a need of accurate measurement of HEV antibody levels to assess immunity and the responses of candidate vaccines. The Centers for Disease Control and Prevention, Atlanta, in collaboration with NIBSC, produced a freeze-dried preparation of serum containing antibodies to HEV as a candidate reference reagent. This report describes the results of a collaborative study to assess the suitability of this material to serve as a reference reagent.
Hepatitis E virus (HEV) is a major causative agent of enterically transmitted non-A non-B hepatitis. Outbreaks have been reported in many developing countries and significant mortality reported in pregnant women. Although not able to be cultured in the laboratory, the genetic organisation of HEV has been determined by molecular cloning and sequencing experiments and the antigenic structure using synthetic peptides and recombinant proteins. Diagnostic tests for the detection of antibody to the virus have been established in several laboratories and commercial kits have also been developed. However, in a study to evaluate 12 assays, although six assays detected anti-HEV in >90% of sera, the overall sensitivity of assays with known positive sera ranged from 17 to 100%. In addition the limit of
Candidate reference reagent
*To whom correspondence should be addressed. E-mail: [email protected]
This is a batch of ampoules coded 95/584 containing freeze-dried preparation of a pool of equal
1045–1056/02/030043+06 $35.00/0
Materials
44
M. Ferguson et al.
Table 1. Assay methods used by participants Laboratory code
Source of kit
Proteins/peptide
1 2 3 4
In-house In-house Commercial Commercial
5 6A 6B 7
In-house In-house In-house In-house
ORF2 C-terminal 267 aa expressed in E. coli ORF2 55 kDa expressed in baculovirus Two recombinant antigens from structural region Two recombinant antigens from structural region of Burmese strain 327 residue carboxy end ORF2 123 residue full length ORF3 Mosaic protein of three antigenic domains from ORF2 and two from ORF3 Synthetic peptide from ORF3 Synthetic peptide from ORF3 Capture antibody—rabbit anti-ORF2 expressed in E. coli Partially purified ORF2 (entire) expressed in baculovirus
volumes of serum obtained on five separate dates over a 45 day period (four to five months after onset of illness) from a patient in the U.S. who developed acute hepatitis E following travel to India. The mean weight of the contents of six ampoules was 33·57 mg with a standard deviation of 0·05%. The mean residual moisture of three ampoules was 1·80%. Ampoule contents are re-constituted in 0·5 ml H2O. Coded serum samples
In order to ensure that the candidate reference reagent behaves in a similar manner to samples against which it will be assayed, representative sera were included in the study. Sample A was obtained approximately six months after illness onset from a patient who acquired hepatitis E in Mexico. Sample B was obtained approximately one week after illness onset from a patient who acquired hepatitis E in India. Sample C was obtained approximately six weeks after illness onset from a second patient who acquired hepatitis E in India. Sample D was a coded duplicate of the candidate reference reagent. Study design
Participants were requested to perform three independent assays for HEV antibody content of the four study samples, on di#erent days, preferably one week apart, using the method in routine use in their laboratory. Participants were requested to assay concurrently a series of dilutions of the candidate reference reagent and each of the coded samples. Participants were also asked to select dilutions of each preparation which resulted in similar ranges of
response and so that at least four points fell on the linear part of the log dose-response. Assay method and diluent
Participants used the method in routine use in their laboratories. These are summarised in Table 1. Participants
Seven laboratories participated in the study, each of which is referred to in this report by an arbitrarily assigned number, not necessarily representing the order of listing in the Appendix. Where a laboratory performed assays using more than one peptide, each one is treated as if performed by di#erent laboratories. For example, laboratory 6 used two di#erent peptide assays which are referred to as 6A and 6B. Statistical analysis
All assays were analysed as multiple parallel line bioassays2 comparing assay response to log concentration. When plotted against log concentration, linear response lines which are parallel for all preparations included in the assay, are essential for this analysis. In some cases linearity was achieved by omitting extreme concentrations. If necessary, the observed responses were transformed into logs to gain linearity and parallelism. For one laboratory, data were transformed to percentages of the upper and lower limits of curves for each assay and the WRANL program used.3 The statistical validity of parallelism and linearity of the assays was assessed by analysis of variance tests.
Reference reagent for antibodies to hepatitis E virus
Parallelism was further assessed by comparing estimates of the slopes of the response lines across all assays. Ratios of the slopes were calculated for each of the samples A–D against the candidate reference reagent. If the ratios were not significantly di#erent from 1, that is, the slopes were not significantly di#erent, then the assumption of parallelism was valid. The potencies of A, B, C and D relative to the candidate reference reagent were calculated for each assay. For each laboratory, combined potency estimates were obtained by taking geometric means of results from all assays and an overall potency estimate was calculated as a geometric mean of the laboratory means. Variability within laboratories (between assays) and between laboratories was measured by calculating geometric coe#icients of variation.4
45
used (referred to in this report as laboratory 6A) for the candidate reference reagent and samples A, B and D are very low and should be treated with caution. The response lines of the candidate reference reagent by assays 2B and 3B of laboratory 2 were noticeably non-linear and nonparallel. In addition, the results from the third assay by laboratory 1 were clearly invalid. These assays were therefore omitted from subsequent analysis. The assumptions of linearity and parallelism held separately in 93% and 63% of the total assays, respectively. However, from comparison of the slopes across all assays, non-parallelism was not detected for the study as a whole. Therefore, all assays were used in subsequent analysis. Potencies
Results Assay data
The seven participants contributed data from a total of 25 assays. Each assay by laboratory 2 consisted of two microtitre plates. The first plate was exposed to the candidate reference reagent and samples A and B, and the second plate to the candidate reference reagent and samples C and D. Since the candidate reference reagent was used on each plate, the di#erent plates were treated as di#erent assays, resulting in a total of six assays performed by laboratory 2. The responses from some of the highest doses used by laboratories 3 and 5 gave higher results than they could read and therefore were excluded from analysis. Laboratory 4 returned raw data from one assay for samples A, C and D, but not for sample B or the candidate reference reagent. Therefore, sample D, the coded duplicate, was treated as the candidate reference reagent and estimates of potencies of samples A and C were calculated based on one assay only. Assay validity
Laboratory 6 did not assay the samples or the candidate reference reagent in replicate or duplicate in each assay with either peptide. Therefore, the assumptions of linearity and parallelism could not be tested for this laboratory. However, from plots of the data, these assumptions appear to be satisfied. The responses given by one of the peptides
Detailed values of individual laboratory mean potencies of samples A, B, C and D are listed in Table 2 along with 95% confidence limits. For sample A, laboratories 1, 2 and 7 appear to agree, but are all much higher than the estimates obtained by the other laboratories. Of the other laboratories, the estimates of laboratories 3 and 5 are similar. For samples B and C, the estimates of potency vary among the laboratories. Laboratory 7 obtained a high estimate for sample B and laboratory 6A for sample C. The cause of these high estimates is not clear. There is better agreement among the laboratories’ estimates of potency for sample D, the coded duplicate of the reference reagent, which is expected when comparing similar materials, with the exception, perhaps, of the slightly higher estimate obtained by laboratory 6A. The data for sample D is also shown in histogram form in Figure 1. Due to the varying estimates of potencies of samples A, B and C, overall mean potencies for each of these three samples were not calculated. However, the overall mean potency for sample D relative to the candidate reference reagent is shown in Table 2. Because sample D is a coded duplicate of the candidate reference reagent, its relative potency should be close to 1·00. As can be seen, the overall mean relative potency of D is calculated as 1·08, a little above 1·00. Looking at Table 2 and Figure 1, however, this could be due to laboratory 6A which, as stated above, gave very low responses. Omitting the estimate of potency obtained by this laboratory, the overall mean potency is reduced to 0·99, which is very close to 1·00.
46
M. Ferguson et al.
Table 2. Potencies of samples A–D relative to the candidate reference reagent Laboratory number 1 2 3 4 5 6A 6B 7
Potency A
95% limits
B
95% limits
C
95% limits
D
95% limits
2·42 2·20 0·13 0·77* 0·14 0·34 0·07 2·05
1·47–3·96 1·93–2·52 0·05–0·34 nc 0·07–0·30 0·15–0·78 0·06–0·08 1·40–2·98
1·46 1·37 0·33 nd 1·52 1·60 0·52 4·65
1·39–1·54 0·65–2·89 0·24–0·46 — 0·51–4·53 1·10–2·33 0·35–0·77 3·88–5·57
1·06 1·23 1·18 1·11* 2·64 45·14 0·28 1·70
0·55–2·02 nc 0·56–2·49 nc 0·76–9·13 23·25–87·66 0·12–0·66 1·02–2·84 Mean Omitting 6A
0·99 1·04 0·86 nd nd 1·67 1·25 0·85 1·08 0·99
0·14–6·77 nc 0·41–1·83 — — 1·05–2·65 0·59–2·63 0·27–2·76
* relative to sample D, the coded duplicate. nd, not done; nc, not able to be calculated since data from only one assay.
Table 3. Intra-laboratory variability (% gcv)
6
Laboratory number
Number of labs
5 4 3 2
07
02
1
03
01
0
06B
06A
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 Potency
Figure 1. Potency of sample D, the coded duplicate of the candidate standard 95/584.
Intra-laboratory variability
The variability within each laboratory, expressed as geometric coe#icients of variation (% gcv) for individual laboratories are given in Table 3. These are higher for sample A by laboratories 3, 5 and 6A, sample B by laboratories 2 and 5, sample C by laboratories 3, 5, 6A and 6B and sample D by laboratories 3, 6B and 7. Overall, the variability within the laboratories is highest for samples C and D. Given that sample D is a coded duplicate of the candidate reference reagent, this is perhaps surprising.
1 2 3 4 5 6A 6B 7 Mean
Sample A
Sample B
Sample C
Sample D
5·7 5·5 46·8 nc 35·1 39·1 6·2 16·4 22·1
0·6 35·0 14·5 nc 55·3 16·3 17·3 7·6 20·9
7·5 nc 34·9 nc 64·9 30·6 41·5 22·9 33·7
23·9 nc 35·4 nc nc 20·3 34·9 60·2 35·0
nc, not able to be calculated since data from only one assay. gcv is defined as (e 1)100% where is the standard deviation of the log potencies.3
Table 4. Inter-laboratory variability (% gcv) Sample A B C D
All laboratories
Omitting Lab 6A
318·9 135·1 329·1 29·1
366·0 152·6 99·4 16·9
Inter-laboratory variability
Variability between laboratories for the potency estimates of samples A, B, C and D relative to the candidate reference reagent is summarised in Table 4. Variability was lowest for sample D, when either
including or excluding laboratory 6A, which is expected when comparing similar materials. When laboratory 6A is omitted, variability between laboratories is markedly highest for sample A.
Reference reagent for antibodies to hepatitis E virus
Discussion The mean potency of the coded duplicate was close to that of the candidate reference reagent (1·08, or 0·99 if the most extreme result, the estimate from laboratory 6A was omitted) despite the wide range of assays used. In most laboratories the coded duplicate gave a mean potency of within 20% of the candidate reference reagent. The candidate reference reagent and coded samples A, B and C were sera from di#erent individuals taken at di#erent times after onset of illness. In addition, sample A was from a patient who acquired the disease in Mexico whereas the candidate reference reagent and samples B and C were from patient who acquired the disease in India. Although di#erences in sero-reactivity have been observed with di#erent geographic strains tested on di#erent kits, the importance of geographic diversity has not yet been fully established. Not all laboratories gave full details of their assays. Nevertheless, the assays used by each participant appeared to contain di#erent antigens and it is therefore not surprising that the potencies of the coded samples relative to the candidate reference reagent are somewhat variable as each sample will have varying proportions of antibodies against individual viral proteins. Although it is di#icult to determine the extent to which the use of di#erent proteins a#ected the results of assays on the candidate reference reagent and coded samples, the results indicate that the proteins used are an important aspect of assay development and selection. An additional factor which may have a#ected variability between laboratories was the use of di#erent diluents. Two laboratories used commercial assays. As the data set from laboratory 4 was incomplete, there is insu#icient data to comment on di#erences between in-house and commercial assays. Sample C gave the most consistent results other than those obtained by Laboratory 6A. In general, results from laboratories 1 and 2 gave similar results for all coded samples as did results from laboratories 3 and 6B for samples A and B but not for C. Laboratories 3, 5 and 6B gave very low potencies relative to the candidate reference reagent for sample A. Most assays are reasonably reproducible within a laboratory for samples A and B but were more variable for sample C and the sample D, the coded duplicate. These result in high intra-laboratory variation. However, this is similar to that observed in the assay of individual serum
47
sample in a collaborative study on Parvovirus B19 where di#erent assays were used.5 Despite the variability in potencies of the coded serum samples when assayed in di#erent tests, the availability of a common reference reagent for HEV serum will be of use in the standardisation of diagnostic tests in sero-prevalence studies and for assessing immunity in the community and in vaccine immunogenicity and e#icacy studies. However caution will have to be exercised in the interpretation of levels of antibody detected in di#erent assays until there is greater experience in their use.
Conclusions
The preparation coded 95/584 has been established as an interim Reference Reagent for Human anti-hepatitis E serum6 by the Expert Committee on Biological Standardisation of the World Health Organisation. Each ampoule contains 50 Units per ampoule.
Acknowledgements The authors are grateful to all of the participants in the collaborative study. 2002 Crown Copyright
References 1. Mast EE, Alter MJ, Holland PV, Purcell RH, for the Hepatitis E Virus Antibody Serum Panel Evaluation Group. Evaluation of assays for antibody to hepatitis E virus by a serum panel. Hepatology 1998; 27: 857–861. 2. Finney DJ. Statistical Methods in Biological Assay. 3rd edit. London, Charles Gri#in, 1978. 3. Gaines Das R, Tydeman MS. Iterative weighted regression analysis of logit responses. A computer program for the analysis of bioassays and immunoassays. Computer Programs in Biomedicine 1980; 15: 13–22. 4. Kirkwood TBL. Geometric means and measures of dispersion. Biometrics 1979; 35: 908–909. 5. Ferguson M, Walker D, Cohen B. Report of a collaborative study to establish the International Standard for Parvovirus B19 serum IgG. Biologicals 1997; 25, 283– 288. 6. World Health Organisation. Anti-hepatitis E serum, human. 1999 WHO Technical Report Series 889, p17.
Received 20 September 2001; accepted 16 October 2001
48
M. Ferguson et al.
Appendix Participants
Dr Robert Purcell and Dr S Tsarev Hepatitis Viruses Section/LID/NIAID National Institutes of Health, Building 7, Room 202 7 Center Drive MSC 0740 Bethesda Maryland 20892-0740 U.S.A. Dr Michael S Balayan Institute of Poliomyelitis and Viral Encephalitis 142782 Moscow Russia Dr Heng Feng Seow and Dr D Anderson Hepatitis Research Unit Macfarlane Burnet Centre for Medical Research Yarra Bend Road, PO Box 254 Fairfield Victoria Australia 3078 Dr N Takedo Department of Epidemiology National Institutes of Health Toyama 1-23-1
Shinjuku Tokyo 172 Japan Dr Michael Favorov and Dr Howard Fields Hepatitis Branch Laboratory Bg 1, Rm 1389 Centers for Disease Control and Prevention 1600 Clifton Road Atlanta Georgia 30333 U.S.A. Dr Pierre Coursaget Laboratoire d’Immunologie des Maladies Infectieuses Faculte de Pharmacie 31 avenue Monge 37200 Tours France Dr Jacques Pillot Service de Bacteriologie-Virologie Hopital Antoine Beclere 157 avenue de la Porte de Trivaux 92141 Clamart Cedex France
Report of a Collaborative Study to Assess the Suitability of a Reference Reagent for Antibodies to Hepatitis E Virus 1
Morag Ferguson1*, Dawn Walker1, Eric Mast2 and Howard Fields2 National Institute for Biological Standards and Control, Blanche Lane, South Mimms, Potters Bar, Herts, EN6 3QG, U.K.; 2Hepatitis Branch, Centers for Disease Control and Prevention, Atlanta, Georgia, 30333, U.S.A.
Abstract. Several commercial and ‘‘in-house’’ assays have been developed for the detection of antibodies to hepatitis E virus, a major causative agent of enterically transmitted non-A non-B hepatitis. As these kits contain a variety of synthetic peptides or recombinant proteins, greater standardisation is required. A collaborative study was therefore carried out to assess the suitability of a freeze dried preparation designated 95/584 to serve as a reference reagent for hepatitis E virus serum IgG. Preparation 95/584, which is a serum from a previously infected individual, was assayed along with four coded samples, one of which D, was a coded duplicate of 95/584, and three individual sera, coded A, B and C. These preparations were sent to seven laboratories in five countries who tested them in eight different enzyme immunoassays. In most laboratories the coded duplicate gave a mean potency of within 20% of the candidate reference reagent despite the wide range of assays used. However, the potencies of the coded samples which were from different individuals gave somewhat variable potencies relative to the candidate reference reagent. This is not surprising as each sample will have varying proportions of antibodies against individual viral proteins and result in the variation in results observed. Nevertheless, this material will be of use in the standardisation of diagnostic tests for use in sero-prevalence studies and for assessing immunity. Preparation 95/584 was found to be suitable to serve as a reference reagent for hepatitis E serum IgG and has been established as an interim Reference Reagent for Human anti-hepatitis E serum. Each ampoule contains 50 Units per ampoule.
Introduction
detection ranged from 1 in 5 to 1 in 160. Greater standardisation is therefore required so that the sensitivity of di#erent assays and the results obtained in di#erent laboratories can be compared. There will also be a need of accurate measurement of HEV antibody levels to assess immunity and the responses of candidate vaccines. The Centers for Disease Control and Prevention, Atlanta, in collaboration with NIBSC, produced a freeze-dried preparation of serum containing antibodies to HEV as a candidate reference reagent. This report describes the results of a collaborative study to assess the suitability of this material to serve as a reference reagent.
Hepatitis E virus (HEV) is a major causative agent of enterically transmitted non-A non-B hepatitis. Outbreaks have been reported in many developing countries and significant mortality reported in pregnant women. Although not able to be cultured in the laboratory, the genetic organisation of HEV has been determined by molecular cloning and sequencing experiments and the antigenic structure using synthetic peptides and recombinant proteins. Diagnostic tests for the detection of antibody to the virus have been established in several laboratories and commercial kits have also been developed. However, in a study to evaluate 12 assays, although six assays detected anti-HEV in >90% of sera, the overall sensitivity of assays with known positive sera ranged from 17 to 100%. In addition the limit of
Candidate reference reagent
*To whom correspondence should be addressed. E-mail: [email protected]
This is a batch of ampoules coded 95/584 containing freeze-dried preparation of a pool of equal
1045–1056/02/030043+06 $35.00/0
Materials
44
M. Ferguson et al.
Table 1. Assay methods used by participants Laboratory code
Source of kit
Proteins/peptide
1 2 3 4
In-house In-house Commercial Commercial
5 6A 6B 7
In-house In-house In-house In-house
ORF2 C-terminal 267 aa expressed in E. coli ORF2 55 kDa expressed in baculovirus Two recombinant antigens from structural region Two recombinant antigens from structural region of Burmese strain 327 residue carboxy end ORF2 123 residue full length ORF3 Mosaic protein of three antigenic domains from ORF2 and two from ORF3 Synthetic peptide from ORF3 Synthetic peptide from ORF3 Capture antibody—rabbit anti-ORF2 expressed in E. coli Partially purified ORF2 (entire) expressed in baculovirus
volumes of serum obtained on five separate dates over a 45 day period (four to five months after onset of illness) from a patient in the U.S. who developed acute hepatitis E following travel to India. The mean weight of the contents of six ampoules was 33·57 mg with a standard deviation of 0·05%. The mean residual moisture of three ampoules was 1·80%. Ampoule contents are re-constituted in 0·5 ml H2O. Coded serum samples
In order to ensure that the candidate reference reagent behaves in a similar manner to samples against which it will be assayed, representative sera were included in the study. Sample A was obtained approximately six months after illness onset from a patient who acquired hepatitis E in Mexico. Sample B was obtained approximately one week after illness onset from a patient who acquired hepatitis E in India. Sample C was obtained approximately six weeks after illness onset from a second patient who acquired hepatitis E in India. Sample D was a coded duplicate of the candidate reference reagent. Study design
Participants were requested to perform three independent assays for HEV antibody content of the four study samples, on di#erent days, preferably one week apart, using the method in routine use in their laboratory. Participants were requested to assay concurrently a series of dilutions of the candidate reference reagent and each of the coded samples. Participants were also asked to select dilutions of each preparation which resulted in similar ranges of
response and so that at least four points fell on the linear part of the log dose-response. Assay method and diluent
Participants used the method in routine use in their laboratories. These are summarised in Table 1. Participants
Seven laboratories participated in the study, each of which is referred to in this report by an arbitrarily assigned number, not necessarily representing the order of listing in the Appendix. Where a laboratory performed assays using more than one peptide, each one is treated as if performed by di#erent laboratories. For example, laboratory 6 used two di#erent peptide assays which are referred to as 6A and 6B. Statistical analysis
All assays were analysed as multiple parallel line bioassays2 comparing assay response to log concentration. When plotted against log concentration, linear response lines which are parallel for all preparations included in the assay, are essential for this analysis. In some cases linearity was achieved by omitting extreme concentrations. If necessary, the observed responses were transformed into logs to gain linearity and parallelism. For one laboratory, data were transformed to percentages of the upper and lower limits of curves for each assay and the WRANL program used.3 The statistical validity of parallelism and linearity of the assays was assessed by analysis of variance tests.
Reference reagent for antibodies to hepatitis E virus
Parallelism was further assessed by comparing estimates of the slopes of the response lines across all assays. Ratios of the slopes were calculated for each of the samples A–D against the candidate reference reagent. If the ratios were not significantly di#erent from 1, that is, the slopes were not significantly di#erent, then the assumption of parallelism was valid. The potencies of A, B, C and D relative to the candidate reference reagent were calculated for each assay. For each laboratory, combined potency estimates were obtained by taking geometric means of results from all assays and an overall potency estimate was calculated as a geometric mean of the laboratory means. Variability within laboratories (between assays) and between laboratories was measured by calculating geometric coe#icients of variation.4
45
used (referred to in this report as laboratory 6A) for the candidate reference reagent and samples A, B and D are very low and should be treated with caution. The response lines of the candidate reference reagent by assays 2B and 3B of laboratory 2 were noticeably non-linear and nonparallel. In addition, the results from the third assay by laboratory 1 were clearly invalid. These assays were therefore omitted from subsequent analysis. The assumptions of linearity and parallelism held separately in 93% and 63% of the total assays, respectively. However, from comparison of the slopes across all assays, non-parallelism was not detected for the study as a whole. Therefore, all assays were used in subsequent analysis. Potencies
Results Assay data
The seven participants contributed data from a total of 25 assays. Each assay by laboratory 2 consisted of two microtitre plates. The first plate was exposed to the candidate reference reagent and samples A and B, and the second plate to the candidate reference reagent and samples C and D. Since the candidate reference reagent was used on each plate, the di#erent plates were treated as di#erent assays, resulting in a total of six assays performed by laboratory 2. The responses from some of the highest doses used by laboratories 3 and 5 gave higher results than they could read and therefore were excluded from analysis. Laboratory 4 returned raw data from one assay for samples A, C and D, but not for sample B or the candidate reference reagent. Therefore, sample D, the coded duplicate, was treated as the candidate reference reagent and estimates of potencies of samples A and C were calculated based on one assay only. Assay validity
Laboratory 6 did not assay the samples or the candidate reference reagent in replicate or duplicate in each assay with either peptide. Therefore, the assumptions of linearity and parallelism could not be tested for this laboratory. However, from plots of the data, these assumptions appear to be satisfied. The responses given by one of the peptides
Detailed values of individual laboratory mean potencies of samples A, B, C and D are listed in Table 2 along with 95% confidence limits. For sample A, laboratories 1, 2 and 7 appear to agree, but are all much higher than the estimates obtained by the other laboratories. Of the other laboratories, the estimates of laboratories 3 and 5 are similar. For samples B and C, the estimates of potency vary among the laboratories. Laboratory 7 obtained a high estimate for sample B and laboratory 6A for sample C. The cause of these high estimates is not clear. There is better agreement among the laboratories’ estimates of potency for sample D, the coded duplicate of the reference reagent, which is expected when comparing similar materials, with the exception, perhaps, of the slightly higher estimate obtained by laboratory 6A. The data for sample D is also shown in histogram form in Figure 1. Due to the varying estimates of potencies of samples A, B and C, overall mean potencies for each of these three samples were not calculated. However, the overall mean potency for sample D relative to the candidate reference reagent is shown in Table 2. Because sample D is a coded duplicate of the candidate reference reagent, its relative potency should be close to 1·00. As can be seen, the overall mean relative potency of D is calculated as 1·08, a little above 1·00. Looking at Table 2 and Figure 1, however, this could be due to laboratory 6A which, as stated above, gave very low responses. Omitting the estimate of potency obtained by this laboratory, the overall mean potency is reduced to 0·99, which is very close to 1·00.
46
M. Ferguson et al.
Table 2. Potencies of samples A–D relative to the candidate reference reagent Laboratory number 1 2 3 4 5 6A 6B 7
Potency A
95% limits
B
95% limits
C
95% limits
D
95% limits
2·42 2·20 0·13 0·77* 0·14 0·34 0·07 2·05
1·47–3·96 1·93–2·52 0·05–0·34 nc 0·07–0·30 0·15–0·78 0·06–0·08 1·40–2·98
1·46 1·37 0·33 nd 1·52 1·60 0·52 4·65
1·39–1·54 0·65–2·89 0·24–0·46 — 0·51–4·53 1·10–2·33 0·35–0·77 3·88–5·57
1·06 1·23 1·18 1·11* 2·64 45·14 0·28 1·70
0·55–2·02 nc 0·56–2·49 nc 0·76–9·13 23·25–87·66 0·12–0·66 1·02–2·84 Mean Omitting 6A
0·99 1·04 0·86 nd nd 1·67 1·25 0·85 1·08 0·99
0·14–6·77 nc 0·41–1·83 — — 1·05–2·65 0·59–2·63 0·27–2·76
* relative to sample D, the coded duplicate. nd, not done; nc, not able to be calculated since data from only one assay.
Table 3. Intra-laboratory variability (% gcv)
6
Laboratory number
Number of labs
5 4 3 2
07
02
1
03
01
0
06B
06A
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 Potency
Figure 1. Potency of sample D, the coded duplicate of the candidate standard 95/584.
Intra-laboratory variability
The variability within each laboratory, expressed as geometric coe#icients of variation (% gcv) for individual laboratories are given in Table 3. These are higher for sample A by laboratories 3, 5 and 6A, sample B by laboratories 2 and 5, sample C by laboratories 3, 5, 6A and 6B and sample D by laboratories 3, 6B and 7. Overall, the variability within the laboratories is highest for samples C and D. Given that sample D is a coded duplicate of the candidate reference reagent, this is perhaps surprising.
1 2 3 4 5 6A 6B 7 Mean
Sample A
Sample B
Sample C
Sample D
5·7 5·5 46·8 nc 35·1 39·1 6·2 16·4 22·1
0·6 35·0 14·5 nc 55·3 16·3 17·3 7·6 20·9
7·5 nc 34·9 nc 64·9 30·6 41·5 22·9 33·7
23·9 nc 35·4 nc nc 20·3 34·9 60·2 35·0
nc, not able to be calculated since data from only one assay. gcv is defined as (e 1)100% where is the standard deviation of the log potencies.3
Table 4. Inter-laboratory variability (% gcv) Sample A B C D
All laboratories
Omitting Lab 6A
318·9 135·1 329·1 29·1
366·0 152·6 99·4 16·9
Inter-laboratory variability
Variability between laboratories for the potency estimates of samples A, B, C and D relative to the candidate reference reagent is summarised in Table 4. Variability was lowest for sample D, when either
including or excluding laboratory 6A, which is expected when comparing similar materials. When laboratory 6A is omitted, variability between laboratories is markedly highest for sample A.
Reference reagent for antibodies to hepatitis E virus
Discussion The mean potency of the coded duplicate was close to that of the candidate reference reagent (1·08, or 0·99 if the most extreme result, the estimate from laboratory 6A was omitted) despite the wide range of assays used. In most laboratories the coded duplicate gave a mean potency of within 20% of the candidate reference reagent. The candidate reference reagent and coded samples A, B and C were sera from di#erent individuals taken at di#erent times after onset of illness. In addition, sample A was from a patient who acquired the disease in Mexico whereas the candidate reference reagent and samples B and C were from patient who acquired the disease in India. Although di#erences in sero-reactivity have been observed with di#erent geographic strains tested on di#erent kits, the importance of geographic diversity has not yet been fully established. Not all laboratories gave full details of their assays. Nevertheless, the assays used by each participant appeared to contain di#erent antigens and it is therefore not surprising that the potencies of the coded samples relative to the candidate reference reagent are somewhat variable as each sample will have varying proportions of antibodies against individual viral proteins. Although it is di#icult to determine the extent to which the use of di#erent proteins a#ected the results of assays on the candidate reference reagent and coded samples, the results indicate that the proteins used are an important aspect of assay development and selection. An additional factor which may have a#ected variability between laboratories was the use of di#erent diluents. Two laboratories used commercial assays. As the data set from laboratory 4 was incomplete, there is insu#icient data to comment on di#erences between in-house and commercial assays. Sample C gave the most consistent results other than those obtained by Laboratory 6A. In general, results from laboratories 1 and 2 gave similar results for all coded samples as did results from laboratories 3 and 6B for samples A and B but not for C. Laboratories 3, 5 and 6B gave very low potencies relative to the candidate reference reagent for sample A. Most assays are reasonably reproducible within a laboratory for samples A and B but were more variable for sample C and the sample D, the coded duplicate. These result in high intra-laboratory variation. However, this is similar to that observed in the assay of individual serum
47
sample in a collaborative study on Parvovirus B19 where di#erent assays were used.5 Despite the variability in potencies of the coded serum samples when assayed in di#erent tests, the availability of a common reference reagent for HEV serum will be of use in the standardisation of diagnostic tests in sero-prevalence studies and for assessing immunity in the community and in vaccine immunogenicity and e#icacy studies. However caution will have to be exercised in the interpretation of levels of antibody detected in di#erent assays until there is greater experience in their use.
Conclusions
The preparation coded 95/584 has been established as an interim Reference Reagent for Human anti-hepatitis E serum6 by the Expert Committee on Biological Standardisation of the World Health Organisation. Each ampoule contains 50 Units per ampoule.
Acknowledgements The authors are grateful to all of the participants in the collaborative study. 2002 Crown Copyright
References 1. Mast EE, Alter MJ, Holland PV, Purcell RH, for the Hepatitis E Virus Antibody Serum Panel Evaluation Group. Evaluation of assays for antibody to hepatitis E virus by a serum panel. Hepatology 1998; 27: 857–861. 2. Finney DJ. Statistical Methods in Biological Assay. 3rd edit. London, Charles Gri#in, 1978. 3. Gaines Das R, Tydeman MS. Iterative weighted regression analysis of logit responses. A computer program for the analysis of bioassays and immunoassays. Computer Programs in Biomedicine 1980; 15: 13–22. 4. Kirkwood TBL. Geometric means and measures of dispersion. Biometrics 1979; 35: 908–909. 5. Ferguson M, Walker D, Cohen B. Report of a collaborative study to establish the International Standard for Parvovirus B19 serum IgG. Biologicals 1997; 25, 283– 288. 6. World Health Organisation. Anti-hepatitis E serum, human. 1999 WHO Technical Report Series 889, p17.
Received 20 September 2001; accepted 16 October 2001
48
M. Ferguson et al.
Appendix Participants
Dr Robert Purcell and Dr S Tsarev Hepatitis Viruses Section/LID/NIAID National Institutes of Health, Building 7, Room 202 7 Center Drive MSC 0740 Bethesda Maryland 20892-0740 U.S.A. Dr Michael S Balayan Institute of Poliomyelitis and Viral Encephalitis 142782 Moscow Russia Dr Heng Feng Seow and Dr D Anderson Hepatitis Research Unit Macfarlane Burnet Centre for Medical Research Yarra Bend Road, PO Box 254 Fairfield Victoria Australia 3078 Dr N Takedo Department of Epidemiology National Institutes of Health Toyama 1-23-1
Shinjuku Tokyo 172 Japan Dr Michael Favorov and Dr Howard Fields Hepatitis Branch Laboratory Bg 1, Rm 1389 Centers for Disease Control and Prevention 1600 Clifton Road Atlanta Georgia 30333 U.S.A. Dr Pierre Coursaget Laboratoire d’Immunologie des Maladies Infectieuses Faculte de Pharmacie 31 avenue Monge 37200 Tours France Dr Jacques Pillot Service de Bacteriologie-Virologie Hopital Antoine Beclere 157 avenue de la Porte de Trivaux 92141 Clamart Cedex France