Biologicok (1990) 18, 345-350

COLLABORATIVE STUDY Report of a Collaborative Study for Assessing the

Potency of Hepatitis B Vaccines

Morag Ferguson,* Alan Heath* and Philip D. Minor* tDepartment of Computing and Statistics

*Department of Virology, National institute for Biological Sfandards and Controt, Blanche Lane, South Mimms, Potters Bar, Herts EN6 3QG, U.K.

Abstract. A collaborative study was carried out to examine the suitability of a hepatitis B vaccine derived from plasma as an immunogenicity reference for vaccines produced by recombinant DNA technology in yeast. The use of a plasma derived vaccine as reference appeared satisfactory, although the use of homologous reference improved agreement in potency estimates. The use of a recombinant standard did not however improve agreement for a recombinant vaccine produced by a different manufacturer.

The variation in the dilution of vaccine required to induce antibodies in 50% of test animals and in potency estimates varied widely between laboratories (25fold and 1 O-fold respectively). However this was similar to the variation found in a previous collaborative study.

Introduction

Hepatitis B vaccines prepared by purification of 22 nm hepatitis B surface antigen (HB,Ag) particles from the plasma of chronically infected individuals or by recombinant DNA technology are now in use throughout Europe. An international reference reagent for hepatitis B vaccine for use in mouse immunogenicity tests has been established. ’ This reagent is an aluminium hydroxide adsorbed vaccine containing plasma derived HB,Ag. The collaborative study described in this report was conducted to exam- ine the suitability of a plasma derived hepatitis B vaccine to serve as an immunogenicity reference for material derived by recombinant DNA methods in yeast or mammalian cells. An additional aim was to establish the consistency of assay methods among control laboratories.

Materials and methods

Vaccines

The vaccines examined in the study were the inter- national reference reagent (IRR) for hepatitis B vac- cine immunogenicity which contains HB,Ag derived from plasma, two vaccines containing recombinant HB,Ag produced in yeast by different manufacturing procedures and a vaccine containing antigen pre- pared in mammalian cells. The vaccines were supplied coded A, B, C and D. One of the yeast derived

vaccines was supplied in duplicate labelled with dif- ferent code letters (B and 0. These vaccines were kindly provided by Smith Kline Biologicals, Merck Sharp and Dohme and Pasteur Vaccins.

lmmunogenicity test

The protocol for the test is outlined in the WHO requirements for hepatitis B vaccines.2T3 Briefly, groups of 15 mice, approximately 5 weeks old, were immunized intraperitoneally with 1 ml volumes of serial dilutions of each vaccine in alum diluent (0.5 mg/ml aluminium hydroxide per ml of normal saline). A group of mice are also immunised with diluent as a negative control. Twenty-eight days post inoculation mice are bled and individual sera samples assayed for antibodies to HB,Ag by the method used routinely in individual laboratories.

Anti HBs assays

The method used by each participant and source of kit for the assay of antibodies to HB,Ag is given in Table 1. The cut-off value used locally for the determi- nation of seroconversion is also given.

Participants

Seven participants from different EC countries participated in the study. In the report they are identified by a random code number not necessarily in the order listed in the appendix.

1045-1056/90/040345+06 $03.00/O @ 1990 The International Association of Biological Standardization

346 M. Ferguson et al.

Table 1. Anti HBs assays used by participants

Local cut off for Laboratory Assay Source seroconversion

ELISA Homemade ELISA Locally produced

commercial ELISA Abbott RIA Abbott RIA Abbott

6 RIA Abbott

7 ELISA Abbott

2-l X mean of neg control mice 20 m IU

? ? Upper 95% limit of mean of neg control mice Upper 95% limit of mean of neg control mice (previously 10 mIU) Mean neg control mice -t 0.05

Design of study

Participants were supplied with vials of the IRR, coded vaccines and aluminium hydroxide for prepara- tion of diluent. They were requested to carry out immunogenicity tests on the five vaccines, concur- rently, on two occasions.

Statistical analysis

This test measures the proportion of mice which develop an antibody response after inoculation with each of a series of dilutions of vaccine. In this study, mice which sero-converted were defined as animals with a response greater than 2.1 x the mean response of the negative control mice immunized with diluent.

For each test, responses obtained with individual preparations were analysed by calculating the regression line of probit of the proportion of animals with positive antibody responses on log dilution using the method described by Finney.4 The EDsc (i.e. the dilution inducing antibodies in 50% of the animals) was then estimated from the regression line, fitted independently for each preparation. In some assays, the probit software failed to fit a line for some prepa- rations, where there were insufficient non-extreme responses (i.e. dose levels where all animals or no animals responded). In these cases, EDscs were calcu- lated from a line fitted with common slope, from the parallel line model.

The potencies of vaccines A, B, C and D were each calculated relative to the IRR, using the parallel line model. The statistical analysis includes tests for deviations from this model of non-linearity and non- parallelism. Any assay demonstrating significant (P < 0.01) deviations was noted.

Overall combinations of estimates of potency and

EDso were calculated as unweighted geometric means of all assays, after first testing for statistical homogeneity.

Results and discussion

A total of 13 assays were received from 7 laboratories (Laboratory 7 only supplying one assay). One assay was unusable, with all dose levels of the IRR giving maximum response (Laboratory 31, and was excluded.

The data from Laboratory 2 were obtained in a locally produced ELISA. Although the negative con- trol sera included in the kit gave low optical density (OD) readings (O-05 approximately), negative control mouse sera gave readings of approximately 1. The resultant cut off value was therefore very high but the mice could be scored as positive or negative. Despite this technical aberration, the assays are included in the subsequent analysis.

Relative potencies

The distribution of potencies (logic relative potency) are shown in Fig. 1, for all assays. Estimates of potency for vaccines A, B, and C vary by around a factor of 10. Vaccine D had higher potency, and estimates varied by around a factor of 100. The esti- mated relative potencies for each assay are shown in Table 2, along with 95% confidence intervals. The assays which were non-parallel are indicated. For some assays, it was not possible to obtain a potency estimate for some individual vaccines, where there were insufficient non-extreme responses.

The overall mean potencies are also shown (unweighted geometric means of all assays), along

Hepatttir B vecclne study 347

8 - Vaccine A 8- Vaccine C 7- 7- 6- 6-

5- 5- 4- 6 4-

7

3- 5 6

.B v) 3-

l5 2- 2

16 2 5 6

I- 1114 71

o- I 4151

n -0.6-04 -0-2 0-O 0.2 0.4 O-6 0.8 I.0 I.2 1.4 -06 -04 -0.2 O-0 0.2 04 O-6 Cb8 I.0 I.2 lb m Log potency Log potency b 5 t al 2 E

2 8- Vaccine D 7-

6- 6- 5- 5- 4- 4-

3- 7 - 3-

2- 5

6 6 - 2- 6

I-

3 5

2 4

I 4 I 2 O-4 41

I 11315111516 O-06 -0.4 -0.2 043 0.2 0.4

2 217141 O-0.6 -04 -0-2 0.0 0.2 O-6 O-8 I.0 I.2 I.4 0% o-8 I.0 I.2 I.4

Log potency Log potency

Figure 1. Frequency distributions of the log potency estimate of vaccines A, B, C and D relative to the IBB obtained from immunogenicity tests. Each square represents the estimate from one assay. The numbers in the squares refer to the code numbers of the laboratories.

Table 2. Potencies with 95% confidence intervals for vaccines A, B, C and D relative to the IRR obtained for individual assays and for all assays combined

Vaccine A Vaccine B Vaccine C Vaccine D

Lab Assay Potency Potency code No.

Potency Potency (o/o> 95% CI” (%o) 95% c1* (%o) 95% c1* (%I 95% c1*

1 1 2t

2 1 2t

3 1°F 4 1

2 5 1

2t 6 1

2 7 1 Overall (All assays) Overall (Excluding non-parallel)

82 41-162 43 23- 80 80 46-140

100 69-145 -

197 65

270 87 76

106 142

- 94-412 30-140

122-597 40-193 47-123 55-203 63-319

100 71-141 99 68-144 96 65-141 229 lOl- 520

113 74-172 124 86-178 114 65-199 383 160- 920

100 50-199 43 23- 82

190 107337 175 121-252

49 30- 80 234 110-499

65 30-140 89 42-186 43 19- 95

133 82-215 97 51-186

167 74-377

81 41-162 47 25- 87 -

130 69

192 35

175 62

146 101 173

90-187 42-l 13 92-401 16 75 81376 28-135 go-237 53-194 76392

84 41- 171 31 16 57

767 368-1599 463 320- 670

49 36 80 1631 416-6392

414 157-1089 132 62- 281 66 30- 145

201 123- 328 345 139 851

1165 362-3753

* CI, confidence interval. t Significant deviation from parallelism (P < 0.01).

348 M. Ferguson et al.

-3.0 -2.6 -2.2 -1.8 -1.4

Log ED,,

Log ED,,

-3-o -2.6 -2-2 -1.8 - l-4

Log EDSO

-3.0 -2.6 -2.2 -1.8 -1.4

Log ED,,

-30 -2.6 -2.2 - I.8 -1.4

Log ED,,

Figure 2. Frequency distributions of the EDsos for IRR and vaccines A, B, C and D from immunogenicity tests. Each square represents the estimate from one assay. The num- bers in the squares refer to the code numbers of the labora- tories.

with associated confidence intervals. The overall means, calculated excluding the non-parallel assays, are also shown. There is little difference between the means, with or without the exclusions. The estimated potencies were statistically non-homogeneous, between assays.

The results for the duplicate vaccine, coded B and

C, show variations in estimated potency up to a factor of two within the same assay. However, none of the assays indicated a statistically significant difference in potency. The overall mean potency figure for vac- cines B and C are very close.

ED5os

The estimated EDsos (logio EDso) are shown in Fig. 2. Again, vaccine D exhibited greater varia- bility of estimates (around loo-fold range) than the other preparations (around 25fold range).

Tables 3 and 4 show the overall estimates of rela- tive potency and ED 80 respectively. Confidence intervals and geometric coefficients of variation (GCV) are also given. All assays are included as exclusion of non-parallel assays makes little dif- ference to the estimates of ED&o.

The results above have again demonstrated the improvement in inter-laboratory variation obtained by estimating potency relative to an accepted reference preparation.

One possible reason for the higher variability in estimates for vaccine D is that the responses tended to be at the higher end of the dose-response curve, close to the plateau (i.e. the majority of animals respond). This makes estimation of EDbos, and relative potencies more difficult. If a different pre-dilution factor had been used for vaccine D, to bring the doses away from the plateau of the dose-response curve, variability of estimation may be reduced.

Use of a recombinant vaccine as standard

All assays were re-analysed, excluding the IRR, and taking vaccine B as a standard. The same four assays were non-parallel. Overall potency estimates and GCVs of vaccines A, C and D relative to vaccine B are shown in Table 5. The overall mean potency estimates are similar to those given in Table 3. The GCV for the duplicate vaccine C is reduced whereas that for heterologous recombinant vaccine A is not. All potency estimates were non homogeneous with the exception of vaccine C when measured relative to vaccine B.

Analysis of data by quantitative criteria

The analysis of the assays using quantitative meas- ures (e.g. OD or cpm) rather than scoring individual mice seropositive or seronegative is being considered. However there are several problems associated with this approach including considerations of the nature of the dose response curve and error distributions. This analysis will be presented in a separate report.

Hepatitis B vaccine study 349

Table 3. Geometric means and coefficients of variations (GCV) of estimates of potencies relative to the IRR

Geometric mean 95% confidence interval GCV (%) Number of assays

Vaccine A 100 71-141 67 11 Vaccine B 99 68-144 82 12 Vaccine C 96 65-141 78 11 Vaccine D 229 101-520 263 12

Table 4. Geometric means and coefficients of variations (GCV) of estimates of EDSes

Geometric mean* 95% confidence interval GCV (%) Number of assays

Vaccine A 127 66-238 161 11 Vaccine B 118 77-178 93 12 Vaccine C 120 69-212 130 11 Vaccine D 270 123-588 244 12 IRR E 119 75-196 119 12

* EDso expressed as the reciprocal of the dilution of the original undiluted material.

Table 5. Geometric means and coefficients of variation (GCV) of estimates of potencies relative to Vaccine B

Geometric mean 95% confidence interval GCV (%) Number of assays

Vaccine A 94 64-137 76 11 Vaccine C 102 80-130 44 11 Vaccine D 239 136-419 142 12

Conclusions

The variation in EDso and potency estimates between laboratories obtained in this study was similar to that found in a previous collaborative study carried out to establish the IRR for hepatitis B vaccine.5 Although it appears that a plasma derived vaccine can be used as a reference in mouse immunogenicity tests on re- combinant hepatitis B vaccines the use of a homolo- gous reference improved agreement in potency estimates. However the use of a recombinant stan- dard did not improve agreement for a heterologous recombinant vaccine (vaccine A).

Conclusions from the previous collaborative study concerning the clinical potency of hepatitis B vaccines remain valid i.e. there is not necessarily a constant relationship between the immunogenicity in animal tests of hepatitis B vaccines prepared by different procedures and their immunological potential in man.

Acknowledgements

This study was supported in part by the Com- munity Bureau of Reference.

References

1.

2.

3.

4.

5.

World Health Organisation. International reference reagent of hepatitis B vaccine (plasma-derived). WHO Tech Rep Ser 1987; 760: 29-30. World Health Organisation. Requirements for hepa- tatis B vaccines made by recombinant DNA techniques in yeast. WHO Tech Rep Ser 1987; 760: 107-138. World Health Organ&&ion. Requirements for hepa- tatis B vaccine prepared from plasma. WHO Tech Rep Ser 1988; 771: 181-207. Finney DJ. Statistical methods in biological assay, 3rd edn. London: Griffin and Co. Ltd. 1978. Ferguson M, Seagroatt V and Schild GC. A collabora- tive study to establish the international reference reagent for hepatitis B vaccine containing plasma derived hepatitis B surface antigen. J Biol Stand 1989; 17:X1-160.

Appendix. List of Participants

Dr R Netter Lab National de la Sante Publique 25 Blvd St Jacques Paris XIV 75104 France

350 M. Ferguson et af.

Dr H van der Donk Rijks Instituut Postbus 1 3720 BA Bilthoven Netherlands

Dr R Dobbelaer Instituut voor Hygiene en Epidemiologie Juliette Wytsmanstraat 14 1050 Brussel Belgium

Dr M Ferguson NIBSC Blanche Lane South Mimms Potters Bar Herts. EN6 3QG U.K.

Dr Weeke Luttman Paul Ehrlich Institute Paul Ehrlich strasse 42 D6000 Frankfurt Main 70 Germany

Dr F Salmero Garcia Department Control de Productos Biologicos Instituto de Salud Carlos III

Carretera de Majadahonda a Pozuelo, Km 2 28220 Madrid, Spain

Dr Ssa Maria Rapicetta Instituto Superiore di Sanita Dept Virology Viale Regina Elena 299 00161 Roma Italy

Received for publication 5 March 1990; accepted 18 June 1990.