Biologicals 42 (2014) 101e108

Contents lists available at ScienceDirect

Biologicals journal homepage: www.elsevier.com/locate/biologicals

An international collaborative study of the effect of active pertussis toxin on the modified Kendrick test for acellular pertussis vaccines Dorothy Xing*, Rose Gaines Das, Alex Douglas-Bardsley, Catpagavalli Asokanathan, Michael Corbel National Institute for Biological Standards and Control, Potters Bar, Hertfordshire EN6 3QG, UK

a r t i c l e i n f o

a b s t r a c t

Article history: Received 13 September 2013 Received in revised form 5 November 2013 Accepted 19 November 2013

Speculation that the Japanese modified intra-cerebral challenge assay, which is used in several countries for control of acellular pertussis vaccines, depends on the presence of small amounts of active pertussis toxin led to an assumption that it may not be appropriate for highly toxoided or genetically detoxified vaccines. Consequently, at the recommendation of a World Health Organisation AD Hoc Working Group on mouse protection models for testing and control of acellular pertussis vaccine, the effect of pertussis toxin on the modified intra-cerebral challenge assay (modified Kendrick, MICA) was evaluated in an international collaborative study. Results of this study showed that for genetically detoxified vaccines both with and without active pertussis toxin the MICA clearly distinguished mice vaccinated with acellular vaccines from unvaccinated mice and gave a significant doseeresponse relationship. However, vaccine samples containing active pertussis toxin (5 or 50 ng/single human dose) appeared to be more potent than the equivalent sample without active pertussis toxin. Similar results were also given by two respiratory infection models (intranasal and aerosol) included in the study. The results also indicated that the effect of pertussis toxin may vary depending on mouse strain. Ó 2013 The International Alliance for Biological Standardization. Published by Elsevier Ltd. All rights reserved.

Keywords: Acellular pertussis vaccine Pertussis toxin Potency

1. Introduction Acellular pertussis vaccines (ACVs) have been shown to be both safe and efficacious in clinical trials in several European countries and Senegal [1e6]. These products are now licensed for routine use in many countries. Although licensed products are expected to be consistent with clinical trial lots, there is no internationally agreed potency assay in general use to assure this. Problems in standardising and controlling ACVs include varied antigenic composition, different detoxification methods, absence of unequivocal correlates of protection and lack of an independently validated and generally accepted animal model [7e9]. Currently, the modified intra-cerebral challenge assay (MICA, modified Kendrick test) is used in Japan, Korea and China and possibly other Asian countries as the potency assay for routine release of acellular pertussis and combination vaccines [10]. For release, vaccines must have potency
4 unit (unit as defined by

* Corresponding author. Division of Bacteriology, National Institute for Biological Standards and Control, Potters Bar, Hertfordshire EN6 3QG, UK. Tel.: þ44 (0) 1707 641 433; fax: þ44 (0) 1707 641 054. E-mail addresses: [email protected], [email protected] (D. Xing).

national reference standard)/dose with lower 95% confidence limit
2 unit/dose [11,12]. The vaccines which meet release criteria in these countries are known to be clinically effective [13e19]. However, it has been claimed that the MICA may not be appropriate for highly toxoided or genetically detoxified vaccines because it depends on the presence of small amounts of active pertussis toxin (PT) [8,20,21]. At a meeting of representatives of vaccine manufacturers, regulatory authorities and members of the World Health Organisation (WHO) AD Hoc Working Group on mouse protection models for testing and control of ACVs, it was recommended that a collaborative investigation on the effect of PT on MICA should be carried out [8]. It was also noted that other factors might affect the MICA, and that these should be investigated if possible. The National Institute for Biological Standards and Control (NIBSC), UK was designated to co-ordinate this investigation. Accordingly, an international collaborative study was initiated with the following aims: 1) To compare, using the MICA, the protective effect of ACV containing different concentrations of PT. 2) To provide WHO with more information about this model for evaluation of ACVs based on the results of the study, the experience and comments of the participants.

1045-1056/$36.00 Ó 2013 The International Alliance for Biological Standardization. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.biologicals.2013.11.008

102

D. Xing et al. / Biologicals 42 (2014) 101e108

3) To compare, if possible, the effect of different mouse strains on determination of active PT content using MICA, and if possible to assess the samples using respiratory challenge models.

Table 2 Sample information. Sample code

Description

A

3-component DTPa vaccine, each 0.5 ml dose contains 5 mg genetically detoxified PT, 2.5 mg FHA, 2.5 mg pertactin, 25 Lf Dip, 10Lf Tet with aluminium hydroxide as adjuvant. Sample A þ 5 ng PT/0.5 ml Aluminium hydroxide (contains 0.2% Al(OH)3) JNIH-3, 2-components freeze-dried acellular pertussis vaccine preparation contains 7.5 mg (PN) PT and 7.5 mg (PN) FHA and have a nominal potency of 40 units of pertussis vaccine per ampoulea. Sample A þ 50 ng PT/0.5 ml

2. Participants A total of eleven laboratories, including vaccine manufacturers and national control laboratories, nine of whom performed MICA and two of whom performed respiratory challenge tests (intranasal and aerosol) for the evaluation of ACVs were invited to participate in the study. Results from eight laboratories performing MICA and two laboratories performing respiratory challenge, listed in Table 1, were reported to NIBSC. Throughout this report participants are identified only by a randomly assigned number from 1 to 11. Separate assays have been numbered sequentially within laboratories.

B C D

E a

Subsequent to this study, JNIH-3 has been established as the First International Standard for Acellular Pertussis Vaccine for use in the MICA and assigned a potency of 34 IU/ampoule [23].

The participating laboratories were asked to evaluate 3 samples of ACV and an aluminium adjuvant control, all of which were coded by letter, together with JNIH-3 (coded D) as working reference preparation. The details of these samples are given in Table 2. Participants were also asked to include their in-house reference (IHR) (if possible).

description of their detailed assay procedure together with the raw data for each assay to permit data analysis using as far as possible a common method. If more than one mouse strain was available, participants were requested where possible to repeat the assay to investigate the effect of different mouse strains. Two assays were also carried out using intranasal challenge (INC) [22], and two assays were carried out using aerosol challenge (AC) [23]. Information about assay methods, as received from participants, is given in Table 3.

4. Study design

5. Statistical analysis

Participants were asked to carry out MICA [11,12] using their own methodology, reagents and animal strains, including challenge strains and controls used routinely in their laboratory. Each participant was requested to carry out at least two independent assays for each sample and to complete and return to NIBSC a

Data from each MICA have been analysed as a multiple parallel line assay [24] relating the probits of the proportion of mice surviving to the log of dose of each vaccine preparation using iterative maximum likelihood estimation. Where data permitted, the lethal dose 50% (LD50) of the challenge strains was also calculated using probit transformed proportions responding. Validity of individual assays has been assessed in terms of total deviations from the linear parallel line model for the multiple comparisons of preparations. Any assays for which the deviations from either linearity or parallelism were significant (p < 0.05) and for which the total deviations from the model were also significant (p < 0.05) were taken to be invalid, and estimates from these assays have not been included in the calculated mean estimates. Data from the AC assays conformed to the conditions for a multiple parallel line assay relating the log of observed colony forming units (CFU) at 7 days to the log of dose of each vaccine preparation. Estimates of potency and weights are shown with estimates of potency for the MICAs. Data for the INC assays consisted of observed CFU at several time points following administration of the same nominal total dose of each vaccine preparation. Since only a single total dose of each vaccine sample has been tested, any estimation of relative potency would necessarily be conditional on assumptions about the (unknown) nature of the relationship of the response to total dose of vaccine. For these assays, and for the AC assays (using responses for only the larger of the two doses of each vaccine to give comparability with the INC assays), ‘potency’ has been expressed as ratios of absolute activities, although it has been shown in previous studies that the values of such relative activities are highly dependent on the individual assay. For the AC assays, both relative activities and relative potency estimates have been calculated and can be compared. Estimates of potency have been assessed for homogeneity using a c2 test. Homogeneous estimates have been combined as weighted geometric means, using as weight the reciprocal of the asymptotic variance of the log potency. The reported weights reflect the

3. Materials

Table 1 Participating laboratories. Dr Alexandre Alves de S.O. Dias National Institute of Quality Control in Health (INCQS), Fundacao Oswaldo Cruz, Av. Brasil 4365-Manguinhos, Rio de Janeiro/RJ, CEP 21. 045-900, Brasil Dr. N. Guiso Laboratoire des Bordetella, Institut Pasteur, 25, Rue du Dr. Roux F-75724, Paris Cedex 15, France Dr. Koichiro Ojima Vaccine Production Department, Niigata Factory DENKA SEIKEN CO., LTD, 1-2-2 Minamihoncho, Gosen-shi, Niigata-ken, Japan 959-1695 Dr. Shin Sakuma Division 2, First Production Department, The Chemo-Sero-Therapeutic Research Institute, 1-6-1 Okubo, Kumamoto 860-8568, Japan Dr. Masaaki Nagai Division of Bacterial vaccines, Research Center for Biologicals The Kitasato Institute, 6-111, Arai, Kitamoto-shi, Saitama 364-0026, Japan Drs Kazunari Kamachi, Yoshinobu Horiuchi, Toshifumi Konda Department of Bacterial and Blood Products, National Institute of Infectious Diseases (NIID), 4-7-1, Gakuen, Musashimurayama-shi, Tokyo 208-0011, Japan Dr Choi, Soon Chul Division 2 of Quality Control, Quality Assurance Department, GreenCross Vaccine Corp., 227-3 Kugal-Ri Kiheung-Eup Yongin, 449-903 Korea Dr. Baokui Liu Department of Quality Control, National Vaccine and Serum Institute, Chaoyangou, Beijing 100024, People’s Republic of China Dr Lei Dianliang and Miss Hou Qiming National Institute for the Control of Pharmaceutical and Biological Products, Temple of Heaven, Beijing 100050, People’s Republic of China Dr D. Xing and Dr. R. Gaines-Das Division of Bacteriology and Informatics, National Institute for Biological Standards and Control, Blanche Lane, South Mimms, Potters Bar, Hertfordshire, EN6 3QG, U.K.

D. Xing et al. / Biologicals 42 (2014) 101e108

103

Table 3 Assay Information provided by participants. Lab code

Model

Mice strain

Age of mice

Immunisation programme

Time of challenge after immunisation

Length/volume of challenge/mouse

Challenge strain

Challenge dose CFU/ml or OD

1

ic

Slc:ddY

4.5 weeks

3 weeks

25 ml

18.323

2.0  105/ml

1

ic

Crj:CD-1(ICR)

4.5 weeks

3 weeks

25 ml

18.323

2.0  105/ml

2

ic

NIH

10e12 g

3 weeks

30 ml

18.323

2.67  106/ml

4

ic

Slc:ddY

4.5 weeks

3 weeks

25 ml

18.323

4  105/ml 6.96  105/ml

5a

ic

Slc:ddY

4.5 weeks

3 weeks

25 ml

18.323

4.96  105/ml 5.36  105/ml

6

ic

Slc:ddY

4 weeks

3 weeks

25 ml

18.323

2  106 cell/ml

7

ic

NIH

14e18 g

3 weeks

30 ml

18.323

Unspecified

8

ic

ICR

4 weeks

3 weeks

30 ml

18.323

OD 0.466 and 0.464 at 584 nml

9

ic

NIH

10e12 g

3 weeks

30 ml

18.323

2.7  106/ml

10

Intranasal

BALB/c

4 weeks

18.323

Aerosol

NIH

3e4 weeks

2 weeks from 2nd immunisation 3 weeks

50 ml

11

1/16 SHD 1/80 SHD 1/400HD 1/8 SHD 1/40 SHD 1/200 SHD 1/5 SHD 1/25 SHD 1/125 SHD 1/200 HD 1/16 SHD 1/80 SHD 1/400 SHD 1/16 SHD 1/8 SHD 1/80 SHD 1/40 SHD 1/400 SHD 1/200 SHD 1/200 HD 1/8 SHD 1/40 SHD 1/200 SHD 1/4SHD 1/16 SHD 1/64 SHD 1/256 SHD 1/8 SHD 1/40 SHD 1/200 SHD 1/5 SHD 1/25 SHD 1/125 SHD 1/4 SHD day 0 1/4 SHD day 14 Day 0 0.25 & 0.025 SHD

5 min

18.323

9.55  107/ml 1.1  108/ml 0.2 OD 5.5  108/ml

a

Two sets of immunisation doses were used. SHD: Single human dose.

variance of logarithm to base 10 of the potency estimates. Additionally, unweighted analysis of variance of log potency estimates has been used to provide a measure of between assay variability and to assess the significance of between laboratory differences of estimates relative to the pooled between assay variability. The variability of unweighted geometric means is expressed in terms of the geometric coefficient of variation (GCV) determined as exp(s) 1 expressed as percent, where s is the standard deviation of the logarithms of the estimates combined. 6. Results 6.1. Data contributed to the study Initial examination of the MICA data showed that in these laboratories, virtually all mice survived for 72 h following challenge. Exceptions were laboratories 2 and 7, where it was not uncommon for 20e25% of a group of mice to fail to survive this period. We do not know the reason(s) for this. However, the data from these two laboratories did not show a significant association of proportions surviving this initial period with any preparations or doses. Preliminary analysis of data showed that in all except laboratory 7 there was no significant regression of response on dose for preparation C (the negative control). Therefore, preparation C has been omitted from the multiple parallel line analysis carried out for the other preparations. After omission of preparation C, there were no statistically significant deviations from the linear parallel line model except in four assays. In assay 4 of laboratory 1 (using (ICR)

Crj:CD-1 mice) the total deviations from the model were marginally significant (p w 0.04) and appeared to result from the occurrence of a flat slope for preparation D and a steep slope for preparation E. These slope differences were not observed in the previous assay of these preparations (statistically valid analysis when 5 preparations were included). If preparation D was omitted, the assay of the remaining 4 preparations was statistically valid. Similarly, if preparation E was omitted, the assay of the remaining 4 preparations was statistically valid. It is likely that this marginal statistical deviation from the model when both preparations are included in the combined analysis reflects artefacts of these specific data rather than nonparallelism. For this reason, these data have been included in the analysis. Subsequent comparison of potency estimates with the previous statistically valid assay shows estimates from the two assays to be consistent within the limits of variability of the assays. In assay 1 of laboratory 4, the data were consistent with a survival rate of significantly greater than 0% for the controls and for the lowest doses of vaccine. The responses to the lowest doses of vaccine were therefore deemed unreliable, and the analysis for this assay has been based on only the two largest doses of each preparation. In assay 1 of laboratory 5, the responses to preparation A gave an exceptionally flat dose e response line, and the response to the largest dose of vaccine A was significantly less in this assay than in the subsequent assay, although this was not the case for any other sample. Preparation A has therefore been deleted from analysis of assay 1 in laboratory 5. In assay 1 of laboratory 7, responses to preparation E were inconsistent with any monotonic dose e response relationship, and this preparation has therefore

104

D. Xing et al. / Biologicals 42 (2014) 101e108

been omitted from analysis of this assay. Except as noted above, all data as reported have been included in analysis. 6.2. Bacterial challenge strain The challenge strain was assessed within the assay in which it was used by laboratories 1 (single group of unvaccinated controls in each assay), 4, 5 and 6 (relationship between survival and CFU of challenge), and laboratories 7 and 8 (relationship between survival and challenge expressed as dilution factor of LD50). Survivals in the unvaccinated control groups in the two assays of laboratory 1 in Slc:ddY mice were 1/10 (noted as a likely failure of the MICA challenge) and 0/20 and in the two assays of laboratory 1 in ICR (Crj:CD-1) mice were 0/15 and 0/14. LD50s of the challenge strain were found to be 121 and 298 CFU per dose in the two assays of laboratory 4, 165 and 224 CFU per dose in the two assays of laboratory 5 and 51 and 60 CFU per dose in the two assays of laboratory 6. The LD50 was found to occur at dilution factors of challenge of 161 and 106 in the two assays of laboratory 7 and at dilution factors of challenge of 548 and 326 in the two assays of laboratory 8. 6.3. Log dose e probit response lines for vaccines in MICAs

Table 4 Estimates of the potency (Pot) of the various in house reference preparations (IHR)a and of samples A, B and Eb. Laboratory code, mouse strain, and assay code

IHR

A

B

E

U/U

U/Shd

U/Shd

U/Shd

1 Crj:CD-1 3 1 Crj:CD-1 4 1 SIc:ddY 1 1 SIc:ddY 2 2 NIH 1 2 NIH 2 4 SIc:ddY 1 4 SIc:ddY 2 5 SIc:ddY 1 5 SIc:ddY 2 6 SIc:ddy 1 6 SIc:ddy 2 7 NIH 1 7 NIH 2 8 ICR 1 8 ICR 2 9 NIH 1 9 NIH 2 Weighted GM, 95% limits Unweighted GM of laboratory GMs 95% limits GCV Omit lab. 7

1.94 33 0.72 31 0.83 46 0.35 37 0.85 27 0.81 28 0.49 39 0.37 29 0.68 52 0.63 36 NA NA NA NA 0.77 54 0.54 45 0.69 22 0.99 19 0.68 0.55e0.84 0.70

Pot

Wt

Pot

Wt

4.28 30 2.13 30 4.38 43 2.44 36 4.12 26 5.46 27 2.67 36 2.66 27 * 2.51 33 3.98 25 8.04 33 2.73 23 4.51 12 3.61 53 3.08 44 5.36 21 2.36 14 3.53 2.89e4.31 3.47

Pot

Wt

Pot

Wt

6.18 32 2.33 30 2.99 39 3.37 37 6.75 26 5.84 27 2.31 34 2.70 27 3.29 46 2.76 34 7.36 28 8.56 33 9.53 21 14.80 10 4.23 53 4.40 45 7.41 22 4.58 16 4.37 3.61e5.28 4.80

5.61 32 4.41 31 7.97 46 4.56 38 9.00 27 11.43 26 3.54 38 5.51 29 4.61 49 3.45 35 8.34 29 8.56 33 ** 24.71 9 3.90 53 2.59 43 6.13 21 7.10 18 5.41 4.46e6.55 6.59

3.25e7.09 66% 4.28 3.06e6.00 50%

4.11e10.6 85% 5.59 4.04e7.74 48%

4.83 6.02

6.46 6.74

The slopes of the individually fitted log dose e probit response lines, excluding sample C, were determined and assessed. All assays conformed satisfactorily to the linear parallel line model (probability of observed total deviations from model >0.1) in all cases with the exception of laboratory 1 assay 4 as discussed above. Considered over all assays, there were no consistent differences of slopes among the preparations A, B, D and the various IHR (data not shown). Analysis of variance of the slopes showed that there was a marginal (0.05 < p < 0.1) tendency for slopes for JNIH-3 and/or the various IHR to be steeper than the slopes for samples A, B and E, which was more marked in assays using mouse strains ICR or Slc:ddY. However, there was no consistent overall tendency for the slopes for samples A, B and E to differ from one another, and no consistent overall difference between slopes for JNIH-3 and the various IHR.

The weight (Wt), determined as the reciprocal of variance of log base 10 of the estimated potency. NA denotes that the preparation was not included in the assay. GM: Geometric mean; GCV: Geometric coefficient variation; * and ** denote preparations which have been deleted from the assay (as discussed in text). a Expressed as Unit of JNIH-3 (nominal 40 units/amp) equivalent in activity to 1 nominal Unit of the IHR. b Expressed as Unit of JNIH-3 (nominal 40 units/amp) equivalent in activity to 1 SHD of sample.

6.4. Between assay variability and assay precision

6.6. Estimates for vaccine samples A, B and E

Within assay variance as reflected in the weights of the log potency estimates was broadly similar across laboratories and for the different preparations, being on average about 32. Nevertheless, data are consistent with more precise assays in some laboratories than in others, in part reflecting the number of mice tested per group. The between assay variance, pooled over all laboratories and preparations, was consistent with a somewhat larger average weight of about 50. Within laboratories estimates for each preparation did not show significant deviations from homogeneity (Table 4). Considered over all laboratories, estimates for any preparation also did not differ significantly from homogeneity. However, unweighted analysis of variance showed significant between laboratory differences for samples B and E, but not for sample A and the various IHR.

Estimates from each assay, and overall mean estimates are shown in Table 4. Within the individual assays, the three samples were not clearly distinguished, since precision of the individual assays is limited. However, when considered over all assays, there was a clear tendency for sample A to be the least potent and for sample E to be the most potent. Estimates of the potency for sample A were homogeneous when considered over all laboratories, and the between laboratory variance for sample A was similar to the pooled within laboratory variance. Although the estimates of potency of samples B and E were homogeneous when considered over all laboratories, it was noted that the between laboratory variances for samples B and E were some 4-fold larger than the between laboratory variance of sample A or than the pooled within laboratory variance. Exclusion of estimates from laboratory 7 substantially reduced the between laboratory variances for samples B and E, but these were still more than 2-fold larger than that for sample A (and significantly, p < 0.05, larger than the pooled within laboratory variance). Estimates from laboratories 1 and 8, using mouse strain ICR, and from laboratories 1, 4 and 5, using mouse strain Slc:ddY were smaller than estimates from laboratories 6, also using mouse strain Slc:ddY, and laboratories 2, 9 and 7 using mouse strain NIH. If estimates from laboratories 6 and 7 are excluded, analysis of variance of log estimates shows statistically significant (p < 0.01)

6.5. Comparisons of JNIH-3 with the various IHR Estimates of the potency of the various IHR relative to JNIH-3 expressed as unit of JNIH-3 equivalent in activity to 1 nominal unit of IHR are given in Table 4. Considered over all laboratories, the estimates of potency for the various IHR were homogeneous. However, the between laboratory variance was about two-fold larger than the pooled within laboratory variance.

11 1 2

0.51e0.95 39%

2.85e4.24 30% 3.47 2.75e4.37 32% AC assays in NIH mice NA 2.62 50 NA 3.02 151

51 144

51 144

D. Xing et al. / Biologicals 42 (2014) 101e108

differences both between samples and between mouse strains. However, inclusion of laboratory 6 (for which assays were valid) suggests that between laboratory variance with nominally the same mouse strain may be as large as the differences between different mouse strains. Estimates of potency of the three vaccine samples were significantly correlated with one another (p < 0.05). Thus, the three vaccine samples have been directly compared with one another, as shown in Table 5. For each of the comparisons among samples A, B and E, the between laboratory variance (excluding estimates from laboratory 7) is not significantly larger than the pooled within laboratory variance, indicating that the estimates are broadly consistent between laboratories. The mean estimates indicate that sample B is some 25% more potent than sample A, and that sample E is some 25% more potent than sample B, in these assay systems. Although the ratios of estimates B/A and E/B are only marginally significantly different from 1, the ratio E/A is significantly larger than 1, considered overall laboratories. The data presented here do not permit generalization about the effect of mouse strain or other assay conditions on the sensitivity of the assay to the presence of PT, although the larger between laboratory variance for samples B and E than for sample A suggests that such effects may occur. For example, in assays in ICR mice in laboratory 8, samples E and A are not distinguished, and the potency of E relative to B is less than 1 although sample E contains a larger concentration of PT than sample B. 6.7. Estimates obtained using AC and INC assays Estimates of potency obtained using the AC assays are shown in Table 4 for comparison with estimates obtained using MICAs. There Table 5 Estimates of the potency of the three vaccine samples relative to one anothera. Laboratory code, mouse strain, and assay code

B/A

E/A

E/B

1 Crj:CD-1 3 1 Crj:CD-1 4 1 SIc:ddY 1 1 SIc:ddY 2 2 NIH 1 2 NIH 2 4 SIc:ddY 1 4 SIc:ddY 2 5 SIc:ddY 1 5 SIc:ddY 2 6 SIc:ddy 1 6 SIc:ddy 2 7 NIH 1 7 NIH 2 8 ICR 1 8 ICR 2 9 NIH 1 9 NIH 2 GM of laboratory GMs 95% limits GCV GM of laboratory GMs, omit lab 7 95% limits GCV 11 AC assays in NIH mice 1 2

1.44 1.09 0.68 1.38 1.64 1.07 0.86 1.01

1.31 2.08 1.82 1.87 2.19 2.09 1.33 2.07

1.10 1.85 1.06 3.49 3.28 1.17 1.43 1.38 1.94

1.37 2.09 1.06

0.91 1.90 2.66 1.35 1.33 1.96 1.54 2.04 1.40 1.25 1.13 1.00

5.48 1.08 0.84 1.14 3.00

1.67 0.92 0.59 0.83 1.55

1.36 1.02e1.84 47%

1.81 1.26e2.61 61%

1.34 1.06e1.69 35%

1.22 1.04e1.43 21%

1.58 1.29e1.94 28%

1.30 1.01e1.69 36%

1.84 1.99

2.46 2.23

1.34 1.12

SHD: single human dose; GM: geometric mean; GCV: geometric coefficient variation. a Expressed as SHD of one sample equivalent to 1SHD of the other, namely as SHD of A equivalent to 1 SHD of B, as SHD of A equivalent to 1 SHD of E, and as SHD of B equivalent to 1SHD of E. Estimates for the AC assays are shown for comparison.

105

was a tendency for the log dose e response lines of samples A and D to be flatter than those for samples B and E in the AC assays, and this difference was statistically significant (p w 0.05) in assay 2. However, this finding may also be influenced by the observed response ranges, as it was also noted that the smaller of the two doses used in these assays gave responses which differed only marginally from the responses to the negative controls, and did not show significant differences among the four vaccine samples. Data for the INC assays did not allow estimation of relative potency, since each of the vaccine samples was tested in the same way and using the same nominal amount of vaccine, although CFU levels were obtained at 4 or 3 times (assays 1 and 2 respectively). There were possible differences among the vaccine preparations (excluding the negative control) only at two of these times (days 5 and 8). The ratios of the geometric mean CFU values are shown for each of the two times at which there was the possibility of some difference among the preparations (Table 6). These ratios of activity provide an indication of the order of activity of the samples, but do not otherwise reflect potency of the samples. Although CFU values were obtained only at a single time for the AC assays, the comparable ratios are shown for comparison in Table 6. The dependence of ratios such as these on the assay specific levels of response and rates of change of response with change in dose can be seen from comparison of these ratios of activity with the estimates of potency for the AC assays. The order of activity of preparations A, B and E differs markedly between days 5 and 8 in the INC assays. This appears to reflect a tendency for samples B and E to show a marked reduction in CFU from controls between days 0 and 5, but to show relatively little change in CFU level between days 5 and 8, whereas samples A and D, although showing a less marked reduction than B and E at day 5, show an additional reduction in CFU levels between days 5 and 8 (as great or greater than the reduction between day 0 and day 5). 7. Discussion 7.1. Assessment of validity This study is limited to consideration of results for one or two assays from any one laboratory. In general, we considered an assay to be valid if deviations from linearity and parallelism were not significant (p > 0.05) or if total deviation from the linear parallel line model was not significant. As noted in the results, estimates

Table 6 Relative activity of the vaccine sample to JNIH-3 (sample coded D) determined by the INC assaysa and the AC assaysb. Assay method

Response variable

INC

Day 5 Day 8 Days 5 and 8 combined Day 5 Day 8 Days 5 and 8 combined Day 7 Day 7

AC*

Assay number

Vaccine sample A

B

E

1

2.81 1.17 2.29 10.1 0.31 1.76 0.29 0.81

24.6 0.24 2.42 36.2 0.04 1.23 2.44 27.3

103 2.29 15.4 29.2 0.35 3.21 4.47 17.3

2

1 2

*Relative activity base on only the larger of the two vaccine doses tested. a Relative activity obtained as the ratio of the ‘absolute activity’ (expressed as geometric mean CFU) of the vaccine sample to the ‘absolute activity’ of JNIH-3 (sample coded D)as determined for the INC assays using day 5 and day 8 responses. b As determined for the AC assays using day 7 responses for only the larger of the two vaccine doses tested. Estimates of relative potency calculated for the AC assays using the methods of parallel line assay analysis are shown for comparison, and indicate the greater reproducibility associated with relative potency based on observed dose e response relationships than associated with ‘absolute activity’.

106

D. Xing et al. / Biologicals 42 (2014) 101e108

from one assay which would have been omitted under strict interpretation of this criterion have been included in the analysis. It is our experience that the marginal significance in this case is likely to have resulted from an artefact of assay design or failure of the theoretical assumptions about the distribution of the assay data. Repeated occurrence of such marginal cases within an individual laboratory would be considered to require further analysis. The observation of a significant ‘dose’ response relationship for the negative control in laboratory 7 raises questions about the validity of assays in this laboratory. If over 90% of mice are protected by a ‘dose’ of the negative control, as occurred for one group in this laboratory, then all responses up to this level could be an artefact of assay design. Thus, combined estimates have been determined both including and excluding estimates from laboratory 7. 7.2. Calibration of IHR preparations Doseeresponse lines did not differ significantly between JNIH-3 and the various IHR. However, the various IHR generally appear to be somewhat less potent than the potency of JNIH-3, with on average 0.7 unit of JNIH-3 equivalent in activity to 1 nominal unit of IHR. The subsequent calibration of JNIH-3 is described elsewhere [25]. The larger between laboratory variance compared to the pooled within laboratory variance suggests that there may be differences in potency of the various IHR between laboratories. This is not unexpected, as the different laboratories use different IHR, which are in some instances whole cell pertussis vaccines. 7.3. Effect of PT on potency estimates The MICAs did not provide evidence of differences among the vaccine samples which were reflected in the dose e response relationships. However, the AC assays suggested the possibility of some difference in response rate between the samples, with the samples containing PT tending to give steeper dose e response lines. Similarly, the two INC assays suggested that the time course of CFU reduction in response to preparations containing PT might differ from that for preparations which do not contain PT (assuming that JNIH-3 does not contain PT). Considered overall, the vaccine samples containing PT appeared to be more potent than the equivalent sample without PT, and the sample containing the larger amount of PT appeared to be more potent than the sample containing the smaller amount of PT (Tables 4 and 6, Fig. 1). The results also indicated that the synergistic effect of PT on protection was not confined to intra-cerebral (ic) challenge but also extended to respiratory infection. This is in agreement with previous findings that PT enhances the protective activity of other antigens of Bordetella pertussis in the ic challenge model [21,26e28]. It has been reported that PT enhances nitric oxide (NO) production through interferon gamma (IFN-g) production [29] and the effect of PT may be exerted through its adjuvant action on acquired immunity, especially on antigen-specific T cell responses, which would activate macrophages but also more non-specifically through the stimulation of antigen-independent IFN-g. The initiation of this process is most effectively achieved by a combination of B. pertussis antigens including pertussis toxoid (PTd), filamentous haemagglutinin (FHA), fimbriaes (Fims), pertactin (Prn) and possibly other antigens, in the presence of small quantities of PT [30]. However, the effects of PT on the estimates may also differ depending on assay conditions e.g. mouse strain as discussed below. 7.4. Effect of MICA conditions on potency estimates The MICAs were carried out in different laboratories with consequently different assay conditions and different strains of

mice. The resulting data indicate that these different conditions may influence the effect of PT on potency. For sample A, containing no PT, estimates in terms of JNIH-3 (also containing no PT) were homogeneous across laboratories, and the between laboratory variance was not larger than the pooled within laboratory variance. However, for samples B and E, each containing PT, the between laboratory variances for estimates in terms of JNIH-3 were larger than that for A, and significantly larger than the pooled within laboratory variance. For the samples containing PT, there is no clear distinction of estimates into groups which contribute to the excess variability. Two-fold differences of estimates for the same sample between assays within the same laboratory are not uncommon. Thus any differences due to assay conditions are not of sufficient magnitude to be apparent given the limited numbers of assays, and the variability of the individual estimates. The comments below, are thus suggestive, but confirmation of the effect of assay conditions would require further study and large numbers of assays. Three different mouse strains were used in MICAs in this study, although strains with the same name may nevertheless differ between laboratories (e.g. different breeding, colonies). The Slc:ddY strain was used in four laboratories, and estimates differed as greatly among these four laboratories as among laboratories using different strains, suggesting that even where a nominally common strain was used, other assay conditions may vary. For example, in laboratories 6 (Slc:ddY mice) and 8 (ICR mice), the potency of the sample with the larger amount of PT was less than or similar to that of the sample with the smaller amount of PT. However, three other laboratories in addition to laboratory 6 also used Slc:ddY mice and laboratory 1 also used ICR mice and did not see this effect, suggesting that other assay conditions may also be influential. Comparison of estimates from two assays in each of two mouse strains in one laboratory (laboratory 1) showed no significant differences, but as noted above, any differences which would be detectable with limited numbers of assays would be greater than two-fold. There was a suggestion of differences in response to PT of the Slc:ddY and NIH strains (Fig. 1). The mean estimate for sample B (5 ng PT) in terms of sample A (no PT) in three of the four laboratories using Slc:ddY mice was about 1.0, with overall mean of the four laboratory means 1.09 Single Human Dose (SHD) of A equivalent to 1 SHD of B. However, for the two laboratories using NIH mice (excluding laboratory 7) the estimates were larger than 1.0 with mean of the two laboratory means 1.47 SHD of A equivalent to 1 SHD of B. Results for sample E (50 ng PT) were means of 1.58 SHD of A equivalent to 1 SHD of E using Slc:ddY mice and 1.99 SHD of A equivalent to 1 SHD of E using NIH mice. It was also noted that the results in the two AC assays carried out in a single laboratory used NIH mice, and the estimates in these assays showed better agreement with the MICAs using the same strain of mice than with the MICAs using different mouse strains. Because of the variability of the individual estimates, and the limited numbers of estimates, the observed differences are suggestive, but are not statistically significant. Nevertheless, the effect of PT in each of the strains of mice was to increase the apparent potency of the vaccine, although the extent of the increase and the amount of PT required may vary depending on assay conditions and mouse strain (Fig. 1). A previous report showed that a range of ACV formulations with different numbers of purified acellular pertussis (aP) components did not show significant dissimilarities from one another in the MICA system, suggesting that apart from co-purified ACVs MICA could potentially be used for assay for purified products [25]. In the present study, estimates for potency of vaccine A (without addition of active PT) in terms of JNIH-3 and IHRs are homogeneous over laboratories. Thus this study indicates that the MICA may also be suitable for genetically detoxified vaccines.

D. Xing et al. / Biologicals 42 (2014) 101e108

107

A(0ng PT)

4 11AC

3 9N

2

7N

1C

3.5 5S

4S 11AC

1S

4S

2N

7N

9N

6S

1S

2N

1

er of Estimates Numbe s

0

1C

8I

8I

6S

B(5 PT) B(5ng 4 4.4

3

11AC

2 1

1C

5S

4S

4S

5S 1S

1S

8I

8I

1C

9N 11AC

2N

9N 2N

6S

6S

7N

7N

6

E(50ng PT)

5 4 5.4

3 1C

6S

2 5S

1C

11AC

6S

1S

4S

9N 11AC 9N

1S

1 0

2.2

8I

5S

2.7

3.3

4S

8I

4.1

5.0

6.0

7.4

2N

2N

9.0

11

7N

15

25

Potency (unit of JNIH-3/SHD) Fig. 1. Estimates of the relative potency, expressed as units of JNIH-3 (coded D), for vaccine samples A (0 ng PT), B (5 ng PT) and E (50 ng PT). The value indicated by the arrow is the weighted geometric mean of all estimates by IC challenge. Each box denotes the estimate from an individual assay. The label in the box gives the laboratory code followed by the first letter of the mouse strain for assays carrying out MICA, and for laboratory 11 identifies the method as aerosol challenge (AC). Solid boxes denote MICAs carried out using mouse strain Slc:ddY; diagonally shaded boxes denote MICAs carried out using mouse strain NIH excluding laboratory 7; grey boxes denote AC challenge assays carried out using NIH mice, which are included in this plot for comparison with MICA estimates, especially those also using NIH mice; open boxes denote strains of mice (ICR (Crj:CD)) each used in a single laboratory except for laboratory 7. SHD ¼ single human dose.

7.5. Comparison with traditional Kendrick test

8. Conclusions

The MICA assay in this study performs similarly to the traditional Kendrick tests specified by various pharmacopoeias for control of whole cell pertussis vaccines (except for the extended immunisation period from 2 weeks to 3 weeks). In particular, the precision of the individual MICA assays was similar to that observed for the Kendrick test in a WHO proficiency study of whole cell pertussis vaccines [31]. In this study, the single laboratory carrying out AC assays obtained results in good agreement with the results from the MICA assays, and individual assays appeared to have slightly improved precision in comparison with the MICA assays. Although the order of activity of samples appeared similar in the single laboratory carrying out INC assays, no estimates of potency were available, as only a single dose of vaccine was used for each sample.

The models included in this study all showed an ability to distinguish mice vaccinated with ACVs from unvaccinated mice (Sample code C) except one laboratory. The MICA appeared to be suitable for control of genetically detoxified vaccines. Vaccine samples containing active PT (5 or 50 ng/SHD) generally appeared to be more potent, as well as having greater interlaboratory variability, than the equivalent sample without active PT, and the sample containing the larger amount of active PT generally appeared to be more potent than the sample containing the smaller amount of PT in MICA. Similar results were obtained in the AC assays in a single laboratory and a similar trend was also found in the INC assays in a single laboratory. Data on the effect of MICA assay conditions, including mouse strain, on the calibration of vaccine samples containing PT is suggestive, but insufficient for any general conclusions. Further study

108

D. Xing et al. / Biologicals 42 (2014) 101e108

to evaluate the sensitivity of different mouse strains to active PT may be of interest.

Acknowledgements Grateful acknowledgements are due to Dr Nicolas Hug, Chiron, Italy, for provision of vaccine samples used in this study. We thank Miss Penny Newland for her efforts in distribution of the samples to participants. We also thank all of the participants for their helpful contributions to the study.

References [1] Greco D, Salmaso S, Mastrantonio P, Giuliano M, Tozzi AE, Anemona A, et al., The Progetto Pertosase Working Group. A controlled trial of two acellular vaccines and one whole cell vaccine against pertussis. N Engl J Med 1996;334:341e8. [2] Gustafsson L, Hallander HO, Olin P, Reizenstein E, Storsaeter J. A controlled trial of two component acellular, a five component acellular, and a whole cell pertussis vaccine. N Engl J Med 1996;334:349e55. [3] Olin P, Rasmussen F, Gustafsson L, Hallander HO, Heijbel H. The Ad Hoc Group for the study of pertussis vaccines. Randomised controlled trial of twocomponent, three-component, and five component acellular pertussis vaccines compared with whole-cell pertussis vaccines. Lancet 1997;350:1569e77. [4] Heininger U, Cherry JD, Christenson PD, Eckhardt T, Göering U, Jakob P, et al. Comparative study of Lederle Takeda acellular and Lederle whole cell pertussis-component diphtheria-tetanus pertussis vaccines in infants in Germany. Vaccine 1994;12:81e6. [5] Trollfors B, Taranger J, Lagergard T, Lind L, Sundh V, Zackrisson G, et al. A placebo-controlled trial of a pertussis toxoid vaccine. N Engl J Med 1995;333:1045e50. [6] Simondon F, Preziosi MP, Yam A, Kane CT, Chabirand L, Iteman I, et al. Randomized double blind trial comparimg a two component acellular to a whole cell pertussis vaccine in Senegal. Vaccine 1997;15:1606e12. [7] Corbel MJ, Xing DKL, Kreeftenberg JG. Informal consultation with manufacturers and the Ad Hoc Working Group on mouse protection models for acellular pertussis vaccines. Biologicals 1999;27:189e93. [8] Corbel MJ, Mastrantonio P, Kreeftenberg JG. Meeting report: Ad Hoc working group on acellular pertussis vaccines World Health Organisation, Geneva, 2728 July 2000. Vaccine 2002;20:288e91. [9] Corbel MJ, Kreeftenberg JG, Knezevic I. WHO working group on the standardisation and control of pertussis vaccines-report of a meeting held on 6e7 May 2003, Ferney Voltaire, France. Vaccine 2004;22:293e300. [10] WHO Expert Committee on Biological Standardization. Recommendations to assure the quality, safety and efficacy of acellular pertussis vaccines. Available from: http://who.int/biologicals/vaccines/pertussis/en/index.html; June 2012. [11] Ministry of Health and Welfare, Japanese Government. Minimum requirements for biological products of Japan; 2006. pp. 146e9. [12] The Chinese pharmacopeia. 8th ed.; 2005. pp. 67e71. [13] Sato Y, Sato H. Further characterization of Japanese acellular pertussis vaccine prepared in 1988 by 6 Japanese manufacturers. Tokai J Exp Clin Med 1988;(Suppl. 13):79e88.

[14] Kamiya H, Nii R. Overview of currently available Japanese acellular pertussis vaccines and future problems. Tokai J Exp Clin Med 1988; (Suppl. 13):45e9. [15] Isomura S. Clinical studies on efficacy and safety of acellular pertussis vaccine. Tokai J Exp Clin Med 1988;(Suppl. 13):39e43. [16] Heininger U, Cherry JD, Stehr K, Schmitt-Grohe S, Uberall M, Laussucq S, et al. Comparative efficacy of the Lederle/Takeda acellular pertussis component DTP (DTaP) vaccine and Lederle whole-cell component DTP vaccine in German children after household exposure. Pertussis Vaccine Study Group. Pediatrics 1998;102:546e53. [17] Kato T, Kaku H, Arimoto Y. Protection against pertussis by Takeda’s acellular pertussis vaccine: household contact studies in Kawasaki City, Japan. Tokai J Exp Clin Med 1988;(13 Suppl.):35e8. [18] Cherry JD, Mortimer EA, Hackell JG, Scott JV. Clinical trials in the United States and Japan with the Lederle-Takeda and Takeda acellular pertussisdiphtheriatetanus (APDT) vaccines. The Multicenter APDT Vaccine Study Groups. Dev Biol Stand 1991;73:51e8. [19] Heininger U, Stehr K, Christenson P, Cherry JD. Evidence of efficacy of the Lederle/Takeda acellular pertussis component diphtheria and tetanus toxoids and pertussis vaccine but not the Lederle whole-cell component diphtheria and tetanus toxoids and pertussis vaccine against Bordetella parapertussis infection. Clin Infect Dis 1999;28:602e4. [20] Robinson A, Irons LI. Synergistic effect of Bordetella pertussis lymphocytosispromoting factor on protective activities of isolated Bordetella antigen in mice. Infect Immun 1983;40:523e8. [21] Robinson A, Funnell SGP. Potency testing of acellular pertussis vaccines. Vaccine 1992;10:139e41. [22] Guiso N, Capiau C, Carletti G, Poolman J, Hauser P. Intranasal murine model of Bordetella pertussis infection. I. Prediction of protection in human infants by acellular vaccines. Vaccine 1999;17:2366e76. [23] Xing D, Das RG, Willams L, Canthaboo C, Tremmil J, Corbel MJ. An aerosol challenge model of Bordetella pertussis infection as a potential bioassay for acellular pertussis vaccines. Vaccine 1999;17:565e76. [24] Finney DJ. Statistical methods in biological assay. 3rd ed. London: Charles Griffin and Co Ltd.; 1978. [25] Gaines-Das R, Horiuchi Y, Zhang SM, Newland P, Kim Y, Corbel M, et al. Modified intra-cerebral challenge assay for acellular pertussis vaccines: comparisons among whole cell and acellular vaccines. Vaccine 2009;27: 6824e32. [26] Munoz JJ, Arai H, Cole RL. Mouse protecting and histamine-sensitizing activities of pertussigen and fimbrial hemagglutinin from Bordetella pertussis. Infect Immun 1981;32:243e50. [27] Munoz JJ, Peacock MG. Role of pertussigen (pertussis toxin) on the mouse protective activity of vaccines made from Bordetella species. Microbiol Immunol 1989;33:341e55. [28] Amiel SA. The effects of Bordetella pertussis vaccination on cerebral vascular permeability. Br J Exp Pathol 1976;57:653e62. [29] Ryan M, Murphy G, Gothefors L, Nilsson L, Storsaeter J, Mills KHG. Bordetella pertussis respiratory infection in children is associated with preferential activation of type 1 T helper cells. JID 1997;175:1246e50. [30] Xing D, Canthaboo C, Corbel MJ. Effect of pertussis toxin on induction of nitric oxide in murine macrophages and protection in vivo. Vaccine 2000;18:2110e9. [31] Xing D, Gaines Das R, O’Neill T, Corbel M, Dellepiane N, Milstien J, et al. Laboratory testing of whole cell pertussis vaccine: a WHO proficiency study using the Kendrick test. Vaccine 2002;20:342e51.