The second international reference preparation of procaine benzylpenicillin in oil with aluminium monostearate*

Bull. Org. mond. Sant) 1973, 48, 91-98Bull. Wld Hith Org.J The second international reference preparation of procaine benzylpenicillin in oil with aluminium monostearate* JILLIAN M. BOND,1 J. W. LIGHTBOWN,1 & MARJORIE V. MUSSETT2 The second International Reference Preparation of Procaine Benzylpenicillin in Oil with Aluminium Monostearate was established in replacement of the first international reference preparation. The blood-level duration test (WHO Requirements for Biological Substances, No. 9), used to evaluate batches of manufactured material in relation to this reference preparation, was revised to relate to the new reference preparation. Thus a batch now passes the test if the lower confidence limit of the ratio is greater than 0.45, instead of0.50-the value that related to the first international reference preparation. This change will ensure that material passing the test in the future will be of the same quality as in the past. The WHO Expert Committee on Biological Stan- dardization (1964b) noted that stocks of the Inter- national Reference Preparation (IRP) of Procaine Benzylpenicillin in Oil with Aluminium Monostear- ate (PBAM) would soon be exhausted and that material that appeared to be suitable to serve as a second international reference preparation was avail- able. The Committee requested the National Institute for Medical Research, London, to arrange for a col- laborative study of that material. A number of batches of PBAM had been exam- ined by the laboratories of the Food and Drug Administration, Washington, D.C., USA, and of the National Institute for Medical Research, with respect to their blood level and other proper- ties. On the basis of that preliminary screening, one batch had been selected as probably being suitable to replace the IRP. The batch consisted of 4000 vials each containing 10 ml of a suspension of 300 000 IU of procaine benzylpenicillin- per milli- litre in arachis oil with aluminium monostearate. The material 3 was received at the National Institute * From the WHO International Laboratory for Biological Standards, National Institute for Medical Research, Mill Hill, London, England. Division of Biological Standards. 'Statistical Services Section. 3Donated by the Upjohn Company, Kalamazoo, Mich., USA, and obtained through the good offices of Dr W. W. Wright, Food and Drug Administration, Department of Health, Education, and Welfare, Washington, D.C., USA. for Medical Research, London, in December 1964 and stored at room temperature in the dark. One thousand further vials were made available to WHO, Geneva, Switzerland, for clinical studies in venereal diseases and treponematoses. THE COLLABORATIVE STUDY In contradistinction to international standards and most international reference preparations, the IRP of PBAM is not intended to define an international unit of activity but is stated to be a specific batch of PBAM having adequate blood level duration properties, with which all other batches should be compared. The properties of the IRP have been described (Bond et al., 1965). The blood-level duration test, to be used in the comparison, is described in Part A, section 5.3.2, of the Require- ments for Procaine Benzylpenicillin in Oil with Aluminium Monostearate (Requirements for Bio- logical Substances No. 9) (WHO Expert Committee on Biological Standardization, 1964a). In this test, equal doses of the IRP and of the test sample are injected into groups of rabbits and the concentration of penicillin in the blood at 72 hours is determined. The ratio (arithmetic mean blood level of sample group)/(arithmetic mean blood level of IRP group)- referred to as the activity ratio in this paper-is cal- culated, together with the 99% confidence limits of 2984 -91- J. M. BOND ET AL. that ratio. If the lower confidence limit was greater than 0.5, the test material was satisfactory. The most important requirement for the replace- ment reference preparation was that it should have blood-level duration ploperties similar to those of the existing reference preparation. Since it was unlikely that the latter could be duplicated exactly, it was considered that the lower confidence limit in the blood-level duration test should be altered so that, if any material to be compared with the new reference preparation passed the test, this would mean that it was of the same minimum quality as materials -found acceptable in the past. The main purpose of the collaborative study was to allow such a revalua- tion of this lower confidence limit to be made if it was found to be necessary. Fifteen laboratories in 9 different countries par- ticipated in the collaborative study (see Annex 1). Every laboratory received 4 vials each of the IRP and of the proposed second international reference preparation (PIRP 2). The laboratories are referred to throughout the present article by code numbers bearing no direct relation to the order in which the laboratories are listed in Annex 1. Contrary to the normal procedure in collaborative assays, the par- ticipants were asked to compare the existing and proposed international reference preparations by a specific test under defined conditions. Details of the procedure to be followed were sent to each labora- tory in the form of a memorandum (Annex 2). This also indicated the information and results required. Since all the laboratories were to carry out the same test, they were supplied with forms on which the information could be entered and submitted for statistical evaluation. RESULTS The results were analysed statistically in an attempt to answer the following questions. (1) Is the proposed second international reference preparation suitable to replace the International Reference Preparation? (2) If so, is it necessary to alter the level at which material passes the blood-level duration test-i.e., a lower 99% confidence limit of 0.5 for the activity ratio? If it is, what level would be satisfactory for the new preparation? (3) Is there evidence that the use of a cross-over design significantly affects the results of such a comparison? (4) Is it more correct to base the calculation of activity ratio and limits on the logarithms of the blood levels rather than on the actual blood levels? STATISTICAL ANALYSIS The basic design of the study consisted of two groups of rabbits, one of which was tested with the IRP and the other simultaneously with the PIRP 2. Three weeks later the treatments were crossed between the same two groups of rabbits. The size of the groups ranged initially from 5 to 16 rabbits, but was sometimes reduced by the death of animals during the course of the experiments. Some laboratories repeated the entire comparison (Table 1). Although most of the laboratories carried out the second part of the comparison 21 days after the first, Laboratory 15 allowed 28 days to elapse and Laboratory 4 (which made the comparison twice), 21 and 31 days. At Laboratory 9, where a cross-over design was not used, IRP was injected into one group of 5 rabbits and, after an interval of 3 or 4 days, PIRP 2 into a second group. This procedure was then repeated with different sets of rabbits. The data from each laboratory were analysed in three ways wherever possible: Method 1. The logarithms of the responses (i.e., the potencies of penicillin in the serum samples 72 hours after injection) were obtained. The differ- ence between each pair of values for each rabbit- i.e., for IRP and PIRP 2-represented the log activ- ity ratio of PIRP 2 to IRP estimated from one rabbit. Table 1 shows the geometric means of these ratios, together with the lower confidence limits, based on within-group variation. It was not possible to analyse the results from Laboratory 9 by this method since, in that laboratory, the comparison was not made according to the usual practice in collaborative studies of international reference preparations (see p. 95). Method 2. The log activity ratio was calculated as the difference between the means of logged responses obtained with PIRP 2 and with IRP. This provided the same activity ratios as calculated by Method 1. The cross-over design was, however, ignored when confidence limits were estimated, the error variance being obtained by pooling the vari- ances for each of the 4 subgroups of logarithmic values (Table 1). 92 PROCAINE BENZYLPENICILLIN Table 1. Activity ratios derived from individual tests Labor- No. of Activity ratio at 72 hours (lower confidence limit: P = 0.99) atory rabbits Cross-over (logged) Not crossed (logged) Not crossed (unlogged) No. Method 1 Method 2 Method 3 1 19 0.882 (0.578) 0.882 (0.598) 0.785 (0.453) 2 24 1.146 (0.674) 1.146 (0.692) 1.073 (0.560) 3 19 1.724 (0.648) 1.724 (0.723) 1.744 (0.090) 4 10 0.847 (0.394) 0.847 (0.409) 0.937 (0.402) 10 1.039 (0.612) 1.039 (0.634) 1.098 (0.588) 5 20 1.453 (0.671) 1.453 (0.671) 1.372 (0.779) 6 20 1.968 (0.712) 1.824 a (1.389) 1.507 a (0.677) 7 18 1.144 (0.822) 1.144 (0.792) 1.063 (0.666) 8 20 1.078 (0.976) 1.078 (0.974) 1.089 (0.991) 9 20 - 0.922 (0.776) 0.926 (0.627) 10 31 0.924 (0.474) 0.924 (0.506) 0.894 (0.460) 11 20 0.984 (0.404) 0.984 (0.419) 1.068 (0.528) 12 8 0.946 (0.100) 0.946 (0.176) 0.929 (0.017) 8 1.259 (0.080) 1.259 (0.167) 1.358 (0.177) 13 10 1.023 (0.645) 1.023 (0.670) 1.044 (0.608) 10 1.074 (0.684) 1.074 (0.722) 1.079 (0.664) 14 10 0.731 (0.290) 0.731 (0.309) 1.019 (0.094) 10 1.195 (0.623) 1.195 (0.713) 1.099 (0.570) 9 0.842 (0.372) 0.842 (0.413) 0.928 (0.300) 9 1.259 (0.643) 1.259 (0.637) 1.332 (0.396) 15 20 1.187 (0.761) 1.187 (0.783) 1.350 (0.754) aUsing responses for Day 1 only. Method 3. The activity ratios were calculated by the method described in the sixteenth report of the WHO Expert Committee on Biological Standardiza- tion (1964a), using the actual responses, not their logarithms. By this method the crossing of the treatments was ignored and ratios of activity were calculated with confidence limits based on the pooled variance within subgroups (see Table 1). Non-measurable penicillin levels were recorded by laboratories as zero responses. Since it was necessary to log the responses in order to analyse the data by Methods 1 and 2, all such values were replaced by 0.005 IU/ml for all methods of calculation. This value represents approximately the minimum measur- able level of penicillin in a sample of serum, when the recommended method of assay is used. Laboratory 6 found no measurable response, in more than half of the rabbits tested, in the second part of the test. This probably resulted from a loss of activity during storage of the serum samples before assay (48 hours at 4°C). It was decided to carry out the analysis by Methods 2 and 3 for this laboratory, using only responses to the first set of injections, since the large proportion of identical values of 0.005 would otherwise have invalidated the estimates of variance. Variances were not unduly affected in this way when Method 1 was used, because in this analysis differences between the two responses in a rabbit gave reasonable distributions, the first response being measurable in all but one instance. Data relating to the animals that died during the experiment were eliminated entirely from the analysis. COMBINATION OF ESTIMATES OF ACTIVITY RATIO Within laboratories Laboratories that used only 5 rabbits per group in the cross-over design repeated the whole test on one or more occasions. Ratios (or log ratios) derived 7 93 J. M. BOND ET AL. Table 2. Weighted mean activity ratios for laboratories that repeated the test Labor- Activity ratio at 72 hours (lower confidence limit: P = 0.99) atory rabbits Cross-over (logged) Not crossed (logged) Not crossed (unlogged) Method I Method 2 Method 3 4 20 0.973 (0.666) 0.974 (0.664) 1.021 (0.676) 12 16 1.066 (0.263) 1.061 (0.327) 1.091 (0.392) 13 20 1.049 (0.789) 1.050 (0.800) 1.062 (0.781) 14 38 1.044 (0.775) 1.045 (0.776) 1.067 (0.749) from the results of these replicate tests were weighted by the reciprocals of their variances and combined to give a weighted mean for each laboratory (Table 2). Data from these same laboratories were analysed as though they had resulted from a series of single experiments carried out as recommended (WHO Expert Committee on Biological Standardization, 1964a)-i.e., groups of rabbits tested on different occasions were divided between the two treatments and the activity ratio was estimated from the accu- mulated results for all the rabbits. The ratios obtained in this way are presented in Table 3, together with the relevant parts of Table 1, which are included for ease of comparison. Table 3. Activity ratios calculated for each laboratory by the method described in the requirements using logged and unlogged values Activity ratio at 72 hours Labor- No. of (lower confidence limit: P = 0.99) atory Strain of rabbit repnsNo. response Not crossed (logged) Not crossed (unlogged) Method 2 Method 3 I New Zealand 38 0.882 (0.598) 0.785 (0.453) 2 (not defined) 48 1.146 (0.692) 1.073 (0.560) 3 Belgian 38 1.724 (0.723) 1.744 (0.090) 4 English type 40 0.938 (0.621) 1.038 (0.686) 5 Fauves de Bourgogne 40 1.453 (0.671) 1.372 (0.779) 6 "Rustique" 20 1.824 (1.389) 1.507 (0.677) 7 New Zealand White 36 1.144 (0.792) 1.063 (0.666) 8 Californian 40 1.078 (0.974) 1.089 (0.991) 9 Male Chinchilla 20 0.922 (0.776) 0.926 (0.627) 10 (not defined) 62 0.924 (0.506) 0.894 (0.460) 11 Chinchilla 40 0.984 (0.419) 1.068 (0.528) 12 New Zealand 32 1.091 (0.328) 1.146 (0.471) 13 Mixed 40 1.048 (0.798) 1.063 (0.790) 14 Common White 76 0.981 (0.713) 1.095 (0.848) 15 Himalayan 40 1.187 (0.783) 1.350 (0.754) 94 PROCAINE BENZYLPENICILLIN Between laboratories In order to test the homogeneity of estimates made by different laboratories, x2 tests were carried out on the logarithms of the activity ratios shown in Table 1 but using the combined values for Labora- tories 4, 12, 13, and 14, reported in Table 2. The tests showed that results were consistent between laboratories and they could be combined to give a weighted mean of about 1.075 with 99% confidence limits of +7%, whether the basic analysis had been done by Method 1, 2, or 3. The actual values of the means and limits were as follows: Method 1: cross-over (logged) 1.077 (1.004-1.155) Method 2: not crossed (logged) 1.076 (1.001-1.159) Method 3: not crossed (unlogged) 1.073 (1.001-1.145) Each of these estimates was based on the results obtained from 14 laboratories since, according to the usual practice in interpreting collaborative studies of international standards or reference preparations, only comparisons in which the two preparations were compared in parallel and at the same time were included in the final value. On these grounds the results from Laboratory 9 had to be excluded. DISCUSSION The material assayed in the collaborative study appeared to be suitable to replace the International Reference Preparation when stocks of this were ex- hausted. The chosen material was shown to have blood-level duration properties comparable to those of the IRP, and the variability in blood levels produced by the two preparations was similar. Whatever method of analysis was used, the activity ratio of the PIRP 2 was about 1.075 with 99% con- fidence limits of 4 7 %. The proposed new reference preparation appeared to be of slightly higher quality than the existing international reference preparation. It was therefore suggested that the level at which material passed the blood-level duration test (i.e., the lower 99% confidence limit) should be lowered from 0.5 to 0.45 when the PIRP 2 came into use. In this way, the material currently passing the test by com- parison with the IRP should be found acceptable when compared with the PIRP 2. It was not possible to assess completely the use- fulness of a cross-over design by analysing the data in two different ways, because the groups ofresponses to the two preparations were obtained from the same rabbits and could never become independent, al- though this was assumed for the purposes of analysis by Method 2. There was, however, little difference in the variances " between rabbits " and " within rabbits" and it seemed unlikely that the use of a cross-over design would improve the precision of the test. There are certain theoretical arguments in favour of basing calculations on the logarithm of the response (blood level at 72 hours) rather than on the response itself. Ratios are expected to be dis- tributed lognormally, and use of the log response should equalize variances of the treatment groups within laboratories. In this collaborative study there were few discrepancies between these variances, and using the logarithmic values seemed to offer no advantage. However, the design of the study was not such that this question could be decided. It would have been necessary for a larger series of experiments to be carried out at each laboratory. In theory, the distribution of log activity ratios might also be expected to be normal, whereas that of the activity ratios would be skew. In practice, however, this did not seem to occur, although the amount of data was insufficient to allow a firm conclusion to be drawn. A difficulty in the use of the log transformation is the inevitability of occasional zeros among the potencies, and it is necessary to justify some criterion for dealing with these. The device used in the analysis of this study is an arbitrary one that cannot be justified on theoretical grounds. Limits to the ratio are simple to calculate when using the logarithm of the response, the variance of the activity ratio being based on the error variance and the number of rabbits in the test. The calcula- tion of fiducial limits is a lengthier procedure when the unlogged response is used, and the calculation of confidence limits, as described in the existing test, is a simpler procedure. The variance of the ratio (R) s2(I±R2)is proportional to ( 2 where S2 is the errorx2 variance and x the mean response to the reference preparation. Thus, the variance will increase if the ratio is large (which is probably unimportant as materials with a high ratio will tend to pass the test in any case) but the variance will also increase if x is small, unless s2 is correlated with x, thus implying the need for a log transformation. In fact the levels of x were fairly constant from one laboratory to another, whereas 82 varied more than 60-fold. In Tables 4 and 5, laboratories are grouped according to the error variances actually calculated from their data; theoretical lower limits of error are also given for varying sizes of the treatment 95 J. M. BOND ET AL. Table 4. Error variances obtained by different labora- tories when log responses were used, together with theoretical lower limits (P = 0.99) assuming an activity ratio of 1.0 Variances obtained Tertcllwrlmtby participating laboratories Theoretical lower limits Error variance Laboratory 5 10 15 20 a (S2) No. 5 I0 15 2 8,9 0.01 0.61 0.74 0.79 0.82 1,4,7,13,14,15 0.34 0.52 0.60 0.64 0.05 2,6 0.10 0.21 0.39 0.48 0.52 3,5,10,11 0.20 0.11 0.27 0.35 0.41 12 0.30 0.06 0.20 0.28 0.34 a Number of responses to each preparation. Table 5. Error variances obtained by different labora- tories when untransformed responses (potencies) were used, together with theoretical lower limits (P = 0.99), assuming an activity ratio of 1.0 and mean blood levels of 0.2 for each preparation Variances obtained Theoretical lower limitsby participating laboratories Tertcllwrlmt Error variance Laboratory 10 15 20 a(S2) No. 5 10 15 2 8 0.0005 0.76 0.84 0.89 0.90 9,13 0.0010 0.66 0.76 0.84 0.86 4,6 0.0050 0.25 0.50 0.64 0.69 2,5,11,12,15 0.0075 0.08 0.43 0.56 0.62 1,7,10,14 0.0100 -0.06 0.29 0.49 0.57 0.0300 -0.84 -0.23 0.12 0.25 3 a Number of responses to each preparation. groups with specific values for the variance. For these theoretical values it was assumed that the activity ratio would always be unity, and for unlogged data it was also necessary to assume a value for x, which was taken to be 0.2. The mean values of x for the IRP obtained by laboratories in the study ranged from 0.08 to 0.24. From the tables it is possible to make a rough estimate of the numbers of animals that would be required to demonstrate that a test sample of a quality comparable to that of the IRP is satisfactory. Group sizes are given for different variances that might be expected to occur in practice. In this study the use of response or log of response had no significant effect on the calculated activity ratio (1.073, using the response; 1.076, using its loga- rithm). The lower 99% confidence limit was raised by using the log transformation in 9 out of the 15 laboratories, but in only 3 of these (Laboratories 1, 3, and 10) did this mean that the PIRP 2 would have passed the blood-level duration test. Indeed, in Laboratory 11, the PIRP 2 would pass if the response were used in the calculations but would not pass if the log of the response were used. The laboratories taking part in the collaborative assay agreed that the PIRP 2 was suitable material to replace the IRP. They also agreed with the con- clusion that, in relation to the former, the level at which material passes the blood-level duration test- i.e., the 99% lower confidence limit of the ratio- should be changed from 0.5 to 0.45. Therefore, as authorized by the WHO Expert Committee on Biological Standardization (1966), the material examined in the collaborative study was established as the second International Reference Preparation of Procaine Benzylpenicillin in Oil with Aluminium Monostearate. In its nineteenth report, the WHO Expert Committee on Biological Stan- dardization (1967) adopted the modifications to the requirements for PBAM, rendered necessary by the establishment of the PIRP 2. These modifications, which had been prepared by the WHO Secretariat in collaboration with various experts, included the change from 0.5 to 0.45, which was suggested by the collaborative study as the minimum criterion for compliance with the blood-level duration test. The WHO Expert Committee on Biological Stan- dardization (1971) noted the objections that had been raised, on mathematical and statistical grounds, in regard to the recommended method of evaluating the results of the blood-level duration test (WHO Expert Committee on Biological Standardization, 1964a). The Committee agreed that it was preferable to allow the use of any suitable statistical analysis that permitted the calculation of the appropriate 99% confidence limit. 96 PROCAINE BENZYLPENICILLIN 97 RESUME DEUXIPME PREPARATION INTERNATIONALE DE REFERENCE DE BENZYLPENICILLINE PROCATNME, DANS L'HUILE, ADDITIONNEE DE MONOSTEARATE D'ALUMINIUM La duree de la p6nicillin6mie r6sultant de l'injection de benzylp6nicilline procaln6e, dans l'huile, additionn6e de monost6arate d'aluminium (PAM) est tres variable suivant l'origine des pr6parations, alors meme que des doses identiques ont 6t6 inject6es. On ne peut pr6voir ces variations par l'etude in vitro des proprietes d'un echantillon donn6 et il est indispensable de recourir a des comparaisons in vivo dans lesquelles 1'echantillon a 6valuer est compare a une preparation de ref6rence dont on connait l'activit6 satisfaisante chez l'homme. Les stocks de la pr6paration internationale de r6f6rence de PAM utilis6e a cet effet 6tant pratiquement 6puises, on a constitu6 une deuxieme pr6paration internationale de r6f6rence sur la base des resultats d'un titrage col- lectif. Les premi6re et deuxi6me pr6parations de r6f6- rence ont ete compar6es dans 15 laboratoires de 9 pays employant la technique d'6preuve enonc6e dans les Normes OMS pour les substances biologiques No 9. La deuxieme pr6paration internationale de reference s'6tant r6v6l6e d'une qualit6 l6gerement superieure a celle de la premiere, il a fallu modifier la limite inferieure de conflance A 99% du rapport entre les penicillin6mies moyennes A laquelle un lot satisfait A l'epreuve de dur6e de la p6nicillinemie et l'abaisser de 0,5 A 0,45. Cette modification a ete accept6e par les participants au titrage comparatif et approuv6e par le Comit6 OMS d'experts de la Standardisation biologique (1967). Le Comit6 (1971) a d6cid6 d'autoriser pour le calcul de cette limite inf6rieure l'utilisation de toute methode d'analyse sta- tistique appropriee. REFERENCES Bond, J. M. et al. (1965) Bull. Wid Hlth Org., 33, 817-836 WHO Expert Committee for Biological Standardization (1964a) Wld Hlth Org. techn. Rep. Ser., No. 274, Anhex 2 WHO Expert Committee for Biological Standardization (1964b) Wld Hlth Org. techn. Rep. Ser., No. 293, p. 11 WHO Expert Committee for Biological Standardization (1966) Wld Hlth Org. techn. Rep. Ser., No. 329, p. 7 WHO Expert Committee for Biological Standardization (1967) Wld Hlth Org. techn. Rep. Ser., No. 361, Annex 3 WHO Expert Committee for Biological Standardization (1971) Wld Hlth Org. techn. Rep. Ser., No. 463, p. 24 Annex 1 PARTICIPATING LABORATORIES Commonwealth Serum Laboratories Parkville, Victoria, Australia (Dr E. H. Baxter) Drug Control Department Belgian Pharmaceutical Association Brussels, Belgium (Dr J. Dony & Mr de Roeck) Department of Antibiotics State Serum Institute Copenhagen, Denmark (Dr J. Bang & Dr Per Hye-Knudsen) Glaxo Laboratories Ltd Greenford, Middlesex, England (Mr D. C. M. Adamson, Dr P. W. Muggleton, & Dr M. L. Hilton) Division of Biological Standards National Institute for Medical Research London, England (Mr J. W. Lightbown, Dr J. M. Bond, & Mr D. H. Hewgill) Microbiology Section National Public Health Laboratory Paris, France (Professor J. Desbordes & Miss Chaniot) Institut Alfred Fournier Paris, France (Dr P. Durel, Dr D. Videau, & Dr (Mrs) V. Roiron) Laboratory of Hygiene Faculty of Medicine Lyons, France (Professor J. Thivolet) J. M. BOND ET AL. Quality Control Laboratory Hindustan Antibiotics Ltd Poona, India (Dr S. R. Sarvotham) Department of Antibiotics National Institute of Health Tokyo, Japan (Dr H. Umezawa) Wyeth Laboratories Inc. West Chester, Pa., USA (Dr W. F. Elias) The Upjohn Company Kalamazoo, Mich., USA (Mr C. L. Graham) Food & Drug Administration Department of Health, Education, & Welfare Washington, D.C., USA (Dr W. W. Wright) Department of Microbiology Central Institute of the Ministry of Health of the USSR for Venereal & Skin Diseases Moscow, USSR (Dr H. M. Ov6innikov) Laboratory of Microbiology The Leningrad Institute for Research on Antibiotics Leningrad, USSR (Dr A. N. Klimov) Annex 2 DETAILS OF THE COMPARISON OF THE INTERNATIONAL REFERENCE PREPARATION AND THE PROPOSED SECOND INTERNATIONAL REFERENCE PREPARATION TO BE MADE IN EACH LABORATORY The test will be carried out as laid down in the WHO Requirements for Biological Substances No. 9, Part A, section 5.3.2., with the following qualifications: 1. At least 20 rabbits should be used for the compari- son, each being used twice in a cross-over design, e.g., Day 1 Day 2 Day 1+21 Day 2+21 IRP Rabbits 1-5 Rabbits 11-15 Rabbits 6-10 Rabbits 16-20 PIRP 2 Rabbits 6-10 Rabbits 16-20 Rabbits 1-5 Rabbits 11-15 or IRP PIRP 2 Day 1 Rabbits 1-10 Rabbits 11-20 Day 1+21 Rabbits 11-20 Rabbits 1-10 2. For the purpose of this experimental study, it is important that the material in the vials should be effi- ciently mixed. It is therefore recommended that the vials be shaken vigorously the day before use, if possible mechanically. Immediately before injection, the vials should be shaken vigorously by hand, but care must be taken to ensure that any air bubbles produced during shaking are allowed to dissipate. 98