The international reference preparation of colistin methane sulfonate*

Bull. Org. mond. Sant} 1973, 48, 75-80 Bull. Wid Hith Org.J The international reference preparation of colistin methane sulfonate* J. W. LIGHTBOWN,1 JILLIAN M. BOND,' & B. GRAB 2 The international unit of colistin methane sulfonate has been defined by collaborative assay as the activity contained in 0.00007874 mg of the international reference preparation. The definition was basedon resultsfrom 7 laboratories in 5 countries which carried out assays against their existing national standards. Because of the complex heterogeneity in composi- tion of this antibiotic the international reference preparation was not designated as an international standard. The WHO Expert Committee on Biological Stan- dardization (1964) decided that there was a need for an international reference preparation of colistin methane sulfonate and requested the National Insti- tute for Medical Research, London, to obtain a suit- able sample and make a preliminary evaluation of it. A small sample of colistin methane sulfonate was obtained through the good offices of Dr Nakamura, Director of the National Institute of Health, Tokyo, Japan, for consideration as the proposed international reference preparation (PIRP). It was distributed to the three national control laboratories holding na- tional standards of this material, and assayed by each of them against their national standard. The poten- cies obtained by the different laboratories for the proposed international standard were in fairly close agreement, i.e., within 5%. However, it had been reported that colistin methane sulfonates of different origins differed in the degree of substitution of the five free amino groups (Barnett et al., 1964). This sug- gested that separate standards might be necessary for these variants. The differently substituted materials can be separated by electrophoresis; therefore com- parisons of a number of commercial batches were carried out. Nine samples of colistin methane sulfo- nate, five of French origin and four of Japanese origin, were compared with the proposed interna- * No International Nonproprietary Name has been recommended for this material. In the United Kingdom it is known as Colistin Sulfomethate and in the USA as Colistomethate. 1 WHO International Laboratory for Biological Stan- dards, National Institute for Medical Research, Mill Hill, London, England. 2 Health Statistical Methodology, World Health Organi- zation, Geneva, Switzerland. tional reference preparation by electrophoresis in agar. The PIRP showed no differences in composition from the Japanese batches but minor differences from the French batches, there being an additional minor component present in the French material. No marked differences were observed between batches from the same country and the Expert Committee on Biological Standardization (1966) concluded that only one international reference preparation would be required for colistin methane sulfonate. It re- quested the National Institute for Medical Research to obtain a sufficient quantity of the material origi- nally examined, to distribute this in ampoules, to establish the material as the International Reference Preparation, and to arrange a collaborative assay. The material could then be established as an international standard with a defined international unit of activity. A sufficient quantity of this material was therefore obtained.3 The sample was received at the National Institute for Medical Research in December 1965 and stored at -10°C until July 1966 when it was distrib- uted in approximately 75-mg quantities into neutral glass ampoules. The ampoules and contents were dried in vacuum over phosphorus(V) oxide to con- stant weight, and the ampoules were filled with pure dry nitrogen and sealed. After testing for leaks, the ampoules were stored at - 20°C. The residual moisture content of the material in the 3The material was supplied by Dr Yasuo Koyama, Director, Kayaku Antibiotics Research Co., Ltd., 1546-2 Chome, Mejiro-machi, Toshima-ku, Tokyo, Japan, through the good offices of Dr Umezawa. The following data were supplied by the manufacturers-batch: lot No. MCB-2169; potency: 12 000 units/mg; pyrogen: pass; toxicity: pass; pH of aqueous solution (25°C): 8.55; and loss in weight at 60°C for 3 hours at a pressure of 5 mmHg or less: 2.8 %. 2981 75- J. W. LIGHTBOWN ET AL. ampoules, measured as loss in weight at 56°C at a pressure of <0.02 mmHg over phosphorus(V) oxide for 5 hours, was found to be 0.44%. When exposed to an atmosphere of 45% relative humidity, the material in the ampoules increased in weight approximately 0.21 % w/w in 5 minutes and 1.67% w/w in 60 mi- nutes. THE COLLABORATIVE ASSAY Seven laboratories in five countries agreed to parti- cipate in the collaborative study, and they were asked to assay the proposed international standard against their own national standard. The participants were requested to provide evidence of linearity of the log dose-response lines for their assay conditions; each assay was to contain sufficient information to pro- vide, from its own internal evidence, an estimate of potency and the fiducial limits to that estimate. The names of the participants are given in Annex 1, but throughout the report they are referred to by a number that has no connexion with the order in which they are listed in the Annex. Methods of statistical analysis The seven collaborating laboratories performed a total of 148 assays, all by the diffusion method. Laboratory 5 provided results performed by three sublaboratories and these were analysed separately. Each laboratory compared the proposed interna- tional standard (PIS) of colistin methane sulfonate with own national standard (NS). The potencies of the national standards used were reported by the participating laboratories as follows: Canadian national standard, 11 300 units/mg; Japan- ese national standard (2221), 12 800 units/mg; Japan- ese working standard (MCB 2446), 12 900 units/mg; Table 1. Details of assay methods used in different laboratories Dose levels of each No.Labora- Test organism a Type of assay preparation Structure of assay of assaystory No. (units/ml) prformed 1 Bordetella diffusion on 100; 150; 225 3 + 3: one of each concentration of each 11 bronchiseptica Petri dishes; preparation on a dish; 9 dishes per ATCC 4617 cylinders assay 2 Escherichia coli diffusion on Approx: 5 000; 20 000 2 + 2: one of each concentration of each 3 (NIHJ) Petri dishes preparation on a dish; 5 dishes per assay 3 Escherichia coli diffusion on 5 000; 10 000 2 + 2: one of each concentration of each 16 (NIHJ) Petri dishes preparation on a dish; 5 dishes per assay 4 Escherichia coil diffusion on 1 250; 2 500; 5 000 3 + 3: one of each concentration of each 17(95 ISM) Petri dishes; preparation on a dish; 9, 10, or 11 punched holes dishes per assay 5a Escherichia coil 1 250; 2 500; 5 000 14 (95 ISM) 5b Bordetella 25; 50; 100 3 + 3: one of each concentration of each 17 bronchiseptica diffusion on preparation on a dish; 9 dishes per(ATCC 4617) Petri dishes; assay b punched holes Escherichia coil 1 250; 2500; 5 000 J18 (95 ISM) 5c Escherichia coli 2 500; 5 000; 10 000 3 + 3: one of each concentration of each 24 (95 ISM) preparation on a dish; 9 dishes per assay 6 Bordetella 625; 2 500 8 bronchiseptica 2 + 2: 16 replicates of each concentration NCTC 8344 diffusion on of each preparation arranged in an 8 x 8 Escherichia coli l 2 500; 10000 pLatin square design; 1 plate per assay 8 (95 ISM) 7 Bordetella diffusion on 3.75; 7.5; 15 3 + 3: 2 replicates of 1 concentration of 12 bronchiseptica Petri dishes; each preparation on each of 4 dishes; (ATCC 4617) cylinders 12 dishes per assay a B. bronchiseptica NCTC 8344 is identical with ATCC 4617. b Except for one assay in Sublaboratory 5b with E. coli in which only 4 dishes were used. 76 COLISTIN METHAN SULFONATE French national standard, 12 500 units/mg; British national standard, 12 500 units/mg; and United States of America FDA standard, 516 "pg "/mg. The French and British standards were prepared from a single batch of material. The potency of the FDA standard was not expressed in terms ofunits/mg but in terms of ",ug "/mg; 1 "M,g " of activity was reported by the FDA as being equivalent to 30 units of activity. The potency of the FDA standard was therefore taken as 15 480 units/mg. Table 1 shows the number of assays performed and summarizes particulars of the methods used in each laboratory. The statistical analyses were carried out in the same way as for the evaluation of the collaborative assay of the International Reference Preparation of Colistin (Lightbown et al., 1973). RESULTS Ofthe 148 assaysagainst standardpreparationsonly 2 were rejected; one, from Sublaboratory 5c, gave a potency that was completely aberrant (most probably because of a weighing error), while the other, from Laboratory 4, was grossly nonlinear and was largely responsible for the intralaboratory heterogeneity of the assays of that laboratory. The lowest potency, 11 354 units/mg, was obtained by Laboratory 6 and the highest potency, 15 552 units/mg, by Sublabora- tory 5a. The number of assays from the different laborato- ries that showed a significant departure from parallel- ism was small and lack of parallelism did not seem to be attributable to the use of one particular micro- organism. Of 112 three-point assays, 23 were non- linear; there was, however, no significant difference between the numbers of convex and concave depar- tures from linearity among these assays nor was lack of linearity associated with the use of a particular test organism. Only one assay demonstrated a significant sum of squares for " difference of quadratics ". In this type of collaborative study, involving assays of high precision, small degrees of nonparallelism and/or curvature-as a result of exceptionally low error-variance-may occur and produce formal sta- tistical invalidity in a proportion of the assays. Under these conditions the treatment of invalid assays will depend largely on the results of a study of the homogeneity of their estimated potencies. The results of assays performed in Laboratories 1, 2, 3, and 7 were homogeneous within each laboratory. For all these laboratories the average log potency Table 2. Average by potencies for the laboratories with homogeneous results Labora- No. Weighted geometric tory No. Microorganism of assays average potency(units/mg) 1 B. bronchiseptlca 11 13 529 2 E. co/i 3 12496 3 E. coli 16 12198 7 B. bronchiseptica 12 12 734 shown in Table 2 was calculated by weighting each log potency value with the reciprocal variance based on the internal evidence of the assay. Significant heterogeneity at the 1% probability level was observed among individual potencies ob- tained by Laboratories 4, 5 (each of the sublabora- tories a, b, and c), and 6. An attempt was made to identify the sources of heterogeneity within the assays of these laboratories by examination of the X2 components for different sources of variation. As mentioned previously, one nonlinear assay from Laboratory 4 was excluded on the basis of this analysis, its exclusion decreasing the x2 value for the collected potencies of this labora- tory by more than half from 108.86 for 16 degrees of freedom to 50.39 for 15 degrees of freedom. In the two laboratories that performed assays with two test organisms, i.e., 5 b and 6, the variation in potency resulting from the use of the different organisms was significant; other sources of variation were day, weighings, and, in some instances, operators. Com- parison of the individual potency estimates obtained by the two laboratories that used two different micro- organisms shows that the lower potencies were gen- erally obtained when Escherichia coli was used. For Laboratories 4, 5 (a, b, and c), and 6, with heterogeneous results, a new series of weights based on both within- and between-assay variances were calculated by the method described by Bliss (1952). The laboratory average log potencies were then com- puted by applying the adjusted weights to the indi- vidual potencies. Table 3 gives the resulting average log potency for each laboratory and separately by microorganism used. From the last two columns the importance of the decrease from original weights to new weights can be appreciated. The final laboratory results are summarized in 6 77 J. W. LIGHTBOWN ET AL. Table 3. Weighted average log potencies for the laboratories with heterogeneous results Weighted geometric average potency a TtlwihLab. (units/mg) Total weight No. E. coil B. bronchiseptica All assays Original Adjusted 4 b 12 935 (16) - 12 935 (16) 421 391 98 784 5a 13 246 (14) - 13 246 (14) 222 572 17 895 5b 13 134 (18) 13 577 (17) 13 315 (35) 254 736 98 787 5c b 12 454 (23) - 12 454 (23) 1 574 209 18 310 6 11 777 (8) 12 208 (8) 12 003 (16) 311 940 92 272 a Calculated with the adjusted weights. The number of assays is given in brackets. b Excluding one invalid assay. Table 4, which shows that the various average po- tency values did not differ considerably from one laboratory to another, the range extending from 12 003 units/mg in Laboratory 6 to 13 529 units/mg in Laboratory 1. However, the X2 for between-laboratory homogeneity was much greater than the critical value at the 1% probability level for 8 degrees of freedom. Column 6 of Table 4 shows that this situation is to be expected since 2/3 of the total weight was provided by the two laboratories with the lowest average potencies (Laboratories 6 and 3). In order to take into account the additional vari- ance component originating from the between-lab- oratory variability, the last two columns of Table 4 were calculated giving the total variance and the corresponding adjusted weight for each laboratory. The combined weighted log potency estimate was computed by applying to each laboratory log potency Table 4. Summary of the results of assays of the proposed international standard for colistin methane sulfonate Weighted Total Labor- No. of Homogeneity geometric Weighted Total within- Within-laboratory Total laboratory adjusted atory assays within average average log laboratory variance of average variance a laboratorylaboratory potency potency weight log potency weight(units/mg)(1) (2) (3) (4) (5) (6) (7) (8) (9) 1 11 homogeneous 13 529 4.13126 26 972 0.000 037 075 0.000 324 992 3 077 2 3 homogeneous 12486 4.09677 8 915 0.000 112 170 0.000 400 087 2 499 3 16 homogeneous 12 198 4.08628 723 097 0.000 001 383 0.000 289 300 3457 4 16 heterogeneous 12 935 4.11175 b 98 784 c 0.000 010 123 0.000 298 040 3 355 5a 14 heterogeneous 13 246 4.12209 b 17 895 c 0.000 055 882 0.000 343 799 2 909 5b 35 heterogeneous 13 315 4.12433 b 98 787 c 0.000 010 123 0.000 298 040 3 355 5c 23 heterogeneous 12 454 4.09532 b 18 310 c 0.000 054 615 0.000 342 532 2 919 6 16 heterogeneous 12 003 4.07928 b 92 272 c 0.000 010 838 0.000 298 755 3 347 7 12 homogeneous 12 734 4.10497 150 041 0.000 006 665 0.000 294 582 3 395 total 146 d 1 235 073 28 313 a Including between-laboratory variance (0.000 287 917). b Weighted with weights adjusted for within-laboratory heterogeneity. c Total of adjusted weights. d One invalid assay was excluded from the results of Laboratory 4 and one from those of Sublaboratory 5c. 78 COLISTIN METHANE SULFONATE Table 5. General unweighted geometric mean potencies Unweighted 95 % confidence Coverage ofNo. geometric mean intervalofassays potency (nt/g(units/mg) (nt/g all assays a 146 12 828 12 696-12 961 valid assays 115 12 689 12 557-12 822 a Excluding one invalid assay from Laboratory 4 and one from Sublaboratory 5c. given in column 5 of Table 4 the corresponding weight shown in column 9. The variance of this general average log potency was estimated as the reciprocal value of the sum of the total adjusted laboratory weights (1/28 313 = 0.000 035 319). The resulting combined weighted mean potency of the proposed international standard for colistin methane sulfonate was 12 754 units/mg with the 95% confidence interval at 12 358-13 163 units/mg. The general unweighted geometric mean potency was calculated separately (a) for all the 146 assays and (b) for the valid assays only (i.e., excluding all assays with statistically significant departure from parallelism and/or linearity) and the results are com- pared in Table 5. DISCUSSION The proposed international standard for colistin methane sulfonate has been compared with five na- tional standards and one working standard for this antibiotic. Preliminary studies had indicated that the three national standards then existing were defining very similar units of activity and this has been confirmed in this more complete study. Colistin methane sulfonate is perhaps the most complex antibiotic mixture for which a biological standard has had to be established. It is possible by electrophoresis in agar to show the presence ofat least 15-20 biologically active components when the mate- rial is in solution. In spite of this complex heterogene- ity, most assays performed in the different laborato- ries were statistically valid, only 10 of 147 assays showing nonparallelism, hardly more than would be expected at the 5% probability level. As mentioned previously, colistin methane sulfo- nate is derived from two main sources of manufac- ture, the two types of material being very similar in composition but with slight differences. It is of inter- est that in this study significant heterogeneity of potency estimates only occurred within a laboratory when the source of manufacture of the national standard was different from that of the PIS, i.e., Laboratories 4, 5 (a, b, c), and 6; however, Labora- tory 7 used a national standard derived from the same source as these laboratories and obtained a homogeneous set of potency estimates. The influence of a particular test organism on the potency estimate is seen in the results of the only two laboratories that made parallel assays using both E. coli and Bordetella bronchiseptica. Both these lab- oratories used the same material as national stan- dard, and both laboratories obtained significantly lower potencies for the proposed international stan- dard when using E. coli as test organism than when using B. bronchiseptica. Beyond emphasizing that the potency obtained in biological assays of sam- ples of colistin methane sulfonate can be expected to vary with the test organism used, the observation has no special significance. It cannot be predicted that any given sample assayed against the proposed inter- national standard will give a higher potency with E. coli than with B. bronchiseptica or vice versa, only that it is possible that different potencies will be obtained. The values obtained for the combined mean po- tency of the proposed standard, in terms of those national standards that are comparable, agree closely using all methods of combination, i.e., with or with- out weighting, including or omitting invalid assays. A value of 12 700 units/mg, which is included in the confidence interval of the combined mean potencies calculated by all methods, was accepted by the parti- cipating laboratories as suitable for defining the international unit, and, in accordance with the author- ization of the WHO Expert Committee on Biological Standardization (1967), the international unit for colistin methane sulfonate was defined as the activity contained in 0.00007874 mg of the International Reference Preparation of Colistin Methane Sulfo- nate. The WHO Expert Committee on Biological Standardization (1969) noted this definition of the international unit for colistin methane sulfonate but did not establish the international reference prepara- tion as an international standard because of the complex heterogeneity of the material. 79 80 J. W. LIGHTBOWN ET AL. RFtSUM2E PRE~PARATION INTERNATIONALE DE REFERENCE DE METHANO-SULFONATE DE COLISTINE On a 6tabli une pr6paration internationale de r6ference de m6thano-sulfonate de colistine a partir d'une pr6pa- ration qui, lors d'un examen preliminaire dans trois labo- ratoires nationaux de controle, s'6tait revel6e propre a cet usage. Bien que des 6chantillons de methano-sulfonate de colistine puissent contenir jusqu'a 20 composants diff6rents bielogiquement actifs, la composition du mate- riel choisi ne s'6cartait guere de celle d'autres preparations similaires. Apr6s avoir ete r6parti en ampoules au National Institute for Medical Research, de Londres, le mat6riel a ete distribu6 a 7 laboratoires de 5 pays pour y etre 6tudie par rapport aux preparations-etalons nationales. Au cours de ce titrage comparatif, on a pratique 148 6preuves par diffusion, en utilisant comme micro-orga- nismes d'epreuve Escherichia coli et Bordetella bronchi- septica, dont 10 seulement ont donne des resultats nota- blement divergents. Une h6t6rog6neite des estimations de l'activit6 a e constat6e lorsqu'un laboratoire a utilis6 un etalon national provenant d'une autre source que le mat6riel propos6. Apres analyse des resultats, les participants au titrage comparatif sont convenus d'attribuer A la pr6paration internationale de r6ference de m6thano-sulfonate de colis- tine une activit6 de 12 700 unites par milligramme. Conform6ment A l'autorisation accordee par le Comit6 OMS d'experts de la Standardisation biologique (1967), l'unite internationale de m6thano-sulfonate de colistine a 'te definie comme l'activite de 0,00007874 mg de la pr6paration internationale de ref6rence. Le Comit6 OMS d'experts de la Standardisation biologique (1970) a pris note de cette definition, mais n'a pas constitu6 la pr6pa- ration internationale de reference en 6talon international en raison de la composition tr6s h6terogene de l'anti- biotique. REFERENCES Bliss, C. I. (1952) The statistics of bioassay with special reference to vitamins, New York, Academic Press, pp. 580-582 Barmett, M. et al. (1964) Brit. J. Pharmacol., 23, 552 Humphrey, J. H. et al. (1953) Bull. Wld Hlth Org., 9, 15 Lightbown, J. W. et al. (1973) Bull. Wld Hlth Org., 48, 65 WHO Expert Committee on Biological Standardization (1964) Wld Hlth Org. techn. Rep. Ser., No. 274, p. 8 WHO Expert Committee on Biological Standardization (1966) Wld Hlth Org. techn. Rep. Ser., No. 329, p. 7 WHO Expert Committee on Biological Standardization (1967) Wld Hlth Org. techn. Rep. Ser., No. 361, p. 11 WHO Expert Committee on Biological Standardization (1969) Wld Hlth Org. techn. Rep. Ser., No. 413, p. 11 Annex PARTICIPATING LABORATORIES Laboratory of Hygiene Department of Health and Welfare Ottawa, Ontario, Canada (Dr L. Greenberg & Miss K. Fitzpatrick) National Public Health Laboratory Paris, France (Professor J. Desbordes) Laboratoire Roger Bellon Neuilly-sur-Seine, France (Dr P. de Lajudie) Kayaku Antibiotics Co., Ltd Mejiro-machi, Toshima-ku Tokyo, Japan (Dr Y. Koyama) National Institute of Health Tokyo, Japan (Dr H. Umezawa) Division of Biological Standards National Institute for Medical Research London, England (Mr J. W. Lightbown & Dr J. M. Bond) Food and Drug Administration Department of Health, Education, and Welfare Washington, D.C., USA (Dr W. W. Wright)