PRODUCTION OF ANTIPNEUMOCOCCIC SERUM 573 



three animals were used for each amount of serum or culture, 

 chance death or survival caused confusion. Trevan (1930) 1418 

 found that the number of animals necessary to offset a given mar- 

 gin of error depends on the ratio of the increment of effect of the 

 serum to the increment of dose of culture. As an example, it was 

 found that if forty mice per dose are used to test two serums, the 

 first may be estimated as equal in potency to the other, although 

 in reality the potency of the first may be as low as 72 per cent or 

 as high as 140 per cent of the second. Trevan suggested computing 

 the survival time of test mice as a means of increasing the accu- 

 racy of results and believed that by this plan not more than one- 

 half the number of mice would be required for the same determina- 

 tion as would be necessary if simple enumeration of the number 

 dead on a given day were made. It was found, however, that this 

 plan might be unduly favorable to the serum, since the distribution 

 of survival times was not normal. Discrepancies in the results of 

 tests were also investigated by Beard and Clapp (1932), 95 who 

 recommended the use of at least ten mice per dose of serum and at 

 least two independent tests on each lot of serum. Results of differ- 

 ent tests were considered together, rather than separately. Barnes, 

 Clarke, and Wight (1936), 82 using the method described in the 

 Appendix, found that when two or more separate tests, each of 

 which included thirty mice per sample of serum, were used in esti- 

 mating the value of a serum as compared with a standard, it was 

 practicable to arrange the amounts of serum in the ratio 4 to 2 

 tol. 



Calculations of the potency of serum were based on the follow- 

 ing equation : 



Units per cubic centimeter in unknown serum = a 



m 



a = units per cubic centimeter in control serum 

 D = total number of mice dead on control serum 

 x = lowest dilution of unknown serum 

 D' =z total number of mice dead on unknown serum 

 y = lowest dilution of control serum 



