In a previous communication (I) the use of the APL/360 system for fitting exponential curves was described. The convenience of this terminal-operated system has prompted investigation of its use in other computational problems in pharmacology, and in this communication probit analysis routines are described.
The problems involved in providing some reproducible measurements of toxicity arise from the sigmoid nature of the curve of percentage mortality against the logarithm of the dose of toxic agent (2). It is usual to express drug toxicities in terms of the LD,,, defined as the dose which will kill half a given population of animals within a certain specified time. The LDSO is thus the dose corresponding to 50% mortality, but the nonlinear relationship between dose and toxicity renders this value difficult to estimate with any accuracy, and does not permit any estimate of confidence limits. It was shown by Gaddum (3) that if the normal equivalent deviation is plotted against the logarithm of the dose, the relationship is linear and hence the LDso and the confidence limits may be obtained by linear regression techniques. Bliss (4, 5) suggested that (normal equivalent deviation + 5) be employed as the ordinate, thus avoiding the use of negative numbers. These units are referred to as probits, and the process of transforming the experimental data to probits and the ensuing regression is referred to as probit analysis. The procedure for this analysis was devised before the advent of computing machinery of any sophistication, and even the treatise by Finney (6) arranges the calculations for manual analytical techniques. The computation involved is extremely tedious, and this has severely limited the application of the method; however, the use of computers can facilitate the analysis to the point where it may be routinely employed. Since the process involves extensive manipulation of limited data, it is ideally suited to a terminal-operated system, and owing to the power and convenience of APL/360, this language has been employed for the probit analysis reported here.