Analysis of worm burdens in experimental schistosomiasis.

The techniques of generalized linear modelling were applied to worm burdens of 2268 mice artificially infected with Schistosoma mansoni and treated with oxamniquine and/or praziquantel. Models which maximized the Poisson likelihood were shown to be superior to those assuming either a log Normal or negative binomial error structure. Use of the models to calculate and compare ED values of 8 strains of S. mansoni, detect abnormal dose response and synergism and to pool data from several experiments is illustrated.