An analysis of the predictive probability of the in vitro contracture test for determining susceptibility to malignant hyperthermia.

An objective estimate of the likelihood of correct designation of malignant hyperthermia (MH) susceptibility from in vitro contracture test (IVCT) results is essential if genetic linkage studies of MH are to be more informative. The aim of this study was to generate and test statistical models that could be used to predict the probability of susceptibility of an individual to MH from the results of their IVCTs. Logistic regression of the IVCT results of an index group of 50 patients (age range 9-73 years; MH susceptible [MHS], n = 13; MH normal [MHN], n = 32; MH equivocal [MHE], n = 5) who were either at low risk of MH or were proband cases were used to generate models to predict probability of MH susceptibility. Models incorporated data from individual contracture tests or from combinations of tests (static halothane, dynamic halothane, caffeine, ryanodine) performed according to the protocols of the European Malignant Hyperthermia Group. Of the individual contracture tests, the ryanodine test was most closely correlated with MH status. Discriminatory ability of the models was assessed using receiver operating characteristic (ROC) curves. Inclusion of predictor variables from the ryanodine, caffeine, and dynamic halothane tests improved upon the discriminatory ability of the models incorporating variables from individuals tests and was considered to be the best model. The reproducibility of this model was confirmed using an ROC curve constructed using data from 47 patients (age range 10-62 years; MHS, n = 15; MHN, n = 28; MHE, n = 4) who were classified in a way similar to the index group. A further group of 153 patients (age range 9-74 years; MHS, n = 44; MHN, n = 92; MHE, n = 17) who were consecutively tested relatives of susceptible individuals was used to assess the generalizability of the best model. The model met the criteria for a useful discriminatory model with this group of patients, 125 of whom (including 9 MHE patients) could be designated as positive or negative for MH with a likelihood of more than 95%. The logistic regression models provide objective likelihoods for the MH phenotype that could be usefully incorporated into genetic linkage studies of the condition.