Quantification of arrhythmias using scoring systems: an examination of seven scores in an in vivo model of regional myocardial ischaemia

Arrhythmia scores have been used in recent years to facilitate the analysis of arrhythmias, particularly in relation to regional myocardial ischaemia. The recent Lambeth Conventions recommended caution in the use of arrhythmia scores since their use may be misleading. In the present study seven scoring systems were examined in an attempt to validate the use of arrhythmia scores. A strong positive correlation was present between all seven scores. Furthermore, the scores all correlated with the incidences of ventricular fibrillation, ventricular tachycardia, and ventricular premature beats in early myocardial ischaemia. All seven scores successfully detected statistically significant reductions in the incidence of ventricular fibrillation resulting from the administration of two drugs. Some of the scores occasionally showed statistically significant reductions when effects on the raw arrhythmia data were not statistically significant. In this respect, parametric statistical analysis of arrhythmia scores may be a more sensitive method of quantifying arrhythmias than non-parametric analysis of binomially distributed raw data such as the incidence of ventricular fibrillation (in accordance with the power of such tests) indicating that the scores have precision. However, none of the scores incorrectly showed a statistically significant reduction when the raw data expressed a statistically significant or non-significant increase, indicating that the scores have accuracy. In conclusion, it is possible to design many arrhythmia scores that show changes in arrhythmia severity when more conventional analyses show only non-statistically significant trends. When used in conjunction with raw arrhythmia data, comprehensive drug dose ranges, and appropriate parametric statistical tests, arrhythmia scores facilitate the quantification of arrhythmias. It is recommended that arrhythmia scores should be used only for quantifying group data and model building and not for prognostic purposes in individuals.