Specificity of the moving average method for detecting alternans

The moving average method of detecting alternans, as originally described by Nearing and Verrier, does not address several factors that may impact its specificity. Omitted factors include the influences of RR interval irregularity and the harmonics of 4:1 processes like respiration. Also missing is a test for statistical significance. The relevance of these factors has been tested in a database of 383 recordings from a large trial on chronic heart failure patients (the GISSI-HF study). This database consists of 12-lead 24-hour Holter recordings. Non-overlapping 128-beat segments were analyzed to obtain the average odd minus even beat difference of normal beat pairs. The database provided 284,755 such segments, each with at least 36 odd/even pairs of normal beats. Only lead V5 was evaluated for alternans. Of these, 27,932 distinct 128-beat segments exhibited an alternans amplitude > 20 microvolts in at least one 20ms part of the average STT window odd/even differences. After applying tests for RR irregularity/instability, statistical significance, subharmonics, RR alternans, and using wider windows, only 268 events remained. These data suggest that the moving average method previously described exhibits poor specificity without application of additional criteria.