STUDY OBJECTIVE: To determine if the uncorrected QT interval (QT(u)) more accurately predicts drug-induced torsade de pointes than QT intervals corrected using the Bazett's (QT(B)), Fridericia (QT(Frid)), or Framingham (QT(Fram)) methods. DESIGN: Retrospective analysis of a previously reported case-control study of risk factors for haloperidol-induced torsade de pointes. SETTING: Large tertiary care teaching hospital. PATIENTS: Forty-six critically ill patients who received intravenous haloperidol for sedation; seven developed torsade de pointes. MEASUREMENTS AND MAIN RESULTS: The QT intervals were measured manually by one investigator from electrocardiograms performed before and during haloperidol therapy. Logistic regression analysis for prediction of torsade de pointes was performed, incorporating QT(u), QT(B), QT(Frid), QT(Fram), and RR intervals measured during treatment. Receiver operating characteristics (ROC) curves were constructed. Primary outcome measures were proportion of explained variation (maximum-rescaled R2); area under the ROC curves for QT(u), QT(B), QT(Frid), QT(Fram), and RR interval; and sensitivity and specificity for prediction of haloperidol-induced torsade de pointes. The QT(u) was associated with the highest R2 compared with QT(Fram), QT(Frid), QT(B), and RR interval (0.77, 0.73, 0.68, 0.53, and 0.30, respectively). No significant differences in areas under the ROC curves were found between any of the QT-interval methods. Areas under the ROC curves for QT(u) and QT(Fram) trended toward being greater than that associated with the RR interval. All QT-interval methods were highly sensitive (100% for each), whereas the RR interval was less sensitive (86%); QT(u) and QT(Fram) were most specific (82%) compared with the QT(Frid) (72%), QT(B) (64%), and RR interval (36%). CONCLUSION: Compared with QT(B) and QT(Frid), the QT(u) and QT(Fram) best predicted haloperidol-induced torsade de pointes in critically ill patients; the QT(Fram) offered no advantage over the QT(u).
STUDY OBJECTIVE: To determine if the uncorrected QT interval (QT(u)) more accurately predicts drug-induced torsade de pointes than QT intervals corrected using the Bazett's (QT(B)), Fridericia (QT(Frid)), or Framingham (QT(Fram)) methods. DESIGN: Retrospective analysis of a previously reported case-control study of risk factors for haloperidol-induced torsade de pointes. SETTING: Large tertiary care teaching hospital. PATIENTS: Forty-six critically illpatients who received intravenous haloperidol for sedation; seven developed torsade de pointes. MEASUREMENTS AND MAIN RESULTS: The QT intervals were measured manually by one investigator from electrocardiograms performed before and during haloperidol therapy. Logistic regression analysis for prediction of torsade de pointes was performed, incorporating QT(u), QT(B), QT(Frid), QT(Fram), and RR intervals measured during treatment. Receiver operating characteristics (ROC) curves were constructed. Primary outcome measures were proportion of explained variation (maximum-rescaled R2); area under the ROC curves for QT(u), QT(B), QT(Frid), QT(Fram), and RR interval; and sensitivity and specificity for prediction of haloperidol-induced torsade de pointes. The QT(u) was associated with the highest R2 compared with QT(Fram), QT(Frid), QT(B), and RR interval (0.77, 0.73, 0.68, 0.53, and 0.30, respectively). No significant differences in areas under the ROC curves were found between any of the QT-interval methods. Areas under the ROC curves for QT(u) and QT(Fram) trended toward being greater than that associated with the RR interval. All QT-interval methods were highly sensitive (100% for each), whereas the RR interval was less sensitive (86%); QT(u) and QT(Fram) were most specific (82%) compared with the QT(Frid) (72%), QT(B) (64%), and RR interval (36%). CONCLUSION: Compared with QT(B) and QT(Frid), the QT(u) and QT(Fram) best predicted haloperidol-induced torsade de pointes in critically illpatients; the QT(Fram) offered no advantage over the QT(u).