Fig 1.
(A) Illustration of the QRSp method applied to lead V5 of a representative patient. Five positive (green circles) and five negative (green squares) abnormal QRS peaks are identified on the local QRS (lQRS; solid black line) after identifying three normal peaks (red diamonds) using the smoothed global QRS (gQRS; dashed red line). The number of positive and negative abnormal peaks is summed to produce a QRSp score of 10. (B) Illustrative example of normal QRS peak classification using a magnified view of the QRS complex in Panel A. A negative peak is identified on the smoothed gQRS (black x). The most negative peak on the lQRS within ±10ms of the gQRS peak (shaded area) is classified as normal.
Table 1.
Patient baseline characteristics.
Fig 2.
Comparison of QRSp and QRSd between CRT response groups (N = 47).
Bar graphs comparing (A) QRSp Max and (B) QRSd between CRT responders and non-responders. (C) Scatter plot illustrating a significant (p<0.004) but weak (R = 0.41) linear correlation between the baseline QRSp Max and QRSd for all 47 CRT patients. The 28 CRT responders are highlighted with red circles.
Table 2.
Logistic regression analysis for prediction of CRT response (N = 47).
Fig 3.
Performance of QRSp and QRSd in predicting CRT response (N = 47).
(A) ROC curves for QRSp Max (red solid line) and QRSd (blue dashed line) as a predictor of CRT response. Black circles highlight the sensitivity and 1-specificity obtained by using a QRSp Max cutpoint ≥7 and a QRSd cutpoint ≥150ms. (B) Bar graphs comparing the prognostic performance of the QRSp Max and QRSd cut-points for identifying CRT responders. (C-E) Bar graphs comparing CRT responder rates between patients stratified by (C) the optimal QRSp Max cutpoint of ≥7, (D) the conventional QRSd cutpoint of ≥150ms, and (E) the combination of both these cutpoints. While nearly all patients with both QRS ≥150ms and QRSp ≥7 responded to CRT, there were no responders amongst patients with both a QRS <150ms and QRSp <7. AUC = area under the curve. 95% CI = 95% confidence interval. PPV = positive predictive value. NPV = negative predictive value.
Fig 4.
Comparison of QRSp and QRSd between baseline and follow-up in CRT responders and non-responders (N = 38).
Bar graphs comparing changes between baseline and follow-up (A) QRSp Max and (B) QRSd in CRT responders and non-responders.
Fig 5.
Examples of QRSp at baseline and follow-up in CRT responders and non-responders with baseline QRSd above and below 150ms.
QRSp Max results at baseline and follow-up for a single 10-beat window from (top left) a CRT responder with a QRSd ≥150ms, (top right) a CRT non-responder with a QRSd ≥150ms, (bottom left) a CRT responder with a QRSd <150ms, and (bottom right) a CRT non-responder with a QRSd <150ms. Irrespective of their baseline QRSd, the CRT responders have greater baseline QRSp Max values than the non-responders. At post-CRT followup, the QRS complexes of the responders become smoother and their QRSp decreases, while the QRS complexes of non-responders become more fractionated and their QRSp increase. Solid black lines demarcate the lQRS and dashed red lines demarcate the gQRS. Normal peaks are annotated with red diamonds, and abnormal peaks are annotated with green circles.