In patients with New York Heart Association (NYHA) Class I to IV heart failure (HF), CRT improves cardiac function and reduces morbidity and mortality. In NYHA Class III to IV HF patients, CRT also improves quality of life. In addition to direct therapeutic effect, modern CRT devices monitor a large number of physiologic parameters, which can be used to fine-tune medical management (
Table 13.1). Monitoring of HRV is readily available on most CRT devices; it is used for HF management.
Device-Based Automatic Monitoring of SDAAM/SDANN
Modern implantable cardioverter-defibrillators (ICD) and CRT devices of major device manufacturers are capable of automatic assessment of HRV. The standard deviation of 5 minutes median atrial-to-atrial intrinsic intervals (SDAAM) is calculated automatically. The device excludes from analysis atrial paced beats, atrial and ventricular premature complexes, and arrhythmic episodes.
16 The percentage of time during the 24-hour collection period in which intrinsic atrial beats were eligible for analysis is reported, as well. In Boston Scientific devices, if there were less than 67% eligible intrinsic atrial intervals, SDAAM is not displayed for that period.
16 In addition to a 24-hour period, average SDAAM is reported weekly. Device-measured SDAAM is an equivalent of Holter-measured standard deviation of 5-minute
average normal-to-normal (NN) intervals (SDANN). SDANN or SDAAM measures long-term HR fluctuations. SDAAM/SDANN is an optimal long-term HRV metric for automated analysis, because averaging several hundred atrial-to-atrial intrinsic intervals is less affected by inaccuracies in device diagnosis of atrial premature complexes and abnormal supraventricular rhythms and provides reasonably reliable HRV estimate. For automated analysis, SDANN is a more accurate long-term HRV measure than standard deviation of NN intervals (SDNN).
2
Braunschweig et al
17 monitored CRT-device-based SDANN in the InSyncIII prospective study participants and reported improvement of SDANN 2 weeks post-CRT implant, especially among the NYHA Class III HF patients. In the NYHA Class II and IV HF patients, SDANN improvement plateaued at 4 weeks and did not improve further. In Class III HF patients, improvement of SDANN continued until 3-month post-CRT implantation.
Adamson et al showed that decreased SDAAM (<50 ms when averaged over 4 weeks) was associated with a greater than 3-fold mortality.
18 SDAAM decreased median 16 days before HF hospitalization and returned to baseline after treatment. Sensitivity for prediction of HF hospitalizations was 70%, while the rate of false-positives was high (2.4 per 1 patient-year of follow-up).
Fantoni et al
16 showed that CRT-responders demonstrated a significant increase of SDANN 3 months after starting CRT. SDANN increase was consistent in patients with ischemic and nonischemic cardiomyopathy. Fantoni et al
16 defined CRT nonresponders as patients in whom SDANN did not improve or even worsened 4 weeks after CRT implantation. SDANN-defined CRT nonresponders had a significantly higher rate of combined clinical outcome events (HF hospitalizations, appropriate ICD therapies, or cardiovascular mortality). SDANN-defined CRT nonresponders showed no significant improvement in exercise capacity and left ventricular (LV) function 1 year post CRT. SDANN was the earliest parameter, indicating functional improvement. Close monitoring of SDANN within the first months after CRT implantation can identify potential nonresponders very early, just 1 month post-CRT implant. Early recognition of CRT nonresponse provides a chance to tailor patient management
19 and improve clinical outcomes. SDANN improvement post CRT is
consistent with a significant reduction of sympathetic neural activity after biventricular pacing in systolic HF,
20 suggesting that the long-term HRV measure SDANN does reflect sympathetic tone.
The Heart Failure-Heart Rate Variability Registry (HF-HRV)
21 prospectively evaluated the relationships between device-measured HRV parameters and clinical characteristics of HF patients.
22,
23 When adjusted by demographics, body mass index, and diastolic blood pressure, Cox regression analyses showed significant association of SDANN and HRV footprint percentage with all-cause mortality in CRT patients.
22 Changes in SDANN and HRV footprint percentage significantly correlated with clinical improvement, measured by quality of life and activity level.
23
Footprint is a graphical rendering of the likelihood of a particular beat-to-beat HR change occurring at each intrinsic sinus rate during a 24-hour period,
24 implemented in Boston Scientific devices. The footprint area is the normalized size of the two-dimensional plot of RR interval variability versus HR; it is reported as a percentage of graph area used. HRV footprint provides a graphical representation of several variables, including frequency; day-to-day rate variability; and maximum, minimum, and mean HR. Molon et al
25 used HRV footprint graph to derive two nonlinear HRV measures: HR complexity and HR entropy. Nonlinear indices represent more structured information than footprint area. HF patients with higher HR-related complexity, representing a less compromised autonomic function, had better clinical response.
25 A significant increase in HR complexity was associated with clinical improvement 1 year post-CRT implantation.
25 However, HR complexity and entropy strongly correlated with traditional time-domain HRV metrics (SDANN). Further study of HR complexity and entropy is needed, to determine whether their addition to traditional HRV metrics carries additional predictive value. Any additional computational feature in implantable devices utilizes device memory and can potentially shorten device battery life. Therefore, complementary predictive value and benefit of any new HRV parameter for continuous monitoring should be documented first.
Risk Scores Using Monitoring of Autonomic Tone for Prediction of HF Outcomes
It is well known that combination of several risk factors into a risk score can improve the accuracy of the prediction. Two up-to-date risk models incorporate data obtained by CRT devices during continuous monitoring of autonomic tone. One, developed and validated by Singh et al,
21 predicts all-cause mortality in HF patients. The other, developed and validated by Cowie et al,
26 predicts HF hospitalizations.
The risk model by Singh et al
21 was developed in The Cardiac Resynchronization Therapy Registry Evaluating Patient Response with RENEWAL Family Devices (CRT RENEWAL) study and was validated in the HF-HRV study of CRT patients. SDANN, HRV footprint, mean 24-hour HR, and activity log data were collected and included in the predictive model. All-cause mortality served as the primary outcome. Monitored HRV parameters were dichotomized with higher risk of death assigned on the basis of device diagnostics obtained at the 2-week visit: SDANN less than 43 ms, mean HR greater than 74 beats/min, HRV footprint less than 29%, and activity level less than 5%. The developed risk model predicted 1-year all-cause mortality with moderate accuracy (C-statistic was 0.678 in the development cohort and 0.704 in the validation cohort). The risk model was well calibrated: mortality in the high-risk group was 13.8% in the development cohort and 13.6% in the validation cohort. The risk score model by Singh et al
21 can be easily implemented in clinical practice.
The risk score by Cowie et al
26 was developed in the combined patient population of OFISSER,
27 Italian ClinicalService Project,
28 and CONNECT
29 studies. The validation cohort included PARTNERS-HF,
30 FAST, PRECEDE-HF, and SENSE-HF
31 studies population. Patient data were included if 90 days of device diagnostic data
were available. Adjudicated HF hospitalization event served as a primary outcome. Intrathoracic impedance, activity level, night HR, HRV, atrial fibrillation burden and characteristics, as well as ventricular arrhythmia burden and treatment were included in the model. A Bayesian Belief Network framework was used to combine the evidence from each diagnostic parameter and generate the risk score. In a high-risk group 2.3% of patients were hospitalized within the next 30 days in the development cohort and 6.8% in the validation cohort. In the validation cohort, the high-risk group was identified with 46% sensitivity (ie, 46% of the months with HF hospitalizations were preceded by a high-risk score) and 90% specificity (ie, 10% of the months with no HF hospitalizations were also preceded by a high-risk score). Patients who achieve a high-risk state on any day in the last 30 days were 10 times more likely to be hospitalized for HF in the next 30 days as compared to patients who had a low risk on each of the last 30 days. The advantage of this risk score is that it is time varying. Each patient can change risk category and transition from low to high risk or from high to low risk at any given time. Two risk scores are compared in
Table 13.2. To simplify implementation in clinical practice, it would be more convenient if authors provided the Excel calculator on a website, or as a supplementary to the published manuscript Excel file. Unfortunately, neither of these two scores provided a user-friendly tool, slowing down the implementation of the risk stratification tool in clinical practice.