Author + information
- Received January 8, 2018
- Revision received March 9, 2018
- Accepted March 22, 2018
- Published online August 20, 2018.
- Junaid A.B. Zaman, MA, BMBCha,
- Kelvin Chua, MDa,
- Ali A. Sovari, MDa,
- Bruce Gunderson, MSb,
- Eli S. Gang, MDa,
- Sylvain Ploux, MDc and
- Charles D. Swerdlow, MDa,∗ ()
- aCardiac Electrophysiology, Cedars-Sinai Heart Center, Cedars-Sinai Medical Center, Los Angeles, California
- bMedtronic PLC, Mounds View, Minnesota
- cHôpital Cardiologique du Haut-Lévêque, CHU Centre Hospitalier Universitaire, Bordeaux, Université Bordeaux, L'Institut de RYthmologie et Modélisation Cardiaque, Bordeaux, France
- ↵∗Address for correspondence:
Dr. Charles D. Swerdlow, Cedar-Sinai Heart Center, Cedars-Sinai Medical Center, 414 North Camden Drive, Suite 1100, Beverly Hills, California 90210.
Objectives This study sought to develop and evaluate an algorithm for early diagnosis of dislodged implantable cardioverter-defibrillator (ICD) leads.
Background Dislodged defibrillation leads may sense atrial and ventricular electrograms (EGMs), triggering shocks in the vulnerable period that induce ventricular fibrillation (VF).
Methods We developed a 2-step algorithm by using experimental lead dislodgements (LDs) at ICD implantation and a control dataset of newly implanted, in situ leads. Step 1 consisted of an alert triggered by abrupt decrease in R-wave amplitude and increase in pacing threshold. Step 2 withheld therapy based on ventricular EGM evidence of LD identified from experimental LD behavior. We estimated the algorithm’s performance using a registry dataset of 3,624 new implantations and an atrial dislodgement dataset of 14 LDs at the atrium.
Results In the registry dataset, the algorithm identified 20 of 21 radiographic LDs (95%) at a median of 11 days before clinical diagnosis. Step 1 had positive predictive values of 57% for radiographic LD and 77% for surgical revision. The false positive rate was 0.4% after step 1 and ≤0.2% after step 2. In the atrial dislodgement dataset, step 1 identified all 14 LDs; step 2 would have prevented inappropriate therapy in all 7 patients with stored EGMs at LD, including 2 patients with fatal, shock-induced VF.
Conclusions An ICD algorithm can facilitate early diagnosis of defibrillation LD. Additional data are needed to determine the safety of withholding shocks based on EGM evidence of LD.
Dislodgement of a right ventricular (RV) defibrillation lead may cause inappropriate shocks that initiate fatal proarrhythmia. Lead dislodgement (LD) at the right atrium may result in simultaneous sensing of both atrial and ventricular electrograms (EGMs), causing inappropriate detection of ventricular tachycardia (VT) or ventricular fibrillation (VF). A shock synchronized to the atrial EGM may often induce VF because atrial EGMs time in the ventricular vulnerable period during sinus tachycardia, and the shock field of a lead in the atrium is usually below the upper limit of vulnerability in the ventricle. Crucially, the dislodged lead may neither sense nor defibrillate VF. Case reports (1–6) and a research review (7) documented inappropriate shocks inducing VF and failing to defibrillate in both sinus rhythm (1–3) and atrial fibrillation (AF) (4–6).
Implantable cardioverter-defibrillator (ICD) algorithms provide early diagnosis of structural lead failures (8–10). We developed a corresponding algorithm for early diagnosis of ventricular defibrillation LD focused on the unique risks of dislodgement to the atrium and evaluated the algorithm’s performance.
We developed the algorithm using data from an experimental dislodgement dataset of LDs simulated at ICD implantations and a control dataset of normally functioning in situ leads. Then we evaluated the algorithm by using remote monitoring data from a registry dataset of unselected ICD implants and an atrial dislodgement dataset of known clinical LDs to the atrium. Table 1 summarizes these datasets.
Experimental lead dislodgement dataset
Patients were eligible if they were undergoing clinically indicated ICD implantation and did not have complete heart block. Eighteen patients gave written, informed consent to a protocol approved by the Committee on Human Research at Cedars-Sinai Medical Center.
A dedicated bipolar active-fixation defibrillation lead was inserted through the axillary vein and advanced until the tip touched the endocardium of the RV apex. We then withdrew the lead approximately 0.5 cm under right anterior oblique fluoroscopy, extended the active-fixation helix, advanced the lead minimally until the near-field EGM (NF-EGM) showed a current of injury, and removed the stylet. The lead was connected to an ICD generator (Evera VR/DR, Medtronic, Edgewater, Maryland), which was placed in the pocket. The minimum ventricular sensitivity was set to 0.3 mV; VT/VF detection was programmed off, and pacing was disabled. We recorded EGMs and electrical measurements through telemetry and then pulled the lead back toward the atrium to record at 5 distinct, potential LD positions, as determined by fluoroscopy in the right and left anterior oblique positions. The right panels of Figures 1B to 1F show these positions: tip dislodged but abutting the endocardium (RV apex dislodged), midpoint of the coil at tricuspid valve (coil at tricuspid valve), tip at tricuspid valve, tip in atrial cavity approximately 0.5 cm proximal to tricuspid valve, and tip abutting the atrial endocardium (atrial wall), respectively. Finally, we recorded with the lead implanted (Figure 1A, RV apex implanted). We defined positions in which the lead tip was in the atrial or ventricular cavity, as cavitary positions.
We recorded surface electrocardiography (ECG); marker channel; wide-band filtered, NF sensing EGM (tip to ring); and far-field (FF) shock EGM (RV Coil to Can) for at least 30 s, as well as pacing impedance, shock impedance (RV Coil to Can), and pacing threshold. EGMs were exported by using model 9790 programmer’s analog output (Medtronic) to a laptop computer-based data acquisition and analysis system (MP 150, BioPac Systems Inc., Goleta California).
At each position, we measured the maximum positive amplitude (R-wave) and negative amplitude (S-wave) of NF and FF ventricular EGMs by using digital calipers for 12 cardiac cycles, excluding premature ventricular complexes. We defined a fixed, sensing threshold for the FF channel of 0.15 mV.
EGM descriptors of LD
We identified 3 EGM descriptors of simulated LDs that could be measured efficiently in real time by using ICD technology. The first descriptor was based on the observation that LD at the RV cavity resulted in a NF, dedicated bipolar EGM, similar to dominant-negative unipolar EGMs recorded in the RV cavity (11). We defined a cavitary NF-EGM as one in which both the R-wave amplitude was lower than the S wave amplitude and the S-wave had absolute amplitude of 1 to 5 mV (Figure 2A), determined by measuring 12 consecutive EGMs, excluding the largest and smallest, and averaging R and S amplitudes for the remaining 10. This analysis was restricted to 15 patients in whom the lead tip was free in the RV cavity on fluoroscopy when the RV coil was at the tricuspid valve.
The remaining descriptors provided indirect evidence of simultaneous recording of atrial and ventricular signals. The second descriptor was a short-long pattern of inter-EGM intervals on either NF or FF ventricular EGMs, confirmed to represent alternating atrial (A) and ventricular (V) signals by the simultaneous surface ECG (Figure 2B). Long intervals were defined as at least 50% greater than short intervals. For each of the 15 patients in sinus rhythm, we analyzed the first 30 s of both NF- and FF-EGMs at each position. A short-long pattern required ≥3 consecutive sequences of short-long intervals (corresponding to sequential A-V-A-V-A-V EGMs) in any 2 segments of 12 sensed EGMs. The third descriptor identified atrial and ventricular signals without requiring a consistent short-long pattern, based on our observation that the atrial signals often had lower amplitude than the ventricular signals. We defined the EGM amplitude variability index as the ratio of the sum of the 3 largest absolute amplitudes to the sum of 3 smallest absolute amplitudes in a sequence of 12 NF or FF-EGMs (Figure 2C). Both of these criteria required that the average of the smallest 3 of 12 EGMs was ≤2 mV to reject interval/amplitude variations in implanted leads (e.g., premature ventricular complexes).
The control dataset consisted of remotely monitored transmissions from 200 randomly selected, de-identified patients on the CareLink Network (Medtronic), recorded 45 to 90 days after implantation. Transmissions included automated electrical measurements and 10-s real-time EGMs. ICD systems met 4 criteria: each included a newly implanted, dedicated-bipolar defibrillation lead programmed to dedicated-bipolar sensing; each day had <10% ventricular paced events; both NF and FF ventricular EGMs were stored; and both the Lead Integrity Alert (Medtronic) (8) and lead noise algorithm (10) were active.
Algorithm validation datasets
The Product Surveillance Registry (Medtronic) is a prospective, active, post-market approval registry (U.S. Food and Drug Administration post-approval study number P010031 S232). The registry dataset includes all 3,624 lead implants from 2008 to 2017 that met inclusion criteria for the control dataset. We evaluated algorithm sensitivity and specificity by using remotely monitored transmissions for 6 months after implantation. LD was diagnosed as radiographically identified intracavitary displacement requiring repositioning or replacement. This diagnosis excluded lead perforation and electrical abnormalities without radiographic LD (“microdislodgement”), which have a lower risk of shock-induced VF.
ICDs in the control and registry datasets measured pacing impedance every 6 h and R-wave amplitude and pacing threshold daily. Pacing thresholds ≥2.5 V were reported as ≥2.5 V. ICDs store daily values for 14 days and weekly minimum/maximum values thereafter.
Atrial dislodgement dataset
We requested data regarding clinical LDs to the atrium from 25 institutions. Fourteen patients met inclusion criteria: LD at the atrium with electrical measurements both at implant and diagnosis of LD. Dislodgements occurred from 2013 to 2016, except for 1 instance in 2008. The study was approved by the Committees on Human Research at participating institutions.
For experimental LDs, continuous variables with a normal distribution were compared by using repeated-measures analysis of variance in Prism software (Graphpad, La Jolla, California) with Tukey post hoc comparisons between groups. For the registry dataset, the time distribution of LDs was modeled by using the Weibull distribution. Graphs display mean ± SEM and mean ± SD in the tables. Dichotomous variables were compared using the chi-square test. Results were considered significant at a p level of <0.05. Confidence intervals (CIs) are indicated at 95% interval.
Experimental lead dislodgement dataset
Characteristics of the 18 study patients are shown in Table 2.
Figures 3A and 3B graph NF- and FF-EGM amplitudes at each position. Table 3 provides numerical values. NF-EGM amplitude (automated R-wave) decreased to a mean <2 mV with the tip at or proximal to the tricuspid valve. Compared with implantation, pacing impedance decreased at cavitary positions (p < 0.05) (Figure 3C) but not with the tip against the atrial wall; shock impedance did not change significantly. Pacing threshold varied when the tip abutted endocardium but exceeded 5 V at cavitary positions (Figure 3D).
Implanted NF-EGMs showed a dominant R-wave with little or no S-wave (Figure 1A). Dislodgement with the tip abutting the RV apex caused a ≥50% reduction in R-wave amplitude, but minimal change in morphology (Figure 1B). In contrast, NF-EGMs recorded with the tip free in the RV (Figure 1C) showed a cavitary EGM, often with an initial Q-wave. With the tip on the tricuspid valve, the NF-EGM showed distinct atrial and ventricular components (Figure 1D); it continued to display both components as the tip was withdrawn into the atrial cavity close to the tricuspid valve (Figure 1E). In the posterior atrium, the ventricular component was diminished, uniformly below the sensing threshold (Online Figure S1C). When the tip contacted atrial endocardium, NF-EGMs displayed a dominant atrial EGM, sometimes with a current of injury (Figure 1F).
In 2 patients with AF and 1 patient with atrial flutter, rapid sensing of atrial EGMs occurred when the tip reached the atrium. In each, VF (AF cases) or VT (atrial flutter case) would have been detected immediately if detection had been programmed on (Online Figure S2).
In FF-EGMs, with the tip in the RV apex, either fixed or dislodged, FF-EGMs had lower amplitudes than the NF-EGMs did (p < 0.001) (Figures 1A and 1B). With the coil at the tricuspid valve, EGM amplitude decreased, and a low-amplitude atrial component appeared (Figure 1C, Online Figure S1A). FF-EGMs continued to show discrete atrial and ventricular components with the tip at the tricuspid valve (Figure 1D, Online Figure S1B), in the atrial cavity (Figure 1E, Online Figure S1C), or against atrial endocardium (Figure 1F, Online Figure S1D).
Cavitary RV EGM
These NF-EGMs had a characteristic dominant negative signal with S waves of 1 to 5 mV in all 15 patients analyzed (79% to 100%) (Figures 1C and 2A, Online Figure S1A). No patient showed this pattern with the tip implanted or abutting RV endocardium. The absolute R-wave was <3.5 mV in 14 of 15 patients (Online Figure S3).
Figure 4A shows that this pattern was more common in FF-EGMs than in NF-EGMs (p < 0.001) (Figure 1C, Online Figure S1A, S1C, and S1D); but, with the tip near the tricuspid valve, NF-EGMs occasionally recorded it when the FF-EGM’s atrial component was too small to sense.
EGM variability index
Figure 4B shows plots of EGM variability. When NF-EGMs recorded both atrial and ventricular signals, their relative amplitudes were inconsistent. In contrast, FF-EGMs usually recorded small atrial signal and larger ventricular signals (Figures 1C, 1E, and 1F, Online Figure S1A to S1D). Thus, NF-EGMs showed little variability, but FF-EGM variability increased when the coil was at or proximal to the tricuspid valve (p < 0.001 vs. implant). The maximum FF-EGM variability index with the lead in contact with the RV apex was 1.26; 16 patients (88%) had indices <1.1. In contrast, 1 or more atrial positions had an FF-EGM variability index >1.5 in 16 patients (88%; 95% CI: 67% to 97%) (Figure 4A). For all patients, the average of the 3 smallest FF-EGM signals was >2 mV at all RV apical positions and <2 mV at all atrial positions (corresponding to atrial signals).
Evidence of atrial oversensing
With the tip at or proximal to the tricuspid valve, 14 of 15 patients in sinus rhythm (93%) had at least 1 abnormal measurement of atrial over-sensing (Figure 4A). All 3 patients with AF or atrial flutter (2 patients with AF and 1 patient with atrial flutter) had FF-EGM variability indices ≥1.5 at all of these positions. Hence, 17 of 18 patients (94%; 95% CI: 74% to 99%) satisfied at least 1 atrial over-sensing criterion at 1 dislodgement position.
EGMs in atrial fibrillation flutter
In the 2 patients with AF and 1 patient with atrial flutter, rapid sensing of atrial EGMs occurred when the lead tip reached the atrium. In each, VF (AF cases) or VT (atrial flutter case) would have been detected immediately if detection had been programmed ON (Online Figure S3).
Two patients (1%) with suspected LD were replaced with randomly selected patients (Online Figure S4). Table 4 shows an incidence of increased pacing threshold or decreased R-wave amplitude of 0% to 3.5%, depending on the criterion. However, impedance decreased ≥25% in 24% of patients (22% in the first week). In real-time EGMs, no patient had either a FF-EGM variability index ≥1.5 or a short-long pattern on either NF- or FF-EGMs. Although 4% of patients showed cavitary EGMs, none showed increased pacing threshold or decreased R-wave amplitude.
Figure 5 shows the 2-step algorithm developed by using the experimental dislodgement and control datasets. It was designed to fulfill 3 requirements: 1) high sensitivity for LD; 2) low false positive rate; and 3) low power consumption in ICDs connected to normally functioning leads. Step 1 measured R-wave amplitude and pacing threshold hourly and triggered an alert for any of the 3 criteria indicating abrupt change from baseline or abnormal initial measurement (Table 5). Triggers stored an EGM and transmitted a remotely monitored alert. Optionally, triggers stored additional EGMs periodically and activated a patient alert (8–10). Step 1 is estimated to have minimal impact on device longevity: hourly measurements increase baseline current drain by 0.1% while the algorithm is active (personal communication, B. Gunderson, August 2017).
Step 2 analyzes EGMs stored by step 1 triggers for 3 characteristics of LD: cavitary NF-EGM, short-long pattern on NF or FF-EGM, or FF-EGM variability index >1.5. If any characteristic was present, the algorithm disabled detection of VT/VF to prevent inappropriate therapy.
A total of 27 patients had radiographically confirmed LDs. Table 2 summarizes clinical characteristics. One patient died suddenly on day 166; electrical parameters were normal and stable on day 153.
LDs occurred a median of 7 days after implant (range: 1 to 56 days). On the basis of their temporal distribution, an algorithm that operates for 90 days post-implant will sample ≥90% of LDs with 95% likelihood. LD was diagnosed pre-discharge in 7 patients, at scheduled clinic follow-up in 18 patients, and by remote monitoring in 2 patients (pacing threshold increase and Lead Integrity Alert [Medtronic], 1 patient each). Neither the Lead Integrity Alert nor the Lead Noise Algorithm detected any other LDs. One patient presented with 3 inappropriate shocks (Online Figure S5).
Automated electric measurements
Six dislodged leads were repositioned on the day of implantation, before ICDs measured electrical parameters. In the remaining 21 patients, for the week of alert, device-measured R-wave amplitude decreased (8.1 ± 5.4 mV vs. 1.3 ± 0.9 mV, respectively; p < 0.00001) (Figure 6A), pacing threshold increased (p < 0.001) (Figure 6B) and was ≥2.5 V in 18 of 21 patients (86%); and pacing impedance decreased (489 ± 139 Ω vs. 329 ± 54 Ω, respectively; p < .001) (Figure 6C).
Twenty of 21 analyzed LDs satisfied step 1 (95% CI: 77% to 99%) a median of 11 days before clinical diagnosis (range: 0 to 128 days), based on the end of the week in which the alert was triggered. In the remaining patient, R-wave amplitude decreased 53%, but pacing threshold only increased from 0.375 V to 1.25 V.
The flowchart in Figure 7 shows that 40 of 3,624 newly implanted leads triggered step 1 alerts (1.1%). Clinical confirmation could not be obtained for 5 leads because study sites were closed. The remaining 35 leads included 20 radiographic LDs and 15 false positives in 7 leads that required repositioning or replacement for other reasons (RV perforations in 2 patients, electrical abnormalities without radiographic dislodgement in 5 patients) and 8 leads that did not undergo reoperation (e.g., see Online Figure S6). Step 1 alerts had a positive predictive value of 20 of 35 for radiographic LD (57%; 95% CI: 41% to 72%) and 27 of 35 (77%; 95% CI: 61% to 88%) for surgical revision. The false positive rate was 0.4% for absence of radiographic LD and 0.2% for absence of surgical revision. Excluding the 2 RV perforations, 10 of the remaining 13 false positives occurred in the first 2 weeks, suggesting that most represented microdislodgement or early changes at the electrode–myocardial interface.
In step 2, only 3 LD patients had stored EGMs corresponding to step 1 alerts. Each showed simultaneous sensing of atrial and ventricular signals detected as sustained VT (n = 1) (Online Figure S5) or nonsustained VT (n = 2). In each satisfied step 2 short-long criterion, the 2 patients with stored FF-EGMs satisfied the EGM variability criterion; 1 patient showed a cavitary ventricular EGM (Online Figure S5). Of 15 step 1 false positives, step 2 could be applied to real-time EGMs for the 8 patients who did not undergo lead revision; none satisfied step 2. Thus, the 7 step 1 false positives requiring revision placed an upper bound on the false positive rate for the complete algorithm of 0.2%.
Atrial dislodgement dataset
LD was diagnosed on post-implantation days 11 to 30 in 6 patients, days 31 to 90 in 4 patients, days 91 to 180 in 3 patients, and day 1,322 in 1 patient with Twiddler syndrome (median: 38 days; range: 11 to 1,322 days). At diagnosis, the lead tip was in the innominate vein in the patient with Twiddler syndrome and in the atrium in the remaining 13 patients (Online Figure S7).
Eight patients presented with shocks from simultaneous sensing of atrial and ventricular EGMs (median of 2 shocks; range: 1 to 81); 6 patients had antitachycardia pacing. Overall, 9 patients (64%) had inappropriate therapy. Three patients had pro-arrhythmic induction of VF by shocks in the vulnerable zone (n = 2) or antitachycardia pacing (1 patient) (Online Figure S8). Two patients died because induced VF was under-sensed during AF (Figure 8A, Online Figure S9) or sinus tachycardia (Figure 8B).
LD was diagnosed in 6 asymptomatic patients by routine follow-up (n = 2), chest radiographs taken for unrelated indications (n = 3), or remote monitoring alert (n = 1). Of 11 ICDs with lead failure alerts, only 1 ICD triggered an alert (Online Figure S10).
Operator-performed electric measurements
From implantation to LD, R-wave amplitude decreased (11.2 ± 3.9 mV vs. 1.7 ± 1.5 mV, respectively; p < 0.00001) (Figure 6D). Pacing threshold increased (p < 0.001) (Figure 6E); no capture occurred in 9 patients (64%). Pacing impedance decreased (606 ± 183 Ω to 408 ± 154 Ω; p = 0.0018) (Figure 6F). Shock impedance did not change (52 ± 10 Ω vs. 54 ± 11 Ω; p = 0.60).
For step 1, we estimated performance by comparing measurements at implant and diagnosis. All 14 patients satisfied step 1 (95% CI: 78% to 100%) (Figures 6D and 6F). ICD-stored trends for electrical measurements were available for 8 patients. In 7 patients, step 1 alerted a median of 15 days prior to diagnosis (range: 2 to 1,126 days).
In step 2, stored EGMs were available for 7 patients with inappropriate shocks. All 7 also had EGMs stored as monitor zone, nonsustained, or aborted shock events after the step 1 trigger and before the inappropriate shock. In each of the 6 patients in sinus rhythm, both preceding EGMs and EGMs from inappropriate shocks showed short-long patterns on both NF and FF channels (Online Figures S8, S10, and S11). Thus, step 2 would have withheld all shocks, including the fatal shock in sinus rhythm. For the patient in AF, a monitor zone episode stored 16 h prior to the shock showed cavitary EGMs that fulfilled step 1 of 50% R-wave amplitude reduction criterion (Online Figure S12). If step 2 analyzed this EGM, it would have withheld the fatal shock.
This experimental and clinical investigation demonstrates that an ICD algorithm based on electrical and EGM descriptors can improve early diagnosis of defibrillation LD and potentially reduce related inappropriate therapies.
Lead dislodgement algorithm
Step 1 performed with high sensitivity for LD and a low false positive rate. We could not evaluate step 2 on EGMs stored by step 1 true-positive alerts. However, step 2 would have withheld inappropriate shocks in all clinical LDs with stored EGMs, including fatal pro-arrhythmia in 2 patients in the atrial dislodgement dataset. In the control dataset, step 2 had few false positives, and in the registry dataset, step 2 correctly classified all false positive step 1 alerts with real-time EGMs.
We propose implementing step 1 in an active mode. The combination of lead failure alerts and remote monitoring improved early diagnosis of lead failure (9,12). Step 1 combined with remote monitoring will likely improve early diagnosis of LD. We propose implementing step 2 in a passive mode until its performance after step 1 alerts has been evaluated.
Step 1 electric measurements
No previous study has determined the accuracy of changes in electrical parameters for differentiating LD from maturation of newly implanted in situ leads. We found that a 50% reduction in R-wave amplitude was sensitive but nonspecific. Combining it with a large increase in pacing threshold enhanced specificity but missed experimental LDs with the tip against endocardium. Adding 75% reduction in R-wave amplitude identified most LDs against the atrial endocardium with minimal decrease in specificity.
Step 2: EGM analysis
We found that characteristic changes in ventricular EGMs during experimental LDs also occurred during clinical LDs.
Battro and Bidoggia (11) first reported that unipolar EGMs recorded from the RV cavity have a dominant negative component. Little attention has been paid to cavitary bipolar EGMs, both because they are not used to implant leads and because devices measure rectified R waves, removing polarity information. In simulated LDs, we found that cavitary NF-EGMs have low absolute amplitude and a dominant negative component, a pattern not recorded from any in situ lead at the RV apex. However, 4% of control patients had cavitary EGMs, possibly reflecting septal lead placement (11).
EGM evidence of atrial over-sensing: short-long pattern and EGM variability
Sequential atrial and ventricular EGMs in a short-long pattern were recorded more commonly on the FF-EGM with its wide field of view than on the NF-EGM with its narrow field of view. Because the atrial signal usually was distinctly smaller than the ventricular signal on the FF channel, FF-EGM variability increased for LDs to the atrium. These 2 measurements are complementary. The short-long pattern applies when atrial and ventricular EGMs have similar amplitude. FF-EGM variability applies if the short-long pattern is absent (e.g., in AF, sinus tachycardia with prolonged PR interval) or present but not detected (e.g., intermittent under-sensing). Bigeminy, R-wave double counting, and T-wave over-sensing may satisfy both measurements on in situ leads, but they may be rejected by sensing enhancements, and none will trigger step 1. It is especially important to reject T-wave over-sensing, which may occur from dynamic adjustment of sensitivity if R-wave amplitude falls 75%.
Additional observations on clinical lead dislodgements
Our 2 validation datasets reflect different samples of the spectrum of clinical LDs. Both sets show that ICD algorithms for early diagnosis of lead failure (8–10) rarely detect LD. The registry dataset permitted us to determine that an algorithm should operate for approximately 90 days after implantation. However, LDs due to twiddling may occur late. The atrial dislodgement dataset of LDs at high risk for atrial over-sensing confirmed the risks of both inappropriate shocks and fatal pro-arrhythmia identified in case reports (1–6) and in a research review (7).
The algorithm is designed to operate using real-time data, but we could only evaluate performance using data stored for reasons other than LD; this required approximations and limited the number of LD EGMs available for analysis. Almost all NF-EGMs used dedicated bipolar sensing; we have insufficient data to evaluate algorithm performance for integrated bipolar sensing. Except for the atrial dislodgement dataset, ICDs were limited to 1 manufacturer. However, our findings likely apply generally because EGMs were recorded using wide-band filtering and analyzed without proprietary software.
We excluded patients with frequent ventricular pacing. However, based on the algorithm’s design, we expect performance for patients with a safe ventricular escape rhythm and cardiac resynchronization patients with intact AV conduction to be comparable to the results presented. For patients without a safe ventricular escape rhythm, step 1 will omit the R-wave amplitude requirement, and step 2 does not apply because the clinical problem is symptomatic bradycardia not pro-arrhythmia.
We have insufficient data to assess performance of step 2 in patients with AF. In the only clinical LD with a stored EGM and AF, the cavitary EGM criterion would have detected LD at least 16 h before fatal proarrhythmia. However, an ideal LD algorithm would operate during detection of VT/VF because the time from LD until the tip enters the atrium likely varies and LD to the atrium during AF may result in immediate, inappropriate detection of VF. Further work is required to discriminate between appropriate detection of VF by in situ leads and inappropriate detection of AF as VF during LD.
This study demonstrates that an ICD algorithm can use electrical measurements and EGM characteristics to facilitate early diagnosis of defibrillation LD with a low false positive rate. Additional data are needed to determine the safety of withholding shocks based on EGM characteristics.
COMPETENCY IN MEDICAL KNOWLEDGE: Although rare, defibrillation LD may cause inappropriate shocks that initiate fatal proarrhythmia. In the first 90 days after implant, the combination of an abrupt decrease in R-wave amplitude and abrupt increase in pacing threshold is highly sensitive and moderately specific for LD. EGM characteristics related to intracavitary tip position and simultaneous sensing of atrial and ventricular signals increase specificity. Using these features, an ICD algorithm can facilitate early diagnosis of defibrillation LD.
TRANSLATIONAL OUTLOOK: Further analytical work is required to discriminate appropriate detection of VF by in situ leads from inappropriate detection of AF as VF during LD. Performance of the algorithm’s alert feature needs confirmation in clinical practice. Additional data are necessary to determine the safety of withholding shocks based on EGM evidence of LD.
The authors acknowledge the following physicians for contributing cases to the Atrial Dislodgement DataSet: Drs. Donna Gallik (Cedars-Sinai Medical Center, Los Angeles, CA), Jose Dizon (Columbia Presbyterian Medical Center, Mohegan Lake, NY), Ronald Berger (Johns Hopkins Hospital, Baltimore, MD), Alan Cheng (Johns Hopkins Hospital, Baltimore, MD), Raoul Bacquelin (Centre Hospitalier De Chambery Université Paris Descartes, France), Ho Kah Leng (National Heart Centre Singapore, Singapore, Singapore), Hemal Nayak (University of Chicago, Chicago, IL), and Nigel Gupta (Southern California Permanente Medical Group, Los Angeles, CA). They also acknowledge Liesa Shanahan (Medtronic, PLC), Jodi Koehler (Medtronic, PLC), and Jeff Lande (Medtronic, PLC) for contributing to analysis of the registry dataset.
Dr. Zaman was supported by British Heart Foundation, Fulbright Commission. Mr. Gunderson is an employee of and holds stock in Medtronic. Dr. Swerdlow has consulted for Medtronic; and has received honoraria for teaching at Medtronic and Boston Scientific. All other authors have reported that they have no relationships with industry relevant to the contents of this paper to disclose.
Drs. Zaman, Chua, and Sovari contributed equally to this work and are joint first authors.
All authors attest they are in compliance with human studies committees and animal welfare regulations of the authors’ institutions and Food and Drug Administration guidelines, including patient consent where appropriate. For more information, visit the JACC: Clinical Electrophysiology author instructions page.
- Abbreviations and Acronyms
- confidence interval
- far field
- implantable cardioverter-defibrillator
- lead dislodgement
- right ventricular
- ventricular fibrillation
- ventricular tachycardia
- Received January 8, 2018.
- Revision received March 9, 2018.
- Accepted March 22, 2018.
- 2018 American College of Cardiology Foundation
- Nair G.M.,
- Nair V.,
- Healey J.S.,
- Morillo C.A.
- Iwasawa J.,
- Miyazaki S.,
- Masuda R.,
- et al.
- Ruiz-Salas A.,
- Datino T.,
- Pena-Hernandez J.,
- Calvo D.,
- Alzueta J.
- Li W.,
- Tanel R.E.
- Morales J.L.,
- Nava S.,
- Márquez M.F.,
- et al.
- Swerdlow C.D.,
- Gunderson B.D.,
- Ousdigian K.T.,
- et al.
- Welte N.,
- Strik M.,
- Eschalier R.,
- et al.