Clinical performance of implantable cardioverter-de ﬁ brillator lead monitoring diagnostics

BACKGROUND Implantable cardioverter-de ﬁ brillator (ICD) lead monitoring diagnostic alerts facilitate the diagnosis of structural lead failure. OBJECTIVE The purpose of this study was to prospectively study the performance of Medtronic ICD lead monitoring alerts. METHODS A false-positive alerts resulted in an unnecessary lead replacement. For the diagnosis of pace-sense lead failure, an alert for oversensing was more sensitive than a pacing impedance threshold alert.


Introduction
Automated, device-based, lead-monitoring alerts were developed to improve the early detection of defibrillation lead failure. [1][2][3][4][5] In retrospective studies, alerts reduced inappropriate shocks in patients with recalled Sprint Fidelis (Medtronic, Inc. Minneapolis, MN) leads 6 and facilitated the diagnosis of lead failure in patients with other leads. 3,7,8 No lead performance diagnostic has been studied prospectively.
The World-Wide Randomized Antibiotic Envelope Infection Prevention Trial (WRAP-IT) assessed an absorbable, antibiotic-eluting envelope for reducing infection associated with device implantations. 9 In a prespecified ancillary study, we prospectively assessed the real-world performance of lead monitoring alerts to detect structural lead failure.

Methods
The WRAP-IT study design, patient population, and results have been reported. 9 The protocol was approved by the ethics committee at each participating institution; all patients provided written informed consent.

Lead monitoring study design
This study was prespecified as an ancillary prospective study using the WRAP-IT study patient population to assess the performance of lead monitoring alerts.

Patient population
WRAP-IT patients with an implantable cardioverterdefibrillator (ICD) or cardiac resynchronization therapy-defibrillator were enrolled. A Medtronic generator was implanted at the index WRAP-IT procedure, but leads from any manufacturer were permitted. The investigators chose pacing and shock impedance alert thresholds, rate thresholds, and number of intervals to detect ventricular tachycardia (VT) or fibrillation (VF). Only those patients with all 4 lead monitoring alerts enabled were included in this analysis.
Alerts included right ventricular (RV) pacing and shock impedance alerts, the Lead Integrity Alert (LIA), and the RV Lead Noise Algorithm (LNA). Impedance alerts are triggered for values outside a programmable range, nominally 200-2000 U for pacing and 20-200 U for shock impedance. LIA is composed of 2 oversensing components and 1 abrupt impedance change component 6,10 and is triggered when threshold criteria are satisfied for any 2 components. LNA identifies oversensing as signals on the near-field sensing channel that do not correspond to simultaneous signals on the far-field sensing channel. 10,11 The Online Supplement describes the components of the lead monitoring diagnostic alerts in more detail.

Data collection
At each study visit, device data including alerts were reviewed and reported; all suspected lead system events (LSEs) were reported, regardless of whether an alert was triggered. Reports included results of diagnostic tests, occurrence of inappropriate shocks, and any action taken in response to the suspected LSE and/or alert.

Data adjudication
The Lead System Events Clinical Events Committee (CEC) comprised 3 physicians. Study report forms and source records (eg, radiographs, procedure notes, and device interrogation files) corresponding to each alert and all other suspected LSEs were independently reviewed to determine the root cause. Majority agreement with high confidence was required for final adjudication. Events were excluded from the analysis if no ICD interrogation data were available for analysis. An LSE was defined as an event affecting the sensing or defibrillation function of the RV defibrillation lead. The CEC used previously defined criteria based on the analysis of returned leads 5,12,13 and/or radiography 12,14 to adjudicate LSEs as structural lead failures or one of the following specific conditions: connection issues (between the lead pin and the header), lead perforations, or lead dislodgments. Findings that were used to diagnose structural lead failure were as follows: Pace-sense lead failure had to include either characteristic oversensing 12,13,15 or abrupt impedance change and connection problem criteria were not fulfilled for either a DF-4 or an IS-1 connector 4,12 : For the shock component, abrupt change in painless (subthreshold) or shock measurement of high-voltage impedance was required and a connection problem was excluded. When available, radiographs showed complete insertion of the lead pin into the header.
Events that did not meet prespecified criteria for LSEs were adjudicated as nonlead system events (NLEs). For NLEs, previously reported criteria were used to determine the specific causes of impedance changes 4,12 and oversensing. 13

Statistical analysis
Standard descriptive statistics were used to summarize patient characteristics. LSE detection sensitivity for each alert was estimated as the number of LSEs detected by the alert divided by the total number of LSEs. Detection sensitivity was estimated in this manner for each type of LSE, including lead failure. The positive predictive value (PPV) of each alert for lead failure was defined as the number of lead failures detected by the alert divided by the number of alerts. Sensitivity and PPV for different alerts were compared using the c 2 test with Yates's correction. Since lead monitoring is constantly engaged and it was assumed that leads were functioning normally in the absence of a lead issue or event, false-positive rate is a more natural metric than specificity for understanding lead monitoring alerts in the absence of lead issues. The false-positive rate at which alerts detect NLEs was estimated by dividing the total number of NLEs by the total number of lead monitoring years and expressed by the number of lead monitoring years required to observe 1 NLE.
The research reported has adhered to the Helsinki Declaration as revised in 2013.

Study population
The patient population studied is shown in Figure 1. An RV defibrillation lead was present in 5208 patients in the WRAP-IT. Patients were excluded from this analysis if they had no interrogation after implant (n5 192), any of the 4 alerts were not enabled (n 5 62), or the status of alerts could not be verified (n 5 12). Therefore, 4942 patients were analyzed. Table 1 and Online Supplemental Table 2 summarize patient and device characteristics of the study population. Patients were followed for a total of 95,685 patient-months (19.4 6 8.7 mo/patient). The average age of leads at study entry was 92.3 6 41.9 months.

LSEs
Of the 84 LSEs, 65 triggered alerts including 48 lead failures and 17 other LSEs. Among the 17 other LSEs, the CEC could not adjudicate between lead failure and a connection issue in 3 instances. All 19 LSEs that did not trigger an alert were lead dislodgments or perforations. Figure 2, Online Supplemental  LSEs, 36 were in the envelope arm and 48 were in the no envelope arm (42.9% vs 57.1%; P 5 .1904). Of the 48 lead failures, 18 were in the envelope arm and 30 were in the no envelope arm (37.5% vs 62.5%; P 5 .0833) (Online Supplemental Table 2).

NLEs with alerts
Of the 59 NLEs that triggered alerts, 36 were oversensing alerts and 23 were impedance alerts ( Figure 3). These included oversensing alerts during VF/VT or agonal rhythm (n 5 17), electromagnetic interference (n 5 11), and physiologic signals (n 5 8). Impedance alerts were triggered by physiologic variations in pacing or shock impedance (n 5 12) or single anomalous pacing or shock impedance measurements (n 5 8). Three cases were due to electrode-myocardial interface pacing impedance changes.

Performance of lead monitoring alerts
Overall performance The sensitivity of the combined suite of all 4 lead monitoring alerts was 77.4% (95% confidence interval 67.0%-85.8%) for the 84 diagnosed LSEs and 100% (95% confidence interval 92.6%-100%) for the 48 diagnosed lead failures. The sensitivity was also 100% for the 7 connection issues and for the 3 events adjudicated as possible lead failure or connection issue. The sensitivity for lead dislodgment or perforation was 3 of 22 (13.6%). The PPV of any lead monitoring alert for lead failure was 38.7% (48 of 124 lead monitoring alerts). Overall, there was 1 true-positive alert per 166 patient-years for definite lead failure; 1 alert per 123 patientyears for any LSE; and 1 false-positive alert per 135 patientyears for any NLE ( Table 2).

Performance of individual alerts
Online Supplement and Online Supplemental Table 5 describe a detailed performance of individual alerts. We emphasize 4 observations: (1) For pace-sense lead failures, LIA was more sensitive than pacing impedance (97.1% vs 26.5%; P , .0001) and had similar PPV (44.6% vs 50.0%; P 5 .8815).
(2) When LIA was activated for lead failure, the 2 oversensing criteria were satisfied first more frequently than the abrupt impedance criterion and 1 oversensing criterion (64.7% vs 32.4%; P 5 .0152); however, the PPV was higher when the abrupt impedance criterion was satisfied (100% vs 34.9%; P 5 .0002). (3) LNA was triggered in only 9% of pace-sense lead failures and always in conjunction with LIA (Table 2). (4) The PPV of a shock impedance alert for lead failure was 33.3%. No high-voltage lead failure presented clinically. All were identified by an abrupt change in shock impedance (Supplemental Figures 6B and 6C). Values are presented as mean 6 SD or n (%). COPD 5 chronic obstructive pulmonary disease; CRT 5 cardiac resynchronization therapy; ICD 5 implantable cardioverter-defibrillator; RV 5 right ventricular. *Patients enrolled because they were planning to receive a new CRTdefibrillator, but received a new ICD instead. † Implant date was not available for all chronic leads.

Clinical considerations
Remote monitoring Of the 124 total alerts, 112 (90.3%) occurred in patients enrolled in remote monitoring; 12 (9.7%) were in patients without remote monitoring and were detected by patient auditory alert or in-person interrogation. Of the 112 alerts in patients with remote monitoring, 81 were transmitted within 2 days (50 transmitted automatically, 31 transmitted manually in response to auditory alerts), 21 were interrogated within 2 days before any transmission occurred, and 10 were transmitted or interrogated after 2 days because of circumstances such as being out of town, not being near remote monitor, or being hospitalized at the time.
Inappropriate shocks There were 6 inappropriate shocks in 6 patients. Three shocks for 2 pace-sense lead failures (LIA alerts), and 1 shock for a lead dislodgment to the atrium (no alert). One additional patient was shocked when LNA timed out (1 minute after LIA) and the final patient had a shock 39 hours after LIA triggered while on vacation without a remote monitor (see Supplemental Figure 4A legend for details).

Physician response to lead-monitoring alerts
Investigators replaced 37 of the 48 failed leads. One lead replacement procedure was unsuccessful, and no replacement was reported before study exit for the remaining 10 failed leads (Online Supplemental Table 6). Of the 59 NLEs with an alert, none resulted in unnecessary lead replacement. Physicians reported reprogramming for 19 of the 59 NLEs (32.2%) and no specific intervention for the remainder.

Temporal distribution of LSEs
As expected, the type of LSE that triggered lead monitoring alerts varied with the age of the lead or time after the index operation for the subset of patients for whom the date of original lead implant was available. The median time from lead implant to alert was 0.6 months for dislodgment or perforation (n 5 22) and 107 months (n 5 28) for lead failure.

Discussion
This prespecified, prospective, ancillary study of WRAP-IT examined the real-world performance of lead monitoring alerts in a large international patient cohort. Our primary finding is that a suite of 4 alerts was 100% sensitive for clinically diagnosed, structural lead failures. Further, the rate of inappropriate shocks for both lead failures and other causes of rapid oversensing was low.
Although the PPV of lead monitoring alerts was only 38.7% for lead failure, false-positive alerts triggered by events other than lead failures did not result in unnecessary lead replacements or other adverse events and some alerted clinicians to other clinically important findings.
Individual lead-monitoring alerts LIA LIA, composed of both oversensing and abrupt impedance change components, was designed to provide early warning for conductor fractures and reduce inappropriate shocks. 2,6 Subsequently, it performed well in retrospective studies of other Medtronic leads 12 and leads from multiple manufacturers. 7,8,16 Our prospective data confirm 3 key aspects of LIA's performance previously reported in a retrospective cohort 7 : (1) Most pace-sense lead failures are diagnosed by rapid oversensing alone without impedance alerts, (2) LIA's oversensing criteria are insensitive to the cause of rapid oversensing, resulting in a low PPV when only oversensing criteria are triggered; and (3) LIA had 100% PPV for lead failure for the few alerts in which the impedance criterion was triggered with an oversensing criterion. In a previous report, connection issues were shown to trigger LIA's impedance criterion. 7 LNA LNA withholds inappropriate shocks if the VF detection is otherwise satisfied by rapid oversensing 10,11 In our study, LNA was triggered rarely and always in conjunction with LIA, suggesting that prompt response to LIAs reduced the role of LNA. LNA's most serious potential limitation is that it could withhold lifesaving therapy, but this has never been reported to our knowledge. 15,17 While arrhythmia events were not adjudicated in this study, we are not aware of any VT/VF event in which therapy was withheld. Oversensing Alerts Figure 3 Nonlead system events (NLEs). All adjudicated NLEs associated with an alert are shown. *Two patients also triggered the Lead Noise Algorithm. EMI 5 electromagnetic interference; LIA 5 Lead Integrity Alert; NA 5 not applicable; VT/VF 5 ventricular tachycardia/fibrillation; Z 5 impedance.

Impedance alerts
Consistent with bench testing 4 and retrospective studies, 6,7,12,16 we found that alerts for high pacing impedance were insensitive for lead failure and had only a modest (50%) PPV. Shock impedance provides the only alert for failure of lead shock components. 4 Bench testing and retrospective reports indicate that abrupt increases in shock impedance are sensitive for clinically significant conductor fractures or connection issue 4,5 but decreases are insensitive for insulation breaches. 4,18 In this prospective study, 13 shock conductor fractures were diagnosed by impedance alerts associated with abrupt impedance increases but only 1 low shock impedance alert occurred.

Clinical implications
False-positive alerts Some alerts triggered by causes other than lead failure provided the first warning of clinically important findings, such as connection issues and T-wave oversensing. Other alerts were not helpful, such as oversensing alerts for VT/ VF and most impedance alerts in normally functioning leads. Such alerts can cause patient anxiety and burden devicemonitoring clinics. Fortunately, in the present study, no physician responded to a false-positive alert by replacing a lead unnecessarily.

Inappropriate shocks
The inappropriate shock rate of 6% for pace-sense lead failures in this study was not significantly different from the rate of 12% (8 of 65) in a cohort of 12,793 lead-device combinations followed by remote monitoring for 13,562 years. 7 Relevance to alerts from other manufacturers While this study analyzed only Medtronic lead monitoring alerts, our primary findings should apply to all ICDs. Methods for measuring pacing and shock impedances are similar for most manufacturers, 4 so all should be subject to the same limitations of sensitivity and PPV. Other manufacturers' oversensing diagnostics have similarities to those studied. Biotronik's RV Lead Monitoring (Lake Oswego, OR) feature is triggered by short intervals only. When triggered by oversensing alone, LIA has low PPV for lead failure even though it requires both rapid nonsustained tachycardia and short intervals. Performance has not been reported, but our data predict that this alert is sensitive but not specific. The oversensing component of Boston Scientific's Latitude Lead Check (St. Paul, MN) is triggered by 4 short intervals, but only within a sustained arrhythmia episode. It is implemented in the remote monitoring network rather than the generator; and it is the only remote monitoring "Red Alert" that is nominally off. Additionally, like LNA, Abbott's Se-cureSense (Little Canada, MN) feature identifies oversensed signals that do not correlate with electrograms on the far-field channel; but unlike LNA, SecureSense can be triggered by  *In 4 patients with LIA, LNA was also triggered. Three were pace-sense lead failures. One was a connection issue. † In 1 lead failure, the SVC, LIA, and shock impedance was triggered, but the patient had a single-coil lead. This patient was counted under both shock impedance and pacing impedance (6LIA). ‡ Of the 14 lead failures identified by a shock impedance alert only, 3 were RV shocks, 7 were SVC shock impedance alerts, and 4 were RV shocks and SVC shocks. nonsustained events. 3 Welte et al 3 reported high sensitivity but low specificity.
Opportunities for improved lead-monitoring diagnostics Our findings suggest possible improvements: we identified false-positive oversensing alerts by specific electrogram characteristics 15,17 and false-positive impedance alerts by their impedance trend patterns. 4 These analyses could be automated by postprocessing source data for existing alerts in a programmer or remote monitoring network before they are presented to the clinician. Remote monitoring facilitates rapid transmission of alerts for prompt intervention, and opportunities to maximize its use are important. 16,19 Our findings also highlight the need for new diagnostics for lead dislodgment or perforation. 14,20 Additionally, previous work has established the insensitivity of shock impedance for potentially catastrophic insulation breaches and proposed novel diagnostics to overcome this limitation. 4,18

Study limitations
This study has 2 fundamental limitations common to clinical studies of lead failure. First, we could only identify lead failures if they triggered alerts or presented with clinical evidence of lead dysfunction, such as abnormal impedance on a delivered shock or failure to capture. Neither this study nor other clinical studies of lead failure could identify subclinical failures that did not trigger alerts or present clinically. A second fundamental limitation is that no lead diagnosed as having failed was returned for analysis, so we could not confirm diagnoses. However, the diagnostic criteria we used were based on rigorous studies in which leads were returned for analysis (Online Supplemental Table 1). 5,12,13 Characteristic oversensing patterns are considered diagnostic of lead failure or connection issue 12,13 ; these occurred in 33 of 34 pace-sense lead failures. Subthreshold impedance or impedance during delivered shocks are the primary clinical measures of shock component integrity. The limitation of subthreshold measurements for underdiagnosis of insulation breaches is widely acknowledged, 4,18,21 but most leads in this study were Medtronic leads, which have a low incidence of insulation breaches. For shock conductor fracture, the abrupt impedance increase criterion we used was validated by returned-product analysis 5 and did not occur in any of the 7796 functioning Medtronic leads in a recent analysis. 4 In analyzing alerts, it is not always possible to distinguish lead conductor fracture from connection issues because both may cause abrupt increases in impedance. 12,13 Thus, we adjudicated 3 alerts as LSEs caused by either conductor fracture or connection issue. We excluded patients who did not have device interrogations and followed patients only to their last interrogation or last available contact point, so it is unlikely that we missed clinical evidence of lead system malfunction that did not trigger an alert. Patients were followed for a mean of only 19.4 months, but many leads were at risk of failure, since the average age of leads, when known, at study entry was 92.3 6 41.9 months. In comparison with an unselected population, our cohort comprised primarily Medtronic leads, 12 which are more prone to conductor fracture than insulation breach, and a higher proportion of older leads after generator change than new leads after the initial implant.

Conclusion
In this first prospective international study, lead monitoring alerts were 100% sensitive for clinically diagnosed, structural lead failures. Use of all lead monitoring alerts was associated with a low rate of inappropriate shocks. For pace-sense lead failure, an alert that incorporates oversensing was more sensitive than an impedance threshold alert and had a comparable PPV. Although the overall PPV of lead monitoring alerts was modest, false-positive alerts were rare and did not result in unnecessary lead replacements or other adverse events.