Author + information
- Received September 4, 2010
- Revision received November 17, 2010
- Accepted December 7, 2010
- Published online June 7, 2011.
- Charles D. Swerdlow, MD⁎,⁎ (, )
- Haresh Sachanandani, MS†,
- Bruce D. Gunderson, MS†,
- Kevin T. Ousdigian, MS†,
- Mark Hjelle, BS† and
- Kenneth A. Ellenbogen, MD‡
- ↵⁎Reprint requests and correspondence:
Dr. Charles D. Swerdlow, Cedars-Sinai Medical Towers, 8635 West Third Street, Suite 1190 W, Los Angeles, California 90048
Objectives This study sought to use implantable cardioverter-defibrillator (ICD) diagnostics to discriminate ICD lead fractures from normally functioning leads with high impedance and from connection problems between the lead and header.
Background ICD diagnostics facilitate identification of fractures, but there are no accepted criteria for discriminating fractures from other causes of high impedance and/or nonphysiological “noise” oversensing.
Methods We analyzed a development set of 91 leads to construct a stepwise algorithm based on ICD diagnostics. It included 40 fractures, 30 connection problems, and 21 functioning leads that triggered high-impedance alerts. Then we applied this algorithm to an independent test set of 100 leads: 70 fractures and 30 intact leads with connection problems that were misdiagnosed clinically as fractures. In the algorithm, either extremely high maximum impedance or noise oversensing with a normal impedance trend indicated a fracture. A short interval from surgery to impedance rise or prolonged stable impedance after an abrupt rise indicated a connection problem. A gradual impedance increase or stable, high impedance indicated a functioning lead.
Results In the test set, the algorithm correctly classified 100% of fractures (95% confidence interval [CI]: 95% to 100%) and 87% of connection problems that were misdiagnosed as fractures (95% CI: 70% to 95%).
Conclusions An algorithm using only ICD diagnostics identifies leads with oversensing or high impedance as fractures or connection problems with a high degree of accuracy.
Pace-sense fractures of implantable cardioverter defibrillator (ICD) leads are diagnosed by oversensing of characteristic, nonphysiological signals (“noise”) and/or an increase in pacing impedance (1–4). Rapid diagnosis is important because the time between abnormal diagnostics and 1 or multiple inappropriate shocks caused by rapid oversensing may be short (1,3,4). However, lead replacement is associated with significant morbidity and rare deaths (5), and an incorrect diagnosis may result in replacement of normally functioning leads (6).
Our goal is to reduce overdiagnosis of lead fractures. Previously, we reported characteristics of noise that discriminate fractures from other causes of rapid oversensing with normal impedance (7,8). The differential diagnosis of high impedance and noise oversensing includes connection problems between the lead and ICD (9,10); the differential diagnosis of high impedance also includes normally functioning leads. Neither requires lead replacement. In the present study, we analyzed characteristics of impedance trends and the relationship between these characteristics and noise oversensing. We hypothesized that these ICD diagnostics can be used to construct a comprehensive algorithm that discriminates lead fractures from normally functioning leads with high impedance and connection problems, including those that were misdiagnosed as fractures.
We analyzed diagnostics from ICD “save-to-disk” or remote-monitoring data files. First, we performed an exploratory analysis in a development set to identify discriminating criteria. We used these criteria to construct a stepwise algorithm. Then we applied this algorithm to an independent test set.
Patient groups: development and test sets
The “development set” of leads from 91 patients consisted of 30 leads with connection problems, 40 Fidelis fractures, and 21 functioning Fidelis leads with high impedance. The “test set” of leads from 100 patients included 30 connection problems and 70 fractures; fractures were divided prospectively into a subgroup of 40 consecutive Fidelis leads and a subgroup of 30 consecutive non-Fidelis, multilumen (Sprint-Fidelis family, Medtronic, Inc., Minneapolis, Minnesota) leads.
We included functioning leads in the development set with the goal of identifying unique features of presumed electrode-myocardial interface problems. Because these leads remained implanted, we could not determine the cause of their high impedance definitively. Thus, we did not include them in the test set.
Reference standards for lead status
This study used both clinical criteria and criteria based on returned product analysis (RPA) of explanted leads. Fractures were determined only by RPA. Normally functioning leads with high impedance were defined by the absence of clinical problems during follow-up. In the test set, connection problems were determined only by RPA. In the development set, they were determined either clinically or by RPA.
RPA Reference Standards
Lead fracture was determined if analysis of an explanted lead or lead fragment identified a fracture. In contrast, a lead was considered intact (not fractured) only if it was returned in its entirety for analysis; no abnormality was identified by visual inspection and low-power (10× to 30×) light microscopy except for minor, typical extraction-related changes; and unipolar electrical continuity measurements were within impedance specifications for each conductor, both at rest and under various positions of mechanical stress. Impedance specifications varied depending on lead model and length (e.g., 21.6 ± 3.6 Ω for the tip electrode and <47 Ω for the ring electrode of a 65-cm, model 6949 Fidelis lead).
If the lead pin is not inserted completely into the header, the ring electrode contacts its corresponding conductor incompletely or intermittently, resulting in a connection problem. Tightened set screws result in identifiable marks on the pin. We used this observation to determine incomplete insertion of the pin into the ICD header. When leads were explanted for clinical misdiagnosis of pace-sense fracture and RPA determined they were intact, the lead pin was inspected under low-power microscopy. A connection problem was diagnosed if the set-screw mark was within 0.165 cm from the tip of the pin (normal distance ≥0.330 cm) (Fig. 1).
Clinical determinations were applied to leads in the development set that were not explanted. Connection problems met 3 criteria: 1) incomplete lead-pin insertion into the generator's header was identified at a subsequent surgical procedure; 2) complete insertion of the pin corrected the clinical problem (high impedance and/or oversensing); and 3) neither the generator nor the lead was replaced.
Leads with high and rising impedance that triggered the high-impedance feature of the lead integrity alert (LIA) (3) were determined to be “functioning” normally if they had no evidence of lead or ICD system malfunction—including any abnormality of pacing, sensing, or high voltage conductors—during minimum follow-up of 1 year after the impedance alert (median: 20 months, range 13 to 36 months). These leads did not undergo RPA because they remained implanted. This functional determination differs qualitatively from a RPA reference standard. We assumed that determination of a functioning lead excluded a clinically significant fracture, but not a connection problem.
Study inclusion criteria
Leads were included only if a save-to-disk or remote-monitoring file had adequate data preceding the patient's clinical presentation. Fractures were consecutive leads identified from the Medtronic RPA database that were explanted because a physician diagnosed pace-sense fracture. Separate database queries were performed for Fidelis and non-Fidelis leads. Leads with connection problems determined by RPA were consecutive leads identified from the RPA database that met study criteria and were extracted because a pace-sense lead fracture was misdiagnosed by a physician due to high impedance or noise (7,8) oversensing (Figs. 2A and 2B). We did not analyze leads with connection problems that were explanted for reasons other than misdiagnosis of fracture. Clinically determined functioning leads included all functioning Fidelis leads that triggered the LIA's impedance alert in a previous report (3). Clinically determined connection problems included all surgically corrected connection problems that triggered the LIA's impedance alert in the same report.
The Medtronic ICDs in this study measure lead impedance once daily within the range 0 to 16,382 Ω, although the programmer and remote-monitoring software truncate the maximum displayed value at 2,500 Ω. The impedance trend stores up to the last 14 daily measurements and up to 80 weekly minimum and maximum values. For impedance monitoring, the threshold for an impedance alert is programmable between 1,000 and 2,500 Ω. For LIA, the threshold is based on the average of the most recent measurements (3).
Impedance values were extracted from impedance trends and analyzed using Matlab (version 7.7, Matlab, Natick, Massachusetts). For leads subjected to RPA, we analyzed trends up to the date of explantation. For functioning leads, we analyzed trends up to the first interrogation after the impedance alert.
We selected variables related to impedance trends and oversensing. Using the development set, we identified discriminatory thresholds for each variable. Then, we evaluated various combinations of criteria to construct an algorithm.
Variables were selected based on clinical conjectures. Five variables were related to the shape of impedance trends, and 2 were related to their time course. Changes at the electrode-myocardial interface typically cause gradual rises to stable, high impedances, but high impedances in fractures and connection problems are abrupt and intermittent. Thus, we analyzed: 1) abrupt versus gradual impedance rise; and 2) stable high impedance. Open circuits resulting in very high impedance occur with fractures but not connection problems; thus we analyzed: 3) maximum impedance. Incomplete insertion of the lead pin into the connector may result in long intervals of baseline impedance after an abrupt rise, but fractures do not show a consistent return to baseline impedance. Thus, we analyzed: 4) longest return to baseline after an abrupt rise. Connection problems typically have at least 1 high impedance measurement over weeks to months. Conversely, 5) characteristic, nonphysiological noise oversensing (7,8) without an impedance rise indicates a fracture. Connection problems are often identified perioperatively, whereas fractures develop late after lead implant. Thus, we analyzed: 6) time from last ICD surgical procedure to first oversensing or abrupt impedance rise, and 7) time from lead implant to first abrupt impedance rise.
Based on analysis of the development set, we used these definitions related to impedance trends: “Initial impedance” was the first, saved minimum value. An “impedance rise” was defined as any (daily or weekly) measurement that exceeded the initial impedance by 350 Ω or 60%. “Baseline impedance” was defined as the median of weekly minimum values for the 3 previous weeks, updated weekly until an abrupt impedance rise occurred. Thereafter, it remained constant at its last value. An “abrupt impedance rise” was any impedance measurement that exceeded the current, moving baseline by 350 Ω or 60% (Figs. 3A,3B, and 3D). An impedance rise that did not qualify as abrupt was defined as “gradual” (Fig. 3C). “Return to baseline” impedance after an abrupt rise was defined as a daily impedance measurement that returned to within 350 Ω and 60% of baseline impedance after an abrupt rise (Fig. 3A). An abrupt impedance rise was considered stable if the maximum value did not return to baseline for the 2 weeks after the rise and the minimum value exceeded 65% of the maximum throughout (Fig. 3D).
The study's prospectively defined endpoints were based on the algorithm's performance on the test set: the fractions of correctly classified fractures and connection problems. Data are presented as mean ± SD. Ninety-five percent confidence intervals (CIs) were calculated using the binomial distribution. Non-normal, continuous variables were compared using the Wilcoxon rank-sum test. Reported p values have not been adjusted for multiple comparisons and are not study endpoints. A p value <0.05 was considered significant for single comparisons.
The Online Appendix provides data for individual leads in the development set. Figure 4 shows the algorithm developed from these data. A lead is classified as having a connection problem if it fulfills 1 of 3 criteria: 1) either noise oversensing or an impedance rise occurs in the first 30 post-operative days after lead implantation or generator change (Step 1); 2) an abrupt impedance rise is followed by a return to baseline longer than 45 days (Step 5); or 3) an abrupt impedance rise occurs within 200 days of lead implantation (Step 6). A lead is classified as a fracture if it fulfills 1of 2 criteria: 1) noise oversensing occurs without an impedance rise (Step 3B); or 2) an abrupt impedance rise exceeds 10,000 Ω (Step 4). In addition, a lead with an abrupt impedance rise is classified as a fracture if it is not given another classification (Step 6). A lead is classified as functioning if it fulfills both of 2 criteria: 1) noise oversensing does not occur; and 2) either the impedance rise is gradual or high impedance is stable after an abrupt rise. Figure 5 shows a conceptual view of the algorithm.
The development set
Two features discriminated functioning leads with high impedance from fractures and connection problems. First, gradual impedance rises or stable high impedance after an abrupt rise occurred in 9 of 21 functioning leads (43%, 95% CI: 24% to 63%) versus 0 of 70 fractures and connection problems (0%, 95% CI: 0% to 5%, p < 0.0001). Second, maximum impedance >2,500 Ω did not occur in any of 21 functioning leads (0%, 95% CI: 0% to 15%) versus 41 of 70 fractures or connection problems (59%, 95% CI: 47% to 69%, p < 0.0001).
The algorithm correctly classified 40 of 40 fractures (100%, 95% CI: 91% to 100%) and 28 of 30 connection problems (93%, 95% CI: 79% to 98%). Of the 21 functioning leads, the algorithm classified 11 as connection problems (52%), 9 as functioning (43%), and 1 as a fracture (5%). Including functioning leads and connection problems, overall accuracy for rejection of lead fracture was 48 of 51 (94%, 95% CI: 84% to 98%). Development set results represent “best case” performance.
Table 1 summarizes clinical presentations. Inappropriate shocks were more common in fractures than connection problems were (51 of 70, 73% vs. 7 of 30, 23%; p < 0.0001). Table 2 shows the number of patients in each group whose values exceeded thresholds for each variable established from the development set. Figure 6 provides individual patient data.
An impedance rise occurred in 43 of 70 fractures (61%, 95% CI: 50% to 72%) and 28 of 30 connection problems (93%, 95% CI: 79% to 98%). All impedance rises were abrupt and all but 1 were stable. The stable rise corresponded to a connection problem.
Figure 6A shows that impedances greater than the threshold 10,000 Ω occurred only in fractures: 15 of 70 fractures (21%, 95% CI: 17% to 40%) versus 0 of 30 connection problems (0%, 95% CI: 0% to 11%, p = 0.0119). Figure 6B plots individual values for the duration of return to baseline after an abrupt rise. Return longer than the threshold of 45 days occurred in 0 of 70 fractures (0%, 95% CI: 0% to 5%) versus 13 of 30 connection problems (43%, 95% CI: 27% to 60%, p < 0.0001).
For leads presenting after post-operative day 30, noise oversensing without an impedance rise occurred in 0 of 25 connection problems (0%, 95% CI: 0% to 13%) versus 27 of 70 fractures (39%, 95% CI: 28% to 50%; p = 0.0006). In these 27 fractures, the median value of maximum impedance was 528 Ω (range 443 to 1,224 Ω). Oversensing without an impedance rise was more common in non-Fidelis fractures (21 of 30, 70%) than in Fidelis fractures (6 of 40, 15%, p < 0.0001). Two connection problems presented with oversensing but no impedance rises, on post-operative days 1 and 14, respectively.
Time from Surgery to First Impedance Rise
Figure 6C plots the interval from the last procedure to the first impedance rise for individual patients with impedance rises. Considering all connection problems with impedance rises (n = 28), the first rise occurred within 30 days of the last surgery in only 3 leads (11%), within the first 90 days in 9 leads (32%), from 91 to 180 days in 5 leads (18%), and after 180 days in 13 leads (46%). The last surgery was the ICD system implantation in 98 patients (98%) and generator change in 2 patients (2%).
In Figure 4, percentages of leads classified correctly and incorrectly (in parentheses) are shown at each step. The algorithm correctly classified 70 of 70 fractures (sensitivity: 100%, 95% CI: 95% to 100%) and 26 of 30 connection problems (specificity: 87%, 95% CI: 70% to 95%). Rates of correct classification were similar in the development and test sets for both fractures (100% vs. 100%) and connection problems (93% vs. 87%, p = 0.6707). In the test set, the positive predictive value for fracture was 70 of 74 (95%, 95% CI: 87% to 98%); the negative predictive value was 26 of 26 (100%, 95% CI: 87% to 100%).
Figure 7A shows the impedance trend for a typical case classified as a connection problem. There is an early abrupt rise, a highly erratic impedance trend, and a long return to baseline. Figures 7B and 7C show impedance trends of the functioning lead that was misclassified as a fracture. Figure 7B shows a late, gradual rise interrupted by an abrupt increase, truncated by the end of data collection 20 days later. The lead was misclassified based on the late abrupt rise. Figure 7C, from a subsequent interrogation, shows a long return to baseline (504 days) that would have been classified as a connection problem. It was not included because we analyzed leads at time of the first data file that showed an impedance rise, resulting in worst-case for misdiagnosis.
All 5 misclassifications of connection problems as fractures occurred because “fracture” is the default classification when the first impedance rise occurs more than 200 days after implant. Both leads in the development set and 2 of 3 in the test set had similar trends: They were extracted shortly after the impedance rise, preventing subsequent evaluation for stable impedance. The remaining lead had highly variable impedances for 210 days between the impedance rise and extraction.
Figure 7D shows the connection problem that was classified as a functioning lead. Oversensing did not occur. There is an abrupt increase during a more gradual impedance rise. The impedance rise is classified as stable because the minimum impedance exceeds 65% of the maximum throughout.
Early diagnosis of ICD lead fractures is important to reduce morbidity from loss of pacing, inappropriate ICD shocks, and/or ineffective treatment of ventricular tachyarrhythmias (1–4); however, overdiagnosis may result in unnecessary lead replacement risking morbidity and rare deaths (5). ICD diagnostics facilitate identification of fractures (1–4,11), but individual diagnostics are nonspecific (3,8,11). Furthermore, there are no evidence-based criteria for discriminating fractures from other causes of high impedance and/or noise oversensing. The principal finding of our study is that diagnostic accuracy can be improved by combining multiple characteristics of impedance trends and the relationship between these characteristics and noise oversensing. This approach can discriminate fractures from connection problems that were misdiagnosed as fractures.
Discriminating fractures from connection problems
Connection problems include loose set screws (9), air trapped in the header escaping through seal plugs (12,13), weakened ICD header bonds (14), adapter problems (15), and incomplete contact between the lead pin and header (10). The literature is limited to reports of a few or single cases, which presented intraoperatively or perioperatively.
In the present study, most connection problems that were misdiagnosed as fractures presented after the perioperative period; 46% presented more than 6 months later. The Online Appendix shows that connection problems that were diagnosed correctly after LIA alerts also presented primarily after the perioperative period. Thus, connection problems are an important cause of late impedance rises. However, fractures did not occur early after implantation. The earliest presented at 256 days.
Our data support 3 criteria related to impedance trends for discriminating fractures from connection problems. First, very high impedance, indicating an open or nearly open circuit, is specific for fracture. Second, a long return to baseline after an abrupt rise occurs only in connection problems. It may occur if stress occasionally alters a mostly adequate connection with incomplete insertion of the pin. This criterion is helpful when the first impedance rise occurs late after surgery or an early rise is insufficient to trigger an impedance alert, but not at the first impedance rise. Third, fractures may result in oversensing without abnormal impedance, but connection problems do not if multiple measurements are made over time. Such fractures may represent breaks involving 1 or a few filars of the multifilar cable conductor to the ring electrode (16); the fractured filars may generate noise signals while the remaining filars conduct sufficiently to prevent an impedance increase at the measured resolution.
Discriminating fractures from functioning leads
Gradual, stable impedance rises in normally functioning leads are presumed to occur at the electrode-myocardial interface. We are unaware of any report that characterizes impedance rises at the electrode-myocardial interface for ICD leads or of evidence-based criteria for discriminating such changes from fractures and connection problems. For example, Kallinen et al. (4) reported 3 patients with normally functioning Fidelis leads who had gradual impedance increases. The leads were replaced because of concern that they might be fractured. By definition, functioning leads do not have noise oversensing.
Our algorithm assumes that noise oversensing is identified accurately (3,8). In reviewing explanted leads, we found three normal leads without impedance rises that were misdiagnosed as fractures because rapid, physiological oversensing was misinterpreted as noise (Fig. 2C) (3,8). These leads were not included in our study.
We found that all impedance rises were abrupt in confirmed fractures or connection problems; stable high impedance occurred in only 1 connection problem, without clinical consequence. In contrast, 43% of functioning leads had either only gradual increases or increases to stable high impedance, suggesting a different mechanism of impedance rise. From a clinical perspective, these leads do not require prompt replacement.
With sufficient follow-up, all functioning leads with abrupt, unstable impedance increases showed impedance trends typical of connection problems. Unfortunately, we do not have radiographs to determine if lead pins were inserted incompletely. All functioning leads also had impedances less than the maximum displayed by the programmer or remote-monitoring (2,500 Ω). This raises the possibility that physicians' decisions to replace functioning leads may be influenced by the range of displayed data.
Because functioning leads remained implanted, we could not analyze them using an RPA reference standard. Thus, we do not know if their impedance increases occurred because of changes at the electrode-myocardial interface, subclinical fractures, or connection problems. We do not know how well our criteria apply to leads or connections from other manufacturers or to other rare types of header or connection problems (9,12–14). We did not analyze criteria for high-voltage fractures. These comprise approximately 10% of Fidelis fractures (17), with or without concomitant pace-sense fractures. The Medtronic RPA database is designed for analyzing returned leads, not for determining the reasons leads were extracted; thus, we do not know what fraction of extracted leads are misdiagnosed as fractures. Algorithm performance may differ when it is applied to implanted leads rather than explanted leads.
Outside the perioperative period, noise oversensing with normal impedance trend indicates a fracture rather than a connection problem. If noise oversensing and abrupt impedance rise occur, our algorithm assists in differentiating fractures from connection problems. A connection problem may be diagnosed by radiography (10), but the orientation of the pulse generator relative to the x-ray beam may preclude definitive diagnosis. Intraoperatively, the ICD should be inspected for incomplete insertion of the lead pin into the header, loose set screws, and other header problems (14) before disconnecting the lead.
This study provides limited data supporting follow-up without operative intervention for leads with a gradual impedance rise. For leads with an abrupt rise and no other evidence of ICD system malfunction, operative intervention is indicated if the algorithm indicates a fracture; however, we cannot provide an evidence-based recommendation if the algorithm indicates a connection problem. About one-half of high-impedance, functioning leads had impedance trends indistinguishable from connection problems with a maximum impedance <2,500 Ω.
Finally, this study highlights uncertainties in clinical diagnosis that may influence reported rates of lead fractures. Studies of lead fractures should indicate the specific criteria used for diagnosis.
The authors thank Linda Chaman, MPH, for manuscript preparation.
For data for individual leads in the development set, please see the online version of this paper.
Preventing Overdiagnosis of Implantable Cardioverter-Defibrillator Lead Fractures Using Device Diagnostics
This study was funded by Medtronic, Inc. Dr. Swerdlow has received consulting fees from, has served on the Speakers' Bureau of, and has received scientific study/trial support from Medtronic, Inc. Mr. Sachanandani, Mr. Gunderson, and Mr. Ousdigian own stock in Medtronic, Inc. Mr. Sachanandani, Mr. Gunderson, Mr. Ousdigian, and Mr. Hjelle are Medtronic, Inc., employees. Dr. Ellenbogen has received honoraria from Medtronic, Inc., Boston Scientific, St. Jude Medical, Sorin Biomedical, and Biotronik; has received scientific study/trial support from Medtronic, Inc., St. Jude Medical, and Boston Scientific; has received consulting fees from St. Jude Medical; has received medical fellowship support from Medtronic, Inc., and Biotronik; and has served on the Speakers' Bureau for Medtronic, Inc., and Boston Scientific.
- Abbreviations and Acronyms
- confidence interval
- implantable cardioverter-defibrillator
- lead integrity alert
- returned product analysis
- Received September 4, 2010.
- Revision received November 17, 2010.
- Accepted December 7, 2010.
- American College of Cardiology Foundation
- Swerdlow C.D.,
- Gunderson B.D.,
- Ousdigian K.T.,
- et al.
- Parkash R.,
- Crystal E.,
- Bashir J.,
- et al.
- Gunderson B.,
- Patel A.,
- Bounds C.
- Kowalski M.,
- Ellenbogen K.A.,
- Wood M.A.,
- Friedman P.L.
- Gunderson B.D.,
- Patel A.S.,
- Bounds C.A.,
- Shepard R.K.,
- Wood M.A.,
- Ellenbogen K.A.
- Samsel T.