Implantable cardioverter-defibrillator (ICD) therapy is proven to reduce mortality (1- 5). However, inappropriate shocks for atrial arrhythmias with rapid ventricular conduction (6- 7) or for abnormal sensing (8- 10) results in multiple adverse effects (11- 14) including impaired quality of life (15), psychiatric disturbances (16), and even provocation of nonfatal (17) or fatal (18) ventricular arrhythmia. Although inappropriate shocks have been studied in some ICD groups (19- 20), no reports have detailed inappropriate ICD therapy in a pure primary prevention group like those in the MADIT II study.

The MADIT II protocol permitted implantation of U.S. Food and Drug Administration–approved single-chamber or dual-chamber Guidant ICDs. Each ICD stored intracardiac electrograms for arrhythmia episodes. Moreover, each unit offered 2 algorithms intended to minimize inappropriate shocks: 1) “stability,” evaluating the regularity of the tachyarrhythmia; and 2) “sudden onset,” the degree to which the arrhythmia began suddenly versus gradually (21). The dual-chamber devices provided additional algorithms evaluating the atrial rate (22). The ICD programming, including such discriminator usage, was left to the discretion of the investigators using standard clinical practice.

The MADIT II study randomly allocated 742 patients to the ICD arm, but 1 withdrew consent and 22 never received an ICD, leaving 719 that could be evaluated for inappropriate shocks. The ICDs were interrogated quarterly and after ICD shocks. The ICD therapy was defined as either antitachycardia pacing (ATP) or ICD shock. Two investigators (J.P.D, W.Z.) categorized the rhythm prompting ATP or shock using the stored electrograms. Any ICD therapy not delivered for VT or VF was deemed inappropriate, and the rhythm triggering therapy categorized as: atrial fibrillation or atrial flutter (AF), supraventricular including sinus tachycardia (SVT), or inappropriate sensing using published criteria (9,19). A small percentage of rhythms triggering ICD therapy (2.2%) were unclassified because of missing or incomplete data. An episode’s termination was defined by the ICD re-detecting sinus rhythm and thus could include more than 1 shock (and/or ATP bursts). As done previously in the AVID (Antiarrhythmics Versus Implantable Defibrillators) study, a subsequent episode beginning <5 min after episode termination was ignored for this analysis (19). Thus, an inappropriate shock episode was defined as an episode during which one or more inappropriate shocks occurred; a separate ICD episode of the same type (inappropriate or appropriate) occurring <5 min later was not counted.

Clinical characteristics were compared using the Wilcoxon rank sum test for continuous variables. The chi-square test was used for dichotomous variables, except the Fisher exact test was used when 25% or more of the cells had expected cell counts fewer than 5 (sudden and nonsudden cardiac death). The Kaplan-Meier life-table method was used to graphically display the time to first event and calculate the cumulative event rates for each group and within each group by risk factors. The results were compared using the log-rank statistic.

Using printouts or ICD discs downloaded after an inappropriate shock prompted interrogation, the major tachyarrhythmia detection settings were compared for the 83 inappropriate shock patients to 83 randomly chosen subjects without inappropriate shock. The latter group was matched with the inappropriate shock group for the presence or absence of 3 variables associated with inappropriate shock, prior atrial fibrillation, smoking, and appropriate shock. Theoretically 8 combinations of these 3 variables existed. In reality, only 7 combinations of these 3 variables contained patients who had received an inappropriate shock. Next, from the pool of patients who did not receive an inappropriate shock, we randomly chose the same number of patients from each of the 7 groups defined by having the same combination of these 3 characteristics. A stratified difference-between-proportions test, with strata weights inversely proportional to variances, was used to derive p values for the binary variables. The Mann-Whitney rank statistic, likewise stratified, was used for the numerical variables. Unstratified versions were also carried out to evaluate consistency of findings.

Using the 83 patients who did experience an inappropriate shock and the 636 patients who did not experience inappropriate shocks, Cox proportional hazards regression models were developed to determine what factors predicted time to inappropriate shock. Similarly, again using the patients with and without inappropriate shock, Cox proportional hazards regression models were developed to determine whether the occurrence of inappropriate shocks predicted time to all-cause mortality. Variables from the clinical characteristics table that met the criteria of p value <0.20 were considered as candidates for the regression model of inappropriate shock. For the inappropriate shock end point, appropriate shock was modeled as a time-dependent covariate.

Both appropriate and inappropriate shock were modeled as time-dependent covariates in the all-cause mortality model. Additional commonly used risk factors were included in this model. Follow-up time began at ICD implantation. Unclassified rhythms (n = 13, 2.2% of total shocks) were not included in this analysis.

The statistical software used for the analyses was SAS version 9.13 (SAS Institute Inc., Cary, North Carolina). A 2-sided probability value of <0.05 identified statistical significance.

One or more inappropriate ICD shocks occurred in 83 of 719 patients (11.5%). After 2 years of follow-up, patients had a 13% likelihood of having experienced one or more inappropriate shocks (Figure 1).(Table 1) shows patient experience with inappropriate and appropriate shocks. Thirty-two patients had more than 1 inappropriate shock episode (Table 1), ranging up to a maximum of 16 inappropriate shock episodes during 18 months of follow-up for 1 patient. Patients experiencing an inappropriate shock had a mean number of 2.2 ± 2.5 inappropriate shock episodes.

(Table 2) delineates the inappropriate shocks by episode rather than by patient and shows the subclassification of inappropriate shock by mechanism. As noted above, shock episodes were counted and multiple shocks within an episode were not tallied, nor were episodes caused by the same mechanism beginning within 5 min of a prior episode (19). (Table 2) thus reflects separate ICD shock episodes rather than the total number of shocks. Atrial fibrillation or atrial flutter was the most common mechanism for inappropriate shock, followed by SVT, with inappropriate sensing the least common mechanism for inappropriate shock in the MADIT II study. The time-dependent occurrence of a patient’s first inappropriate shock caused by any of the 3 mechanisms is shown in (Figure 2), illustrating that the time course of the occurrence of the 3 types of inappropriate shocks is similar. In the cohort of 719 patients with ICDs included in this analysis, 79 (11.0%) of patients had one type of inappropriate shock, 3 (0.4%) had 2 types of inappropriate shock mechanism, and 1 (0.1%) of the patients experienced ICD shock for all 3 types of inappropriate therapy mechanism.

We analyzed the heart rate at the time of a patient’s first inappropriate shock for either AF or SVT. Inappropriate sensing episodes were excluded from this analysis only, because the true ventricular rate was actually normal for these, and not what the device reported. The mean ventricular rate triggering inappropriate shock for AF or SVT was 174 ± 22 beats/min (Figure 3). The number of patients experiencing inappropriate shock at a given rate is shown on the y axis. For instance, 3 patients had a (first) inappropriate shock for atrial fibrillation with a heart rate of 200 beats/min, and 1 patient had a heart rate of 200 beats/min because of SVT prompting inappropriate shock. At the time of inappropriate shock, the ventricular rate exceeded 160 beats/min in 78% of episodes.

Considering any inappropriate therapy, that is inappropriate shock or ATP, this occurred in 100 of 719 (13.9%) patients (Figure 1B), of whom 17 patients experienced inappropriate ATP therapy without having at least 1 inappropriate shock during follow-up.

Inappropriate shock recipients differed statistically from nonrecipients (Table 3) for 3 baseline clinical characteristics: 1) prior atrial fibrillation (18.1% vs. 7.4%, p = 0.001), 2) smoking history (89.2% vs. 78.4%, p = 0.022), and 3) diastolic blood pressure ≥80 mm Hg (37.3% vs. 26.3%, p = 0.033). Considering interim events, appropriate therapy occurred more commonly in the inappropriate shock group than in those without inappropriate shocks (42.2% vs. 21.1%, p < 0.001). Notably, the proportion of patients receiving a dual-chamber ICD in the inappropriate shock group versus those without inappropriate shock was not appreciably different (Table 3), suggesting that dual-chamber ICD systems were not less prone to inappropriate shocks in MADIT II.

Cox proportional hazards analysis found the baseline characteristics of prior atrial fibrillation (hazard ratio [HR] 2.90, p < 0.01), smoking history (HR 2.18, p = 0.03), and diastolic blood pressure ≥80 (HR 1.61, p = 0.04) independently predicted the occurrence of inappropriate shock (Table 4). Considering time-dependent factors, prior interim appropriate shock predicted the occurrence of an inappropriate shock (HR 2.25, p = 0.03), as did interim atrial fibrillation (HR 3.45, 95% confidence interval 1.55 to 7.69, p < 0.01).

The SVT–VT discriminator stability function had been activated less frequently in inappropriate shock recipients (17%) than in those not receiving inappropriate shocks (36%, p = 0.030). Among dual-chamber devices, a trend toward the V>A criterion being used less frequently was seen in patients with inappropriate shocks (31% vs. 50%, p = 0.054). We did not observe other significant differences in these parameters, such as the lowest zone detection rate (Table 5).

Inappropriate shock recipients tended to have a higher percent total mortality than those patients not experiencing inappropriate shocks (16.9% vs. 12.9%, p = 0.317, Table 3). Because appropriate shock occurrence was one predictor of inappropriate shocks (Table 4), Cox proportional hazards regression analysis was used to evaluate the independent survival impact of inappropriate ICD shocks (Table 6). In multivariate analysis, mortality was predicted by blood urea nitrogen >25 (HR 2.07, p < 0.01), absence of beta-blocker therapy (HR 1.64, p = 0.02), and interim congestive heart failure (CHF) hospitalization (HR 4.23, p < 0.01) Because inappropriate and appropriate shocks were intertwined in their occurrence, we examined them independently and together (Table 6). The occurrence of both inappropriate and appropriate shock was associated with an over 4-fold increase in probability of mortality for a given follow-up period (HR 4.08, p < 0.01). The occurrence of an inappropriate shock was associated with an HR for mortality of 2.29 (p = 0.02). Prior appropriate shock alone portended an HR of 3.36 for mortality (p < 0.01). On the other hand, neither appropriate nor inappropriate ATP was associated with a significant mortality increase (Table 6). Inappropriate shocks were not a predictor of subsequent CHF hospitalization (data not shown), although appropriate shocks were identified as a predictor of CHF events as previously reported (23).

In the MADIT II study, inappropriate ICD shocks were common, occurring in 11.5% of patients and with a cumulative 1- and 2-year event rate of 10% and 13%, respectively (Figure 1A). Overall, 184 inappropriate shock episodes occurred compared with a total of 393 appropriate shock episodes. An AF caused 44.0% of inappropriate shock episodes; SVT and inappropriate sensing were less common.

Frequent inappropriate ICD shocks and ATP occurred in other series (7,19- 20), such as the AVID study with 20% of patients experiencing an inappropriate shock or ATP in follow-up and 9% by 2 years of follow-up (19). Enrolling patients with prior sustained VT or VF, a higher proportion of patients experienced appropriate therapy, so the overall percent of events that were inappropriate was lower in the AVID study. Swerdlow et al. (24) had previously predicted that the proportion of inappropriate shocks of the total shock count would be higher in primary prevention ICD patients because of a lower appropriate shock rate. Enrolling a diverse ICD population, the Pain Free Study found that 15% of patients experienced inappropriate therapy (20), with the proportion of all events being inappropriate trending higher in the primary prevention subset than in the secondary prevention group (46% vs. 34%, p = 0.09) (20). In the Multicenter InSync ICD Randomized Clinical Evaluation (MIRACLE ICD) study, 32% of primary prevention biventricular ICD patients had inappropriate detections, although not all of these led to shocks, and unlike in the MADIT II study and most studies, inappropriate detections for sinus tachycardia exceeded atrial arrhythmias (19- 20,25- 26).

Inappropriate shock patients more commonly had a history of atrial fibrillation, smoking, and/or diastolic hypertension and were more likely to also have had a prior appropriate ICD shock (Table 3). Alter et al. (27) reported that young age and nonischemic cardiomyopathy predicted inappropriate shocks. Other investigators have found that prior history of atrial fibrillation and appropriate therapy were predictors of inappropriate shocks, similar to our findings (26). Because AF was the most common inappropriate shock mechanism, its association as a predictive factor is expected. Smoking was recently found to increase the incidence of both appropriate and inappropriate shocks in the MADIT II study, possibly because of a myriad of adverse consequences such as sympathetic stimulation, increased platelet reactivity, vasoconstriction, endothelial dysfunction, and tachycardia (28). Hypertension is a potent risk factor for AF and may act in this way to promote the likelihood of an inappropriate shock. The link between appropriate and inappropriate therapy likely stems from a combination of several factors: 1) ventricular arrhythmia or its treatment (ATP or shock) provoking atrial fibrillation (29- 30); 2) inappropriate therapy causing VT, that is, proarrhythmia (18,27); 3) a common factor or factors predisposing to both VT/VF and AF or SVT; or 4) incorrect categorization of some appropriate episodes in a given patient as inappropriate or vice versa.

The prior occurrence of an inappropriate shock was associated with a doubled risk of total mortality in this study even after accounting for the known association between appropriate shock and increased mortality (23) (Table 6). Possible explanations for the increased mortality in the cohort with inappropriate shocks include: 1) potential direct mechanical, arrhythmic, or hemodynamic adverse effect of the shocks themselves, such as fatal proarrhythmia (18,27) or other effects (11); and 2) baseline characteristics or interim events, such as AF, causing both inappropriate shocks and an increased risk of mortality. Because shocks bore a greater association with mortality than ATP (Table 6), explanation one may be favored. However, an alternative inference is that a more recalcitrant, persistent, rapid, or recurrent arrhythmia may have led to a shock, whereas a brief flurry of the arrhythmia may have led to ATP and not required a shock.

We found a statistically different, lower utilization rate of the stability SVT–VT discriminator in patients with inappropriate shocks compared with patients not having inappropriate shocks, raising the possibility that more aggressive use of the available SVT discriminators could reduce the incidence of inappropriate detection. Stability was used in a minority of the overall population (Table 5), likely reflecting standard clinical practice at the time of this study. The SVT–VT discriminator usage carries the theoretical risk of underdetection of true ventricular arrhythmias, although current data suggest that underdetection due to stability usage is infrequent (9,21,31- 32). Nevertheless, the stability discriminator’s effectiveness in preventing therapy for atrial fibrillation is markedly reduced at rates above 170 beats/min (9). In fact, (Figure 3) shows that most of the rapidly conducted atrial fibrillation episodes causing inappropriate shocks were indeed 170 beats/min or faster. Thus, even if this discriminator had been uniformly programmed, it may not have had a major impact. Other parameters were not statistically different in the 2 groups, further arguing against programming aberrations explaining the inappropriate shocks (Table 5). Although programming a higher detection rate likely reduces the detection of atrial arrhythmias and misinterpretation of them as ventricular arrhythmias, it is clear that programming too high a rate cutoff could lead to underdetection of relatively slow monomorphic VTs or underdetection of polymorphic ventricular arrhythmias or ventricular fibrillation due to intermittent undersensing (31,33- 34). Although there is an incidence of sudden death in any ICD population, the extent to which this could have been prevented by the ICD or whether some of these sudden deaths are caused by underdetection of ventricular arrhythmias remains unknown.

Dual-chamber ICD systems were used approximately as commonly in patients who experienced an inappropriate shock (38.6%) as in those who did not (44.4%, Table 3). Prior studies have often failed to show a benefit of dual-chamber devices in preventing inappropriate detections of supraventricular arrhythmias (22,26,35- 36). The Detect Supraventricular Tachycardia Study (DETECT SVT) recently showed a reduction in the percent of total SVT episodes that were inappropriately classified as VT in the dual-chamber detection group (30.9%) than the single-chamber group (39.5%) (37). However, this means that almost one-third of the SVT episodes were still misclassified despite the dual-chamber criteria, and furthermore that total inappropriate shocks were not reduced in the dual-chamber arm (37- 38). Future enhancements of SVT–VT discrimination systems may yield improvements (9,33,38- 42).

The role of medications in preventing inappropriate shocks is mixed, with a prior study failing to find beta-blocker therapy protective from inappropriate shocks in the MADIT II study (43). However, sotalol and amiodarone have reduced inappropriate shocks by as much as 78% in a secondary prevention population (27,44- 45). Recently introduced continuous wireless ICD monitoring (46- 47) could reduce inappropriate therapies.

Total shocks were not counted, but rather ICD episodes containing one or more shocks were tallied. The potential additional prognostic importance of the total number of shocks cannot be evaluated. The ICD programming was not protocol specified, and although the VT cutoff rate was collected prospectively, the other parameters analyzed (Table 5) were obtained retrospectively, and the difference in stability programming must be interpreted cautiously. Moreover, parameter changes and possible effects on inappropriate shocks could not be analyzed in a time-dependent fashion. Electrogram interpretation to adjudicate appropriate versus inappropriate shocks was based on available data, which differs for single-chamber and dual-chamber systems (48). Furthermore, errors in classification may occur (49). This MADIT II substudy’s findings may not pertain to other ICD populations. Lastly, although quality of life data were collected in MADIT II and initial analysis has been performed (50), the association of inappropriate shocks and quality of life is not yet available.

Inappropriate shocks were common in the MADIT II study, as in other recently reported trials, occurring in 83 (11.5%) of the 719 MADIT II ICD patients and constituting 31.2% of all shock episodes. Smoking, atrial fibrillation, diastolic hypertension, and prior appropriate shocks were associated with an increased chance of inappropriate shock. Inappropriate shock occurrence was associated with increased probability of mortality in follow-up. Coupled with potential effects on quality of life, this association with increased mortality heightens the importance of efforts to reduce the occurrence of inappropriate shocks.

The investigators acknowledge the invaluable assistance of Ms. Jodie Palma in the review of inappropriate shock electrograms and device programming and of Mr. Thomas Ross in preparation of the manuscript.