Increasing age does not affect time to appropriate therapy in primary prevention ICD/CRT-D: a competing risks analysis

Introduction

The decision to implant an implantable cardioverter defibrillator (ICD) for a primary prevention indication in elderly patients can be complex due to the competing risk of death associated with increasing age. Multiple randomized controlled trials have demonstrated that ICD and cardiac resynchronization therapy defibrillator (CRT-D) are superior to pharmaceutical strategies in the prevention of SCD from ventricular arrhythmias.^1–6 However, elderly patients were under-represented in these studies. In contrast, real-world practice shows us that pacemakers and ICDs are predominantly implanted in the elderly as they have a high disease burden satisfying implant indications and a high incidence of sudden cardiac death.⁷ The UK has an ageing population and in the next 25 years it is predicted that 1 in 8 of the population will be aged 75 years and over—an increase of 64% from current numbers (United Kingdom Office of National statistics, 2014).

We therefore sought to evaluate the impact of age on the clinical outcomes in a primary ICD/CRT-D population.

Methods

A retrospective, multicentre analysis was performed. All patients aged 60 years and over at the time of implant with primary prevention ICD/CRT-D devices implanted between 1 January 2006 and 1 November 2014 were identified. Pulse generator replacements were included. Where a duplicate patient exists (i.e. had a de novo then had a replacement ICD during the study period) the most recent implant date was used, in order to maximize the proportion of elderly patients in the analysis. (In total, 17 duplicate patients removed.) We analysed the prospectively collected ICD follow-up database (Medtronic Paceart^® System) for patients collected in Southampton [392 patients (92.4% of total)]. The electrograms from the ICD follow-up clinic device interrogation were routinely scanned and uploaded onto the database. All entries from the date of implantation to primary outcome, death, or follow-up were reviewed. In the event of appropriate therapy being delivered, the electrograms were scrutinized by one of three un-blinded investigators to determine the rate of the arrhythmia and appropriateness of the therapy. For patients from the Bristol Heart Institute, 32 (7.6%) patients, ICD files were reviewed and data on appropriate therapy were collected. Printed electrograms were scrutinized, where available, by an un-blinded investigator.

Appropriate therapy was defined any anti-tachycardia pacing (ATP)/shock delivered for a ventricular arrhythmia with ≥200 b.p.m. across all age groups. Outcomes were categorized as follows: survival to follow-up with no therapy (T1), death prior to follow-up with no therapy (T2), delivery of appropriate therapy with survival to follow-up (T3) and delivery of appropriate therapy with death prior to follow-up (T4). Data on co-morbidities were also collected.

Programming

Programming was standardized according to hospital protocols. The University Hospital of Southampton NHS Foundation Trust's (UHS) protocol was implemented in July 2012 (Supplementary material online) prior to the publication of MADIT-RIT.⁸ Primary prevention ICD/CRT-Ds were programmed with three tachycardia zones (Zone 1 ventricular rates 165–167 b.p.m. to 200 b.p.m., Zone 2 ventricular rates 200 b.p.m. to 240–250 b.p.m., and Zone 3 ventricular rates >240–250 b.p.m.). Data on therapy delivered in Zone 1 were collected but not used as part of the analysis in order to standardize therapy rates with MADIT-RIT programming, high rate, and to reflect the notion that therapy delivered for lower rates might not necessarily be potentially life-prolonging and to standardize with the second centre, Bristol Heart Institute (Supplementary material online). Primary prevention ICD/CRT-Ds were programmed with two tachycardia zones (Zone 1 ventricular rates 200–240 b.p.m. and Zone 2 ventricular rates >240 b.p.m.) (Supplementary material online).

Statistical analysis

Patients were analysed according to age decile (Decile 1 = 60–69.9 years, Decile 2 = 70–79.9 years, and Decile 3 = 80 years and over). The primary outcome variable was time to appropriate therapy. The secondary outcome variable was all-cause mortality. Categorical variables are expressed as absolute numbers and percentages. Continuous variables were presented as mean standard deviation. Baseline characteristics were compared using independent sample t-test for continuous variables and χ² test for categorical variables. Last follow-up date was 1 January 2015. Timelines and cumulative incidence frequency curves were constructed. Sub-hazard ratios (SHR) for time to appropriate therapy when all-cause death was the competing risk was performed in all three age decile groups. Sub-hazard ratios for time to death were performed when appropriate therapy was the competing risk. Competing risks were modelled using the Fine–Gray model. Predictors of all-cause mortality were identified using Cox proportionate hazards model. The following variables were entered in the univariate analysis: age at implant, gender, device type, baseline estimated glomerular filtration rate (eGFR), baseline haemoglobin, presence of diabetes mellitus, and ischaemic aetiology. All the predictors had a P-value of <0.2–0.25 in the univariate analyses. All statistical analyses were performed using Stata 13 (StataCorp., College Station, TX, USA).

Results

In total, 424 cases were analysed. Baseline device and clinical characteristics are presented in Table 1. Overall 366 patients (86.3%) were male, 247 (57.3%) were CRT-D devices, and 98 (23.1%) were generator exchanges. The age decile groups are as follows: Decile 1, n = 189 (44.6%); Decile 2, n = 161 (38.0%); and Decile 3, n = 74 (17.5%).

Median follow-up time was 32.6 months during which time 44 patients (10.1%) received appropriate therapy (Supplementary material online). Of these 7 (1.6%) were ATP alone, the remainder being a combination of ATP and shock therapy. Thirty patients received inappropriate therapy [ATP or shock (6.9%)]. Implantable cardioverter defibrillators and CRT-D from all major device manufactures were used (Medtronic 62.4%, St Jude Medical 19.5%, Boston Scientific/Guidant 11.8 and 6.3%, and Sorin 6.3%). Overall, 73 patients died (17.2%). Of these, 20 were in Decile 1 (10.6%), 34 were in Decile 2 (21.1%), and 29 were in Decile 3 (39.2%).

Competing risks analysis

The SHR for the cumulative incidence of appropriate therapy according to age at implant was 1.00 (P = 0.851; 95% CI 0.96–1.04). Thus, there is no extra hazard in the cumulative incidence of appropriate therapy with every year increment in age. The SHR for cumulative incidence of death according to age at implant was 1.06 (P < 0.001; 95% CI 1.03–1.01). Thus, there is a 6% increase in the incidence of death with every year increase in age at implant. These finding remained unchanged when the type of device was included as a covariate. Cumulative incidence curves for the three age decile groups are shown in Figure 1.

Figure 1

Estimated cumulative incidence function for time to death prior to appropriate therapy (T2) and time to appropriate therapy (T3) in age Decile 1 (A), age Decile 2 (B), and age Decile 3 (C). Time is represented in months on the horizontal and axis cumulative incidence function on the vertical axis. In (A) (60- to 70-year-old group), at 36 months, the cumulative incidence function of death with no therapy is ∼0.1 (10%) and the CIF of appropriate therapy is similar, 0.1 (10%). In contrast, in the (C) (>80-year-old group), the CIF of death with no appropriate therapy at 36 months is much greater at ∼0.3 (30%) and the CIF of appropriate therapy is similar to the 60- to 70-year-old group at 0.1 (10%).

Open in new tabDownload slide

Cardiac resynchronization therapy defibrillator appeared to be associated with reduced rates of appropriate therapy and death. The SHR for the cumulative incidence of CRT-D on appropriate therapy when death is the competing event 0.73 (95% CI 0.41–1.31). The SHR for the cumulative CRT-D and on death when potentially life-prolonging therapy is the competing event 0.87 (95% CI 0.55–1.37).

Lifelines

Lifelines were constructed for all three age deciles (Figure 2). Patients are grouped by clinical outcome: survival to follow-up with no therapy (T1), death prior to follow-up with no therapy (T2), delivery of appropriate therapy with survival to follow-up (T3), and delivery of appropriate therapy followed by death prior to follow-up (T4). Each horizontal line represents a single patient, starting at the time of device implant and ending at follow-up date for T1 and T3 and date of death for groups T2 and T4.

Figure 2

Timeline of implantable cardioverter defibrillator/cardiac resynchronization therapy defibrillator outcomes according to Decile 1 (A), Decile 2 (B), Decile 3 (C), and Decile 3 with time of appropriate therapy represented by in (D). This is a unique graphical representation of the clinical outcomes of each patients in the study. Each horizontal line represents a single patient. Time in months is represented on the horizontal axis. Each time line starts at the time on device implant and follows the patient until study end (follow-up or death). Patients are grouped in clinical outcomes (T1–T3). This novel presentation of the data allows easy comparison of the outcomes of patients in each age category and illustrates how the proportion of patients who die prior to follow-up increases with age while there is little/no difference in the proportion who receive appropriate therapy between the groups. FU, follow-up.

Open in new tabDownload slide

Of those patients who did not receive any appropriate therapy, the proportion who died (T2) to those who survived to follow-up (T1) increased significantly with age: Decile 1 compared with the Decile 3 [20/178 (11.2%) vs. 24/64 (37.5%); P< 0.0001]; Decile 2 vs. Decile 3: [28/149 (18.8%) vs. 24/64 (37.5%); P= 0.006]. The difference between the Decile 1 and Decile 2 groups approached significant difference 20/178 (11.2%) vs. 28/149 (18.8%) (P= 0.08).

Of those patients who did receive appropriate therapy, there were no significant differences between those who died subsequently (T4) and those who survived to follow-up between the age decile groups. However, numbers in this category were relatively small.

Predictors of mortality and appropriate therapy

A univariate analysis of baseline device and clinical characteristics was performed to identify predictors of mortality using the following variables: gender, decile group, age at implant, ischaemic aetiology, baseline eGRF, baseline haemoglobin, and diabetes mellitus. Variables with a P-value of ≤0.1 were selected for further analysis in a Cox proportionate hazards model. Results of the regression model are presented in Table 2. Significant predictors were identified in the multivariate analysis with ischaemic aetiology, age at implant, and haemoglobin at implant but not with diabetes mellitus and eGFR at baseline.

A univariate analysis of baseline device and clinical characteristics was performed to identify predictors of appropriate therapy using the following variables: gender, decile group, age at implant, ischaemic aetiology, baseline eGRF, baseline haemoglobin, and diabetes mellitus. Variables with a P-value of ≤0.1 were selected for further analysis in a Cox proportionate hazards model. Results of the regression model are presented in Table 3. No significant predictors were identified in the multivariate analysis though, ischaemic aetiology and male gender were associated with appropriate therapy.

Discussion

Due to an ageing population, there is increasing interest in determining the outcomes of ICD therapy in elderly patients.^7–15 Older patients have a higher mortality. As the risk of death in ICD/CRT-D patients increases, it competes with the risk of patients receiving appropriate therapy. In this two centre study of 424 primary prevention ICD/CRT-D patients, we used a competing risks analysis¹⁶ to show that age does not influence the time to delivery appropriate therapy [SHR = 1.00 (P = 0.851; 95% CI 0.96–1.04)]. This study supports results from two national registry data sets and one pooled analysis of five primary prevention ICD trials that show similar rates of appropriate therapy across age brackets.^8,11,12 Other studies have performed somewhat similar analyses.^17,18 However, this study adds a novel perspective evaluating ICD outcomes. First, we have clearly defined ICD outcomes by categorizing patients into three broad categories: those who received appropriate therapy, those who have not received appropriate therapy (yet), and those who died before being able to receive appropriate therapy. Second, we used novel graphic illustrations of ICD outcomes (using timelines) which we feel easily convey the impact of age on ICD outcomes. The timelines were constructed using purposely written software code, and we are not aware of any other ICD studies that have chosen to portray results in such a way. Finally, whilst we feel that the use of a competing risk analysis is novel (but not unique), the result of the cumulative incidence function of appropriate therapy given death as a competing risk is 1.00 with very narrow confidence intervals is a remarkable result. This strongly suggests that increasing age does not change the time to receiving appropriate therapy and this reproduces the other much larger analysis that used competing risks analysis.¹² Our study adds to the comparatively small amount of data on outcomes in octogenarian ICDs/CRT-Ds.

Our study also supports the concept that it is the substrate (non-sustained VT, fractionated ECG, T-wave alternans, and QT variability etc.) that governs future ventricular arrhythmic episodes in a primary prevention population rather than age.^19–22 One of the clinical implications of this study is that elderly patients are at similar risk of ventricular arrhythmias as younger patients and therefore ICD treatment should not be declined on the basis of age alone provided the patient is not deemed to be at high risk of non-arrhythmic mortality. Clinical decision-making in implanting defibrillators in elderly patients can be challenging, and some of the issues in the area are well summarized by Barra et al.^23,24

Our study adds useful additional information. We characterized the clinical outcomes into four groups as follows: survival to follow-up with no therapy (T1), death prior to follow-up with no appropriate therapy (T2), delivery of appropriate (life-prolonging) therapy (T3), and death following appropriate therapy (T4). As patients age, mortality increases (by 6% per annum in this study). We devised a novel timeline illustration to demonstrate clearly the clinical outcomes in ICD recipients. Our study clearly demonstrates that the proportion of patients in T2 (death prior to receiving appropriate therapy) increases significantly as from those aged 60–69.9 years compared with the over 80-year-old group. Importantly, this appears not to be at the cost of those who receive appropriate therapy (T3) but at the cost of those who have not received appropriate therapy (T1). In other words, the pool of patients who are alive but who have not yet received appropriate therapy diminishes as age increases whereas those who receive appropriate therapy remains constant.

This raises two important questions. This first is how to identify those patients who are likely to die before appropriate therapy is delivered. This is important as defibrillator implantation in this category represents a potential wasted resource for healthcare organizations and unnecessary exposure of elderly persons to ICD-related morbidity at the end of their life. The second question is to ask whether the practice of withholding ICD therapy from elderly patients, who are likely to die prior to receiving appropriate therapy, will result in improved clinical outcomes in the elderly at a population level.

Using a Cox proportionate hazards model we identified the following variables as predictors of mortality: age at implant, ischaemic aetiology, baseline haemoglobin, and the presence of diabetes mellitus. There exist a number of clinical risk stratification tools that can be employed to attempt to identify this cohort: the FADES (Functional status, Age, Diabetes mellitus, Ejection fraction and Smoking) Risk score,²⁵ the MADIT-II (Multicenter Automatic Defibrillator Trial) Risk score,²⁶ and the Seattle Heart Failure Model-differential ICD (SHFM-D) risk scores.²⁷ The FADES Risk score stratifies patients into low, intermediate, or high risk of death without prior appropriate therapy and is based on the following clinical variables: New York Heart Association (NYHA) class >III, advanced age, diabetes mellitus, left ventricular ejection fraction <25%, and a smoking history. The strongest predictor was age ≥75 years (OR 2.95, 95% CI 1.7–5.1). In the MADIT-II risk score, a point is allocated if any of the following variables are present (NYHA >II, age >70 years, renal impairment, QRS duration >120 ms, and atrial fibrillation). A U-shaped pattern for ICD benefit was seen with very low and very high risk patients deriving no significant benefit from ICD therapy and only those with an intermediate mortality risk deriving benefit. The SHRM is a multi-variable risk model that predicts all-cause and cause-specific mortality in heart failure patients and the SHRM-D is a derivative of this based on the SCD-Heft trial using the following variables: age, gender, systolic blood pressure, ischaemic aetiology, NYHA class, ejection fraction, ACEi/ARB use, β-blocker use, statin use, digoxin use, carvedilol use, loop diuretic use, serum sodium, and serum creatinine. The maximum benefit derived from the ICD was seen in those with intermediate SHFM-D scores.²⁷ No benefit of ICD treatment was seen in the highest risk patients—those with an annual predicted mortality of >20%. These scores have been validated and the SHRM-D has been found to have the greatest discriminatory power compared with the FADES and MADIT risk scores.²⁷ Another study has used data from over 18 000 patients enrolled in the ICD Registry of the National Cardiovascular Disease Registries to identify six clinical variables to predict 1-year mortality in primary prevention ICD recipients.

While these risk scores are helpful in identifying those who may be at risk of death prior to appropriate therapy, we believe that they lack an easily measurable and important clinical determinant of cardiovascular and all-cause mortality: frailty. It has been recommended that frailty as well as health status, disability, and cognition are incorporated into risk prediction models when dealing with an elderly population.^29,30 Clinical studies are required, which will help identify which patients die prior to receiving appropriate therapy using clinical risk factors such as frailty and functional status indices as well as co-morbidity scores. The information from this study may help inform appropriate conversations when counselling patients on the merits and disadvantages of ICD therapy. It is always important to keep the individual in mind when making these important clinical decisions.

Limitations

The principal limitation is that these data comprise a retrospective rather than a prospective cohort. Detailed data on medication at the time of device implant is therefore not available and therefore the reader must bear this in mind when considering the results of the predictors of mortality and appropriate therapy. Device programming included a tachycardia zone for relatively slow ventricular rates. Appropriate therapy delivered in this zone was collected but not used in the analysis. This may have resulted in a lower than expected rate of therapy in this study as some of the ventricular arrhythmias may have otherwise accelerated and been treated in a faster tachycardia zone had they not been treated in the slower zone. However, the programming was consistent over all age groups and thus we do not believe that it affected the results of this study. In the Cox proportionate hazard model used to identify clinical predictors, diabetes mellitus was associated with a non-significant increased risk in mortality and ischaemic aetiology and male gender were non-significantly associated with appropriate therapy. This suggests that the study may have been underpowered to identify these predictors. However, the primary outcome of interest, appropriate therapy was collected and recorded prospectively in all patients; therefore, we are confident that these data are reliable.

Conclusion

Age has no impact on the time to appropriate ICD therapy, whereas it increases risk of death prior to therapy by 6% for every year increment. As the ICD population ages, the proportion who die without receiving potentially life-prolonging therapy increases. Studies are required to help identify clinical risk factors that are able to distinguish those elderly ICD patients who die prior to receiving appropriate therapy and those who do not in order to aid clinical decision-making and maximize the overall efficacy of ICD therapy in this patient population.

Supplementary material

Supplementary material is available at Europace online.

Funding

D.G.W. has received an educational grant from Medtronic, UK (eCATS agreement number: A 1102694).

Conflict of interest: D.G.W. was supported by educational grants from Medtronic Inc. J.M.M., P.R.R., and A.Y. have received honoraria and research grants from Medtronic, St Jude, Sorin, and Boston Scientific. J.M.M. has become the Chief Medical Officer of Boston Scientific cardiac rhythm management in Europe.

Acknowledgements

David Culliford, senior medical statistician at the University of Southampton, for his advice on aspects of the data analysis.

References