Key inclusion criteria were nonischemic HF (preferably determined using coronary angiography, but a normal computed tomographic angiogram or nuclear myocardial perfusion imaging was accepted), a left ventricular ejection fraction (LVEF) <35%, New York Heart Association (NYHA) functional class II or III (or class IV if cardiac resynchronization therapy [CRT] was planned), an n-terminal pro-B-type natriuretic peptide (NT-proBNP) concentration >200 pg/ml, and optimal treatment with medical therapy for HF. Patients with pre-existing pacemakers or CRT pacemakers could also be included in the study if they were willing to accept a potential upgrade. Patients fulfilling the indication for a CRT device received a CRT defibrillator (if randomized to the ICD arm) or CRT pacemaker (if randomized to the usual clinical care arm), and the decision to implant a CRT device had to be made before randomization. Key exclusion criteria were a resting heart rate >100 beats per minute in patients with permanent atrial fibrillation and renal failure that was treated with dialysis. A complete list of exclusion criteria is provided in the primary publication. ^,

In the present analysis, frailty was assessed using the Rockwood cumulative deficit approach, and this approach has been described in detail previously. ^,,,,,, Standard criteria for constructing a frailty index (FI) using this approach are the following: at least 30 items are required; items must be associated with health status; items must cover a range of body systems and not be isolated to 1 system; items must not be part of normal ageing or saturate too early (e.g., presbyopia), but they should generally increase with age. We created a 30-item FI, and these items were derived from medical history, vital signs, laboratory data, and the EuroQoL-5 Domain (EQ-5D) questionnaire (health-related quality of life measures, including functional status) ( Supplementary Table 1 ). A score was assigned for each nonmissing item, and the FI score was calculated as the sum of these scores divided by the total number of nonmissing items, with higher scores indicating greater frailty. Binary variables (e.g., a history of chronic obstructive pulmonary disease) were scored 0 of 1 (absent/present); ordinal variables (e.g., quality of life measures) were scored from 0 to 1 (in increments of 0.50, with a score of 1 indicating the greatest severity); and continuous variables (e.g., creatinine) were categorized and scored as 0 of 1 (normal/abnormal). Patients with >20% missing items were excluded. ^,,,, Patients were divided into the following 2 subgroups: FI <0.210 (FI class 1, i.e., nonfrail patients, as defined previously); FI >0.211 (FI class 2 to 3, i.e., frail patients).

In the main trial, patients were followed from randomization (the first patient was enrolled on February 7, 2008) until June 30, 2016. In the present study with extended follow-up, patients were followed from randomization until May 18, 2020, and no patients were lost to follow-up. As in the main trial, the primary outcome in the present study was death from any cause, and secondary outcomes were cardiovascular death, sudden cardiovascular death, and noncardiovascular death (the latter only in the analysis examining the association between FI class and outcomes). All outcomes were assessed by review of patients’ electronic and hard copy medical files and adjudicated by an event committee blinded to treatment allocation.

Although available only for the original follow-up duration, device complications and shock therapy among patients randomized to ICD implantation were also examined.

Baseline characteristics were summarized as frequencies with percentages or medians with 25 ^th to 75 ^th percentiles. Differences in baseline characteristics were tested using the chi-square test for categorical and the Wilcoxon test for continuous variables.

The association between FI class and outcomes was evaluated using Cox proportional hazards regression models, adjusted for age, sex, randomized treatment, center, heart rate, duration of HF, prior HF hospitalization, NYHA functional class, log of NT-proBNP, LVEF, and smoking status, were reported; variables which were part of the FI were not adjusted for, since the categorization of FI into the 3 classes were conditioned on these variables. The association between continuous FI and outcomes was examined using Cox proportional hazards regression models, adjusted for the variables mentioned above and a restricted cubic spline of FI with 3 knots.

The effect of ICD implantation versus usual clinical care according to FI class was examined with the Kaplan-Meier estimator, Aalen-Johansen estimator, and Cox proportional hazards regression models, stratified according to center. The effect of ICD implantation on outcomes was also examined according to continuous FI using fractional polynomial models. Fractional polynomial models were chosen because these models had lower Akaike Information Criterion scores compared with other models (e.g., restricted cubic spline and linear regression models). In addition, a “complete case” analysis of the effect of ICD implantation according to FI class was performed; this analysis included only patients with no missing data for any of the components of the FI.

To test for interaction between the treatment effect of ICD implantation and the FI, the Wald test was used for the Cox proportional-hazards models. The proportional hazards assumption was examined with log(−log[survival]) curves and scaled Schoenfeld residuals, and the assumption was not violated for any of the models.

Data were analyzed according to the intention-to-treat principle. All analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC) and STATA version 17.0 (College Station, TX). A p value of 0.05 was considered statistically significant.

Of the 1,116 patients randomized in DANISH, FI could be calculated for 1,109 patients. The distribution of the FI is shown in Supplementary Figure 1 , and the numbers of patients with missing data for individual components of the FI are shown in Supplementary Tables 2 and 3 . Mean FI was 0.199 (standard deviation, 0.090) and median FI was 0.200 (27 ^th to 75 ^th percentile, 0.133 to 0.260; range 0 to 0.560). In total, 618 (55.7%) patients were in frailty class 1 (FI <0.210, i.e., not frail), 361 (32.6%) in frailty class 2 (FI 0.211 to 0.310, i.e., more frail), and 130 (11.7%) in frailty class 3 (FI >0.311, i.e., most frail).

Baseline characteristics according to FI class are presented in Table 1 . Compared with patients with lower FI (i.e., less frailty), those with higher FI (i.e., worse frailty) were older and more likely to have cardiovascular and noncardiovascular comorbidities. They also had higher systolic blood pressure and BMI, but lower eGFR, sodium, and hemoglobin levels. Although patients with higher FI were more likely to have a longer duration of HF, higher NT-proBNP, worse NYHA functional class, and HF due to valvular heart disease and hypertension, they were not more likely to have a prior HF hospitalization, and they did not have a significantly different LVEF, compared with those with lower frailty. With respect to HF therapy, patients with higher FI were less often treated with ACEI/ARB, but more often with CRT.

Stay updated, free articles. Join our Telegram channel

Join