To the best of our knowledge, this is the first study to analyze survival in all patients undergoing ICD and CRT-D implantation within a National Health System. This design allowed to avoiding selection bias and increasing the generalizability of effectiveness results.

Overall 1-year, 2-year and 3-year survival rates in the ICD-alone group were similar to those reported in the AVID study (89.3%, 81.6% and 75.4%, respectively) [17]. In addition, the rate of cardiac deaths was similar to that reported in the same study [18]. One-year, 2-year, 3-year, 4-year, and 5-year survival rates in the CRT-D and ICD groups were similar to those described in a large case series in the US [7]. The overall 5-year survival in our study was similar to that reported in the RAFT Trial (65.4%) among ICD patients, but considerably lower than that reported in CRT-D patients [19]. This difference in long-term prognosis may be explained by the inclusion of patients with NYHA functional class II or III in this study. The one-year survival in the CRT-D group was lower than the 88% described in the COMPANION trial, even when the group undergoing transvenous implant was considered individually [2].

The 5-year survival found among ICD patients in our study was very similar to that observed in a study of patients with Chagas’ heart disease [10]. Although Chagas disease was the underlying cause in one-third of all patients who eventually died, we cannot infer that these patients had a poorer prognosis. Our survival curves by underlying disease suggests that etiology of heart disease (Chagas vs. ischemic heart disease) was not a prognostic factor. The ICD-LABOR study that also included patients with Chagas’ heart disease found similar results [20]. Barbosa et al. suggested that Chagas’ heart disease patients are more likely to have ventricular arrhythmias than patients with other cardiopathies, but on the other hand, they have higher rates of appropriate ICD therapy [21].

The in-hospital case fatality of 0.33% observed in the ICD group in our study was only slightly higher than the 0.2% reported in a systematic review of nonthoracotomy ICD trials [22] but considerably lower than that reported in other clinical data analyses [12,23,24]. Nevertheless, the 1.1% 30-day case fatality observed was identical to that reported in the AVID trial, but higher than the 0.6% found in a systematic review of nonthoracotomy ICD trials by Rees et al. [22]. The mean length of hospital stay among ICD patients was shorter than that reported in a large study of administrative databases in the US [12].

The in-hospital case fatality of 2.9% seen in the CRT-D group was higher than the 0.5% case fatality reported in a systematic review [25], 0.9% reported by Swindle et al. [23] and 1.1% reported in Medicare patients [24]. The COMPANION and MADIT-CRT trials also found much lower in-hospital case fatality (0.6% and 0.1%, respectively) than that found in the present study; however, both studies excluded patients with implantation with thoracotomy and the MADIT-CRT trial included only patients with NYHA Class I and II [22]. Even the in-hospital case fatality in the CRT-D sub-group undergoing transvenous implantation (1.6%) was higher than those reported in these studies. The 8.5% in-hospital case fatality in CRT-D patients older than 70 was much higher than that reported by Swindle et al. in elderly patients [23]. The 30-day mortality in those undergoing transvenous CRT-D implantation in our study (4.1%) was higher than the 1.8% observed in the COMPANION trial [2]. Even among younger patients, the mean length of stay in the CRT-D group was higher within SUS than that reported in Zhan et al. study [12]. The median length of hospital stay in our study was the same regardless of the implant technique.

Our study did not find any deaths in children undergoing ICD implantation. Other study corroborate the good prognosis in this age group [9]. Differences in survival times between ICD-alone or CRT-D patients were expected because the indication criteria for CRT-D implant included poor functional status (NYHA Class III or IV) and ventricular dyssynchrony.

The decrease in survival among CRT-D patients around the fourth year of follow-up may suggest the impact of the disease natural history or device-related problems. A study by Cleland and colleagues with patients with heart failure and dyssynchrony found among those treated with medical therapy alone a pattern of decline in survival mainly due to sudden death that is similar to that observed in our study [26]. However, this pattern of survival was not observed in Saxon and colleagues study that also assessed long-term outcomes [7]. One explanation for this pattern would be the effect of a factor affecting the long-term effectiveness of CRT-D therapy and its impact in the disease natural history. Horlbeck and colleagues showed that mean lifetime of CRT-D devices was 4 years [27]. Likewise, Biffi and colleagues demonstrated that median lifetime of CRT-D devices was approximately 4 years [28]. Thijssen and colleagues reported that mean battery lifetime of CRT-D devices was 4.7 years [29]. In an earlier study by Hauser and colleagues (1998–2005) only 4% of CRT-D pulse generators were operating normally within four years of the implant [30]. A recent study by Landolina and colleagues reported that at three years slightly more than 10% of patients underwent surgical revision for battery depletion and at four years this rate rose to about 50% [31]. This same study showed that patients undergoing replacement had double the risk of infection, which could also explain the mortality observed in our study. Only ongoing monitoring can establish whether this finding was exclusive to the initial cohort of patients undergoing CRT-D implantation (long follow-up) and to what extent it was impacted by the small number of patients at-risk in the fourth year of follow-up.

Dual-chamber devices accounted for 65% of all ICD implants In the last three years studied, a percentage that is similar to that reported in the US National ICD Registry (62% from 2006 to 2007) [8]. In this study only 40.4% of patients implanted with dual-chamber ICD devices met the indications for pacemaker therapy and the use of dual-chamber ICD devices was associated with increased in-hospital complications and in-hospital case fatality rates [8]. In our study we found no differences in in-hospital case fatality and short-term, medium-term and long-term survival rates among those implanted with single- and dual-chamber devices. Those undergoing dual-chamber ICD implantation were less likely to have supraventricular tachycardia, which contrasts with that reported by Dewland and colleagues [8], suggesting that this is not a common indication for atrial lead placement in Brazilian patients.

The present study showed that the proportion of CRT-D with surgically implanted left ventricle leads as first approach (28%) seems slightly greater than that reported in other studies (24.1% to 24.9%) [32,33]. The percentage of transvenous implantation failure in our study (13.7%) was also higher than that reported in other studies with CRT-D (5.9% to 11%) [32-34]. In patients undergoing CRT-D implantation, epicardial lead placement was found associated with increased in-hospital case fatality, as suggested in other studies with CRT [35] and CRT-D [32]. Despite increased in-hospital case fatality and 30-day case-fatality, we found similar medium-term and long-term survival rates for both implantation techniques suggesting similar effectiveness of these techniques in patients surviving the initial post-implantation period, which corroborates that reported by Miller and colleagues [35] with CRT. However, conflicting results have been reported. One study showed poorer short-term and long-term prognosis with the epicardial technique [32] while other studies found no significant differences in short-term or long-term results [33,36]. The results found in our study regarding the epicardial technique may be explained by the “learning curve” of surgical teams, quality of postoperative care and potentially suboptimal drug therapy prescribed to patients with advanced heart failure referred to surgery [37].

Our study has some limitations common to studies relying on administrative databases. The lack of information on patient variables such as NYHA functional class, left ventricle ejection fraction, and history of sudden cardiac death does not allow proper adjustment for the patients’ baseline risk. The proportion of underlying diseases may not reflect their actual distribution in the study population due to missing information on this variable in SIH database. Information on the underlying condition is often missing in hospital admission authorization forms because providers are required to fill out two different diagnosis fields including diagnostic codes to support the patient’s eligibility for device implantation (e.g., type of arrhythmia and heart failure). The sensitivity of the probabilistic record linkage method used to find deaths registries (90.6%) may have potentially missed some death records in the database. However, there was 100% specificity in identifying deaths records and probably random loss [16].