Sudden death and pump failure death in heart failure

Sudden death and pump failure death are two major modes of death in patients with heart failure and reduced ejection fraction (HF-REF) and in patients with heart failure and preserved ejection fraction (HF-PEF). There have been advances in evidence-based treatments in patients with HF-REF over the last two decades, along with the changing patient characteristics in both HF-REF and HF-PEF populations. It is of great interest and significance to discover if these changes have translated into temporal changes (and corresponding trends over time) in the risks of sudden death and pump failure death in both populations. Apart from examining any changes in the rates of mode-specific death in population level, it is also of interest and importance to estimate the risks for sudden death and pump failure death in individual patients. Accurate risk prediction can aid in better risk stratification. In patients with HF-REF, identifying high-risk subgroups would help target the device therapy to those most likely to benefit and identifying low-risk subgroups would avoid unnecessary implantation, thus improving the cost-effectiveness of the therapy. In patients with HF-PEF, identifying high-risk subgroups would enable further research into the efficacy of device therapy in this population. The aims of this work were to examine the trends in the rates of sudden death and pump failure death over time in patients with HF-REF and in patients with HF-PEF, and to separately develop validated models to predict sudden death and pump failure death in both populations. Given that there are limited data on mode-specific death from community-based studies, I used data from clinical trials which have more detailed and standardised sub-classification and adjudication of mortal events. Besides, compared to community-based studies, clinical trials have more detailed baseline characterisation, which allows more complete multivariable adjustment to account for confounding and between-study differences. Therefore, a cohort of 46,163 patients with HF-REF enrolled in 13 clinical trials conducted between 1995-2015 and a cohort of 10,517 patients with HF-PEF in 3 clinical trials over the period 1999-2013 were included in this thesis. Multiple linear regression analysis was used to examine the trends in the rates of sudden death and pump failure death over time in both populations respectively. The cumulative incidences for sudden death and pump failure death in each trial at different time points during follow-up were calculated with the cumulative incidence function method, counting the competing risk of death from other causes. The risk for each mode of death across trial arms and by HF duration was examined using the Cox regression models, with further adjustment for a number of confounding variables. The models to predict sudden death and pump failure death in patients with HF-REF were separately developed in PARADIGM-HF and validated in ATMOSPHERE. Models for both modes of death in HF-PEF were developed in I-PRESERVE and validated in CHARM-Preserved as well as TOPCAT. These models were constructed using a competing risk approach with the Fine-Gray sub-distributional hazards regression analysis. Model performance was examined by assessing calibration (i.e. the agreement between the observed and predicted cumulative incidences over time) and discrimination (i.e. the ability to separate patients at higher risk from those at lower risk). I found that the risks of sudden death and pump failure death in patients with HF-REF have fallen across 13 clinical trials over the period 1995-2015, consistent with a cumulative use of evidence-based therapies in this population. The absolute rates of sudden death and pump failure death were very low in the early follow-up after randomisation in patients with HF-REF who received modern evidence-based treatment. Longer standing HF was associated with greater risks of sudden death and, particularly, pump failure death in HF-REF. The risks of sudden death and pump failure death were consistently low across the 3 largest clinical trials in patients with HF-PEF, with little difference by experimental treatment in any trial. There was a downward trend in the rates of sudden death and pump failure death across these trials over time, in parallel with a changing characteristic of patients enrolled in these trials. Nevertheless, sudden death and pump failure death remained the most common modes of death, altogether accounting for the majority of CV death. The absolute rates of sudden death and pump failure death in patients with HF-PEF were extremely low in the early follow-up after randomisation. Longer standing HF was associated with a slightly higher risk of sudden death and a substantially higher risk of pump failure death in HF-PEF. The sudden death and pump failure death models in patients with HF-REF I developed in the largest and most contemporary cohort (PARADIGM-HF), included a number of variables collected in routine clinical practice, and accounted for the prognostic impact of the competing risk of death from other causes. The discriminating ability was modest for the sudden death model but excellent for the pump failure death model. Both models showed good calibration and were robust when externally validated in ATMOSPHERE. The prognostic models in patients with HF-PEF I developed in I-PRESERVE, using simple demographic and clinical variables, showed good discrimination and calibration for both sudden death and pump failure death, and were robust in external validation in CHARM-Preserved and TOPCAT. The performance of both models was further improved with the inclusion of NT-proBNP. In conclusion, I have found that the risks of sudden death and pump failure death have declined over time both in patients with HF-REF and in patients with HF-PEF based on clinical trial data. The patterns of change in the rates of both modes of death over time need to be examined in community-based populations. The prognostic models for both modes of death, showing reasonable performance, can be considered for use in risk stratification for mode-specific death in both populations, aiding in decision making in device therapy in similar patients in HF-REF and helping with patient selection for device interventions in future trials in HF-PEF