Ventricular assist device (VAD) therapy has proven to be a life-saving therapeutic option in patients with end-stage heart failure.1 The clinical outcomes with this therapy have assumed an even higher level of success in the continuous-flow left ventricular assist device (LVAD) era. With wider application of this therapy, not only to higher-risk patients, but also potentially to patients with earlier stages of heart failure, an emphasis on patient selection is of great importance. Tools that are easily used by clinicians who do not practice at VAD implant centers may facilitate the identification and referral of patients with an optimal risk/benefit profile. Many models designed for risk stratification based on preoperative variables have sought to fill this need for refining and improving patient selection and timing of LVAD implantation. These include the Lietz-Miller model, Acute Physiology and Chronic Health Evaluation (APACHE), Columbia University risk score, the HeartMate II Risk Score, and the Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS) scores.2 Although all these models have shown varying degrees of success and applicability to heart failure patient populations, the Seattle heart failure model (SHFM) has had the best success as a predictive model for heart failure patients requiring LVAD implantation.3,4
The SHFM was originally derived from a cohort of outpatients with heart failure. This has sometimes been considered a limitation of this model for evaluation of patients with advanced heart failure, and efforts have been made to incorporate variables such as intraaortic balloon pump (IABP) usage, inotropes, and mechanical ventilation.5 Recently, the SHFM has been applied successfully to both single and multicenter studies as well as in both bridge-to-transplant and destination therapy populations, demonstrating value even in the end-stage heart failure population. However, waiting to apply the SHFM model until someone is actively undergoing evaluation for LVAD may suggest that we are waiting until too late in the course of disease to evaluate risk of mortality. As mortality after LVAD implantation improves, it is only natural to consider whether lower-risk patients may also have a favorable risk/benefit profile. This highlights a major advantage of the SHFM over other risk scores that were derived from the small group of patients who were actually implanted with a VAD—the ability to produce a continuous risk score for every heart failure patient that can be expressed as predicted life expectancy and percentage chance of survival to a particular year via a simple one-step exponential transformation. This feature might prove advantageous not only in aiding individual patient selection, but also in deciding optimal timing of implant.
The thoughtful study by Pamboukian et al.6 in this issue provides additional evidence of the utility of the SHFM, applied shortly before implant, in predicting far superior 1-year survival with LVAD therapy as compared to medical management (82 vs 49%) in patients with end-stage heart failure. Although effective in this population (14 of the patients were on ≥3 pressors/inotropes), it may be time to expand our focus from refining risk stratification immediately before implant to evaluation of mortality risk earlier in the course of illness. This may be particularly important, as the field has shifted to almost exclusive use of continuous-flow devices.
It is unknown whether risk scores are used routinely in many centers on a prospective basis to calculate risk scores on individual patients to aid patient selection (or recommend continued medical therapy) or whether their major role has been to draw attention to preoperative variables that place patients at higher postoperative risks. It is clear that the development of risk models, primarily, SFHM, and INTERMACS, has allowed investigators to study the impact of various risk factors on outcomes to aid not only patient selection, but also timing of implantation. Although objective prospective data on the use of these risk scores are currently not available, the ideal paradigm should also consider whether such scores could be used to guide modification of risk factors before LVAD placement. For example, a patient with a prohibitive risk score at initial evaluation might be an unacceptable candidate for LVAD placement, but after intensive medical therapy (improving nutrition, correcting abnormal coagulation parameters, treatment of infections) and optimizing hemodynamics (ultrafiltration, inotropes, IABP support), this patient may have an improved risk profile and could then potentially undergo LVAD placement with acceptable outcomes.
Unfortunately, the SHFM was an extremely poor prognostic model in severely ill heart failure patients requiring biventricular assist device placement in this study. This limitation of the SHFM hampers its applicability (which would be ideal) to all patient populations considered for VAD therapy. It should be noted that there are several additional risk scores to determine the risk of right ventricular failure and the need for right ventricular assist device support after LVAD placement, highlighting the particular challenge of stratifying this population.7 Whether some of these risk scores or additional preoperative variables could be incorporated into the SHFM to further improve its utility remains to be seen, although it would be difficult to substantially improve the overall performance of the model solely by refining accuracy in this relatively small population of patients with severe biventricular heart failure.
Ventricular assist device therapy has clearly evolved over the past decade with improved survival with continuous-flow devices as well as an increased emphasis on patient selection. The ability to present survival data and an accurate risk assessment to patients and families based on preoperative risk variables is an extremely useful tool. However, the challenge remains to be able do this in a prospective fashion with a model that is easy to use and retains its reliability when applied to a broad range of heart failure patients.
1. Miller LW, Pagani FD, Russell SD, et al.HeartMate II Clinical Investigators. Use of a continuous-flow device in patients awaiting heart transplantation. N Engl J Med. 2007;357:885–896
2. Schaffer JM, Allen JG, Weiss ES, et al. Evaluation of risk indices in continuous-flow left ventricular assist device patients. Ann Thorac Surg. 2009;88:1889–1896
3. Levy WC, Mozaffarian D, Linker DT, Farrar DJ, Miller LWREMATCH Investigators. . Can the Seattle heart failure model be used to risk-stratify heart failure patients for potential left ventricular assist device therapy? J Heart Lung Transplant. 2009;28:231–236
4. Ketchum ES, Moorman AJ, Fishbein DP, et al. Predictive value of the Seattle Heart Failure Model in patients undergoing left ventricular assist device placement. J Heart Lung Transplant. 2010;29:1021–1025
5. Levy WC, Mozaffarian D, Linker DT, et al. The Seattle Heart Failure Model: Prediction of survival in heart failure. Circulation. 2006;113:1424–1433
6. SalpyJoseRobert, et al. Comparison of observed survival after ventricular assist device placement versus predicted survival without assist device using the seattle heart failure model. ASAIO J. 2012;58:xx
7. Matthews JC, Koelling TM, Pagani FD, Aaronson KD. The right ventricular failure risk score a pre-operative tool for assessing the risk of right ventricular failure in left ventricular assist device candidates. J Am Coll Cardiol. 2008;51:2163–2172