NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Rector TS, Taylor BC, Greer N, et al. Use of Left Ventricular Assist Devices as Destination Therapy in End-Stage Congestive Heart Failure: A Systematic Review [Internet]. Washington (DC): Department of Veterans Affairs (US); 2012 May.

Cover of Use of Left Ventricular Assist Devices as Destination Therapy in End-Stage Congestive Heart Failure: A Systematic Review

Use of Left Ventricular Assist Devices as Destination Therapy in End-Stage Congestive Heart Failure: A Systematic Review [Internet].

Show details

METHODS

TOPIC DEVELOPMENT

This project was nominated by Dr. Chester Good, Chief, Section of General Medicine. The key questions were developed with input from a technical expert panel (see Appendix A).

The final key questions were:

Key Question #1.

How does use of an FDA-approved, current generation LVAD as destination therapy (i.e., the HeartMate II left ventricular assist device) effect patient outcomes?

Key Question #2.

What patient or site characteristics have been associated with patient benefits or harms when the FDA-approved, current generation LVAD is used as destination therapy?

Key Question #3.

What is the range of cost-effectiveness estimates of using the FDA-approved, current generation LVAD as destination therapy in end-stage heart failure and what explains variation in these estimates?

Figure 1 depicts the analytic framework for these questions.

Figure 1. Analytic Framework.

Figure 1

Analytic Framework.

SEARCH STRATEGY

We searched MEDLINE (OVID) for studies that reported patient outcomes, articles about patient selection or prediction of patient outcomes, systematic reviews or cost-effectiveness analyses from 1995 to October 2011 using standard search terms. The 1995 start date is well before the first randomized clinical trial that used a left ventricular assist device for destination therapy. The search was limited to articles involving human subjects and published in the English language. Search terms included: heart-assist devices, heart failure and ventricular dysfunction (See Appendix B for the MEDLINE search strategy). We also examined reference lists to identify other pertinent publications and asked our panel of experts to identify additional reports.

Other searches included the Cochrane Database of Systematic Reviews, the Translating Research into Practice (TRIP) database for systematic reviews and technology assessments, the Center for Medicare and Medicaid Services (CMS) Web site and the NIH Clinical Trials Web site.

STUDY SELECTION

Generally, abstracts and full text articles were excluded whenever

  1. The report did not provide data about the only continuous flow device currently approved by FDA for destination therapy, the HeartMate II ventricular assist device.
  2. The report did not provide data about use of the device as destination therapy. Short-term use of the device as a bridge to transplant was excluded.
  3. The report did not provide data about patient outcomes of interest such as survival, hospitalizations, daily function, health-related quality of life or harms.
  4. The report did not provide data about adults defined as age 18 years or older.

The following discusses more specific study selection criteria for each key question.

Key Question #1. Patient Outcomes

Randomized controlled trials were sought as the highest quality of evidence for the first key question about patient outcomes. At a minimum, randomization helps assure unbiased allocation of subjects to the groups being compared. However, the number of patients currently eligible for destination device therapy is limited and randomization might not balance baseline characteristics in small studies. Subjects that withdraw from the study after randomization or crossover to the comparison group may also bias the comparison, particularly if any changes in treatment were related to study outcomes. Obviously a comparison of a surgically implanted device to continuation of optimal medical therapy can’t be blinded, and endpoint assessments, especially subjective assessments and device-related harms, may be biased as a result. However, more objective endpoints such as maximal exercise tests are affected by patient effort and difficult to interpret in terms of how they translate into affects on patients’ lives. Given the high morbidity and mortality of patients with refractory end-stage heart failure, randomization to continuation of non-surgical therapies might not be acceptable to patients and providers who believe a ventricular assist device is a reasonable alternative. Therefore, randomized clinical trials are being designed as a non-inferiority comparison of a new ventricular assist device to an approved device.17,18 Non-inferiority studies introduce additional concerns that the characteristics of the enrolled subjects, including how they were treated, might not be similar to previous studies that demonstrated the ‘control’ device is effective with acceptable risks.19 Furthermore, the magnitude of the differences between devices excluded by the statistical analysis of non-inferiority needs to be small enough to rule out clinically important differences. Since high quality evidence from randomized clinical trials of the FDA-approved continuous flow ventricular assist device for destination therapy is very limited, we did not restrict our review to randomized controlled trials, and considered cohort studies that could provide estimates of the likelihood of patient outcomes.

Key Question #2. Patient Selection

Given numerous differences in outcomes including device malfunction between the two ventricular assist devices currently approved for use as destination therapy, the HeartMate XVE and HeartMate II, and the current exclusive use of the HeartMate II device for destination therapy in the INTERMACS registry, our search for the second key question concerning selection of patients sought specific analyses about the HeartMate II ventricular assist device. As previously mentioned, we also focused on selection of patients for destination therapy rather than bridge therapy because the criteria and outcomes including competing risks such as heart transplantation are not the same. Thus, studies were selected if they provided evidence specifically or predominantly about the selection of patients for use of the HeartMate II device as destination therapy. Subgroup analyses that focused on this specific therapy were considered including regression analysis that included variables indicating the type of ventricular assist device and/or therapy.

Patient selection criteria for destination therapy are based primarily on the selection criteria used in the studies that supported FDA approval and therefore define patients eligible for the approved indication. However, regression or subgroup analyses are often conducted using study data or other patient cohorts in an effort to better define which patients are more likely to benefit or be harmed. Studies were sought that provided statistical evidence for significant differences in patient outcomes between groups defined by preoperative patient characteristics. Expert reviews of patient selection criteria were read in search of additional scientific evidence about patient selection criteria. Because estimates of patient survival without the ventricular assist device are used to select patients, we also included studies that evaluated survival prediction models in a sample of patients eligible for destination therapy.

Key Question #3. Cost Effectiveness

We included studies that provided cost-effectiveness estimates for the use of HeartMate II ventricular assist device as destination therapy.

DATA ABSTRACTION

For reports that provided pertinent evidence about patient outcomes and selection (Key Questions #1 and #2), we extracted information about the study sites and sponsor, subject inclusion and exclusion criteria, sample characteristics, intervention(s) including the comparison group(s), if any, length of follow-up, patient outcome(s) of interest and quality of the evidence.

For reports that provided pertinent evidence about cost effectiveness (Key Question #3), we extracted information about the overall estimate of cost effectiveness, the uncertainty in the cost effectiveness estimate, the base case assumptions for the cost effectiveness model and the results of sensitivity analyses that varied the assumptions in the base case model.

QUALITY ASSESSMENT

Key Question #1. Patient Outcomes

The quality of clinical trials was judged based on the potential for bias in the estimates of treatment effects according to the following criteria: 1) random assignment to treatment with adequate concealment of assignment, 2) blinding of key study personnel (i.e., providers, study personnel and/or patients) who determined outcomes to assigned treatment, 3) analysis by intention-to-treat (i.e., all subjects counted in group to which they were randomized in the analysis of outcomes), 4) reporting of number of withdrawals/dropouts by group assignment along with reasons for any losses to follow-up that may be related to beneficial or adverse treatment effects and 5) the size of the treatment effects (larger effects are less likely to be explained by baseline differences between treatment groups or differential losses to follow-up).20 Studies were rated as good, fair, or poor quality. A rating of ‘good quality’ generally required that the investigators randomly assigned patients to treatments and reported adequate concealment of assignments, blinded or objective outcome assessments, an intent-to-treat analysis, an adequate description of reasons for dropouts/attrition and a sizable treatment effect. The quality of a study was generally considered poor if the method of allocation concealment was inadequate or not defined, blinding was not reported or possible, analysis by intent-to-treat was not reported and reasons for dropouts/attrition were not reported and/or there was a high rate of attrition or the estimated treatment effect was small.

Key Question #2. Patient Selection

Criteria to assess the quality of evidence concerning variables and multivariable models to predict patient outcomes have not been firmly established. We relied on our, as yet unpublished, guidance for conducting systematic reviews of prognostic tests commissioned by the Agency for Healthcare Research and Quality (Rector TS, Taylor BG, Wilt TJ. Chapter 12: Systematic Reviews of Prognostic Tests in Methods Guide for Comparative Effectiveness Reviews). Specific criteria: 1) were patients in the analysis similar to those who would receive an FDA-approved ventricular assist device as destination therapy?, 2) were the variables used to make outcome predictions measured shortly before implantation of the device and not affected by the procedure, subsequent care or knowledge of the outcome being predicted?, 3) were the measurements of the potential predictor variables and outcomes reliable, valid and routinely available in clinical practice?, 4) did losses to follow-up bias the assessment of outcomes?, 5) was the duration of follow-up adequate?, 6) were the number of patients that had the outcome being predicted adequate for the number of predictors tested?, 7) were predicted outcome probabilities reported for patient subgroups that would be included or excluded from destination therapy?, 8) how closely did outcome predictions agree with the observed outcomes?, 9) were the outcome prediction somehow validated? and 10) did the analysis demonstrate that the outcome predictions could be used to improve patient outcomes?

Key Question #3. Cost Effectiveness

There are no well-accepted criteria for evaluating the quality of cost-effectiveness analyses, however, there are long lists of factors related to the analytical model and assumptions that can be considered.21 In order to assess quality we extracted information on the cost-effectiveness model structure and assumptions.

RATING THE BODY OF EVIDENCE

The overall evidence for a key question was graded using the method proposed by Owens et al.23 using the following criteria:

  • High grade evidence: Further research is very unlikely to change the confidence in the estimated treatment effect on patient outcomes. Generally, a high grade requires more than one good quality study and consistent estimates with statistical confidence intervals that exclude clinically meaningful differences.
  • Moderate grade evidence: Further research may change the estimate of the size of the effect or the level of uncertainty.
  • Low grade evidence: Further research is likely to change the estimate of the size of the effect and or the confidence interval.
  • Insufficient evidence: Sufficient evidence was not found to answer the question.

DATA SYNTHESIS

We constructed evidence tables for Key Questions #1, #2 and #3, and drew our conclusions based on a qualitative synthesis of the evidence available to answer each key question. Not finding several reports that provided independent estimates of patient outcomes, relationships between baseline variables and patient outcomes or cost effectiveness, we did not do any meta-analyses to pool evidence from different studies.

PEER REVIEW

A draft version of this report was reviewed by clinical experts including the Technical Expert Panel. Their comments are presented in Appendix C as are our responses to any suggestions to modify the report itself.

Views

  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (755K)

Other titles in this collection

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...