Translating stem cell research to cardiac disease therapies: pitfalls and prospects for improvement

Abstract

Over the past 2 decades, there have been numerous stem cell studies focused on cardiac diseases, ranging from proof-of-concept to phase 2 trials. This series of papers focuses on the legacy of these studies and the outlook for future treatment of cardiac diseases with stem cell therapies. The first section by Drs. Rosen and Myerburg is an independent review that analyzes the basic science and translational strategies supporting the rapid advance of stem cell technology to the clinic, the philosophies behind them, trial designs, and means for going forward that may impact favorably on progress. The second and third sections were collected as responses to the initial section of this review. The commentary by Drs. Francis and Cole discusses the review by Drs. Rosen and Myerburg and details how trial outcomes can be affected by noise, poor trial design (particularly the absence of blinding), and normal human tendencies toward optimism and denial. The final, independent paper by Dr. Marbán takes a different perspective concerning the potential for positive impact of stem cell research applied to heart disease and future prospects for its clinical application. (Compiled by the JACC editors).

Keywords: cell-based therapy; heart diseases; research design; stem cell research; stem cell transplantation; tissue-based therapy.

Figure 1. Strategies to Evaluate Benefits of Tsherapy

Clinical testing strategies range from rigidly controlled randomized trials to observational strategies for defining association, causation, clinical effects, and safety.

Figure 2. How the Size of Random Noise and the Presence of Bias Towards Optimism Interact to Generate Statistically Significant, But False Positive, Results

Each data point is the measured improvement in a single patient of a clinical variable after a therapy (which happens to be ineffective). Across all patients, the average change is 0, but only if the measurements are acquired dispassionately. If, for example, 2 measurements are made and the “best” chosen, this optimism contributes a positive bias that may be (correctly) detected as statistically significant even though it is illusory. Noisier measurements are more vulnerable. Large study size paradoxically makes the problem worse (96, 97).

Figure 3. Contributors to the 66% symptomatic response widely described for CRT

Within RCTs the response rate in the active arm averages 51%. Placebo control arms (which receive a device but not CRT pacing) show a response rate averaging 35%, indicating the incremental effect of pacing to be 16 percentage points. The controls in blinded trials report a better response than those in unblinded trials: we can use this to estimate the size of the placebo effect of device implantation. The response in the unblinded controls is not due to pacing or implantation placebo, and might be described as spontaneous improvement. Figure based on Sohaib et al. (101)