Residuals: Mitral Regurgitation, Bias, and Effects of Medical Therapy

Justin A. Ezekowitz, MBBCh, MSc


Circulation. 2021;144(6):438-440.. 

If ever there is a question that has plagued clinicians dealing with patients with left ventricular dysfunction and led to great debates, it is regarding the mitral valve. Whether to intervene early or late, whether it is primary or secondary, proportionate or disproportionate, or even whether it is related to clinical outcomes including the patient's reported symptoms. In the case of mitral regurgitation (MR), estimating the MR severity by echocardiogram has vastly informed this debate, but equipoise remains on the these unanswered questions. Considering the recent technical therapeutic advances such as the development of transcatheter edge-to-edge repair (TEER) devices, these have now been studied in patients with secondary MR.

The COAPT trial (Cardiovascular Outcomes Assessment of the MitraClip Percutaneous Therapy for Heart Failure Patients with Functional Mitral Regurgitation), the largest trial of its kind, demonstrated that, on top of medical therapy, using a TEER device improved clinical outcomes and specifically heart failure hospitalization (HFH).[1] However, on examination of the COAPT trial results, residual questions remain.

The term "residuals" is an apt descriptor for the questions that remain because a residual, in statistical terms, is defined as the distance between a single data point and the regression line, or, in other words, the distance from an individual point (patient) of data to the groups (lots of patients) best fit of data. In clinical terms, it may represent an individual patient compared with the grouped result of a clinical trial. Residual questions regarding the effect of medical therapy on MR, how successful TEER is at reducing MR, the association of residual MR with outcomes, and the durability of the findings of COAPT are among the many faced by both patient and clinician. In this issue of Circulation, we gain some answers, but some residual questions remain.[2]

Of the 614 patients originally enrolled, 87% had echocardiograms done 1 month after randomization and MR was evaluated. Using echocardiographic techniques to evaluate MR after a TEER (there are 2 jets and, thus, conventional techniques are rendered inadequate), the COAPT investigators now present new insight into the outcomes of patients who had either persistent severe (3+/4+) or moderate (2+) or minimal (0/1+) MR 30 days after enrollment. They identified that the likelihood of improved MR was, unsurprisingly, strongly correlated with the intervention, and the severity of MR was related to the primary composite outcome of all-cause mortality/HFH, and the quality of life as measured by the Kansas City Cardiomyopathy Questionnaire, as well. Because more patients in the intervention arm had less MR, it goes without saying that the intervention group had lower risk for future events overall.

The COAPT trial showed that ≈30% of patients in the medical therapy arm also improved to ≤2+MR by 30 days without any further invasive intervention; this reflects the nature of MR. This is inherent in research involving a dynamic measurement of a biological variable such as MR, which is subject to the effects of the myocardium, changes in intracardiac pressure, volume loading, atrial mechanics, and many other factors, as well. However, this does not mean that the reduction of the MR is on the causal pathway. The reduction in MR may simply reflect improved neurohormonal balance, the timing effect of medical therapy, or other factors independent and unrelated to the measured MR. Because there were no changes in left or right ventricular structure or function in patients who improved their MR by 30 days and out to 2 years, should we be concerned? More sophisticated analytical methods including joint modeling or mediation analysis could shed light on this topic, and, although they too remain imperfect, would offer insight about what is causal versus simply an observation.

A few caveats deserve consideration. By 30 days, 13% of the trial population was not included in this analysis because of death (n=20), withdrawal or lost to follow-up (n=7), or because the echocardiogram was unavailable (n=53). It is important to note that, as outlined in Table II in the Data Supplement, these excluded patients were at far greater risk with higher NT-proBNP (N-terminal pro-B-type natriuretic peptide; triple the value of the patients who were included), higher Society of Thoracic Surgeons score, worse kidney function, and a one-third to two-thirds split of being in the intervention versus medical therapy group, respectively. How might this affect the results? In the first 30 days of the overall trial, there were more deaths and unplanned cardiac interventions in the intervention group than in the medical therapy group. An assumption could be made that these individuals who had an early untoward event did not have improvement in MR and thus would bias the current findings to the null, or, alternatively, the MR and the event were unrelated. It is especially unfortunate, however, that the echocardiographic data are unavailable in a trial so reliant on this information. Clinicians must be aware that the trial environment represents the best-case scenario and offer carefully selected patients a realistic assessment of risk and possible outcomes.

NT-proBNP values at baseline offer some insight into this population of patients with MR and heart failure. In patients who had mild, moderate, or severe MR, the median NT-proBNP values of 3050, 2287, and 3255 pg/mL, respectively, indicate a higher-risk population more akin to patients enrolled after a worsening heart failure event such as the Guiding Evidence Based Therapy Using Biomarker Intensified Treatment in Heart Failure[3] or Vericiguat Global Study in Subjects with Heart Failure with Reduced Ejection Fraction[4] trials. This is important because it appears that there is a lack of a mortality benefit in COAPT (the primary end point is driven by HFH), and it is hard to move the needle on this outcome, even with advanced therapies. Why is this important? First, if improving MR is on the causal pathway for mortal outcomes (either all-cause or cardiovascular), then improving MR should lead to a reduction in 1- or 2-year mortality using an intervention directed at reducing MR directly (TEER) rather than indirectly (medical therapy). However, this is not the case for TEER as is evident in the data provided up to 2-years; the 5-year follow-up may also show this, but that remains to be seen. Second, a reduction in HFH is important for patients, payors, and clinicians, and recurrent admissions are particularly useful to examine as an outcome. We and others have previously examined this ever shortening gap[5] and risk between HFHs and identified that, if one could break the cycle of readmissions, it might offer an opportunity to improve outcomes, because patients often convey that it takes weeks to months to recover each time from such an unfortunate event.

In case we needed a further reminder as to the value of clinical trials using device or surgical interventions, this provides another lesson regarding why blinding (where possible) and randomization remain critical parts of the evidence generation process. One might surmise that an intervention will improve quality of life only by way of the intervention itself, but this can be a faulty thought. For example, in patients with persistent severe MR by 30 days (inferring the intervention did not work as anticipated), the Kansas City Cardiomyopathy Questionnaire score improved by 11 points in the TEER arm and 1.6 points in the medical therapy arm: a key reminder that there is a strong placebo effect of intervening, even if the intervention does not work. This is not to indicate that, in totality, the intervention overall does not work, but should move the field away from single-arm, nonrandomized or pre–post comparisons where the evidence generation may provide little replicable insight.

Supportive of the relevant residual effects of medical therapy is that, by 12 months and 24 months, 33% and 40% of patients, respectively, in the medical therapy arm who did not initially improve by 30 days (in terms of MR), went on to improve and downgrade the MR to mild or moderate. Medical therapy for heart failure with reduced ejection fraction has evolved substantially, and guidelines[6] (based on a data set of thousands of patients in landmark trials using mortality as an outcome) now endorse an early upfront approach of medications with little waiting for poor outcomes. This is a phase shift from the slower, staccato methods of one-at-a-time medications over many months (or, sadly, years) to an earlier rapid uptitration approach.[7] What we need to decide is when to reimage a patient to evaluate structure and function and decide on elements such as MR. If the COAPT trial was started today, we would likely want far greater use of sacubitril/valsartan (<5% in COAPT), robust use of a mineralocorticoid antagonist (50% in COAPT), and sodium-glucose cotransporter 2 inhibitor use (0% in COAPT). We remain uncertain as to the doses achieved before and after randomization, because these data remain unpublished, but it is reasonable to assume that they match other registry data identifying suboptimal dosing, even at expert centers. In patient selection, optimizing medical therapy with newer therapies, and doses of existing therapies, as well, rigorous and high-quality imaging and enough time to accrue for medical therapy to work must be part of the pathway before considering a surgical or percutaneous intervention. We previously identified that ≈1% of patients with heart failure with reduced ejection fraction would have been eligible for COAPT,[8] without applying stricter dosing criteria.

What should we think of the residual MR and the residual questions? How long is long enough to wait before TEER will remain an unanswered question for the time being, but to intervene too early is an unfortunate exposure to the risk of a procedure (and no known benefit), and too late exposes a risk of the underlying disease process. Clinicians should therefore remember the benefit of patience for their patients and how much residual distance exists between their patients and the trial outcomes.