Left Ventricular Dysfunction in Potential Heart Donors After Brain Death: When is it Reversible?

Adult heart transplant (HT) listings in the United States (US) have consistently outpaced the number of transplants performed. As a result, many candidates wait several months or longer for transplant and hundreds die annually while waiting. Increasing HT volume is the obvious solution, but this is constrained by the supply and utilization of available heart donors – only one-third of which are actually used for HT.1 Left ventricular (LV) dysfunction is one of the most common reasons for donor heart non-use, accounting for up to 25% of all declined heart offers.2 Confirming prior reports in smaller cohorts, we found that donor LV dysfunction was both common (occurring in 13% of potential heart donors) and often reversible prior to organ recovery (in 58% of cases) in the Donor Heart Study (DHS) cohort.3 When such hearts are used for transplant, they appear to confer acceptable recipient outcomes.3-5 These findings argue for a ‘wait and see’ approach to donors with initial LV dysfunction, using serial echocardiograms to assess for its reversibility. But for the organ procurement organizations (OPOs) that manage donors prior to organ procurement, this ‘wait and see’ approach can be resource-intensive and can compromise the quality of other solid organs, particularly when the donor is haemodynamically unstable.6 Offering OPOs and transplant centres the ability to predict which donors hearts are likely (vs. unlikely) to recover normal LV function – and thus warrant a ‘wait and see’ approach – was the aim of our current study. We used data from the DHS, an observational prospective cohort study of 4333 potential heart donors enrolled at eight US OPOs (listed in online supplementary Appendix S1) from February 2015 to May 2020. Adult (age 18–65 years) donors were eligible for study inclusion if they experienced brain death and tested negative for human immunodeficiency virus. Those lacking authorization for research from the donor's family and those not being considered for cardiac donation (e.g. due to coroner restriction or pre-existing cardiovascular disease) were excluded. An initial transthoracic echocardiogram (TTE) was performed on all enrolled donors. If real-time local interpretation identified LV dysfunction, defined as a LV ejection fraction (LVEF) <50%, a second TTE was performed 24 ± 6 h later. All echocardiograms were later interpreted by an expert ‘core’ reviewer (JZ), who was blinded to the local interpretation and all clinical data. Our analytic sample included all DHS donors who (i) had initial LV dysfunction per core TTE interpretation, and (ii) had a second TTE performed per study protocol. Some DHS enrollees (n = 269) met the first criterion but not the second, for reasons detailed previously.3 For the purposes of this analysis, we excluded donors in whom initial LV dysfunction was only ‘borderline’ (i.e. LVEF 46–50%); given that LVEF measures are subject to inherent uncertainty, such borderline cases might not definitively represent true LV dysfunction. Our primary outcome was ‘reversibility’, deemed present if core interpretation of the second TTE showed either (i) LVEF ≥50% or (ii) an increase (compared to initial TTE) in LVEF by ≥10 percentage points. We computed standardized mean differences (SMDs) to assess the magnitude of differences in donor characteristics by outcome status. Of note, troponin and N-terminal pro-B-type natriuretic peptide (NT-proBNP) were measured using varying assays at different centres; each was converted to a single measure as detailed in online supplementary Appendix S1. A multivariable logistic regression model was fit on the primary outcome using a limited subset of candidate predictors (listed in Table 1), all obtained at the beginning of donor management. The rationale for their selection is further detailed in online supplementary Appendix S1. A conditional inference (‘decision’) tree was constructed based on candidate predictors, as detailed in online supplementary Appendix S1. Donors with reversible (n = 130; 70.2%) and non-reversible (n = 55; 29.8%) LV dysfunction are compared in Table 1. Mean NT-proBNP was more than double in non-reversible (vs. reversible) cases (10 270 vs. 4454 pg/ml; SMD = 0.51). Lactate, troponin, and the prevalence of acidaemia were also higher in non-reversible cases, but these differences were modest (SMD <0.2 for all). Each of these associations persisted after multivariate adjustment (Figure 1A), but only that with NT-proBNP was statistically significant (p = 0.04). Mean LVEF (unadjusted) differed minimally by subgroup (35.8% vs. 35.6%; SMD = 0.02). The conditional inference algorithm identified NT-proBNP level as a significant predictor of reversibility, at a threshold value of 11 136 pg/ml. In Figure 1B, this threshold is ‘rounded’ to 11 000 pg/ml for ease of clinical application. Donors with NT-proBNP below this threshold (n = 167) exhibited reversibility in 74% of cases. Donors with NT-proBNP above this threshold (n = 18) exhibited reversibility in only 33% of cares. NT-proBNP below this threshold was associated with significantly higher acceptance for HT (47% vs. 5.6%; p <0.001 by chi-squared test). With respect to the outcome of reversibility, our algorithm has a measured sensitivity, specificity, positive predictive value, and negative predictive value of 0.95, 0.22, 0.74, and 0.67, respectively. In summary, we present a simple algorithm using two commonly measured donor clinical variables that can help predict whether donor LV dysfunction after brain death is reversible. Its use by OPOs and transplant centres could inform decisions made during donor evaluation and potentially avoid discard of viable donor hearts. That LVEF and NT-proBNP predict reversibility aligns with clinical intuition in a way that lends credence to our model. But exactly why they predict reversibility is unclear, as LV dysfunction after brain death is a poorly understood phenomenon. According to the ‘neurocardiogenic’ hypothesis, brain death prompts profound sympathetic nervous system activation and catecholamine release, resulting in myocardial injury.7 NT-proBNP is a non-specific biomarker of cardiac injury and may simply reflect the severity of this neurocardiogenic cascade. But its profound association with reversibility – independent of LVEF itself and larger than that of other non-specific biomarkers (e.g. lactate, troponin) – suggests that it may be more than a bystander. Its broader prognostic utility (e.g. in assessing donors with normal LV function but other risk factors) warrants further study. Some limitations regarding the measurement and interpretation of NT-proBNP and B-type natriuretic peptide (BNP) should be acknowledged. Both can vary temporally in association with haemodynamic changes, which our use of a single measure could not capture. Combining BNP and NT-proBNP into a single variable was necessary but not ideal; these biomarkers have different half-times, stabilities, and metabolic pathways and the conversion formula we used has not been validated in multiple populations. Our prediction algorithm is derived from (and thus applicable to) only donors who were considered for HT and lacked obvious contraindications. Sample size precluded our consideration of factors (e.g. age, body mass index, cause of death) that are weakly associated with reversibility and may be relevant to consider alongside the broader clinical context. While we identified initial LVEF as weakly predictive of reversibility, we lacked TTE data on global longitudinal strain and Doppler parameters; these markers of LV function may be more sensitive and predictive of myocardial recovery. Finally, we caution against use of our algorithm to ‘rule out’ potential donors with NT-proBNP ≥11 000 pg/ml, given their scarcity (n = 18) in our sample. Prospective validation of our algorithm in a larger cohort will increase the confidence of this finding. For now, a more appropriate use is to tentatively ‘rule in’ those with the more common phenotype (NT-proBNP <11 000 pg/ml), and thereby expand the use of available donor hearts for transplantation. This work was supported by funding from the National Heart, Lung and Blood Institute at the National Institutes of Health (F32 HL154750 to B.W. and R01 HL125303 to K.K.). Conflict of interest: none declared. Appendix S1. Supporting Information. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.