Utility of Right Ventricular Mass Ratios in Recipients With Pulmonary Hypertension: A United Network for Organ‐Sharing Database Analysis

ABSTRACT

Purpose

There is a lack of data reviewing the role of predicted cardiac mass ratio (PCMR) in transplant recipients with pulmonary hypertension (PH). In particular, the right ventricular (RV) mass ratio (RVMR) is a component of PCMR not yet studied separately for this purpose. We aimed to review the relationship of PCMR and RVMR with PH in terms of post‐transplant overall survival.

Methods

The United Network for Organ‐Sharing (UNOS) database was queried for primary heart transplant (HTx) patients from 2004 to 2023. Predicted RVMR was calculated for donors and recipients. A parsimonious multivariable model Cox model was formed for risk adjustment. Covariates with statistically significant relationships to the outcomes were further evaluated with tensor product spline models.

Results

A total of 37 904 patients were identified. After risk adjustment, diastolic pulmonary artery pressure (dPAP) was protective (HR 0.993 [0.989–0.997], p = 0.001), whereas increasing systolic pulmonary artery pressures (sPAPs) (HR 1.007 [1.005–1.009], p < 0.001) and diastolic pulmonary gradient (DPG) (HR 1.011 [1.007–1.015], p < 0.001) were associated with increased overall mortality. Increasing RV mass ratio (RVMR) was protective (OR 0.663 [0.589–0.747], p < 0.001)). This was appreciated in a restricted cubic spline analysis as a nonlinear relation was evident with any RVMR less than 1 (p < 0.001). Subsequent tensor product spline models evaluating the complex relationship between RVMR and pulmonary pressures demonstrated increasing relative hazards in patients with PH and smaller RVMR (p < 0.001).

Conclusion

In HTx patients with PH, RVMR ≥ 1 is significantly associated with a reduced hazard of overall mortality. This indicates RV undersizing may be harmful in the population.

Abbreviations

dPAP

diastolic pulmonary artery pressure

DPG

diastolic pulmonary gradient

HTx

heart transplant

mPAP

mean pulmonary artery pressure

PCMR

predicted cardiac mass ratio

PH

pulmonary hypertension

RV

Right Ventricle

RVMR

right ventricular mass ratio

sPAP

systolic pulmonary artery pressure

TPG

transpulmonary gradient

UNOS

United Network for Organ‐Sharing

1. Introduction

Preoperative pulmonary hypertension (PH) has been associated with a greater risk of morbidity and mortality post‐heart transplant (HTx) [1, 2]. This is partially a result of elevated pulmonary pressures increasing afterload on the right ventricle, leading to maladaptive myocardial remodeling and consequential right heart failure [3]. For this reason, severe PH is a relative contraindication for HTx [4]. Al this relationship has been studied previously, equivocal results have been obtained with dichotomized categories of the severity of PH and its effect on short and long‐term outcomes after HTx [5, 6, 7]. Due to this, it has been difficult to characterize the type and extent of the relationship between preoperative PH and outcomes in HTx recipients.

Nevertheless, for patients with PH undergoing HTx, there has been a historical preference for oversizing donor hearts [8] despite a paucity of data supporting this practice. Although the advantage of donor oversizing for cardiac transplant recipients with moderate PH is plausible, limited work has been done in this regard. For example, Shah et al. showed that donor under‐sizing was associated with a higher 90‐day and 1‐year mortality compared to oversized or matched donors in recipients with moderate PH [8].

In the context of donor heart sizing, PH, and HTx outcomes, many important factors are involved, including sizing metrics such as donor to recipient ratio of predicted cardiac mass (PCM)—subdivided into donor to recipient left ventricular mass ratio (LVMR) and right ventricular mass ratio (RVMR), measures of PH such as transpulmonary gradient (TPG), diastolic pulmonary artery pressure (dPAP), and mean pulmonary artery pressure (mPAP), and measures of outcome such as survival and primary graft dysfunction (PGD). In such a complex interplay of factors, arbitrary categorization of PH is likely not granular enough to illustrate the trends of association.

Thus, we attempted to depict how the relationship between cardiac mass ratios and pulmonary pressures interacts with post‐HTx outcomes. We took the total donor to recipient PCM ratio (PCMR) and RVMR as measures of cardiac size given the utility of PCMR for predicting mortality post‐HTx [9, 10] and the direct effect of elevated pulmonary pressures on the RV [3, 11]. This was assessed in the context of PA pressures, TPG, and diastolic pulmonary gradient (DPG) [12, 13, 14]. Our primary outcome measure was overall mortality. With this, we aimed to provide a unique perspective in illustrating the relationship between donor heart sizing, recipient preoperative PH, and outcomes post‐HTx in a continuous manner.

2. Methods

2.1. Database Utilization

A thorough search of deidentified clinical data from the United Network for Organ‐Sharing (UNOS) database was conducted to identify all adult patients undergoing primary HTx from January 1, 2004 to December 31, 2023. Primary HTx was defined as a patient's initial heart‐only transplant. Multi‐organ transplants were excluded. Waitlist and follow‐up databases were combined using transplant case identifying numbers leading to the creation of a single dataset. This cohort was utilized further in the analysis of clinical characteristics, outcomes, and survival data. Due to the deidentified nature of the study was exempt from the Institutional Review Board.

2.2. Clinical Calculations

Using the validated MESA equations [15, 16], recipient and donor left ventricular, right ventricular (RV), and total PCMs were estimated using gender, height, weight, and age. Total PCM was calculated by adding left and RV mass. PCMR was calculated by dividing donor mass by recipient mass. TPG was calculated by subtracting pulmonary capillary wedge pressure (PCWP) from mPAP. DPG was calculated by obtaining the difference of dPAP from PCWP.

2.3. Handling of Missing Data

Missing data were characterized using the validated missing at random (MAR) and missing not at random (MNAR) approach [17]. MNAR data were addressed using practical and clinical acumen to adjust the reported data. MAR data (primarily continuous variables) went through a data imputation process. This process was executed using a supervised machine learning technique with the creation of a Bayesian network [18].

2.4. Handling of Outliers

All included variables were plotted for visual inspection as well as reviewed manually by creating summary statistics to identify miscoded data as well as obvious outliers that were not clinically plausible. Simple outliers that were considered plausible in a clinical context were included in the analysis.

2.5. Initial Modeling and Factor Selection

Cox proportional hazard models were used for survival analysis. To assure parsimonious models, a backward stepwise regression technique was applied to identify clinically important variables and select the most valuable model. Relevant baseline, clinical, operative, and hemodynamic parameters were included in the initial model. This included: recipient age, donor age, ischemic time, days on waitlist, preoperative creatinine, cigarette use, diabetes history of recipient, prior sternotomy, dialysis requirement, recipient and donor sex match, preoperative diastolic PA pressure, preoperative systolic PA pressure, total cardiac mass ratio, donor to recipient RV mass ratio, and DPG.

2.6. Restricted Cubic Spline and Tensor Spline Analysis With Local Effect Curve

Covariates with significant relationships to our main outcome (survival) were subsequently included in initial restricted cubic spline and subsequent tensor spline analysis to assess for complex nonlinear relationships between relevant continuous variables [19]. The foundation of this analysis is built on multiple restricted cubic splines of each covariate placed in three‐dimensional graphical representation to model the association of RVMR and pulmonary pressures with patient survival. The local effect, also known as the conditional effect, of RVMR in setting of the spectrum of systolic pulmonary artery pressure (sPAP) was calculated. Hazard ratio was calculated at a baseline RVMR of 1.0 for every 0.2 unit increases across the entire documented spectrum of sPAP. This was then plotted on a curve. Values less than 1.0, indicated by the dotted line, would be deemed protective. All included covariates had significant total as well as nonlinear p values illustrating the validity of the relationship between them and outcomes. Given associations with cardiac mass sizing, analyses were adjusted for patient age and sex‐matching. Statistical analysis and modeling were done using R statistical software, version 4.2.3 (R Foundation for Statistical Computing, Vienna, Austria) and p values < 0.05 were considered statistically significant for all analysis.

3. RESULTS

3.1. Demographics and Clinical Characteristics

A total of 37 904 patients meeting the inclusion criteria were identified in the UNOS database. Median age was 55 (IQR 46, 61) years, and 23.8% were female. The most common indications for cardiac transplant were idiopathic dilated cardiomyopathy (34.9%) and ischemic heart disease 29.8%. Median total ischemic time was 3 h (IQR 2.3, 3.7), while median time on waitlist was 88 (IQR 26, 248) days. Further demographic details are provided in Table 1.

TABLE 1.

Demographic characteristics.

Variable	Total
Age, median [IQR]	55 [46, 61]
Females, n (%)	14 833 (23.8)
Height (cm)
Recipient, median [IQR]	175.0 [167.6, 180.3]
Donor, median [IQR]	175.0 [168.0, 180.3]
Weight (kg)
Recipient, median [IQR]	79.4 [68.0, 91.0]
Donor, median [IQR]	77.1 [68.0, 89.4]
Ischemic time (h), median [IQR]	3.0 [2.3, 3.7]
Days on waitlist, median [IQR]	88 [26, 248]
Recipient creatinine (mg/dL), median [IQR]	1.2 [1.0, 1.5]
Prior cardiac surgery, n (%)	7841 (12.6)
Ethnicity, n (%)
White	46 086 (74.1)
Black	10 009 (16.1)
Hispanic	4026 (6.5)
Asian	1433 (2.3)
Multiracial	246 (0.4)
American Indian/Alaska Native	197 (0.3)
Native Hawaiian/Other Pacific Islander	144 (0.2)
Unknown	28 (0.0)
Recipient primary diagnosis, n (%)
Idiopathic dilated cardiomyopathy	21 688 (34.9)
Ischemic cardiomyopathy	18 522 (29.8)
Coronary artery disease	8405 (13.5)
Other	4542 (7.3)
Unspecified cardiomyopathy	2610 (4.2)
Valvular heart disease	1436 (2.3)
Familial cardiomyopathy	1268 (2.0)
Hypertrophic cardiomyopathy	1107 (1.8)
Post‐partum cardiomyopathy	767 (1.2)
Viral cardiomyopathy	708 (1.1)
Congenital heart disease	544 (0.9)
Idiopathic restrictive myopathy	371 (0.6)
Alcohol related cardiomyopathy	253 (0.4)
Life support at listing, n (%)	29 110 (46.8)
Not requiring dialysis between listing and transplant, n (%)	48 958 (78.7)
Cigarette use, n (%)	14 850 (23.9)
Cigarette use > 10 pack years, n (%)	8251 (13.3)
Diabetes, n (%)	24 480 (39.4)
Length of stay (days), median [IQR]	18 [12, 23]
Time to last follow‐up (years), median [IQR]	5.7 [1.9, 10.9]

3.2. Cox Proportional Hazards

The parsimonious multivariable model derived from backward stepwise logistic regression is shown in Table 2. TPG, donor age, history of diabetes, and prior sternotomy were not selected via the stepwise regression. In terms of patient characteristics, increasing age (OR 1.013 [1.011–1.014], p < 0.001), elevated preoperative creatinine (OR 1.091 [1.074–1.108], p<0.001), history of cigarette use (OR 1.270 [1.222–1.319], p<0.001), and dialysis requirement preoperatively (OR 1.808 [1.609–2.030], p<0.001) were associated with increased mortality. In terms of sex matching of donor and recipient, male to male (OR 1.069 [1.006–1.139], p = 0.030) transplants were associated with worse survival compared to female to female transplants. In terms of PA pressures, increasing dPAP was protective (OR 0.993 [0.989–0.997], p = 0.001), whereas increasing sPAP (OR 1.007 [1.005–1.009], p < 0.001) and DPG (OR 1.011 [1.007–1.015], p < 0.001) was associated with increasing mortality. Finally, increasing RV mass ratio was protective (OR 0.675 [0.599–0.761], p < 0.001).

TABLE 2.

Cox proportional hazard model.

Variable	OR	Lower 95% CI	Upper 95% CI	p value
Recipient age	1.013	1.011	1.014	<0.001
Ischemic time	1.076	1.058	1.095	<0.001
Days on wait list	1.000	1.000	1.000	0.010
Preoperative creatinine	1.091	1.074	1.108	<0.001
Hx of cigarette use	1.270	1.222	1.319	<0.001
Dialysis requirement	1.808	1.609	2.030	<0.001
Sex match: F to M	1.049	0.979	1.125	0.177
Sex match: M to M	1.069	1.006	1.136	0.030
Sex match: M to F	1.036	0.960	1.118	0.366
Diastolic PA pressure	0.993	0.989	0.997	<0.001
Systolic PA pressure	1.007	1.005	1.009	<0.001
RV mass ratio	0.675	0.599	0.761	<0.001
DPG	1.011	1.007	1.015	<0.001

3.3. Nonlinear Analysis

Given the significant associations described above, restricted cubic spline and tensor spline analysis was executed to look at complex relationships between these variables across a spectrum. In individual analysis, RVMR <1 had significantly worse survival in a nonlinear relationship (Figure 1, p < 0.001). In an isolated unadjusted analysis of total PCMR, a nonlinear analysis was not demonstrated and as such was not included in the final analysis (Supporting Information S1). In tensor spline analysis, RVMR was protective and was reviewed across the spectrum of sPAP and DPG values. As age was a statistically associated predictor, it was adjusted for in this analysis.

FIGURE 1.

Restricted cubic spline of survival across the spectrum of right ventricle (RV) mass ratio.

In Figure 2, patients with any degree of PH (∼>30 mmHg) only had improved survival with equivalent or oversized RV mass ratios. For patients with RMVR <1, only normal—sub 25 mmHg patients did well. Finally, a similar but more exaggerated trend was seen in reviewing DPG with RVMR (Figure 3). In this context, all patients with roughly an RVMR <0.9 had worse survival, and true survival benefit was conveyed at an RVMR roughly >1.1.

FIGURE 2.

Right ventricular mass ratio (RVMR) in relation to systolic pulmonary artery pressure (sPAP) and hazard of death adjusted for age.

FIGURE 3.

Right ventricular mass ratio (RVMR) in relation to diastolic pulmonary gradient (DPG) and hazards of death adjusted for age.

In Figure 4, it is appreciated that minimally impacted survival at all sPAPs below roughly 45 mmHG. At approximately 50–52 mmHg, the conditional effect hazard ratio curve crosses 1.0 and therefore represents a protective effect. At its peak at severely elevated sPAP, increasing RVMR by 0.2 units was up to 22% more protective.

FIGURE 4.

Local effect curve of increasing right ventricular mass ratio (RVMR) across the spectrum of systolic pulmonary artery pressure (sPAP).

4. DISCUSSION

Our work here clearly demonstrates that in risk‐adjusted analyses, RV undersizing is associated with worse overall mortality. In the context of clinically relevant measures of PH, those patients with PH in particular benefit from an RVMR >1. The association with RV size matching is complex and likely plays a role in the conflicting messages published previously on the topic.

Studies assessing the pros and cons of oversizing or undersizing the heart in transplant, particularly in patients with PH, have conflicting results [20, 21, 22, 23]. Importantly, a well‐cited study by Shah et al. has shown no survival benefit for oversizing donors with pHTN [8]. However, a new study by Pasrija et al. does demonstrate that there is potential utility in upsizing the RV to mitigate the negative physiologic consequences of elevated peripheral vascular resistance in PH [24]. Our hypothesized explanation for this variation in outcomes, and difference in our reported results, is centered around the methodology and approach.

From the authors perspective, two areas of the research should likely be addressed:

1. Overall cardiac mass ratio is a complex metric that likely is not one‐size‐fits‐all. PCMR has been validated as a useful measure of cardiac size [8, 25], which has been associated with outcomes after HTx [26]. However, cardiac mass ratio as a whole is potentially too crude of a metric to assess post‐transplant survival in specific clinical situations such as PH. Assessing the RV and LV as individual entities may be more appropriate in certain patient populations. This has been recently demonstrated, as our group has shown that the RV may play an independent role and perhaps oversized importance in size matching compared to its weight relative to the LV [27]. This work reemphasizes the independent significance of the RV with the results of the reported Cox model, where increasing isolated RVMR is protective. This is best demonstrated with the results of the reported Cox model where increasing isolated RVMR is protective. Traditionally, total cardiac mass has been utilized for matching. However, for example, in two PH patients, one with an oversized RV + undersized LV and one patient with the commensurate inverse sizing would have the potentially exact same total cardiac mass ratio but very different ventricular sizes and associated outcomes. This may lead others to believe cardiac size matching does not impact outcomes in this population, where assess the RV in isolation may elucidate important findings and improve outcomes.

2. The current utilization of PH metrics are insufficient to assess nuanced variations in clinical condition. A recurring theme in these studies is the varied definition and approach of PH relevant pressure variables (mPAP, dPAP, sPAP, TPG, DPG, etc.). Most frequently, these continuous variables are binned or placed into categories without consistency. Analyzing these variables in such a way greatly limits the granularity of the analysis [28, 29, 30]. The interplay of these pressures requires thoughtful review. For example, this study demonstrates that when accounting for sPAP and DPG, increasing dPAP is deemed protective. This finding is somewhat counterintuitive, however likely explained when for a given sPAP, elevated dPAP may be beneficial. When looking at dPAP in isolation, increasing dPAP would be associated with worse mortality. This may also be an example where the spectrum of dPAP impact on survival is likely nonlinear. Continuous analysis such as restricted cubic splines and tensor product spline analysis are free from this limitation and may represent the data more accurately [19].

As a result, the authors hypothesize that when reviewing this complex clinical question through a more granular lens, the following findings can be appreciated. As seen in Figures 1, 2, 3, patients grossly perform better with an RV that is oversized. Similar benefit of oversizing RV in PH has been reported by Cheshire et al. [10]. One can appreciate that when PH is present an RVMR <1 is substantially associated with increasing risk‐adjusted overall mortality. This result has biologic plausibility as an oversized donor RV may already be remodeled in an already hypertensive environment, or simply a de novo larger RV is more ready to compensate early after implantation. Most significantly, based on the findings of demonstrated in Figure 4, a specific clinical cutoff of approximately sPAP of greater than 50 mmHg is where RVMR is noted to be statistically protective. This builds on the existing literature supporting oversized RV and adds novel features to this study that can allow for clinical action by providers [24].

The etiology of increased mortality with a lower RVMR is difficult to discern; however, the hypothesized cause would be primary graft dysfunction. Unfortunately, a limitation of this database is the historically irregular definitions of PGD and poor coding on the variable, making a more detailed assessment not feasible. However, prior literature would support this perspective. The initial survival disadvantage of undersizing might be due to the lag in RV adapting to the new afterload. Over time, as the RV adapts to the elevated pulmonary pressures, survival also may become comparable to patients without PH [6, 8], but not totally reach equivalent levels as demonstrated in our analysis.

The role of donor heart sizing in recipients with PH may not be the same as in patients without it since PH is an independent predictor [6, 31, 32, 33, 34, 35] of post‐HTx outcomes. In those patients with PA pressures within normal limits may not require a focus on heart sizing. However, it appears the findings presented above are important with real‐world application and utility. Patients with PH are a unique candidate for HTx and require a different perspective when organ selection takes place, and a strong focus on RV sizing may improve outcomes in this population.

4.1. Limitations

This study is not without its limitations. Inherently, database analyses struggle with data granularity and missing information. This was dealt with appropriately to assure a representative sample was created to address missing data. Moreover, patients with severe PH usually do not get listed and donor hearts also do not usually get too undersized. Both these realities could hint at the presence of selection bias in the database and as such, findings should be taken in this context. Additionally, there is a complex ecosystem of pharmaceutical and technical methods including mechanical circulatory support that impacts outcomes in patients with elevated pulmonary pressures. This was unable to be adequately adjusted for in this analysis given the nature of the data provided. Also, as the world of deceased cardiac death donors grows, how PH impacts these grafts and the potential need for changes in size‐matching protocols will need to be addressed. Unfortunately, this question could not be addressed as data were not organized based on brain or cardiac death donors. Finally, as mentioned in the discussion, there is a variability in the definition of PGD. Importantly, the International Society of Heart and Lung Transplantation (ISHLT) standardized the definition of PGD less than a decade ago. However some transplant centers had been defining it according to their own criteria impacting the accuracy of the variable as presented—limiting analysis [36].

5. CONCLUSION

Utilizing a novel, more granular statistical approach, a strong survival benefit is appreciated when oversizing the donor RV across increasing pulmonary pressures in HTx recipients. This indicates that in the unique population of patients with PH, isolated ventricular size matching, in addition to total cardiac mass, may be more clinically applicable. Thus, oversizing the RV and should be considered in HTx candidates with PH.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting Information S1: Restricted cubic spline of survival across the spectrum of total cardiac mass ratio.

Data Availability Statement

The data that support the findings of this study are available from UNOS. Restrictions apply to the availability of these data, which were used under license for this study. Data are available from the author(s) with the permission of UNOS.