Optimal patient selection for simultaneous heart-kidney transplant: a modified cost-effectiveness analysis

Abstract

Increasing rates of simultaneous heart-kidney (SHK) transplant in the United States exacerbate the overall shortage of deceased donor kidneys (DDK). Current allocation policy does not impose constraints on SHK eligibility, and how best to do so remains unknown. We apply a decision analytic model to evaluate options for heart transplant (HT) candidates with comorbid kidney dysfunction. We compare SHK with a “Safety Net” strategy, in which DDK transplant is performed six months after HT, only if native kidneys do not recover. We identify patient subsets for whom SHK using a DDK is efficient, considering the quality-adjusted life year (QALY) gains from DDKs instead allocated for kidney transplant-only. For an average-aged candidate with 50% probability of kidney recovery after HT-only, SHK produces 0.64 more QALYs than Safety Net at a cost of 0.58 more kidneys used. SHK is inefficient in this scenario, producing fewer QALYs per DDK used (1.1) than a DDK allocated for KT-only (2.2). SHK is preferred to Safety Net only for candidates with a lower probability of native kidney recovery (24 – 38%, varying by recipient age). This finding favors implementation of a Safety Net provision and should inform the establishment of objective criteria for SHK transplant eligibility.

1. Introduction

Kidney dysfunction is a common sequela of end-stage heart failure and afflicts many patients awaiting heart transplant (HT). Kidney function may recover after heart transplant alone (HT-only), particularly when acute and mild or moderate in severity (1). More chronic and severe kidney dysfunction may not be reversible after HT-only and often prompts consideration of simultaneous heart-kidney (SHK) transplant, the rates of which have doubled over the last decade (2).

This rise in SHK transplants has been facilitated by the current allocation system, in which all SHK candidates are prioritized above all candidates waiting for a kidney transplant alone (KT-only), with no standard criteria for SHK eligibility (3). In contrast, those with persistent kidney dysfunction after HT-only receive no such priority and, in the absence of a living donor, face the same expected wait time for a deceased donor kidney (DDK) as KT-only candidates (4). Such a policy produces an obvious incentive to favor SHK over HT-only whenever the reversibility of a HT candidate’s kidney dysfunction is in question.

A similar situation existed in liver transplantation until a new allocation policy for simultaneous liver-kidney (SLK) transplant was introduced in August 2017 (5). The SLK policy established a set of objective eligibility requirements for SLK and a “Safety Net” system, in which patients exhibiting irreversible kidney failure within the first year of liver transplant gain priority access to DDKs. Consistent with results of a prior modeling study (6), this policy reduced kidney utilization for SLK, particularly among those with mild or moderate kidney dysfunction, and with no adverse impact on kidney graft outcomes (7) or liver transplant recipient survival (8).

From the standpoint of an individual SHK candidate, the incentive favoring SHK is harmless and potentially beneficial (9–14). But on a system-wide level, the liberal utilization of SHK transplant comes at a substantial cost. Each kidney used in the over 200 SHK transplants performed annually (3) is one fewer available to the over 100,000 patients waiting for KT-only (4) - for most of whom a kidney transplant will improve life expectancy and quality of life (15,16). Those KT-only candidates who “lose out” on a DDK due to its allocation instead to SHK suffer measurably worse outcomes (17).

The Final Rule of transplantation (18) states that each organ should be allocated on the basis of medical judgement and not systemic biases (19). The allocation of a DDK to a HT recipient with reversible kidney dysfunction - and thus unclear benefit - ahead of a KT-only candidate with higher expected benefit reflects a systemic bias which violates this principle. To stem such “overuse” of kidneys for SHK, many have advocated for standardized SHK eligibility requirements (3,14,17,20–24) and an accompanying Safety Net provision. (20–22)

No studies have provided quantitative support for this policy. As a randomized controlled trial would be infeasible, we apply a decision analytic model to quantify the trade-off between costs and benefits of SHK and to identify subgroups for which SHK is optimal from a societal standpoint.

2. Methods

2.1. Overview

We apply a deterministic decision analytic model to assess expected outcomes for a single HT candidate with concurrent kidney dysfunction, starting from the time of their first transplant (t = 0, “time zero”) until death. We compare outcomes under two strategies. In the “SHK” strategy, SHK transplant is performed at time zero. In the “Safety Net” strategy, HT-only is performed at time zero and, if kidney failure ensues, kidney transplant is performed at t = 6 months. After the index transplant (either HT-only or SHK), a patient can experience early death or graft failure, which we define as occurring in the first six months post-transplant.

While real-world HT recipients can exhibit a varying degree and duration of kidney dysfunction after transplant, we make the simplifying assumption that kidney function after transplant is a binary variable representing either “persistent kidney failure” (K−) or “kidney recovery” (K+). We define “persistent kidney failure” as dialysis-dependence at t = 6 months and “kidney recovery” as the converse at t = 6 months. We acknowledge that “persistent” and “recovery” are misnomers for patients who have kidney dysfunction, but are not dialysis-dependent, prior to transplant; we adopt these terms as an accessible shorthand. Our model also ignores the varying degrees of heart allograft dysfunction that can occur post-HT, and defines a patient as having intact graft function (H+) or graft failure (H−).

Real-world candidates for SHK (or the “Safety Net” strategy) also exhibit varying degrees of kidney dysfunction prior to transplant. Our model does not include pre-transplant GFR or any other explicit measure of kidney dysfunction; however, it does represent the degree of pre-transplant kidney dysfunction using two parameters: “reversibility” and “SHK benefit ratio”. We define “reversibility” as the probability that a given HT-only recipient with kidney dysfunction enters the H+K+ state, in the absence of death or graft failure. “SHK benefit ratio”, defined mathematically below, indicates the expected survival benefit that SHK (vs. HT-only) will confer to a given patient. Clinical intuition and prior reports suggest that reversibility will be lower and benefit will be greater for those with more severe kidney dysfunction, including those already on dialysis before transplant (12). However, by modifying the “reversibility” and “benefit” parameters (as in the sensitivity analyses described below), our model is applicable to HT candidates with any degree of kidney dysfunction and/or dialysis-dependence.

Figure 1 shows the model schematic. We apply a decision tree which specifies outcomes at 6 and 12 months after the index transplant. Beyond 12 months, we apply population means for life expectancy and quality of life (QoL) weight to calculate the net benefit – in quality-adjusted life years (QALYs) - and cost – in number of DDKs used - of each transplant strategy. We account for diminished functional status and comorbidity in the early post-transplant period by applying a lower QoL weight in the first six months after index transplant. Our model employs simplifying assumptions which are detailed in Supporting Information (S1). We test the influence of selected assumptions in the sensitivity analyses described below.

Figure 1. Decision tree model schematic showing post-transplant health states and events under the a) SHK and b) Safety Net strategies.

Orange boxes represent transplant events and blue boxes represent health states occupied by patients after transplant. The short time to KT assumes the general availability of a Safety Net provision. A patient who does not experience early death may undergo a maximum of two total heart transplants and two total kidney transplants before entering one of three terminal states at 6 months or 1 year after index transplant. H+K+ indicates a “thriving” state with intact heart and kidney function, H+K− indicates a patient on long-term dialysis until death, and H− (K+ or−) represents a patient with end-stage heart failure (+/− dialysis-dependence).

SHK: simultaneous heart-kidney transplant; HT: heart transplant; KT: kidney transplant.

Reversibility, as defined above, is dependent on the chronicity and severity of kidney dysfunction along with other patient characteristics (25) and thus will vary significantly among SHK candidates. Little evidence exists to inform the extent of and predictors of reversibility after heart transplantation (21). Rather than specify an arbitrary single “base case” assumption, our analysis assesses outcomes across a wide range of potential values for reversibility. Such an approach enables us to identify the level of reversibility at which SHK is justified, accounting for the costs (DDKs used) of the SHK strategy.

2.2. Model Parameterization

To derive our model parameters, we use primarily the Scientific Registry for Transplant Recipients (SRTR), which contains data on all solid organ transplants in the United States including dates of transplant, graft failure, and death and serum creatinine and dialysis need pre-transplant and annually post-transplant (26). We supplemented these using data from existing literature where available, and clinical intuition when necessary. Table 1 details each parameter assumption and the corresponding reference or derivation.

Table 1.

Model parameterization for base case and sensitivity analyses.

Parameter	Base case value (plausible rangeg)	Derivation / reference
Epidemiologic parameters
SHK benefit ratioa	0.56 (0.45 – 0.80)	Represents a range of estimates from prior studies measuring the survival benefit of SHK. Base case, lower bound, and upper bound values are derived from Chou et al (2018),9 Gill et al (2008),12 Schaffer et al (2014),13 respectively.
Probability of earlyb death after SHK	9.0% (6.0 – 12.0%)	Reflects observed mortality after adult SHK transplants (2014 – 2018; n = 770)
Probability of early death after Safety Net KT	6.0% (4.0 – 8.0%)	Reflects observed mortality after adult kidney transplants (August 2017 – February 2019; n = 370) occurring within one year of liver transplant; rationale is further detailed in Supporting Information (S4).
Relative risk of early death after re-transplant, compared to initial transplant of same organ	1.37 (1.25 – 1.49)	Barghash and Pinney (2020)32
Probability of early graft failure after HTc	0.3% (0.2 – 0.4%)	Reflects observed heart graft failure in a cohort including all 1) SHK transplants 2) HT among patients on dialysis pre-transplant 3) HT among patients with creatinine ≥ 2.3 pre-transplant (2014 –2018; n = 1064)
Probability of early graft failure after kidney transplantc	2.1% (1.4 – 2.8%)	Reflects observed kidney graft failure in a cohort including all 1) SHK transplants 2) kidney transplants occurring within one year of HT (2014 –2018; n = 872)
Expected survival duration (years), conditional upon survival to six months post-transplant, by stated
Thriving state (H+K+), average patient	17.7 (13.9 – 22.5)	Derived from observed post-HT survival among HT recipients surviving to six months post-transplant, as detailed in Methods. Sensitivity analyses were performed assuming a plausible range for H+K+ survival of 13.9 to 22.5 years, corresponding to old and young patients, respectively. Noting the small sample used to derive H+K- survival and resulting uncertainty, we assumed an especially wide plausible range (+/− 67%).
Thriving state, younge patient	22.5
Thriving stage, olde patient	13.9
On long-term dialysis (H+K−)	3.5 (2.0 – 5.0)
In end-stage heart failure (H−)	0.25 (0.1 – 0.5)	Derived from a reported mortality incidence after delisting from the HT waitlist due to clinical deterioration (VanderPluym et al 2014).33 Given absence of data specific to the post-HT population, we assume a wide plausible range.
Quality of life parameters
QoL weight in thriving (H+K+) state	0.76 (0.68 – 0.84)	Based on Emin et al (2016),34 Sharples et al (2006),35 Saeed et al (2008),36 Buendia et al (2011).37 Given consistency of estimates across multiple studies, we employed a narrower (+/− 10%) plausible range.
QoL weight in dialysis-dependent (H+K−) state	0.52 (0.44 – 0.60)	Dominguez et al (2011),38 Held et al (2016)31
QoL weight in end-stage heart failure (H−) state	0.35 (0.30 – 0.40)	Holland et al (2010)39
% decrease in QoL in first six months post-transplantf	20% (13 – 27%)	Based on Butler et al (2003);40 given the small sample size in this study, we employ a wider (+/− 33%) plausible range.

^aRefers to the odds ratio: [odds of early death among SHK recipients] / [odds of early death among HT-only recipients with no renal recovery after HT]; defined in this way, a lower value for this parameter represents a greater benefit to SHK, in terms of mortality reduction.

^b”Early” death and graft failure refer to events occurring in the first six months post-transplant.

^cConditional on survival to six months post-transplant

^dListed values refer to duration of survival beyond (i.e. not including) the first six months post-transplant.

^e“Young” and “old” refer to HT recipients in the lowest and highest age tertiles, respectively.

^fRefers to the decrement in QoL weight compared to that in the equivalent state after six months post-transplant

^gPlausible range is assumed to be +/− 15% for quality of life parameters and +/− 33% for all other parameters, unless otherwise indicated.

SHK: simultaneous heart-kidney transplant; KT: kidney transplant; HT: heart transplant; QoL: quality of life; OR: odds ratio

We derive probabilities of early death and graft failure by measuring the incidence of these events in the first six months post-transplant among relevant cohorts in SRTR. For example, our assumed probability of early death after SHK transplant (9.0%) reflects the prevalence of this event in real-world SHK transplant recipients (n = 770) from 2014 – 2018. We assume the same probability of early death for the subgroup of HT-only recipients who do experience renal recovery. For HT-only recipients who do not experience kidney recovery, the probability of early death is higher, reflecting the putative survival benefit of SHK in our population of interest. More precisely, we define a parameter “SHK benefit ratio” as the following odds ratio:

$$SHKbenefitratio=\frac{\left[oddsofearlydeathamongSHKrecipients\right]}{[oddsofearlydeathamongHT-onlyrecipientswithpersistentkidneyfailure]}$$

As defined above, a lower value for “SHK benefit ratio” represents a greater effect of SHK on survival, relative to HT-only. Where reversibility equals 1, 100% of HT-only recipients recover kidney function by 6 months and all are assumed to have the same probability of early death as SHK recipients (9.0%). SHK benefit ratio does not enter into this calculation at all; it only affects the probability of early death for those who do not recovery kidney function. In the base case, we assume SHK benefit ratio is 0.56 – this is mathematically equivalent to the hazard ratio (0.58) observed in a large recent study measuring the mortality effect of SHK (9), applied over a six-month period.

Patients who survive to six months after their last transplant are assigned a subsequent life expectancy and QoL weight, both of which depend on their heart and kidney function as detailed in Table 1. For the H+K+ state, we separately derive life expectancy for a “young”, “old”, and “average” patient using observed survival among the lowest, highest, and all age tertiles, respectively. The derivation of each life expectancy is detailed in Supporting Information (S2).

2.3. Analyses

Our primary outcomes include 1) probability of death within one year of the initial transplant, 2) QALYs, and 3) expected number of kidneys used for transplant, each calculated on a per patient basis. Initial SHK transplant, compared with HT-only followed by Safety Net KT, should produce higher expected QALYs under reasonable parameter assumptions but at a greater cost of more kidneys used. To express the tradeoff between this benefit and “cost” we utilize a “modified incremental cost-effectiveness ratio (ICER),” as defined previously (6):

$$modifiedICER=\frac{\left[\left(QALYswithSHK\right)-\left(QALYswithSafetyNet\right)\right]}{[\left(kidneysusedwithSHK\right)-(kidneysusedwithSafetyNet\left)\right]}$$

This reflects the expected number of QALYs gained per additional kidney used by pursuing SHK instead of Safety Net. A 3% annual discount rate is applied to accumulated QALYs.

We specify a willingness-to-transplant threshold, a previously defined metric (27), which when applied here represents the number of additional QALYs that a candidate must gain from the SHK strategy (compared to Safety Net) to offset its cost in terms of additional DDKs used. Per prior recommendations, we set our willingness-to-transplant threshold equal to the minimum QALY benefit that each incremental kidney would confer if allocated instead to someone on the KT-only waitlist (27). In other words, this threshold represents the “opportunity cost” of using a DDK for SHK. As detailed in Supporting Information (S3), we estimate the willingness-to-transplant threshold (opportunity cost) at 2.22 QALYs per DDK. Thus, for SHK to be the preferred strategy, it must confer at least 2.22 additional QALYs (relative to Safety Net) per additional kidney used.

We first calculate outcomes and modified ICERs separately for patients of varying levels of reversibility, ranging from 0 to 1, while holding SHK benefit ratio constant at its base case value (0.56). We highlight results for selected values of reversibility including “low” (10%), “moderate” (25%), and “high” (50%). We perform each of these analyses separately using life expectancies for an “average”, “young”, and “old” patient as defined above. As reversibility varies by patient and SHK benefit ratio is uncertain, we conduct a two-way sensitivity analysis in which both parameters are varied over a wide range. We identify those combinations of reversibility and SHK benefit ratio in which the SHK strategy is preferred and those combinations in which the Safety Net strategy is preferred – the two are delineated graphically by an indifference curve.

We test the sensitivity of our findings to assumptions for all input parameters across their plausible ranges (Table 1). We vary each input one at a time, otherwise applying base case assumptions, average post-transplant life expectancy, and reversibility equal to its “moderate” value (0.25). We conduct further scenario analyses testing the sensitivity of our findings to the model’s simplifying assumptions, which are detailed in Supplemental Table 1.

Analyses were conducted using SAS version 9.4 and Microsoft Excel 2016. The data reported here have been supplied by the Minneapolis Medical Research Foundation as the contractor for SRTR. The interpretation and reporting of these data are the responsibility of the authors and in no way should be seen as an official policy of or interpretation by the SRTR or the United States government.

3. Results

Table 2 shows outcomes under the base case assumptions and varying levels of life expectancy and reversibility of renal dysfunction. For a patient of average life expectancy (age 50–59 years at transplant) and moderate reversibility (25%), the Safety Net strategy results in an expected 9.3 QALYs, 0.65 kidneys used, and 17% one-year mortality. Compared to Safety Net, the SHK strategy results in 0.98 more QALYs, 0.37 more kidneys used, and an absolute reduction in one-year mortality of 8.1%. This corresponds to a modified ICER of 2.66 QALYs per additional kidney used.

Table 2.

Outcomes of SHK compared to Safety Net strategy at different levels of reversibility and life expectancy. Each outcome is expressed as the difference between SHK and Safety Net on a per patient basis. “Reversibility” refers to the probability that a patient will have kidney recovery after HT-only. “Young” and “old” refer to life expectancy among HT recipients in the lowest and highest age tertiles, respectively.

	Reversibility
	Low (10%)	Moderate (25%)	High (50%)
Absolute difference in one-year mortality (SHK vs. Safety Net)	−9.8%	−8.1%	−5.4%
Additional kidneys used (SHK vs. Safety Net)	+0.24	+0.37	+0.58
Expected QALYs gained:
Young (18 – 49 years)	1.41	1.17	0.77
Average age (50 – 59 years)	1.18	0.98	0.64
Old (≥ 60 years)	0.98	0.81	0.54
Modified ICER (QALYs/kidney):
Young (18 – 49 years)	5.89	3.17	1.31
Average age (50 – 59 years)	4.94	2.65	1.10
Old (≥ 60 years)	4.11	2.21	0.92

SHK: simultaneous heart-kidney transplant; HT: heart transplant

For a patient with low reversibility (10%), more kidneys are used and slightly poorer survival results under both strategies. However, such a patient experiences greater benefits (defined by a greater gain in QALYs, greater reduction in one-year mortality, and higher modified ICER) from the SHK (compared to Safety Net) strategy than a patient with moderate or high reversibility. These benefits of SHK (vs. Safety Net) decrease continuously as reversibility increases from 0 to 1 (Figure 2). A young (age 18 – 49 years at transplant) patient experiences greater than average benefits (as defined above) from SHK; the opposite is true for old patients (age ≥ 60 years at transplant).

Figure 2. How reversibility modifies the effects of SHK (compared to Safety Net) strategy on selected outcomes.

Outcomes include QALYs gained (a), number of kidneys used per patient (b), QALYs gained per kidney used (c), and absolute reduction in one-year mortality (d), all expressed on a per patient basis. Each is plotted as a function of “reversibility”, defined here as the probability that a patient will have recovery of native kidney function after HT-only. Results are obtained assuming average post-transplant life expectancy and the base case values of all other model parameters, as listed in Table 1.

SHK: simultaneous heart-kidney transplant; HT: heart transplant; QALY: quality-adjusted life year

Figure 3 displays indifference curves, which delineate conditions under which SHK is preferred to Safety Net. Areas to the upper-right of each curve represent scenarios in which the Safety Net strategy is preferred, based on the specified willingness-to-transplant threshold (as defined in Methods). The indifference curves shift to the lower-left with increasing age, indicating that older patients have a broader range of scenarios in which Safety Net is preferred to SHK. As Figure 3 demonstrates, SHK is consistently preferred for patients with low (10%) reversibility and Safety Net is consistently preferred for patients with high (50%) reversibility, across all plausible values of age and SHK benefit ratio.

Figure 3. Indifference curves for the comparison of SHK and Safety Net strategies.

Each indifference curve is derived using a distinct set of assumptions for 1) the age of the potential SHK recipient and 2) ) the assumed willingness-to-transplant threshold (“opportunity cost”) per DDK used (as defined in Methods). Each panel shows indifference curves corresponding to the base case (2.22 QALYs per DDK), upper bound (2.77 QALYs per DDK), and lower bound (1.66 QALYs per DDK) values for willingness-to-transplant threshold. Selected points on each curve are labeled with the value of reversibility corresponding to the base case value of SHK benefit ratio (0.56).

SHK: simultaneous heart-kidney transplant; QALY: quality-adjusted life year; OR: odds ratio; DDK: deceased donor kidney

For patients of moderate (25%) reversibility, the base case preferred strategy varies depending on assumed post-transplant life expectancy and SHK benefit ratio but is robust to plausible variation in all other input parameters (Figure 4). Our base case findings are also robust to scenarios in which 1) the timing of early death and re-transplant are varied, 2) kidney re-transplants are not possible, 3) the risks of kidney and heart graft failure are correlated, and 4) graft failure risk differs for SHK vs. single-organ transplant. Conditions and outcomes in each of these scenarios are further detailed in Supplemental Table 1.

Figure 4. Tornado diagram of the results of sensitivity analyses.

For plausible ranges of each input variable (shown in parentheses), the range in modified ICER estimates for the SHK (compared to Safety Net) strategy is shown. The chart includes all input variables for which the corresponding range in ICERs is greater than 0.1 QALYs per kidney. Each parameter is varied in isolation while otherwise applying base case assumptions, average post-transplant life expectancy, and reversibility equal to its “moderate” value (0.25). The base case willingness-to-transplant threshold (2.22 QALYs per DDK) is indicated by a dashed line.

SHK: simultaneous heart-kidney transplant; ICER: incremental cost-effectiveness ratio; QALY: quality-adjusted life year; OR: odds ratio; QoL: quality of life; DDK: deceased donor kidney

4. Discussion

In this study, we evaluate transplant strategies for HT candidates with concurrent kidney failure who could be considered for either SHK or HT-only followed by expedited KT (Safety Net strategy). We define the parameters under which each strategy is preferred, assuming that a Safety Net provision is implemented and accounting for benefits accrued by the alternative use of each DDK for KT-only. We estimate this “opportunity cost” at 2.22 QALYs per kidney, which represents the estimated benefit to transplant for the marginal KT-only candidate. We find that for a large subgroup of SHK candidates, the benefits of SHK over Safety Net are relatively small compared to the opportunity cost. Specifically, deceased donor SHK transplant is preferred to Safety Net only when the probability of kidney recovery after HT-only (“reversibility”) is less than 30%. If we adopt a lower bound estimate for the opportunity cost (1.66 QALYs per kidney), then SHK is preferred for a larger subgroup of candidates – all with reversibility less than 38%.

We find that the threshold at which SHK is preferred to Safety Net varies by the SHK candidate’s age. Candidates aged 18 – 49 accrue greater than average QALY gains from SHK; the opposite is true among candidates aged 59 and older. As a result, SHK is preferred for the younger subgroup whenever the probability of kidney recovery after HT-only (“reversibility”) is less than 35%; for older patients, SHK is preferred only when reversibility is less than 25%. This finding reflects the observation that patients with longer expected post-transplant survival accrue greater survival benefit from transplantation (15). While we examined only age, other patient characteristics affecting long-term survival would similarly modify our results.

Our study uses the base case assumption per Chou et al (9) that SHK reduces the odds of early mortality by 44%. Estimates of this effect are similar in a smaller, more recent study (14), but higher (12) and lower (13) in other studies. All are observational and subject to potential selection bias, which could exaggerate the apparent benefit of SHK (20). A consistent finding is that this benefit is greater among patients with more advanced kidney dysfunction prior to transplant (12–14,20). Our model incorporates such heterogeneity by assuming that SHK reduces mortality only for patients who do not experience kidney recovery with HT-only. Acknowledging that this parameter is both influential and uncertain, we present a two-way sensitivity analysis that varies the SHK mortality benefit over a wide range.

4.2. Future directions for research and policy

Our study is the first to demonstrate quantitatively what many have argued on qualitative grounds (3,14,17,20–24) - that unconstrained use of multi-organ transplants, including SHK, fails to maximize the benefit from a limited supply of deceased donor kidneys. This finding supports the adoption of standard eligibility criteria for SHK combined with a Safety Net provision.

The design of eligibility criteria poses a challenge for policymakers. We find that post-transplant life expectancy and reversibility of kidney dysfunction are the pivotal factors that dictate the relative costs and benefits of SHK. As the extent and predictors of reversibility are not yet known, defining eligibility for SHK will require synthesizing prior and future research to identify predictors of kidney recovery after HT, perhaps in the form of a prediction score. Prior studies indicate that up to 50% of SLK recipients have recovery of significant native kidney function after SLK, when assessed by radionuclide scan (25,28). If roughly the same is true for SHK recipients under the status quo, then an application of our “reversibility threshold” to limit SHK eligibility would significantly reduce SHK volume and ensure allocation of organs to patients who truly need them.

From a utilitarian perspective, our findings support differential prioritization of SHK candidates based on age and other prognostic factors, perhaps by assigning them to “tiers” based on their expected post-transplant survival (29). While equity needs to be considered, there are widely accepted and publicly vetted paradigms for incorporating post-transplant outcomes into allocation decisions. For example, the Kidney Allocation System preferentially assigns the highest quality kidneys to candidates with the highest post-transplant survival and similar policies are under development for the allocation of other solid organs. On a local programmatic level, the decision to list a candidate for transplant is often informed by their expected post-transplant survival (30).

Our analysis takes a systems-level perspective with the objective of maximizing the societal benefit from a finite supply of organs under conditions of scarcity. We acknowledge that this perspective can produce findings that are at odds with the interests of specific individuals. For example, we find that Safety Net is preferred for a patient with relatively high (50%) probability of renal recovery after HT-only - despite the fact that such a patient would still benefit from SHK, gaining 0.64 QALYs and a 5% absolute reduction in one-year mortality. Implementing a new SHK allocation policy that balances these competing interests – while practically and ethically challenging – would clearly improve on the status quo which systematically discounts the interests of KT-only candidates by prioritizing all SHK candidates first.

4.3. Study Limitations

Our deterministic decision tree model requires some simplifying assumptions. Relaxing these assumptions produced no significant difference in our findings, suggesting that a more complex model is unlikely to produce more accurate results. We acknowledge that HT recipients with renal dysfunction have a highly variable post-transplant course. Precisely replicating this would have required a microsimulation model with more parameters and attending assumptions. This is particularly true over the first six months, in which for example, one patient may require dialysis for just hours or days and another may require dialysis throughout, complicated by multiple hospital readmissions.

Our model does not explicitly capture this variation, instead defining patients only by whether or not they are dialysis-dependent at the end of this complex period. However, the salient features of this early post-transplant course – for the purposes of our modified cost-effectiveness analysis – are a patient’s morbidity (represented by QoL weight) and mortality (represented by “probability of early death”). Even when using a broad plausible range for these parameters in our sensitivity analysis (e.g. +/− 33% for mortality parameters), our base case findings are qualitatively unchanged. This suggests that instead using a microsimulation model, which (in theory) could more precisely characterize morbidity and mortality for a given patient, would not have changed our primary findings.

We show in sensitivity analyses (Supplemental Table 1) that our findings are robust to many of the simplifying assumptions used in constructing the model. There are other simplifying assumptions that we could not test in this manner, as doing so would require a prohibitively complex model. For example, we estimated the “opportunity cost” of each DDK used for SHK by assuming that the alternative is its allocation for KT-only to the “marginal” recipient (i.e. a diabetic of age ≥ 65). In reality, the “marginal” recipient will vary by context and in most cases be younger and healthier; particularly as the DDKs used for SHK tend to be of the “high-quality” sort that are often reserved for younger KT-only recipients (23). This would translate into a higher “opportunity cost” and imply that Safety Net is preferred for a larger majority of patients than our results suggest. This majority would be even larger, had we accounted for other “real-world” possibilities including 1) some Safety Net transplants could be performed using living donor kidneys (with no attached “opportunity cost”) and 2) SHK recipients experiencing kidney graft failure may not get priority for prompt re-transplant (as in our model), but instead may wait as long as KT-only candidates (as is current practice for SLK recipients).

As in any modeling study, our findings are subject to uncertainty in the parameter assumptions. Our sensitivity analyses demonstrate that for most parameters, this uncertainty has little bearing on our results and their policy implications. For those parameter assumptions that are influential and also heterogeneous across patients, we present our primary findings for variously defined patient subgroups to demonstrate their relative effects.

One such parameter - “SHK benefit ratio” – was estimated from observational studies that subject to confounding by indication. This warranted use of a wide plausible range (0.45 – 0.8) in our study, which spans the range of reported effect sizes (9–14). However, if all such studies are biased in the same direction, the true effect size may fall outside this range. This inherent limitation highlights the pressing need for further (ideally randomized) studies to better characterize the mortality benefit of SHK and how it varies by patient subgroup.

Our analysis does not account for economic costs, and the potential economic cost savings from doing two procedures simultaneously (i.e. SHK) rather than on separate hospitalizations. Doing so might make SHK appear more attractive from an individual patient and transplant program perspective. However, from a societal perspective, these “immediate” cost savings must be balanced against the savings accrued by doing more KT-only transplants under the Safety Net strategy. Per a recent estimate (31), a KT recipient incurs $146k less in lifetime costs than one who remains on chronic dialysis. The inclusion of economic costs would likely further reinforce our findings, which favor SHK and Safety Net at lower and higher levels of reversibility, respectively.

In conclusion, we find that a Safety Net strategy is preferred to SHK for most HT candidates with concurrent kidney failure, excepting only those with a low probability of renal recovery after HT-only. This finding favors implementation of a Safety Net provision along with objective criteria for SHK eligibility; however, doing so hinges upon the ability to better predict renal recovery, an area where further research is needed.

Abbreviations:

DDK

deceased donor kidney

HT

heart transplant

ICER

incremental cost effectiveness ratio

KT

kidney transplant

QALY

quality-adjusted life year

QoL

quality of life

SHK

simultaneous heart-kidney

SLK

simultaneous liver-kidney

SRTR

Scientific Registry of Transplant Recipients

Data availability statement:

Data sharing is not applicable to this article as no new data were created or analyzed in this study.