Association of Serum Uric Acid With Graft Survival After Kidney Transplantation: A Time-Varying Analysis

Authors


Abdolreza Haririan, ahariria@medicine.umaryland.edu

Abstract

The association of serum uric acid (UA) with kidney transplant outcomes is uncertain. We examined the predictive value of UA during the first year posttransplant as a time-varying factor for graft survival after adjustment for time-dependent and independent confounding factors. Four hundred and eighty-eight renal allograft recipients transplanted from January 2004 to June 2006 and followed for 41.1 ± 17.7 months were included. Data on UA, estimated glomerular filtration rate (eGFR), tacrolimus level, mycophenolate mofetil (MMF) and prednisone doses, use of allopurinol, angiotensin-converting enzyme-inhibitor/angiotensin-receptor-blocker (ACEi/ARB) and diuretics at 1, 3, 6, 9 and 12 months were collected. Primary endpoint of the study was graft loss, defined as graft failure and death. Cox proportional hazard models and generalized estimating equations were used for analysis. UA level was associated with eGFR, gender, retransplantation, decease-donor organ, delayed graft function, diuretics, ACEi/ARB and MMF dose. After adjustment for these confounders, UA was independently associated with increased risk of graft loss (HR: 1.15, p = 0.003; 95% CI: 1.05–1.27). Interestingly, UA interacted with eGFR (HR: 0.996, p < 0.05; 95% CI: 0.993–0.999 for interaction term). Here, we report a significant association between serum UA during first year posttransplant and graft loss, after adjustment for corresponding values of time-varying variables including eGFR, immunosuppressive drug regimen and other confounding factors. Its negative impact seems to be worse with lower eGFR.

Abbreviations: 
ACEi/ARB

ACE inhibitor/angiotensin receptor blocker

ACR

acute cellular rejection

AA

African American

ATG

thymoglobulin

BSX

basiliximab

CV

cardiovascular

DD

deceased-donor

DGF

delayed graft function

eGFR

estimated glomerular filtration rate

GEE

generalized estimating equations

MMF

mycophenolate mofetil

POD

postoperative day

SCr

serum creatinine

TAC

tacrolimus

UA

uric acid

Introduction

Despite the impressive improvement in short-term renal transplant outcomes with the advent of newer immunosuppressive agents, the long-term graft and patient survival has not significantly improved. Efforts to develop effective therapeutic measures to improve long-term outcomes require recognition of novel risk factors, which directly or indirectly impact the results of transplantation.

There has been increasing experimental and epidemiological data that suggest uric acid (UA) plays a role in progression of cardiovascular (CV) and renal disease. UA impairs endothelial cell function (1–3), stimulates vascular smooth muscle cell proliferation and profibrotic and inflammatory cytokines (4,5), promotes T-cell activation through macrophage/monocyte stimulation (6) and is associated with increase in inflammatory markers (7). UA has been associated with the genesis of hypertension (8) and leads to an upregulation of the renin–angiotensin system (9).

Increased serum UA levels has been associated with coronary artery calcification and carotid intimal thickening (10,11), and is independently associated with myocardial infarction, ischemic stroke, CV events and all-cause and CV mortality (12–16).

Elevated UA level has been shown to be predictive of incident kidney disease and progression of disease to ESRD (17–23). In nonproteinuric patients with type 1 diabetes, high-normal serum UA is associated with decreased renal function (24). Moreover, reduction of serum UA with allopurinol has been associated with slowing of the progression of renal disease, and reduction in blood pressure (24,25).

In experimental models, mild hyperuricemia induces glomerular hypertension and blood pressure-independent small vessel disease, and accelerates progression of renal disease (9,26–28). In cyclosporine-treated rats, hyperuricemia causes worsening of arteriolar hyalinosis, tubular injury and interstitial fibrosis, and is associated with increased interstitial and glomerular macrophage accumulation (27).

Kidney transplant recipients, particularly those with impaired glomerular filtration rate (GFR), often have higher levels of UA. Others and we have hypothesized that UA level is associated with kidney transplant outcomes. There are only a limited number of studies that have examined the impact of UA on various outcomes after transplantation. These studies, including our previous report, suffer from a major shortcoming that weakens the strength of the findings. They all examine a single value or the average value of UA as the predictor variable. Since UA level depends on a large number of factors and varies over time, using a single value would not provide a strong evidence of association. Moreover, the majority of the reported observations were not adjusted for important variables including estimated (e) GFR or immunosuppressive drugs. Therefore, we sought to evaluate the independent association of serum UA level, measured as multiple time points, as a time-varying factor during the first year after kidney transplantation with long-term graft survival. The results were adjusted for graft function, as well as the immunosuppressive and other medications, all as time-varying covariates. The patients were transplanted during a 30-month span and received a uniform drug regimen posttransplant.

Methods

This is a retrospective cohort study of kidney transplant outcome in adult renal allograft recipients who were transplanted at our center from January 2004 through June 2006 and had recoverable study-related posttransplant data. Recipients of combined organs or with positive flow cytometry cross-match were excluded. All patients received induction therapy with thymoglobulin (ATG) or basiliximab (BSX). ATG was dosed at 1.5 mg/kg/day for 4–7 days starting intraoperatively, with dose adjustment as needed to maintain CD3 lymphocyte counts <50 cells/mm3 and for leukopenia and/or thrombocytopenia. BSX was administered as two doses of 20 mg, on the day of surgery and postoperative day (POD) 4. Patients received methylprednisolone 500 mg IV intraoperatively, followed by 250 and 125 mg on subsequent days, with tapering to oral prednisone at 1 mg/kg by 1 week. Prednisone was further tapered off over the following 2 weeks. For patients with high panel-reactive antibody (PRA), previous transplants and those with early acute rejection prednisone was tapered to 10 mg/day at 3 months and 5 mg/day at 6–12 months. All patients were started on mycophenolate mofetil (MMF) 2 g/day postoperatively, with subsequent dose adjustment for gastrointestinal side effects or leukopenia. Tacrolimus (TAC) was started on POD 1–2 with immediate and 4–7 with delayed graft function (DGF), with target 12-h trough level 8–9 ng/mL during first year and 6–8 ng/mL afterwards. All patients received prophylaxis with trimethoprim/sulfamethoxazole, clotrimazole and valganciclovir.

Follow-up

After hospital discharge, patients were followed according to the recommended guidelines (29). Patients with DGF, defined as the need for hemodialysis in the first week posttransplantation, underwent surveillance biopsies on POD 7–10 and weekly thereafter, while dialysis dependant. Unexplained increase in serum creatinine (SCr) was investigated by graft biopsy as indicated. During the study period, all recipients were screened for polyoma virus by urine cytology and graft biopsy, as needed (30). All episodes of acute cellular rejection (ACR) were confirmed by biopsy and graded according to Banff 97 criteria (31). Grade I rejection episodes were treated with methylprednisolone 500 mg IV for 3 days. Steroid-resistant grade I and grade II rejections were treated with 5–7 daily doses of ATG.

Outcomes

The primary outcome of interest for this study was graft loss, defined as return to dialysis retransplantation, or death with a functioning graft. The secondary endpoint was death-censored graft loss. Recovered data included recipient age, race, gender, pretransplant diabetes, hepatitis C virus serostatus, history of previous renal transplantation, peak (P)-PRA, donor source, donor/recipient HLA-mismatch (MM), initial graft function and SCr at 1, 3, 6, 9 and 12 months posttransplant. Data were collected on serum UA and TAC blood levels, use of allopurinol, ACE inhibitor and/or angiotensin receptor blocker (ACEi/ARB), and diuretics at 1, 3, 6, 9 and 12 months and MMF and prednisone doses at 1, 3, 6 and 12 months after transplantation. All biopsies performed during the study period were reviewed and episodes of acute rejection were identified. Because we did not have access to blood pressure readings and weights in a large fraction of patients in the later time points, we did not include them in our analysis.

Statistical methods

Multiple imputation method was used to estimate the missing values for UA, TAC level and MMF or prednisone dose (32). To estimate the missing values, regression models using corresponding values at the previous time point, age, gender, race and retransplantation, donor source, induction agent, DGF, HLA-MM and P-PRA as the predictor variables were used and the average of 10 imputations were used for analyses. The UA values were missing in 21.9–29.8% of the patients at different time points. For the other variables, values were missing in generally less than 10% of the cases at the study time points.

Data are reported as mean ± SD. Comparison between the variables was performed using Student's t-test, as appropriate. The Cox Proportional Hazard Model was used to evaluate the independent association of UA level measure at multiple time points, as a time-varying predictor variable, with outcome after adjustment for other time-dependent and independent potential confounding factors. We also examined the possibility of interaction between UA and eGFR, to examine if the association of UA with the outcome is affected by the level of eGFR. Therefore, we included the product of UA and eGFR as a separate covariate in the model. Because the value of zero was meaningless for either variable, to make the interpretation of the results meaningful, we replaced the observed values with the values minus 5th percentile value (3.63 and 24.24, respectively). Generalized estimating equations (GEE), using independent correlation with robust standard errors, were used to identify factors that are associated with UA level (33).

To estimate GFR, CKD-EPI formulas were used using SCr levels at the described time points (34). The distribution of continuous variables was examined by distributional and normal probability plots. Variables with nonGaussian distribution were categorized. Regression model assumptions were examined with proper diagnostics.

Intercooled Stata 11.1 software package, (Stata Corporation, College Station, TX, USA) was used for statistical analysis.

Results

There were 488 patients eligible for this study. Patient and transplanted related characteristics are summarized in Table 1. The mean age of the cases was 52.6 years. Near two thirds were men and 45% were African American (AA). Close to two-thirds of the organs were from deceased donors (DD). Sixty-one percent of the patients received BSX and 39% received ATG induction therapy. Nearly 28% of the recipients experienced DGF and 18% had at least one episode of ACR during the follow-up.

Table 1.  Patient and transplant-related characteristics
Age (years)52.6 ± 13.1Retransplant (%)14.7
Male (%)63.8Deceased donor (%)66.7
African American (%)44.8HLA MM4.5 ± 1.6
Diabetes (%)45.2Induction (%) 
 rATG39.1
 BSX60.9
Hepatitis C + (%) 9.2Delayed graft function (%)27.6
Peak PRA (%) Acute cellular rejection (%)18.4
 <1071.4  
 10–4010.4  
 40–80 8.8  
 >80 9.4  

The time-varying values of UA, eGFR, TAC blood levels, MMF and prednisone doses, use of allopurinol, ACEi/ARB and diuretics at different time points after transplantation are summarized in Table 2. As depicted in Figure 1, UA levels had an upward trend, increasing from 5.5 ± 1.7 mg/dL at 1 month to 6.7 ± 2.2 mg/dL at 6 months (p < 0.0001), along with small increase in eGFR during the first few months posttransplant. There was a downward trend in the immunosuppressive agent levels/doses during the first year. Less than 10% of the patients were on allopurinol, 20–30% were taking diuretics, and increasing number of cases were treated with ACEi/ARB as months passed.

Table 2.  Values of the time-dependent variables during the first year posttransplant
 1 month3 months6 months9 months12 months
Uric acid (mg/dL)5.5 ± 1.76.2 ± 1.86.7 ± 2.26.5 ± 1.76.6 ± 1.7
Hyperuricemia (%)13.530.739.539.942.0
eGFR (mL/min)53.1 ± 21.658.0 ± 19.256.4 ± 19.358.3 ± 21.757.2 ± 17.6
Tacrolimus (ng/mL)8.9 ± 4.67.9 ± 4.86.3 ± 4.37.0 ± 2.45.6 ± 3.7
MMF (mg/day)1546 ± 6401388 ± 7001157 ± 7331036 ± 724
Prednisone (mg/day)10.0 ± 14.85.5 ± 7.33.6 ± 4.52.4 ± 4.1
Allopurinol (%)4.18.27.57.47.9
ACEi/ARB (%)5.913.518.826.227.7
Diuretic (%)20.522.625.126.027.7
Figure 1.

Time course of mean uric acid and estimated GFR, and the corresponding 95% CI of the means.

We examined the association of UA as a time-varying element with the time-dependent and independent variables listed in the tables. Male gender was associated with higher UA levels (β= 1.09, p < 0.001, 95% CI: 0.85–1.34), as was DD grafts (β= 1.07, p < 0.001, 95% CI: 0.83–1.30). Previous history of transplantation was associated with higher UA (β= 0.35, p < 0.03, 95% CI: 0.04–0.66). Those who received ATG induction had on average 0.33 mg/dL lower UA level than BSX recipients (p = 0.017, 95% CI: 0.06–0.59), and patients who experienced DGF had higher UA levels (β= 1.03, p < 0.001, 95% CI: 0.72–1.33). Higher eGFR was predictive of lower UA (β=–0.04, p < 0.001, 95% CI: –0.049–0.037). Patients treated with diuretics had on average 1.14 mg/dL higher UA (p < 0.001, 95% CI: 0.84–1.43), and those taking ACEi/ARB overall had a mean 0.40 mg/dL higher levels (p = 0.007, 95% CI: 0.11–0.68). MMF dose < 1000 mg/day was associated with higher UA, as compared with >1500 mg/day (β= 0.46, p = 0.001, 95% CI: 0.19–0.73). There was no significant association with age, race, DM, HCV status, P-PRA, HLA MM, TAC level, prednisone dose or allopurinol use. To determine the independent predictors of UA value, we included all the variables with significant associations in univariate analyses in the final GEE regression model. All of these factors except induction agent were associated with UA level after adjustment for the other covariates, including eGFR (β=–0.04, p < 0.001, 95% CI: –0.46–0.33), male gender (β= 0.78, p < 0.001, 95% CI: 0.58–0.97), retransplantation (β= 0.42, p = 0.001, 95% CI: 0.17–0.68), DD organ (β= 0.33, p = 0.002, 95% CI: 0.12–0.54), DGF (β= 0.29, p = 0.02, 95% CI: 0.04–0.53), diuretics (β= 0.65, p < 0.001, 95% CI: 0.40–0.90), ACEi/ARB (β= 0.49, p < 0.001, 95% CI: 0.25–0.73) and MMF dose (β= 0.29, p = 0.007, 95% CI: 0.08–0.50 for <1000 mg/day compared with >1500 mg/dL). The regression coefficient (β) represents the change in UA for each one-unit increase in eGFR as a continuous predictor variable, and presence versus lack of the dichotomous variables.

During the 41.1 ± 17.7 months follow-up period, there were 115 incidents of graft loss defined as death or graft failure, i.e., return to dialysis or retransplantation. We examined the association of UA and other variables listed in the tables with risk of graft loss. The factors that were individually associated with this outcome included UA (HR: 1.30, p < 0.001; 95% CI: 1.204–1.40), eGFR (HR: 0.96, p < 0.001; 95% CI: 0.94–0.97), AA race (HR: 1.63; p = 0.009; 95% CI: 1.13–2.36), DD organ (HR: 2.49; p < 0.001; 95% CI: 1.57–3.97), P-PRA (HR: 1.65; p = 0.02; 95% CI: 1.07–2.52 for P-PRA > 80%), HLA-MM (HR: 1.17; p < 0.02; 95% CI: 1.03–1.34), DGF (HR: 2.70; p < 0.001; 95% CI: 1.87–3.89), ACR (HR: 2.18; p < 0.001; 95% CI: 1.47–3.23), TAC level (HR: 0.90; p < 0.001; 95% CI: 0.85–0.95), MMF dose (HR: 2.29; p < 0.001; 95% CI: 1.57–3.32 for MMF < 1000 mg/day) and ACEi/ARB (HR: 0.51; p = 0.009; 95% CI: 0.31–0.85). We then examined the possibility of interaction between UA and eGFR. By including UA, eGFR and the interaction term in the model, we observed a significant interaction between the two (HR: 1.15, p = 0.002; 95% CI: 1.05, 1.25 for UA; HR: 0.97, p < 0.001; 95% CI: 0.95–0.99 for eGFR and HR: 0.996, p = 0.03; 95% CI: 0.993–0.999 for interaction term). This suggests that the hazard of graft loss for increasing UA levels decreases at higher GFR values. When we examined the other time-dependent and independent variables, there was no significant interaction between UA level and these variables (data not shown). To evaluate the independent association of UA with graft loss, we included all the variables listed earlier including interaction term between UA and eGFR in the final model. After adjustment for the confounders, UA was independently associated with increased risk of graft loss (HR: 1.15; p = 0.003; 95% CI: 1.05–1.27) and interacted with eGFR (HR: 0.996; p < 0.05; 95% CI: 0.993–0.999 for interaction term; Figure 2). For sensitivity analysis, we limited the analysis to 179 patients with complete nonimputed UA data and observed similar association with graft failure (HR: 1.13); however, as would be expected, this association was not significant.

Figure 2.

Change in the adjusted hazard ratio of graft loss for each 1 mg/dL increase in uric acid level for 5 mm/min increments of eGFR as a result of interaction between the two variables.

To corroborate the observed association using UA as a continuous variable, we categorized UA into sex-specific quartiles (cutoff values 4.4, 5.4, 6.3 for women and 5.3, 5.4 and 6.4 for men). Including the UA quartiles along with other covariates listed above and interaction term with eGFR in the model, we observed an independent association between UA quartiles and graft survival (HR: 1.87; p = 0.25; 95% CI: 0.64–5.41 for quartile 2, HR: 3.5; p < 0.06; 95% CI: 0.96–12.8 for quartile 3 and HR: 4.1; p < 0.05; 95% CI: 1.0–17.5 for quartile 4, compared with quartile 1; Figure 3). In this model as well, there was a significant interaction between eGFR and UA quartiles (HR: 0.985; p < 0.02; 95% CI: 0.974–0.998).

Figure 3.

Unadjusted graft survival by UA quartiles.

Among the cases with graft loss, 69 experienced death-censored graft loss. We examined the association of the UA and other variables listed in the tables with risk of graft failure. The factors that were associated with graft failure included UA (HR: 1.36; p < 0.001; 95% CI: 1.24–1.48), eGFR (HR: 0.94; p < 0.001; 95% CI: 0.93–0.96), AA race, diabetes, DD organ, P-PRA, HLA-MM, DGF, ACR, TAC level, MMF dose, prednisone dose and use of ACEi/ARB (data not shown). Even after adjustment for eGFR, higher UA level was predictive of higher risk of graft failure (HR: 1.16; p = 0.04; 95% CI: 1.00–1.33) and there was no significant interaction between the two. However, after adjusting for all the listed variables that were associated with this outcome, UA was not significantly associated with increased risk of graft failure (HR: 1.14; p = 0.16; 95% CI: 0.95–1.36).

Discussion

In this study, we examined the association of serum UA values during the first year posttransplant as a time-varying predictor with medium to long-term graft survival after adjustment for major time-varying and fixed confounding factors in a relatively uniformly treated large cohort of kidney transplant recipients. Furthermore, we evaluated the association of UA level with a large number of patient and transplant-related variables.

We observed a significant association between UA value and graft survival, after adjustment for potential confounding variables including eGFR. Notably a 15% increase in the risk of graft loss during the study period was observed for each 1 mg/dL increase in UA when eGFR is around 25 mL/min. An interesting observation in this analysis was the evidence of interaction between UA level and eGFR, suggesting that the adverse impact of UA on graft outcome decreases with higher levels of graft function. Therefore, one could postulate that UA reduction should be more aggressive with lower GFR. Categorizing the UA levels into sex-specific quartiles, quartiles 3 and 4 had significantly higher risk of graft loss compared to lowest quartile (HR: 3.5 and 4.1, respectively). When we evaluated graft failure as the study endpoint, similar association was observed, however, this was no longer statistically significant. We believe that this is because of loss of study power to detect the lower observed number of graft failures. Considering the role of UA in CV disease and mortality (10–16), this also could be because of the independent association of UA with mortality in this patient population.

There are a limited number of reports in the literature that have studied the impact of UA on graft survival or graft function, with mixed results. Gores et al. (35) studied hyperuricemia in 262 patients enrolled in a prospective, randomized trial of cyclosporine. Among patients in the cyclosporine group, severely hyperuricemic cases had a mean SCr similar to those with normal UA, and their 4-year graft survival rate was 90%, suggesting that UA level was not associated with graft function or survival. Akgul et al. (36) conducted a retrospective study of 133 patients with at least 6 months follow-up. The investigators did not find any difference in the rate of CAN at the end of each of the 3 years of follow-up between hyperuricemic and normouricemic recipients. More recently, in a retrospective study Meier-Kriesche et al. (37) investigated the association between UA level at 1 month and eGFR at 3 years posttransplant in 852 patients who were enrolled in the Symphony study. When corrected for baseline renal function, 1-month UA was not independently associated with 3-year renal allograft function. It should be noted that the outcome of this study was graft function, different from graft survival presented here. Although the authors have shown that UA levels increase over the course of first year after transplantation, they have evaluated only UA level at one early time point. Moreover, different immunosuppressive agents and doses were used and no adjustment was made for drug levels.

On the contrary, Gerhardt et al. (38) reported an association between hyperuricemia and decreased graft survival during 5 years of follow-up compared with normouricemia (68.8% vs. 83.3%, respectively) among 350 kidney transplant recipients with functioning grafts at 1 year posttransplant. In addition, patients with hyperuricemia had higher SCr levels. Armstrong et al. (39) studied the association of a single UA level measured at a variable period after transplantation (median 7.1 years) with eGFR after 2 years of follow-up in 90 kidney transplant recipients. The investigators found an inverse correlation between baseline UA and eGFR at follow-up (β: –22.2; p = 0.02) after adjusting for baseline eGFR. Moreover, it was the only independent factor associated with requirement for three or more antihypertensive agents. Akalin et al. (40) studied 307 renal allograft recipients with a functioning graft at 6 months. They observed an association between hyperuricemia and composite endpoint including death, graft failure, new CV events, and biopsy-proven CAN over a 4.3-years mean follow-up. After adjusting for a number of variables including eGFR, UA level (HR: 1.12; p = 0.053) and hyperuricemia (HR: 1.69; p = 0.047) were associated with this endpoint. When the investigators examined the interaction with eGFR, hyperuricemia was associated with the composite endpoint only in those with eGFR less than 50 mL/min/1.7 m2. In our previous study, we examined the association of the average of UA during the first 6 months posttransplant and graft survival during 68 months (mean) follow-up in 212 living donor kidney transplant recipients (41). UA level, as a continuous variable, and hyperuricemia, a dichotomous variable, were associated with graft loss (HR: 1.26; p = 0.026 and HR: 1.92; p = 0.029, respectively) and 1-year eGFR independent of mean 6-month eGFR and other confounders.

All studies summarized above suffer from different important weaknesses, which include the use of single UA value, variable time of testing, lack of adjustment for a number of confounders including graft function and relatively small sample size. The current study has removed many of these pitfalls and provides stronger evidence for the association between UA and kidney transplant outcomes.

The major distinguishing aspect of the current analysis from the previously published reports is the use of UA levels at multiple time points as a time-varying factor and adjustment for corresponding eGFR. Moreover, we also adjusted for the important confounding effect of immunosuppressive regimen, including TAC level and MMF and prednisone doses during the first year. In addition, we have adjusted for the use of important drugs that can influence serum UA levels and influence graft survival, including allopurinol, diuretics and the use of rennin–angiotensin system blocking drugs. An interesting finding in our study was the observation of interaction between UA and eGFR in predicting graft survival, which corroborates the crude observation in the study by Akalin et al. (40).

In this cohort, the factors that were independently predictive of UA level included eGFR, male gender, retransplantation, transplantation with a DD organ, DGF, use of diuretics and ACEi/ARB and MMF dose. Although losartan, an ARB, causes uricosuria and is expected to reduce serum UA (42), we observed an increase with the combined group of all ACEi and ARB drugs. Armstrong et al. (39) in a more limited analysis had reported the association of UA level with male gender, history of hypertension, use of cyclosporine, prednisolone dose and eGFR.

The findings of our study, albeit more robust compared with prior studies, should be interpreted cautiously. Because of its retrospective design, residual confounding cannot be excluded. Moreover, we limited the measures of UA and other time-dependent variables to the first year posttransplant as the available data were less consistent afterwards. Extending the follow-up period would also provide observation of a larger number of events to provide sufficient power for examination of graft failure as the single endpoint. Moreover, we did not include the blood pressure and weight values during the follow-up period in our analysis. Despite these limitations, this study has notable strengths and unique features as detailed above.

In summary, in this retrospective cohort study we observed a significant, association between serum UA level during first year posttransplant and graft loss, including graft failure and death with functioning graft, after adjustment for corresponding values of eGFR, immunosuppressive drug regimen, and other confounding factors. A unique feature of this study is the use of UA and a number of confounding variables at different time points as time-varying factors. This finding raises the important question whether specific UA lowering treatment would be beneficial in improving long-term outcomes after kidney transplantation. Further studies are needed to examine the causal relationship between UA and these outcomes.

Disclosure

The authors of this manuscript have no conflict of interest to disclose as described by the American Journal of Transplantation.

Ancillary