Early steroid withdrawal and kidney transplant outcomes in first-transplant and retransplant recipients

ABSTRACT

Background

Early steroid withdrawal (ESW) is often preferred over conventional steroid maintenance (CSM) therapy for kidney transplant recipients with low immunological risks because it may minimize immunosuppression-related adverse events while achieving similar transplant outcomes. However, the risk–benefit balance of ESW could be less favorable in retransplant recipients given their unique immunological risk profile. We hypothesized that the association of ESW with transplant outcomes would differ between first-transplant and retransplant recipients.

Methods

To assess whether the impact of ESW differs between first and retransplant recipients, we studied 210 086 adult deceased-donor kidney transplant recipients using the Scientific Registry of Transplant Recipients. Recipients who discontinued maintenance steroids before discharge from transplant admission were classified with ESW; all others were classified with CSM. We quantified the association of ESW (vs CSM) with acute rejection, death-censored graft failure and death, addressing retransplant as an effect modifier, using logistic/Cox regression with inverse probability weights to control for confounders.

Results

In our cohort, 26 248 (12%) were retransplant recipients. ESW was used in 30% of first-transplant and 20% of retransplant recipients. Among first-transplant recipients, ESW was associated with no significant difference in acute rejection {adjusted odds ratio (aOR) = 1.04 [95% confidence interval (CI) = 1.00–1.09]}, slightly higher hazard of graft failure [hazard ratio (HR) = 1.09 (95% CI = 1.05–1.12)] and slightly lower mortality [HR = 0.93 (95% CI = 0.91–0.95)] compared with CSM. Nonetheless, among retransplant recipients, ESW was associated with notably higher risk of acute rejection [OR = 1.42 (95% CI = 1.29–1.57); interaction P < .001] and graft failure [HR = 1.24 (95% CI = 1.14–1.34); interaction P = .003], and similar mortality [HR = 1.01 (95% CI = 0.94–1.08); interaction P = .04].

Conclusions

In retransplant recipients, the negative impacts of ESW on transplant outcomes appear to be non-negligible. A more conservatively tailored approach to ESW might be necessary for retransplant recipients.

Graphical Abstract

Graphical Abstract.

KEY LEARNING POINTS.

What was known:

• Early steroid withdrawal (ESW) after kidney transplantation reduces in immunosuppression-related side effects while achieving similar transplant outcomes in selected recipients.

• Retransplant recipients have previous long-term exposure to immunosuppression, but also show higher risk of rejection and graft failure.

• The risk–benefit balance of ESW in retransplant recipients remains unclear.

This study adds:

• In first-transplant recipients, ESW was associated with no significant difference in rejection and 9% higher hazard of graft failure.

• In retransplant recipients, ESW was associated with 42% higher odds of rejection and 24% higher hazard of graft failure, even among subgroups with low immunological risk (0 HLA mismatches or PRA = 0).

Potential impact:

• Retransplant status may adversely alter the risk–benefit balance of ESW by amplifying ESW's impact on rejection and graft failure.

• Our findings support a more conservatively tailored approach to ESW for retransplant recipients.

INTRODUCTION

Early steroid withdrawal (ESW) is an immunosuppression minimization approach that is generally recommended for kidney transplant recipients with low immunological risk [1–7]. In properly selected recipients, ESW is associated with decreases in long-term steroid-related adverse events [8–11], but with only minimal increases in acute rejection and graft failure [2, 11]. These risks and benefits of ESW should be considered on balance for individualization of maintenance immunosuppression [12]. In other words, when the risks of ESW may outweigh the benefits, conventional steroid maintenance (CSM) would be a more risk-appropriate strategy for the individual recipient. Currently, approximately 30% of adult kidney transplant recipients in the USA undergo ESW [13].

Among kidney transplant recipients who experience graft failure, retransplantation is the preferrable treatment option as it offers superior survival and quality of life over returning to chronic dialysis [14–19]. In 2021, 11% of adult kidney transplant recipients had prior history of graft failure [13]. Retransplant recipients show higher risk of acute rejection and graft failure compared with first-transplant recipients [20–23], which might render CSM a more risk-appropriate approach compared with ESW for this subgroup. Indeed, most of the landmark clinical trials that shaped the current ESW practice explicitly listed retransplant as one of the exclusion criteria [3, 24–28]. On the other hand, retransplant recipients could be more prone to immunosuppression-related adverse events, including cardiovascular events, osteoporosis and new-onset diabetes after transplantation [8–11], as they often have higher prevalence of comorbidities and previous long-term exposure to immunosuppression [16, 29]. Therefore, reducing steroid-related adverse events can be particularly important in retransplant recipients considering the high comorbidity burden and previous long-term exposure to immunosuppression in this subgroup.

Due to these competing issues, the clinical implication of retransplantation on the risk–benefit balance of ESW remains unclear. We posit that ESW could be a feasible approach in carefully selected retransplant recipients, provided that it results in only minimal increases in acute rejection and graft failure in retransplant recipients, as it does in the overall kidney transplant recipient population [1–5]. To test this hypothesis, we conducted a national cohort study to quantify the association of ESW (versus CSM) with post-transplant outcomes and to compare the strength of this association by retransplant status.

MATERIALS AND METHODS

Data source

This study used data from the Scientific Registry of Transplant Recipients (SRTR). The SRTR data system includes data on all donors, waitlisted candidates and transplant recipients in the USA, submitted by the members of the Organ Procurement and Transplantation Network (OPTN). The Health Resources and Services Administration (HRSA), US Department of Health and Human Services, provides oversight to the activities of the OPTN and SRTR contractors. A detailed description of the data has been provided elsewhere [30]. This manuscript followed the STrengthening the Reporting of OBservational Studies in Epidemiology (STROBE) guidelines for cohort studies [31].

Study population

We studied all adult (18 years or older) deceased-donor kidney-only transplant recipients from 1 January 2005 to 31 December 2022. After excluding 2735 (1%) recipients whose discharge immunosuppression records were unavailable, 210 086 recipients were included in the study population. Recipients with previous kidney transplants as reported on the OPTN Adult Kidney Transplant Recipient Registration Worksheet were labeled as retransplant recipients.

Early steroid withdrawal

ESW was defined as withdrawal of maintenance steroids prior to the discharge from the transplant admission. All other recipients were considered CSM. We used the inverse probability of treatment weights (IPTW) to control for confounding, especially confounding by indication, in the transplant outcomes analyses [32]. We used the extreme gradient boosting, a general-purpose machine learning algorithm, to characterize the treatment assignment mechanism. The use of ESW (versus CSM) was the outcome variable of this model. Covariables included were donor factors (age, race, sex, ABO blood type, height, weight, body mass index, cause of death, terminal serum creatinine, cytomegalovirus, hepatitis C virus, diabetes, hypertension, donation after cardiac death, cardiac arrest, machine perfusion, history of cigarette smoking and history of cocaine use), recipient factors (age, race, sex, primary cause of kidney failure, ABO blood type, height, weight, body mass index, HIV, cytomegalovirus, hepatitis B virus, hepatitis C virus, Epstein–Barr virus, time on dialysis, calculated panel reactive antibody, diabetes, hypertension, total serum albumin, education level, functional status, malignancy, peripheral vascular disease, angina or coronary artery disease and previous pregnancy), induction immunosuppression, HLA-A/B/DR mismatches and cold ischemic time.

We used this model to calculate the propensity score for undergoing ESW within each patient. We examined the distribution of the IPTW values and did not observe extreme weights. We assessed covariable imbalance before and after IPTW across all covariables using the standardized mean difference [33]. Additionally, we measured the variable importance of each covariable using the mean of the absolute value of the Shapley additive explanation value for the covariable [34].

Acute rejection

Acute rejection was ascertained using the Adult Kidney Transplant Recipient Follow-Up Worksheet. This worksheet is submitted by transplant centers to the OPTN to report clinical episodes during each post-transplant periods (0–6 months, 7–12 months and annual thereafter); however, the exact dates of the acute rejection episodes are not reported. Therefore, we treated acute rejection during the first year after transplant as a binary outcome, in accordance with the previous studies of the national registry data [30, 35]. We used logistic regression to compare the odds of acute rejection between the ESW and CSM groups, after adjusting for confounders using the IPTW as described above. These models included an interaction term between ESW and retransplantation to measure the effect measure modification on ESW by retransplantation.

Graft and patient survival

Graft and patient survival were treated as survival outcomes as the exact date of the events were available in our dataset. Graft survival was defined as the time from kidney transplant to graft failure, censoring for death or the end of follow-up (31 December 2022). Patient survival was defined as the time from kidney transplant to death, censoring for the end of follow-up. Deaths were collected by OPTN and supplemented by SRTR using multiple sources, including follow-up reports from transplant centers, Centers for Medicare & Medicaid Services ESRD Death Notification Form (CMS 2746) and Social Security Death Master File. We used Cox regression to compare the hazard of graft failure and mortality between the ESW and CSM groups. These models were also adjusted for confounders using IPTW and accounted for effect measure modifications using interactions terms between ESW and retransplantation.

Statistical analysis

We used a stepwise modeling approach for mechanistic understanding of the interaction between ESW and retransplantation. Model 1 measured the association of ESW with post-transplant outcomes in the entire cohort irrespective of retransplant status and therefore included no interaction terms. Model 2 included interactions terms between ESW and retransplant status to measure the association of ESW with post-transplant outcomes separately in the first-transplant and retransplant recipients. We used the Wald test on the interaction term to determine the statistical significance of the effect measure modifications. A significant effect measure modification suggests that the association of ESW with post-transplant outcome differs between the first-transplant and retransplant subgroups. We used the estimates from Model 2 as our primary outcome measurements.

To explore the potential mechanism of the observed effect measure modification, we expanded Model 2 by further stratifying the cohort. First, we created Model 3 which added further stratification by panel reactive antibody (PRA). Since a high PRA value indicates immune sensitization, it is a potential mediator between retransplantation and worse post-transplant outcomes [18, 36], and an effect measure modifier that may indicate CSM over ESW [37]. Therefore, among retransplant recipients with sufficiently low PRA values (e.g. PRA = 0), ESW could still be associated with acceptable post-transplant outcomes. PRA was calculated centrally by OPTN across all antigens (classes I and II). Recipients with missing PRA data (n = 64) were excluded from this model.

Second, Model 4A/4B/4DR added stratification by the number of HLA-A/B/DR mismatches, respectively. We aimed to explore if retransplant recipients with 0 HLA mismatches respond to ESW differently when compared with retransplant recipients with 1 or 2 mismatches.

Third, Model 5 added further stratification by recipient age. First-transplant recipients tended to be older than retransplant recipients (median age 56 vs 47 years; Supplementary data, Table S1), corroborating the previous reports of limited access to retransplantation in older adults [18]. To examine whether this age disparity explains the effect modification, we stratified the cohort by recipient age of 50 years (18–50 vs >50) such that the younger and older strata are similar in sample size.

Lastly, Model 6 added further stratification by the cause of end-stage kidney disease (ESKD). The causes were categorized into immunoglobulin A nephropathy, other glomerulonephritis, diabetes, hypertension and others.

For Models 3–6, we stratified the cohort by retransplant status and by the additional stratifying variable and included interaction terms with ESW. We used the Wald test to examine whether the association of ESW with post-transplant outcome was significantly different between the first-transplant and retransplant subgroups within each stratum.

In addition, we conducted two sensitivity analyses in which we further stratified the retransplant group by the number of previous transplants [19] and excluded recipients with previous non-kidney solid organ transplants. Our main findings were mostly unchanged (Supplementary data, Tables S2 and S3).

Stratified analyses in retransplant recipients

After the main analyses, we conducted additional analyses with further stratifications in the retransplant subgroup to understand how other clinical factors could influence on the suitability of ESW versus CSM in this population. First, we explored whether lymphocyte-depleting induction immunosuppression could attenuate the association of ESW with the increase in acute rejection in retransplant recipients. Lymphocyte-depleting agents included rabbit anti-thymocyte globulin and alemtuzumab. The comparators included basiliximab, daclizumab and no antibody-based induction. This analysis excluded 1689 retransplant recipients who received any other agents or more than one agent for induction immunosuppression. We included an interaction term between ESW and lymphocyte-depleting induction immunosuppression. Additionally, we examined whether the impact of ESW in retransplant recipients is more favorable among those who had undergone ESW for their previous transplant as well. We performed this comparison using an interaction term between ESW for retransplant and ESW for the previous transplant.

RESULTS

Study population

Among the 210 086 recipients included in our study, 59 977 (29%) underwent ESW. Compared with those who underwent CSM, the recipients who underwent ESW were more likely to be white (44.4% vs 38.0%) and less likely to be retransplant recipients (8.7% vs 14.0%). The ESW group had slightly more recipients with 0 PRA (68.5% vs 60.2%) with a lower PRA distribution [median (interquartile range) 0.0% (0.0%–9.5%) vs 0.0% (0.0%–43.5%)]. Other clinical characteristics were similar between the two groups (Table 1).

Table 1:

Population characteristics of 210 086 kidney first-transplant and retransplant recipients by maintenance steroid regimen.

	ESW (n = 59 977)	CSM (n = 150 109)
Donor characteristics
Age, years	41 (27, 52)	40 (27, 51)
Race
Black/African American	13.7	14.1
Hispanic/Latino	12.6	14.8
White	70.6	67.2
Others	3.0	3.8
Female sex	39.3	38.5
ABO blood type
A	37.6	36.7
AB	3.8	3.5
B	11.6	11.9
O	47.1	47.8
Body mass index, kg/m2	26.9 (23.2, 31.5)	26.8 (23.2, 31.4)
Recipient characteristics
Age, years	55 (45, 64)	54 (43, 63)
Race
Black/African American	30.4	34.7
Hispanic/Latino	16.3	18.0
White	44.4	38.0
Others	8.9	9.3
Female sex	37.1	40.8
Cause of kidney failure
Glomerulonephritis	19.3	22.7
Diabetes	29.9	28.0
Hypertension	24.0	23.4
Others	26.8	25.9
Body mass index, kg/m2	28.0 (24.4, 32.1)	27.8 (24.2, 31.9)
Time on dialysis, years	3.4 (1.5, 5.6)	3.6 (1.6, 6.0)
Zero cPRA	68.5	60.2
cPRA, %	0.0 (0.0, 9.5)	0.0 (0.0, 43.5)
Diabetes	37.5	35.5
Hypertension	88.9	87.5
Cold ischemic time, h	17.4 (12.0, 23.5)	17.4 (12.1, 23.0)
Kidney retransplant	8.7	14.0
Immunosuppression
Induction immunosuppression
ATG	51.7	56.0
IL2RA	7.0	19.7
Alemtuzumab	26.5	7.2
Multiple/others	7.2	5.0
None	7.6	12.0
Tacrolimus	92.5	92.7
Cyclosporin	3.2	3.4
Mychophenolates	93.8	96.5
mTORi	3.7	1.7

Numbers shown are median (interquartile range) for categorical variables and % for categorical variables.

cPRA, calculated PRA; ATG, anti-thymocyte globulin; IL2RA, interleukin-2 receptor antagonists; mTORi, mammalian target of rapamycin inhibitors.

After IPTW, the observed differences between the two groups were mostly removed, as indicated by the substantial reduction in standardized mean difference across all variables (Supplementary data, Table S4). White recipients constituted 40% of both ESW and CSM groups, and retransplant recipients did 12% of the ESW group and 13% of the CSM group. PRA showed similar distributions with median (interquartile range) of 0.0% (0.0%–29.5%) in the ESW group and 0.0% (0.0%–32.4%) in the CSM group.

Retransplant and use of early steroid withdrawal

When stratified by retransplant status, 54 745 of 183 838 first-transplant (30%) and 5232 of 26 248 retransplant recipients (20%) underwent ESW. Variable importance analysis suggested that retransplant status ranked eighth among 49 variables supplied to the model. Other variables with high importance included induction agent, PRA and functional status (Table 2).

Table 2:

Ten variables with the highest importance on maintenance steroid use.

Rank	Variable	Importance	Relative value
1	Induction agent	0.452	Reference
2	Calculated PRA	0.175	39%
3	Functional status (Karnofsky Performance Scale)	0.149	33%
4	Recipient race	0.132	29%
5	Cold ischemic time	0.102	23%
6	Recipient hypertension	0.073	16%
7	Recipient HBV core antibody	0.072	16%
8	Retransplant	0.071	16%
9	Machine perfusion	0.071	16%
10	Recipient total serum albumin	0.063	14%

Importance estimates are the mean of the absolute value of the Shapley additive explanation values from an extreme gradient boosting model to characterize the choice of early steroid withdrawal versus conventional steroid maintenance. Relative values are scaled to the highest-ranked variable.

HBV, hepatitis B virus.

Acute rejection, graft failure and patient death

Overall, recipients who underwent ESW showed increased acute rejection and graft failure. These adverse associations were substantially more pronounced among retransplant recipients compared with first-transplant recipients, suggesting an effect measure modification on ESW by retransplant status (Fig. 1).

Figure 1:

Association of ESW versus CSM with post-transplant outcomes by first-transplant versus retransplant. Model 1 includes the entire cohort. Model 2 is stratified by first-transplant versus retransplant. Rejection indicates 12-month acute rejection. Graft failure indicates death-censored graft failure. Estimates are ORs for acute rejection and HRs for graft failure and death. 95% CI are shown in parentheses and as horizontal bars. All estimates are weighted for various confounders using IPTW. KT, kidney transplant.

Over the entire study cohort (Model 1), ESW was associated with slightly higher risk of acute rejection {odds ratio (OR) = 1.10 [95% confidence interval (CI) 1.05–1.14]} and graft failure [hazard ratio (HR) = 1.11 (95% CI 1.07–1.14)], but with lower risk of mortality [HR = 0.94 (95% CI 0.92–0.96)] compared with CSM.

When stratified by retransplant status (Model 2), the associations of ESW with post-transplant outcomes were generally favorable among the first-transplant recipients. ESW was not statistically significantly associated with acute rejection [OR = 1.04 (95% CI 0.998–1.09)]. ESW was associated with slightly higher risk of graft failure [HR = 1.09 (95% CI 1.05–1.12)], but with lower risk of mortality [HR = 0.93 (95% CI 0.91–0.95)]. On the other hand, the associations of ESW with post-transplant outcomes were more adverse among the retransplant recipients. ESW was associated with markedly higher risk of acute rejection [OR = 1.42 (95% CI 1.29–1.57]; P for effect measure modification (P_EMM) <.001] and graft failure [HR = 1.24 (95% CI 1.14–1.34); P_EMM = .003]. ESW was not associated with mortality [HR = 1.01 (95% CI 0.94–1.08); P_EMM = 0.04].

Further stratification by PRA, HLA mismatch and recipient age

In Model 3, ESW was adversely associated with post-transplant outcomes in retransplant recipients even when their PRA was 0 (Fig. 2). Among the retransplant recipients with PRA = 0, ESW was associated with markedly higher risk of acute rejection [OR = 1.47 (95% CI 1.23–1.76)] and graft failure [HR = 1.26 (95% CI 1.11–1.44)]. These adverse associations were absent or subdued among the first-transplant recipients with PRA = 0. Similar effect measure modifications were observed in recipients with PRA > 0.

Figure 2:

Association of ESW versus CSM with post-transplant outcomes by first-transplant versus retransplant, stratified by PRA, HLA-DR mismatch and recipient age. Model 3 adds further stratification by calculated PRA (0 vs >0) to Model 2; 64 recipients with missing PRA were excluded. Model 4DR adds further stratification by the number of HLA-DR mismatches to Model 2; 14 recipients with missing HLA-DR match data were excluded. Model 5 adds further stratification by recipient age (18–50 vs >50 years) to Model 2. Rejection indicates 12-month acute rejection. Graft failure indicates death-censored graft failure. Estimates are ORs for acute rejection and HRs for graft failure and death. 95% CI are shown in parentheses and as horizontal bars. All estimates are weighted for various confounders using IPTW. MM, mismatch; KT, kidney transplant.

In Model 4DR, ESW was associated with higher odds of acute rejection [OR = 1.23 (95% CI 1.03–1.48)] and graft failure [HR = 1.29 (95% CI 1.12–1.48)] in retransplant recipients even when they had 0 HLA-DR mismatches (Fig. 2). Similar trends were observed when stratified by HLA-A or B (Supplementary data, Fig. S1).

In Model 5, the effect measure modification between retransplant and ESW was consistent across recipient age strata (Fig. 2). ESW was associated with significantly higher odds of acute rejection in retransplant recipients of younger [18–50 years; OR = 1.44 (95% CI 1.27–1.62)] and older (>50 years; OR = 1.39 (95% CI 1.18–1.65)] age, but these associations were not observed in first-transplant recipients of the same age strata. Similar effect measure modifications were observed for graft failure and mortality.

In Model 6, the effect measure modification between retransplant and ESW on rejection was consistent across ESKD pathologies. However, the effect measure modification on graft failure was more evident among those with ESKD due to hypertension or other causes (P_EMM = .02 and .057, respectively), and that on death was more evident among those with ESKD due to immunoglobulin A nephropathy or other glomerulonephritis (P_EMM = .04 for both subgroups) (Supplementary data, Fig. S2).

Stratified analyses in retransplant recipients

The association of ESW with transplant outcomes did not differ by induction agent in retransplant recipients (Supplementary data, Table S5a), suggesting that lymphocyte-depleting induction agents might not mitigate the adverse association of ESW with transplant outcomes in retransplant recipients.

The association of ESW with acute rejection was more pronounced among the retransplant recipients who underwent CSM for the previous kidney transplant [OR = 1.58 (95% CI 1.40–1.78)] compared with those who also underwent ESW for the previous kidney transplant [OR = 1.12 (95% CI 0.91–1.37); P_EMM = .004]. On the other hand, ESW was not associated with mortality among the retransplant recipients who underwent CSM for the previous kidney transplant [HR = 0.98 (95% CI 0.90–1.07)], whereas it was significantly associated with increased mortality among those who underwent ESW for the previous kidney transplant [OR = 1.21 (95% CI 1.04–1.21); P_EMM = .02] (Supplementary data, Table S5b).

DISCUSSION

In our national study of 210 086 kidney transplants recipients, the association of ESW with post-transplant outcomes differed significantly between first-transplant and retransplant recipients. Risks associated with ESW were minimal for first-transplant recipients, showing no difference in rejection and only 9% higher hazard of graft failure compared with CSM. However, for retransplant recipients, ESW posed more significant risks, including 42% higher odds of rejection and 24% higher hazard of graft failure. Moreover, ESW was associated with 7% lower hazard of mortality among first-transplant recipients, suggesting ESW's potential benefits, which were absent in retransplant recipients. We explored several low-risk subgroups within retransplant recipients, including 0 HLA mismatches or 0 PRA. ESW was associated with significantly increased risk of rejection and graft failure in all subgroups. These findings suggest that CSM is potentially a more risk-appropriate strategy over ESW for most retransplant recipients.

Retransplant has been widely viewed as a risk factor for suboptimal post-transplant outcomes, mainly increased acute rejection and graft failure [20–23]. These findings may have influenced clinical practice favoring CSM over ESW for retransplant recipients as a part of immunosuppression tailoring, although little clinical evidence had supported this practice [38]. Correspondingly, we observed that fewer retransplant recipients underwent ESW compared with first-time recipients (20% vs 30%) and that retransplant status was the 8th important variable out of 49 that influence the selection between ESW and CSM. The effect measure modifications observed in our study provide quantitative evidence supporting these clinical practices.

In our stratified analyses, ESW did not appear to be a preferrable approach even in lower-risk subgroups of retransplant recipients, such as those with 0 PRA, 0 HLA mismatches or those who underwent lymphocyte-depleting induction immunosuppression. Although low PRA is a possible indication to ESW in the overall kidney transplant recipient population [37], retransplant recipients with 0 PRA still showed an adverse association of ESW with acute rejection (OR = 1.47), suggesting that ESW might not be an optimal approach in retransplant recipients regardless of their PRA (Fig. 1). In addition, unlike first-transplant recipients where there is evidence to support ESW especially when combined with lymphocyte-depleting induction immunosuppression [1, 3, 39], we observed that the association of ESW with post-transplant outcomes did not differ by induction agent in retransplant recipients (Supplementary data, Table S5a).

Our study has several limitations. First, due to the observational design of our study, confounding by individual-level clinical factors or by center-level protocol cannot be completely ruled out. To minimize this risk, we used causal inference methods, mainly IPTW, leveraging a machine learning algorithm to create a propensity score model based on a wide array of clinical factors. Moreover, our primary finding regarding the effect measure modification would be biased only when the impact of confounding differed between first-transplant and retransplant recipients. Second, since we defined ESW as withdrawal of maintenance steroids by discharge, the recipients who withdrew or restarted maintenance steroids after discharge could have been misclassified. However, such regimen changes are relatively uncommon [40], and our definition corresponds to the intention-to-treat approach, which typically results in a conservatively biased estimation of the association. Third, our dataset did not include clinical information that may represent the immunological risk more specifically, such as donor-specific antibody or pre-transplant desensitization. If those who showed lower risk in these unmeasured factors were more likely to undergo ESW, the association of ESW with increased acute rejection and graft failure observed in retransplant recipients might become slightly stronger after adjusting for these factors. Fourth, the scope of our study did not cover the entire range of the risk–benefit balance of ESW. Our study focused on three key transplant outcomes, rejection, graft failure and death, whereas the benefits of ESW can manifest as risk reduction in non-transplant-related outcomes such as diabetes, cardiovascular disease or bone mineral disorder [8–11]. Lastly, our study used US data and focused on deceased-donor recipients for homogeneity. It remains unclear whether our findings would be generalizable to populations outside the USA or to living-donor recipients.

In summary, ESW was associated with 1.42-fold odds of acute rejection, 1.24-fold hazard of graft failure and no difference in mortality among retransplant recipients. This risk–benefit balance significantly differed from that in first-transplant recipients, among whom ESW was associated with minimal increases in acute rejection and graft failure, and with a statistically significant decrease in mortality. ESW was strongly associated with acute rejection and graft failure even among lower-risk subgroups of retransplant recipients, such as those with 0 PRA, 0 HLA mismatches or lymphocyte-depleting induction immunosuppression. These effect measure modifications suggest that CSM might be a more favorable maintenance regimen over ESW for most retransplant recipients.

FUNDING

This work was supported by K01DK132490 (PI: S.B.) and R01DK120518 (PI: M.A.M.D.) from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), K02AG076883 (PI: M.A.M.D.) from the National Institute on Aging (NIA) and K24AI144954 (PI: D.L.S.) from the National Institute of Allergy and Infectious Diseases (NIAID).

AUTHORS’ CONTRIBUTIONS

S.B. participated in research design, the writing of the paper, the performance of the research and data analysis. Y.C. participated in research design, the writing of the paper and data analysis. S.S. participated in research design and the writing of the paper. K.L.L. participated in research design and the writing of the paper. M.S. participated in research design and the writing of the paper. D.L.S. participated in research design, the writing of the paper and the performance of the research. M.A.M.D. participated in research design, the writing of the paper and the performance of the research.

DATA AVAILABILITY STATEMENT

The data used in this study is publicly available from the Scientific Registry of Transplant Recipients.

CONFLICT OF INTEREST STATEMENT

S.S. has received an education grant from Amgen Canada to improve the experiences of patients with graft failure. K.L.L. receives consulting fees from CareDx, speaker honoraria from Sanofi, and author honoraria from UptoDate. D.L.S. has received consulting and speaking honoraria from Sanofi, Novartis, CLS Behring, Jazz Pharmaceuticals, Veloxis, Mallinckrodt and Thermo Fisher Scientific. M.A.M.D. has received honoraria from Chiesi and UptoDate. All remaining authors have nothing to disclose.