This retrospective, multicenter study was performed at four French adult KT centers (Clermont-Ferrand, Bordeaux, Rouen and Poitiers). Patients were selected using the Cristal prospective database. The inclusion criteria were patients over 18 years old who had undergone a second KT between 1 January 2008 and 31 December 2020 at the Clermont-Ferrand, Rouen, or Poitiers transplant centers or between 1 January 2016 and 31 December 2020 at the Bordeaux transplant center (because of a change of the computerized patient record systems). The exclusion criteria consisted of second preemptive transplantations, continuation of IS treatment without CNIs, and multiorgan transplantations.

The following demographic, clinical, and biological data were collected: i) at the time of KTF—age, sex, body mass index, initial kidney disease, first transplant outcome and cause of allograft failure, PRA level, and the eventual presence of donor-specific anti-HLA antibodies (DSAs); ii) at the inscription on the waiting list for a second KT—PRA level and comorbidities (diabetes, stroke, ischemic heart disease, lower limb revascularization, neoplasia, and persistent post-KTF infection); iii) during the dialysis period—potential allograft nephrectomy, severe infection defined as an opportunistic infection [30] or requiring hospitalization [31], major cardiovascular events (hospitalization for acute coronary syndrome, cardiac arrhythmia, heart failure, lower limb revascularization, and stroke), and whether IS with CNIs was maintained; iv) at the retransplant initial hospitalization—induction therapy modalities, PRA level, eventual presence of DSAs (against the new KT), the type of donor (expanded criteria donor [32] or living donor), residual diuresis, and delayed graft function defined as the requirement of at least one dialysis session during the first week after transplantation [33]; and v) during the follow-up after the second KT—graft rejection episodes (Banff 2019 [34]), the appearance of DSAs, severe infection, major cardiovascular events, neoplasia, graft, and patient survival. Detection of anti-HLA antibodies was performed using the Luminex Single Antigen method (One Lambda, Canoga Park, CA) at the Clermont-Ferrand, Bordeaux, and Poitiers centers or Immucor Lifecodes (Immucor, Stamford, CT) at the Rouen center [35].

Oversight and study approval were provided by the Committee for Protection of Human Subjects (CPP SUD-EST VI) on 3 September 2019 (institutional review board 00008526) and by the National Consultative Committee on the Use of Health Research Information (14.510). No written consent was required for this study, but a non-opposition letter was sent to all patients in accordance with national legislation (MR-004 reference methodology) [36].

Two groups of patients were defined according to the modality of management of IS in the period between the two KTs: i) the CNI group (G-CNI), defined by the continuation of IS including CNIs either as monotherapy or in combination with other IS (i.e., steroids, mycophenolate mofetil, azathioprine, and mTOR pathway inhibitors) during the entire period between the two KTs, and ii) the stop group (G-STOP), defined by the cessation of all IS during the intertransplant period.

Statistical analyses were performed with Stata software (version 15; StataCorp, College Station, Texas, USA). All tests were two-sided with an alpha level set at 5%. Categorical variables were expressed as number of patients and associated percentages, and continuous variables as mean ± standard deviation or median [25th; 75th percentiles], according to their statistical distribution.

Demographic and first transplant characteristics were compared between G-STOP and G-CNI using usual statistical tests: chi-squared test or Fisher’s exact test for categorical variables and Student’s t-test or Mann-Whitney test for continuous variables.

To assess the relationship between the group (G-STOP and G-CNI) and the primary and secondary endpoints, a propensity score (PS) analysis was implemented using the inverse probability of treatment weighting (IPTW) method [37, 38]. The PS was derived from the probability that treatment with a CNI would be continued for a given patient (G-CNI) conditional on confounders. The IPTW method consists of creating a “pseudo sample” of treated (G-CNI) and untreated (G-STOP) patients, weighting each patient by the inverse probability of receiving the treatment he or she actually received as follows: 1/PS in the G-CNI and 1/(1-PS) in the G-STOP. In practice, the probability of continuing CNI therapy was modeled using multiple logistic regression, and the estimated probability was used as the PS. Variables were selected for the PS based on clinical relevance: age at the end of the first transplant, cardiovascular comorbidities at the end of the first transplant, cause of first transplant failure, and cPRA level at the inscription on the waiting list for retransplant. Patients with missing cPRA levels at the inscription date were excluded from the analysis, as were patients with diabetes or infections because they were all G-STOP patients. Balance between groups was measured by standardized mean differences, calculated before and after weighting, and expressed as absolute values. A value greater than 0.2 was considered a sign of imbalance.

The primary outcome was a composite of dialysis and death after the second KT, presented as survival free of dialysis and death. This outcome was expressed as censored data and was estimated with the Kaplan-Meier method, and the groups were compared by the log-rank statistic. Multivariable analyses were performed with a Cox model (with the center as a random effect) considering covariables in terms of their significant results in univariate analysis (p < 0.10) as well as their clinical relevance^14,32. The results are expressed as hazard ratios (HRs) and 95% confidence intervals (CIs).

Secondary outcomes were compared in both groups by mixed models, considering the center as a random effect: linear mixed models were used for continuous outcomes and generalized linear mixed models with the logit link function were used for binary outcomes.

Finally, exposure-adjusted rates were calculated as the total number of event episodes (including recurrent events) over the total duration of follow-up and are expressed per 100 patient-years (p-y).

Among the 3246 KTs performed at the four centers during the study period, 407 patients (12.5%) received a second KT (Figure 1). Five patients with multiorgan transplantation were excluded, as well as 44 patients with preemptive KT, 31 patients with IS without CNIs, and 52 due to lack of data. A total of 275 patients were included, 216 in the G-STOP and 59 in the G-CNI. The median follow-up time after the second KT was 3.6 years [2.0; 7.0].

The characteristics of the patients before IPTW are depicted in Table 1. The 275 patients included were mainly men (64.7%), aged 49.7 ± 13.6 years at KTF. The primary cause of the first graft loss was rejection (61.8%). G-STOP patients compared to G-CNI patients had more diabetes at the end of the first KT (11.6% vs. 1.7%, p = 0.02), a shorter transplantation survival (92 months [34; 163] vs. 133 [87; 220], p = 0.003), and a higher cPRA at inscription on the waiting list for retransplant (51% [0; 86] vs. 5% [0; 70], p = 0.02). The rate of hyperimmunized patients (cPRA ≥85%) in the G-STOP and G-CNI was 26.7% and 13.2%, respectively (p = 0.04). Among G-CNI patients, 36 (61.0%) were treated with tacrolimus and 23 (39.0%) with cyclosporine. IS maintenance until the second KT consisted of CNI monotherapy in 19 patients (32.2%) and CNI combined with an antimetabolite or corticosteroid therapy in 30 patients (50.8%). Ten patients (17.0%) received triple IS.

The characteristics at the time of KTF of the 205 patients included in the PS analysis are summarized in Table 2. After applying the IPTW method, the G-STOP and G-CNI were well balanced (standardized mean differences <20%) for the variables included in the IPTW model: age, cardiovascular comorbidities, the cause of first-transplant failure, and the cPRA level at inscription on the waiting list for retransplant.

The median time on dialysis until the second KT was 21 months [11; 43]. This value was significantly lower in the G-CNI than in the G-STOP (16 [5; 26] vs. 37 [22; 64], respectively, p < 0.001). The waiting times from relisting to the second KT were 16 months [8; 23] in the G-CNI and 27 months [13; 48] in the G-STOP (p = 0.06) (Table 3).

G-CNI patients, compared to G-STOP patients, had a lower median cPRA level at the time of the second KT (67% [0; 84] vs. 87% [55; 96], p = 0.001). The rate of hyperimmunized patients was also lower in the G-CNI: 23.9% versus 55.2% in the G-STOP (p < 0.001). The numbers of patients transplanted with preformed DSAs and induction treatment were comparable between the groups (Table 3).

Patients in the G-STOP were more frequently transplanted with an expanded criteria donor graft (43.2% vs. 29.9% in the G-CNI, p = 0.01). Hyperimmunized patients, compared with patients with cPRA levels <85%, were more likely to receive a kidney transplant from an expanded criteria donor [44.1% and 32%, respectively (p = 0.005)]. On the day of the second KT, 44.1% of G-CNI patients had a residual diuresis ≥500 mL compared to 13.8% in G-STOP (p = 0.06). The delayed graft function rate was 9.9% in the G-CNI and 25.8% in the G-STOP (p = 0.001). The numbers of patients transplanted with preformed DSA and an induction treatment were comparable between the groups (Table 3). Data before IPTW are presented in Supplementary Table SI.

After the second KT, 10 years survival free of dialysis and death was significantly better in the G-CNI than in the G-STOP (HR: 0.06, 95% CI: 0.01–0.30, p = 0.001) (Figure 2), with 10 years survival rates of 98.7% and 59.5%, respectively. In multivariable analysis after adjustment for expanded criteria donor, rejection, delayed graft function, age at second KT, graft survival time from the primary transplant, and rejection as etiology of first graft failure, continuation of CNIs between the two KTs was associated with a better 10 years survival free of dialysis and death (HR: 0.08, 95% CI: 0.01–0.58, p = 0.01) (Table 4). The difference in survival also remained significant after sensitivity analysis excluding second living donor transplants, with a 10 years survival rate of 98.5% in the G-CNI versus 56.4% in the G-STOP (HR: 0.06, 95% CI: 0.01–0.30, p = 0.001). Data on survival before IPTW are presented in Supplementary Figure S1.

A return to dialysis was observed in 18.3% of G-STOP patients compared to 1.3% of G-CNI patients (p = 0.004). The main cause of graft loss was rejection (45.1%). The number of humoral rejections and the occurrence of DSA were comparable in the two groups, but there was less cellular rejection in the G-CNI than in the G-STOP (2.1% vs. 8.8%, respectively, p < 0.001). All deaths were observed in the G-STOP (Table 3; Supplementary Table SI).

In the period between the two KTs, the serious infectious event rates and their exposure-adjusted rates (patient-years) in the G-CNI and G-STOP were similar (Figure 3A; Supplementary Figure S2). The rates of cardiovascular events and neoplasia and their exposure-adjusted rates were significantly lower in the G-CNI than in the G-STOP (Figure 3A; Supplementary Figure S2).

At the last follow-up after the second KT, the rates of patients with neoplastic events were similar in the G-CNI and G-STOP (Figure 3B). The rate of cardiovascular events was lower in the G-CNI than in the G-STOP (7.9% and 15.9%, respectively, p = 0.04) (Figure 3B). The serious infectious event rates were similar in the G-CNI and G-STOP (Figure 3B), but the exposure-adjusted rate was higher in the G-CNI than in the G-STOP (28.2/100 p-y and 22.8/100 p-y, respectively; p = 0.02) (Supplementary Figure S2).

Overall, after the first KTF, patients in the G-CNI and the G-STOP had a higher exposure-adjusted rate of serious infection (22.0/100 p-y and 15.3/100 p-y, respectively; p < 0.001) but a lower rate of major cardiovascular events (1.5/100 p-y and 4.8/100 p-y, respectively; p < 0.001). The exposure-adjusted rate of neoplasia was similar in both groups (Supplementary Figure S2).