This study was a retrospective analysis of all adult kidney-only transplantations in Finland between January 2006 and December 2021. The data were obtained from the national Finnish Transplant Registry, and/or electronic patient records from Helsinki University Hospital (HUS), which is the only transplant center in Finland. HLA antibody analyses were made at the Finnish Red Cross Blood Service, which process this data on behalf of HUS. This study was approved by the Institutional Review Board of HUS, Abdominal Center (HUS/136/2024)

The inclusion and exclusion criteria of the patient material are presented in Figure 1. Living donor transplants and multi-organ transplantations were excluded. All transplantations were from donors after brain death, as donors after circulatory death were not used in our center during the study period. All transplantations were complement dependent cytotoxicity crossmatch negative, examined with donor splenocytes between 2006 and 2015 and peripheral T- and B-cells thereafter. No desensitization was used. According to the local immunosuppression protocol, the baseline immunosuppression was a combination of a calcineurin inhibitor (cyclosporine or tacrolimus, of which cyclosporine was used in earlier years, and tacrolimus increasingly in the later years), mycophenolate (1,000 mg twice daily with cyclosporine or 500 mg twice daily with tacrolimus) and steroids. Induction was with basiliximab for recipients with retransplantations or HLA mismatch grade over 3 in the A, B, and 1 DR loci. Anti-T-lymphocyte globulin was used for induction after 2015 in patients with known or presumed (panel reactive antibody (PRA) percentage over 80%) DSAs at the time of transplantation. Target trough levels for tacrolimus are 7–10 μg/L for the first 3 months and later 4–6 μg/L. As for cyclosporine, target levels are 160–200 μg/L for the first 3 months and 80–110 μg/L until 1 year after transplantation. Steroids were withdrawn after the first posttransplant year for most stable patients with no history rejections or immunological baseline kidney disease.

Data of the recipients included number of previous kidney transplants, time on the waitlist, age at transplantation, sex, height, weight, body-mass index, ABO blood type, smoking status, ICD-10 code of kidney disease, cytomegalovirus (CMV) status, highest and last calculated PRA percentage for class I and II HLAs pre-transplantation, dialysis time and last dialysis type (hemodialysis or peritoneal dialysis). Donor data included gender, age, height, weight and CMV status. Transplantation data included HLA AB and DR mismatch grades, cold ischemia time, follow-up time and status of graft survival, biopsy-proven acute rejection (yes or no), and rejection time and type (T-cell or antibody mediated), as categorized by the respective Banff classification at the time of biopsy [9]. Borderline findings were registered as rejections if they were treated accordingly. Mixed phenotypes were classified as AMR.

The DSA status of sensitized patients, including specificity and MFI, were collected from posttransplant tissue compatibility reports. One Lambda Labscreen® mixed and single antigen beads with Luminex® were used for HLA antibody screening and identification with the use of HLA Fusion software (One Lambda Inc., Canoga Park, CA). MFI level >1,000 was considered positive. Highly sensitized recipients with no available DSA data (n = 7) were excluded from the study. Donor HLA typing was performed at intermediate resolution, complemented with haplotype analyses.

Recipients were compared in different groups based on their DSA status (DSAs vs. no DSAs). Recipients with DSAs were also divided into different subgroups based on their cumulative overall DSA MFI: 1,000–4,999, 5,000–9,999, or at least 10,000. To find out the significance of class I vs. class II DSAs, the DSA group was also categorized based on the separate class I and class II cumulative MFIs, using the same above-mentioned limits. DSA MFI was also tested for the models as a continuous variable, but as the linearity assumption was not met, categorized MFI was used instead.

The outcomes examined included DCGS, defined as return to dialysis or retransplantation, and censoring for deaths; biopsy-proven acute rejections (BPAR) and AMR. For multivariable Cox regression analyzing factors associated with DCGS, confounder analysis for the model was constructed as a directed acyclic graph (DAG) [10]. The DAG (Figure 2) presents factors possibly affecting the presence of DSAs at the time of transplantation and confounding the associations with graft survival. For Cox regression models, we included cumulative MFI (categorized into four groups (no DSAs, 1,000–4,999, 5,000–9,999, at least 10,000)), donor age, recipient sex, age and kidney disease (diabetic kidney disease, glomerulonephritis, polycystic kidney disease or other), retransplantation (yes or no), time on the waitlist, cold ischemia time, use of induction immunosuppression (other than iv steroids), and use of cyclosporine (vs. tacrolimus), as covariates. Relevant first-degree interactions between DSAs and other variables were tested, and the assumption for proportional hazard was supported for DSAs. In addition, interaction between retransplantation and cumulative MFI, was examined (and found nonsignificant). Graft survival was estimated with the Kaplan-Meier method where graft loss (return to dialysis or retransplantation) was used as the event while death and end of follow-up time were censored. Factor comparisons for DCGS were calculated with the Log Rank test.

For multivariable analysis of BPAR and AMR, a Cox regression analysis was run with the same covariates as the analysis regarding DCGS, presented in the DAG, taking rejection time into account. Interaction between retransplantation and cumulative MFI was examined and found insignificant for BPARs and significant for AMRs. To compare the risk of rejections within different groups, univariable analyses were calculated using Cox regression, for DSA status (any DSAs, and class I and class II DSAs separately), and categorized cumulative MFI for total MFI, and class I and class II separately. The assumption for proportional hazard was again supported for DSAs.

As a sensitivity analysis, all statistical tests were repeated for immunodominant MFIs for class I, class II and all DSAs.

All statistical analyses were done using IBM SPSS Statistics, version 28.0. P-value of under 0.05 was considered statistically significant. Missing data were assumed to be missing at random, and as the number of missing data were low for the variables of interest for the multivariable analyses, cases with missing values were excluded from the respective analyses.

A total of 3209 transplantations were included in this study, of which 226 (7%) recipients were DSA positive at the time of transplantation. The patient characteristics are described in Table 1. The only missing data were for recipient BMI (n = 361), time on the waitlist (n = 4), recipient CMV status (n = 29), donor CMV status (n = 53), donor age (n = 1), cold ischemia time (n = 2), last higher and total PRA% before transplantation (n = 294), average maximum PRA% before transplantation (n = 238), HLA AB mismatch (n = 1) and DR mismatch (n = 1). Statistically significant differences were found for recipient sex, retransplantation, baseline kidney disease, recipient time on waitlist, recipient CMV positivity, donor age, being on tacrolimus (vs. cyclosporine), induction immunosuppression, delayed graft function and HLA AB mismatch. The maximum follow-up time was 16.0 years, with a median of 5.4 years.

Among recipients with DSAs, the mean cumulative MFI was 11,563 (SD 13 886). Altogether 90 recipients had a cumulative MFI of 1,000–4,999, 46 had 5,000–9,999, and 86 had at least 10,000. Four DSA positive recipients did not have the complete MFI data available, of which three recipients had only class I DSAs with missing MFI, and one recipient had both class I and II DSAs with MFI available for class I but unavailable for class II, and these recipients were excluded from the analyses concerning missing data.

Altogether 162 recipients had class I DSAs and 119 had class II DSAs, while 55 recipients had both. Categorized into groups based on class I DSAs, 77 recipients had a cumulative MFI of 1,000–4,999, 43 had 5,000–9,999 and 40 had at least 10,000. As for class II DSAs, 48 patients had a cumulative MFI of 1,000–4,999, 25 had 5,000–9,999, and 45 had at least 10,000.

DSAs were associated with worse death-censored graft survival (p < 0.001) in comparison to no DSAs (Figure 3). Recipients with DSAs had a 5-year DCGS of 84% and 10-year survival of 74% while recipients without DSAs had a 92% 5-year DCGS and an 83% 10-year DCGS.

Higher cumulative MFI indicated worse 5-year DCGS (no-DSA: 92%, MFI 1,000–4,999: 89%, 5,000–9,999: 81%, at least 10,000: 79%, p < 0.001) and 10-year graft survival (no-DSA 83%, 1,000–4,999: 76%, 5,000–9,999: 78%, at least 10,000: 66%) (Figure 4). However, there were no statistically significant differences between the groups with different cumulative MFI levels (p = 0.441).

The results of multivariable Cox regression analysis regarding DCGS are presented in Table 2. Recipient age, baseline kidney disease, cold ischemia time, donor age and retransplantation were independently associated with DCGS. A trend was seen in the association of categorized cumulative MFI with worse DCGS, but this association did not reach statistical significance.

Regarding class I DSAs, there was no statistically significant (p = 0.841) association of higher cumulative MFI indicating worse 5-year DCGS or 10-year DCGS (Figure 5). Similarly, regarding class II DSAs, there was no statistically significant (p = 0.069) association of higher cumulative MFI indicating worse 5-year DCGS or 10-year DCGS (Figure 6).

Rejection frequencies for BPAR, AMR and TCMR are described in Table 3. Hazard ratios for BPAR and AMR are presented in supplementary material, Supplementary Table S1.

DSAs were associated with more BPAR (HR 2.0, 95% CI 1.6–2.5, p < 0.001), with even higher HR for acute AMR (HR 22.7, 95% CI 14.3–36.2, p < 0.001), as 40% of DSA positive recipients had BPARs and 26% had AMR. Comparing only the class I or class II DSA status to those without the above-mentioned DSAs, both acute overall rejections (class I 39.5% vs. class II 47.9%) and acute AMR (class I 25.9% vs. class II 34.5%) were more prevalent in recipients with class II DSAs, but the HRs were higher for class I. Comparing recipients with class I DSAs to those without them, HR for BPAR was 2.2 (95% CI 1.7–2.9, p < 0.001) and HR for acute AMR was 16.4 (95% CI 10.5–25.6, p < 0.001), while comparing recipients with class II DSAs to those without them, the HR for BPAR was 1.8 (95% CI 1.4–2.4, p < 0.001) and the HR for acute AMR was 13.9 (95% CI 9.0–21.4, p < 0.001).

Examining the cumulative effect of DSAs, both the hazard ratio of BPAR (p < 0.001) and acute AMR (p < 0.001) was increased with the increasing cumulative MFI, categorized into 1,000–4,999, 5,000–9,999 and at least 10,000. Regarding cumulative class I MFIs, the frequency of acute AMR (p = 0.0499) was increased with the increasing categorized cumulative MFI. There was no statistically significant difference for the frequency of BPAR, but the HR increased for every MFI limit for both BPAR and AMR. As for different categories of cumulative class II MFIs, the frequency of BPAR (p = 0.057) and AMR (p < 0.001) was decreased at the limit of 10,000, which was also seen in hazard ratios at the same limit.

The multivariable Cox regression analyses for the risk of acute rejections using the categorized cumulative total MFI are presented in Table 4. For acute rejection, statistically significant risk factors were categorized cumulative MFI, donor age, induction immunosuppression, and being on cyclosporine vs. tacrolimus recipient age, while categorized kidney disease, cold ischemia time, recipient sex, retransplantation and time on the waitlist were statistically unsignificant. Regarding acute AMR, the only statistically significant risk factors were categorized cumulative MFI and recipient age.

As a sensitivity analysis, all analyses were repeated for immunodominant MFIs for class I, class II and all DSAs. The results were considered similar to the primary analyses, as higher MFI for the immunodominant DSA was similarly associated with increased risk of rejection and inferior graft survival in all analyses. The results of these sensitivity analyses are presented in supplementary material, Supplementary Tables S2, S3 and Supplementary Figures S1–S3.