This was a single-center, observational, retrospective cohort study. Circulating anti-HLA alloantibodies were prospectively detected in 1236 consecutive patients who underwent kidney transplantation between January 2000 and December 2021. The DSA screening period spanned from January 2005 to December 2022. Anti-HLA antibody screening was routinely performed at 0, 1, 3, 6, and 12 months post-transplant and then annually as part of a routine surveillance protocol. Anti-HLA antibody detection was also conducted at the time of biopsy or following any sensitizing events. None of the patients had DSA at the time of transplantation. Patients in whom dnDSA was detected after graft loss were excluded from the study.

In addition to examining the cohort with dnDSAs, a matched control group was constructed comprising patients without dnDSAs. Controls were randomly sampled at a 1:1 ratio from the source population based on the following matching criteria: age (± 5 years), rank of transplant, and date of transplantation (± 6 months). To ensure that controls were sampled from an at-risk population, patients were considered potential controls only if they had a functioning graft at the time of dnDSA detection in the corresponding case.

Serum samples were screened for dnDSA by flow cytometry using FlowPRA beads (One Lambda, Canoga Park, CA, USA). Between 2005 and 2012, the identification was performed using single-flow antigen beads. Since 2012, the Luminex Platform has been used to identify HLA antibodies using LABScreen single-antigen beads (One Lambda). A normalized mean fluorescence intensity (MFI) cut-off value of ≥ 1500 was used as a guide to detect dnDSA and their presence was confirmed in each case by the HLA laboratory director by eplet analysis. When donor HLA-DQB1 typing data were unavailable, HLA-DQB1 typing was assigned based on frequency associations within the donor ethnic group.

Daily average tacrolimus exposure was calculated based on the blood tacrolimus levels measured at 0, 1, 3, 6, 12, and 24 months after dnDSA detection. Four patients received cyclosporine rather than tacrolimus. In such cases, C2 blood levels were converted to C0 tacrolimus equivalents using a 1/115 correction factor (8, 13, 14). Four patients did not receive tacrolimus at the time of DSA detection but were administered medication afterward.

Biopsy samples were prospectively graded by local attending renal pathologists (E.L. and J.R.) according to the Banff 1997 criteria and their subsequent updates; the diagnosis of borderline (BL) changes suspicious for acute T cell-mediated rejection (TCMR) was made using the Banff 2019 definition (15–19). All biopsies were reviewed by pathologists and nephrologists during weekly clinicopathological transplant conferences in which challenging cases were discussed.

This study was approved by our Institutional Ethics Committee (Project 2024-6949). The reported clinical and research activities are consistent with the Principles of the Declaration of Istanbul.

The Mann–Whitney test, Fisher’s exact test, or Chi-squared test was used to compare baseline clinical characteristics between patients according to the HLA class of dnDSAs. The relationship between HLA class of dnDSA, tacrolimus levels and endpoints was assessed using the Kaplan–Meier method, log-rank test, and Cox proportional hazard model. When patients had more than one DSA, the date of appearance of the first dnDSA was used as a reference to analyze time post-dnDSA. Statistical analyses were performed using STATA version 11.0 (StataCorp, College Station, TX, USA) and SPSS Statistics version 29.0 (IBM Corp., Armonk, NY, USA). All tests were two-tailed, and a p-value <0.05 was considered statistically significant.

dnDSA was detected in the sera of 55 patients during the screening period. These patients underwent kidney transplantation between January 2000 and December 2021, during which a total of 1236 patients received a kidney. The incidence of dnDSAs in the entire cohort was thus 4.4% during the period, with an incidence ratio of 0.46/100 person-years. The patients were predominantly recipients of first kidney transplant from a deceased donor ( Table 1 ).

The median time to dnDSA detection was 61 months post-transplant (25^th–75^th percentiles, 28–101). In total, 18 (33%) patients had HLA class I dnDSAs only, 33 (60%) had HLA class II dnDSAs only, and 4 (7%) had dnDSAs of both HLA classes ( Table 1 ). There were no differences in baseline clinical characteristics between these groups of patients, except for a non-significant but clear trend towards shorter time to dnDSA detection post-transplant in patients with both HLA class I and II dnDSAs.

Among the 34 patients who underwent biopsy at the time of dnDSA detection, 20 biopsies were done solely because of the DSA. There was no difference in pathological scores between those with dnDSAs of HLA class I only versus HLA class II only versus both HLA classes ( Table 1 ). Overall, rejection was observed in 24/34 (71%) of the patients who underwent biopsy: 3/34 (9%) with AMR only, 6/34 (18%) with TCMR/BL only, and 15/34 (44%) with mixed rejection. We observed no difference in the proportion of patients with a diagnosis of rejection between the HLA class specificities of dnDSA, whether class I, class II or classes I and II.

During follow-up after dnDSA detection, 30 patients experienced graft loss: 9/18 (50%) with HLA class I dnDSAs only, 17/33 (52%) with HLA class II dnDSAs only, and 4/4 (100%) with dnDSAs of both HLA classes. The median follow-up time was 39 months in patients who experienced graft loss [25^th–75^th percentiles, 10–59 months] versus 79 months in those who retained a functioning graft [39–105 months]. Survival analysis using Kaplan–Meier curves and log-rank tests showed that, compared with the presence of dnDSAs of HLA class I only or HLA class II only, the presence of DSAs of both HLA classes I and II substantially increased the risk of graft loss ( Figure 1A ). In contrast, survival was comparable for those with DSAs of either class I or class II only. Cox modeling suggested that the presence of dnDSA of both HLA classes was a strong predictor of graft loss, even after adjustment for age, sex, serum creatinine level at DSA detection, and time post-transplant at DSA detection (adjusted hazard ratio [aHR] 10.3, 95% CI 2.5–42.2, p<0.01; Table 2 ). These data did not indicate a different prognosis for patients with dnDSAs of HLA class I only compared with those with dnDSAs of HLA class II only (aHR 1.2, 95% CI 0.5–3.0, p=0.68; Table 2 ).

Twenty-eight patients had a biopsy during their follow-up after dnDSA detection. Among them, 9 patients had evidence of de novo AMR; 6 of those patients experienced graft loss during follow-up. In addition, 13 patients had signs of AMR at the time of dnDSA detection, that persisted on the follow-up biopsy; 7 of them lost their graft. 6 patients had no signs of rejection on the biopsy performed after dnDSA detection.

To further assess the prognosis of patients with dnDSA, we established a control group without dnDSA. Using the criteria mentioned previously, 51 patients were successfully matched. No major differences were observed in the baseline characteristics between cases and controls ( Supplementary Table S1 ). Kaplan–Meier curves and log-rank analysis revealed that compared with matched controls, patients with dnDSA had substantially shorter graft survival ( Figure 1B ). Similarly, adjusted models showed that all groups with dnDSAs had higher risk estimates for graft loss than the matched controls ( Table 3 ). Additionally, compared with the controls, the risk estimates for graft loss were similar between patients with dnDSAs of HLA class I only and those with dnDSA of HLA class II only (aHR 2.7, 95% CI 1.1–6.6; p=0.04, and aHR, 3.1, 95% CI, 1.5–6.6; p<0.01 respectively; Table 3 ).

Next, we examined whether the previous findings regarding the association between tacrolimus trough levels and outcomes in patients with dnDSA were valid in the current cohort (n=55). At the time of dnDSA detection, the mean (± standard deviation) tacrolimus level was 5.4±2.5 ng/ml for the entire cohort. For each patient, the average daily exposure to tacrolimus was derived from the blood levels at time 0, 3, 6, 12, and 24 months after DSA detection. Tacrolimus levels were first analyzed as a continuous predictor. Cox models revealed that the risk of graft loss decreased with higher tacrolimus trough levels following dnDSA detection (HR, 0.8, 95% CI, 0.7–0.99; p=0.04; Table 4 ), an association that remained significant following adjustments for age, sex, serum creatinine at detection and time post-transplant at detection (aHR, 0.7; 95% CI 0.6–0.9; p=0.02) as well as following adjustment of antibody removal therapies in addition to these covariates (aHR, 0.7; 95% CI 0.6–0.9; p=0.02). We found was no association between the prednisone dose and graft loss (HR 1.0 95% CI 0.9–1.2, p=0.59), nor between the MMF dose and graft loss (HR 1.0 95% CI 0.999–1.001, p=0.94).

Finally, we aimed to validate the previously established tacrolimus trough cut-off values derived from the tertiles of our previous cohort, which were as follows: <5.3 ng/mL, 5.3–6.3 ng/mL, and >6.3 ng/mL (8). In the adjusted Cox model, individuals with a mean tacrolimus level <5.3 ng/mL had a higher risk of graft loss compared with those with a tacrolimus level >6.3 ng/mL (aHR, 3.0, 95% CI 1.2–7.5; p=0.02). However, there was no significant difference in patients with tacrolimus levels of 5.3–6.3 ng/mL. Taken together, these data validate the association between tacrolimus trough levels and graft loss in this population. It corroborates that tacrolimus levels below 5.3 ng/ml predict poor outcomes in these patients.

We first sought to determine whether restricting the analysis to DSAs with higher MFIs would result in similar observations. We first examined patients using an MFI threshold of 5000 and obtained a limited number of positive DSA for each class: only 1, 16 and 1 patients respectively had DSA of class I only, class II only and both classes using this threshold. We then examined a threshold of 3000. As depicted in Supplementary Figure S1 , the number of patients with class I only, class II only and both classes with this threshold were 3, 23, and 2 respectively. With the limitation of small numbers in some of these groups, patients with DSA class I did not seem to have a better prognosis than those with class II only.

We next analyzed the impact of antibody removal therapies. First, we focused on the 18 patients with AMR. Among them, 6 patients received plasma exchanges, IVIG and rituximab in accordance to a protocol previously reported by Lefaucheur et al. ( 20 ), 2 patients received the same protocol except for rituximab, and 1 patient received only solumedrol; the remaining 9 patients did not receive additional treatment. Supplementary Table S2 displays the outcome according to the treatment. The median graft survival for the 10 patients who received solumedrol only or no additional treatment was 43 months post DSA detection, versus 52 months for those who received antibody removal therapy; the difference was not significant (p=0.61). Supplementary Table S2 shows the proportional hazards model adjusted for age, sex, serum creatinine at DSA detection, and time post-transplant; there was no signal for a difference between groups. Second, we examined the impact of antibody removal therapies according to the HLA class of DSA at the cohort level. We found no impact of such treatment on graft survival for patients with class I only and class II only ( Supplementary Figures S3A, B ); survival was non-significantly better in the four patients with DSA of both classes ( Supplementary Figure S3C ). Given the small number of patients in each subgroup, these results need to be interpreted with caution.