The study was conducted in accordance with the Declaration of Helsinki and it was approved by the Bioethical Committee of Wroclaw Medical University (certificate no. KB 334/2025). This study integrates both retrospective and prospective approaches. The study population included all patients who had undergone KTx and had a biopsy of the transplanted kidney between January 2018 and July 2025.

Initially, 218 records were identified, but these did not include non-HLA antibodies. At our clinic, routine assessment for anti-HLA antibodies was conducted at the time of biopsy, and if patients consent, their serum samples are stored in the clinic’s freezers after being collected for anti-HLA testing. After reviewing the availability of serum samples from the freezer, the final cohort consisted of 151 biopsy results. Thus, the retrospective analysis focused on biopsy results, while the prospective component involved measuring anti-AT1R antibodies with IgG subclasses in the stored serum samples.

In summary, the inclusion criteria were post-renal transplantation status, availability of a biopsy of the transplanted kidney performed no earlier than 2018, availability of frozen serum samples collected at the time of the biopsy, and access to the patient’s medical history along with other routinely assessed laboratory and clinical data. Patients who did not meet these criteria were excluded from the analysis.

To ensure the reliability of the histopathological assessments, only biopsy results from 2018 onwards were included, as earlier Banff classifications were considered outdated. All evaluations were performed by experienced nephropathologists (PD and AH) from the University Hospital in Wroclaw.

The assessments followed the most recent Banff classifications available at the time of each biopsy, specifically Banff 2017, Banff 2019, and Banff 2022, which were incorporated into the final analysis (3, 18, 19). Tissue samples were processed according to a standardized protocol: paraffin-embedded specimens were cut into 3-μm sections and stained with haematoxylin and eosin (H&E), periodic acid–Schiff (PAS), Jones methenamine silver (JMS), Masson’s trichrome, and immunohistochemically for C4d, ensuring consistency and reproducibility.

The evaluation focused on glomerular, interstitial, tubular, vascular, and peritubular capillary compartments. Each compartment was scored on a standardized scale from 0 (no lesions) to 3 (severe lesions), reflecting the extent of pathological changes. The mean time between transplantation and biopsy was 4.73 years, ranging from 10 days to 22.9 years.

Blood samples were obtained as part of routine standard-of-care testing. The mean interval between biopsy and blood collection was 28 days; in 71 cases, samples were collected within ≤7 days of biopsy. Venous blood was drawn in the morning into serum separator tubes (fasting not required), centrifuged at 1500 × g for 10 min at room temperature, and serum was aliquoted and stored at −80 °C until analysis.

Total anti-AT1R IgG and anti-AT1R IgG1–IgG4 subclasses were measured by CellTrend GmbH (Luckenwalde, Germany) using ELISA kits (total IgG: #12000; IgG1: #12010; IgG2: #12020; IgG3: #12030; IgG4: #12040). Plates were pre-coated with AT1R antigen; bound antibodies were detected with subclass-specific enzyme-conjugated anti-human IgG (or anti-human IgG subclass) secondary antibodies, followed by a chromogenic substrate reaction. Total anti-AT1R IgG was reported in U/mL, whereas subclass results were reported as relative optical density (OD). All samples were assayed in duplicate and mean values were used for downstream analyses.

All statistical analysis was performed using Python version 3.13. The following libraries were used: pandas and numpy for data handling (20, 21), scipy.stats for standard statistical tests (22), statsmodels for regression analyses (23), and scikit-learn for ROC curve and AUC calculations (24). Data visualization was performed using matplotlib and seaborn to generate figures (25, 26).

Categorical variables were analyzed using the chi-square (χ²) test to assess associations between groups. For comparisons between two groups, either Student’s t-test (for normally distributed continuous data) or the Mann–Whitney U test (for non-normally distributed data) was applied. For comparisons across more than two groups, one-way ANOVA or the Kruskal–Wallis test was used.

Associations between continuous variables were evaluated using Pearson correlation coefficients for normally distributed data and Spearman rank correlation coefficients for non-parametric data. To assess antibody levels to predict the likelihood of antibody-mediated rejection, receiver operating characteristic (ROC) curves were constructed, and the area under the curve (AUC) was calculated.

Logistic regression models were used to quantify the association between antibody levels and the probability of AMR occurrence. Regression coefficients were transformed into odds ratios (ORs), which reflect the relative change in the odds of AMR per 1 SD increase pertaining to antibody level or optical density (OD) (standardization purposes). Statistical analyses were performed using Mann-Whitney U tests to compare AMR and non-AMR groups. Odds ratios comparing the 75th to 25th percentile (interquartile range) were calculated to provide clinically interpretable effect sizes.

Antibody levels were also categorized into quartiles to evaluate trends in AMR prevalence across different ranges of antibody concentration or OD. Differences between quartile groups were analyzed using the chi-square test.

The study comprised 151 biopsy results of 143 patients with available biobanked serum, Banff-classified diagnostic biopsies, and complete clinical data. Patients were divided into the following diagnostic groups: no rejection or borderline changes (N = 83), any form of T-cell–mediated rejection (TCMR; N = 35), and those with antibody-mediated rejection (AMR; N = 33). The main characteristics of the patients are summarized in Table 1. These cohorts were evaluated for total anti-AT1R AAB levels and IgG subclasses (IgG1–4). No relevant associations were observed for IgG1, IgG2, IgG4 (results not shown), hence only total anti-AT1R and IgG3 subclass were subjected to further analysis.

As shown in Table 1, C4d positivity and anti-HLA2 antibody positivity were strongly associated with AMR. To assess whether anti-AT1R AABs are independent from these factors a further analysis was performed. For each C4d score (0-3) the mean level and median value of anti-AT1R AABs (total and IgG3) was assessed. The analysis revealed no association indicating that antibody levels do not correlate with C4d positivity (Table 2) (p values of 0.56 and 0.22 for AT1R-IgG and IgG3 respectively).

As anti-HLA2 antibodies were strongly associated with AMR occurrence, we assessed whether mean and median anti-AT1R AAB values differ depending on anti-HLA2 status. For both, total IgG and for IgG3 subclass no associations were observed (Table 3) (p values 0.91 and 0.72 respectively). Considering that both, total anti-AT1R and IgG3 subclass were independent from C4d and anti-HLA2 status, indicated their potential as individual AMR diagnostic markers, which was analysed further.

For total AT1R-IgG, the lowest antibody levels were observed in the no-rejection/borderline cohort (mean 11.3; median 10.37). Higher levels were detected in both the AMR and TCMR groups (mean 14.13 and 15.01; median 11.09 and 14.09, respectively) (Figure 1). However, none of these differences reached statistical significance. For the IgG3 subclass, the lowest mean absorbance was again observed in the no-rejection/borderline group (mean 0.81; median 0.73). Higher absorbance values were found in the TCMR (mean 0.88) and AMR (mean 0.96; median 0.98) cohorts (Figure 1). In this case, statistically significant differences were detected using the Kruskal–Wallis test (p = 0.0396).

AT1R and AT1R-IgG3 were assessed for correlation with AMR diagnosis using both Pearson and Spearman coefficients. No association was observed for total AT1R (Pearson r = 0.0854, p = 0.2971; Spearman ρ = 0.0507, p = 0.5361). For the IgG3 subclass, a positive association was detected by Spearman correlation (ρ = 0.1890, p = 0.0201), whereas Pearson correlation did not confirm this association (r = 0.1320, p = 0.1063). This discrepancy suggests that the relationship between the variables may be monotonic but not linear.

Anti-AT1R-IgG (total) levels and AT1R-IgG3 were evaluated for their predictive performance for AMR using ROC AUC curves. The diagnostic performance was assessed by the AUC, which measures the ability of each antibody to discriminate AMR occurrence. The AUC for total AT1R IgG was 0.535 (95% CI: 0.42–0.65), indicating no predictive ability. In contrast, IgG3 showed an AUC of 0.632 (95% CI: 0.52–0.74), suggesting limited but potentially additive predictive value (Figure 2).

Logistic regression analyses were performed to evaluate the association between AT1R-IgG (total) and AT1R-IgG3 levels and the risk of antibody-mediated rejection. For total AT1R-IgG, the regression coefficient was 0.178, corresponding to an odds ratio (OR) of 1.19, indicating that higher AT1R-IgG levels were modestly associated with increased risk of AMR. However, the predictive performance of the model was limited, with a cross-validated area under the receiver operating characteristic curve (AUC) of 0.53 ± 0.08.

In contrast, AT1R-IgG3 showed a stronger association with AMR. The regression coefficient was 0.287, yielding an OR of 1.33, which means that the chances for AMR occurrence rise by 33% per 1 SD increase … The predictive ability of AT1R-IgG3 was higher than total AT1R-IgG, as reflected by a cross-validated AUC of 0.63 ± 0.08, suggesting moderate discrimination between patients with and without AMR.

To better facilitate clinical interpretation, we calculated ORs comparing patients at the 75th percentile to those at the 25th percentile (IQR (interquartile range)) for each antibody. OD of AT1R-IgG3 antibodies was significantly higher in AMR patients (median 0.98 vs 0.75, p = 0.021). Comparing patients at the 75th percentile (1.01) to those at the 25th percentile (0.58), the IQR was 0,4297 with OR 1.39 (39% increase in odds). AT1R-IgG (total) antibodies showed no significant association with AMR (p = 0.536), with IQR 6,8232 and OR = 1.16.

It must be noted that the analyses were univariate and reflect rather associative than independent effects.

We assessed the prevalence of AMR according to quartiles of total AT1R-IgG levels and AT1R-IgG3. AMR prevalence across total AT1R-IgG quartiles was not consistent, varying from 18.4% in Q1 to 26.3% in Q4, with no clear linear trend (slope = 0.019, p = 0.53, not significant) (Figure 3).

In contrast, the prevalence of AMR increased progressively across AT1R-IgG3 quartiles, ranging from 10.5% in the lowest quartile (Q1) to 31.6% in the highest quartile (Q4) (Figure 4). Trend analysis confirmed a statistically significant increase in AMR risk with higher AT1R-IgG3 levels (slope = 0.072, p = 0.017). These results suggest that AMR risk is more strongly associated with AT1R-IgG3 subclass levels than with total AT1R-IgG.

Additional comparative analyses were performed across AT1R-IgG3 quartiles to further evaluate the pattern of association with AMR. AMR prevalence was significantly higher in the upper half of the distribution (Q3–Q4: 29.3%) compared with the lower half (Q1–Q2: 14.5%), p = 0.044; OR = 2.45. The statistically significant difference between the extremes (Q1 vs Q4, p = 0.049, OR = 3.92) also confirmed that at the ends of the distribution, the discriminatory capacity is substantial. However, the lack of statistical significance in pairwise comparisons between adjacent quartiles suggests that the gradient is not strictly linear, and the differences between consecutive quartiles are relatively modest. Taken together, these results suggest that the association between AT1R-IgG3 levels and AMR exhibits characteristics of both a dose-response relationship (with increasing risk apparent from Q3 onward) and a gradual threshold effect (with the most pronounced differences at higher quartiles). The clinically relevant distinction appears to occur between patients with values below versus above the median, with those in the upper half of the distribution (Q3-Q4) demonstrating 2.45-fold increased odds of AMR compared to those in the lower half (Q1-Q2).