The study was approved by the local ethics committee (EKNZ 2026-00071). This is a retrospective study based on all renal transplantations performed at the University Hospital Basel from January 2014 until August 2024. During this time frame, 825 transplantations were performed. All consenting adult kidney transplant recipients who had an informative 1-year surveillance biopsy and complete standard urine sample data were included (Figure 1). The standard urine analysis at the timepoint of the 1-year surveillance biopsy included a urine sediment and measurements of total proteinuria (protein/creatinine ratio; normal range <11 mg/mmol), albuminuria (albumin/creatinine ratio; normal range <3 mg/mmol), and α1-microglobulinuria (α1-microglobulin/creatinine ratio; normal range <1.58 mg/mmol). Aliquots of these urine samples were stored at −75 °C for later use.

Urine CXCL10 was measured with the MSD V-Plex Human IP-10 Gen. B Kits (Meso Scale Discovery) and is given as CXCL10/creatinine ratios [ng/mmol creatinine] as previously reported [15]. Urine creatinine was measured by an enzymatic method on a Roche Cobas 8100 instrument. Urine cell components were analyzed by flowcytometry. BKPyV- and JCPyV-DNAuria were quantified retrospectively from stored aliquots using well-validated, species-specific laboratory-developed quantitative PCR assays as previously described [16–18]. Significant leukocyturia, relevant BKV- or JCV-DNAuria, and subclinical rejection were considered as individual inflammatory processes. If several such processes occurred simultaneously, they were summarized as “combination” of inflammatory processes. The individual inflammatory processes were defined as follows: (i) relevant DNAuria was defined as >10⁶ copies per milliliter [c/mL] for BKV and >10⁷ c/mL for JCV in accordance with the recently published international consensus guidelines [19]; (ii) significant leukocyturia was defined as >10 leukocytes per high power field (HPF), which corresponds in our laboratory to 56 leukocytes/μL; and (iii) rejection was defined by the Banff 2022 classification [20].

If the measured values were below the detection threshold (e.g., protein/creatinine ratio), for analytical purposes we used a value of half the detection threshold instead. In addition, urinary leukocyte counts of <1/HPF were set as zero.

Categorical data are shown as count (percentage) and analyzed by Χ²-test. In the absence of a normal distribution, continuous data are presented as median (interquartile range). Receiver operating characteristics (ROC) were performed to assess the diagnostic characteristics of urinary CXCL10/creatinine ratio, serum creatinine, protein/creatinine ratio, albumin/creatinine ratio, and α1-microglobulin/creatinine ratio to detect inflammatory processes. A multivariable model was used to assess independently associated parameters with subclinical allograft rejection. It included all clinical and laboratory parameters that had a p-value <0.10 in the univariate analysis. The variable selection for the univariate analysis was based on established risk factors for allograft rejection as well as known confounders for urinary CXCL10/creatinine levels (i.e., relevant BKV-DNAuria, relevant JCV-DNAuria, and significant leukocyturia). A two-sided p-value of <0.05 was considered statistically significant. For statistical analysis we used IBM SPSS Statistics version 31.0.0.0 (117), JMP Pro 18, and R statistical software ¹ .

In the study, 405 patients were included (Figure 1). The median recipient age was 55 years (44-63), and a third of all participants were female. Deceased donors accounted for 58% of the allografts and median donor age was 56 years (47-63). Twenty-two percent of the transplantations were considered as high-immunological risk (i.e., pretransplant donor-specific HLA-antibodies [HLA-DSA], ABO incompatibility, or husband-to-wife transplantation with mutual children). Induction therapy was mainly basiliximab (84%) and ATG (15%); the initial maintenance therapy was a triple therapy consisting of tacrolimus, mycophenolic-acid, and steroids in 99%. De novo HLA-DSA at 1-year posttransplant were detected in 11/405 patients (2.7%). The detailed population characteristics are summarized in Table 1.

Two hundred and eighty-one of the 405 patients (69%) had no inflammatory process at the 1-year surveillance biopsy. The most frequent individual inflammatory process was rejection with 15.8% (n = 64), followed by leukocyturia with 11.6% (n = 47), BKV-DNAuria with 5.7% (n = 23), and JCV-DNAuria with 2.2% (n = 9). Eighteen patients (4.4%) showed an overlap of inflammatory processes. The most frequent combinations were leukocyturia with rejection in 12 patients and three patients with BKV-DNAuria and rejection. The distribution of individual inflammatory processes and their overlaps are shown in Figure 2A.

The 64 rejection phenotypes were classified as follows in descending order of frequency: borderline changes (n = 15, 3.7%), microvascular inflammation (MVI) without HLA-DSA and negative C4d (n = 12, 3.0%), active antibody-mediated rejection (AMR) (n = 11, 2.7%), chronic active AMR (n = 8, 2.0%), acute T-cell mediated rejection (TCMR) ≥ Banff IA (n = 7, 1.7%), C4d staining without evidence of rejection (n = 5, 1.2%), mixed rejection (n = 5, 1.2%), and chronic active TCMR (n = 1, 0.2%).

The lowest urinary CXCL10/creatinine levels were observed in patients without any inflammatory processes (median CXCL10/creatinine 1.8 ng/mmol [1.1–3.2]). Urinary CXCL10/creatinine levels were significantly higher in all inflammatory processes compared to the group without any inflammation (all p < 0.0001; Figure 2B). The highest values were observed in patients with a combination of inflammatory processes (median CXCL10/creatinine 15.3 ng/mmol [7.1–39.7]). Interestingly, patients with isolated subclinical rejection had the lowest CXCL10/creatinine among those with any individual inflammatory processes (median CXCL10/creatinine 4.2 ng/mmol [1.7–6.5]). No correlation was found between C reactive protein (CRP) and urinary CXCL10/creatinine levels (r² < 0.001, p = 0.91; data not shown). CXCL10/creatinine levels were not different between the eleven patients with de novo HLA-DSA at 1-year posttransplant and those without (3.6 ng/mmol [1.1–4.6] vs. 2.4 ng/mmol [1.2–5.0]; p = 0.83).

Of the cohort, 124 patients (31%) had at least one inflammation process. The ROC-analysis revealed an AUC of 0.786 (95%-CI: 0.727–0.836) for CXCL10/creatinine (p < 0.0001) to detect any inflammation process. In contrast, none of the standard laboratory parameters had a statistically significant AUC: 0.499 (95%-CI: 0.436–0.562) for serum creatinine (p = 0.68), 0.561 (95%-CI: 0.498–0.621) for eGFR (p = 0.07), 0.636 (95%-CI: 0.575–0.692) for protein/creatinine (p = 0.05), 0.615 (95%-CI: 0.555–0.672) for albumin/creatinine (p = 0.18), and 0.553 (95%-CI: 0.491–0.613) for α1-microglobulin/creatinine (p = 0.26). The AUC for CXCL10/creatinine was significantly higher than for all other parameters (all p < 0.0001; Figure 3A). Next, we performed the same analyses for specific subgroups. The results were almost identical, if 47 cases with leukocyturia were excluded (AUC = 0.783 [95%-CI: 0.713–0.840]; Figure 3B). For detection of isolated relevant BKV-DNAuria (n = 18), the AUC was 0.936 (95%-CI: 0.886–0.965) and significantly higher than all other parameters (all p < 0.0001; Figure 3C). For detection of isolated subclinical rejection (n = 47), the AUC was lower at 0.703 (95%-CI: 0.607–0.783) but still significantly higher than all other parameters (all p < 0.006; Figure 3D). The AUC for isolated relevant JCV-DNAuria (n = 8) and isolated significant leukocyturia (n = 33) were 0.808 (95%-CI: 0.585–0.926) and 0.736 (95%-CI: 0.597–0.840), respectively (data not shown). In a subgroup analysis of immunological standard-risk (n = 315) and high-risk patients (n = 90), we found almost identical results for the standard-risk group (AUC = 0.804 [95%-CI: 0.736–0.857]; p < 0.0001 vs. all other parameters), while the AUC was lower in the high-risk group (AUC = 0.737 [95%-CI: 0.604–0.837]; p < 0.05 vs. all parameters except protein/creatinine ratio) (Supplementary Figure 1).

The details of the diagnostic characteristics of urinary CXCL10/creatinine to detect inflammatory processes at various cutoff levels are summarized in Table 2. The optimal cutoff according to the Youden-index was 3.97 ng/mmol with a sensitivity of 69% and a specificity of 84%. The negative likelihood ratio (LR-) was 0.36, the positive likelihood ratio (LR+) was 4.34. Next, we defined two arbitrary cutoffs to rule-in or rule-out inflammatory processes in an efficient way. At a low rule-out cutoff of 3 ng/mmol with a LR- of 0.36, 51% of uninformative surveillance biopsies could be avoided while maintaining a sensitivity of 73% for detecting inflammatory processes. Importantly, 33/124 (27%) inflammatory processes would be missed, among which subclinical rejection was the most frequent (n = 20). The most commonly missed rejection phenotypes were borderline changes (n = 5) and various forms of AMR/MVI (see footnotes in Table 2). No relevant BKV-DNAuria ere missed, and only 2/9 cases with relevant JCV-DNAuria were missed. At a high rule-in cutoff of 10 ng/mmol with a LR+ of 12.92, only 2% had no inflammatory process. The same analyses were also performed excluding patients with relevant leukocyturia (n = 47) and are summarized in Table 3. The results were very similar to the whole cohort.

Finally, we investigated independently associated parameters with subclinical rejection in a multivariable model. Urinary CXCL10/creatinine, proteinuria, eGFR, patient age, immunological high-risk (i.e., pretransplant HLA-DSA and/or ABO incompatible transplant), leukocyturia, and de novo HLA-DSA had p-values <0.10 in the univariate analysis. In the multivariable analysis, only urinary CXCL10/creatinine (odds ratio 1.03 [95% CI: 1.01–1.05]; p = 0.0008) and immunological high-risk (odds ratio 2.14 [95% CI: 1.16–3.96]; p = 0.02) were independently associated with subclinical rejection (Table 4). Forcing relevant BKV-DNAuria and JCV-DNAuria as known confounders for urinary CXCL10/creatinine levels into the multivariate model did not change the results. Urinary CXCL10/creatinine (odds ratio 1.03 [95% CI: 1.01–1.06]; p = 0.0006) and immunological high-risk (odds ratio 2.14 [95% CI: 1.16–3.97]; p = 0.02) were still the only significantly associated parameters with subclinical rejection, while relevant BKV-DNAuria (odds ratio 0.58 [95% CI: 0.15–2.28]; p = 0.41) and JCV-DNAuria (odds ratio 0.80 [95% CI: 0.10–6.69]; p = 0.83) were not.