Torque teno virus viral load and hospitalisation for infections in long-term kidney transplant recipients: a single-centre prospective cohort study

Abstract

Background

Torque teno virus (TTV) viral loads may reflect patients’ net state of immunosuppression and identify patients at risk of infections and rejection early after solid organ transplantation. However, its role in long-term kidney recipients (KTRs) is less clear.

Methods

This single-centre prospective cohort study evaluated the association between TTV viral load at recruitment and severe infections requiring hospitalisation within 12 months as the primary outcome in kidney recipients > 1-year post-transplant and on stable immunosuppression for > 3 months. Participants were followed up for 12 months, until change of immunosuppression, graft loss, or death. Pre-specified, exploratory secondary outcomes, including opportunistic infections, malignancy, composite outcomes were also analysed.

Results

Median time after transplant amongst the 171 included participants was 10.3 years (interquartile range (IQR) 4.9–16.0). Thirty-one developed severe infections 105 days (IQR 65–204) after baseline TTV measurement. Higher baseline log10-transformed TTV viral load (logTTV) was associated with severe infections within 12 months (odds ratio (OR) 1.37, 95% confidence interval (CI) 1.01–1.86, p = 0.045) with a modest area under receiver operating characteristics curve of 0.60 (95% CI 0.50–0.71). Compared to logTTV < 4.6 copies/mL, logTTV > 6.6 copies/mL (hazard ratio (HR) 3.45, 95% CI 1.10–10.9, p = 0.04) was associated with time to hospitalisations for infections but not logTTV 4.6–6.6 copies/mL (HR 2.30, 95% CI 0.97–5.47, p = 0.06). The ORs for the association between logTTV and malignancy, opportunistic infection and biopsy-proven rejection were 1.86 (95% CI 1.09–3.16), 1.21 (95% CI 0.69–2.13) and 1.09 (95% CI 0.70–1.69) respectively.

Conclusion

Elevated TTV viral loads were associated with severe infections amongst our cohort of long-term KTRs and may be a useful biomarker in this population.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12882-026-05038-1.

Introduction

Torque teno virus (TTV) has been investigated as a potential biomarker for assessing the net state of immunosuppression in solid organ transplantation [1–5]. Elevated TTV viral loads have been shown to identify kidney transplant recipients (KTRs) at risk of infections, particularly within the first-year post-transplant, while low TTV levels have been linked to rejection [1, 6–14]. In addition, changes in viral loads over time has also been shown to correlate with infections and rejection [7, 14, 15]. Moreover, TTV loads are not affected by conventional antiviral therapies and remain relatively stable under steady immunosuppressive exposure [10, 14, 16, 17]. Therefore, while TTV is not a diagnostic tool, it may be useful to guide optimisation of immunosuppression dosing to balance between preventing rejection from inadequate immunosuppression and avoiding immunosuppression-related adverse events (iRAE) such as infection and malignancy from over-immunosuppression [18–23].

Compared with TTV, other tools proposed for immune monitoring in kidney transplant recipients (KTRs) may have important limitations [24, 25]. Therapeutic drug monitoring of immunosuppressants is routinely used in the clinical setting but may not precisely reflect the overall immunological effect in individual patients. Protocol biopsies can detect subclinical rejection, particularly in high-risk patients, but are invasive and can lead to procedure-related complications [26–28].

Numerous biomarkers are under investigation to detect subclinical rejection or to identify patients at risk of rejection [29]. Development of de novo donor-specific antibodies (DSAs) may identify at-risk patients but represents a significant immunological event that is associated with graft failure. Blood donor-derived cell-free DNA (ddcfDNA) reflects graft injury, while urinary chemokines, such as C-X-C motif ligand 9 and 10 (CXCL9 and CXCL10) indicate kidney inflammation. Importantly, they may not reliably differentiate rejection from other conditions that cause graft injury [29–31]. Blood-based gene expression profile assays, which measure the expression of a panel of rejection-associated genes in peripheral blood, may be more specific [29, 32–34]. However, evidence supporting their ability to predict rejection remains limited and further validation in larger, multi-centre studies is required [29, 35]. Therefore, there remains a need for non-invasive biomarkers can assess the adequacy of immunosuppression, predict future adverse events before alloimmune activation and prevent rejection and adverse graft outcomes [4, 36].

Multiple approaches have also been evaluated to identify patients at risk of post-transplant infections. Measuring routine leukocyte counts, peripheral lymphocyte subpopulation counts, complement levels or immunoglobulin concentrations assess only isolated aspects of the immune system and generally lack the sensitivity and specificity to predict clinical events [37–40]. Clinical composite scoring systems incorporating demographic and laboratory values have also been developed [41, 42]. However, these models are typically applicable only to the specific post-transplant time points at which they were derived and have not been externally validated for widespread use. More recently, virus-specific cell immunity assays are proposed to predict viral infections and guide decisions regarding viral prophylaxis [43, 44]. However, these assays are technically complex, require assay standardisation and may need further external validation before widespread adoption [35, 45].

Importantly, most of the strategies outlined have not been developed to predict both rejection and iRAEs. A cellular functional assay that measured intracellular adenosine triphosphate levels following T cell stimulation was previously proposed to predict both infections and rejection, but its utility has not consistently shown in studies [46].

While TTV has been studied in KTRs beyond the immediate post-transplant period, they have primarily been limited to the use of stored blood samples from registry-based or previous prospective studies [15, 17, 47, 48]. These studies may be subject to bias from the inclusion and exclusion criteria of the original cohort and the dependence on the pre-defined or diagnostic code-based outcomes from the parent study. The accuracy of TTV DNA quantification may be uncertain due to the use of stored samples that may be insufficient, stored for a prolonged duration or were potentially subjected the multiple freeze-thaws cycles [49]. Despite these limitations, TTV loads have been found to be associated with vaccine responses, infections, rejection, malignancy, graft loss and mortality [15, 17, 47, 48].

We hypothesised that TTV DNA loads are associated with the risk of severe infections and iRAEs in long-term kidney transplant recipients. Therefore, we aimed to determine the association between plasma TTV DNA loads taken at recruitment and severe infections, defined as any infection requiring hospitalisation, within 1 year from TTV measurement in our KTRs beyond the first year post-transplant in a prospective, single-centre cohort study.

Materials and methods

Study design and population

A summary of the study design is described here, while the detailed study protocol has been published elsewhere [2]. This was a single-centre, prospective cohort study conducted at a tertiary transplant centre, Singapore General Hospital. Adult KTRs (21 years old or older) were screened between July 2023 and February 2024 for eligibility. Patients were eligible if they were more than 1-year post-transplant on stable doses of immunosuppression for more than 3 months. Participants who required titration of immunosuppression (e.g. for rejection or infection) less than 3 months prior or any clinical event which may significantly influence the participant’s net state of immunosuppression (e.g. hospitalisations, uncontrolled infections, newly diagnosed malignancy) were excluded.

Clinical data collection and laboratory measurements

Baseline clinical data including demographics, co-morbidities and transplant data were collected from the institutional database upon recruitment. Results of routine laboratory investigations performed at the same time including white cell count, absolute neutrophil count, absolute lymphocyte count, immunosuppression drug level, liver panel, serum creatinine and estimated glomerular filtration rate (by the 2009 Chronic Kidney Disease Epidemiology Collaboration, CKD-EPI equation), urine protein-to-creatinine ratio were collected. Cytomegalovirus (CMV) infection was defined, at our centre, as having 2 samples of CMV DNAemia > 137 copies/mL or any detectable CMV pp65 antigenaemia at least 1 week apart and/or histological evidence of CMV disease [50, 51]. BK polyomavirus infection was defined as having 2 samples of any detectable BK polyomavirus DNAemia at least 2 weeks apart and/or histological evidence of BK polyomavirus-associated nephropathy [52, 53]. Participants were not routinely screened for infections including CMV or BK polyomavirus virus and testing was performed based on clinical indications.

Plasma TTV viral loads were measured from venous blood samples collected upon recruitment. In brief, plasma was separated by centrifugation for 10 min at 3,000 rpm within 24 h of collection and preserved at − 30 °C. Genomic DNA was subsequently extracted from 200µL of plasma samples using the NUCLISENS^® easyMAG^® platform (bioMérieux), according to the manufacturer’s protocol. TTV DNA loads (in copy numbers per mL) were quantified by real-time PCR assay (TTV R‐gene^®, bioMérieux) using the CFX96 Real-time System (Bio-Rad, Hercules, CA) with a standard curve derived from known quantitative copy numbers. Samples that had detectable TTV DNA viral load but below the lower limit of quantification of 250 copies/mL were set to one-half the lower limit of the range of quantification (i.e. 125 copies/mL or logTTV 2.10 copies/mL) for analysis, while participants with undetectable baseline TTV DNA viral loads were excluded from analysis [15].

Additionally, peripheral lymphocyte subpopulations for CD4, CD8, CD19 and CD56 were quantified via flow cytometry (BD Multitest™, FACSCanto™ and FACSCanto™ II) and serum IgG (Tina-quant IgG Gen.2, Roche) and C3 (Tina-quant complement C3c v2, Roche) were measured via immunoturbidimetry (Cobas c502, Roche). Treating physicians were blinded to TTV viral loads, peripheral lymphocyte subpopulations, and serum IgG and C3 results throughout the study.

If a study participant required hospitalisation or a kidney allograft biopsy, the indication for admission or biopsy, diagnoses, together with the laboratory values upon admission, including full blood count, serum creatinine and estimated glomerular filtration rate (eGFR), urine protein-to-creatinine ratio, immunosuppression drug level were recorded. Additional venous blood samples were collected for measurement TTV DNA viral loads as described above. Donor-specific antibodies were measured when participants require kidney allograft biopsies (LABScreen Single Antigen, One Lambda).

All haematological, biochemistry and molecular laboratory investigations were performed in central laboratories which were accredited by the College of American Pathologists. Donor-specific antibodies assessments were performed in a central laboratory accredited by the American Society for Histocompatibility and Immunogenetics.

Outcome measures

The primary and secondary endpoints were pre-specified [2]. The primary outcome was severe infections defined as any infection requiring hospitalisation [42]. Secondary outcomes included (1) opportunistic infections (OIs) [24], (2) malignancy, (3) biopsy-proven calcineurin inhibitor nephrotoxicity (CNIn), (4) biopsy-proven rejection (BPAR), (5) biopsy-proven glomerulonephritis, (6) graft loss and (7) mortality. Composite outcomes including immunosuppression-related adverse events (iRAE), defined as a combination of any infection requiring hospitalisation, OIs, malignancy, and CNIn, and immune-mediated adverse events, defined as a combination of secondary endpoints BPAR and biopsy-proven glomerulonephritis, were analysed [41].

Outcomes, including diagnoses after hospitalisations and biopsies, were determined by consensus after independent review by 2 study members after reviewing case report forms, the institutional database and electronic medical records of the study participants and resolved after discussion with a third study member if consensus could not be reached. All kidney allograft biopsies were reported by 3 experienced pathologists within the institution according to the Banff 2019 Kidney Meeting Report [54]. Diagnoses of malignancy were based on histology on tissue biopsy.

All participants were followed up for 12 months, or until change in immunosuppression, graft loss (defined as return to dialysis or re-transplantation), or death, whichever occurred earlier.

Sample size calculation and statistical analysis

The calculated sample size was 172 participants (25 in hospitalisation and 147 in non-hospitalisation group) based on the following parameters: expected area under receiver operating characteristics curve (AUROC) of 0.75 with a margin of error of ± 0.125 (i.e. expected AUROC would lie between 0.627 and 0.873 with a two-sided 95% confidence interval), 1:6 allocation ratio and 10% loss to follow-up. The 1:6 ratio was derived from review of our internal database, which showed that 15.7% of patients more than one-year post-transplant were hospitalized for infection within a 12-month period. This estimate is similar to infection-related hospitalisation rates reported in previous studies [42, 55, 56].

Results were expressed as mean and standard deviation for parametric continuous data, median and interquartile range (25th, 75th percentile) for non-parametric continuous data and as frequency and percentage for categorical data. TTV viral loads were log10-transformed (logTTV) and categorized into < 4.6, 4.6–6.6 and > 6.6 copies/mL as previously described [1, 2].

T-tests and one-way ANOVA were used to compare parametric continuous variables, Mann-Whitney U and Kruskal-Wallis tests for non-parametric continuous variables, and Pearson Chi-square and Fisher’s exact tests for categorical variables.

Logistic regression analysis was used to investigate the association between baseline TTV viral load and hospitalisation due to infections within 12 months of follow-up. Receiver operating characteristics (ROC) curves were plotted and the area under ROC curves (AUROC) calculated to assess the discriminatory value of baseline TTV viral loads for infection-related hospitalisation. The optimal cut-off value (Youden’s index) to identify patients at risk of hospitalisations due to infections was calculated balancing sensitivity and specificity (sensitivity + specificity − 1) [57]. Cox proportional hazards model was used to analyse factors associated with time from blood sample collection for baseline TTV viral load to the first hospitalisation due to infections. Multivariable analyses were conducted by initially including variables with p values < 0.10 from univariable analyses, followed by backward selection.

Participants who have been withdrawn or lost to follow-up were censored at their last known observation. All missing values were handled by exclusion from relevant analysis without imputation. A 2-tailed p value < 0.05 was considered statistically significant, and 95% confidence intervals (CIs) were reported when appropriate. Analyses of secondary endpoints were pre-specified, exploratory, and reported as effect estimates with corresponding 95% CIs. All statistical analyses were performed using StataBE version 17 (StataCorp LLC, USA).

Results

Baseline characteristics of included participants

Of 177 patients screened, 2 were not eligible, 3 withdrew consent before baseline blood collection and 1 was excluded due to undetectable TTV DNA levels. The baseline characteristics of the 171 included participants were summarized in Table 1. The median age of the participants was 57 years (interquartile range (IQR) 47–64). The most common primary kidney disease was glomerulonephritis (74.9%). The median time after transplant was 10.3 years (IQR 4.9–16.0). Most participants were maintained on calcineurin inhibitor-based triple immunosuppression – prednisolone, mycophenolate and tacrolimus (50.9%) or prednisolone, mycophenolate, and cyclosporine (21.6%). Median logTTV at baseline was 4.84 copies/mL (IQR 4.13–5.54, range 2.10–8.83) – 6 samples had detectable TTV DNA levels but below the lower limit of quantification.

Table 1.

Baseline characteristics of included participants

Age (IQR), years	Total (n = 171)	No hospitalisation for infection (n = 140)	Hospitalisation for infection (n = 31)
	57 (47–64)	56 (47–64)	60 (48–68)
Female, n (%)	77 (45.0%)	61 (43.6%)	16 (51.6%)
Ethnicity, n (%)
Chinese Malay Indian Others	135 (79.0%) 18 (10.5%) 8 (4.7%) 10 (5.9%)	115 (82.1%) 12 (8.6%) 7 (5.0%) 6 (4.3%)	20 (64.5%) 6 (19.4%) 1 (3.2%) 4 (12.9%)
History of diabetes mellitus, n (%)	48 (28.1%)	38 (27.1%)	10 (32.3%)
History of ischaemic heart disease, n (%)	21 (12.3%)	17 (12.1%)	4 (12.9%)
History of stroke, n (%)	4 (2.3%)	2 (1.4%)	2 (6.5%)
History of liver cirrhosis, n (%)	1 (0.6%)	0 (0%)	1 (3.2%)
History of malignancy, n (%)	14 (8.2%)	12 (8.6%)	2 (6.5%)
History of autoimmune disease, n (%)	4 (2.3%)	2 (1.4%)	2 (6.5%)
BMI (IQR), kg/m2	24.1 (21.2–27.8)	23.9 (21.0-27.5)	25.7 (23.0-28.6)
Primary kidney disease, n (%)
Glomerulonephritis Diabetes Hypertension ADPKD Others/Unknown	128 (74.9%) 18 (10.5%) 5 (2.9%) 7 (4.1%) 13 (7.6%)	106 (75.7%) 15 (10.7%) 4 (2.9%) 4 (2.9%) 11 (7.9%)	22 (71.0%) 3 (9.7%) 1 (3.2%) 3 (9.7%) 2 (6.5%)
Living donor kidney transplant, n (%)	98 (57.3%)	82 (58.6%)	16 (51.6%)
Time after transplant (IQR), years	10.3 (4.9–16.0)	10.6 (4.6–16.3)	10.5 (5.5–14.9)
ABO incompatible, n (%)	12 (7.0%)	11 (7.9%)	1 (3.2%)
HLA mismatch (ABDR) (SD)	3.3 (3–4)	3.3 (1.5)	3.3 (1.7)
CDC crossmatch positive at time of transplant, n (%)	11 (6.4%)	9 (6.4%)	2 (6.5%)
Flow crossmatch positive at time of transplant, n (%)	17 (9.9%)	14 (10.0%)	3 (9.7%)
Donor-specific antibody positive at time of transplant, n (%)	23 (13.5%)	21 (15.0%)	2 (6.5%)
Previous rejection, n (%)	38 (22.2%)	30 (21.4%)	8 (25.8%)
Previous CMV infection, n (%)	35 (20.5%)	33 (23.6%)	2 (6.5%)
Previous BK polyomavirus infection, n (%)	18 (10.6%)	13 (9.3%)	5 (16.1%)
Maintenance immunosuppression, n (%)
Pred/MMF/Tac Pred/MMF/CsA Pred/Tac Pred/CsA Pred/Aza/Tac Pred/Aza/CsA Others	87 (50.9%) 37 (21.6%) 10 (5.9%) 10 (5.9%) 5 (2.9%) 5 (2.9%) 17 (9.9%)	69 (49.3%) 33 (23.6%) 7 (5.0%) 9 (6.4%) 5 (3.6%) 4 (2.9%) 13 (9.3%)	18 (58.1%) 4 (12.9%) 3 (9.7%) 1 (3.2%) 0 (0%) 1 (3.2%) 4 (9.3%)
Baseline eGFR (CKD-EPI), mL/min/1.73m2 (IQR)	53.4 (37.9–69.9)	52.0 (37.3–69.6)	57.8 (44.4–70.1)
Baseline proteinuria, g/day (IQR)	0.22 (0.11–0.49)	0.22 (0.10–0.50)	0.23 (0.12–0.45)
Albumin, g/L (IQR)	41 (39–44)	42 (39–44)	40 (37–42)
WBC, x109/L (IQR)	7.43 (5.98-9.00)	7.46 (6.01–8.98)	6.97 (5.74-10.0)
ANC, x109/L (IQR)	4.62 (3.70–6.16)	4.58 (3.69–5.76)	5.57 (3.74–7.84)
ALC, x109/L (IQR)	1.70 (1.22–2.23)	1.74 (1.31–2.25)	1.29 (1.04–2.19)
CD4 count, cells/µL (IQR)	593 (365–893)	588 (365–867)	740 (367–932)
CD8 count, cells/µL (IQR)	473 (314–694)	455 (312–679)	521 (354–695)
CD19 count, cells/µL (IQR)	65 (31–133)	65 (31–133)	66 (37–124)
CD56 count, cells/µL (IQR)	216 (140–308)	207 (140–305)	237 (140–340)
Serum IgG, g/L (IQR)	10.7 (9.0-12.7)	10.7 (9.1–12.8)	10.3 (8.6–12.2)
Serum C3, g/L (IQR)	1.07 (0.94–1.24)	1.07 (0.96–1.22)	1.07 (0.93–1.29)
Log10-transformed TTV viral load, copies/mL (IQR)	4.84 (4.13–5.54)	4.79 (4.04–5.46)	5.03 (4.62–5.81)
TTV viral load, copies/mL (IQR)	68,475 (13,475 − 343,000)	60,588 (10,981 − 287,625)	107,025 (41,300–648,250)

IQR, interquartile range; BMI, body mass index; ADPKD, autosomal dominant polycystic kidney disease; HLA, human leukocyte antigen; CDC, complement dependent cytotoxicity; CMV, cytomegalovirus; Pred, prednisolone; MMF, mycophenolate; Tac, tacrolimus; CsA, cyclosporine A; Aza, azathioprine; eGFR, estimated glomerular filtration rate; CKD-EPI, Chronic Kidney Disease Epidemiology Collaboration; WBC, white blood cell count; ANC, absolute neutrophil count; ALC, absolute lymphocyte count; C3, complement component 3; TTV, Torque teno virus

Primary outcome – hospitalisation for infections

Of the 171 included participants, 31 were hospitalised due to infections (Table S1) including 1 admission to the intensive care unit due to acute gangrenous cholecystitis, 115 completed 1-year follow-up, 24 changed immunosuppression (12 rejections, 1 concomitant rejection and BK polyomavirus-associated nephropathy, 2 malignancies, 1 IgA nephropathy, 1 thrombotic microangiopathy, 1 CNIn) and 1 demised (aortic dissection). No participants were lost to follow-up. Median time to the first hospitalisation for infection from TTV measurement at recruitment was 105 days (IQR 65–204).

Baseline logTTV, as a continuous variable, was associated with hospitalisation due to infections (odds ratio (OR) 1.37, 95% confidence interval (CI) 1.01–1.86, p = 0.045). ROC analysis, performed to assess the discriminatory value of logTTV to identify patients at risk of infection-related hospitalisation, revealed an AUROC of 0.60 (95% CI 0.50–0.71, Fig. 1). The optimal cut-off value for identifying patients at risk of hospitalisations due to infections in our cohort, balancing sensitivity and specificity (Youden’s index), was logTTV 4.61 copies/mL.

Fig. 1.

Receiver operator characteristics curve of log-transformed Torque teno virus viral load for hospitalisations due to infections

Compared to logTTV < 4.6 copies/mL, logTTV > 6.6 copies/mL (OR 4.64, 95% CI 1.18–18.6, p = 0.03, Table 2) was associated with hospitalisations for infections but not logTTV 4.6–6.6 copies/mL (OR 2.57, 95% CI 1.00-6.61, p = 0.051).

Table 2.

Primary and secondary outcomes of included participants

	Total (n = 171)	Odds ratio for logTTV, copies/mL (95% CI) (vs. logTTV < 4.6, copies/mL)		P value
Primary Outcome
Hospitalisation for infection	31 (18.1%)	4.6–6.6	2.57 (1.00-6.61)	0.051
		> 6.6	4.64 (1.18–18.2)	0.03
Secondary Outcomes
Opportunistic infections	7 (4.1%)	4.6–6.6	1.65 (0.29–9.33)	-
		> 6.6	2.55 (0.21–30.6)	-
Malignancy	5 (2.9%)	4.6–6.6	1.26 (0.20–7.80)	-
		> 6.6	5.27 (0.67–41.5)	-
Biopsy-proven calcineurin inhibitor nephrotoxicity	9 (5.2%)	4.6–6.6	0.60 (0.13–2.80)	-
		> 6.6	2.38 (0.39–14.5)	-
iRAE (hospitalisation for infection, malignancy or calcineurin inhibitor nephrotoxicity)	43 (25.1%)	4.6–6.6	1.47 (0.67–3.24)	-
		> 6.6	4.95 (1.55–15.8)	-
Biopsy-proven acute rejection	13 (7.6%)	4.6–6.6	1.31 (0.41–4.21)	-
		> 6.6	NA	-
Biopsy-proven acute rejection or glomerulonephritis	15 (8.8%)	4.6–6.6	1.48 (0.47–4.66)	-
		> 6.6	0.87 (0.09–8.06)	-

LogTTV, log10-transformed torque teno virus viral load; 95% CI, 95% confidence interval; iRAE, immunosuppression related adverse event; NA, not applicable

Age, proteinuria, serum albumin, absolute lymphocyte counts and logTTV as a continuous and logTTV > 6.6 copies/mL (compared to logTTV < 4.6 copies/mL) were associated with hospitalisations due to infection on univariable analyses (Supplementary Table S2). On multivariable analyses with logTTV, only serum albumin remained associated with hospitalisations due to infections.

For time-to-event analyses, logTTV at baseline was not associated to time to hospitalisation due to infections (hazard ratio (HR) 1.27, 95% CI 0.98–1.64, p = 0.07). Compared to logTTV < 4.6 copies/mL, logTTV > 6.6 copies/mL (HR 3.45, 95% CI 1.10–10.9, p = 0.04) was associated with time to hospitalisations for infections but not logTTV 4.6–6.6 copies/mL (HR 2.30, 95% CI 0.97–5.47, p = 0.06, Fig. 2).

Fig. 2.

Survival curves for hospitalisation due to infections based on log-transformed Torque teno virus viral load categories

Secondary outcomes

Seven participants developed opportunistic infections (OIs) including 4 multi-dermatomal herpes zoster, 1 CMV gastritis, 1 non-tuberculous mycobacterial lung infection with concomitant oesophageal candidiasis and 1 neutropenic sepsis with incidental EBV viraemia with no evidence of post-transplant lymphoproliferative disease after evaluation. Median time to development of OIs from TTV measurement at recruitment was 116 days (IQR 66–156). The OR for the association between baseline logTTV and the development of OIs was 1.21 (95% CI 0.69–2.13). Compared to logTTV < 4.6 copies/mL, the OR for the development of OIs was 1.65 (95% CI 0.29–9.33) for logTTV 4.6–6.6 copies/mL and 2.55 (95% CI 0.21–30.6) for logTTV > 6.6 copies/mL.

A total of 5 participants developed malignancy during follow-up – 2 post-transplant lymphoproliferative disease, 1 lung adenocarcinoma, 1 cholangiocarcinoma and 1 hepatocellular carcinoma. Median time to development of malignancy from baseline TTV measurement was 106 days (IQR 35–111). The OR for the association between logTTV and the development of malignancy was 1.86 (95% CI 1.09–3.16). With logTTV < 4.6 copies/mL as the reference group, OR for the development of malignancy was 1.26 (95% CI 0.20–7.80) for logTTV 4.6–6.6 copies/mL and 5.27 (95% CI 0.67–41.5) for logTTV > 6.6 copies/mL.

A total of 9 participants developed CNIn. Median time to diagnosis of CNIn from TTV measurement at recruitment was 41 days (IQR 8–194). The OR for the association between baseline logTTV and CNIn was OR 1.15 (95% CI 0.70–1.89). Relative to logTTV < 4.6 copies/mL, logTTV 4.6–6.6 copies/mL was associated to CNIn with an OR of 0.60 (95% CI 0.13–2.80). while logTTV > 6.6 copies/mL had an OR of 2.38 (95% CI 0.39–14.5).

Forty-three participants developed immunosuppression-related adverse events (iRAE), a composite endpoint of hospitalisation due to infection, malignancy or CNIn. The OR for the association between baseline logTTV and iRAE was 1.41 (95% CI 1.07–1.86). Compared to logTTV < 4.6 copies/mL, the OR for the development of iRAE was 1.47 (95% CI 0.67–3.24) for logTTV 4.6–6.6 copies/mL and 4.95 (95% CI 1.55–15.8) for logTTV > 6.6 copies/mL.

There were 13 biopsy-proven rejections (BPAR), including 2 borderline rejections, 2 acute T cell mediated rejection (TCMR), 1 concomitant acute TCMR with BK polyomavirus-associated nephropathy, 2 active antibody-mediated rejection (ABMR), 6 chronic active ABMR, during the follow-up period. Median time to diagnosis of BPAR from TTV measurement at recruitment was 29 days (IQR 8 –59). The OR for the association between baseline logTTV and BPAR was 1.09 (95% CI 0.70–1.69). The OR for BPAR was 1.31 (95% CI 0.41–4.21) for logTTV 4.6–6.6 copies/mL compared to logTTV < 4.6 copies/mL. There were no rejection episodes in participants with logTTV > 6.6 copies/mL. LogTTV did not differ between participants who developed rejection and those who did not (4.79 copies/mL, IQR 4.35–5.56 vs. 4.82 copies/mL, IQR 4.04–5.39, p = 0.64).

One patient developed IgA nephropathy, while another developed thrombotic microangiopathy attributable to recurrent lupus nephritis. There was 1 graft loss presumed to be due to chronic rejection after reduction of immunosuppression due to immunosuppression-related adverse effects and 1 death due to aortic dissection.

TTV viral loads based on baseline characteristics and changes in immunosuppression

Participants with diabetes mellitus had higher baseline logTTV levels, while baseline logTTV was inversely associated with baseline eGFR, absolute neutrophil count, absolute lymphocyte count, mTOR inhibitor use and cyclosporine use after multivariable analysis (Supplementary Table S3). Age, sex, and time after transplant were not significantly associated with logTTV on multivariable analysis.

Discussion

Our study demonstrated that high TTV viral loads in long-term kidney transplant recipients were associated with hospitalisations due to infections with an overall modest discriminative performance. While multiple studies have demonstrated the prognostic value of TTV within 1-year post-transplant, studies on long-term KTRs are less common [1, 7, 10, 13]. Our findings corroborate and extend previous studies on long-term kidney transplant recipients, which have primarily analysed stored serum samples from registries or previous studies [17, 47]. The current study adds prospectively collected data on the association of TTV viral loads with infections and other outcomes in a cohort of KTRs at a median of more than 10 years after transplant, a population that remains underrepresented in the existing literature.

In a cohort of 34 KTRs with a mean of 2 years after transplant, Chauvelot et al. showed logTTV > 5.1 copies/mL was associated with a composite endpoint of infection and cancer over a follow-up of 50 months [47]. In another retrospective study of 666 patients with a median 4.9 years post-transplant, logTTV of the highest tertile (median logTTV 5.52 copies/mL) was associated with the risk of all-cause mortality and infection-related mortality over a period of 7 years [17]. LogTTV > 6.6 copies/mL was previously proposed to identify kidney transplant recipients at increased risk of infection within 1 year after transplant. The optimal cut-off in our analysis to identify patients at risk of infection-related hospitalisations over 1 year was lower at logTTV 4.61 copies/mL. Cut-off values derived from different studies may apply to different endpoints and be difficult to compare given differences in patient characteristics and study design and therefore may only apply to patients who are represented by the individual studies. This may also reflect the longer time post-transplant and lower immunosuppression burden in our cohort and suggests that optimal TTV cut-off values may differ amongst different transplant populations. Importantly, TTV viral loads measured using different assays can differ significantly and standardised kits can allow comparability amongst TTV viral loads measured in different laboratories [23, 58, 59].

Similar to previous studies, older age, lower absolute lymphocyte count, lower serum albumin and higher proteinuria were associated with infections after 1 year post-transplant, in addition to TTV [60, 61]. In our cohort, only serum albumin remained independently associated with severe infections after multivariable analysis. This finding should be interpreted cautiously given the modest number of infection events, which limits the statistical power to detect independent effects of multiple biomarkers.

In our cohort, TTV viral loads were found to be associated with the development of malignancy. However, this analysis should similarly be interpreted cautiously, as it is exploratory and based on a small number of events. Previous studies have reported an association between high TTV viral loads and malignancy, both as part of a composite endpoint including infection [14, 47] and independently in a cohort of 625 long-term liver transplant recipients [62]. A high degree of immunosuppression, which can be reflected by high TTV viral loads, has been linked with increased risk of post-transplant malignancy [63, 64]. Conversely, malignancy itself may impair systemic immune function prior to diagnosis, potentially leading to increased TTV viral loads [65]. Therefore, larger prospective and interventional studies are necessary to define the clinical utility of TTV for risk stratification and potential risk reduction of post-transplant malignancy.

Our study did not demonstrate an association between TTV viral loads and biopsy-proven acute rejection. In a previous study of 715 KTRs with a median of 4.4 years after transplant, higher logTTV were independently associated with a reduced risk of antibody-mediated rejection [48]. In another study of 96 KTRs 12 months after transplant, having logTTV < 3.75 copies/mL in > 50% of the samples taken at 12, 24 and 36 months after transplant was associated with appearance of de novo donor-specific antibodies, biopsy-proven rejection or graft loss [47]. Singla also demonstrated in a retrospective analysis of blood samples of the CTOT-8 study of patients within 3 years post-transplant with protocol biopsies, that a single logTTV measurement was associated with subclinical rejection. Additionally, a decreasing trend of logTTV combined with current low logTTV and historical high logTTV was associated with subclinical rejection [15]. Our study was not powered to detect differences in BPAR and the small sample size precluded adjustment for potential confounders. Moreover, our study included stable patients who had been transplanted for a longer duration who may be at lower risk of acute rejection. Serial measurements of TTV viral loads may also be more able to detect deviations of TTV viral loads from a target range and non-compliance to immunosuppression [15].

The current study was conducted in a single centre. Therefore, the findings may not be generalisable to other transplant populations. Baseline characteristics of this cohort, including the high rates of glomerulonephritis and differences in rates of living donor kidney transplant, crossmatch positivity and graft function, likely reflect local disease epidemiology and transplant practices [21, 66–70]. Glomerulonephritis, especially IgA nephropathy, is more prevalent in Singapore and across East Asia, with similar distributions reported in other Asian transplant cohorts [71–74]. In particular, stringent eligibility criteria and prolonged waiting times for deceased donor transplantation introduces a degree of selection bias at the population level [20, 75, 76]. While these differences should be considered when interpreting our findings, their impact on the observed relationship between TTV and clinical outcomes remains uncertain. Close to 10% of the participants of our study were maintained on calcineurin inhibitor-free or mTOR inhibitor-based regimens with reduced calcineurin exposure, for which the performance of TTV is less well-established [4, 77]. Data on length of hospital stay and antibiotics use were not collected. Together with the small sample size and low number of primary outcome events, further assessment based on severity of infection and healthcare utilisation was limited. While the secondary outcomes were pre-specified, event rates were low and the study was not powered to detect differences in these endpoints. Therefore, these analyses remain exploratory and the results need to be interpreted cautiously and validated in larger prospective studies.

The current study is also limited by the lack of systematic repeat TTV sampling due to funding and logistical constraints. The association between TTV levels and outcomes in our study may have been underestimated since the correlation between TTV and the net state of immunosuppression may diminish over time. Regular longitudinal monitoring may uncover temporal trends which can correlate with clinical outcomes and possible drug non-adherence [10, 15, 48]. The optimal frequency for monitoring TTV viral load has been proposed to be every 2–4 months [4].

However, our study investigated the utility of TTV viral loads among long-term kidney transplant recipients, with most of our patients > 10 years post-transplant and no loss to follow-up. While hospitalisations for infections may not precisely correlate with the degree of immunosuppression, it represents a clinically relevant outcome.

The study was observational in nature and therefore unable to draw conclusions regarding specific interventions to reduce the risk of severe infections. Results from the TTV-GUIDE-IT clinical trial will provide useful data on whether TTV can be used for adjustment of immunosuppression in early post-transplant KTRs, while the multi-centre randomised controlled study (TTV-guided mAnagement Of long-term ImmunosuppreSsion in kidney Transplantation, TAOIST) will be useful to clarify the role of TTV-guided immunosuppression in longer term kidney transplant recipients [4, 23, 77].

Conclusion

Elevated TTV viral loads were associated with hospitalisations due to infections and on secondary exploratory analysis may be associated with malignancy amongst long-term kidney transplant recipients. These findings support further investigation into the use of TTV to individualise immunosuppression to mitigate overimmunosuppression in this population.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

95% CI

95% confidence interval

ABMR

Antibody-mediated rejection

AUROC

Area under receiver operating characteristic curve

BPAR

Biopsy-proven acute rejection

C3

Complement component 3

CKD-EPI

Chronic Kidney Disease Epidemiology Collaboration

CNIn

Calcineurin inhibitor nephrotoxicity

ddcfDNA

Donor-derived cell-free DNA

DNA

Deoxyribonucleic acid

DSA

Donor-specific antibodies

eGFR

Estimated glomerular filtration rate

HLA

Human leukocyte antigen

HR

Hazards ratio

IgG

Immunoglobulin G

iRAE

Immunosuppression-related adverse event

KTR

Kidney transplant recipient

LogTTV

Log10-transformed Torque teno virus viral load

OI

Opportunistic infection

OR

Odds ratio

TCMR

T cell-mediated rejection

TTV

Torque teno virus

Author contributions

QYH, conceptualisation; QYH, DCML, ITL, KLL, SJC, ST, CSTien, LLEO, TK methodology; QYH, ITL, CSTien, KLL, WMA, XLJK, ML, JL, RWL, YTY, PHL data collection; QYH, ITL, CSTien, SJC data analysis and interpretation; QYH, DCML, LLEO, CSTan, TK, funding acquisition; LLEO, TK, supervision; QYH, ITL wrote the first draft and all authors participated in revising the manuscript.

Funding

This work was supported by the SingHealth Duke-NUS Academic Medicine Research Grant (AM/SU085/2023, SRDUKAMR2385).

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethical approval

This study was conducted in accordance with The Code of Ethics of the World Medical Association (Declaration of Helsinki) and was approved by the SingHealth Central Institutional Review Board (CIRBE 2023/2170). Written informed consent was obtained from all study participants. In accordance with international ethical standards for transplantation research and national legislation, no organs or tissues used in this study were obtained from prisoners. Organs and tissues procured in Singapore were obtained through legally authorised and ethically regulated channels under Singapore’s organ-donation framework, namely the Human Organ Transplant Act and the Medical (Therapy, Education and Research) Act.

Consent for publication

Not applicable.

Use of AI and AI-assisted technologies in the writing process

The authors used ChatGPT (OpenAI, GPT-5.2; accessed February 2026) to improve grammar and phrasing. The authors reviewed and edited the output as necessary and take full responsibility for the content of the publication.

Competing interests

The authors declare no competing interests.