Transition from Antigenemia to Quantitative Nucleic Acid Ampli cation Testing in Kidney Transplant Recipients Receiving Preemptive Therapy for Cytomegalovirus Infection

Mônica Rika Nakamura Universidade Federal de São Paulo Lúcio R. Requião-Moura (  lucio.requiao@gmail.com ) Universidade Federal de São Paulo Roberto Mayer Gallo Hospital do Rim, Fundação Oswaldo Ramos Camila Botelho Hospital do Rim, Fundação Oswaldo Ramos Júlia Taddeo Hospital do Rim, Fundação Oswaldo Ramos Laila Almeida Viana Hospital do Rim, Fundação Oswaldo Ramos Cláudia Rosso Felipe Universidade Federal de São Paulo José Medina-Pestana Hospital do Rim, Fundação Oswaldo Ramos Hélio Tedesco-Silva Universidade Federal de São Paulo


Introduction
The cytomegalovirus (CMV) infection is one of the most common infectious events after solid organ transplantation, affecting 20 to 60% of kidney transplant recipients 1-3 , increasing morbidity, costs and leading to a possible negative impact on graft survival 3 . The effects of the CMV infection have been traditionally characterized as direct and indirect 4 . Although the indirect effects have been questionable recently, the direct effects, such as symptoms and laboratory changes attributable to CMV and the invasive disease, are still a eld of concern after kidney transplantation 4,5 . Cytomegalovirus replication occurs mainly in the rst three months after the transplant, and the clinical presentation is now well de ned according to international guidelines in infection, disease, and invasive disease 4,6 .
Considering the latent CMV infection is widely detected among candidates for kidney engraftment 7 , the risk of CMV active infection after transplantation should be evaluated, and a strategy to reduce the impact of direct effect has to be adopted 4 . Currently, there are two e cacy and safe alternatives for preventing outcomes related to CMV after transplantation: universal pharmacological prophylaxis or preemptive treatment 8- 10 . Although universal prophylaxis seems to be associated with lower CMV-related effects, some disadvantages have been highlighted: toxicity, late-onset CMV disease, risk of resistance, and costs 4 . In Brazil, for instance, prophylaxis with oral valganciclovir for three months can cost 3 to 7 times more than the preemptive treatment, depending on graft function and frequency of monitoring 11 .
Thus, due to the high cost, the way to reduce the impact of CMV involves targeted prevention through preemptive treatment, especially in low and middle-income countries.
For preemptive treatment, patients must be strictly monitored for CMV replication throughout a laboratory method to detect viral load. For many years, many services performed a semi-quantitative test, an immuno uorescence assay based on monoclonal antibodies that detect the viral antigen, such as the pp65 antigenemia [12][13][14] . However, in the last two decades, it has been replaced by quantitative nucleic acid testing, especially by standardization ultra-sensitivity real-time polymerase chain reaction (RT-PCR), such that it is currently the preferred method for CMV management 4 . In 2017, we started implementing standardized RT-PCR for the preemptive treatment in our center, replacing the antigenemia completely one year later. This change in the clinical routine designed a natural experiment with the potential to measure CMV-related events as outcomes in two different eras. Therefore, in the present study, we aimed to compare the performance of antigenemia transitioned to quantitative nucleic acid ampli cation testing, RT-PCR, in kidney transplant recipients receiving preemptive treatment and to evaluate the potential clinical predictors of the CMV-related events.

Results
Between March 2016 and August 2018, a total of 2,294 kidney transplants were performed in our center. Initially, 905 patients were excluded because they were transplanted in a transition period (from March 2017 to February 2018). In the antigenemia era (March 2016 to March 2017), 932 patients had been transplanted; however, 488 recipients did not present inclusion criteria, and 267 had exclusion criteria, as depicted in gure 1. On the other hand, in the RT-PCR era (February 2018 to August 2018), 457 patients had been transplanted; however, 130 recipients did not present inclusion criteria, and 141 had exclusion criteria ( gure 1). Among patients excluded due to death or graft loss (n= 65), 38 had died or had graft loss within 60 days of transplantation. All 17 deaths were veri ed, and no one was attributed to the CMV event. Therefore, 177 patients were enrolled for the antigenemia era, whereas 186 were for the RT-PCR era.
Demography data according to testing era: antigenemia and RT-PCR Demographic data are shown in Table 1. Patients were 49.0 years old, 54.8% males and 54.8% whites.
The etiology for chronic kidney disease was unknown for 44.6%, and 93.7% had undergone hemodialysis as a renal replacement treatment before transplantation; only 12.7% of patients had been submitted to a retransplantation. Donors were 52.0 years old, 52.9% male, and 52.1% white. Most transplants were performed with a deceased donor (96.1%), whose median KDPI value was 80.0. The cold ischemia time was 23.1 hours, and the delayed graft function (DGF) occurred in 187 patients (51.5%). The demographic data were compared between the era (Table 1). There is no difference in the recipients age, however in the antigenemia era, they were less frequently white (46.  Only one patient had an invasive disease. The length of treatment was 22.0 days. One-year cumulative incidence of the rst CMV-related event was not different according to the era: 50.8% in the antigenemia era vs. 44.1% in the PCR, P= 0.20 ( Figure 2.A). The time between transplantation and the rst event was not different too: 47.0 (37.5; 60.2) vs. 47.0 (36.7; 64.5) days, respectively, P=0.93. There also was no difference regarding the cumulative incidence of CMV disease (23.7% vs. 19.1%, respectively, P=0.41, Figure 2.B). In the rst era, the median antigenemia when the treatment was started was 18 cells, and 20 patients (21.5%) had to be treated with less than 10 cells. In the second era, the viral load when the treatment was started was 7,093 UI/mL (5,247; 12,327), while 19 patients (22.4%) had to be treated with less than 5,000 UI/mL. Furthermore, the length of treatment was longer in the RT-PCR (  One-year cumulative incidence of acute rejection (Figure 2.C) was 12.4% in the antigenemia era and 16.1% in the RT-PCR (P=0.35). Figure 2.D shows the eGFR over the follow-up time according to both eras.
Owing to the difference in the DGF incidence, the eGFR was lower in the rst era from the baseline (day 21) to day 42 (antigenemia and RT-PCR, respectively; values expressed in mL/min/

Discussion
Despite improved kidney transplantation clinical management, the CMV infection is still a concern 4 . According to the best clinical guidelines, both strategies available for preventing the consequences of CMV infection, universal prophylaxis or preemptive treatment, present advantages and some disadvantages, and centers should opt for one or another, considering their characteristics 4 . For example, in Brazil, more than 90% of kidney transplantation is supported by the public health system, and the costs of universal prophylaxis are not disbursed, which can occur in other low and mid-income countries 11 . In the present study, we compared the main CMV-related outcomes when we transitioned from pp65 to RT-PCR in kidney transplant recipients receiving preemptive treatment as a strategy to prevent CMV infection and identi ed potential clinical predictors of the CMV-related events.
In our primary hypothesis, we were expecting to detect a reduction in the rate of symptomatic patients, considering that the RT-PCR has high sensitivity to detect low viral load 15,16 . Moreover, antigenemia presents several limitations in CMV treatment, highlighted in the updated international consensus in 2013 as the lack of standardization, the dependence of the subjective interpretation, and its performance when the count of neutrophils is low 14 . Consequently, since that, quantitative nucleic acid ampli cation testing has been established as the "cornerstone for diagnosis and monitoring for CMV infection and disease" 4,14 . Indeed, using a threshold of 5,000 UI/mL to start the preemptive treatment in the RT-PCR era, the frequency of treatment in our cohort was not different from those observed with pp65 antigenemia, and the time to treatment onset was precisely the same. Additionally, we did not nd the expected reduction in the rate of patients who had CMV disease.
When the treatment is started, the viral load seems to be associated with the clinical resolution 17 . In an exploratory analysis from the VICTOR study, where plasma samples of 267 participants were retested, and the viral load was calibrated based on the CMV World Health Organization, the faster resolution of CMV disease after treatment with valganciclovir was 57% more likely when the initial viral load was lower than 18,200 UI/mL 17 . Different from the VICTOR study, in our cohort, all patients were conducted under the preemptive treatment with closer viral load screening; therefore, the main target of the clinical management was to avoid the symptomatic infection. After that, the median of viral load was 7,093 UI/mL, and 75% of patients had a viral load lower than 12,327 UI/mL. Last, despite the low frequency of invasive disease, we consider that the rate of symptomatic infection was higher than we expected when we transitioned from antigenemia to RT-PCR.
Although the local clinical approach in the rst era had preconized one-week extension in the treatment after the last negative antigenemia, the duration of treatment to reach a viral load suppression was longer with RT-PCR. There was an initial concern that highly sensitive assays to manage the CMV infection resulted in prolonged treatments and unnecessary exposure to antiviral therapy 14 . However, a shorter time of treatment using standardized quantitative nucleic acid testing has been demonstrated in a previous study 17,18 . Additionally, as a direct consequence of a more prolonged treatment time with RT-PCR observed in our study, we expected that the rate of the retreating requirement was reduced, considering that reaching a virological suppression seems to be predictive of clinical response 17 . Indeed, the need for retreatment was slightly lower in the PCR era (11.8% vs. 14.7%); however, this difference was not signi cant. Taking these results together and the indirect evidence gured in the present study, it is possible to speculate that a cut-off lower than 5,000 UI/mL could reach lower rates of symptomatic patients than we observed. On the other hand, it seems to be that a cut-off of 10 positive cells associated with a seven-day extended treatment would be equivalent to the viral load suppression achieved by treatment guided for PCR, and this nd can be helpful for centers that have only antigenemia as the option to conduct the preemptive treatment.
In a secondary analysis, we sought predictors of CMV-related events in patients receiving preemptive treatment. More recently, the quanti cation of the T-cell-speci c response against CMV antigens has been considered a promisor tool for predicting the CMV-related events 19 , and it would be helpful for preemptive treatment optimization 20 . However, its use is not standardized for wieldy clinical use. Here, two clinical predictors were associated with the probability of the rst CMV-related event: early acute rejection and 30day graft function.
The association between acute rejection and CMV replication is mainly supported by immunosuppression intensi cation to treat the immunological event 21 . Therefore, we included the acute rejection within 30 days as an independent variable in the multivariable model. Of note, early acute rejection was also an independent predictor for CMV disease, which supports the correlation between the intensity of immunodepression and the spectrum of infection. Furthermore, the association of early graft function and CMV-related events has been previously reported 22 , which was con rmed by our group in an independent cohort of 938 patients transplanted between 2014 and 2015, where the 30-day eGFR was a strong predictor of CMV infection or disease: OR for each mL/min/1.73m 2 = 0.98; 95% CI, 0.97-0.99 23 . How could these three variables converge to the likelihood of the CMV-related events? It is known that some cytokines expressed in tissue damage, especially in the ischemia and reperfusion injury, and acute rejection (TNFα, IL-6, and IL-1β), promote CMV replication 24 . Nevertheless, the initial high exposure to tacrolimus and the levels of mycophenolate in uenced by the immunological compatibility could increase the risk of CMV replication.
Our study has several limitations. First, a historical study, carried out in a single-center, with groups followed in two different eras, is associated with some biases. Second, in both periods, the thresholds for starting the preemptive treatment were de ned by the clinical routines due to the lack of robust evidence to support a prespeci ed cut-off. Third, some differences in the baseline characteristics were observed when both eras were compared, mainly in the donor's demography, which could be associated with worse graft function 30 days after transplantation, although the model to evaluate predictors of CMV-related events has been adjusted for eras. Last, the adherence to the local approach was not directly measured.
In conclusion, in the present study, we did not observe a reduction in the frequency and in the time for CMV-related events, as well as in the requirement for retreatments when the antigenemia was transitioned to quantitative nucleic acid ampli cation testing, using a threshold of 5,000 UI/mL in the standardization RT-PCR for starting the preemptive treatment in kidney transplant recipients. Finally, we de ned 30-day graft function and early acute rejection as clinical predictors of CMV replication after transplantation.

Study Design and population
This was a retrospective sequential single-center cohort study carried out at Hospital do Rim -São Paulo, Brazil. Considering that the study was aimed to evaluate a transition in methods chosen to assess CMV viremia in the preemptive treatment (antigenemia or PCR), patients were grouped in two different eras: the use of antigenemia in the rst era and the use of RT-PCR in the second one. The study was conducted following the Declaration of Helsinki and was approved by the Ethics Committee at Federal University of São Paulo (identi cation number CAEE 05677618.6.0000.5505, and approval number 3.164.538). Being a retrospective study, the informed consent form was waived by the Ethics Committee at Federal University of São Paulo.
The eligible participants were kidney transplant recipients who underwent kidney transplants between March 2016 and August 2018, under preemptive treatment for the risk reduction of CMV disease, and who completed one year of follow-up. Other inclusion criteria were: age at the transplantation time older than 18 years, immunological induction with thymoglobulin, and the immunosuppression regime of maintenance based on tacrolimus and mycophenolate, owing to it was the main indication for the preemptive treatment according to the local approach. In addition, recipients of kidney transplants combined with another solid organ or negative CMV serology were excluded. The time for transition from antigenemia to RT-OCR was from March 2017 to February 2018, when patients started treatment based on antigenemia, but the antiviral was usually interrupted based on the result of PCR. Therefore, patients transplanted in this period were excluded too.

Immunosuppression and prophylaxis
All patients received a single dose of 3.0 mg/kg of Thymoglobulin as an induction strategy, following the local practice, which was previously published 25 . The maintenance immunosuppression regime consisted of a combination of tacrolimus, prednisone, and acid mycophenolate. The dose of tacrolimus was adjusted to maintain C 0 levels between 5-15 ng/mL. In addition, all patients underwent prophylaxis with albendazole for parasitic infections and sulfamethoxazole-trimethoprim for Pneumocystis jirovecii.

Monitoring and treatment of CMV infection
For the preemptive treatment, viremia was collected every two weeks from the 21st after transplantation. The CMV tests results were available one day after sample collection. When patients presented the preemptive treatment criteria, the antiviral was started two or three days after the sample collection. For pp65 antigenemia, after peripheral blood extraction, leukocytes were incubated with C10/C11 antibodies and other reagents from the CMV Brite Turbo kit (IQ® Products, Groningen, Netherlands). The presence of pp65 antigen was detected by a homogeneous yellow-green nuclear pattern in a uorescence microscope, and the nal result was expressed by the number of positive cells per 200,000 leukocytes 26 . The RT-PCR was performed with a commercial Abbott RealTime CMV kit. The DNA extraction, ampli cation, and detection were performed in the automated Real-Time m2000 system (Abbott Molecular Inc), having the DNA sequences of the UL34 and UL80.5 CMV genes as targets 15 . The procedure consisted of a real-time ampli cation reaction on a microplate, with programmable temperature control and variation, and simultaneously an optical uorescence detection system with the reaction in a thermocycler 16,27 . The reported limits of detection and quanti cation were 31.2 IU/mL.
In the antigenemia era, the preemptive treatment was indicated in the presence of 10 or more positive cells in asymptomatic patients or in patients who presented symptoms attributable to CMV infection, independent of the number of positive cells. In the RT-PCR era, the preemptive treatment was indicated in the presence of 5.000 UI/mL or more in asymptomatic patients or in patients who presented symptoms attributable to CMV infection, independent of viral load in the RT-PCR. The treatment consisted of intravenous ganciclovir 5 mg/kg twice a day, adjusted for renal function. During treatment, monitoring was carried out weekly. For the antigenemia era, the treatment was extended for seven following days from the rst negative result. On the other hand, in the RT-PCR era, the treatment was interrupted when the result was undetectable (>31 UI/mL) 28 . Monitoring after the treatment interruption was maintained over the following three months.

De nitions
Cytomegalovirus infection was classi ed according to the Third International Cytomegalovirus Consensus as CMV infection, CMV disease, and invasive disease 4 : infection was de ned by the evidence of viral replication in the absence of symptoms attributable to the viral activity, whereas disease was determined by evidence of CMV replication, associated with attributable symptoms or laboratory abnormalities, and the invasive disease was de ned by the presence of the virus in the histological analysis of any tissue regardless of the result of the viremia or by retinitis, meningitis, or encephalitis.
Recurrences of CMV infection or disease were de ned by the need for a new treatment after the complete remission of the previous episode. Delayed graft function was de ned by the need for dialysis during the rst week, and acute rejection (AR) as treated rejections, proven by biopsy or not, according to Banff's classi cation 29 . The estimated glomerular ltration rate (eGFR) was calculated by the CKD-EPI equation 30 .

Outcomes
Page 12/17 The primary outcome was a CMV-related event de ned as infection or CMV disease, which occurs rst. Additionally, the incidence of disease, the time for detecting events, the length of treatment with ganciclovir, and the frequency of retreatment requirement were compared according to the era. The incidence of acute rejection and 1-year graft function were secondarily evaluated.

Statistical analysis
Continuous variables are summarized as the median and interquartile range (1st; 3rd) and compared by test U of Mann-Whitney, and categorical variables are summarized as absolute and relative frequencies and compared by the X 2 test or Fisher's exact test. These comparisons were tted for the era (antigenemia vs. RT-PCR).
The cumulative incidence of CMV-related events, CMV disease, and acute rejection were calculated by Kaplan-Meier and compared by log-rank test. The frequency of retreatment requirement was compared by the X 2 test. Time for detecting CMV-related events and the length of treatment with ganciclovir according to era were compared by test U of Mann-Whitney. For graft function, a generalized estimated equation was performed to compare the mean of eGFR between eras (antigenemia and RT-PCR). The model was adjusted by the Bonferroni test. The same approach was performed to compare tacrolimus levels between eras.
The potential clinical predictors for the primary outcomes (CMV-related events and CMV disease) were analyzed by the proportional hazard ratios (HR) throughout the Cox regression modeling (backward stepwise). The variables for the model were selected in bivariable analyses comparisons of patients who had CMV-related events with those who did not (supplementary material). The same approach was performed to select candidates variables related to CMV disease. Variables that reached a P-value < 0.20 were considered for the nal modeling. The median was imputed for the only variable with missing values, 30-day eGFR (1.38%). The accuracy of the nal model was assessed by the area under a receiver operating characteristic (AUC-ROC). Statistical analyses were performed using Statistical Package for the Social Sciences (version 26; IBM, Armonk, NY, USA), and statistical signi cance was de ned as P<0.05, with the 95% con dence interval.