Cytomegalovirus viraemia and mortality in renal transplant recipients in the era of antiviral prophylaxis. Lessons from the western Australian experience

All adult (age ≥ 18 years) patients who received a live or deceased donor kidney transplant in Perth, Western Australia (WA), between January 2007 and December 2012 were included in this study. Data were extracted from local health records and linked to data recorded in the Australia and New Zealand Dialysis and Transplant (ANZDATA) registry. If a patient received more than one kidney transplant during the study period, data regarding both grafts were extracted and both transplants were considered as separate grafts for analysis.

In WA, most patients received basiliximab induction and routine triple immunosuppression with a calcineurin inhibitor (CyA/tacrolimus), mycophenolate sodium or mycophenolate mofetil and oral prednisolone. Patients at lower risk of rejection did not receive either basiliximab or thymoglobulin induction and those at highest risk (around 10%) received thymoglobulin induction instead of basiliximab. Target tacrolimus trough levels are typically between 5 and 10 μg/L for the first 3 months, with the dose of prednisolone reduced to 5–10 mg daily by 3 months. Biopsy-proven T cell mediated rejection episodes were managed with intravenous methylprednisolone, with T cell depleting antibody being prescribed for steroid-resistant rejection episodes.

CMV prophylaxis with valganciclovir was administered for 100 days to R⁺ patients, 200 days to D⁺/R⁻ grafts, and no prophylaxis was given to D⁻/R⁻ grafts. CMV testing at both centres during or after prophylaxis is usually performed on a monthly basis throughout the first year post-transplant. After that time, testing was done on an adhoc basis, usually only if there was a clinical indication. Valganciclovir at induction doses was routinely given following CMV reactivation at a viral load between 10³ and 10⁵ copies/mL. CMV treatment was routinely stopped after CMV was undetectable for two consecutive tests over at least a 1 month period.

The following data were obtained from ANZDATA (to December 2013): age at transplantation; graft number; date of transplantation; date of birth; racial origin; weight and height at transplant; maximum panel reactive antibody (PRA) status; recipient smoking status; recipient CMV serology status; donor source; donor age; human leukocyte antigen (HLA) mismatches; occurrence of acute rejection (including rejection type, severity and treatment); induction therapy (interleukin-2 receptor antibody or T cell depleting antibody); immunosuppression prescribed at 0, 3, 6 months and at each year post-transplant; comorbidities of cardiovascular disease, chronic lung disease, diabetes or cancer (post-transplant or recurrent); whether or not graft failure occurred and the date of failure; and whether or not the patient died (to December 2014) and the date of death. Information on treatment for CMV viraemia was not available.

Donor CMV serology status was obtained from individual patient charts at the two transplant hospitals. Recipient CMV serology status was confirmed from individual patient records and where differences between patient records and ANZDATA were identified, the patient record information was used. Any data missing from the ANZDATA dataset were also obtained from patient records where possible.

CMV DNA plasma viral load testing was performed at PathWest Laboratory Medicine WA using the COBAS Amplicor CMV MONITOR assay (Roche Diagnostics, Branchburg, N.J., USA) from January 2007 to 12 May 2011, which has a linear range of 6 × 10² to 1 × 10⁵ copies/mL (not referenced to the international CMV quantitative standard, so not reportable in IU/mL). From 13 May 2011 the Abbott RealTime CMV assay (Abbott Molecular Inc., Des Plaines, IL, USA) on the m2000 RealTime platform was introduced, which has a linear range of 2 × 10¹ to 1 × 10⁸ copies/mL (multiplication of copies/mL by 1.56 converts the result to IU/mL according to the WHO international CMV quantitative standard). We obtained testing data (date of test, test result) up to December 2014 inclusive and CMV results were reported in copies/ml.

Incidence rates and 95% confidence intervals were calculated for all variables of interest for the outcomes: CMV viraemia with greater than or equal to 600 copies/ml; and patient death prior to 31 December 2014. The start time was the date of the first recorded transplant for calculating the incidence of CMV viraemia; and the date of the most recent recorded transplant for calculating the incidence of patient death prior to 2014. When calculating incidence for time-dependent variables, the start time was the time of transplant for the period when the variable equalled zero. If and when the categorical variables became greater than zero, this was the start time for incidence. Patient days prior to that time were counted as patient days for that variable being zero. The Exact method was used to estimate p-values for the significance of effect for binomial variables and univariable Cox Proportional Hazards analysis was used to estimate p-values for categorical variables with more than two possible values. Time-dependent variables, such as CMV, were fitted as time-dependent variables in the univariable Cox Proportional Hazards models. Age at transplantation was categorised to <65 and ≥65 on the basis of 65 being an age cut-off for increased risk of death and adverse outcomes in other studies [15‐17]. The cut-off for categorisation of CMV levels was set at 600 copies/ml on the basis that 600 copies/ml was the limit of detection for the less sensitive Roche COBAS Amplicor assay. The cut-off for categorisation of donor age was set at 60 years of age, and for categorisation of the total number of HLA mismatches the cut-off was set at two. Patient death outcome data were censored as of December 2014 and CMV viraemia outcome data were censored as of December 2013. The earlier censoring of CMV viraemia outcome data was because information about rejection and immunosuppression was only available up to December 2013.

Multivariable survival analysis was performed using Cox Proportional Hazards regression. Tied failures were calculated using the Breslow method for ties. Robust standard errors were used to estimate p values and 95% confidence intervals. Covariates with a p-value of 0.100 or less in the univariable models or covariates previously reported to correlate with study outcomes were included in the initial models. Variables were excluded from the model if their p-values for their adjusted hazard ratio was >0.05. For the model analysing CMV viraemia as the outcome, immunosuppression was categorised into three groups: calcineurin inhibitors (cyclosporin A or tacrolimus); antimetabolites (azathioprine, mycophenolate mofetil, or mycophenolate sodium); and mammalian target of rapamycin (mTOR) inhibitors (everolimus or sirolimus), and fitted as time-dependent covariates for each time period for which immunosuppression data were recorded in ANZDATA (3 months, 6 months, 1, 2, 3, 4, and 5 years). Anti-T lymphocyte antibody treatment after transplant was fitted as a time-dependant variable and was compared to having no treatment or treatment at or before transplantation. All rejection episodes were biopsy-proven with severe rejection defined as any vascular rejection or any severe humoral, glomerular or cellular rejection. Rejection variables were also fitted as time-dependent covariates. The following variables were included in the initial model for CMV viraemia: recipient age (categorised at 65 years), sex, donor age (categorised at 60 years), donor/recipient CMV status, immunosuppression, rejection, severe rejection, numbers of HLA mismatches (categorised as two or less or greater than two), and donor status (deceased or live). For the model examining death as the outcome, CMV status was fitted as a time-dependent covariate as one or more episodes (combined). The variables included in the initial model for death as an outcome were age (categorised at 65 years), sex, donor age (categorised at 60 years), donor status (deceased or live), rejection, severe rejection, CMV status, Aboriginality, smoking status, graft failure and the following comorbidities: diabetes; cerebrovascular disease, any cardiovascular disease other than cerebrovascular disease in patients with no cerebrovascular disease; chronic lung disease and cancer. The proportional hazards assumption was tested for all variables in every final model based on the analysis of Schoenfeld residuals.

A Kaplan-Meier plot showing death by CMV viral load was generated with CMV viral load being fitted as a time-dependent variable.

Data analysis was performed using STATA v14 (StataCorp LP, College Station, Texas, USA).

One hundred and thirty four (30.6%) patients had at least one episode of detectable viraemia during the study period. Of these, 74 (55.2%) patients had at least one episode of viraemia at 600 copies/ml or more. Eleven (14.9%) of the 74 patients had two episodes of CMV viraemia at 600 copies/ml or more with intervening periods of no CMV DNA being detected. One patient had a single CMV viraemia episode of 600 copies/ml or more before transplantation. This viraemic episode was excluded, leaving in the analysis 84 CMV viraemia episodes of 600 copies/ml or more from 73 patients over 461,961 patient days. The incidence of CMV viraemia of 600 copies/ml or more was 18.2 per 100,000 person days (95%CI 14.7–22.5). Forty-six patients died before 31 December 2013 (Fig. 1). There were three patients who developed CMV viraemia after graft failure and had no record of a subsequent transplant. These patients were excluded from the univariable analysis, and the CMV episodes were excluded from the Cox Proportional Hazards analysis as they were censored at graft failure.

Table 1 presents the numbers and rates of CMV viraemia at 600 copies/ml or more stratified by covariate categories. Being 65 years or above at the time of transplant, having D⁺/R⁻ status, having at least one episode of severe rejection, and having a transplant from a deceased donor or a donor aged 60 years and over were all significantly associated with having CMV viraemia in univariable analysis (Table 1).

Of the 438 transplant recipients, 55 (12.6%) died before 31 December 2014. Having at least one episode of CMV detectable at 600 copies/ml or greater, being aged 65 or over at the time of transplant, receiving a graft from a deceased donor, lung disease, cerebrovascular disease, and Aboriginal race were all predictors of death in univariable analysis (Table 3).

We next sought to examine the relationship between a number of thresholds for CMV viral load, the duration of viraemia and all-cause mortality. There were 134 (30.6%) patients who had at least one episode of detectable CMV viraemia. The median CMV viral load among these cases was 656 copies/ml (IQR 68, 4590), and levels ranged from 2.6 to 5,710,000 copies/ml. Using the quartile values as cut-offs, the CMV viral load was categorised as <68, 68–655, 656–4589 and 4590+ copies/ml. A Cox Regression model showed a significant HR for death only for those cases with a CMV viral load of ≥656 copies/ml (Table 4). Kaplan Meier survival plots for death by CMV viral load are shown in Fig. 2.

Using Cox Proportional Hazards regression analysis, having at least one episode of CMV viraemia with viral loads of 656 copies/ml or more, Aboriginal race, aged 65 or more at transplant, having cerebrovascular disease, and having any other vascular disease were all associated with death (Table 5). An interaction term of CMV with age was not significant and were therefore not included in the final model.

Of the 55 patients who died, 47 died with a functioning graft. A new Cox Proportional Hazards regression model was generated censoring at graft failure the eight cases who died after graft failure. In this model, having at least one episode of CMV viraemia with levels of 656–4589 copies/ml (aHR 3.36, 95% CI 1.49, 7.55, p = 0.003), having at least one episode of CMV viraemia with levels of 4590 copies/ml or more (aHR 3.96, 95% CI 1.85, 8.49, p < 0.001), Aboriginal race (aHR 3.33, 95% CI 1.74, 6.36, p < 0.001), aged ≥65 at transplant (aHR 2.38, 95% CI 1.11, 5.13, p = 0.026), cerebrovascular disease (aHR 4.24, 95% CI 1.87, 9.66, p = 0.001), and other vascular disease (aHR 2.23, 95% CI 1.22, 4.08, p = 0.009) were associated with death with a functioning graft.

The duration of CMV viraemia was not associated with death. The incidence of death from most recent transplant was 5.8 (95% CI 4.2, 8.1) for patients with no episode of CMV viraemia of 656 copies/ml or more; 20.9 (95% CI 11.2, 38.8) for patients with CMV viraemia of 656 copies/ml or more for a duration of up to 7 days; and 21.0 (95% CI 11.3, 39.1) for patients with CMV viraemia of 656 copies/ml or more for a duration of at least 7 days. Further analysis comparing death with duration of CMV viraemia with a threshold of 14 days and with 28 days also did not show an association between duration of CMV viraemia and death (data not shown).

There were 73 patients who had at least one episode of CMV viraemia of 656 copies/ml or greater post-transplant. Eleven of these patients had two episodes of CMV viraemia of 656 copies/ml or greater (a total of 84 episodes in 73 patients). Of those viraemic episodes, 31 (37%) occurred during the period when the patient was expected to be on CMV prophylaxis.