Clinical Significance of Low Cytomegalovirus DNA Levels in Human Plasma

Abstract

The clinical significance of the detection of low copy numbers of cytomegalovirus (CMV) DNA in immune-suppressed patients remains unclear. In this study, we compared the artus CMV Rotor-Gene PCR, utilizing an automated nucleic acid extraction and assay setup (the artus CMV protocol), with the COBAS Amplicor CMV Monitor test (our reference protocol). We then analyzed the results of all CMV PCR tests ordered following the implementation of the artus CMV protocol at our institution and followed 91 adult patients with positive test results. The artus CMV protocol had a linear range extending from 2.0 to 7.0 log₁₀ copies/ml and had a lower limit of 95% detection of 57 copies/ml. With archived plasma samples, this protocol demonstrated 100% sensitivity and 94% specificity for the detection of CMV DNA. Following implementation of the artus CMV protocol, 320 of 1,403 (22.8%) plasma samples tested positive (compared with 323/3,579 [9.0%] samples in the preceding 6 months), and 227 (16.2%) samples had copy numbers of <400/ml. Ninety-one adult patients had at least one positive test. The data were analyzed using a threshold of 200 copies/ml, and in 22 episodes, the viral load increased from <200 copies/ml to ≥200 copies/ml on sequential tests. In 21 of these 22 episodes, either the viral load continued to increase or antiviral treatment was initiated in response to the repeat value. In summary, we evaluate the performance characteristics of a protocol utilizing the artus CMV PCR and identify clinically meaningful changes in CMV DNA copy numbers even when they are initially detected at a low level.

INTRODUCTION

Cytomegalovirus (CMV) remains a major cause of morbidity following solid-organ transplantation (SOT) and hematopoietic stem cell transplantation (HSCT), and even with antiviral treatment, mortality from certain forms of CMV end-organ disease, such as CMV pneumonitis, remains high (14, 17, 18). Viremia has been associated with end-organ disease, and an increase in CMV DNA titers in serial blood samples, typically detected using a quantitative PCR assay, can predict disease development (10, 20). When used in a preemptive treatment protocol, quantitative PCR helps decrease the incidence of CMV disease and the use of antiviral therapy relative to those with the use of viral culture (9). However, despite the widespread use of quantitative CMV PCR assays, no thresholds for the diagnosis of disease or the initiation of preemptive therapy have been established (9, 14, 17). This situation has resulted from the use of different clinical diagnostic tests for CMV (both commercially available and lab developed), the broad range of values over which CMV disease can occur, and, until recently, the lack of a universal standard for CMV DNA quantitation (8, 13, 24).

The COBAS Amplicor CMV Monitor test was the first commercially available quantitative CMV PCR assay and remains commonly used in clinical virology laboratories (7). It has a lower limit of detection (LLOD) of 400 copies/ml of plasma and a linear range from 2.78 log₁₀ to 5.0 log₁₀ copies/ml (600 to 100,000 copies/ml) (5). The limits of detection and quantitation of this assay result in a large number of patients with ongoing yet unquantifiable or even undetectable viremia.

The artus CMV Rotor-Gene (RG) PCR is a commercially available quantitative CMV PCR assay. Initial reports have demonstrated a broad linear range and a lower limit of detection below that of the COBAS CMV Monitor assay (4). While copy numbers of CMV DNA as low as those detected with the artus assay have been reported, the clinical significance of these “low-positive” levels in a population of immune-suppressed patients has not been well described (3, 11, 15).

In this study, we report the analytical performance of a protocol utilizing the artus CMV PCR combined with automated sample preparation and assay setup (SP/AS) on the QIAsymphony SP/AS device (referred to below as the artus CMV protocol). Using archived human plasma samples, we determined the sensitivity and specificity of the artus CMV protocol for the detection of CMV DNA compared to those of the COBAS CMV Monitor assay following extraction on the MagNA Pure LC system (referred to below as the reference protocol). We documented an increased rate of detection of CMV in patient plasma samples following implementation of the artus CMV protocol at our institution. We then followed 91 adult patients with positive test results, and for a group of serially tested transplant recipients, we identified clinically relevant increases in CMV DNA copy numbers, even when quantified at low levels, previously undetectable with the reference protocol.

MATERIALS AND METHODS

Samples and control material.

Eighty-two archived plasma samples that had been tested by the reference protocol previously were tested using the artus CMV protocol. Basematrix 53 defibrinated human plasma (DHP; SeraCare, Milford, MA) was used as a negative control in all PCR runs. CMV AD-169 stocks (ATCC, Manassas, VA) diluted to 5.0 and 3.0 log₁₀ copies/ml (100,000 and 1,000 copies/ml) in DHP were used as high- and low-positive controls, respectively. The concentration of the original stock was determined using the reference protocol.

DNA extraction and assay setup.

DNA for the artus CMV protocol was extracted on 1.0 ml of acid citrate dextrose (ACD) plasma (1.2 ml of input volume required to account for the instrument dead volume) using the Qiagen Virus/Bacteria Midi kit on the QIAsymphony SP device (both from Qiagen, Valencia, CA). DNA was eluted in a final volume of 95 μl. Following DNA extraction, the artus CMV PCR was set up using the QIAsymphony AS. Ten microliters of extracted nucleic acids, 12.5 μl CMV RG master mix, and 2.5 μl of CMV magnesium solution were used, for a final reaction volume of 25 μl.

For the reference protocol, 100 μl of each specimen was extracted on the MagNA Pure LC system using the MagNA Pure LC DNA isolation kit with the DNA I Blood Cell High Performance protocol (Roche, Indianapolis, IN). Reactions were manually set up by pipetting 50 μl out of the 100-μl elution into COBAS Amplicor amplification tubes (A-tubes) containing 50 μl of the COBAS CMV Monitor master mix with magnesium.

Quantitative PCR assays.

The artus CMV PCR (Qiagen, Valencia, CA) is a real-time, hydrolysis-probe-based PCR targeting the CMV major immediate-early gene. Reactions were carried out on the Rotor-Gene Q (RGQ) instrument. Reaction mixtures underwent an initial 10 min at 95°C, followed by 10 cycles of touchdown PCR (95°C for 15 s, 65°C for 30 s, and 72°C for 20 s, with the annealing step decreasing by 1°C each cycle). This was followed by 35 cycles of 95°C for 15 s, 56°C for 30 s, and 72°C for 20 s. Data were collected on the green (CMV) and yellow (internal-control) channels during annealing. An internal control is added to each primary sample prior to extraction, and amplification is performed with specific primers and hydrolysis probes contained in the artus CMV master mix to ensure adequate extraction efficiency and the absence of inhibitors. Positive samples produced a signal above a threshold value of 0.1 in the green channel. The number of CMV copies per milliliter was calculated by comparing the cycle number at which the fluorescent signal crossed this threshold (C_T) with the four-point standard curve included on each run. The standard curve was generated using four quantitation standards (Qiagen, Valencia, CA) with concentrations of 10,000, 1,000, 100, and 10 copies/μl. Results were reported in copies per milliliter of the original plasma sample. Samples with detectable virus quantified at <150 copies/ml were reported as “Detected, <150 copies/ml.”

The artus CMV protocol was subsequently calibrated to the 1st WHO International Standard for human cytomegalovirus for nucleic acid testing (NAT)-based assays (NIBSC code 09/162), obtained from the National Institute for Biological Standards and Controls (NIBSC; Hertfordshire, United Kingdom). The WHO International Standard was diluted to 5.0, 4.7, 4.0, 3.7, and 3.0 log₁₀ IU/ml, and 6 replicates at each concentration were run on 4 separate days. The data collected in copies per milliliter were compared to the expected international units per milliliter. A conversion factor was calculated by taking the mean ratio of international units per milliliter to copies per milliliter for all data points.

The COBAS Amplicor CMV Monitor assay (Roche, Indianapolis, IN) utilized in the reference protocol is a nonsaturating endpoint PCR targeting the CMV polymerase UL54 gene, followed by quantitation with an automated enzyme-linked oligosorbent assay (ELOSA). A separate internal quantitation standard is added to the primary specimen prior to extraction. Amplification and detection were performed as described previously (23).

Samples that yielded discordant results by the artus CMV and reference protocols were tested using the CMV R-gene real-time PCR assay (referred to below as the CMV R-gene assay) (Argene, Shirley, NY), which targets the pp65 UL83 gene. This assay has a reported lower limit of detection (LLOD) of 50 copies/ml and a linear range extending to 7.0 log₁₀ copies/ml (1). Ten microliters of DNA extracted using the Qiagen SP device was added to 15 μl of amplification premix. Reaction conditions on the RGQ instrument were 95°C for 15 min, followed by 45 cycles at 95°C for 10 s and 60°C for 40 s. Positive samples produced a signal above a threshold value of 0.025 in the green channel. The number of CMV copies per milliliter was calculated by comparing the cycle number at which the fluorescent signal crossed this threshold (C_T) with the four-point standard curve included on each run. The standard curve was generated using four quantitation standards (Argene, Shirley, NY) with concentrations of 5,000, 500, 50, and 5 copies/μl.

Linearity, lower limit of detection, and accuracy.

The linearity of the artus CMV protocol was determined using serial 10-fold dilutions of CMV AD-169 in DHP. Replicates of each dilution were taken through the entire assay, from extraction to quantitation. To determine the LLOD, OptiQuant CMV panel samples (AcroMetrix, Benicia, CA) were diluted to 100, 50, and 25 copies/ml. Twenty replicates were run at each concentration, and probit analysis was performed to determine the lower limit of 95% detection. The quantitative results obtained with the artus CMV and reference protocols were compared using 82 previously tested clinical samples. Samples that yielded discordant results were retested using the CMV R-gene assay.

Clinical study design.

This research was approved by the Stanford University Institutional Review Board. We analyzed the general test characteristics for plasma CMV PCR assays performed at our institution during the study period, which extended from 20 June 2011, the date on which the artus CMV protocol was implemented, to 12 September 2011. Adult patients with a single positive result by the artus CMV protocol during this interval were then identified. We collected relevant clinical data, including patient age, sex, and significant medical history. For transplant recipients, we documented the type of transplant, the indication for transplantation, the CMV status at transplantation, and antiviral prophylaxis. We also documented any treatment administered during the study. The death summary was used for the documentation of cause of death where applicable.

The definitions used for CMV infection and end-organ disease were consistent with previously published recommendations (14, 17, 19). CMV infection was defined by the detection of CMV DNA in patient plasma. The confirmation of CMV disease required a biopsy with consistent histopathology, as well as virus isolation or positive immunohistochemistry (IHC). The diagnosis of “CMV syndrome” was considered only for SOT recipients. Recurrent infection was defined as a rising CMV load following the completion of a course of antiviral treatment, with the documented absence of detectable CMV DNA or the presence of CMV DNA that was detectable but at <200 copies/ml at the completion of treatment.

Immune suppression regimens varied by transplant organ, time from transplantation, and evidence of graft rejection or graft-versus-host disease. Transplant patients received antiviral prophylaxis according to protocols employed by the different transplant services at our institution. Two HSCT patients remained on treatment for CMV infection throughout the study period and were not included in the tally of patients receiving prophylaxis. Preemptive therapy and treatment for CMV infection or disease were initiated, and the duration of therapy was determined by the attending physician caring for the patient. Repeat CMV testing during therapy and follow-up were performed by the clinical laboratory at the discretion of the treatment team.

Statistical analysis.

Agreement between the artus CMV and reference protocols was assessed using a Bland-Altman plot (2). Standard statistical analyses were performed using Excel (Microsoft, Redmond, WA). Probit analysis was performed using SPSS (IBM, Armonk, NY). Two-tailed Fisher exact tests and unpaired t tests were performed using GraphPad software (GraphPad, La Jolla, CA).

RESULTS

Analytical characteristics of the artus CMV protocol.

The linear range of the artus CMV protocol extended from 2.0 log₁₀ to 7.0 log₁₀ copies/ml (100 to 10,000,000 copies/ml). By use of probit analysis at 95% detection, the LLOD was found to be 57 copies/ml. Assay precision was further evaluated using samples with concentrations of 5.0 log₁₀, 3.0 log₁₀, and 2.30 log₁₀ copies/ml (100,000, 1,000, and 200 copies/ml). The mean values were 5.08 log₁₀, 2.97 log₁₀, and 2.27 log₁₀ copies/ml (120,000, 933, and 195 copies/ml), and the interrun coefficients of variation, expressed as percentages (%COV), were 1.4, 3.7, and 5.8, respectively (calculated for logarithmic values). A checkerboard experiment using 24 CMV-negative plasma samples and 24 CMV AD-169-spiked plasma samples (concentration, 6.0 log₁₀ copies/ml [1,000,000 copies/ml]) was performed to evaluate the QIAsymphony SP/AS device; no evidence of cross-contamination was observed (data not shown).

The artus CMV protocol was calibrated to the 1st WHO International Standard for human cytomegalovirus for NAT-based assays, resulting in a conversion factor of 0.76 IU/copy of CMV DNA.

Comparison of the artus CMV and reference protocols.

A total of 82 archived clinical samples (42 positive, 40 negative) that had been tested previously using the reference protocol were retested using the artus CMV protocol on the QIAsymphony SP/AS/RGQ system. All 42 samples with detectable CMV DNA by the reference protocol were positive by the artus CMV protocol (sensitivity, 100%) (Table 1). There was good agreement in the quantitative results between the artus CMV and reference protocols (Fig. 1). Except for two samples (discussed below), all samples yielded results that differed by <2 standard deviations (SD) of the mean difference.

Table 1.

Comparison of the artus CMV protocol to the reference protocol using 82 clinical samples

Result by the artus CMV protocol	No. of samples with the following result by the reference protocol:		Total
	Positive	Negative
Positive	42	10a	52
Negative	0	30	30
Total	42	40	82

^aThe 10 samples positive by the artus CMV protocol and negative by the reference protocol were retested using the CMV R-gene assay. Eight of 10 samples were positive by the CMV R-gene assay.

Fig 1.

Bland-Altman plot for the comparison of the artus CMV and reference protocols using 42 clinical samples. The mean difference (boldface horizontal line) was zero; dotted lines indicate the mean ± 2 standard deviations.

Of 40 samples with CMV DNA undetectable by the reference protocol, 10 were positive by the artus CMV protocol (specificity, 75%) (Table 1). The mean concentration in the 10 discordant samples by use of the artus CMV protocol was 255 copies/ml (range, 13 to 986 copies/ml), and 8 samples had copy numbers of <400/ml. Seven patients had a positive CMV PCR result by the reference protocol prior to (4 patients) or subsequent to (3 patients) the collection of the discordant sample that was used in this validation. On retesting using the CMV R-gene assay, 8 of 10 discordant samples (80%) had detectable CMV DNA, and 2 remained undetectable. The revised sensitivity and specificity for the artus CMV protocol, by using any two positive tests as the “gold standard,” were 100% and 94%, respectively.

Two samples, collected from a single patient at different time points, were detected but underquantitated by the artus CMV protocol compared to the reference protocol (2.82 versus 3.73 log₁₀ copies/ml and 2.44 versus 3.22 log₁₀ copies/ml for the artus CMV and reference protocols, respectively). The results differed by ≥2 SD from the mean difference (Fig. 2). For these discordant samples, the CMV R-gene assay results were consistent with the reference protocol and remained elevated over those for the artus CMV protocol.

Fig 2.

Flow chart of the 91 adult patients with at least one positive plasma CMV PCR result during the study period.

Clinical results.

During the study period, 1,403 plasma CMV assays were performed at our institution using the artus CMV protocol. Of these, 320 showed detectable CMV DNA (22.8%), with 93 quantitated at >400 copies/ml (6.6%) and 227 (16.2%) detected at <400 copies/ml. In the 6 months prior to this change, using the reference protocol, 323 out of 3,579 tests showed detectable CMV DNA (9.0%; P, <0.0001). All of these samples had copy numbers of ≥400/ml. The number of patients with at least one positive test also increased with the use of the artus CMV protocol over the same time interval (17.8% versus 6.9% of patients per month; P, <0.0001). The average number of tests did not differ significantly throughout this period (593 versus 539 tests per month, respectively; P = 0.052).

We identified and collected clinical information on the 91 adult patients who had at least one plasma sample with CMV DNA detectable by the artus CMV protocol. The mean age for patients was 51.2 years (standard deviation, 14.8 years), and 53 patients were male. The majority of patients had received a transplant (n = 82; 90.1%); of these, 55 (60.4%) had HSCTs and 27 (29.7%) had SOTs (lung [n = 9], renal [n = 6], heart [n = 5], liver [n = 3], or combined [n = 4] transplants). Table 2 shows the characteristics of the transplant recipients, including CMV status at the time of transplantation and antiviral prophylaxis following transplantation.

Table 2.

Characteristics of 82 transplant recipients with positive plasma CMV PCR results during the study period

Transplant type and characteristic	No. of patientsa
Total	82 (100)
SOT	27 (32.9)
CMV status at transplantation
R+/D+	15
R+/D−	5
R−/D+	6
R+/D unknown	1
Prophylaxis
Acyclovir	3
Valganciclovir	15
None	9
HSCT	55 (67.1)
Allogeneic	53
Autologousb	2
CMV status at transplantation
R+/D+	29
R+/D−	22
R−/D+	4
Prophylaxis	53
Acyclovir	51
Valganciclovir	0
None	2

^aNumbers in parentheses are percentages of patients.

^bThe two autologous HSCT transplants were CMV positive (R⁺) at transplantation.

The number of tests per patient during the study period (20 June through 12 September 2011) differed significantly based on the underlying condition. HSCT recipients had a mean of 7.0 tests performed per patient (standard deviation, 3.8), whereas non-HSCT patients had a mean of 2.7 tests per patient (standard deviation, 2.5) (P, <0.0001).

Treatment outcomes and recurrence.

Forty-five patients (49.5%) received antiviral treatment for CMV infection or disease. There was no difference in the proportion of patients treated for CMV based on the type of transplant (58.2% for HSCT versus 40.7% for SOT) (P, 0.163). All patients received valganciclovir or ganciclovir as treatment. Four patients also received foscarnet during the study (8.9%). Twenty-eight treatment episodes were started and completed during the study. The average duration of antiviral therapy for these episodes was 30.5 days (standard deviation, 15.7 days; range, 10 to 78 days). Twenty-one patients had at least 2 weeks of follow-up off therapy, and eight patients had recurrent CMV infection. The rates of recurrence were not significantly different for patients whose CMV DNA remained detectable but at <200 copies/ml at the end of treatment (50%) (n = 10) and those for whom CMV DNA had become undetectable (27%) (n = 11) (P, 0.387).

Two patients had confirmed end-organ disease. The first patient, without a transplant, presented with diabetic ketoacidosis (DKA) and odynophagia and was found to have esophageal ulcerations positive for CMV by IHC. The serum viral load was 3.78 log₁₀ copies/ml (6,020 copies/ml). The second patient received a renal transplant (R⁻/D⁺) 17 months before presenting with abdominal pain and diarrhea. Serum CMV testing showed 5.41 log₁₀ copies/ml (259,000 copies/ml), and colon biopsies revealed viral inclusions positive for CMV staining by IHC. Three other patients received treatment for possible CMV disease (colitis [2 patients] and CMV syndrome [1 patient]). One patient with possible colitis did not undergo a colonoscopy until after treatment had started. The second patient had a colonoscopy at an outside facility, the results of which were not available to us. The patient with possible CMV syndrome improved with antiviral treatment; a limited workup did not reveal other possible causes for his illness.

Eight patients died during the study; causes of death included intracerebral hemorrhage, end-stage liver disease (2 patients), graft-versus-host disease (2 patients), pulmonary embolism and pneumonia, congestive heart failure, and acute myeloid leukemia (AML) with blast crisis. None of these patients had confirmed CMV disease. Three patients were receiving treatment for CMV infection at the end of life.

Patients with detectable CMV DNA at low levels.

To determine the clinical significance of low-positive CMV DNA levels, we evaluated our data based on a threshold of 200 copies/ml. This threshold was chosen after initial review of our data for 91 patients with positive test results. Eighty percent (4/5) of patients with quantified viral loads between 150 and 200 copies/ml had a spontaneous decline in the viral load to <150 copies/ml, whereas 21/22 patients with viral loads that increased to ≥200 copies/ml either had viral loads that continued to increase or received treatment (P, 0.0014).

As shown in Fig. 2, 86 (94.5%) patients in this study had at least one sample in which CMV DNA was detected at <200 copies/ml. Five (5.5%) of the patients always had elevated CMV DNA levels of >1,000 copies/ml during the study period and were not evaluated further. Sixteen (17.6%) patients had samples with viral loads of >200 copies/ml at the beginning of the study period (median viral load, 1,348 copies/ml; range, 354 to 9,320 copies/ml). Fifteen of these patients received antiviral treatment. A single patient had an increased viral load in the setting of a disseminated varicella-zoster virus infection, and the CMV copy number declined with high-dose acyclovir treatment. The transition from a viral load of <200 to >200 copies/ml was not captured during the study period for these 16 patients.

Forty-nine (53.8%) patients had samples in which CMV DNA was detected but only at <200 copies/ml. Of these 49 patients, 16 had a single test performed during the study period. Among the 33 patients with multiple tests, the rate of positive tests per patient ranged from 8.3% to 81.8% (from 1 out of 12 to 9 out of 11 samples). None of these patients received treatment during the study period, though 15 had been treated for CMV infection or disease previously.

Twenty-one patients developed 22 clinical episodes during which the CMV DNA copy number increased from a level of <200 copies/ml (either undetected or detected but at <200 copies/ml) to a level of ≥200 copies/ml on sequential tests (Table 3). All 21 patients had received HSCTs (10 had a CMV status of R⁺/D⁺; 9 had R⁺/D⁻; and 2 had R⁻/D⁺). The median time between sequential tests was 7 days (range, 1 to 27 days).

Table 3.

Outcomes for 22 clinical episodes during which the CMV viral load increased from <200 to >200 copies/ml on sequential tests

Viral load on the second test (copies/ml)	Total no. of episodes	No. of episodes in which:
		Viral load continued to increase	Viral load decreased without intervention	Patient received treatment
200–400	9	8	1	8
400–600	3	2	0	3
>600	10	5	0	10
Total	22	15	1	21

We further categorized these 22 episodes based on quantification by the sequential tests. In 45.5% (10/22) of episodes, the viral load subsequent to initial testing was ≥600 copies/ml, levels quantifiable by the reference assay. In five of these episodes, antiviral treatment was initiated in response to this viral load (median viral load, 999 copies/ml; range, 937 to 8,390 copies/ml). In the other five episodes, the CMV level continued to increase, and all five patients were eventually treated (median viral load at the start of treatment, 946 copies/ml; range, 763 to 2,570 copies/ml). In 13.6% (3/22) of episodes, the subsequent viral load was 400 to 600 copies/ml, a range detectable but not quantifiable by the reference method. A CMV level of 541 copies/ml prompted antiviral treatment during one episode. In two episodes, the viral load continued to increase; both patients received antiviral treatment (viral loads at the start of treatment, 1,160 and 761 copies/ml, respectively).

In 40.9% (9/22) of episodes, the subsequent viral load ranged from 200 to 400 copies/ml, a range that was not detectable by the reference method. In 8 of these episodes, the viral load continued to increase. Seven of these episodes were ultimately treated with antivirals, and one was treated with reduced immune suppression (median viral load at the start of treatment, 888 copies/ml; range, 323 to 7,950 copies/ml). There was a single episode in which the viral load increased from <200 to >200 copies/ml on sequential tests (262 copies/ml) and then decreased without intervention.

In 72.2% (16/22) of episodes, patients did not immediately receive treatment in response to their subsequent CMV loads. In 15 of these clinical episodes, CMV levels continued to increase and ultimately required antiviral therapy or reduced immune suppression.

DISCUSSION

In this study, we evaluated a CMV testing method that involves automated sample preparation and assay setup and real-time quantitative PCR using the artus CMV PCR. This protocol showed good numerical agreement with the reference protocol, and the artus CMV protocol has a broader linear range and lower LLOD. In addition, we also report the finding of a significant increase in the rate of positive tests following the implementation of the artus CMV protocol for the detection and monitoring of CMV in patient plasma samples at our institution. This increase appears to be accounted for by the detection of viral loads undetectable by the reference method, though this cannot be confirmed without testing all samples by both methods.

The artus CMV PCR has been evaluated in a number of previous studies (4, 6, 7, 12, 22). Our results agree with the findings of the only previous report to compare artus CMV PCR reagents to the COBAS Amplicor CMV Monitor assay (4). However, we demonstrate a lower LLOD and extend the previous results to include the use of a different DNA extraction protocol and smaller reaction volumes.

We did identify two specimens that amplified poorly in the artus CMV protocol. These samples were drawn at separate times from a single patient, and repeat testing using the reference protocol and the CMV R-gene assay showed consistent and elevated quantitative results. The three assays used in this study target different regions of the CMV genome. One potential explanation for the discrepant quantitation is that the CMV isolate from this patient contains a mutation in the proprietary major immediate-early gene sequences targeted by the artus CMV PCR. Such mutations have been reported to cause decreased amplification in other assays that target this gene, but they have not been reported previously for the artus CMV PCR (6, 16).

Recently, the WHO approved an international quantitation standard for CMV in an effort to address concerns regarding the agreement of quantitative results for CMV PCR testing between centers (14, 21, 24). Our results were generated using a commercially available set of quantitation standards, and a conversion factor of 0.76 IU/copy of CMV DNA was subsequently calculated by calibrating the artus CMV protocol to the 1st WHO International Standard for human cytomegalovirus. To our knowledge, this is one of the first reports to provide such a conversion factor. Reporting clinical results in international units per milliliter should allow for more-accurate comparisons of quantitative results generated using different assays.

Previous studies have documented the ability of PCR assays to detect CMV DNA copy numbers below 100/ml, though the clinical significance of such values remained unclear (3, 15). Boeckh et al. designed a sensitive real-time PCR assay and reported on two patients diagnosed with CMV disease for whom viral loads were detectable in stored samples as early as 7 weeks prior to diagnosis (3). Kaiser et al. modified the COBAS Amplicor CMV Monitor assay such that they could detect CMV DNA down to 20 copies/ml of patient plasma (15). They then identified 16 HSCT recipients with CMV DNA detectable by their new protocol. Three patients in that study had low detectable viral loads that resolved without therapy. However, most patients received treatment following the detection of CMV DNA, making it difficult to draw conclusions regarding changes in the viral load from this report.

For HIV-positive patients with low CD4 counts, Erice et al. studied the use of a CMV antigen assay (limit of detection reported as 200 copies/ml) and the reference protocol used in our study (11). In univariate analysis after adjustment for HIV load and CD4 count, an increase in the CMV copy number to >200/ml during the study was significantly associated with decreased survival. While the numerical cutoff used in their study cannot be generalized to our population given the significant differences in testing modalities, these results support the conclusion that increases in CMV load are significant even when CMV DNA is initially detected at a very low level.

During our study, 94.5% (n = 86) of patients had a CMV load that was detectable but at <200 copies/ml, and many patients had repeatedly detectable viral loads at <200 copies/ml. The finding of detectable virus at this level is difficult to interpret in isolation. However, the CMV load continued to rise, or antiviral treatment was initiated, in 21 out of 22 clinical episodes during which the viral load increased from <200 to >200 copies/ml. Twenty of these patients eventually received antiviral treatment (Table 3), and in only a single episode did the viral load decline without any intervention (from 262 to <200 copies/ml). All 21 patients in this subgroup had received HSCTs and were being followed using a preemptive treatment protocol; none of them developed confirmed CMV disease.

Recommendations for the duration of treatment for CMV infection and disease are available, but these have not been studied in a randomized setting (14, 17). Detectable CMV loads at the end of treatment have been shown to predict recurrence in SOT patients, though that report utilized a less analytically sensitive assay than the test used here (23). In our study, the rates of recurrent infection for patients with detectable viral loads at the end of therapy did not differ significantly from those for patients with undetectable CMV at the end of therapy. The number of patients in this analysis was small (n = 21), however, and this merits further study in a larger population.

Limitations of our study include the retrospective design and the reliance on treating physicians to determine when to initiate antiviral therapy during periods of viremia. Despite these limitations, we were able to demonstrate that episodes of CMV viremia eventually treated with antiviral therapy can be detected earlier using a more sensitive testing method. A prospective study would better define the clinical benefits that may result from using the CMV threshold identified here (i.e., a shorter duration of antiviral therapy or decreased toxicity). While the incidence of end-organ disease would provide a more rigorous endpoint, its occurrence is likely too rare for it to be used in this fashion.

In summary, we report the performance characteristics of an automated protocol using the artus CMV PCR in comparison with a widely used reference protocol. We also report our findings from the study of a cohort of immunocompromised patients monitored with this test. In a subgroup of HSCT recipients, we observed clinically meaningful increases in viral loads, even when quantified at levels previously undetectable in our reference protocol. These findings support the use of a low threshold for the initiation of preemptive therapy in transplant patients.