Performance evaluation of fully automated cobas® 6800 CMV PCR for the detection and quantification of cytomegalovirus DNA in neonatal urine and saliva, and adult urine, saliva, and vaginal secretion

Laboratory testing for cytomegalovirus (CMV) in bodily fluids is essential to manage congenital and prenatal CMV infection. The rapid and fully automated cobas® CMV PCR is approved only for the testing of plasma in transplant patients. To evaluate the performance of the cobas® CMV to detect and quantify CMV DNA in neonatal and adult female urine, saliva, and vaginal secretion, the limit of detection (LoD), limit of quantification (LoQ), imprecision, linearity, PCR efficiency, bias, analytical specificity, cross‐reactivity, and cross‐contamination of the cobas® CMV for urine, saliva, and vaginal secretion was determined. The performance of the assay was evaluated prospectively with two laboratory‐developed PCR assays using neonatal and adult urine, saliva swabs, and vaginal swabs. The LoD and LoQ were 31 and 100 IU/mL, respectively, for urine, and 81 and 100 IU/mL, respectively, for vaginal secretion. The LoD and LoQ for saliva were the same (200 IU/mL). The cobas® CMV was precise (coefficient of variation ≤10%), linear (R2 ≥ 0.995), and efficient (1.07 and 1.09) between 100 and 250,000 IU/mL for the sample types. The bias and analytical specificity was <±0.30 log10 IU/mL and 100%, respectively. Cross‐reactivity with non‐CMV pathogens was not detected. Cross‐contamination rate was 0.28%. The diagnostic accuracy, sensitivity, and specificity of the cobas® CMV for neonatal urine and saliva were ≥95.0%, ≥93.3%, and ≥90.4%, respectively. The overall percent agreement for adult urine, saliva, and vaginal secretion was 86.6%, 94.5%, and 89.4%, respectively. Taken together, the cobas® CMV demonstrated acceptable analytical and diagnostic performance, and is suitable for routine diagnostic laboratory investigation of CMV infection in neonates and adults.


| INTRODUCTION
[5][6][7] Rapid laboratory testing for CMV in neonatal urine and saliva is recommended to diagnose cCMV and guide clinical management. 80][11] The need to improve CMV diagnostics to support timely patient management and universal screening has been reported. 12,13In pregnant women, testing for CMV shedding in urine, saliva, and vaginal secretion not only allows the detection of maternal infection and prediction of the risks of vertical transmission, but may also support early discussion for further investigation and potential medical referral. 5,6,14til very recently, a rapid, fully automated, and high throughput CMV test to support clinicians' decision-making and patient management for congenital and prenatal CMV was not available.[16] In addition, the inefficient laboratory processes and long test turnaround time have contributed to suboptimal patient care. 12r laboratories that do not have access to a locally validated test (such as ours), samples were referred to reference laboratories for testing which resulted in further delays.
The cobas ® 6800 is a rapid and fully automated sample-toresult high throughput PCR system that was implemented in our laboratory in 2016 for the detection and quantification of viral load. 17The cobas ® 6800 and assays were developed to improve laboratory efficiency, test turnaround times, and analytical and clinical performance. 18One of the implemented tests was cobas ® CMV.However, the assay is FDA-approved and IVDR-certified only for the detection and quantification of CMV DNAemia in transplant patients, which impedes its use for nonplasma samples to support a timely management of congenital and prenatal CMV infection unless its fitness for intended purpose has been validated and evaluated. 19,20Therefore, the aim of this study was to validate and evaluate the diagnostic performance of the cobas ® CMV assay for neonatal urine and saliva, and adult female urine, saliva, and vaginal secretion, to facilitate a rapid, accurate and efficient laboratory detection and quantification of CMV DNA.Further manual intervention of the testing process was not required.
The instrument performs fully automated sample extraction, PCR amplification, and result analysis using proprietary algorithms.To evaluate the performance of the cobas ® CMV with ISO 15189accredited PCR assays, samples were tested also by reference laboratory M (within 20 h of receipt, urine and vaginal swabs) or reference laboratory R (within 5 days of receipt, saliva swabs).
Samples were stored at 2-8°C at the respective reference laboratory before testing.The reference laboratories were blinded to the cobas ® CMV results.The PCR protocol, workflow, and turnaround time for the cobas ® and reference laboratory assays are summarized in Supporting Information: Table S1.

| Collection of samples
For the validation of the cobas ® CMV, nonmenstrual and CMV PCR negative urine and vaginal swabs were collected from five nonpregnant female volunteers as described previously. 21Urine preserved in CPM was prepared by adding 5 mL urine to 4.3 mL CPM.CMV PCR negative saliva fluids and swabs were collected and prepared from three of the five aforementioned volunteers as described previously. 21For the crossreactivity experiment, self-collected saliva swabs were obtained from four of the five volunteers.All samples were homogenized before being tested or pooled for spiking experiments.

| Limit of detection (LoD)
Eight CMV titers (0-90 IU/mL, urine and vaginal secretion preserved in CPM; 0-250 IU/mL, saliva preserved in eNAT) were prepared using the 1 st WHO standard for CMV. 22Samples were tested in line with the Clinical and Laboratory Standards Institute (CLSI) EP17-A2 guidance. 23The procedural variations are described in Supporting Information: Tables S2-S4.

| Bias, imprecision, limit of quantification (LoQ), linearity, and PCR efficiency
Eight CMV titers (20-250,000 IU/mL) in urine, saliva, and vaginal secretion preserved in PCR media were prepared using the 1 st WHO standard for CMV. 224][25] The procedural variations are described in Supporting Information: Tables S5-S7.

| Analytical specificity
CMV DNA negative urine, saliva, and vaginal secretion preserved in PCR media was tested with eight cobas ® CMV reagent lots.

| Cross-contamination
CMV DNA negative cobas ® Diluent and urine, saliva, and vaginal secretion preserved in PCR media was tested whenever the cobas ® CMV PCR was performed.The negative samples were placed randomly among high titers CMV samples.

| Performance evaluation of the cobas ® CMV for neonatal urine and saliva
The evaluation was performed prospectively by testing neonatal urine and/or saliva swabs submitted to the diagnostic laboratory between 2020 and 2023 for CMV PCR as part of the clinical investigation of, or follow-up for, cCMV infection from three acute tertiary referral hospitals in London (Croydon University Hospital, Kingston Hospital [KH], and St George's Hospital [SGH]).Seven neonatal saliva swabs submitted between April and July 2021 for the investigation of a possible cCMV infection as part of the Cytomegalovirus Shedding Characteristics in Pregnant Women (cCHIPS) study (NCT04021628) were also included. 279 | Performance evaluation of the cobas ® CMV for adult urine, saliva, and vaginal secretion The evaluation was performed prospectively by testing pregnant and nonpregnant adult female urine, saliva swabs, and/or vaginal swabs collected between 2019 and 2023 for the diagnostic investigation of, or follow-up for, maternal CMV infection from two hospitals (KH and SGH), cobas ® CMV assay validation, and cCHIPS study.

| Data analysis
CMV DNA titer was log 10 -transformed before analysis.Data were summarized as mean ± SD, or median, and the minimum and maximum range, as appropriate.LoD was estimated using probit regression at 95% probability. 23For the determination of bias, the difference between the measured and nominal CMV titers was determined, using a goal of ≤±0.30 log 10 IU/mL acceptable error. 24Imprecision was determined for the repeatability and intermediate precision conditions of measurement and expressed as coefficient of variation (CV). 25The LoQ was the lowest titer with a hit rate of 100%, CV ≤ 10%, and ≤0.50 log 10 IU/mL total analytical error (TAE). 23Linearity was assessed by regression and Kolmogorov-Smirnov-CUSUM test, using a goal of ≤0.50 log 10 IU/mL allowable difference. 28PCR efficiency was determined with the formula 29 Analytical specificity and cross-contamination rate with 95% confidence interval (CI) was determined.For the performance evaluation of the cobas ® CMV for neonatal urine and saliva, the diagnostic accuracy, sensitivity, specificity, positive predictive value, negative predictive value, positive likelihood ratio (LR+), and negative likelihood ratio (LR-) with 95% CI of the cobas ® CMV was determined. 30For adult samples, because currently there is neither a consensus definition of CMV "shedding", nor an established reference assay for the detection and interpretation of CMV DNA in urine, saliva, and vaginal secretion, the performance of the cobas ® CMV, using reference laboratory M and R assays as comparators, was reported as overall percent agreement, positive percent agreement (PPA), and negative percent agreement (NPA) with 95% Cl. 30 Data analysis was performed using Analyse-it software (Analyse-it Software Ltd).
p values < 0.05 were considered statistically significant.

| Bias, imprecision, LoQ, linearity, and PCR efficiency
Between 100 and 250,000 IU/mL, the detection and quantification of CMV DNA in urine, saliva, and vaginal secretion was accurate.At these titers, the bias between the measured and nominal titers was <±0.30 log 10 IU/mL (Table 1).The detection and quantification of CMV DNA was less reliable at low titers (≤50 IU/mL) (Supporting Information:

Tables S5-S7
).There was a satisfactory imprecision for all sample types; for CMV titers ≥1000 IU/mL, the CVs for repeatability and intermediate precision were ≤5%, and for titers <1000 IU/mL, the CVs for repeatability and intermediate precision (if quantifiable) were ≤10% (Table 1).Using the criteria of a 100% hit rate, CV ≤10% and ≤0.50 log 10 IU/mL TAE, the LoQ was designated as 100 IU/mL for urine and vaginal secretion.For saliva, as the LoD was 200 IU/mL, the LoQ was designated as 200 IU/mL.The cobas ® CMV was found to be linear and efficient between 100 and 250,000 IU/mL for all sample types.At these titers, the coefficient of determination and PCR efficiency were 0.998 and 1.07, respectively, for urine, 0.996 and 1.09 for saliva, and 0.995 and 1.07 for vaginal secretion (Table 1, Supporting Information: Figure S1).

| Analytical specificity
Sixty-nine vials of each sample type were tested with eight cobas ® CMV reagent lots.All samples tested negative for CMV DNA.The analytical specificity of the cobas ® CMV for each sample type was 100% (95% CI 94.8-100.0).

| Performance evaluation of the cobas ® CMV for neonatal urine and saliva
One hundred and forty-eight urine and 100 saliva swabs from 25 and 12 neonates, respectively, were tested by the cobas ® CMV in parallel with reference laboratory PCR assays (Supporting Information: Table S11).As the cobas ® CMV was not validated for the testing of urine and saliva swabs, only the reference laboratory PCR results were reported to clinicians for patient management.The most common indications for a CMV PCR request were "congenital T A B L E 1 Analytical performance characteristics of the cobas ® CMV for urine, saliva, and vaginal secretion.CMV", "intrauterine growth restriction", "suspected sepsis" (urine and saliva), "premature baby" (urine), and "failed newborn hearing test" (saliva).Overall, the performance of the cobas ® CMV was excellent (   S12 and S13).Of these samples, 8 urine, 5 saliva swabs, and 5 vaginal swabs were collected from four patients for the diagnostic investigation of, or follow up for, CMV infection in pregnancy.As the detection of CMV in these samples to support patient consultation was not part of a routine clinical practice, CMV PCR results were reported with qualification to requesting clinicians following an informed consent of the nonapproved use of the test.Overall, the cobas ® CMV demonstrated satisfactory agreement with reference laboratory assays (Table 3).

| DISCUSSION
Laboratory testing for CMV infection in neonatal urine and saliva is essential to diagnose and manage cCMV. 85][6][7]14 We found the cobas ® CMV PCR demonstrated the speed, accuracy, and efficiency required for diagnostic laboratory detection and quantification of CMV DNA in urine, saliva, and vaginal secretion.If implemented, the assay may reduce the delay in initiating antiviral therapy for cCMV, and support a timely management of CMV infection in neonates and adults. 8,12,13 the context of diagnosing cCMV in neonates, a sensitive and specific test is required.The cobas ® assay was found to be fit for this purpose.The discordant samples at low CMV titers were detected only by the cobas ® or reference assay, and could be due to a laboratory contamination and/or differences in assay sensitivity.The discordant results may also be due to PCR stochastic variations or subsampling error inherent to low template concentration. 31Due to a limited sample volume and requirement for a higher CMV titer for sequencing (≥10 4 IU/mL), the samples were not retested to resolve the discrepancy.
It is worth noting that the practice of discrepant resolution, that is, revising original data following a repeat testing of discordant samples, has recently been questioned. 32The significance of the discordant results is uncertain, but is unlikely to be a major concern as CMV shedding in urine and saliva is often at high levels in cCMV, and the cobas ® CMV is unlikely to miss a detection due its high sensitivity. 33,344][35][36] To avoid confusion with low-level false positives, saliva swabs should be taken ≥1 h after breastfeeding. 36,37[40] Diagnostic laboratories should also optimize preanalytical handling of samples to avoid false negatives due to CMV DNA degradation. 21,38,39r the detection of CMV shedding in adult females, the cobas ® assay also has demonstrated satisfactory concordance with comparator assays for urine, saliva, and vaginal secretion.As aforementioned, a discrepant resolution was not performed partially due to limited sample availability but also to avoid analysis bias. 32Although detecting CMV shedding is not part of routine practice, clinicians and/or pregnant women do request CMV PCR to support patient discussions about current or future pregnancies.[44] In agreement with analytical performance data and the manufacturer's cross-reactivity results, a significant molecular interference with non-CMV pathogens or sample matrices (in CPM or eNAT transport media) was not detected. 19The largest difference between the theoretical and the quantified CMV DNA load was 0.53 log 10 IU/mL for CMV-BK virus in saliva sample (an equivalent of 1.77 cycle threshold shift), which is not considered to be a significant interference or competitive inhibition. 45Consistent with our previous studies, a PCR inhibition caused by biological samples was not observed in the current study, presumably due to the use of validated swab types and transport media that permit an optimal stabilization and recovery of CMV DNA. 21,39Therefore, the lower NPA between the cobas ® CMV and reference laboratory M assays for adult urine samples was unlikely to be due to CMV DNA degradation or PCR inhibition, and more likely to be due to differences in extraction efficiency between the cobas ® 6800 and Kingfisher Flex extractors as reported. 21Nevertheless, we cannot rule out the possibility of the differences in detection capability between the assays as a more concentrated eluate and a higher ratio of PCR input were used in the cobas ® CMV assay.The possibility of detecting small amplicons or fragmented CMV DNA by the cobas ® assay also cannot be excluded, though unlikely as a similar rate of disagreement was not observed for saliva and vaginal secretion. 46,47 our knowledge, this is the first study that investigated the clinical utility of a rapid and fully automated CMV PCR to support T A B L E 3 Performance of the cobas ® CMV for adult urine, saliva, and vaginal secretion.timely decision-making and management of congenital and prenatal CMV infection.Although Roh and colleagues also investigated the utility of the cobas ® CMV assay to detect and quantify CMV DNA in urine samples, the group did not assess saliva and vaginal secretion samples. 48In addition, differences in methodology and the population investigated prevent a direct comparison with our results.
Nonetheless, both studies have demonstrated the utility of the cobas ® CMV assay for samples to inform patient care.Compared with reference laboratory M and R, the cobas ® CMV PCR test demonstrates several significant improvements in laboratory workflow and efficiency.These include (1) a shorter turnaround time due to the removal of the need to refer samples to, and a delay in receiving results from, reference laboratories, (2) a reduction in the number of user's intervention and risk of errors as samples are pipetted directly from primary containers and processed by fully automated sample-to-result processes, and (3) the removal of the need to remove reagents for storage which reduces hands-on time and prevents contamination (Supporting Information: Table S1).
These improvements will support a timely management of patients and universal screening for cCMV. 12,13However, the benefits are contingent on a laboratory access to the cobas ® CMV assay and a minimum of 350 µL sample volume, which may not be realistic for some laboratories or patients such as low birth weight infants.In addition, as with all automated solutions, loss of clinical samples is possible due to instrument failure.Fortunately, the failure rate of the cobas ® instrument or assay is very low in our laboratory.
The present study has limitations.First, our laboratory method evaluation study cannot be regarded as a true diagnostic accuracy study.The diagnostic performance of the cobas ® CMV needs to be verified in clinical studies with predefined eligibility and inclusion criteria that fulfill the requirements of the Standards for Reporting Diagnostic Accuracy Studies (STARD). 49However, before such studies can be performed, a consensus definition of CMV shedding will first need to be established.Similarly, the analytical performance data also need to be confirmed in larger studies.Notwithstanding the limitations, we believe our results may inform future clinical and laboratory studies as currently there is a paucity of data on the validation and evaluation of the cobas ® CMV in neonatal and adult women cohorts.Second, amniotic fluid has not been assessed in the current study due to a limited availability of clinical samples and the administrative and financial constraints to source amniotic fluid.As the detection of CMV DNA in amniotic fluid is vital for the diagnosis of cCMV in pregnancy, validation of this sample type on a rapid and accurate assay such as the cobas ® CMV is warranted. 8[42][43][44] Finally, our performance data were based on the use of the described sample collection devices and assays, and cannot be generalized to other sampling and PCR methods.
Laboratories that employ different sample collection devices and/or nucleic acid amplification technologies will need to conduct their own performance validation and evaluation before implementation into routine use. 20 conclusion, we have demonstrated the analytical and diagnostic performance of the cobas ® CMV for the detection and quantification of CMV DNA in urine, saliva, and vaginal secretion.The assay demonstrated excellent performance and is suitable for a rapid, accurate, and efficient laboratory investigation of CMV infection in neonates and adults.Further clinical and laboratory studies to confirm the clinical utility of the cobas ® CMV assay are warranted.

2 |
MATERIALS AND METHODS 2.1 | CMV PCR assays cobas ® CMV PCR was performed on the cobas ® 6800 instrument in line with the manufacturer's instructions (Roche Diagnostics).Briefly, samples (urine and vaginal swabs preserved in cobas ® PCR media [CPM] [Roche Diagnostics], and saliva swabs preserved in eNAT ® PCR media [eNAT] [Copan Italia SpA, Italy]) were loaded directly onto the instrument following homogenization and removal of swabs.

3. 7 |
Performance evaluation of the cobas ® CMV for adult urine, saliva, and vaginal secretion One hundred and forty-nine urine, 145 saliva swabs, and 161 vaginal swabs from 36, 60, and 43 adult females, respectively, were tested by the cobas ® CMV in parallel with reference laboratory PCR assays (Supporting Information: Tables