Implementation of pooled saliva tests for universal screening of cCMV infection

Congenital cytomegalovirus is the most common intrauterine infection, with worldwide birth prevalence estimated at 6.4–7 per 1,000 and substantial risk of neurodevelopmental disabilities and sensorineural hearing loss. Targeted newborn screening (for infants failing hearing screens, with maternal infection, or clinical suspicion) misses the majority of infected infants who are asymptomatic at birth but at risk for late sequelae. Alternative screening methods have limitations: dried blood spot PCR leverages existing infrastructure but shows variable and suboptimal sensitivity (around 75% on average), while saliva PCR is highly sensitive but prone to false positives due to peripartum contamination from breast milk, necessitating urine confirmation and increasing costs for universal use. Sample pooling (Dorfman method) can increase throughput and reduce resource use, and given cCMV’s typically high saliva viral loads and low birth prevalence, it is theoretically well-suited for pooling with minimal sensitivity loss. Building on prior large-scale SARS-CoV-2 pooling experience, this study aimed to validate and implement an eight-sample pooled saliva RT-PCR pipeline for universal newborn cCMV screening, and to evaluate its sensitivity, efficiency, feasibility, and effect on false positives, compared to prior targeted screening.

The paper situates the work within: (1) epidemiology of cCMV with variable regional prevalence (2–13 per 1,000) and significant long-term morbidity including SNHL; (2) benefits of early diagnosis and antiviral therapy for symptomatic cCMV and early intervention for hearing outcomes; (3) limitations of targeted screening that misses most asymptomatic cases; (4) technical performance of DBS PCR with variable sensitivities and recent estimates around 75%, potentially inadequate for universal screening; (5) saliva PCR as a sensitive reference for targeted screening but with high false positive rates primarily due to breast milk contamination, requiring urine confirmation; (6) pooling theory (Dorfman) and prior small proof-of-concept cCMV pooling studies (hundreds to ~7,000 infants) that mainly established feasibility without comprehensive sensitivity/efficiency analyses; (7) successful large-scale pooling implementations for SARS-CoV-2 testing. These works collectively motivate evaluating pooled saliva PCR as a scalable, accurate universal cCMV screening strategy.

Design: Prospective, population-based cohort at two hospitals of Hadassah Medical Center (seven newborn nurseries, two intermediate units, two NICUs), April 2022–April 2023, following a 3-month pilot (January–March 2022). Inclusion: All newborns with parental written informed consent; specimens collected within 1–3 days of life (all by ≤10 days). Ethics: IRB approval #0272-22-HMO. Screening algorithm: Saliva samples from all consented infants were tested in pools. Eight-sample (1:8) pools were the default; occasionally 5–7 samples per pool or individual testing was used for logistical reasons. Pool testing employed DNA extraction and RT-PCR. If a pool was negative, all constituent samples were deemed negative. If a pool was positive (Ct ≤40 for either CMV target), the pool was opened and all individual samples in that pool were tested by saliva RT-PCR. Infants with a positive individual saliva result underwent confirmatory urine RT-PCR; only urine-confirmed cases were considered true cCMV positives. Laboratory methods: Saliva collected by buccal swab into 3 ml Universal Transport Medium (COPAN); urine collected in sterile bags. Samples stored at 4 °C and tested the same day or within 72 h. DNA extraction from 200 µl sample with Roche MagnaPure, eluted in 50 µl. RT-PCR used 5 µl eluate in 25 µl reactions with TaqMan Fast Advanced Master Mix on QuantStudio 5. CMV targets: glycoprotein B (gB, for quantitation) and immediate early 1 (IE1) genes; internal control: human ERV3. Primer/probe sequences are provided. Assay quantitation was linear over 6 logs with analytic sensitivity 50–100 copies/ml; ≥50 copies/ml considered positive. Pooling operations: Pools of eight were created using a Tecan robotic system; equal volumes pooled to 3.2 ml. DNA extraction and RT-PCR as per individual testing. IT system supported sample tracking, pool composition, retest assignment, reporting, and data analysis. Quality control included ongoing individual retesting of approximately 4.8–5% of negative pools during routine implementation to monitor negative predictive value. Experimental validation: Sensitivity assessed by spiking one positive sample with seven negatives across a dilution series (5,000–250 copies/ml; Ct ~33–37); all detected in pooled testing. Additionally, 160 saliva samples were formed into twenty 1:8 pools (17 with one positive sample each spanning Ct 20–37, three all-negative); all positive pools detected for gB or IE1, and all-negative pools remained negative, indicating no loss of diagnostic sensitivity. Pilot phase parallel testing: During January–March 2022, all saliva samples were tested both via pooling and individually. Of 196 pools, eight were positive and 188 negative. All 1,414 individual samples from negative pools were negative (NPV 100%); seven of eight positive pools (88%) contained at least one positive sample upon deconvolution; all individually positive samples were detected in pooled testing. Theoretical analysis for pool size selection: Retrospective analysis of 2014–2021 targeted testing predicted, for pool size n=8, 99.5% sensitivity (1/188 true positives potentially missed) and filtering of 57% (66/115) of false-positive saliva results. The potentially missed case had very low viral loads in saliva and urine and was asymptomatic at 7 years. Statistical analysis: Pooling efficiency defined as number of samples tested per single RT-PCR reaction. Linear regression (R, lm) constrained slope=1 compared pool Ct vs corresponding individual positive sample Ct to estimate Ct shift; expected shift log2(n)=3 for n=8. Data handling used R packages dplyr, stats, ggplot2.

• Coverage and throughput: 15,805 neonates screened over 13 months (93.6% of 16,884 live births); 15,273 screened via pooling (96.6% of screened; 90.5% of all live births), 532 screened individually due to logistics.
• Prevalence and case detection: 54 infants confirmed with cCMV (urine PCR), birth prevalence 3.4 per 1,000 (95% CI 2.6–4.3). Of these, 30 (55.6%) were identified only because of universal screening and would have been missed by targeted screening; one asymptomatic newborn had intracranial involvement leading to early valganciclovir treatment.
• Pool performance: 1,990 pools were run; 76 pools (3.82%) were positive for IE or gB. Of positive pools, 85.5% contained at least one positive individual sample upon deconvolution; 14.5% yielded no positive individual sample (all had Ct >34 or single-gene positivity), reflecting permissive pool-opening to maximize sensitivity.
• Sensitivity (Ct shift): Among 47 eight-sample pools each containing a single positive sample, the mean Ct increase for the pool vs individual was 3.05 cycles (regression y = x + 3.05; r² = 0.948; p = 7.25×10⁻18), consistent with theoretical log2(8)=3.
• Negative predictive value: Pilot parallel testing—188 negative pools encompassing 1,414 individual samples all negative (NPV 100%). During routine implementation, 92 randomly selected negative pools (4.8% of all negative pools; 736 samples) were opened and tested individually; no true positives were found (one false-positive saliva sample: IE Ct 37, gB negative).
• Efficiency: 15,273 pooled saliva samples tested using 2,578 RT-PCR reactions (including deconvolutions), yielding empirical efficiency of 5.92 samples per RT-PCR reaction. Compared to the previous year’s individual testing (1,275 samples with 1,275 reactions), pooling enabled testing 12-fold more samples with roughly twice as many PCR reactions, sparing approximately 83% of tests.
• False positives filtering (theoretical): Eight-sample pooling predicted to filter 57% of false-positive saliva results from targeted era data.

The implementation of eight-sample pooled saliva RT-PCR for universal newborn cCMV screening achieved high practical efficiency and coverage while maintaining clinically acceptable sensitivity. The observed Ct shift of approximately three cycles matches theoretical expectations for 1:8 dilution and did not materially compromise detection given typically high saliva viral loads in cCMV. Empirical quality control indicated an NPV effectively at 100% in both pilot and routine sampling, supporting the reliability of negative pooled results. Pooling also reduced downstream burden by filtering a substantial proportion of low-level, false-positive saliva results, thereby decreasing unnecessary urine confirmations and associated logistics. Operationally, leveraging an existing robotic pooling pipeline and IT tracking facilitated rapid turnaround (often enabling confirmation and clinical evaluation within 2–3 days of life) and smooth integration into routine practice. The approach enabled near-universal consented screening (~94%), similar to other standard newborn screens, and identified more than half of cases that would have been missed by targeted strategies, including at least one case benefiting from early treatment. Epidemiologically, the observed birth prevalence (3.4/1,000) aligns with Western European estimates and underscores the value of sustained large-scale screening to refine population-level burden estimates, especially in heterogeneous populations like Jerusalem’s.

Pooled saliva RT-PCR using eight-sample Dorfman pooling is a feasible, efficient, and sensitive strategy for universal newborn cCMV screening. In real-world implementation across two hospitals, it substantially increased screening coverage, reduced testing burden by approximately 83%, minimized false-positive confirmations, and maintained a clinically negligible sensitivity loss. This pooling setup can be readily adopted by medium-to-large diagnostic laboratories with similar birth prevalence, providing an expected efficiency of about six samples per PCR reaction. Future work should assess cost-effectiveness, validate scalability across multiple centers and populations, explore subgroup prevalence differences, and conduct longitudinal follow-up to define the burden, risk factors, and outcomes among universally screened infants.

Not all samples within negative pools were individually tested during routine implementation, precluding a full empirical estimate of the false-negative rate (though random deconvolution of ~5% negative pools found no true positives). The study did not perform a formal cost-effectiveness analysis despite the high empirical efficiency. The findings derive from two hospitals in Jerusalem and may not fully represent national or international populations, though the large, heterogeneous cohort supports generalizability. Large-scale multicenter implementation could potentially miss more than a single case, and clinical testing should still be performed individually if suspicion for cCMV remains.