Development,Testing,and Calibration of LINCS: A New Microsimulation Model of Maternal and Fetal Cytomegalovirus Infection in the United States

Abstract

Background. Congenital cytomegalovirus (cCMV) is a leading cause of birth defects and the most common cause of nongenetic sensorineural hearing loss in children. There is a lack of decision-modeling frameworks that can project cytomegalovirus (CMV)–related patient outcomes and inform health policy. We created, tested, and calibrated a model of CMV acquisition and transmission in pregnancy using linked mother–infant dyads. Methods. We developed the Linking INfants and Mothers in Cytomegalovirus Simulation (LINCS) dyad-level Monte Carlo microsimulation model of CMV infection among pregnant people and fetuses throughout pregnancy. We parameterized the model with data from North America and Europe from the existing literature, implemented rigorous code-testing procedures, and calibrated a key set of parameters to match the model output to external data on cCMV prevalence and symptom risk. Results. A fully parameterized model for CMV among pregnant people and fetuses was developed, and the model code was confirmed to perform as specified. The calibration procedure identified parameter sets that generated model output closely matching the target values from the available data on cCMV prevalence and symptom risk. Conclusions. The LINCS model’s ability to simulate the natural history of CMV infection during pregnancy was described and demonstrated, and the model was tested and calibrated to ensure proper functioning. Base-case parameters were derived for CMV infection natural history to be used in future decision analyses of CMV testing and treatment strategies.

Highlights

Congenital cytomegalovirus (cCMV) is a leading cause of birth defects, but there are limited empirical data available and conflicting guidelines around the best approach to prevention, screening, and treatment. The Linking INfants and Mothers in Cytomegalovirus Simulation (LINCS) model is a comprehensive decision analytic model for cCMV, capturing the full dynamics of the incidence and transmission of CMV among both pregnant people and infants using a microsimulation model centered around the mother–infant dyad.

Despite limited data around cCMV, key parameters in the LINCS model were successfully calibrated to available estimates of cCMV prevalence and symptom risk.

The LINCS model can be used to project outcomes under available strategies to prevent, screen for, and treat cCMV, providing a robust framework to conduct decision analyses, including cost-effectiveness and comparative effectiveness analyses, and inform better policy and clinical guidelines to reduce the impact of cCMV.

Keywords

cytomegalovirus congenital cytomegalovirus simulation modeling maternal health calibration

Congenital cytomegalovirus (cCMV) is the most frequent infectious cause of birth defects and the leading cause of nongenetic sensorineural hearing loss in newborns and infants.¹ Pregnant people are often exposed to and infected with CMV during pregnancy (“maternal infection”), frequently through contact with saliva or urine from children in daycare or preschool.² CMV infections in adults often present with minimal or no symptoms, leading to frequently missed diagnosis and treatment of infection during pregnancy and increasing the risk of birth defects or other fetal complications. It is estimated that 0.3% to 0.6% of all newborns in the United States are born with cCMV, which in 2022 would have been between 10,000 and 22,000 cCMV cases in newborns.³ Of infants with cCMV, up to 30% are born with or will go on to develop cognitive impairment, vision loss, cerebral palsy, or hearing loss, leading to a demonstrated substantial need for costly health care services across settings.^3–5

Despite the major health and economic burdens of cCMV, the best practices for preventing, diagnosing, and treating CMV in pregnant people and infants remain uncertain.⁶ Interventions to reduce risks of maternal infection, vertical transmission (VT), and disease severity among affected newborns have recently emerged,⁷ but the lack of accurate diagnostic tools and the uncertain benefits of treatment pose challenges in the prevention and management of cCMV. Some organizations such as the European Society for Clinical Virology-European Congenital CMV Initiative (ESCV-ECCI) and the Society of Obstetricians and Gynaecologists of Canada (SCOG) recommend screening for CMV serostatus among all pregnant women.^8,9 However, current guidelines from major US professional societies do not recommend universal cCMV screening or treatment in pregnancy, primarily due to concerns about imperfect diagnostic tests and limited confidence in trial data supporting novel therapies for the treatment of maternal or fetal CMV infection.^10,11

Disease simulation modeling can synthesize the best available data from multiple sources, project long-term outcomes beyond the horizon of clinical trials or cohort studies, evaluate multiple strategies simultaneously, and explicitly evaluate the tradeoffs in benefits and harms.^12–15 Modeling can add substantial value to traditional study designs such as trials or observational cohorts and can inform the development of clinical care guidelines.^16–18 However, there are limited published models of CMV infection that use discrete time step agent-based microsimulation methods, with previous studies mostly focused on vaccination.^19–21 Previous models capable of simulating CMV and cCMV screening strategies primarily use deterministic or probabilistic state-transition decision trees.^22–26 To our knowledge, no agent-based discrete-time microsimulation models have been developed to simulate such a wide array of testing strategies or emerging therapies for maternal, fetal, or infant cCMV. Agent-based microsimulation modeling has the advantage of allowing for more agent-specific heterogeneity than with cohort-based models, and the discrete-time structure allows for explicit control over the timing of events.²⁷ There are very few models focusing on CMV acquisition during pregnancy in the United States setting.^22,23 Although all modeling studies are necessarily limited by key assumptions, previous prenatal CMV modeling studies suggest that prenatal screening during the first trimester for CMV would be cost-effective if treatment with valacyclovir followed a diagnosis of a primary maternal infection in the first trimester.^22,23,28 Unlike these previous models of CMV testing and treatment during pregnancy in the United States, the Linking INfants and Mothers in Cytomegalovirus Simulation (LINCS) model also accounts for nonprimary infection (reinfection or reactivation), which can, in theory, affect the utility and therefore cost-effectiveness of maternal CMV tests, as risks of VT may be different for nonprimary infections. Although nonprimary infection is frequently described as conferring markedly lower risk of VT, empirical data on differences in transmission dynamics between primary and nonprimary infection are very limited.^29,30 The model allows us to perform calibration and sensitivity analyses on this uncertain parameter. The discrete time step nature also allows for the simulation of serologic testing–based algorithms, such as immunoglobulin G (IgG) avidity, which can approximately estimate the timing of infection. Since key cCMV outcome risks can be affected by type (primary v. nonprimary infection) and timing of infection, identifying infection type and timing is important to clinical decisions. This model structure allows us to simulate how different testing algorithms interact with the kinetics of viral load and serologic response throughout the course of pregnancy.

Our objective was to calibrate the key parameters of the model to data on the epidemiology of cCMV in the United States as a foundation to conduct future prenatal CMV screening and treatment analyses and project clinical outcomes of CMV infection during pregnancy, primarily within the United States but among other populations of interest as well.

Methods

Overview

We developed the LINCS microsimulation model to project clinical outcomes and costs among pregnant people at risk for CMV infection during pregnancy and their infants. The LINCS model can be used to evaluate current and novel testing and treatment strategies in decision analyses, including cost-effectiveness and comparative effectiveness analyses. Cohort characteristics in the model can be adjusted to reflect the epidemiological dynamics and demographic characteristics of different populations. We derived model input parameters from cohort studies and clinical trials, including rates of nonprimary CMV infection in pregnancy, accuracy of diagnostic tests for CMV, and effectiveness of currently available treatments (Table 1). We also used a Bayesian calibration approach to estimate key uncertain model inputs: we used published estimates of neonatal cCMV prevalence and symptomatic cCMV risk in North American settings as calibration targets and calibrated 6 model parameters, including rates of primary (first-ever) CMV infection in pregnancy and risk of VT following primary and nonprimary infection. We verified the code used to implement the model structure and checked for errors by comparing model outputs to the data used as model inputs, evaluating the model at extreme parameter values, and inspecting traces of individual patient clinical trajectories. Additional code quality testing strategies were performed to verify that each part of the model and the interactions between them were behaving as specified.

Table 1

Selected Data Inputs for the LINCS Model of CMV in Pregnancy

Input	Value	Source
Cohort
Maternal age, y, mean (SD)	26 (5)	31
Maternal primary CMV incidence rate, per week, %	Included as calibration parameter; final calibrated values shown in Table 3
Maternal CMV seroprevalence, %	Calculated based on calibrated primary CMV incidence rate; final calibrated values shown in Table 3
Probability of vertical transmission (by maternal infection type and timing), %	Included as calibration parameter; final calibrated values shown in Table 3
Proportion of cCMV that is symptomatic at birth, by timing of maternal infection, %	Included as calibration parameter; final calibrated values shown in Table 3
Weekly probability of premature birth, range by maternal age, %	Weeks 24–31:Age <24: 0.7%Age 25–29: 0.6%Age 30–34: 0.6%Age 35–39: 0.8%Age >40: 1.2%Weeks 32–36:Age <24: 4.7%Age 25–29: 4.1%Age 30–34: 3.9%Age 35–39: 4.4%Age >40: 5.2%	32
Probability of fetal demise, range by maternal age in years, %	Before 20 wk (miscarriage):Age <20: 16.2Age 20–24: 11.2%Age 25–29: 9.7%Age 30–34: 10.8%Age 35–39: 16.7%Age 40–44: 33.2%Age >45: 56.9At/after 20 wk (stillbirth):Age <20: 0.2%Age 20–24: 0.2%Age 25–29: 0.3%Age 30–34: 0.3%Age 35–39: 0.4%Age 40–44: 0.3%Age >45: 0.3%	33
Baseline prenatal visit schedule
Obstetric visit frequency, by trimester	Weeks 6–28: 1/moWeeks 29–36: 1–2 wkWeeks 36–39: 1/wk	34
Timing of first anatomic scan ultrasound	20 wk	34
Testing
Assays to detect maternal primary CMV infection during pregnancy
Serum IgM sensitivity/specificity, %	90/96	35
Serum IgG sensitivity/specificity, %	97/96	36
Serum IgG avidity sensitivity/specificity %	94.3/100	37
Serum CMV DNA PCR sensitivity/specificity, %	94/99	38
Assays to detect fetal cCMV infection
Amniotic fluid CMV DNA PCR sensitivity/specificity, %	86/100	39
Probability of follow-up testing after abnormal ultrasound result, %	60–100	Assumption
Probability of follow-up testing after abnormal IgM–IgG result, %	95–100	Assumption
Therapies
Valacyclovir for primary infection in periconception and first trimester, odds ratio of vertical transmission risk	0.34	40

CMV, cytomegalovirus; Ig, immunoglobulin; LINCS, Linking INfants and Mothers in Cytomegalovirus Simulation; PCR, polymerase chain reaction; SD, standard deviation.

Model Structure

Mother–infant dyads

The LINCS model is a dyad-level Monte Carlo microsimulation model with a weekly time step that begins for each mother–child pair at conception and ends at delivery. All patients are modeled as a linked mother–child dyad throughout pregnancy and delivery and enter the simulation at the time of conception (week 2 of gestational age). The representation of the mother and child as linked allows for more detailed simulation of both biological and clinical interactions between the mother and fetus during pregnancy and helps account for the interactions between different prevention, screening, and treatment strategies in decision analyses. A literature review found that previous microsimulation models of CMV in pregnancy modeled the mother and child separately, which limited the ability of those models to assess interactions between strategies involving both the mother and child.^24,26,28,41

Maternal CMV infection

At model start, all pregnant people are assigned an age and previous CMV infection history. CMV infection history includes infection that occurred remotely before conception (for this analysis, >3 mo before conception), infection that occurred recently preconception (for this analysis, ≤3 mo to 3 wk prior), infection that occurred periconception (within 3 wk prior to conception), or no prior CMV infection. This differentiation enables the model to account for clinical and diagnostic implications of type and timing of CMV infection (and can be varied as needed). For example, this structure can be used when assessing serologic tests that do not reliably distinguish between infection 3 to 12 mo prior to conception (conferring no risk to the fetus) or <3 mo (when the risk of cCMV is present, conferring a substantial risk to the fetus).^29,42,43

Pregnant people with a recent preconception or periconception infection are assumed not to be at risk for another maternal CMV infection during pregnancy. Although data on strain-specific immunity are sparse, we assume a 12-mo duration of protection from reinfection.⁴⁴ All others face a weekly probability of CMV infection. Those with remote preconception infection are at risk for nonprimary CMV infection (reinfection or reactivation); those with no prior CMV infection are at risk for primary infection (Table 1).⁴³

Vertical transmission

The risk of vertical transmission (VT) to modeled fetuses depends on the type and timing of maternal CMV infection. With recent preconception or periconception maternal infection, simulated fetuses face a risk of VT modeled to occur at week 6, based on expert opinion that sufficient embryonic development has occurred to permit transmission.^45–47 For maternal infection that occurs during pregnancy (either primary or nonprimary infection), modeled fetuses face a 1-time risk of VT 6 wk after maternal infection occurred, allowing time for maternal viremia, placental infection, and fetal infection. VT risks are lower for maternal infections that occur in the first trimester compared with later in pregnancy, although there is a higher risk of symptomatic disease if cCMV does occur following first trimester maternal infection. VT risks are higher for primary compared with secondary maternal infections.

The model divides cCMV into 5 mutually exclusive “phenotypes” to reflect multiple possible manifestations of cCMV for infants/children, including 1) fully asymptomatic cCMV infection at birth and for life, 2) transient laboratory abnormalities or fetal growth restriction (FGR) at birth, with no lifelong symptoms, 3) hearing loss present at birth or developing later (these infants may additionally develop vestibular dysfunction [impaired balance and gait], speech and language delay, or sensory integration disorders, which are not necessarily caused by hearing loss but are associated with cCMV infection^48–50), 4) mild to moderate neurodevelopmental delay (these infants may have mild clinical symptoms [hypotonia, hepatosplenomegaly] and/or nonspecific neuroimaging findings and may develop sensorineural hearing loss [SNHL]), and 5) severe neurodevelopmental delay (e.g., severe autism; these infants are nonambulatory and/or nonverbal lifelong and may have extensive abnormalities on neuroimaging, clinical signs of severe cCMV [e.g., microcephaly, severe FGR], with or without SNHL and/or severe laboratory abnormalities [e.g., thrombocytopenia, neutropenia, jaundice]).² For this analysis, we populated phenotypes based on visible symptoms at birth, in 2 categories: asymptomatic (1 and 2) and symptomatic (3, 4, and 5), although the model program is flexible to simulate additional phenotypes.⁴³

Diagnostic testing

Maternal CMV and fetal cCMV can be diagnosed via user-specified prenatal assays. Tests to detect maternal CMV infection include serum assays such as CMV DNA polymerase chain reaction (PCR), IgM antibody, IgG antibody, and IgG avidity (feasible to assess when IgG is present; greater avidity of the IgG antibody to CMV antigens reflects longer time since infection occurred).²⁹ We derived the sensitivity and specificity for each of these diagnostic tests from the published literature (Table 1). In the model, these sensitivity and specificities are applied alongside modeled viral load and serology kinetics in pregnant people (Figure 1). For example, following maternal infection, CMV viremia is modeled to occur after 1 wk and resolve by 3 wk; CMV DNA PCR can detect this with the listed sensitivity. Detection of CMV DNA in maternal serum is generally considered diagnostic of maternal CMV infection. IgM antibody is modeled to develop within 2 wk and persist for a mean of 52 wk after infection (standard deviation [SD] 4 wk). IgG is modeled to develop within 3 wk and persist lifelong. IgG avidity is modeled as “low” when IgG first develops and rises to “high” at 22 wk (SD: 2 wk) after infection, reflecting the cutoff values and estimated timing from commercial assays.⁵¹

Figure 1

Timing of maternal biological markers in primary cytomegalovirus (CMV) infection in the Linking INfants and Mothers in Cytomegalovirus Simulation (LINCS) model. Base-case representation of changes in anti-CMV immunoglobulin M (IgM) and immunoglobulin G (IgG) antibody levels, as well as IgG avidity, following a primary CMV infection.^29,41 IgM production occurs during both primary infection and reinfection and can persist in some cases for more than a year following an acute primary infection.²⁹ IgG antibodies persist for life following acute primary infection, and IgG avidity gradually increases during the months following primary infection. Maternal CMV viremia is detectable by polymerase chain reaction testing between 1 and 3 wk after infection in the base case.^44,52 This timing of when each biological marker meets the threshold for positivity can be defined by the model user.

If VT has occurred, tests that may identify cCMV infection in the fetus include routine diagnostic ultrasound, detailed diagnostic ultrasound, and amniocentesis followed by CMV DNA PCR performed on amniotic fluid.²⁹ The sensitivity and specificity of these tests are also applied alongside biologic events in the fetus.³⁹ Following VT, infant infection can lead to impaired embryogenesis or fetal development, resulting in abnormalities detectable on ultrasound after 18 wk of gestational age.⁵³ These abnormalities are not specific to CMV and can occur with many other conditions, and they also do not occur in all fetal cCMV infections, creating imperfect sensitivity and specificity of ultrasounds for CMV.⁵⁴ CMV DNA may become present in amniotic fluid (which is composed largely of fetal urine) after 20 wk of gestation, allowing for sufficient fetal renal development to permit excretion of DNA in fetal urine, and persists until delivery.³⁹ Detection of CMV DNA in amniotic fluid is generally considered to confirm fetal infection.⁵⁵

The model user can specify the proportion of primary and nonprimary maternal infections that are symptomatic and the proportion of symptomatic maternal infections that are diagnosed as CMV. The user can also specify diagnostic testing algorithms to reflect current clinical care practices or novel diagnostic algorithms. For example, IgM/IgG serologic testing can be offered as part of routine screening of all pregnant people, as recommended by ECCI and SCOG, or as part of focused testing following known exposure, maternal symptoms consistent with CMV, or fetal ultrasound findings concerning for cCMV, as is more commonly done in the United States (Table 1).^6,9,10 Amniocentesis with amniotic fluid CMV DNA PCR can be offered following confirmed maternal primary infection or abnormal findings on fetal ultrasonography. These diagnostic algorithms and the clinical interpretation of test results are shown in Table 2.

Table 2

Diagnostic Algorithms and Interpretations for CMV in Pregnancy and Congenital CMV⁹

IgM	IgG	Avidity	Interpretation	Follow-up Test	Follow-up Result	Follow-up Interpretation
Negative	Negative	—	No prior CMV infection	n/a	n/a	n/a
Negative	Positive	—	Remote PI	No follow-up	n/a	n/a
Positive	Negative	—	Recent PI or FP IgM	Serum CMV DNA PCR	Positive	Recent PI
Positive	Negative	—	Recent PI or FP IgM	Serum CMV DNA PCR	Negative	No prior infection, FP IgM
Positive	Positive	Low	Recent PI	—	—	—
		High	Remote PI, with recent NPI or FP IgM	Amniocentesis with amniotic fluid CMV DNA PCR	Positive	Vertical transmission
		High	Remote PI, with recent NPI or FP IgM	Amniocentesis with amniotic fluid CMV DNA PCR	Negative	No vertical transmission

CMV, cytomegalovirus; FP, false positive; Ig, immunoglobulin; n/a, not applicable; NPI, nonprimary infection; PCR, polymerase chain reaction; PI, primary infection.

Therapies

We calibrated the model assuming no antiviral therapy during pregnancy, reflecting the settings in which the data used as calibration targets were collected. However, for future policy analyses, the structure of the model permits the user to specify therapeutic interventions in pregnancy following diagnosed maternal or fetal CMV infection. For example, treatment with high-dose valacyclovir (8 g/d) is recommended in some settings following confirmed maternal primary infection in the first trimester with the goal of preventing fetal infection; this is continued through the time of amniocentesis.^8,9 We will model the effectiveness of this intervention as a reduction in VT risk (Table 1).⁴⁰ Following confirmed fetal infection, high-dose valacyclovir can also be offered to a pregnant woman to treat fetal infection and reduce the severity of infant cCMV disease.^9,40,56 Although there is no randomized controlled trial evidence currently available to parameterize the effectiveness of this treatment on cCMV severity, the model structure permits sensitivity and scenario analyses on fetal treatment by adjusting the phenotype distribution for newborns with cCMV, shifting the distribution under treatment toward milder, less symptomatic cCMV phenotypes.

Model outcomes

In each week of the simulation, the model tracks true maternal CMV infection status, true fetal cCMV status, true biomarker status (eg, IgM, IgG, DNA presence in maternal serum; DNA presence in amniotic fluid), and fetal survival. At the end of each week, there are 3 possible pregnancy-related health states: continuation of the pregnancy into the following week, live birth (considered premature if before week 37), or fetal demise (considered spontaneous abortion/miscarriage if before 20 wk and intrauterine fetal demise/stillbirth if after 20 wk; the probability depends on true fetal cCMV status and week of gestation). If neither live birth nor fetal demise occurs, the pregnancy continues into the next week (Figure 2). After birth, a user-specified proportion of infants with hearing loss or visible symptoms will be diagnosed with cCMV, and a probability of immediate neonatal death depends on true cCMV status and the week of gestation at which birth occurred. After the entire cohort is simulated, summary statistics are tallied, including number of live births, number of preterm deliveries, neonates with cCMV, distribution of neonatal cCMV phenotypes, number of maternal infections diagnosed during pregnancy, and number of neonates diagnosed with cCMV.

Figure 2

Schematic diagram of the structure of the Linking INfants and Mothers in Cytomegalovirus Simulation (LINCS) model.

Unit Testing and Functional Testing of Model Dynamics

Unit testing and functional testing are code and software development methodologies to ensure code quality and correct function in accordance with specifications. Testing was performed at each step in the development of the simulation model to ensure code quality and proper function. Descriptions of these methods can be found in the supplemental materials.

Model Calibration

Overview of calibration approach

We implemented a Bayesian calibration approach to calibrate 5 groups of key parameters in the model.⁵⁷ Our 2 calibration targets were the observed cCMV prevalence in a universal screening program in Minnesota and the observed fraction of cCMV cases that are symptomatic in the same universal screening program.⁵⁸ The 5 calibrated parameters were: 1) rate of maternal infection and maternal seroprevalence, 2) risk of VT following maternal primary infection (VT:primary), 3) relative risk of VT following nonprimary maternal infection (RR-VT:nonprimary) compared with risk following primary infection, 4) proportion of cCMV that is symptomatic at birth following maternal primary infection (symptomatic:primary), and 5) relative proportion of cCMV that is symptomatic at birth following maternal nonprimary infection (RR-symptomatic:nonprimary) compared with symptomatic cCMV following maternal primary infection. VT and symptomatic cCMV risks are stratified by timing of maternal infection – preconception (1 to 3 mo prior to conception), periconception (0 to 1 mo prior to conception), first trimester, second trimester, and third trimester. Following the guidance outlined by Menzies et al^57,59 for Bayesian calibration using the sampling importance resampling algorithm, we sampled from prior distributions for each of the parameters, calculated the likelihood of the model output given the observed data for the calibration targets, and then resampled from the original parameter draws with replacement using the calibration target likelihood as sample weights to obtain posterior distributions of the model parameters listed above. We provide a summary of this process below, and a detailed description of the derivation of the prior distributions for the input parameters is included in the supplemental materials.

Prior distributions for input parameters

Maternal CMV infection incidence rates and seroprevalence

The incidence rates of primary maternal CMV infection were drawn from serology data collected by the National Health and Nutrition Examination Survey (NHANES) from 1999 to 2004.⁶⁰ To derive age-specific primary incidence rates from the cross-sectional CMV seroprevalence data provided by NHANES (ie, positive or negative for CMV antibodies on the survey date), we estimated an interval censored hazard regression model, since we know incident CMV infection occurred for individuals with CMV antibodies between birth and their age when the data were collected.⁶¹ We fit the hazard regression using natural cubic splines for age with 3 knots, implemented using the flexsurv package in R.⁶² Based on the estimated hazard model, we also calculated seroprevalence given maternal age at model start. To generate a distribution for the incidence rate that we could use as a prior in the calibration procedure, we used a bootstrap approach that accounted for the complex survey design in NHANES: we generated 10,000 replicate datasets using the svrep R package and reestimated the hazard model in each replicate to generate a distribution for the maternal incidence rate and seroprevalence estimates.⁶³

To estimate the incidence rate of nonprimary infection, we calculated the relative risk of nonprimary compared with primary maternal infection using results from Paris et al.⁶⁴ This prospective study recorded 22 primary and 12 nonprimary (defined as >4-fold IgG increase) infections over 36 mo out of 149 and 204 adolescent girls who were seronegative and seropositive, respectively, at baseline. A Poisson regression with an offset term was used to estimate a prior distribution for the relative risk of nonprimary to primary infection.

VT risk and symptomatic cCMV infection risk

The priors for VT:primary and symptomatic:primary were drawn from the results of studies reported by Chatzakis et al.⁴³ We fit Bayesian binomial regressions with study-level random effects to estimate the VT:primary and symptomatic:primary risks stratified by the timing of maternal infection using the rstanarm R package⁶⁵; we used the results of these regressions as the priors for the VT:primary and symptomatic:primary risks. We set the priors for RR-VT:nonprimary and RR-symptomatic:nonprimary as uniform distributions between 0 and 1, based on previous suggestions that the VT risk from nonprimary infection, and the risk of infant symptoms if infection does occur, is lower than from primary infection.^29,30

Calibration targets

Neonatal cCMV prevalence

To derive the calibration target for neonatal cCMV prevalence, we used the results from a universal newborn cCMV screening program in Minnesota reported by Kaye⁵⁸; out of 60,115 tests, they reported 174 positive results within the first 21 d of life, giving an observed prevalence of 0.29%. To compare LINCS output to the Kaye et al data, we first adjusted this observed prevalence for the imperfect sensitivity of the dried blood spot (DBS) test used in the Minnesota screening program in the following way: Dollard et al⁶⁶ separately reported that the relevant type of DBS-based assay correctly identified 41 of 56 cases (73% sensitivity), with perfect specificity. We constructed a likelihood function for the underlying cCMV prevalence in the Minnesota cohort, assuming a binomial distribution with prevalence p for the Kaye et al results and a beta distribution for the test sensitivity s from Dollard et al (with beta distribution parameters α = 42 and β = 16): $L_{p} (p) = \int_{0}^{1} C (60, 115; 174) {(ps)}^{174} {(1 - ps)}^{60, 115 - 174} f (s) ds$ . Here, C(n;k) is the binomial coefficient, and f is the probability density function for the beta distribution. This approach takes into account both the observed prevalence of cCMV from the Minnesota cohort (174/60,115 = 0.289% observed prevalence) as well as the imperfect test sensitivity (73% sensitivity), giving a maximum likelihood estimate for the underlying cCMV prevalence at 0.394%.

Proportion of cCMV that is symptomatic at birth

To derive the calibration targets for the proportions of cCMV infections that are symptomatic at birth, we used results from the Minnesota universal screening program.⁵⁸ Kaye et al reported that out of 176 infants with confirmed cCMV infection and/or disease from the Minnesota universal screening program, 21 (12%) had cCMV disease according to the CDC definition.⁶⁷ Assuming perfect sensitivity and specificity of symptomatic screening, these results provide the following likelihood function for the model-predicted probability r of symptoms at birth among infants with cCMV infection: $L_{r} (r) = C (176; 21) r^{21} {(1 - r)}^{176 - 21}$ .

Overall likelihood function of calibration targets

We define the overall likelihood function for the calibration targets as the product of the individual likelihood functions for the prevalence of cCMV and the proportion of cCMV that is symptomatic at birth listed above (ie, $L_{p}$ and $L_{r} (r)$ ), assuming independence between each target. This is the function used to evaluate the fit of the model output for the proportion of newborns with cCMV and the proportion of cCMV newborns with symptomatic CMV, when compared to the calibration targets.

Calibration runs

We drew 10,000 parameter sets from the prior distributions listed above. We then ran the model for each of these parameter sets; a cohort size of 20 million mother–infant dyads was used (an initial analysis showed that model outcomes were stable across parameters at a cohort size of greater than 15 million). The ages of pregnant people at model start (15 to 49 y) were drawn from a truncated normal distribution with a mean of 26 y and a standard deviation of 5 y,³¹ and the base-case parameter values from Table 1 were used (aside from the parameters being calibrated) with no treatments applied. We simulated only singleton, viable pregnancies for this analysis. All modeled pregnant people were assumed to be otherwise healthy and remain alive throughout the duration of the simulation.

After running the model, we extracted the overall cCMV prevalence at birth and the proportion of newborns with symptomatic cCMV for each run and calculated the corresponding likelihood for each of the 10,000 parameter sets using the overall likelihood function discussed above. We then resampled from these parameter sets (with replacement), using the likelihood values as sampling weights to obtain a posterior distribution of the calibrated input parameters. All calculations were conducted in R 4.4.1.

Results

Unit and Functional Testing

Unit and functional testing were satisfactory to ensure the logic and performance of the model; selected input parameters were all accurately reproduced from the model output within 1% of the true value. Extreme value testing and trace file examination verified that the model was performing as expected. These results are reported in Appendix Table S1.

Calibration

The best-fitting parameter sets generated by the calibration procedures are reported in Table 3. Likelihood values from the simulations to calibrate the prevalence of cCMV at birth are reported in Appendix Table S2. Using the best-fitting parameter sets, the modeled prevalence of cCMV at birth had a mean of 0.398% (compared with the mean target value of 0.394% described in methods above). The mean modeled proportion of symptomatic cCMV was 9.66% (compared with the target value of 12% described above).

Table 3

LINCS Model Calibration Results

Input Parameters
Parameter	Prior Distribution (95% CI)	Posterior Distribution (95% CI)
Maternal primary infection rate at age 26 y (rate of maternal primary infection)	1.96 (1.59, 2.44) cases/100 person-years	1.97 (1.63, 2.38) cases/100 person-years
Maternal seroprevalence at age 26 y	55.8% (52.5, 59.9%)	55.7% (53.1, 59.9%)
Risk of vertical transmission following maternal primary infection, stratified by timing of maternal primary infection (% risk)^a (VT:primary)	Recent preconception: 5.4% (2.9–8.8)Periconception: 17.2% (12.3, 22.9)First trimester: 36.9% (31.5, 43.4)Second trimester: 41.3% (35.3, 48.1)Third trimester: 66.2% (58.6, 73.0)	Recent preconception: 5.6% (3.1, 9.1)Periconception: 17.6% (12.6, 23.0)First trimester: 37.4% (32.0, 43.4)Second trimester: 41.5% (35.3, 47.8)Third trimester: 66.3% (59.5, 72.5)
Relative risk of VT following nonprimary maternal infection compared to risk following primary infection (RR-VT:nonprimary)	0.5 (0.025, 0.975)	0.439 (0.017, 0.957)
Proportion of cCMV that is symptomatic at birth following maternal primary infection, stratified by timing of maternal primary infection (% risk)^a(symptomatic:primary)	Pre-/periconception^b and first trimester: 21.8% (12.7, 39.5)Second trimester: 3.4% (1.0, 8.3)Third trimester: 0.6% (0.002, 3.4)	Pre-/periconception^b and first trimester: 32.6% (19.7, 50.2)Second trimester: 5.7% (1.9, 12.0)Third trimester: 1.2% (0.0037, 6.0)
Relative proportion of cCMV that is symptomatic at birth following maternal nonprimary infection (RR-symptomatic:nonprimary)	0.5 (0.025, 0.975)	0.583 (0.047, 0.978)
Calibration Targets
Outcome	Empirical Target	Model-Estimated Result (95% CI)
Neonatal cCMV Prevalence	0.394%	0.398% (0.332, 0.480)
Proportion of all cCMV infections that are symptomatic at birth	11.9%	9.66% (5.99, 14.49)

cCMV, congenital cytomegalovirus; CI, confidence interval; LINCS, Linking INfants and Mothers in Cytomegalovirus Simulation; VT, vertical transmission.

Estimates for prior drawn from binomial regression of results reported by studies listed in Chatzakis et al.⁴³ Note there are slight differences between the priors listed here and the meta-analysis results reported by Chatzakis et al⁴³ because of differences in the analytic approach.

The pre-/periconception and first-trimester time periods are grouped here because studies were available from Chatzakis et al⁴³ only for the first-trimester time period (when including only prospective studies that used IgG avidity to determine the timing of maternal infection and defined symptoms as severe sensorineural hearing loss and/or neurodevelopmental delay). We assume that the symptomatic risk for the pre-/periconception period is the same as it is for the first-trimester period.

In the secondary comparisons, our calibrated input parameters were similar to the mean estimates from the literature. For VT:primary risk, our calibrated values ranged from 5.6% to 66.3%, depending on the timing of maternal infection, which were very close to the mean estimates of 5.5% to 66.2% from Chatzakis et al.⁴³ For the maternal primary infection rate, our calibrated value was 1.96/100PY, compared with the 1.95/100PY obtained from our analysis of the CDC NHANES data.^68,69 For symptomatic:primary, our calibrated values ranged from 1.2% to 32.5%, depending on the timing of maternal infection, which were higher than the mean estimates of 0.6% to 21.8% derived from the studies reported in Chatzakis et al.⁴³

The relative risk of VT from nonprimary maternal infection compared with the risk from primary maternal infection (RR-VT:nonprimary) was calibrated to be 0.438. Similarly, the relative risk of symptomatic cCMV for infant infections that follow nonprimary maternal infection compared with those following primary maternal infection (RR-symptomatic:nonprimary) was calibrated to be 0.583. Thus, for both VT and symptomatic risk, calibrated values for the risks following nonprimary maternal infections were substantially lower than the risks following primary maternal infection.

Discussion

We developed a patient-level simulation model of CMV disease progression, testing, and treatment among mother–infant pairs during pregnancy. We tested and calibrated our modeling framework using published estimates from various cohort studies. The LINCS model presents several novel contributions to the CMV simulation modeling literature. First, the simulation of mother–infant dyads enables us to easily model interactions between maternal and fetal health states and allows for the incorporation of epidemiologic data for both maternal and fetal CMV outcomes during pregnancy. This not only results in more detailed and realistic simulation dynamics but also will facilitate analyses that identify cCMV testing and treatment strategies that are optimal for both the pregnant person and child. The LINCS model structure additionally provides flexibility in its ability to specify testing and treatment algorithms, the characteristics of diagnostic assays, and the effectiveness of therapies, allowing for comparison of the clinical impact of alternative testing and treatment strategies. This flexibility will allow the model to evaluate a range of potential clinical practices and guidelines, including in the context of comparative or cost-effectiveness analyses, to help guide clinicians and policymakers.

Data on many aspects of the biology of cCMV are limited, so we calibrated some particularly uncertain and influential parameters in our model. In our calibration analyses, the best-fitting parameter sets produced model output that matched the targets from the literature well (Table 3). In addition, for the calibration to prevalence of cCMV at birth, we generated likelihood values representing the fit between the parameter sets and targets; in planned policy analyses using the LINCS model, these likelihood values can be used to conduct probabilistic sensitivity analyses that more fully capture the uncertainty in the available data.

Through the calibration, we found that the calibrated risks of VT after maternal primary infection match previously reported values but that the proportion of cCMV infections that are symptomatic when they occur after maternal primary infection needed to be substantially higher than values previously reported in the literature for model output to match the calibration targets. We also found that the risks of VT and the risk of symptomatic cCMV following maternal nonprimary infection were lower than those following maternal primary infection (relative risks of 0.438 and 0.583, respectively) but that these relative risks are higher than values often quoted in the literature.^30,70–72 This may suggest that these underlying biologic risks are higher than the risks that are able to be observed in clinical trials or cohort studies or that definitions of phenotypes used in clinical studies differ from those used in our model. Future analyses should conduct appropriate sensitivity analyses to assess the robustness of model output to changes in these parameters or potentially perform additional calibration if new targets become available.

Limitations

This analysis has several limitations. The lack of comprehensive data for many of the model inputs required deriving data from several different studies, often in different geographic settings. The extrapolation of data between geographic settings was felt to be a reasonable assumption for biological parameters such as length of time of infection and test sensitivities and specificities, which are likely similar across regions and populations. In contrast, the calibration targets for the prevalence of neonatal cCMV and the proportion of symptomatic cases following cCMV infection were drawn from a single US state (Minnesota); these values might differ across populations.^58,73 Future analyses may require further calibration with regional data to obtain appropriate and accurate results. In addition, we do not explicitly model forms of transmission other than vertical. The transmission of CMV to pregnant people is modeled through a calibrated incidence value, not through dynamic transmission modeling. The current LINCS model is thus well-suited to address critical clinical policy questions about testing and treatment for mother–infant pairs, rather than the population-level impact of these strategies on all people who may transmit or acquire CMV or the impact of prevention strategies such as vaccination.^21,74

Conclusion

We report the development, testing, and calibration of the LINCS model of CMV infection during pregnancy. We have demonstrated that the model provides a reasonable simulation of the risks of CMV VT, compared with estimates in the literature. We anticipate future expansions of the model to include details of prevention and treatment as well as long-term infant outcomes, diagnosis, and care. LINCS will provide a framework that can be used in future work evaluating the clinical and economic effects of prenatal CMV screening strategies, VT prevention methods, and novel treatments in development.

Supplemental Material

sj-pdf-1-mpp-10.1177_23814683261444049 – Supplemental material for Development, Testing, and Calibration of LINCS: A New Microsimulation Model of Maternal and Fetal Cytomegalovirus Infection in the United States

Supplemental material, sj-pdf-1-mpp-10.1177_23814683261444049 for Development, Testing, and Calibration of LINCS: A New Microsimulation Model of Maternal and Fetal Cytomegalovirus Infection in the United States by Aaron S. Wu, Elif Coskun, Malavika Prabhu, Emily M. Santos, Fatima Kakkar, Clare F. Flanagan, Caitlin M. Dugdale, Megan Pesch, Andrea L. Ciaranello and John C. Giardina in MDM Policy & Practice

Supplemental Material

sj-xlsx-2-mpp-10.1177_23814683261444049 – Supplemental material for Development, Testing, and Calibration of LINCS: A New Microsimulation Model of Maternal and Fetal Cytomegalovirus Infection in the United States

Supplemental material, sj-xlsx-2-mpp-10.1177_23814683261444049 for Development, Testing, and Calibration of LINCS: A New Microsimulation Model of Maternal and Fetal Cytomegalovirus Infection in the United States by Aaron S. Wu, Elif Coskun, Malavika Prabhu, Emily M. Santos, Fatima Kakkar, Clare F. Flanagan, Caitlin M. Dugdale, Megan Pesch, Andrea L. Ciaranello and John C. Giardina in MDM Policy & Practice

Footnotes

The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Dr. Prabhu reports royalties from UpToDate, is a consultant for BabyList, Inc, and created a webinar on CMV for Medscape. Dr. Kakkar received research support from Altona Diagnostics. Dr. Pesch serves as the Executive Director of the National CMV Foundation (unpaid) and is a paid consultant for Moderna Tx, WebMD, and 2ndMD. No other conflicting interests are reported. The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Financial support for this study was provided by the James and Audrey Foster MGH Research Scholars Award. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the MGH Executive Committee on Research. The funding agreement ensured the authors’ independence in designing the study, interpreting the data, writing, and publishing the report.

Ethical Considerations

The Massachusetts General Hospital Institutional Review Board has determined that this study was not human subject research.

Consent to Participate

Not applicable.

Consent for Publication

Not applicable.

ORCID iD

John C. Giardina

Data Availability

All data used in this article are provided in the article and supplemental material. All materials will be made available upon request to the corresponding author.

References

Centers for Disease Control and Prevention. Congenital CMV infection. May 31, 2019. Available from: https://www.cdc.gov/cmv/clinical/congenital-cmv.html. [Accessed 16 July, 2025].

Pesch

Kuboushek

McKee

Thorne

Weinberg

JB.

Congenital cytomegalovirus infection. BMJ. 2021;373:n1212. DOI:10.1136/bmj.n1212

Pesch

Schleiss

MR.

Emerging concepts in congenital cytomegalovirus. Pediatrics. 2022;150(2):e2021055896. DOI:10.1542/peds.2021-055896

Fowler

KB.

Congenital cytomegalovirus infection: audiologic outcome. Clin Infect Dis. 2013;57(suppl 4):S182–4. DOI:10.1093/cid/cit609

Grosse

Dollard

Ortega-Sanchez

IR.

Economic assessments of the burden of congenital cytomegalovirus infection and the cost-effectiveness of prevention strategies. Semin Perinatol. 2021;45(3):151393. DOI:10.1016/j.semperi.2021.151393

Society for Maternal-Fetal Medicine (SMFM); Hughes

Gyamfi-Bannerman

Diagnosis and antenatal management of congenital cytomegalovirus infection. Am J Obstet Gynecol. 2016;214(6):B5–B11. DOI:10.1016/j.ajog.2016.02.042

Sartori

Egloff

Hcini

, et al. Primary, secondary, and tertiary prevention of congenital cytomegalovirus infection. Viruses. 2023;15(4):819. DOI:10.3390/v15040819

Boucoiran

Yudin

Poliquin

Caddy

Gantt

Castillo

Guideline

no.

420: cytomegalovirus infection in pregnancy. J Obstet Gynaecol Can. 2021;43(7):893–908. DOI:10.1016/j.jogc.2021.05.015

Leruez-Ville

Chatzakis

Lilleri

, et al. Consensus recommendation for prenatal, neonatal and postnatal management of congenital cytomegalovirus infection from the European congenital infection initiative (ECCI). Lancet Reg Health Eur. 2024;40:100892. DOI:10.1016/j.lanepe.2024.100892

10.

Practice bulletin no. 151: cytomegalovirus, parvovirus B19, varicella zoster, and toxoplasmosis in pregnancy. Obstet Gynecol. 2015;125(6):1510–25. DOI:10.1097/01.AOG.0000466430.19823.53

11.

Kimberlin

Banerjee

Barnett

Lynfield

Sawyer

, eds. Cytomegalovirus infection. In: Red Book: 2024–2027 Report of the Committee on Infectious Diseases. 33rd ed. Itasca (IL): American Academy of Pediatrics; 2024. p 344–52. DOI:10.1542/9781610027373-S3_003_015

12.

Hunink

MGM

. Decision Making in Health and Medicine: Integrating Evidence and Values. 2nd ed. Cambridge (UK): Cambridge University Press; 2014.

13.

Roberts

Russell

Paltiel

Chambers

McEwan

Krahn

Conceptualizing a model: a report of the ISPOR-SMDM Modeling Good Research Practices Task Force–2. Med Decis Making. 2012;32(5):678–89. DOI:10.1177/0272989X12454941

14.

Caro

Briggs

Siebert

Kuntz

KM.

Modeling good research practices—overview: a report of the ISPOR-SMDM modeling good research practices task force–1. Med Decis Making. 2012;32(5):667–77. DOI:10.1177/0272989X12454577

15.

Briggs

Weinstein

Fenwick

EAL

, et al. Model parameter estimation and uncertainty analysis: a report of the ISPOR-SMDM Modeling Good Research Practices Task Force Working Group-6. Med Decis Making. 2012;32(5):722–32. DOI:10.1177/0272989X12458348

16.

Francke

Penazzato

Hou

, et al. Clinical impact and cost-effectiveness of diagnosing HIV infection during early infancy in South Africa: test timing and frequency. J Infect Dis. 2016;214(9):1319–28. DOI:10.1093/infdis/jiw379

17.

World Health Organization. Evidence, policy, impact. WHO guide for evidence-informed decision-making. 2021. https://iris.who.int/bitstream/handle/10665/350994/9789240039872-eng.pdf?sequence=1

18.

Mumtaz

Jabbour

Makhoul

Harb

El-Jardali

Modeling COVID-19 vaccine rollout in Lebanon for better impact (full version). American University of Beirut Knowledge to Policy Center. Published online May 10, 2021. https://www.aub.edu.lb/k2p/Documents/Modeling%20COVID-19%20vaccine%20rollout_full%20document_.pdf

19.

Wang

Chen

Ventresca

Optimizing age specific strategies of vaccination for prevention of cytomegalovirus infection in the US using agent-based simulation. Epidemics. 2023;44:100698. DOI:10.1016/j.epidem.2023.100698

20.

Byrne

Coombs

Gantt

Modestly protective cytomegalovirus vaccination of young children effectively prevents congenital infection at the population level. Vaccine. 2022;40(35):5179–88. DOI:10.1016/j.vaccine.2022.07.026

21.

Rozhnova

Kretzschmar

van der Klis

van Baarle

Korndewal

Short- and long-term impact of vaccination against cytomegalovirus: a modeling study. BMC Med. 2020;18:174. doi:10.1186/s12916-020-01629-3

22.

Dzubay

Gagliuso

Arora

Doshi

Caughey

AB.

Universal screening and valacyclovir for first trimester primary cytomegalovirus: a cost-effectiveness analysis. Am J Obstet Gynecol. 2025;233(3):191.e1–9. DOI:10.1016/j.ajog.2025.02.009

23.

El Hachem

Poder

Carey

, et al. Prenatal screening for CMV primary infection: a cost-utility model. BJOG. 2025;132(6):805–15. DOI:10.1111/1471-0528.18080

24.

Albright

Werner

Hughes

BL.

Cytomegalovirus screening in pregnancy: a cost-effectiveness and threshold analysis. Am J Perinatol. 2019;36(7):678–87. DOI:10.1055/s-0038-1676495

25.

Aoki

Bitnun

Kitano

The cost-effectiveness of maternal and neonatal screening for congenital cytomegalovirus infection in Japan. J Med Virol. 2023;95(1):e28391. DOI:10.1002/jmv.28391

26.

Fisher

Miller

Yee

Grobman

Premkumar

Universal first-trimester cytomegalovirus screening and valaciclovir prophylaxis in pregnant persons: a cost-effectiveness analysis. Am J Obstet Gynecol MFM. 2022;4(5):100676. DOI:10.1016/j.ajogmf.2022.100676

27.

Graves

Garbett

Zhou

Schildcrout

Peterson

Comparison of decision modeling approaches for health technology and policy evaluation. Med Decis Making. 2021;41(4):453–64. DOI:10.1177/0272989X21995805

28.

Périllaud-Dubois

Hachicha-Maalej

Lepers

, et al. Cost-effectiveness of screening and valaciclovir-based treatment strategies for first-trimester cytomegalovirus primary infection in pregnant women in France. Ultrasound Obstet Gynecol. 2023;62(4):573–84. DOI:10.1002/uog.26226

29.

Iijima

Pitfalls in the serological evaluation of maternal cytomegalovirus infection as a potential cause of fetal and neonatal involvements: a narrative literature review. J Clin Med. 2022;11(17):5006. DOI:10.3390/jcm11175006

30.

Hadar

Dorfman

Bardin

Gabbay-Benziv

Amir

Pardo

Symptomatic congenital cytomegalovirus disease following non-primary maternal infection: a retrospective cohort study. BMC Infect Dis. 2017;17(1):31. DOI:10.1186/s12879-016-2161-3

31.

Balegamire

Renaud

Mâsse

, et al. Frequency, timing and risk factors for primary maternal cytomegalovirus infection during pregnancy in Quebec. PLoS One. 2021;16(6):e0252309. DOI:10.1371/journal.pone.0252309

32.

Waldenström

Cnattingius

Vixner

Norman

Advanced maternal age increases the risk of very preterm birth, irrespective of parity: a population-based register study. BJOG. 2017;124(8):1235–44. DOI:10.1111/1471-0528.14368

33.

Magnus

Wilcox

Morken

Weinberg

Håberg

SE.

Role of maternal age and pregnancy history in risk of miscarriage: prospective register based study. BMJ. 2019;364:l869. DOI:10.1136/bmj.l869

34.

American Academy of Pediatrics, American College of Obstetricians and Gynecologists, eds. Guidelines for Perinatal Care. 8th ed. Itasca (IL): American Academy of Pediatrics; The American College of Obstetricians and Gynecologists; 2017.

35.

Périllaud-Dubois

Bouthry

Jadoui

Leng

Roque-Afonso

Vauloup-Fellous

Positive predictive values of CMV-IgM and importance of CMV-IgG avidity testing in detecting primary infection in three different clinical settings. A French retrospective cohort study. J Clin Virol. 2020;132:104641. DOI:10.1016/j.jcv.2020.104641

36.

Leruez-Ville

Foulon

Pass

Ville

Cytomegalovirus infection during pregnancy: state of the science. Am J Obstet Gynecol. 2020;223(3):330–49. DOI:10.1016/j.ajog.2020.02.018

37.

Abdullahi Nasir

Babayo

Shehu

. Clinical significance of IgG avidity testing and other considerations in the diagnosis of congenital cytomegalovirus infection: a review update. Med Sci (Basel). 2016;4(1):5. DOI:10.3390/medsci4010005

38.

Berth

Benoy

Christensen

Evaluation of a standardised real-time PCR based DNA-detection method (Realstar^®) in whole blood for the diagnosis of primary human cytomegalovirus (CMV) infections in immunocompetent patients. Eur J Clin Microbiol Infect Dis. 2016;35(2):245–9. DOI:10.1007/s10096-015-2537-0

39.

Dinsmoor

Fette

Hughes

, et al. Amniocentesis to diagnose congenital cytomegalovirus infection following maternal primary infection. Am J Obstet Gynecol MFM. 2022;4(4):100641. DOI:10.1016/j.ajogmf.2022.100641

40.

Chatzakis

Shahar-Nissan

Faure-Bardon

, et al. The effect of valacyclovir on secondary prevention of congenital cytomegalovirus infection, following primary maternal infection acquired periconceptionally or in the first trimester of pregnancy. An individual patient data meta-analysis. Am J Obstet Gynecol. 2024;230(2):109–17.e2. DOI:10.1016/j.ajog.2023.07.022

41.

Phillips

Park

Gantt

Dedhia

The cost-effectiveness of targeted screening for congenital cytomegalovirus in newborns compared to clinical diagnosis in the US. Int J Pediatr Otorhinolaryngol. 2023;166:111450. DOI:10.1016/j.ijporl.2023.111450

42.

Enders

Daiminger

Bäder

Exler

Enders

Intrauterine transmission and clinical outcome of 248 pregnancies with primary cytomegalovirus infection in relation to gestational age. J Clin Virol. 2011;52(3):244–6. DOI:10.1016/j.jcv.2011.07.005

43.

Chatzakis

Ville

Makrydimas

Dinas

Zavlanos

Sotiriadis

Timing of primary maternal cytomegalovirus infection and rates of vertical transmission and fetal consequences. Am J Obstet Gynecol. 2020;223(6):870–83.e11. DOI:10.1016/j.ajog.2020.05.038

44.

Leruez-Ville

How to diagnose prenatal cCMV. Conference abstract presented at: European Congenital Cytomegalovirus Initiative; October 9, 2024; Leiden, Netherlands.

45.

Davis

King

Kourtis

AP.

Cytomegalovirus infection in pregnancy. Birth Defects Res. 2017;109(5):336–46. DOI:10.1002/bdra.23601

46.

Benoist

Leruez-Ville

Magny

Jacquemard

Salomon

Ville

Management of pregnancies with confirmed cytomegalovirus fetal infection. Fetal Diagn Ther. 2013;33(4):203–14. DOI:10.1159/000342752

47.

Salomè

Corrado

Mazzarelli

, et al. Congenital cytomegalovirus infection: the state of the art and future perspectives. Front Pediatr. 2023;11:1276912. DOI:10.3389/fped.2023.1276912

48.

Anindya

Sekartini

Ariyanto

, et al. Cytomegalovirus-reactive IgG correlates with increased IL-6 and IL-1β levels, affecting eating behaviours and tactile sensitivity in children with autism. Biomedicines. 2025;13(2):338. DOI:10.3390/biomedicines13020338

49.

Pollick

Noorulla

Harrison

, et al. Neonatal characteristics and neurodevelopmental phenotypes in congenital cytomegalovirus. Pediatr Res. Published online September 9, 2025. DOI:10.1038/s41390-025-04327-z

50.

Carmona

Gantt

Chicoine

Caron

Kakkar

Renaud

Targeted therapies for behavioral disorders among children with congenital CMV infection. 2023. https://cmv.usu.edu/schedule/grid-details.cfm?pg=none&aid=16088&ts=3342&ty=grid&des=reg. [Accessed 6 November, 2025].

51.

Prince

Lapé-Nixon

Role of cytomegalovirus (CMV) IgG avidity testing in diagnosing primary CMV infection during pregnancy. Clin Vaccine Immunol. 2014;21(10):1377–84. DOI:10.1128/CVI.00487-14

52.

Ziemann

Unmack

Steppat

Juhl

Görg

Hennig

The natural course of primary cytomegalovirus infection in blood donors. Vox Sanguinis. 2010;99(1):24–33. DOI:10.1111/j.1423-0410.2009.01306.x

53.

Pass

Arav-Boger

Maternal and fetal cytomegalovirus infection: diagnosis, management, and prevention. F1000Res. 2018;7:255. DOI:10.12688/f1000research.12517.1

54.

Leruez-Ville

Ren

Magny

, et al. Accuracy of prenatal ultrasound screening to identify fetuses infected by cytomegalovirus which will develop severe long-term sequelae. Ultrasound Obstet Gynecol. 2021;57(1):97–104. DOI:10.1002/uog.22056

55.

Pesch

Saunders

Abdelnabi

Cytomegalovirus infection in pregnancy: Prevention, presentation, management and neonatal outcomes. J Midwifery Womens Health. 2021;66(3):397–402. DOI:10.1111/jmwh.13228

56.

Leruez-Ville

Ghout

Bussières

, et al. In utero treatment of congenital cytomegalovirus infection with valacyclovir in a multicenter, open-label, phase II study. Am J Obstet Gynecol. 2016;215(4):462.e1–10. DOI:10.1016/j.ajog.2016.04.003

57.

Menzies

Soeteman

Pandya

Kim

JJ.

Bayesian methods for calibrating health policy models: a tutorial. PharmacoEconomics. 2017;35(6):613–24. DOI:10.1007/s40273-017-0494-4

58.

Kaye

Notes from the field: universal newborn screening and surveillance for congenital cytomegalovirus—Minnesota, 2023–2024. MMWR Morb Mortal Wkly Rep. 2024;73(32):703–5. DOI:10.15585/mmwr.mm7332a2

59.

Chrysanthopoulou

Rutter

Gatsonis

CA.

Bayesian versus empirical calibration of microsimulation models: a comparative analysis. Med Decis Making. 2021;41(6):714–26. DOI:10.1177/0272989X211009161

60.

Centers for Disease Control and Prevention. NHANES questionnaires, datasets, and related documentation. National Health and Nutrition Examination Survey. https://wwwn.cdc.gov/nchs/nhanes/default.aspx. [Accessed 22 July, 2025].

61.

Zhang

Sun

Interval censoring. Stat Methods Med Res. 2010;19(1):53–70. DOI:10.1177/0962280209105023

62.

Jackson

flexsurv: a platform for parametric survival modeling in R. J Stat Softw. 2016;70:1–33. DOI:10.18637/jss.v070.i08

63.

Schneider

svrep: tools for creating, updating, and analyzing survey replicate weights. Published online May 18, 2025. https://cran.r-project.org/web/packages/svrep/index.html. [Accessed 22 July, 2025.

64.

Paris

Apter

Boppana

, et al. Incidence of cytomegalovirus primary and secondary infection in adolescent girls: results from a prospective study. J Infect Dis. 2023;228(11):1491–5. DOI:10.1093/infdis/jiad182

65.

rstanarm. Bayesian applied regression modeling via stan. https://mc-stan.org/rstanarm/. [Accessed 2 August, 2025].

66.

Dollard

Dreon

Hernandez-Alvarado

, et al. Sensitivity of dried blood spot testing for detection of congenital cytomegalovirus infection. JAMA Pediatr. 2021;175(3):e205441. DOI:10.1001/jamapediatrics.2020.5441

67.

Centers for Disease Control. Congenital cytomegalovirus (cCMV) infection and disease 2024 case definition. February 26, 2024. Available from: https://ndc.services.cdc.gov/case-definitions/congenital-cytomegalovirus-ccmv-infection-and-disease-2024/#print

68.

Petersen

Patel

Abraham

Quinn

Tobian

AAR

. Changes in cytomegalovirus seroprevalence among U.S. children aged 1-5 years: the National Health and Nutrition Examination Surveys. Clin Infect Dis. 2021;72(9):e408–11. DOI:10.1093/cid/ciaa1168

69.

Centers for Disease Control, Prevention (CDC). National Center for Health Statistics (NCHS). National Health and Nutrition Examination Survey Questionnaire (or Examination Protocol, or Laboratory Protocol). Published online 2006. Available from: http://www.cdc.gov/nchs/nhanes.htm; https://cir.nii.ac.jp/crid/1573387450791452800. [Accessed 21 July, 2025].

70.

Scaramuzzino

Di Pastena

Chiurchiu

, et al. Secondary cytomegalovirus infections: How much do we still not know? Comparison of children with symptomatic congenital cytomegalovirus born to mothers with primary and secondary infection. Front Pediatr. 2022;10:885926. DOI:10.3389/fped.2022.885926

71.

Leruez-Ville

Magny

Couderc

, et al. Risk factors for congenital cytomegalovirus infection following primary and nonprimary maternal infection: a prospective neonatal screening study using polymerase chain reaction in saliva. Clin Infect Dis. 2017;65(3):398–404. DOI:10.1093/cid/cix337

72.

de Vries

JJC

van Zwet

Dekker

Kroes

ACM

Verkerk

Vossen

ACTM.

The apparent paradox of maternal seropositivity as a risk factor for congenital cytomegalovirus infection: a population-based prediction model. Rev Med Virol. 2013;23(4):241–9. DOI:10.1002/rmv.1744

73.

Ben Shoham

Schlesinger

Miskin

Kalderon

Michaelson-Cohen

Wiener-Well

. Cytomegalovirus (CMV) seroprevalence among women at childbearing age, maternal and congenital CMV infection: policy implications of a descriptive, retrospective, community-based study. Isr J Health Policy Res. 2023;12(1):16. DOI:10.1186/s13584-023-00566-9

74.

Lanzieri

Gastañaduy

Gambhir

Plotkin

SA.

Review of mathematical models of vaccination for preventing congenital cytomegalovirus infection. J Infect Dis. 2020;221(suppl 1):S86–93. DOI:10.1093/infdis/jiz402

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.26 MB

1.92 MB