Estimated cytomegalovirus seroprevalence in the general population of the United States and Canada
Abstract
Seroprevalence data for cytomegalovirus (CMV), a widespread virus causing lifelong infection, vary widely, and contemporary data from the United States (US) and Canada are limited. Utilizing a modeling approach based on a literature review (conducted August, 2022) of data published since 2005, we determine age-, sex-, and country-specific CMV seroprevalence in the general US and Canadian populations. Sex-specific data were extracted by age categories, and a random-effects meta-regression model was used to fit the reported data (incorporating splines for the US). Seven studies reported US CMV seroprevalence (both sexes, aged 1‒89 years); all used National Health and Nutrition Examination Survey data. Due to limited population-based studies, Canadian estimates were modeled using other limited country data. In both countries, modeled seroprevalence estimates increased with age and were higher in females versus males (US: 49.0% vs. 41.6% at 18‒19 years; 61.5% vs. 50.0% at 38‒39 years; Canada: 23.7% vs. 13.7% at 18‒19 years; 32.6% vs. 22.6% at 38‒39 years). Notably, by young adulthood, one-half of US and one-quarter of Canadian females have acquired CMV. The observed differences in CMV seroprevalence in the US and Canada may partially reflect variations in general population characteristics.
1 INTRODUCTION
Previously described as a “silent burden”, cytomegalovirus (CMV) is a common pathogen affecting people of all ages worldwide.1, 2 During the course of the disease, primary infection leads to latency, which can then reactivate, or alternatively, reinfection may occur with an entirely different strain.3 Typically, CMV infections are subclinical in immunocompetent individuals, although mild symptoms have been reported.3 However, in individuals who are immunocompromised, such as those receiving organ transplants, CMV infection may lead to various outcomes including uncontrolled CMV replication, viremia, systemic infection, and varied end-organ diseases such as pneumonitis, retinitis, hepatitis, gastroenteritis, and even death.4, 5
Similar to other herpes viruses, CMV transmission occurs through person-to-person contact via bodily fluids.6 Notably, during pregnancy, CMV can be transmitted vertically from the placenta to the developing fetus, causing congenital CMV (cCMV) infection, the leading infectious cause of birth defects in the US and Canada. About 1 in 200 babies are born with cCMV, and 1 in 5 babies will have long-term health problems as a result of congenital infection.6-8 cCMV symptoms include microcephaly, low birth weight, rash, jaundice, hepatosplenomegaly, and retinitis.8 Sensorineural hearing loss (SNHL) and neurodevelopmental delays are the most common serious long-term consequences.8 General symptomology of cCMV can be present at birth, but in some instances may not appear until after the postnatal period, which makes screening and diagnosing cCMV at birth a clinical challenge, especially for mild symptoms.8
In adults, CMV seroprevalence ranges from 45% to 100% globally with variations across geographic regions and socioeconomic populations, all influenced by existing risk factors including age, race, and ethnicity.9, 10 In the US, data on CMV seroprevalence in the general population is critically informed by the National Health and Nutrition Examination Survey (NHANES) program. Since the 1960s, NHANES has annually enrolled a nationally representative sample of the US general population, collecting participant data from health interviews, physical examinations, and laboratory analyses providing insights into population health and informing burden of disease, including seroprevalence of infectious diseases such as CMV. In recent years, NHANES has frequently reported CMV seroprevalence in young children11, 12; however, data in adults were last reported 2 decades ago (2004).13 In Canada, other challenges exist, namely data on CMV seroprevalence in the general population are sparse, as national serological surveys do not exist and epidemiological efforts are largely driven by regional academic groups in specialized populations rather than national governmental initiatives.14 A thorough understanding of the national epidemiology of CMV in the US and Canada can support the current understanding of the burden of disease, and provide evidence to support policy changes to improve newborn screening, as well as encourage vaccine development and provide justification for introduction into immunization schedules.15
Given the lack of granular data in the US across different population groups, and the limited population-based data in Canada, we developed country-specific models to estimate CMV seroprevalence in the general population, stratified across the full age spectrum (when possible) and by sex in the US and Canada. These estimates were informed by existing literature, specifically from research that reported population-based estimates. Our study provides novel contributions to the literature and fill in gaps in North American data.
2 MATERIALS AND METHODS
2.1 Study selection
Literature searches were conducted on August 5, 2022, in PubMed and EMBASE (Ovid) and included publications from 2005 to 2022. This time frame, which was selected as the last available CMV seroprevalence data for the general US population, was from NHANES 2004.16 Earlier data were excluded from the search as CMV seroprevalence patterns have likely changed over the last few decades.11, 17, 18 Congress abstracts/proceedings were excluded. Searches were limited to peer-reviewed, English-language publications that reported seroprevalence in humans and included adults. A combination of the relevant subject headings and text words were included; “human CMV”, “cytomegalovirus”, and “seroprevalence” were the primary search terms. Citation records were downloaded into a single EndNote library and duplicates were removed. Editorials, comments, and letters were also removed.
Titles and abstracts of all articles in the search were reviewed. CMV immunoglobulin G (IgG) seroprevalence estimates for the US and Canada were provided in 27 publications which were selected for full text review and data extraction. To reduce heterogeneity, studies were considered sufficiently rigorous if methods for participant inclusion were clear and representative of the underlying source population, and if the numerator was clearly derived from the representative study population. Articles reporting on special populations for which prevalence may not have been representative of the country as a whole were excluded to control for bias. We also used random-effects models in our meta-regression analyses to reflect the study-to-study variation. Articles selected for full text review were used to identify additional literature that may have been missed in the original search. Six US studies with overlap of data (reuse of the same CMV cohort) were excluded after review.
Rigorous US population-based CMV seroprevalence data were available from NHANES and published in 7 unique articles (Supporting information: Table 1).11, 13, 16, 19-22 Rigorous estimates of Canadian CMV seroprevalence were available from data collected from pregnant women in 3 studies (Supporting information: Table 2)14, 23, 24 and from 1 study among Canadian blood donors.25 Since previous reports26-30 indicated CMV seroprevalence data from blood donors were not representative, the study from blood donors was not included.
2.2 Data analysis
For estimating CMV seroprevalence by 2-year age groups, sex-specific data by reported age categories were extracted; midpoint of the age category was assigned as the age for analysis. In general, when 2 or more studies reported population-based estimates of CMV seroprevalence for the same country, the model included a random-effect term for the study. Using restricted maximum likelihood, a random-effects meta-regression was fit to the reported data, with a random-effect term for the study, using the mixmeta function and package (https://cran.r-project.org/package=mixmeta) in R version 4.0.5 (R foundation for Statistical Computing, Vienna, Austria).31 CMV seroprevalence estimates reported in the literature for each country varied greatly, depending on the study and the population. Therefore, the analytic methods for this report were tailored to the available information for each country and are further described in the Results section.
Reported seroprevalence was plotted at the midpoint of the age range for which it applied, separately by sex, where applicable. Modeled seroprevalence (and 95% confidence interval [CI]) was plotted by age, using 2-year intervals between ages, also separately for males and females.
2.3 Data analysis: United States
A regression model with restricted cubic splines for age (6 knots) and a linear term for the last year in which blood samples were collected was used to estimate CMV seroprevalence in the US, but year was not included in the final models, as it was found to have a small, nonsignificant, annual change in seroprevalence, compatible with no change. Splines were created using a priori selected knots and the default settings of the rcsplines. eval function in the R package Hmisc (https://cran.r-project.org/package=Hmisc). Model fit was evaluated using plots of observed and expected seroprevalence and Akaike information criterion (AIC). Although the AIC suggested that fewer knots would have been acceptable, 6 were retained for fitting so that the model-predicted values for each age group would be closer to the observed values reported for that age group. 2020 was the latest year for which samples were collected in the available studies, so we assumed seroprevalence had not meaningfully changed since this time.
2.4 Data analysis: Canada
No estimates of CMV seroprevalence for the general Canadian population were found in the literature; data were reported only for pregnant females. Accordingly, data from other representative countries were used to provide extrapolated estimates for CMV seroprevalence in Canada. Studies from the UK,32, 33 France,34, 35 and Japan27, 36 were used to estimate the difference in seroprevalence between pregnant and nonpregnant females within the age range of 18 to <50 years, as these studies were the only published population-based studies with a clear description of study methods and no evident source of bias, and were assumed to be the most comparable to Canada with respect to the difference between pregnant and nonpregnant women. The summary difference was estimated using a meta-regression model implemented with the mixmeta function with a linear term for age, and fixed-effect terms for country and pregnancy. Assuming the difference in seroprevalence between pregnant and nonpregnant females was approximately the same in Canada as the summary difference in these 3 countries (0.10), the seroprevalence for nonpregnant females in Canada was estimated by subtracting 10% from each age-specific estimate for pregnant females. Similarly, US studies11, 13, 16, 19-22 were used to estimate the differences between females and males, as the United States is close geographically and comparable socioeconomically, and high-quality estimates were available. Assuming the difference in seroprevalence was approximately the same in Canada as in the US, CMV seroprevalence estimates for males were estimated by subtracting 10% from the estimates in females. In the Canadian analysis, CMV seroprevalence was estimated within the ages of 18–49 years, without splines due to the relatively narrow range. Thus, the age-specific estimates for the general Canadian population involved key assumptions or extrapolations: first the difference between pregnant and all women in Canada was assumed to be the same as the summary difference in the UK, France, and Japan; and second, the Canadian male-female difference was assumed to be the same as that in the US.
3 RESULTS
3.1 CMV seroprevalence in the US
Reported and estimated seroprevalences in the US are shown in Figure 1A for females and Figure 1B for males. For both sexes, close agreement was observed between reported and estimated values, reflecting a good fit of the model. The 95% CIs of the estimated values were narrow indicating good fit, which included nearly all the observed values.

Reported and estimated CMV seroprevalence in (A) females and (B) males in the US. Gray shaded area shows 95% CI of the modeled data. CI, confidence interval; CMV, cytomegalovirus; US, United States.
Estimated CMV seroprevalence for all 2-year age groups was higher in females compared with males in the US (Figure 2 and Supporting information: Table 3). In infants aged ≤1 years, estimated seroprevalence was 24.7% in females and 21.9% in males. Seroprevalence estimates increased to 49.0% (females) and 41.6% (males) by age 18 to <20 years, to 63.3% and 51.5%, respectively, by age 40 to <42 years, and to 94.5% and 81.6%, respectively, by the age of 84 to <86 years. The difference between seroprevalence in females versus males increased modestly with age, from 4.6% at age 4 to <6 years, increasing to 7.5% at age 20 to <22 years, and 11.8% at age 40 to <42 years, increasing only slightly thereafter to 12.4% at age 80 to <82 years.

3.2 CMV seroprevalence in Canada
Reported and estimated seroprevalences in pregnant females in Canada are shown in Figure 3. The τ2 (denoting between-trial heterogeneity) was 0.122, and Ι2 (describing the percentage of total variation across studies due to differences between the included trials rather than by sampling error) was 97%, indicating substantial residual variation. Given the need to also extrapolate between countries to obtain estimates for all females aged 18 to <50 years, the 95% CIs of the estimated values were wide, reflecting the uncertainty of the model.


Although data from Canada reporting seroprevalence in males were not identified, we extrapolated from Canadian data on pregnant females, first to the overall Canadian female population, and then, using the difference observed across the NHANES data, estimated seroprevalence in Canadian males. As the model extrapolated a constant difference of 10% between sexes as described above, estimated CMV seroprevalence in Canada was also consistently higher in females compared with males (Figure 4). Estimated seroprevalence was 23.7% (females) and 13.7% (males) in individuals aged 18 to <20 years, with a moderate increase to 37.1% and 27.1%, respectively, by age 48 to <50 years (Supporting information: Table 4).
4 DISCUSSION
In any discussion on CMV burden of disease, understanding of seroprevalence and risk factors, such as age and sex, in the general population is important, especially in the context of the adverse impact CMV has on at-risk populations such as pregnant women and immunocompromised individuals.3, 6, 37 In our models, estimated CMV IgG seroprevalence within the US increased with age and was higher in females than males at all ages. While in Canada, using data from other countries to extrapolate from the limited seroprevalence data in pregnant women, we report that age and sex patterns were like those observed in the US, with higher seroprevalence in females than males and an increase in estimated seroprevalence with age. Necessary differences between data modeling methodology in the 2 countries limit a direct comparison of estimated seroprevalence between them; however, descriptively, the patterns suggest a consistently higher modeled seroprevalence in the US compared with Canada across various age ranges. To our knowledge, this is the first report to provide extrapolated estimates of seroprevalence rates for males and for the general female population in Canada, which is a major step in revealing true population-based estimates in the general population.
Our findings suggest that approximately half of the female population in the US and three-quarters of the same population in Canada enter their reproductive years immunologically naïve to CMV. If they become pregnant, these individuals are at risk for primary CMV infection, which in turn places their pregnancy at risk for transmitting CMV to their developing fetus leading to cCMV.14, 20, 21, 23, 24 It is worth noting again that CMV is the causative agent of cCMV, which is the leading infectious cause of congenital defects and nongenetic cause of SNHL in children and a significant cause of neurodevelopmental delay.2 One Canadian study in our analysis reported that 2.1% of women seroconverted during their first and third trimesters, highlighting that a high proportion of pregnant women are at risk for CMV infection.23 Despite worse health outcomes associated with seroconversion during pregnancy, individuals immunologically primed against CMV infection (i.e., CMV seropositive) can also give birth to newborns with cCMV.38 Reports suggest an approximately 10-fold difference in transmission rates between CMV-seronegative and CMV-seropositive mothers (30%‒40% vs. 3%‒4%)38; other reports suggest that 32% of primary and 1.4% of recurrent CMV infections during pregnancy lead to congenital infection.39 Previous reports have confirmed that maternal CMV seroprevalence is a significant determinant of cCMV, and while we report higher modeled seroprevalence among US females compared with Canadian females across various age ranges, cCMV birth prevalence is reported to be 0.36%‒2.45% in the US and 0.42‒0.55% in Canada, suggesting other risk factors may be driving these differences.40
Age and sex are strong indicators of CMV serostatus. The increase in CMV seroprevalence with age observed in our analysis of US and Canadian data is consistent with studies in other countries.29, 32, 33, 35 In one of the first reports from NHANES, with samples taken between 1988 and 1994 from more than 20 000 individuals aged ≥6 years, Staras et al also identified the linear relationship between age (and sex) and CMV seropositivity, reporting that females were more likely than males to be CMV seropositive when adjusting for age only, or age in combination with other sociodemographic risk factors.13 Higher CMV seroprevalence in females compared with males has also been observed in other countries, including Germany and France.29, 35 In our report, the gender difference in CMV seroprevalence was maintained among the youngest age groups, which is generally consistent with previous studies.11, 17, 41, 42 However, an understanding of factors that may underlie these differences between genders early in life is limited.11 While potential risk factors among adolescent populations aged 12 to 17 years, including exposure to group living situations and exchange of saliva by kissing or sharing utensils and make-up, have been evaluated and found to be insignificant contributors to gender differences, higher CMV seroprevalence among young females may be due to increased exposure to infected children ≤3 years of age.41, 42 Females are more prone than males to exposure to very young children who are infected with CMV through greater involvement in their care in household or group settings.41, 42 Exposure to young children in the home, feeding, and changing diapers are activities associated with increased CMV infection among young females,42 with children's saliva as the primary source of CMV on home surfaces.43
Previous studies have also shown that CMV seroprevalence in the US and Canada varies substantially by race and ethnicity. NHANES data suggest that seroprevalence was lowest in the non-Hispanic White population and highest in the Hispanic/Mexican-American population11, 13, 16, 21, 22; in Canada, nonwhite pregnant females were twice as likely as White pregnant females to be seropositive (odds ratio 2.0 [95% CI: 1.1‒3.8]).14 These trends confirm other reports of racial and ethnic differences in cCMV birth prevalence estimates in the US; Black infants were reported to have the highest cCMV birth prevalence compared with other racial and ethnic groups.44 In Canada, no CMV seroprevalence or cCMV birth prevalence estimates have been reported for indigenous groups. While beyond the scope of our objective to model age- and sex-specific CMV seroprevalence estimates in the US and Canada, an interesting future analysis would include consideration of the impact of race and ethnicity in the model.
Other risk factors for CMV identified in North America and in other regions, such as France, include an inverse association between a higher level of education and CMV seroprevalence16, 21-23, 35 and lower household income level associated with increased risk of CMV seropositivity.13, 16, 22-24 Exposure to young children, especially those attending daycare, is considered an important source of CMV transmission through contact with saliva, urine, blood, and tears of seropositive individuals. Consequently, CMV is considered an occupational risk for daycare workers.45 For example, a Canadian study of approximately 2000 pregnant females found that working in daycare was associated with a > fourfold risk of being CMV seropositive.23 Several studies in the US and Canada and other regions have generally reported increased CMV seroprevalence in females who have more than 1 child,14, 21, 23, 33, 34 presumably because some of these children would be attending daycare. Indeed, a global systematic review and meta-analysis reported a significant association (odds ratio 2.69, 95% CI: 1.68‒4.30) between CMV infection in children and daycare center attendance.46
Our analysis focused on modeling CMV seroprevalence in the US and Canada. Although NHANES is considered by many as the gold standard for understanding burden of disease for many conditions in the US, some NHANES data, especially for CMV, are now outdated, particularly in adults. In fact, as of 2003‒2004, NHANES no longer includes routine laboratory assessment of CMV seropositivity in adults, and only includes this measure in young children aged 1‒5 years. Unfortunately, in Canada, the reported CMV seroprevalence data are very sparse. This translated to the major limitation of our analyses, which necessitated various extrapolations to estimate seroprevalence specifically in nonpregnant females and in males. The resulting model for the Canadian data therefore has a greater degree of uncertainty when estimating CMV seroprevalence in these populations.
In conclusion, our understanding of the seroprevalence of CMV in the general population in the US and Canada is currently limited. However, the present analysis demonstrates how existing data can be utilized to provide additional insights to estimate data in specific countries, as shown here for the US and cautiously based on extrapolation, for Canada. As development efforts for an effective CMV vaccine continue, this study serves as a call for more comprehensive data to inform public health policies and interventions, ultimately aiming to mitigate the ‘silent global burden’ of cCMV.2
AUTHOR CONTRIBUTIONS
W. Dana Flanders contributed to the concept design of the study, analysis/interpretation of data, writing/review/intellectual contributions, and approved the final draft of the manuscript. Cathy Lally, Anne Dilley, and John Diaz-Decaro contributed towards the concept design of the study, data collection, analysis/interpretation of data, writing/review/intellectual contributions, and approved the final draft of the manuscript.
ACKNOWLEDGMENTS
Medical writing and editorial assistance, under the direction of the authors, were provided by Sarah Feeny, BMedSci, of MEDiSTRAVA in accordance with Good Publication Practice (GPP 2022) guidelines and funded by Moderna, Inc. The authors also thank Colin Kunzweiler MS, PhD (Moderna, Inc.), for their review and comments on early drafts of the manuscript and Kosuke Kawai (Moderna, Inc. [at the time of project initiation]) for their early contributions to the systematic literature review. The data summarized in this review are from published articles and are publicly available.
FUNDING
This work was supported by Moderna, Inc.
CONFLICT OF INTEREST STATEMENT
W.D.F is a consultant for Epidemiologic Research & Methods, a consulting company with several clients, including Moderna, Inc. C.L. and A.D. are Co-Owners of Epidemiologic Research & Methods, LLC which does consulting work for several companies. J.D.D is an employee of Moderna, Inc., and may hold stock/stock options in the company.
ETHICS STATEMENT
Not applicable.
Open Research
DATA AVAILABILITY STATEMENT
The data summarized from this review are from published articles and are publicly available.