Estimation of the worldwide seroprevalence of cytomegalovirus: A systematic review and meta‐analysis

Cytomegalovirus (CMV) infection does not usually produce symptoms when it causes primary infection, reinfection, or reactivation because these three types of infection are all controlled by the normal immune system. However, CMV becomes an important pathogen in individuals whose immune system is immature or compromised, such as the unborn child. Several vaccines against CMV are currently in clinical trials that aim to induce immunity in seronegative individuals and/or to boost the immunity of those with prior natural infection (seropositives). To facilitate estimation of the burden of disease and the need for vaccines that induce de novo immune responses or that boost pre‐existing immunity to CMV, we conducted a systematic survey of the published literature to describe the global seroprevalence of CMV IgG antibodies. We estimated a global CMV seroprevalence of 83% (95%UI: 78‐88) in the general population, 86% (95%UI: 83‐89) in women of childbearing age, and 86% (95%UI: 82‐89) in donors of blood or organs. For each of these three groups, the highest seroprevalence was seen in the World Health Organisation (WHO) Eastern Mediterranean region 90% (95%UI: 85‐94) and the lowest in WHO European region 66% (95%UI: 56‐74). These estimates of the worldwide CMV distribution will help develop national and regional burden of disease models and inform future vaccine development efforts.


Summary
Cytomegalovirus (CMV) infection does not usually produce symptoms when it causes primary infection, reinfection, or reactivation because these three types of infection are all controlled by the normal immune system. However, CMV becomes an important pathogen in individuals whose immune system is immature or compromised, such as the unborn child. Several vaccines against CMV are currently in clinical trials that aim to induce immunity in seronegative individuals and/or to boost the immunity of those with prior natural infection (seropositives). To facilitate estimation of the burden of disease and the need for vaccines that induce de novo immune responses or that boost pre-existing immunity to CMV, we conducted a systematic survey of the pub- However, in patients who are immunocompromised, CMV replication may be uncontrolled and lead to high viral loads in the urine, which are associated with viraemia, dissemination to multiple organs, and endorgan diseases such as pneumonitis, retinitis, hepatitis, or gastroenteritis. 2 CMV is the most common intrauterine infection and a high priority for vaccine development. 3,4 Disease can occur when pregnant women have active CMV infection with viraemia leading to involvement of the placenta and then the fetus. 5 Intrauterine infection can occur following primary infection, reactivation, or reinfection in a pregnant woman. 6,7 In addition to these direct effects that can be seen in individuals and documented histopathologically in biopsies of organs, CMV is associated with indirect effects which are apparent at a populationlevel rather than an individual-level. This phenomenon was first List of abbreviations: CMV, cytomegalovirus; IgG, immunoglobulin G; MeSH, medical subject headings; NHANES, national health and nutrition examination survey The copyright line for this article was changed on 01 March 2019 after original online publication. described after solid organ transplant where an excess of graft rejection, atherosclerosis, and secondary bacterial or fungal infections was seen among those with past history of detectable viraemia. 8 These indirect effects partly explain the survival disadvantage that is seen in solid organ transplant or stem cell transplant patient populations who are CMV seropositive. 9 A recent randomized controlled trial showed that stem cell transplant patients given prophylaxis with letermovir had improved survival compared to recipients of placebo, so supporting CMV as an underlying contributor to all-cause mortality. 10 Indirect effects may also be seen in people living with HIV, manifesting as more rapid progression to AIDS and/or death. 11,12 Most recently, the indirect effect of excess mortality has been defined in members of the general population without classical risk factors for CMV disease. 13,14 Given the complex natural history of CMV infection, the emerging need to plan the evaluation of candidate vaccines 15

| Search strategy
A systematic review was conducted and reported in line with the criteria set out by the PRISMA guidelines. 16 We reviewed articles pub- In consultation with an expert medical librarian, we developed a search strategy and adapted it to each database. A combination of medical subject headings (MeSH) and free text searches was used to search for terms relating to CMV and seroprevalence (full search strategy available in Appendix A).
Publications were screened and catalogued on Endnote X6. A strategy involving "auto-search" and "hand-search" was used to identify duplicates, before two authors (M.Z. and G.W.) systematically screened the search results independently and applied the inclusion and exclusion criteria.
In brief, studies that included data on CMV seroprevalence (based on detection of CMV specific immunoglobulin G [IgG]), assessed in the general population, women of childbearing age, and blood and organ donors were included in this systematic review. Studies were excluded if they were systematic reviews, surveillance reports, case studies, letters, correspondence, or did not include data on CMV IgG. Studies were also excluded if they reported seroprevalence data on high-risk groups or children only or had a sample size <80, this was to avoid low quality studies and selection bias. A full list of the inclusion and exclusion criteria is available in Appendix B. We translated non-English papers by asking colleagues proficient in the language in question or by using "google translate." For studies from the United States, we considered the national health and nutrition examination survey (NHANES) as a representative sample of the general population. 17 Nevertheless, the same full text screening was applied for all studies from the United States, but only original data from NHANES were included.

| Data extraction
Following full text review, we extracted the following variables from each study: study characteristics (number of participants, study dates, and location), participant characteristics (age, sex, population group), sample size and number of positive individuals, or, when the latter was not specified, CMV seroprevalence. Three population groups were considered: the general population (healthy individuals), women of reproductive age (pregnant women and women of reproductive age), and blood and organ donors. Several studies presented stratified results, which were merged if data allowed. However, three types of stratification did not allow merging: different population groups (general population, women of reproductive age, and/or blood and organ donors); groups with non-representative sample sizes; and different sampling timeframes.

| DISCUSSION
Our results summarize extensive information on the prevalence of CMV in different populations around the world. The results are consistent with a socio-economic link with CMV that has been well established in multiple studies showing that, at any given age, CMV prevalence is higher in individuals of lower socio-economic group. 21 It has also been reported that the children of individuals born in a developing country with a high prevalence of CMV have a lower prevalence once they become established in their adopted country. 22 It therefore seems likely that unidentified cultural and behavioural factors also interact with socio-economic group.
The observed increased seroprevalence for women of reproductive age is likely because of their exposure to children as reported previously. 21,23 However, the reason behind an increased seroprevalence among blood donors was counter intuitive, and the causes are less clear.
These results are important for several reasons. First, they will help to estimate the burden of disease attributable to congenital  Although it is believed that women who acquire primary infection during pregnancy are most at risk of delivering babies severely damaged by intrauterine infection, such women are in a minority in many countries of the world. 24 Instead, most babies are born to women who are seropositive prior to conceiving because of their abundance in the community. Even in countries with relatively low seroprevalence, the majority of babies may be infected as a result of CMV reactivation or reinfection in the mother. 25 A high prevalence of infection in a community therefore contributes to all three types of maternal infection. 26 Vaccine studies have been conducted in seronegative and seropositive women. 27,28 Second, the results may help predict the incidence of CMV infection and disease after solid organ transplant which may originate from the donor or the recipient. 1 Phase 2 randomized clinical trials have been conducted with vaccine candidates given to recipients awaiting solid organ transplant. 22,29 The situation is more complex following stem cell transplantation where vaccine may be given to recipients or donors or both to take advantage of adoptive transfer of immunity from a donor with or without prior immunity. 30,31 Many countries in the world do not undertake organ transplantation or stem cell transplantation because high-technology medicine is not available. However, it is those same countries that often have a burden of HIV infection, which allows CMV to become an important opportunistic infection. CMV may therefore act as an important pathogen in patients with impaired immunity whether this results from iatrogenic or HIV induced immunosuppression. The incidence of congenital or perinatal CMV is also greatly increased in women with underlying HIV infection. 32,33 We anticipate that the results presented here will improve understanding of diseases attributable to CMV in different parts of the world and will allow vaccine manufacturers to consider where to recruit individuals for clinical trials. These may either administer vaccine to seronegatives to prevent primary infection or administer vaccine to seropositives to boost immunity. 15 This work had some limitations. There was a paucity of prevalence studies within some countries worldwide and the quality of reported studies also varied. Many of the studies were focused on adults and did not provide age or sex-specific prevalence. The lack of age stratification may be a cause of variation between studies, as CMV seroprevalence is known to increase with age. 34 Some prevalence data collected were also country specific, whereas a substantial withincountry variability could be expected. In large countries with high geographic variation, seroprevalence estimates may not be representative of the national level. In some countries, there was a lack of available up-to-date seroepidemiological studies. Relying on data from older studies may risk generating seroprevalence estimates that do not represent the current epidemiological state of a country.

| CONCLUSION
Knowledge of worldwide CMV distribution will be helpful in informing national health care models of burden of disease and help future vaccine development.

FUNDING
Authors declare no contributions.

CONFLICT OF INTEREST
The authors have no competing interest.