Epidemiology of Cytomegalovirus
Meropi Nassikas, George Tsaples
Faculty of Tec gy, Policy and Delft University of Technology
Jaffalaan 5, 2628 BX, Delft, The Netherlands
m.nassikas @student.tudelft.nl
Paper for the 31st International Conference of the System Dynamics Society
Cambridge, MA — July 21-25, 2013
ABSTRACT
Cytomegalovirus (aka CMV) is the leading cause of neurological disabilities among children causing
among others mental retardation and hearing losses. This article approaches the uncertainty of CMV
with the use of system dynamics to test the effect of existing and hypothesized policies in order to
reduce the number of children affected by logical disabilities because of I infection.
A baseline system dynamics model built on deep uncertainty reveals that in the case of the USA,
the spread of the epidemic can be ined by two ping a vaccine currently under
trials and combining existing policies.
Key words: cytomegalovirus, epidemiology, USA, public health, policy design.
1. Introduction
The preval tr modes and sympt are p in section 1, then system boundaries and modeling choices
are developed in section 2. The third section analyzes the goodness of fit of the model while public health policies are tested
under laboratory conditions in section 4.
1.1 Nature and prevalence of the virus
Congenital cytomegalovirus (CMV) is the leading cause of neurological disabilities in children (Ludwig and Hengel, 2009);
therefore it represents a major public health concern. Human CMV is a large DNA virus belonging to the family Herpes
viridae (cf Figure 1). Like all herpes viruses, CMV establishes a lifelong latency in the host, with periodic re-activations
(Cheeran et al., 2009).
CMV infection is ubiqui in the human population and most i
in the the USA adjusted for ages is for example 60% (Cheeran et al., 2009). Seroprevalence increases among individuals with
are eventually infected. The overall seroprevalence
proximity to infected children or working in childcare facilities, but the virus remains invisible and undetected.
Figure 1: Schematic representation of the CMV virus (Crough and Khanna, 2009)
1.2 Transmission modes
The different modes of transmission are: close interpersonal contact ( body fluids like urine and saliva), sexual activity,
breastfeeding, blood transfusions and organ transplantation (Colugnati et al., 2007). So CMV is not transmitted via the
respiratory system. After the initial infection, the virus stays in saliva, tears, semen, urine, cervical secretions, and blood for
months to years. Seroconversion occurs at mucosal surfaces via infected urine, saliva or other bodily fluids, making children
an excellent host for the virus (Cheeran et al., 2009), especially in day care settings. However all transmission modes are
not equivalent in terms of risk. The most d ission mode is the ital one, in which a pregnant mother
transmits the virus via the placenta to the fetus.
1.3 Congenital infection and timing
It is well documented that the risk of congenital CMV is greatest from a primary infection of the mother during pregnancy
(Cheeran et al., 2009). Seronegative women of child-bearing age (15-44 years) undergoing primary infection have the highest
tisk of trans-placental transmission of CMV to the fetus. The transmission rate to the fetus is 32% in the case of primary
infection, contrary to 1,4% in the case of prior immunity of the mother (Kenneson and Cannon, 2007). In non-primary
infection the fetus is thought to be partially protected by maternal immunity (Ludwig and Hengel, 2009). Furthermore,
the timing of primary infection relative to the pregnancy is a crucial factor in establishing the risk to the fetus for in utero
transmission (Cheeran et al., 2009). Congenital CMV infection during the first trimester is more likely to cause CMV disease,
since organogenesis takes place in this period (Ludwig and Hengel, 2009).
90
80
83
probability (%6)
|
a
3 4 5 6 7
Primary maternal infection time
°
~¢
¢
Figure 2: Chance to develop CNS sequelae as a function of primary infection time (Kenneson and Cannon, 2007)
Figure 2 illustrates that the probability to develop Central Nervous System (CNS) sequelae for the child born after a primary
infection drops significantly from 80 to 8% between the first and the second p 'y trimester. The ivation of aCMV
i I infection; however, the risk is lower, as preexisting maternal
human CMV antibodies have a protective role against intrauterine transmission (Crough and Khanna, 2009). The two main
infection p. are in Figure 3 . The y-axis indicates the percent chance to move from one stage to
the other.
infection during preg: can still cause s
1.Primary infection 2.Reactivation
seronegative pregnant women seropositive pregnant women
primary infection
ae reactivation ot
sppetingection rate
er eae [reactivation or superinfection during preenancy |
Eee (ee
= et wets transplacental tranplacental ian —_
(Kennesson 2007) 32%HH) o itsion rate 1 transmission rate 2 1,4% (Kennesson 2007)
Infected Infants 2
85% (Crough, 2009)
asymptomatic
fants 2
symptomatic
symptomatic
infants 1 2
fants 2
5% Pass 2006)
Mentally retarded
i N
im SNHL children
Figure 3: The congenital infection chain with 2 pathways
1.4 Symptoms
For all transmission modes, except congenital infection, CMV is a rather benign disease, whose worse symptoms are similar
to those of a mild flu. While acquired infection is harmless for children, CMV can have serious consequences for immune
system deficient patients, such as a HIV seropositive patient, as well as for congenitally infected infants. The latter develop
serious permanent impairments which mostly affect the central nervous system and include progressive hearing loss, spastic
tetraplegia , mental retardation and visual impairments (Ludwig and Hengel, 2009). The etymology of cytomegalovirus stems
from “cyto” which stands for cell and “megalo” for surdimensionate, indicating that CMV infection has been associated with
pro inflammatory cytokine increases (Crumpacker, 2010). Cytokines are cell-signaling protein molecules involved in inter
cellular communication.
Figure 4: Central Nervous System outcomes of congenital CMV infection (Cheeran et al., 2009)
The computed tomography (A) and magnetic resonance imaging (B and C) shown in Figure 4 illustrate outcomes of congenital
CMV infection in infants. These examples of CNS sequelae, ranked in severity order are periventricular calcification (A),
polymicrogyria (B), and porencephalic cysts (C), which are a profound structural abnormality (Cheeran et al., 2009).CMV-
damage during the fetal period may cause spontaneous abortion or premature birth, but these cases are rare. While the majority
of congenitally infected children appear asymptomatic at birth, neurological sequelae may develop after months or even years
(Ludwig and Hengel, 2009).
2 Model description
“There is an urgent need for interventions that can reduce the substantial burden of this often overlooked disease.” (Kenneson
and Cannon, 2007)
2.1 Modeling question
Based on previous research done by Prevots et al. (1998)on the epidemic outbreak of polio in Albania in 1996, the hypotheses
is that CMV follows an oscillatory pattern over time, with an inter-peak distance in the order of magnitude of 10 years (cf
Figure 5).
80 | 100
g ze
8 so =
3%]
g we
g | g
8 40 8
% z
3 fo)
<4
8
“tile Jin | F
(6061 62 63 64 6s 60 87 €8 89 7071 72 737475 76 7 7B 70.0 81 #280 84 BS 86 87 88 20 00 91 9203.04 05,
Year
Figure 5: Reported cases of poliomyelitis in Albania (Prevots et al., 1998)
The purpose of this model is thus to first model the epidemiological behavior of the CMV virus, then test the effect of 3
types of policies: prevention, screening and therapy. Since some plausible policies are not yet implemented, the model also
provides laboratory test conditi So the modeling question is the following: what are the effects of each health policy on the
population of CMV-induced disabled children as well as on the infected fraction of the total American population? Different
modeling techniques exist, but system dynamics is an appropriate method to study the epidemiology of CMV because the
system feeds back on itself, it has a dynamic character from which oscillations are expected, and the population approach
is aggregated. This model could thus be used by public policy makers such as health ministries or by pharmaceutical firms
during the trial phase of vaccines and therapeutic drugs. Since the epidemiology is not country specific, the parameters of the
model can be changed to suit another geographic area.
2.2 System boundaries
The geographic boundary of this model is the USA, with its population of more than 314 million people in 2012. The time
horizon is set at 100 years, in order to take into account the delays iated with p ics. Also, disabilities due
to other congenital infections and genetic anomalies are ignored. For cimplification reasons, the evolution of the American
population is only dictated by natural growth, i.e. immigration flows are ignored (cf Figure 6). Two other simplification
assumptions are made: the seroprevalence rate does not vary among ethnic origins (Colugnati et al., 2007), and the virus does
not reactivate when in contact with new strains.
Secilburden |
Figure 6: Conceptual model
2.3 Linking ageing and infection chains
The systems dynamic model created is built around 3 processes: population ageing, post natal infection via contact rates and
infectivity, and the prenatal ital i i The latter i ions are ivided into 2 p: : seropositive i
(i.e. pregnant women whose seroconversion occurred before the pregnancy) and primary infections (where seroconversion
occurred during pregnancy). Only seroconversions during the first two trimesters (T1 and T2) are considered as dangerous to
the fetus (cf Figure 2).
Infectivity
Contact rate
Thfectious but
Senso Neurinal healthy
Hearing Loss (SNEL)
Figure 7: Simplified stock flow structure
2.3.1 The population ageing chain
Seronegative children can either grow up and become seronegative adults, or become infected at day care centers and at
school. Seropositive children then grow up and become seropositive adults (cf Figure8 ). As a result, two death pathways
exist: seropositive and seronegative deaths. A feedback loop connects the adults to the birth sub-model.
2.3.2 The post natal infection chain
Infections after birth are based on a Susceptible Infected Recovered model with deaths (Sterman, 2000), but there is no recov-
ered population in the case of CMV (cf Figure 8) .The key variable in modeling the infections due to close interpersonal contact
is the risk of acquiring infection in adults: risk of acquiring infection in adults = infectivity + contact rate * infected fraction.
Non infctious
seropositive children
Non infectious seropositiv:
adults
growth with cay seropositive adult
death
[phildren becontin:
uniniectious
Inictious children
becdfbin:
tion of tha ———"Pbaingsttions
infection
li expectancy
Infectious adults
-
fjakections in
children
Non infected children
Figure 8: Population ageing chain and post natal infection chain
2.3.3 The prenatal congenital infection chain
Instead of a single “birth” inflow the entire process of congenital infection is modeled. Pregnant women are first divided
into two categories according to their serology: seronegative and seropositive pregnant women. Because of the different
transmission and gravity of sequelae p ilities, two ital infection p: can be seen in green and brown on
Figure 9.
The chain of the seronegative pregnant women is modeled as an infection chain, with the pregnant women further divided into
groups according to their pregnancy trimester. The reason behind this decision is that seronegative women who get infected in
the first trimester present higher chances of transmission to the fetus and subsequent development of CNS sequelae (cf Figure
2). The seropositive pregnant women are modeled as a simple stock because the possibility of a virus reactivation and of a
transmission to the fetus is relatively low (1.4%, cf Figure 3). Besides, reactivation with new virus strains mechanisms and
the on the fetus ciated with their timing are currently still unknown. Also, the probability that an infected
fetus in the reactivation chain develops sequelae is lower than in the case of primary infection (15 compared to 82%).
After birth, infants are not modeled as stocks in the model, because it is possible to calculate directly at the end of pregnancy,
with a second order delay, the distribution of deliveries into 4 children stocks. CMV infected infants can mature and become
part of the following categories according to the severity of the outcomes to:
¢ Children with CMV-induced mental retardation
¢ Children with CMV-induced SNHL sequelae.
, a
=
seropositive frepositive motheqeropositive
pregnancies deliveries
Seronezative roa]
egnancies T]| on infected 2nd T non infected T3 eznancies T3] seronegative
deliveries
primary
fections T2
Figure 9: The congenital infection sub-model
e Infectious healthy children.They do not show any symptoms but are carriers of CMV.
e Non infected children. They are seronegative.
Tf a seronegative mother acquires a primary infection and the virus is transmitted to the fetus, then the fetus can be born either
with symptoms or not. If it is born with symptoms then it has 80% chance to develop a CNS (central nervous system sequelae)
and depending on the damage the virus caused, the child can be either mentally retarded or deaf (Senso Neurinal Hearing
Loss, aka SNHL). Therefore the equation, for the inflow “MR after primary infection in T1” (annual rate of children who
develop mental retardation after a primary infection in the first trimester) is a first order delay with initial value and delay time
of 2,5 years:
MR after primary infection in TI=DELAY N (primary infections T1*transmission rate after primary infec-
tion*(percentage of symptomatic infant*(1-fatality ratio of symptomatic infant)*percentage of symptomatic after
TI that shows CNS signs+(1-percentage of symptomatic infant)*percentage of asymptomatic after T1 that shows
signs of CNS)*percentage of CNS that leads to serious MR, 2.5,
primary infections T1*transmission rate after primary infection*(percentage of primary symptomatic infant*(1-
fatality ratio of symptomatic infant )*percentage of symptomatic after T1 that shows CNS signs+ (1-percentage
of primary symptomatic infant)*percentage of asymptomatic after T1 that shows signs of CNS )*percentage of
CNS that leads to serious MR, 1)
duration of infancy
MR after primary infection in T1 Children with
re x ge CMV-induced Mental | severely disabled
i death
‘MER after primary infection in T2 picteacisnon
‘MR after seropositive transmission
Figure 10: Stock flow diagram for children affected by Serious Mental Retardation
A second order delay is used because the levels of “fetus” and “infant” are omitted, and the probabilities that an infant is born
with symptoms, the chances that it will develop CNS sequelae and also the probability that the CNS sequela lead to mental
retardation or SNHL are known. The same reasoning lies behind the other inflow equations in the sub-model for the 3 other
children’s stocks.
The stocks of SNHL sequelae children, infectious healthy children and non infected children are emptied into the general
population. Given the low severity of the SNHL handicap it is assumed that children who suffer from SNHL sequelae can
also be incorporated in the general population. Because of their severe mental impairments, it is assumed in the model that
children suffering from serious mental retardation do not integrate into the general population chain.
3 Behavior of the Model
The simulation of the model during 100 years is used first to verify the dynamic hypotheses and then to differentiate between
two states: transient and permanent. The model was run on Vensim ( integration method: RK 4 Auto, time step:0.0078125,
units for time: years).
3.1 Similarity with polio: an oscillating pattern
In section 2.1 a comparison between CMV and polio was assumed. The oscillation behavior visible on Figure] 1 confirms the
expectations that CMV would follow the same pattern as polio. CMV’s inter peak duration is however longer than for polio:
whereas the latter is 10 years (cf Figure 5) , CMV infection peaks approximately every 70 years.
Figure 11: Infectious fraction of the population in the baseline model
3.2 Behavioral validation
The behavior of the model is coherent with expected results concerning the evolution of the American population and the
number of disabled children due to CMV. First, the total American population is slowly increasing, which is justified by the
fact that the birth rate (0,0135 birth/person/year) is slightly superior to the death rate ( 0,0133 death/person/year, corresponding
to a life expectancy of 75 years). Since migration patterns are not taken into account in the model, the behavior of the ageing
chain agrees with the hypotheses .
Secondly, the evolution in time of disabled children due to CMV is coherent with the structure of the model. Figure 12
displays the evolution of the number of CMV-induced mental retardation and senso-neurinal hearing losses (SNHL). The
comparison of the evolutions of the infectious fraction ( cf Figure 11) and the CMV-induced disabilitites among children leads
to the following observation. The peak of CMV-induced disabilities is delayed of approximately 2 years after the peak of the
infection in the total population., which corresponds to the structure of the model.
3.3 Sensitivity analysis on key parameters
To gain a better understanding of the roles and relative importance of several key factors, sensitivity tests were performed.
Since a person infected with CMV person does not exhibit severe symptoms but mild and flu-like symptoms, the dynamics
of the virus is not easily accessible. This assumption is supported by the fact that the seroprevalence of the virus in the
Tine (Yeas)
Figure 12: CMY-induced disabilities among children at the age of 2, baseline case
general population reaches the high level of 80% by the age of 60(Cheeran et al., 2009) and the social awareness still remains
at a minimum level. Since the spread of the virus in the model depends on the parameters “infectivity”, “contact rate” and
“duration of the infection”, the series of sensitivity tests addresses the uncertain values of those parameters. The baseline
values are the following:
infectivity = 0.04 person/contact (Colugnati et al., 2007)
inter adult contact rate = 27 contact /year/ person
durationof the in fection = 1,5 year (Cheeran et al., 2009)
The specifications of sensitivity analysis for the 3 parameters are : 1000 simulations, latin hypercube and a random uniform
distribution of the parameter. The ranges of the parameters are the following :
© infectivity : [0.01 -2]
© contact rate : [15 - 90] person/year
© duration of the infection [1 - 10] year.
As shown in Figures 15,16 and 17 presented in appendix, the infectious fraction of the population is, as expected, highly af-
fected by the 3 parameters. For the 3 sensitive parameter’s analysis, the infectious fraction of the total population demonstrates
a high peak in the early years and a value above the baseline scenario in the later years. When the infectivity rate of the virus
varies in the range i above, the i i fraction of the reaches 50% instead of 1,1% in the baseline
case. So the behavior of the CMV can in the worst case scenario follow the pattern of an acute disease. Consequently the
ty, the
parameter infectivity seems to be a leverage for public health policy makers since the spread of the disease is highly sensitive
number of children born with a disability due to a congenital CMV infection increases. To conclude the tests on sens:
to its variation.
4 Policy Testing
Existing and future measures to fight againt CMV are retrieved from litterature. Each policy tested is presented individually,
and then a final comparison of all proposed policies is performed. The specifications of each policies are presented in Appendix
in table 3.
4.1 Overview of prevention measures
The nature of the policies tested is inspired by the existing prevention techniques currently implemented. The overview of the
range of tools available to public health decision makers reveals that there are three types of prevention measures:
1. PRIMARY PREVENTION has the aim to avoid an infection. It consists in a set of prenatal hygiene measures and change
of behavior. Pregnant women, especially high risk seropositive women working in close contact with young children,
are encouraged to limit close interpersonal contacts during the first two trimesters of their pregnancy.
N
SECONDARY PREVENTION has the aim to identify infected patients early, be it pregnant women or infants. Four main
measures can be taken :
21. ing for maternal antibodies in the mother. Especially the avidity index of the protein Immunoglobulin G
can be used to detect seroconversion among pregnant women(Kenneson and Cannon, 2007). Those maternal antibodies
screenings must be done during the first trimester of pregnancy.
2.2. Prenatal ultrasound for the presence of fetal abnormalities caused by CMV .
2.3. Amniocentesis to detect viral genome in the amniotic fluid. Viral culture or Polymerase Chain Reaction (PCR) of
the amniotic fluid can predict fetal transmission and symptomatic infection (Naessens et al., 2005). The sensitivity of
PCR used to detect viral DNA is very good if amniotic fluid is collected at least six weeks after seroconversion and
around the 22nd week of pregnancy (ECCI, 2012).
2.4 Screening of all newborns for CMV shedding in the urine and monitoring of all congenitally CMV infected newborns
in long-term audiologic follow-ups up to 5 years (Ludwig and Hengel, 2009).
By combining serologic screening of IgG and IgM antibodies in the beginning of pregnancy and on cord blood with a
culture of urine of the identified high risk infants, Naessens et al. reached a sensitivity of detection of congenital CMV
of at least 82% (Naessens et al., 2005).
w
. TERTIARY PREVENTION in the case of symptomatic disease. Tertiary prevention involves post natal treatments, which
are antiviral drugs given to infected infants. The most famous drug is called Ganciclovir, and there is evidence that
therapy ameliorates the severity of one of the CNS complications : SNHL(Cheeran et al., 2009). Ganciclovir therapy
begun in the neonatal period in infants showing CNS symptoms prevents hearing deterioration at 6 months and may
prevent hearing deterioration at 1 year (Kimberlin et al., 2003). Another promising therapy is the injection of CMV
immune globulin in utero for infected fetuses. Uncontrolled studies of this method have suggested an impact on neu-
ropathogenesis, and a decrease transmission to the fetus. Controlled trials should be conducted with pregnant women
(Cheeran et al., 2009).
4.2 Testing of available policies
4.2.1 Policy 1: Prevention for pregnant women
In the first policy, which belongs to the category primary prevention, the propagation of the virus is slowed among pregnant
women. If for example, a pregnant woman limits her contacts with people from the general population she can reduce the risk
of infection. The “pregnant women contact rate” is reduced by 50% (from 27 to 13.5 contact/person*year ). The number of
disabled children alters only in a numerical and not in a behavioral way. A sensitivity analysis on the policy with values of
the new contact rate ranging from 20%-80% of the initial contact rate value (ie 5.4 to 21.6 contact/person*year) results in a
disappointing curb of the outbreak peak, which at year 60 is only numerically dampened.
4.2.2 Policy 2: Decrease the transmission rate to the fetus
The second policy refers to the tertiary prevention method of immunoglobulin injection in utero. The transmission rates of
the virus from mother to fetus are decreased (rates for both primary infection and seropositive reactivation transmissions).
Policy 2 can however only be implemented provided maternal antibodies have been analyzed beforehand. By combining
maternal antibody screening and immunoglobin, Policy 2 protects the fetuses by providing them antibodies. In the model,
both ission rates from seropositive and seronegative pregnant women to fetus are reduced by 20%. The implementation
of policy 2 entails a decrease of the number the disabled children changes in a numerical way. A sensitivity analysis on Policy
2 with new values for the transmission rate ranging from 30 to 95% of initial values shows that Policy 2 only has a weak
numerical reducing impact.
policies (2 to 4)
Figure 13: Comparison of existing policies on CMV-induced Mental Retardation at the age of 2
4.2.3 Policy 3: Improving ing p
The third policy refers to the secondary prevention methods of detection. Policy 3 is divided into Policy 3.1 and Policy 3.2.
Policy 3.1 : The choice of abortion Policy 3.1 addresses the first 3 screening methods: maternal antibody check,
ultrasound and amniocentesis. Pregnant women who are confronted with a seroconversion during pregnancy are then faced
with the choice of abortion. An increase in the fatality ratio of the infected fetus is then implemented. Policy3.1 brings a 50%
increase of the fatality ratio. Sensitivity analysis on Policy 3 with values ranging from a 50% to 100% increase in fatality ratio
reveals that it has no major influence on the number of disabled children.
Policy 3.2 : Monitoring and follow-ups of infected infants Policy 3.2 addresses particularly the last screening
method: screening of all newborns for CMV shedding in the urine and monitoring of all congenitally CMV infected new-
borns in long-term audiologic follow-ups. It is modeled by a drop of 40% in the percentage of children (symptomatic or
asymptomatic at birth) that show CNS signs. Supposing that CMV was detected at birth, those children receive extra-attention
and care, thus decreasing their sequelae. The sensitivity analysis attributes values for the chances of sequelae development
between 20% and 90% of the initial values. The result is similar to policies 1, 2 and 3.1 : there is a numerical impact, but not
behavioral.
4.2.4 Policy 4: Prevention in the general population
The spread of the virus can also be contained in the total population. If less people get infected, then the danger for seropositive
and seronegative pregnant women will be reduced. Therefore, Policy 4 reduces the contact rate among adults and also among
children by 20% (from 27 to 13.5 contact/person/year). A significant numerical and behavioral difference with the baseline
scenario is observed, both in the number of disabled children as well as the infectious fraction of the population. This confirms
the original assumption for this policy since less contact implies that the virus does not spread easily in the population. The
infectious fraction of the population reaches its lowest level at 2,5*10~> (instead of 11*10~%) after 40 years. The periods
during which the CMV virus is almost inactive is the longest of all policies tested so far.
However, an unexpected behavior is observed in the long-run, that is to say after 40 years. The increase of disabled children
due to CMV after 40 years (cf the peak of CMV-induced Mental Retardation cases on the red curve of Figure 14) is due to a
spike in primary congenital infections. The peak even exceeds the peak of the baseline scenario. Since contact rates between
adults are reduced, the i ion of adults in rep ing age increases. Over the years, a pool of seronegative
adults is created Consequently, more seronegative pregnant women undergo the chance of a seropositive conversion during
their pregnancy. Policy 4 is therefore efficient in reducing the number of congenitally induced CMV diseases in the short term,
but has adverse long term impacts.
Sensitivity analysis on Policy 4 with values ranging from 20%-95% down from the initial adult and children contact rates
has only a numerical effect, thus confirming that Policy 4 is highly effective in the short run, but has dangerous long term
consequences.
4.2.5 Combined existing policies
Tf all the policies that were presented are combined, the number of CMV-induced Mental Retardation reaches the lowest point
(approximately 700 children) of all tested policies at at year 60. Even if the combination of all existing policies is not feasible
in reality, the drastic positive effect on the number of disabled children at the age of 2 that room for
exists in the fight against CMV.
4.3 Policy 5: the future promising vaccine
A safe and effective CMV vaccine for seronegative women is not available so far (Ludwig and Hengel, 2009). If that vaccine,
that is currently under animal-trial (Cheeran et al., 2009), were to be approved for humans and massively administered, the
infectivity of the virus would be reduced by 50%. Policy 5 thus introduces a new infectivity rate of 0,02 person/contact. The
effect on the number of disabled children is significant in the long-term : for example the number of Mental Retardation cases
induced by CMV falls under 500 cases in year 100 (cf Figure 14). Such a drop in infectivity has a significant impact on the
number of disabled children: after 100 years, the cumulative number of disabled children drops by 84 % compared to the
baseline scenario.(correspoonding to 331 706 disabilities avoided, cf Table1).
The spread of the virus in the population drops exponentially and after 15 years the virus seems to be eliminated from
the American population. A sensitivity analysis of policy 5 with values ranging from 20 to 80% of the initial infectivity
demonstrates that even a drop of 20% of the infectivity (ie to 0.032 people/contact) after the introduction of a vaccine would
yield a decrease of 46% of cumulative disabled children compared to the baseline scenario.
4.4 Comparison of policies
In order to evaluate the magnitude of each policy’s impact, the chosen criterion is the number of disabled children due to CMV
congenital infection ( which includes both SNHL and Mental Retardation sequelae). The evolution of the number of Mental
Retardations at the age of 2 due to CMV presented in Figure 14 emphasizes that in the long run policy 5 (vaccination) performs
better than the combination of existing policies and brings the number of disabled children under 500. The difference between
the base scenario and the different policy scenarios is presented in the last column of Table | and confirms the previous
observation. The cumulative number of disabled children due to CMV is indeed dramatically reduced by vaccination: if the
vaccine were to be clinically approved and distributed on a large scale, more than 330 000 child disabilities would be avoided
in the time span of 100 years.
Figure 14: Comparison of policy effects on the number of CMV-induced Mental Retardation at the age of 2
Cumulative Difference with
Scenario number of disabled % decrease
Fi base model
children
Base Model 344 992
381 383 457 11583 3%
2 314 307 80 685 20%
4 301 543 93 449 24%
32 269 314 125 678 32%
1 234 256 160 736 40%
Combined existing
policies (1;2;3.1;3.2;4) Br181 313.811 80%
5 63 286 331 706 84%
Table 1: Comparison of the policies tested
5 Conclusions
The dangerous character of the cytomegalovirus stems from its “invisibility” in the general population. Since an infected
individual does not manifest severe symptoms, Cytomegalo virus infections are misunderstood for a common cold. Therefore
awareness about the disease is low. Even if the system dynamics model presented in this report indicates that combining
existing policies offers great potential to limit the spread of the virus, this policy can only be implemented if awareness of
CMV and its consequences increases.
Uncertainty remains concerning the mechanisms of trans: ion to the fetus, particularly in the case of reactivation with
new strains of CMV. Besides, even in cases of primary infections, medical practitionners cannot predict with accuracy the
development of symptoms for the infants, and the predictions are even more uncertain for the children’s sequelae. Despite the
ambiguity of di the epidemiol CMV model here shows that there is room for maneuver to reduce the
negative effects of CMV by associating different public health policies:
1), efficient screening methods in utero (Policy 3), better hygiene measures for everyone (Policy 4) and in utero antiviral
increasing awareness among pregnant woman (policy
treatment (Policy 2). Nevertheless the model may have oversimplified the late development of sequelae, which can appear
after 2 years.
Since CMV infection is the leading infection cause of mental retardation and hearing loss among congenitally infected chil-
dren (Colugnati et al., 2007), the absence of a vaccine that would prevent infection from the virus is the biggest challenge
so far. The model’s behavior supports the “sense of urgency in clinical vaccine trials” mentioned by Cheeran (2009). The
vaccine presents indeed the highest potential to reduce the number of disabled children to the extent that it would bring a 84%
decrease of the cumulative number of disabled children due to CMV.
Further research efforts could be directed at first evaluating the cost of the policies presented in this report on a societal level,
and then finding the optimal policy mix. The model can also be refined by distinguishing different classes of pregnant women
according to their seroprevalence, which is correlated to ethnic origin (Colugnati et al., 2007). Finally, Exploratory Modeling
and Analysis (EMA) developed by E. Pruyt and J.H. Kwakkel at TU Delft (Pruyt and J.H., 2013) could help to explore and
analyze the effectiveness and robustness of adaptative policies over time under deep uncertainty.
References
Cheeran, M. C.-J., Lokensgard, J. R., and Schleiss, M. R. (2009). is of ital irus infecti
Disease mechanisms and prospects for intervention. Clinical Microbiology Reviews, 22(1):99-126.
Colugnati, F., Staras, S., Dollard, S., and Cannon, M. (2007). Incidence of cytomegalovirus infection among the general
population and pregnant women in the united states. BMC Infectious Diseases, 7:71.
Crough, T. and Khanna, R. (2009). I bi of human lovirus: from bench to bedside. Clinical Microbiology
Reviews, 22(1):76-98.
Crumpacker, C. S. (2010). Invited ‘y: Human i irus, i ion, cardiovascular disease, and mortality.
American Journal of Epidemiology, 172(4):372-374.
ECCI (2012). European congenital cytomegalovirus initiative.
Kenneson, A. and Cannon, M. J. (2007). Review and meta-analysis of the epidemiol of fe; 8 irus (cmv)
infection. Reviews in Medical Virology, 17(4):253-276.
Kimberlin, D. W., Lin, C.-Y., SAjnchez, P. J., Demmler, G. J., Dankner, W., Shelton, M., Jacobs, R. F., Vaudry, W., Pass,
R. F, Kiell, J. M., Soong, S.-j., Whitley, R. J., for the National Institute of, A., and Infectious Diseases Collaborative
Antiviral Study, G. (2003). Effect of ganciclovir therapy on hearing in ip galovirus disease
involving the central nervous system: a randomized, controlled trial. The Journal of Pediatrics, aes 16-25.
Ludwig, A. and Hengel, H. (2009). Epidemiological impact and disease burden of congenital cytomegalovirus infection in
europe. Eurosurveillance, 14(9).
Naessens, A., Casteels, A., Decatte, L., and Foulon, W. (2005). A serologic strategy for detecting neonates at risk for
congenital cytomegalovirus infection. The Journal of Pediatrics, 146(2):194—197.
Prevots, D. R., Atti, M. L. C. d., Sallabanda, A., Diamante, E., Aylward, R. B., Kakariqqi, E., Fiore, L., Ylli, A., Avoort, H.
v. d., Sutter, R. W., Tozzi, A. E., Panei, P., Schinaia, N., Genovese, D., Oblapenko, G., Greco, D., and Wassilak, S. G. F.
C288): Outbreak of paralytic poliomyelitis in albania, 1996: High attack rate among adults and apparent interruption of
ide mass ination. Clinical Diseases, 26(2):419-425.
Pruyt, E. and J.H., K. (2013). Exploratory modeling and analysis, an approach for model-based foresight under deep uncer-
tainty. Technological Forecasting and Social Change, 80(3)(431.doi:10.1016/j.techfore.2012.10.005):419.
Sterman, J. D. (2000). Business Dynamics, Systems thinking and modeling for a complex world. Mc Graw Hill.
Appendix
Stock Initial value (person)
Pregnancies of seropositive mothers 1.2*10°
Pregancies of seronegative mothers in trimerster 1 4*10°
Children at the age of 2 with serious Mental Retardation due to CMV | 3 500
Children at the age of 2 with SNHL sequelae 10 500
Infectious healthy children at the age of 2 4 800
Non infected children at the age of 2 3*10°
Non infectious seropositive children at the age of 2 12*10°
Infectious children at the age of 3.5 400 000
Non infected children at the age of 3.5 15*10°
Non infectious seropositive adults 115*106
Infectious adults 2*10°
Seronegative adults 156* 10°
Table 2: Stock variables and initial values
Tene (Yeu)
Figure 15: Sensitivity analysis of the parameter “infectivity” on the stock variable “infectious fraction of the population”
So oom
lndecious part of the poputation
O23
os
Figure 16: Sensitivity analysis of the parameter “contact rate” on the stock variable “infectious fraction of the population”
es ee]
Fracnce of the mfcctean part of Be populate
Figure 17: Sensitivity analysis of the parameter “duration of infection” on the stock variable “infectious fraction of the population”
Value after
policy imple-
Policy Parameter Base value [unit] it Sensitivity range*
1 Contact rate among pregnant | 27 [contact/person*year] | 13.5 5.4-21.6
women
2 Transmission rate to foetus 0.32 [Dmnl] 0.256 0.096-0.304
Mtenqarishanysmfesitiono 0.014 [Dmnl] 0.0112 0.0042-0.0133
foetus for seropositive
pregnancies
3:1 Fatality ratio of symptomatic | 0.07 [Dmnl] 0.14 0.105-0.14
infant
Percentage of asymptomatic | 0.5 [Dmnl] 0.3 0.1-0.45
32. after T1 that shows signs of
CNS
Percentage of symptomatic 0.8 [Dmnl] 0.48 0.16-0.72
after T1 that shows CNS
signs
Percentage of infant T2 that 0.08 [Dmnl] 0.048 0.016-0.072
shows CNS signs
4 Inter-adult contact rate 27 [contact/person*year] | 13.5 5.4-21.6
Children contact rate 27 [contact/person*year] | 13.5 5.4-21.6
all of the
Combined all of the above all of the above above all of the above
existing
policies
5 Infectivity 0.04 [person/contact] 0.002 0.008-0.032
* All sensitivity tests were performed with latin hypercube, 200 simulations and a random uniform parameter distribution
Table 3: Summary of policy tests,
