Although the outcomes of toxoplasmosis have been well documented, an integrated estimate of the impact of this infection on the health status of the population is not available. “Disability-adjusted life years” are the sum of years of life lost and years lived with disability, weighted for the severity of the illness. The estimated disease burden of congenital toxoplasmosis in The Netherlands is 620 (range, 220–1900) disability-adjusted life years per year, which is similar to that for salmonellosis and is mainly caused by fetal loss and chorioretinitis. However, there is considerable uncertainty in this estimate. Scenario analysis indicates that the true burden may be underestimated. In other countries, the disease burden is expected to vary with the incidence of congenital infection, but it may also depend on the health care system. In countries that actively screen for toxoplasmosis, such as France, there may be a lower burden of morbidity but a higher burden of mortality.
Although Toxoplasma gondii infects a large proportion of the world's human population, it is an uncommon cause of manifest disease. Infection may be congenitally or postnatally acquired. Congenital infection occurs when a woman becomes infected during pregnancy. The parasite enters the fetal circulation through infection of the placenta . Infants who are born to mothers who acquire an infection in the first and second trimester of pregnancy more frequently show severe symptoms of congenital toxoplasmosis. Alternately, the infection may cause miscarriage or still birth. In contrast, the majority of children who are born to women who acquire an infection during the third trimester of a pregnancy are born with the subclinical form of the infection [1–5].
Although the clinical and subclinical signs of toxoplasmosis have been well documented, an integrated estimate of its disease burden is not available. For the development of evidence-based public health policy, such information is essential. Therefore, in the present study, the disease burden for congenital toxoplasmosis in The Netherlands was estimated. Because not all information needed for the calculations was available, this study was also used to identify the most important data gaps. An international comparison was made, on the basis of data from the literature, that specifically estimates the incidence and outcomes of toxoplasmosis in France.
The different outcomes of infectious disease can be combined in 1 single measure, the disability-adjusted life year (DALY), according to the method proposed by Murray et al. [6, 7]: DALY = number of years of life lost (YLL) due to mortality + number of years lived with a disability (YLD), weighted with a factor between 0 and 1 for the severity of the disability. YLL due to a specific disease in a specified population is calculated by the summation of all fatal cases (d) due to the health outcomes (l) of that specific disease, where each case is multiplied by the expected individual life span (e) at the age of death. Thus:
YLD is calculated by summation over all cases (n) and all health outcomes (l) of the product of the duration of the illness (t) and the severity weight (w) of a specific disease:
The burden of disease attributable to 1 agent is determined by accumulating the impact of all health outcomes associated with this agent.
The health outcomes following T. gondii infection were defined using an outcome tree. Each block represents a health outcome and transition probabilities between all blocks were established. The disease burden is defined as the present expected sum of current and future DALYs resulting from all incident cases of disease in a specific time period, taking into account lifetime probabilities of moving to each disease state [6, 8]. We extracted data for the outcome trees from the international literature.
We estimated the incidence rate from figures from 2 special studies performed in The Netherlands, the Toxoplasma Infection Prevention (TIP)  and Peiling Immunisatie Effect Nederland Ter Evaluatie van het Rijksvaccinatieprogramma (Surveillance Immunization Effect for Evaluation of the National Immunization Program; PIENTER)  studies, and applied these rates to data regarding the number of pregnancies and births that occurred in The Netherlands in the year 2004. To estimate the incidence of the different health outcomes due to congenital toxoplasmosis, we reviewed the relevant studies in the literature and pooled the results, weighting them for study size. We restricted our search to European and North American studies involving cases identified by prenatal or neonatal screening, because the incidence of health outcomes in referred cases is higher and is not representative for all children with congenital toxoplasmosis. The lowest and highest results from these studies were used to estimate the attendant uncertainty.
Estimates of the duration of the adverse health outcomes were preferably derived from Dutch studies and, when necessary, were supplemented with data from international studies. We derived the expected life span of fatal cases for 2004 from standard life tables published by Statistics Netherlands.
The DALY uses explicit preference weights for health status. As a first approximation, we applied severity weights for similar outcomes (table 1), but not for chorioretinitis, which was included in a special panel elicitation . Visual impairment occurs in only a fraction of children with chorioretinitis. In France, Kodjikian et al.  found that, among 130 children, 22% had unilateral visual impairment and only 1.5% had bilateral impairment. In the United States, Mets et al.  reported much higher values of 22% and 29%, respectively. The favorable outcomes in France may have been the result of a treatment effect, but the unfavorable data from the United States may include referral bias. For The Netherlands, where no screening is performed, we decided to consider these 2 studies as the low and high estimates, and we defined our most likely value as their midpoint.
To account for uncertainty, low values, most-likely values, and high values were used for parameters. These were either the fifth, 50th, or 95th percentile of a statistically uncertain parameter or an optimistic, most likely, or pessimistic parameter value for systematic uncertainty. For some specific parameters, however, the uncertainty was too large, and scenario analysis was used to analyze these uncertainties. Scenario analysis was also applied to explore the impact of some controversial choices, such as discounting years of future life or the valuation of the life of an unborn child.
Outcome tree. Clinical manifestations of congenital toxoplasmosis in the first year of life vary from signs of the classic triad of toxoplasmosis (chorioretinitis, intracranial calcifications, and hydrocephalus) to abnormalities of the CNS that lead to neurological deficiencies—that is, psychomotor or other neurological deficiencies, convulsions, and mental retardation —and neonatal death (figure 1). Other signs of congenital toxoplasmosis include a number of nonspecific illnesses, such as anemia, jaundice, rash, encephalitis, pneumonitis, and diarrhea . However, the incidence of these health outcomes is low and most of these disorders are self-limiting. Therefore, their influence on the disease burden is negligible, and we decided not to include these illnesses in the outcome tree. Infants who had experienced subclinical infection during their first year of life may develop clinical signs and deficiencies later in life (mostly affecting the eyes) [17–20].
Incidence. Baseline incidence was derived from the TIP study , a prospective cohort study that started in 1987 and involved 28,000 pregnant women in the southwestern region of The Netherlands. Of 15,170 seronegative women, 44 (0.29%) became infected with T. gondii during pregnancy (90% CI, 0.22%–0.36%), and 42 children could be observed to the age of 1 year. Twelve newborn children had a congenital T. gondii infection, representing a transmission rate of 28% (90% CI, 18%–40%). Thus, the estimated incidence of congenital toxoplasmosis is 8.1 (90% CI, 4.7–12) cases per 10,000 live births by women who were seronegative at the start of their pregnancy. In 2004, there were 194,000 live births . By combining the age-specific seroprevalence rates from the PIENTER study  with the age distribution of pregnant women, we estimated that, among these, a total of 129,000 children (67%; 90% CI, 61%–72%) were born to seronegative mothers. This implies that 8.1 × 129,000/10,000 = 105 (90% CI, 56–167) cases of congenital toxoplasmosis occur in The Netherlands annually (or 105 × 10,000/194,000 = 5.4 cases per 10,000 births).
The PIENTER project  was a cross-sectional seroprevalence study performed in 1996. In the age range of 15–50 years, the observed seroprevalence of toxoplasmosis increases linearly with age at a rate of 1.42% (90% CI, 1.21%–1.63%) per year. Sixty-seven percent of mothers are seronegative at the start of pregnancy. Assuming that the risk of infection for pregnant women is the same as for the general population, the risk of infection during pregnancy is (9 months of pregnancy/12 months per year) × 1.42% × 67% = 0.71%. The risk of transmission to the fetuses of women who are infected 4–6 weeks after conception is <1%, and it is only 6% at 13 weeks of gestation. Thereafter, the frequency of transmission increases sharply and, roughly, linearly, reaching 72% at 36 weeks of gestation. Overall, Dunn et al.  reported a transmission rate of 29% (90% CI, 26%–32%), leading to an estimated incidence rate of 0.71% × 29% = 0.21% or 21 (90% CI, 14–28) cases per 10,000 live births to seronegative mothers. This is ∼2.5-fold higher than the estimate that was based on data from the TIP study .
Incidence of health outcomes. We found no evidence that prenatal and postnatal treatment of the mother and/or child had a significant effect on health outcomes, and we used the pooled estimation as the most likely value (table 2). We found 4 studies in the literature that reported the frequency of fetal losses, in total involving 1863 pregnancies (19 in Gratzl et al. , 144 in Foulon et al. , 194 in Gras et al. , and 1506 in Binquet et al. ). These authors reported, in total, 75 fatal outcomes (4%; 90% CI, 3.3%– 4.8%; i.e., due to spontaneous abortions, induced terminations, intrauterine death, and stillbirths). These data may also include fetal losses due to other causes, and are considered to be the maximum estimate. Our most likely estimate is based on data collected by Agence Française de Securité Sanitaire des Aliments (the French Food Safety Authority; AFSSA) , who estimated that, in France, 47 fetal losses due to toxoplasmosis occur annually among 2676 seroconverting mothers (1.7%).
The minimum estimate was based on Pathologisch-Anatomisch Landelijk Geautomatiseerd Archief (PALGA), the nationwide network and registry of histological and cytopathological findings in The Netherlands where, on average, an annual 0.13 diagnoses of toxoplasmosis are recorded per 100 excerpts from intrauterine fetal deaths (A. Hofhuis, National Institute for Public Health and the Environment, Bilthoven, The Netherlands, and Mariël Casparie, Prismant, Utrecht, The Netherlands; personal communication). It is estimated that 1200 intrauterine fetal deaths occur annually (Dr. P. G. Nikkels, Pediatric Pathology, University Medical Centre, Utrecht, The Netherlands; personal communication). This would result in a minimum estimate of 1.5 fetal losses due to toxoplasmosis per year.
To calculate the number of fetal losses, we assessed the total number of pregnancies in The Netherlands in 2004: 194,007 live-born children  + 1013 stillbirths after 24 weeks of gestation  + 34,515 spontaneous abortions occurring up to 20 weeks of gestation  = 229,435 pregnancies). The number of women who had a primary T. gondii infection was 229,435 × 67% (the percentage of women who were seronegative at the start of their pregnancy, calculated on the basis of data from the PIENTER study ) × 0.29% (on the basis of data from the TIP study ) = 430 women (90% CI, 300–600 women). In the DALY calculation, we took only fetal losses from 24 weeks of gestation to the end of the pregnancy into account, because this is currently the age at which a prematurely born infant is considered to be viable. We assumed that the number of fetal losses over time of pregnancy is equally divided, resulting in an inclusion of 16 (i.e., 40–24 weeks)/40 of the cases. Therefore, the most likely number of fetal losses was estimated to be 430 × 1.7% × 16/40 = 3 (90% CI, 1–10) fetal losses per year.
The most likely estimate of developing chorioretinitis later in life was based on the relatively large studies performed by Binquet et al.  and Gras et al. . The proportion of subjects who were symptom-free at the age of 1 year who developed chorioretinitis at age i was calculated as 1 - e-λ(i-1), where λ is the hazard rate per year. Follow-up in these studies was performed in patients up to age 14 and 10 years, respectively, and determined the hazard rate of developing chorioretinitis later in life to be 2% per year. The reported rate in other studies varied between 1% and 3%, which we applied as low and high estimates, respectively, in our analysis. The studies of Binquet et al.  and Gras et al.  suggested that the incidence was constant throughout the years and did not level off at later ages. A study by Koppe et al.  indicated that new cases can arise in patients aged at least 20 years; however, that number of observations was relatively small. The duration of this disorder is the life expectancy of the general population minus the mean age of developing chorioretinitis (79 - 10 = 69 years; table 3). We used this value as our most-likely estimate as well as our low estimate. As a high estimate, we arbitrarily doubled the period at risk to 40 years.
Disease burden. Table 3 presents the burden of disease per health outcome for congenital toxoplasmosis. It shows a total disease burden of ∼620 DALYs per year. This number is mainly caused by fetal loss (39%) and chorioretinitis (29%) (figure 2A and figure 2B). Uncertainty ranges between 220 and 1900 DALYs per year (figure 2A).
Scenario analysis. To explore the impact of uncertain assumptions on the final result, 6 different scenario analyses were performed (figure 3). (1) When we used the estimated incidence from the PIENTER study , the disease burden increased to 1200 DALYs per year (range, 440–3400 DALYs per year). (2) Economists prefer to discount the value of future life years because an immediate profit is generally preferred over a profit at a later time. Taking a discount factor of 4% into account reduced the number of DALYs to 190 (range, 70–590) per year. (3) We excluded the impact of fetal loss from the calculation, because many women may soon become pregnant again ("replacement pregnancy"). In this scenario, the number of DALYs per year is reduced to 380 (range, 110–1200). (4) Using a severity weight based on a favorable impact on visual function  reduced the disease burden to 480 (range, 180–1500) DALYs per year. Using a severity weight based on unfavorable outcomes  resulted in a disease burden of 760 (range, 250–2000) DALYs per year. (5) The impact of intracranial calcifications has been assigned a very low weight (0.01). However, arbitrarily raising this value to 0.1 only increased the most-likely disease burden to 700 (range, 250–2000) DALYs per year.
International comparison. In most industrialized countries, incidence of congenital toxoplasmosis varies between 1 and 10 per 10,000 live births. The EUROTOXO study  summarizes 8 European studies. Of these, per 10,000 live births, 3 were considered to be representative and reported incidences of 0.73 cases (in Sweden from 1997 to 1998), 5 cases (in Switzerland from 1983 to 1989), and 7 cases (in Switzerland from 1986 to 1994). In the other studies, incidence per 10,000 live births ranged from a low value of 3 cases in Denmark (from 1992 to 1996) to a high value of 13 cases in Germany (from 1988 to 1996). More recently, Gilbert et al.  reported an incidence of 3.4 cases per 10,000 live births in the United Kingdom. In the United States, the incidence of congenital toxoplasmosis is estimated to be 400–4000 cases per year, or a rate of 1–10 cases per 10,000 live births . The disease burden may be expected to vary with the rate of incidence of congenital infection, but it may also depend on the health care system. In France, an active screening program has been established and medical care is offered to infected mothers and their children. The incidence of congenital toxoplasmosis in France is estimated to be 600 cases per year , an incidence of 8.2 cases per 10,000 live births, which is ∼50% higher than in The Netherlands. On the basis of French estimates of clinical outcomes, the disease burden (per 10,000 live births) is approximately twice the burden found in The Netherlands (figure 4). This is particularly the result of a higher estimate of YLL because of fetal losses. Paradoxically, this high burden is a result of the prenatal screening program in France, leading to voluntary or medically indicated terminations of pregnancy. The impact of these abortions depends on the value assigned to the death of an unborn child, as discussed before. The burden of morbidity in France is estimated to be lower than in The Netherlands, in particular because we assigned a low severity weight to chorioretinitis on the basis of favorable outcomes presented by Kodjikian et al. . It must be noted that the French investigators also expressed considerable uncertainty in their estimates, partly because of differences among individual studies and partly because of difficulties in interpreting the available data.
Although it was already known that infection with T. gondii may have serious consequences for individual patients, the disease burden on a population level has not been calculated before. The results of this study show a cumulative disease burden of congenital toxoplasmosis in the Netherlands of ∼620 DALYs per year. Despite the low incidence of toxoplasmosis, the disease burden is similar to that of the major foodborne pathogen, Salmonella species (670 DALYs per year) . Our result is a conservative estimate. Scenario analyses have indicated that the incidence of congenital toxoplasmosis and, thus, the true disease burden, might be considerably higher.
The incidence of congenital toxoplasmosis in The Netherlands is not well known. The incidence in The Netherlands as calculated in the TIP study  was in the same range as that reported for the surrounding European countries. The estimate based on seroprevalence data from the PIENTER study  (21 per 10,000 live births) was higher than in surrounding countries. We believe that the true incidence in The Netherlands is between the results of the TIP  and PIENTER  studies. The TIP study  may have underestimated the national incidence because of intensive counseling of participants. When analyzing data from the PIENTER study , we assumed that seroconversion in the general population was a proxy for pregnant women (i.e., that there was no risk-avoiding behavior). Also, our analysis did not account for time-specific changes in infection rates, which may incur an upwards bias in the estimate of the current rate . On the other hand, Van Druten et al.  have suggested that seropositivity to Toxoplasma species may not be lifelong, leading to higher estimates of the rate of infection.
We found it very difficult to estimate the incidence of clinical outcomes of congenital toxoplasmosis on the basis of findings in the current literature. Many studies refer to clinically referred cases, and may overestimate the burden in unselected populations. Furthermore, differences between countries in prenatal and neonatal screening and the related timing of treatment may result in different outcomes. Our outcome estimates were based on pooled estimates and disregarded the impact of screening and treatment of infected mothers. We did not find evidence that prenatal treatment has a clinically important effect on the risk of transmission, and there is a lack of evidence on the effect of postnatal treatment in children (table 2). This conclusion corresponds with a recent report from the Systematic Review on Congenital Toxoplasmosis (SYROCOT) study that it is “unclear whether prenatal treatment has any effect on transmission or clinical signs” [p. 27, 38]. The difference in burden between the 2 extreme severity weights for chorioretinitis (scenarios 4a and 4b, from 480 to 760 DALYS per year) can be interpreted as reflecting the impact of early treatment (based on neonatal screening) or late treatment (based on clinical detection of children with chorioretinitis).
Including the impact of fetal loss in the disease burden estimate was difficult, because the incidence was difficult to establish. Furthermore, there are no established methods to account for such cases. We chose to only take losses of a viable fetus into account (the age limit was recently lowered to 24 weeks of gestation in The Netherlands) and assumed that fetal loss is equally divided over the duration of pregnancy. This is probably an overestimate, because infections acquired in the first and second trimester of pregnancy are more severe than infections acquired in the third trimester. According to Freeman et al. , the rate of spontaneous fetal loss before 28 weeks of gestation in the European Multicentre Study on Congenital Toxoplasmosis (EMSCOT) cohort was not greater than that in an unselected population, suggesting that fetal loss should not be included in the calculations. On the other hand, analysis of the PALGA registry, even if incomplete, shows that confirmed cases of intrauterine fetal death due to T. gondii infection do occur, and that this serious outcome cannot be neglected. This is especially the case if pregnancies are terminated on the basis of the results of a screening program.
In conclusion, our results demonstrate that toxoplasmosis is a serious infection that has an important impact on public health. The lack of data on the burden of the disease is crucial for any future decision regarding public health intervention against Toxoplasma species infection. Better estimation of the burden, using better data, is necessary, and it is advisable to perform identical studies in other countries to stimulate public health interventions against this disease.
We thank Agnetha Hofhuis (National Institute for Public Health and the Environment, Bilthoven, The Netherlands) and Mariël Casparie (Prismant, Utrecht, The Netherlands) for analysis of the Pathologisch-Anatomisch Landelijk Geautomatiseerd Archief (PALGA) data. Joke van der Giessen (National Institute for Public Health and the Environment, Bilthoven, The Netherlands), Gouke Bonsel (Institute of Social Medicine, Academic Medical Center, Amsterdam, The Netherlands), Ruth Gilbert (Centre for Paediatric Epidemiology and Biostatistics, Institute of Child Health, London, United Kingdom), Fred Angulo (Centers for Disease Control and Prevention, Atlanta, Georgia), and 2 anonymous reviewers provided valuable advice and critical comments.
Financial support. Ministry of Health, Welfare, and Sports, The Hague, The Netherlands.
Potential conflicts of interest. All authors: no conflicts.