Risk Factors for Parvovirus B19 Infection in Pregnancy FREE

In 1983, Anderson et al¹ identified parvovirus B19 infection as the probable cause of erythema infectiosum, also known as fifth disease. Parvovirus B19 also causes chronic anemia in immunocompromised hosts, transient aplastic crisis in patients with increased red blood cell turnover, and may be a rare cause of arthritis.^{2 - 3} Perhaps most public attention has focused on the finding that infection during pregnancy may increase a woman's risk of fetal death.^{4 - 5}

Parvovirus B19 infection is believed to be transmitted mainly through respiratory secretions.^{6 - 7} The importance of contact patterns in the spread of this infection is supported by several reports of local outbreaks of erythema infectiosum in schools,^{8 - 9} and of high attack rates among household contacts of cases of parvovirus B19 infection.¹⁰ In addition, it has been reported that day-care personnel and elementary school teachers are at high risk of infection, probably because of their close contact with children.^{11 - 12} The current opinion is that parvovirus B19 infection during pregnancy is a relatively rare event, but epidemic incidence rates in pregnant women have not been established. Furthermore, there are no data on the relative importance on a population level of the different risk factors for infection.

The aims of this study were to assess endemic and epidemic rates of parvovirus B19 infection in a population-based sample of pregnant women and to quantify the relative importance of previously hypothesized risk factors for infection, such as family structure, housing density, length and type of education (reflecting occupation), and socioeconomic status.

The study population was restricted to pregnant women who lived in Copenhagen, Frederiksborg, or Southern Jutland counties in Denmark. These counties have a population-based, compulsory screening program for syphilis among pregnant women and for inborn errors of metabolism in newborn infants, which is centralized at Statens Serum Institut in Copenhagen. For this study, all serum samples drawn during the first trimester for syphilis testing between November 1992 and June 1994 were stored at −20°C until parvovirus testing took place. Dried blood spots (DBS samples) on filter paper sent for newborn screening were obtained for all pregnancies.

A total of 32,223 serum specimens were received during the study period. To avoid problems with dependence between observations, the material was restricted to the first specimen in case a woman contributed more than 1 sample, leaving 30,946 specimens from women aged 13 to 47 years (mean age, 28.9 years).

The DBS samples from newborns from the 30,946 pregnancies were identified based on a specific linked maternal identification number. The DBS sample was taken a median of 5 days after birth (95% of the material ranged from 4-9 days after birth). Through linkage with national registers, the outcomes of the 30,946 pregnancies were obtained (Table 1). A total of 27,572 pregnancies ended with at least 1 live birth, and in 26,219 (95.1%) of these, a DBS sample from at least 1 infant was identified.

Presence of parvovirus B19 IgG antibodies was determined by enzyme-linked immunoassay (ELISA) using the IDEIA Parvovirus B19 IgG kit (Dako A/S, Copenhagen, Denmark) in accordance with the manufacturer's instructions. To test DBS samples, antibodies were eluted from a 3.2-mm-diameter disk of the DBS sample in 120 µL of the sample diluent buffer supplied with the kit. This eluate was tested as described for serum samples in the manufacturer's instructions.

A unique 10-digit Central Registration System (CRS) number is assigned to all Danish residents shortly after birth or immigration by the CRS registry, which keeps information about date and place of birth, updated vital status, and a cross-reference between children and parents. Because other national registers keep personal information under the CRS number, identity-secure linkage of records is possible.

From the CRS registry we obtained complete information on a pregnant woman's spouse and the outcome of her present and previous pregnancies. Information on siblings, including their birthdates, was obtained on pregnant women (n=20,162) whose mothers themselves were born after April 1, 1935. Thus, the CRS registry contains complete linkage between children and mothers only for mothers born since that date.^{13 - 14}

Information on spontaneous abortions in the study population was obtained through register linkage to the Danish National Patient Register, which contains discharge diagnoses from all Danish hospitals. Information on induced abortions was obtained through the Danish National Board of Health Register of Induced Abortions. This register is based on mandatory reporting of all induced abortions in Denmark.¹⁵ Cases of stillbirth were identified through a linkage with the Danish Medical Birth Registry, which is a central register based on midwives' mandatory registration of all births.¹⁶

Type of education and scheduled length of education by January 1, 1994, were obtained through the Integrated Register of Education and Training. This registry contains person-specific information on ongoing and finished educations reported annually by all educational institutions in Denmark. The type of education was grouped by the expected level of occupational contact with children in different age groups.

The annual income in 1994 of the woman and her spouse was obtained from the tax authorities who register the net taxable income per year. We used the highest of these incomes as a measure of the family's economic status, referred to as the highest taxable income in the household.

We calculated housing density measures, including the number of persons per room and the number of square meters per person, by linking information from the Register of Building and Dwelling with information from the CRS registry on the number of persons living in the same household as the pregnant woman. The National Building and Dwelling Register holds information about all buildings in Denmark, including address, number of rooms, and number of square meters used for living.

Presence of specific parvovirus IgG antibodies during the first trimester of pregnancy was considered evidence of past infection. Seroconversion in women who were seronegative at this early point in pregnancy was determined by detection of maternal IgG antibodies in the DBS sample from the newborn. The beginning and end dates of epidemic periods and seroconversion rates in endemic (R_end) and epidemic (R_epi) periods were estimated simultaneously under the assumption of constant intensities for seroconversion during each of these periods. Thus, the beginning and end dates of the epidemic period were estimated without any initial identification of the epidemic period. The probability of seroconversion for a woman who spends time T_end during an endemic period and time T_epi during an epidemic period is 1−exp([−R_endT_end]−[R_epiT_epi]). This means that for fixed beginning and end dates of an epidemic period, R_end and R_epi may be estimated using a generalized linear model with binomially distributed response and logarithmic link function.^{17 - 18} The likelihood of this model was further maximized according to beginning and end dates of epidemic periods using the Newton Raphson algorithm.¹⁹ Confidence intervals (CIs) for these 4 parameters were estimated by a bootstrap resampling procedure in which each woman was used as sampling unit (2000 bootstraps).

Risk factors for past and current infection were analyzed in logistic regression models. In the analyses, the explanatory variables were grouped as shown in Table 2, Table 3, Table 4, through Table 5. All analyses of risk factors for past infection were adjusted for maternal age at pregnancy and date of test. Maximum likelihood estimation was performed using the GENMOD procedure in the SAS 6.12 statistical software package.²⁰ All P values were based on likelihood ratio tests, and CIs for the various risk factors were calculated by the use of the Wald test. Trends for the variables of interest were scored as their exact values.

The population-attributable risk for current infection was estimated as the fraction of the total number of seroconversions in the population that would not have occurred if the effect of a specific risk factor had been eliminated. The estimation was carried out as described by Bruzzi and colleagues²¹ using the adjusted relative risks and the exposure distribution among seroconverters.

In total, 20,122 (65.0%) of the 30,946 pregnant women were IgG antibody positive for parvovirus B19 during the first trimester. As shown in Table 1, 9693 (89.6%) of the 10,824 women who were seronegative at this stage of pregnancy had a recorded live birth, and a DBS sample was available for antibody testing on 9221 (95.1%) of these newborns. The mean time interval between the serum specimen taken during pregnancy and the DBS sample was 211 days. A total of 220 (2.4%) of the DBS samples were found to be IgG antibody positive, ie, the mothers had seroconverted. An epidemic period was identified from February 17 to June 15, 1994. During endemic periods, the seroconversion rate was estimated to be 1.5% (95% CI, 0.2%-1.9%) per year compared with 13.0% (95% CI, 8.7%-23.1%) per year in the epidemic period.

As shown in Table 2, the prevalence of parvovirus B19–specific IgG antibodies increased only modestly by age from 63.6% among women younger than 20 years to 68.3% among those 35 years or older. However, it increased significantly with the woman's number of siblings at age 10 years. For each extra member of the sibship, the relative increase in the odds ratio (OR) was 16% (95% CI, 11%-20%).

The age interval to the nearest sibling also affected the risk of past infection, ie, women born within 2 years of a sibling had a higher risk than women with siblings at all other age intervals. The pregnant woman's number of own children was another strong predictor of a positive IgG test result.

Another strong determinant of previous infection was the type of education (P<.001). As shown in Table 2, women who were educated to take care of children younger than 7 years (nursery school teachers) had an OR of 1.82 (95% CI, 1.43-2.33), whereas women whose education implied occupational contact with children aged 7 to 16 years (educated personnel at after-school clubs and day-care centers for children with behavioral problems and teachers) had an OR of 1.30 (95% CI, 1.09-1.55). Women educated as licensed vocational nurses or registered nurses had an OR of 0.88 (95% CI, 0.79-0.99). Table 3 gives risk estimates associated with type of education in the most detailed classification available.

The risk of past infection significantly declined with increasing length of education. However, age at the beginning of the pregnancy, highest taxable income, number of square meters per household member, and number of persons per room had no influence on the risk of past infection in the fully adjusted model.

As shown in Table 4, the risk of acute parvovirus B19 infection during pregnancy increased significantly with the number of children already born into the family. Compared with nulliparous women, women with 1 child had a more than 3-fold greater risk, increasing to 7.54-fold (95% CI, 3.80-14.94) in women with 3 or more children. Women educated as nursery school teachers had a risk of 3.09 (95% CI, 1.62-5.89), and the group of women educated as teachers or as personnel at after-school clubs and day-care centers for children with behavioral problems had a risk of 1.64 (95% CI, 0.95-2.85). While age at pregnancy had a significant effect on the risk of infection in the univariate analysis, this effect was no longer present when taking into account the number of children at the beginning of pregnancy and the type of education.

Among pregnant women with only 1 previous child, the age of that child was significantly correlated with the woman's risk of acute infection during pregnancy (P<.001) (Table 5). The highest risk of acute infection was observed in mothers of children aged 5 to 7 years. These women had a more than 4-fold increased risk of acute infection compared with the risk of mothers with a child younger than 2 years.

The exposed population-attributable risks were estimated to be 55.4% for the number of children and 6.0% for the group of education, ie, 55.4% of the seroconversions would not have occurred had all women been nulliparous, and 6.0% would not have occurred had all women had an education in the reference category.

Our results indicate that the current perception of pregnant women being at modest risk of parvovirus B19 infection may apply only to endemic situations. During epidemics, susceptible pregnant women are at a substantial risk of B19 infection, a risk closely associated with their level of contact with children. Most infections during pregnancy were attributable to exposure from a woman's own children and much less so to occupational exposure.

Sixty-five percent of the pregnant women were immune at the time of pregnancy, which is in line with studies from the United States, Germany, and the United Kingdom on seroprevalence in persons of childbearing age.^{22 - 24} Previous studies on the transmission of parvovirus B19 have primarily focused on secondary attack assessments in connection with specific outbreaks of erythema infectiosum in schools, hospital departments, or day-care centers.^{8 - 9 ,11 ,25 - 26} Such studies have reported attack rates for susceptible household contacts of between 18% and 70%.^{9 - 10 ,27 - 28} The present study, which is 25 to 100 times larger than any previous study, addressed the risk of parvovirus infection in a population-based sample of pregnant women during both endemic and epidemic periods. The nature of the study design excluded biases related to differential misclassification and selection bias. Thus, all covariate information was obtained independently of the exposure information and prior to the time of the exposure.

In our study the incidence among pregnant susceptible women during nonepidemic periods was as low as 1.5% per year, whereas it increased to 13% per year during epidemic periods. In a small study of 235 initially seronegative women of childbearing age who were studied through a period without community epidemics, Koch and Adler²⁹ found a seroconversion rate similar to our nonepidemic rate. Other reports obtained during nonepidemic periods have found seroconversion to vary between 2.8% and 6.8% in susceptible pregnant women.^{30 - 31} Our rates probably represent a slight underestimate of the true incidence of parvovirus B19 infection in pregnant women, as a few of the women who were IgG seropositive on their first sample may have just become infected. Furthermore, our incidence figures were based exclusively on seroconversions in successful pregnancies in which the neonate survived through the first 5 days of life, as no follow-up sample was obtained in the event of fetal loss or neonatal death prior to the DBS sampling. Fetal loss has been estimated to occur in 1% to 9% of the cases of maternal infection during pregnancy.^{5 ,30 ,32} Considering a worst-case scenario of 9% fetal loss, a corrected endemic incidence would be 1.6% per year (1.5/0.91) instead of 1.5%, and for the epidemic period, 14.3% per year (13.0/0.91). In contrast to the missing DBS samples in cases of fetal loss or perinatal death, the missing DBS samples in 4.9% of reported live births in the present study are not likely to have influenced our incidence figures. We found no indication that missing samples represented morbidity-selected pregnancies.

Important risk factors for past infection at time of pregnancy were the woman's number of siblings and an interval to the nearest sibling of less than 2 years. This suggests that transmission is more frequent in siblings of approximately the same age, who may have more and closer contact with each other than siblings born at greater age intervals. Number of children already born into the house significantly increased the pregnant woman's risk of past and acute infection. Furthermore, pregnant women with a child in the household aged 5 to 7 years had a much higher risk of becoming infected with parvovirus B19 than pregnant women with children younger than 2 years. This is in line with previous indications that children during early school years represent a particular hazard for infection.¹⁰

The intensity of child exposure at work was another significant risk factor for past and acute parvovirus B19 infection. In the present study we used information on the woman's educational background as a measure of her occupation, because it is more precisely registered than a person's occupation. Using education as a measure of occupation has undoubtedly introduced some misclassification, but such a misclassification is likely to be limited for several reasons. The Danish labor market is strictly regulated, and each of the occupations studied requires a specific education. Thus, persons without such education will not qualify; the trade unions carefully guard these rules. Second, our study group was rather young, with a mean age of 28.9 years. Thus, these women, at this time in their careers, are most likely in an occupational situation reflecting their educational background. The highest risk of past and acute infection was seen in women educated as nursery school teachers and as personnel in after-school clubs, which in Denmark implies occupational contact with children younger than 7 years, and between 6 and 9 years, respectively. This finding is in line with our observation that children aged 5 to 7 years may be most infectious.

Previous studies have indicated increased risks of parvovirus B19 infection in health care personnel in certain wards.^{22 - 23} Although we found a 61% increased risk of acute infection among licensed vocational nurses, registered nurses were not at increased risk of infection. With the exception of outbreak situations in a ward, our observation suggests that occupational exposure in health care personnel in general is limited.

Evidence that socioeconomic factors influence the risk of parvovirus B19 infection was seen only for increasing length of education, which lowered the risk of past infection. Neither taxable income nor crowding measured as square meters per person and number of persons per room had a significant effect on the risk of past infection. Whereas the quality of the information on taxable income and square meters per person is of high validity owing to the reliable registration of the figures in Denmark, the number of rooms is more likely to be associated with some misclassification. Reporting a change in number of rooms is not obligatory unless it involves an expansion of the total number of square meters used for living. Although our ability to detect small effects of crowding was limited, it is reasonable to conclude that number of children in a household is more important to the hazard of parvovirus B19 transmission than the number of square meters available for the family.

In conclusion, susceptible women are at substantial risk of B19 infection during epidemics. In general the importance of occupational exposure appears to be specifically restricted to nursery school teachers and persons in contact with children aged 5 to 7 years. However, the number of children in the household seemed by far the most important risk factor for parvovirus B19 infection during pregnancy.