Socioeconomic position and the risk of spontaneous abortion: a study within the Danish National Birth Cohort.

OBJECTIVES: To investigate the relationship between different indicators of socioeconomic position and the risk of spontaneous abortion.
DESIGN: Cohort study.
SETTING: 1996-2002, Denmark. PARTICIPANTS: All first time participants, a total of 89 829 pregnant women, enrolled in the Danish National Birth Cohort were included in the present study. Overall, 4062 pregnancies ended in spontaneous abortion. Information on education, income and labour market attachment in the year before pregnancy was drawn from national registers. MAIN OUTCOME MEASURE: Spontaneous abortion, that is, fetal death within the first 22 weeks of pregnancy, was the outcome of interest. The authors estimated HRs of spontaneous abortion using Cox regression analysis with gestational age as the underlying time scale.
RESULTS: Women with <10 years of education had an elevated risk of spontaneous abortion when compared with women with >12 years of education (HR 1.19 (95% CI 1.05 to 1.34)). The HR estimates for the four lowest income quintiles were all increased (HRs between 1.09 and 1.15) as compared with the upper quintile but did not differ considerably from each other. In general, no statistically significant association was found between labour market attachment and the risk of spontaneous abortion; however, the group of women on disability pension had an increased HR of spontaneous abortion when compared with women who were employed (HR 1.32 (95% CI 0.82 to 2.13)).
CONCLUSIONS: Educational level and income were inversely associated with the risk of spontaneous abortion. As these factors most likely are non-causally related to spontaneous abortion, the findings indicate that factors related to social position, probably of the environmental and behavioural type, may affect spontaneous abortion risk. The study highlights the need for studies addressing such exposures in order to prevent spontaneous abortions.

Introduction

Social inequality is demonstrated in most reproductive outcomes, such as preterm birth, intrauterine growth retardation and stillbirth.1–4 This inequality indicates a preventive potential since the minimum level of these outcomes, in theory, should be attainable for all groups in society. Spontaneous abortion, that is, fetal death before 22 gestational weeks, is the most frequent adverse pregnancy outcome and affects many women and their relatives. Approximately one of six clinically recognised pregnancies result in spontaneous abortion5 6 and identification of even a small potential for prevention may have significant impact for public health.

Given the commonness of spontaneous abortion, surprisingly few studies have investigated the relationship with socioeconomic position, and no consensus about any association has been established. It has been shown that women with low socioeconomic position have an increased risk of spontaneous abortion when measured by educational attainment,7–10 while other studies have not supported this finding.11–13 When social position has been measured by labour market attachment, the associations seem even more unclear.7 12–14 Two studies that used income as a proxy measure of socioeconomic position did not find an association with the risk of spontaneous abortion.10 13 The risk of spontaneous abortion according to potential risk factors that are known to display a social gradient, such as alcohol drinking during pregnancy and smoking, have been examined individually12 15–19 but with no consistent results.

By examining how different measures of socioeconomic position are associated with the risk of spontaneous abortion, we might be able to come closer to identification of more proximal causal risk factors for spontaneous abortion.

The aim of this study is to describe how educational level, income and labour market attachment, respectively, are related to the risk of spontaneous abortion in a large cohort study.

Methods

We used data from the Danish National Birth Cohort (DNBC), which comprises 100 418 pregnancies recruited in the years 1996–2002. The pregnant women were invited to participate in the cohort at the first antenatal visit at the general practitioner. The women were included in the study if they had posted the informed consent form before gestational week 24, intended to carry the pregnancy to term and were able to complete a telephone interview in Danish. The DNBC is described in details elsewhere.20

For this study, we excluded women with no information on the date of the consent agreement (n=34), women who entered the cohort after 22 weeks of pregnancy (n=1661), women with no information of the date of the pregnancy outcome (n=81) and women with ectopic pregnancies (n=66) or mola hydatidosa (n=48). Furthermore, in cases where women participated in the cohort more than once (n=8699), we only included the women's first pregnancy in the analyses to meet the criteria of independent observations in the statistical model. Consequently, 89 829 pregnancies were eligible for analyses in this study (see figure 1).

Figure 1

Flow-chart of the population included in this study.

We used educational level, maternal income and labour market attachment as indicators of the pregnant women's socioeconomic position. This information was retrieved from national registers where the information of highest educational attainment, yearly income and predominant attachment to the labour market are registered on an individual level every year. We used the last information registered before the date of last menstruation period of the actual pregnancy. The ISCED (International Standard Class of Education) codes from Statistics Denmark were converted into four educational groups, reflecting the highest completed academic educational attainment or—for women in education—the level completion of the actual education would lead to (see table 1). Labour market attachment was categorised as: employed, students, unemployed, disability retired and unknown. To be categorised as unemployed, one had to be unemployed for more than 50% of the time in the year preceding conception. If the period of unemployment was less than half the year, one was categorised as employed. Income level was grouped into quintiles, and was based on the taxable income of the women the calendar year preceding the year of the last menstruation period before pregnancy.

Table 1

Socioeconomic characteristics of 89 829 women in the Danish National Birth Cohort according to age, Denmark, 1996–2002

	n	%	Age (years)
			<25	25–29	30–34	35–39	40+
			%	%	%	%	%
Educational level
BA or more (>12 years)	12 378	13.8	4.94	13.5	16.9	16.3	16.5
Higher education (less than BA degree)	26 627	29.6	13.8	31.8	31.9	34.1	36.7
Upper secondary education and vocational training	39 221	43.7	52.4	44.1	41.8	37.7	20.1
Compulsory school (<10 years)	10 753	11.9	27.9	9.7	8.4	10.9	13.9
Unknown	850	0.9	0.9	0.9	1.0	1.0	2.8
n	89 829	100	12 449	37 114	29 864	9420	982
Income quintile
>80%	17 728	19.7	2.2	13.5	29.3	35.5	38.6
60%–80%	17 725	19.7	8.9	19.8	23.4	22.4	19.5
40%–60%	17 723	19.7	16.5	21.2	19.9	17.8	18.7
20%–40%	17 724	19.7	28.3	22.5	15.2	12.8	10.8
<20%	17 772	19.7	42.9	21.9	10.8	9.9	9.7
Unknown	1207	1.3	1.3	1.2	1.4	1.7	2.8
n	89 829	100	12 449	37 114	29 864	9420	982
Labour market attachment
Employed	74 738	83.2	68.7	82.5	88.6	88.0	85.0
Student	6584	7.3	16.0	9.2	3.2	2.1	1.1
Unemployed (>50% of the year)	2766	3.1	2.2	3,0	3.3	3.8	5.0
Disability pension	238	0.3	0.1	0.2	0.3	0.7	1.7
Unknown	5503	6.1	13.4	5.2	4.6	5.5	7.1
n	89 829	100	12 449	37 114	29 864	9420	982

The outcome measure of interest was spontaneous abortion, defined as death and expulsion of an intrauterine pregnancy before 22 weeks of pregnancy.21 The gestational age was calculated using the self-reported first day in the woman's last menstrual period, which was stated on the informed consent form. Information about the occurrence of spontaneous abortion came primarily from the Danish National Patient Registry, where all women who had been diagnosed or treated in a hospital setting were registered. A small minority of the women had no pregnancy outcome in the Danish National Patient Registry, and for these, we used information from the participant herself.

To study whether the possible association between socioeconomic position and the risk of spontaneous abortion differed according to gestational age, spontaneous abortion was divided into first and second trimester of pregnancy, including spontaneous abortions at 84 days of gestation in the first trimester spontaneous abortions.

Statistical analyses

The HRs of spontaneous abortion according to three different measures of socioeconomic position were estimated using Cox regression analysis. Gestational age in days was used as the underlying time variable. We used a model with delayed entry, so that women entered the cohort on the gestational day of inclusion in the study. The follow-up ended at the gestational age at date of spontaneous abortion, emigration, maternal death or the day the woman completed the first 22 weeks of gestation (154 days), whichever came first. We conducted three sets of analyses estimating the relations between maternal educational level, maternal labour market attachment, maternal income level, respectively, and spontaneous abortion. Individual-level HRs were calculated for each category in comparison to a reference category, defined as the category hypothesised to have the lowest risk of spontaneous abortion. The risk association between the three measures of socioeconomic position and spontaneous abortion was adjusted for maternal age at the time of conception (<25, 25–29, 30–34, 35–39, ≥40 years). Furthermore, we conducted a multivariate analysis where we included all three indicators of socioeconomic position and maternal age. We conducted trend tests for the association between the socioeconomic variables and the rate of spontaneous abortion using Wald's test for trend. All statistical analyses were performed with the SAS software package V.9.2.

The DNBC data collection was approved by the Danish Scientific Ethics Committee, and this particular study was, according to Danish legislation, approved by the Danish Data Protection Agency.

Results

The mean gestational age of recruitment to the study was 78 days, 10% were recruited before 49 days and 10% were recruited after 112 days of gestation. A total of 4062 pregnancies resulted in a spontaneous abortion. Of these, 2146 were first trimester spontaneous abortions and 1916 were second trimester spontaneous abortions.

Table 1 shows the distribution of the three socioeconomic indicators according to maternal age and demonstrates the expected strong association between age and socioeconomic position.

When examining the effect of age on the risk of spontaneous abortion, we found different effects in the first and the second trimester, that is, the proportional hazards assumption was not fulfilled for age. In the final regression analyses, we therefore stratified the effect of age according to trimester.

The association between the three measures of socioeconomic position and spontaneous abortion were not different within the first and second trimester of pregnancy, while the overall associations between educational level, income and labour market attachment, respectively, and spontaneous abortion are presented. We found an inverse association between educational level and the risk of spontaneous abortion (table 2). Women with compulsory school as the highest educational level had an age-adjusted HR of 1.19 (95% CI 1.05 to 1.34) for spontaneous abortion when compared with those with a bachelor level or more (table 2). Women in the four lowest income quintiles had an increased risk of spontaneous abortion when compared with the group with the highest income level, though only two of the estimates reached statistical significance. Unemployed women and students had the same risk of spontaneous abortion as the employed women. However, the group of women on disability pension had an increased risk of spontaneous abortion compared with women who were employed (HR 1.32 (95% CI 0.82 to 2.13)).

Table 2

The risk of spontaneous abortion according to educational level, income level and labour market attachment, respectively, in the Danish National Birth Cohort

	No. of events	Crude	Age adjusted
		HR (95% CI)	HR (95% CI)
Educational level (n=88 958)
BA or more (>12 years)	589	1
Higher education (less than BA degree)	1239	1.01 (0.92 to 1.11)	1.02 (0.93 to 1.13)
Upper secondary education and vocational training	1668	0.97 (0.88 to 1.06)	1.01 (0.92 to 1.11)
Compulsory school (<10 years)	527	1.14 (1.01 to 1.28)	1.19 (1.05 to 1.34)
Income quintile (n=88 602)
>80%	785	1	1
60%–80%	787	1.01 (0.91 to 1.11)	1.1 (0.99 to 1.21)
40%–60%	796	1.03 (0.93 to 1.13)	1.15 (1.04 to 1.27)
20%–40%	740	0.93 (0.84 to 1.03)	1.09 (0.99 to 1.22)
<20%	773	0.95 (0.86 to 1.05)	1.15 (1.03 to 1.27)
Employment status (n=84 306)
Employed	3398	1	1
Student	295	0.91 (0.81 to 1.03)	1.03 (0.91 to 1.16)
Unemployed (>50% of the year)	128	1.04 (0.87 to 1.25)	1.01 (0.84 to 1.20)
Disability pension	17	1.61 (1.00 to 2.60)	1.32 (0.82 to 2.13)

Risks are expressed in HR.

For educational level and income level, we found significant trends (p values =0.01 and 0.04, respectively), while for labour market attachment, there was no clear trend (p value =0.50).

A multivariate model where all three measures of socioeconomic position and maternal age were included revealed essentially the same results for income and educational level, while the elevated risk for women on disability pension disappeared (table 3).

Table 3

The association between spontaneous abortion and educational level, income level and labour market attachment, respectively, in the Danish National Birth Cohort

	HR (95% CI)
Educational level
BA or more (>12 years)	1
Higher education (less than BA degree)	1.03 (0.91 to 1.11)
Upper secondary education and vocational training	1.01 (0.89 to 1.09)
Compulsory school (<10 years)	1.13 (0.98 to 1.29)
Income quintile
>80%	1
60%–80%	1.09 (0.99 to 1.21)
40%–60%	1.13 (1.02 to 1.26)
20%–40%	1.11 (1.00 to 1.23)
<20%	1.13 (1.01 to 1.27)
Employment status
Employed	1
Student	1.02 (0.89 to 1.16)
Unemployed (>50% of the year)	0.95 (0.79 to 1.14)
Disability pension	0.99 (0.57 to 1.72)

Risks are expressed in HR and are adjusted for maternal age and mutually adjusted for the different indicators of socioeconomic position (n=83 470).

Discussion

This study, based on data from all 89 829 women in the DNBC, displayed a social pattern in the risk of spontaneous abortion. Educational level and income were inversely associated with the risk of spontaneous abortion.

Apart from maternal age, no lifestyle risk factors for spontaneous abortion are well established and few studies have examined the association between socioeconomic position and risk of spontaneous abortion. A few previous studies have reported an association between socioeconomic position and spontaneous abortion7–10 when socioeconomic position was measured by educational level and labour market attachment, and others found no such association.11–14 There are several possible explanations for this. Some studies are small with a diminished possibility of detecting a smaller association. Furthermore, two of these studies adjust for earlier spontaneous abortion in their analyses.11 12 Previous spontaneous abortion is associated with a 60% higher risk of spontaneous abortion,22 suggesting that women vary in their baseline risk for this negative pregnancy outcome. Adjusting for earlier spontaneous abortion may therefore distort the possible association between socioeconomic position and (baseline) risk of spontaneous abortion, as we find it less likely that previous spontaneous abortion is a determinant of social position. Another possible explanation may be that only one of the studies uses prospectively collected data.

Why do women with lower socioeconomic position have an increased risk of spontaneous abortion? According to the association we have found between educational level and the risk of spontaneous abortion, it is possible that an overall healthier lifestyle among well-educated women may explain part of the effect. This is not a fulfilling explanation though, since some of the typical lifestyle factors seem to be socially patterned in opposite directions in the DNBC. Smoking for instance is socially patterned, with women of low socioeconomic position smoking more than women of higher socioeconomic position, but studies on the effect of smoking on spontaneous abortion risk are not consistent, and there are several studies reporting no effect of smoking on spontaneous abortion risk.12 18 23 Alcohol intake during pregnancy is also socially patterned and this exposure is strongly associated with the risk of spontaneous abortion.19 In the DNBC though women with high socioeconomic position more frequently reported to have an alcohol intake during their pregnancies.24 Therefore, typical lifestyle factors cannot solely explain the difference in spontaneous abortion risk we find according to socioeconomic position.

It is known that the ability to make use of the health system depends on educational level.25 26 However, it is not clear how this may affect the risk of spontaneous abortion since no preventive or curative measures for this negative pregnancy outcome is known. The association we found between income level and the risk of spontaneous abortion is difficult to interpret. The risk is increased at more or less the same scale for all the income groups compared with the group with the highest income level. A possible explanation could be that the Danish population is relatively homogeneous according to income as compared with other populations,27 and therefore, an association would be clearer in countries with a greater distinction in living circumstances between rich and poor.28 What we wished to examine, looking at the women's employment status before they got pregnant, was whether being outside the labour market had any influence on ones risk of spontaneous abortion. This seems not to be the case. One reasonable explanation might be that in Denmark there is a high social security for people outside the labour market. Another speculation could be that being outside the labour market pose a social risk, while being at the labour market pose several occupational risks and that these risks outweigh each other. The apparently elevated risk of spontaneous abortion for women on disability pension is not surprising, given these women are of remarkable worse health than the rest of the study population. However, the association seems to disappear when we adjust for income and educational level. A possible explanation for this may be that these women all have a low income and that mutually adjustment in this case probably is over adjustment.

The present study is based upon a large population and a considerable number of spontaneous abortions, which offers a good foundation to examine the association between different indicators of socioeconomic position and spontaneous abortion. Cohort studies are potentially subject to selection bias due to loss to follow-up. This is a minor issue in this study since 99.9% of the pregnancy outcomes have been identified. The information we have on the exposure measures is based on register data, which cover almost the whole population, and is therefore not dependent upon the outcome of the pregnancy. To study spontaneous abortion is difficult since a great part of spontaneous abortions happen very early in the pregnancy period—many even before the women themselves know that they are pregnant.29 This implies that many women do not have a chance to be recruited for pregnancy cohorts before the spontaneous abortion. The potential bias arising from that fact is taken care of by applying survival analyses with left truncation and gestational age as time variable, but this is why the proportion of pregnancies ending in spontaneous abortion in the DNBC is less than reported in the background population, and this is also the explanation why we are restricted from being able to conclude anything about the association between socioeconomic position and the very early spontaneous abortions before gestational week 6.

The women participating in the DNBC seem to be somewhat healthier than the rest of the population,30 though the difference is very moderate and the estimated effect upon the risk estimates obtained in internal comparisons are small. This means that it should be possible to transfer findings based on the DNBC to the background population. It cannot be excluded though that there is some bias related to selection and, if so, our results are most likely underestimated.

In this study, we were interested in the overall effect of socioeconomic position on the risk of spontaneous abortion. We did not adjust our analyses for typical lifestyle factors, for example, smoking, alcohol consumption and body mass index, since we believe them to play a role as mediating factors between socioeconomic position and the risk of spontaneous abortion. Conditioning on an intermediate will only be of interest if one wishes to examine something different from the overall effect, that is, the direct effects of the exposure on the outcome.31 In contrast, maternal age is a strong independent risk factor for spontaneous abortion32 and is also causally related to social position, and consequently, we believe that the age-adjusted analyses provide the most accurate estimates.

Conclusions

In this large cohort study, we found an inverse association between measures of socioeconomic position and the risk of spontaneous abortion. These findings indicate that at least some of the spontaneous abortions are preventable and highlight the need for further studies addressing which behavioural and environmental exposures, concentrated in groups with lower socioeconomic position, that are causal risk factors for spontaneous abortion.