A prospective longitudinal population-based study of clinical miscarriage in an urban Swedish population

Authors

  • F Blohm,

    Corresponding author
    1. Department of Obstetrics and Gynecology, Sahlgrenska Academy at Göteborg University, Göteborg, Sweden
      Dr F Blohm, Department of Obstetrics and Gynecology, Sahlgrenska University Hospital/Östra, SE-416 85 Göteborg, Sweden. Email febe.blohm@vgregion.se
    Search for more papers by this author
  • B Fridén,

    1. Department of Obstetrics and Gynecology, Sahlgrenska Academy at Göteborg University, Göteborg, Sweden
    Search for more papers by this author
  • I Milsom

    1. Department of Obstetrics and Gynecology, Sahlgrenska Academy at Göteborg University, Göteborg, Sweden
    Search for more papers by this author

Dr F Blohm, Department of Obstetrics and Gynecology, Sahlgrenska University Hospital/Östra, SE-416 85 Göteborg, Sweden. Email febe.blohm@vgregion.se

Abstract

Objective  To describe the incidence of clinical miscarriage and to investigate the factors influencing the occurrence of clinical miscarriage.

Design  Prospective study with both cross-sectional and longitudinal comparisons.

Setting  City of Göteborg, Sweden.

Population  Population-based study in cohorts of 19-year-old women followed longitudinally.

Main outcome measures  Incidence of miscarriage and pregnancy outcome.

Material and methods  A postal questionnaire was sent to women born in 1962 and resident in the city of Göteborg in 1981 (n= 656) regarding pregnancy outcome, clinical miscarriage and other reproductive health factors. Responders in 1981 were contacted again and requested to answer a similar questionnaire every fifth year up to 2001. The same process was repeated in 1991 with women born in 1972 (n= 780) with follow up of these responders in 1996 and 2001. A third cohort of 19-year-old women born in 1982 (n= 666) was interviewed in 2001. The self-reported pregnancy data were verified from hospital files.

Results  Complete data were available for 341 women born in 1962 and assessed up to the age of 39 years (ever pregnant, n= 320, 94%). There were in total 887 pregnancies (live birth, n= 590, 67%; miscarriage, n= 108, 12%; legal abortion, n= 173, 20% and ectopic pregnancy, n= 16, 2%). Of the 320 ‘ever pregnant’ women, 80 women (25%) had experienced a miscarriage. 76.3% had experienced one miscarriage, 16.3% had two miscarriages and 7.4% had three or more miscarriages. The clinical miscarriage rates in women at different ages were as follows: 20–24 years 13.5%, 25–29 years 12.3%, 30–34 years 10.3% and 35–39 years 17.5%. The corresponding miscarriage rate in the 1972 cohort followed from 19 to 29 years of age was 11%, and in the 1982 cohort assessed at 19 years of age, the miscarriage rate was 9%. No risk factor for miscarriage could be reliably identified.

Conclusions  Clinical miscarriage constituted 12% of all pregnancies, and one in four women who had been pregnant up to 39 years of age had experienced a miscarriage. Three or more miscarriages were experienced by 7.4%. The occurrence of a miscarriage was not influenced by the order of the pregnancy.

Introduction

Miscarriage is a frequent complication of human pregnancy—it has been estimated that one in four women who become pregnant will experience a miscarriage during their reproductive years.1 Clinical miscarriage has been estimated to account for approximately 15–20% of all pregnancies.1–8 Pregnancy tests, using immunological techniques based on antibodies against parts of the urinary human chorionic gonadotrophin molecule,9 became available for broad clinical use during the 1960s and 1970s and with their introduction the failure of human conception has been shown to be considerably higher.4,5,10,11

Several factors have been found to influence the occurrence of clinical miscarriage.1,2,6,12–22 Increasing maternal age has often been reported to be a risk factor for miscarriage,6,7,12 and a relationship between smoking and miscarriage has also been reported.14–16,21 Other factors that have been implicated in the aetiology of miscarriage are exercise during pregnancy,18 body weight20,22 and previous use of oral contraceptives.7,17

In 1981, we initiated a prospective longitudinal study of contraception and pregnancy outcome in a random sample of 19-year-old women resident at that time in the city of Göteborg. The initial results were published in 1982 as a cross-sectional study.23 The responders in 1981 have thereafter been contacted again every fifth year.24,25 Thus, the same women have been followed longitudinally from 19 to 39 years of age. New cohorts of 19-year-old women were recruited in 1991 and 2001, and these women have also been followed in a similar fashion every fifth year.

The primary aim of this study was to describe the incidence of clinically recognised miscarriage in these populations of women. A secondary aim was to evaluate possible factors influencing the occurrence of clinical miscarriage.

Material and methods

Study population

In 1981, a prospective longitudinal population study of women resident in the city of Göteborg was initiated.23 The women included were all born in 1962 and were 19 years of age (the 1962 cohort). This age was chosen to be able to deal with women of legal age, enabling information to be collected through the individuals themselves without the necessity of consent from their parents.

There were 2621 women aged 19 years living in Göteborg in 1981. A random sample of every fourth woman was obtained from the population register (n= 656). They were contacted by letter and asked to return a questionnaire concerning reproductive history and related factors. The study design was approved by the Ethics Committee, Faculty of Medicine, University of Göteborg and the National Data Inspection Board. The questionnaire was completed and returned by 594 women (91%). The responders were considered to be representative, as there were no significant differences between responders and nonresponders regarding civil status and immigrant and descendent status or other measurable background factors.23 Women who returned the questionnaire in 1981 and who were still available in the National Population Database were contacted again every fifth year, i.e. in 1986, 1991, 1996 and 2001 and requested to answer questions on possible pregnancies and reproductive health.

In 1991, a new cohort of 19-year-old women (born in 1972) was recruited and they have been followed in a similar fashion every fifth year (1996 and 2001) since inclusion (the 1972 cohort). In 2001, a third cohort of 19-year-old women (born in 1982, the 1982 cohort) was recruited, and they were assessed using the same techniques as for the cohorts born in 1962 and 1972.

Assessment

The women were invited by letter every fifth year to complete and return a postal questionnaire concerning contraceptive use, reproductive history and related factors. If no reply was received, reminder letters were sent after 2 and 4 weeks. Reproductive history and pregnancy outcome were also obtained from the woman’s hospital records. In this respect, it is important to note that all legal abortions, management of clinical miscarriage and deliveries were carried out in hospitals in Sweden during this time period. Twenty-seven women in 1962 cohort, 5 in 1972 cohort and 3 in 1982 cohort were found to have not reported their legal abortions compared with the hospital records. One woman who was nulliparous had reported three children, but the hospital records showed that these had all been adopted. Also, one woman in 1972 cohort had not reported a miscarriage. The pregnancy outcome in these cases was reported as found in the hospital records. A clinical miscarriage was defined as a miscarriage where the pregnancy had been confirmed by the patient’s doctor or by a conventional pregnancy test.4,26 Recurrent miscarriage was also defined according to the ‘Updated and revised nomenclature for description of early pregnancy events’.26

Data analysis and statistical methods

The accuracy of the data entry was checked on an individual basis for each parameter in all subjects. Fisher’s exact test, Student’s t test and analysis of variants with Tukey post hoc comparisons were used in the study of comparison of clinical miscarriage and pregnancy outcome. An initial analysis of the influence of a number of individual factors (e.g. smoking, exercise habits, body mass index [BMI] and ever use of the combined oral contraceptive [COC] pill) on the occurrence of clinical miscarriage was performed using Fisher’s exact test and odds ratio ± 95% confidence interval. This type of analysis assesses the individual factor independent of all other factors and does not in any way take into account the possible confounding influence of other factors. Therefore, a multiple regression analysis (SAS 9.1.3, SAS Institute Inc., Cary, NC, USA) was performed to take into account the possible confounding influence of these different factors when included simultaneously.

Results

In 1981, the questionnaire was completed and returned by 594 of the 656 women born in 1962 who were contacted (response rate 91%; Figure 1). The first follow up was performed in 1986 using the same postal questionnaire technique. Addresses were available for 589 of the original 594 responders (1 woman had died and 4 women had emigrated, and no forwarding addresses were available). The questionnaire was completed and returned by 489 women, which constituted 75% of the original 656 women contacted 5 years earlier. In 1991, a new attempt was made to contact the responders from 1981 using the same postal questionnaire technique. Addresses were available for 580 of the 594 responders from 1981 (1 woman had died, 2 women were temporarily living abroad and 11 women were no longer available in the register). Three hundred and seventy of the 580 women were still resident in city of Göteborg. In 1996 and 2001, an attempt was made to contact the responders from 1981 who were still available in the National Population Register. In 1996, 502 women were available and 430 answered (86%), and in 2001, 427 were available and 393 responded (92%). In summary, 341 women born in 1962 completed the questionnaire on all five occasions (i.e. 1981, 1986, 1991, 1996 and 2001) and are included in the longitudinal analyses included in this study. They constitute 52% of the original sample of 656 women first assessed in 1981.

Figure 1.

Flow chart illustrating the inclusion of the women from the three birth cohorts and their continued participation in this longitudinal study.

In 1991, a new cohort of 19-year-old women (born in 1972) was obtained at random from the Population Register. There were 2342 19-year-old women resident in the city of Göteborg in 1991, and every third woman (n= 780) was sent the same postal questionnaire. The response rate was 82%, and responders from 1991 have been reassessed in 1996 and 2001. In 2001, a third cohort of 19-year-old women (born in 1982) was enrolled. There were 1998 19-year-old women resident in the city of Göteborg in 2001, and every third woman (n= 666) was sent the same postal questionnaire. The response rate was 77%. A flow chart describing the three cohorts of 19-year-old women and their participation in this study appears in Figure 1.

A comparison of pregnancy outcome in all the women who answered the first questionnaire distributed at 19 years of age in the three cohorts was performed The total number of women who had been pregnant at least once at 19 years of age in the 1982 cohort (38/511, 7.4%) was lower (P < 0.01) than those in the 1962 cohort (68/594, 11.4%) and the 1972 cohort (83/641, 12.9%). The proportion of live births up to 19 years of age was higher in the 1962 cohort (34.0%) than that in the 1972 cohort (25.7%) and the 1982 cohort (23.6%), whereas the situation regarding the proportion of legal abortions was reversed. The proportion of miscarriages up to 19 years of age was higher in 1972 cohort than that in the two other cohorts, but the number of miscarriages was too low to allow statistical evaluation.

The cumulative pregnancy outcome in the women from the three birth cohorts is shown in Table 1. A detailed description regarding pregnancy outcome in the women from the 1962 cohort (n= 341) who had been followed to the age of 39 years appears in Figure 2. There were in total 887 pregnancies among the 320 women (94%) who had been pregnant at least once up to the age of 39 years (live birth, n= 590, 66.5%; miscarriage, n= 108, 12.2%; legal abortion, n= 173, 19.5% and ectopic pregnancy, n= 16, 1.8%). In this cohort, 80 of the 320 women (25%) who had been pregnant at 39 years of age had experienced at least one miscarriage. The total number of miscarriages per woman in the 1962 and 1972 cohorts at 29 and 39 years of age is described in Table 2. The percentage distribution of miscarriages per woman born in 1962 at 39 years of age was as follows (Figure 2): one miscarriage 76.3%; two miscarriages 16.3% and three or more miscarriages 7.4%.

Table 1.  Cumulative pregnancy outcome in the three birth cohorts
 19811986199119962001
1962 cohort (n= 341)
Live birth22 (29.7)127 (50.4)356 (63.5)523 (66.7)590 (66.5)
Miscarriage5 (6.8)29 (11.5)67 (11.9)90 (11.5)108 (12.2)
Legal abortion47 (63.5)93 (36.9)131 (23.4)158 (20.2)173 (19.5)
Ectopic pregnancy0 (0)3 (1.2)7 (1.2)13 (1.6)16 (1.8)
Total number of pregnancies74 (100)252 (100)561 (100)784 (100)887 (100)
1972 cohort (n= 392)
Live birth 10 (21.7)89 (50.8)257 (66.2)
Miscarriage 8 (17.4)16 (9.2)42 (10.8)
Legal abortion 28 (60.9)70 (40.0)88 (22.7)
Ectopic pregnancy 0 (0)0 (0)1 (0.3)
Total number of pregnancies 46 (100)175 (100)388 (100)
1982 cohort (n= 511)
Live birth 13 (23.6)
Miscarriage 5 (9.1)
Legal abortion 37 (67.3)
Ectopic pregnancy 0 (0)
Total number of pregnancies 55 (100)
Figure 2.

Flow chart illustrating a detailed description of pregnancy outcome in the same women (n= 341) from the 1962 cohort who were followed up to 39 years of age.

Table 2.  The total number of miscarriages per woman in the 1962 and 1972 cohorts at 29 and 39 years of age
Number of miscarriages/pregnant women1972 cohort at 29 years of age (n= 242), n (%)1962 cohort at 29 years of age (n= 232), n (%)1962 cohort at 39 years of age (n= 320), n (%)
125 (75.8)32 (76.2)61 (76.3)
27 (21.2)8 (19.0)13 (16.3)
31 (3.0)0 (0.0)4 (5.0)
40 (0.0)1 (2.4)1 (1.2)
50 (0.0)1 (2.4)1 (1.2)
Total33 (100)42 (100)80 (100)

The number of pregnancies per 1000 women years and the number of miscarriages per 1000 women years was calculated to compare miscarriage rates at different ages (Table 3). The clinical miscarriage rates in women at different ages were as follows: 20–24 years 13.5%, 25–29 years 12.3%, 30–34 years 10.3% and 35–39 years 17.5% [not significant (NS)].

Table 3.  The number of pregnancies per 1000 women years and the number of miscarriages per 1000 women years during the different 5-year periods in the 1962 cohort
1962 cohortPregnancies per 1000 women yearsMiscarriages per 1000 women yearsPercentage of miscarriages among total number of pregnancies (%)
20–24 years of age104.4/100014.1/100013.5
25–29 years of age181.2/100022.3/100012.3
30–34 years of age130.8/100013.5/100010.3
35–39 years of age60.4/100010.6/100017.5

The influence of several factors (e.g. pregnancy order, smoking, exercise habits, BMI and ever use of the COC pill) on the occurrence of clinical miscarriage was investigated at 29 years old in the 1962 and 1972 cohorts. In 1962 cohort, 35 of the 80 women (44%) who had experienced a miscarriage had it in their first pregnancy. The corresponding figure for 1972 cohort was 42%. The influence of smoking, BMI, exercising regularly and ever use of the COC pill were analysed separately (Table 4). There was some evidence in these univariate analyses that smoking and ever use of the COC pill influenced the miscarriage rate. Therefore, these factors were then further analysed using multiple regression analysis (SAS) to identify risk factors influencing the occurrence of clinical miscarriage. In the multiple regression analysis, none of these factors was shown to significantly influence the occurrence of clinical miscarriage.

Table 4.  The influence of individual factors on the occurrence of clinical miscarriage in women from the 1962 and 1972 cohorts at 29 years of age. Difference assessed using Fisher’s exact test and Student’s t test
FactorWomen who have not had a miscarriage, n (%)Women who have had a miscarriage, n (%)Odds ratio 95% CI)Significance
Nonsmoker (n= 547)499 (91.2)48 (8.8) 
Smoker (n= 186)159 (85.5)27 (14.5)1.8 (1.1–2.9)P < 0.05
No exercise (n= 172)148 (86.1)24 (14.0) 
Exercise (n= 561)510 (90.9)51 (9.1)0.6 (0.4–1.0)Not significant
Never used COC (n= 112)108 (96.4)4 (3.6%) 
Ever use COC (n= 621)550 (88.6)71 (11.43)3.5 (1.3–9.7)P < 0.05
BMI, mean (SD)22.9 ± 3.923.4 ± 3.9 NS

Discussion

Clinical miscarriage constituted 12% of all pregnancies, and 25% of women who had been pregnant by 39 years of age had experienced at least one miscarriage. The occurrence of three or more miscarriages was reported by 7.4% of women. Miscarriage was not influenced by order of the pregnancy.

Conditions in Sweden are favourable for prospective longitudinal population studies. The Swedish Population Register with its personal identification number system provides up-to-date information on the total population and can be used to obtain random subgroups of the total population for the purpose of longitudinal epidemiological studies. Given these favourable circumstances for population studies, we started in 1981 a prospective longitudinal study of pregnancy outcome, contraceptive use and other reproductive issues in a random sample of 19-year-old women resident at that time in the city of Göteborg. As far as we are aware, there are no previous longitudinal studies describing reproductive history and related factors in the same women from the ages of 19 to 39 years.

The population register includes demographic data that makes it possible to compare the responders with the nonresponders in each sample. In the birth cohort of women born in 1962 initiated in 1981, there were no differences between responders and nonresponders regarding marital status or immigrant and descendent status at any of the first three assessments. However, in the later cohorts and with the progression of time, there were more women with immigrant or descendent status or who were single among the nonresponders. There were no differences between responders and nonresponders regarding the proportion of women still resident in the city of Göteborg at the follow-up assessments.

Pregnancy tests have been freely available in Sweden at Antenatal Clinics, Contraceptive Clinics, Youth Clinics and Abortion Clinics since the 1970s. Women have been encouraged to test for pregnancy if menstruation did not occur after 2 weeks of the expected date. This national policy also assisted in the validity of our assessments. However, during the course of this study, the sensitivity of pregnancy tests has improved, and this could theoretically have influenced the results. In addition, the study design was facilitated by the fact that the women themselves reported their pregnancies with confirmation of the data from the patients’ hospital case sheets.

In this population study, 94% of women who were 39 years of age had been pregnant and 67% of the pregnancies resulted in a live birth. The ectopic pregnancy rate of 2% did not differ from that previously described in Sweden.27

The clinical miscarriage rate in this study (12%) was somewhat lower than previously described in the literature where the majority of studies have reported a miscarriage rate of 15–20%.1–8 There are several possible explanations for this lower rate. Legal abortion was freely available in Sweden during the whole time period of this study and 20% of all pregnancies were terminated in legal abortion. In Sweden, 95% of legal abortions are performed during the first trimester. It is thus highly likely that pregnancies that otherwise would have ultimately resulted in a clinical miscarriage were terminated as legal abortions prior to the miscarriage occurring. In older publications reported in the literature,28 the clinical miscarriage rate was higher than in more recent reports, and this may in part possibly be due to interference with the pregnancy prior to legal abortion becoming more freely available.

Several factors have previously been shown to influence the occurrence of clinical miscarriage.1,2,6,12–22 Several studies have demonstrated an increased risk of miscarriage with increasing maternal age.2,3,6,12 This was confirmed in the present study where the miscarriage rate was higher in women > 35 years. Fetal loss has been reported to be already increased among those in their 30 years of age, and in women ≥ 40 years old, the risk for miscarriage has been reported to be five times higher than in women of age 31–35 years. This marked increase has been attributed to the increased incidence of chromosomal aberrations with increasing age.6,12,19

There is some evidence to suggest that miscarriage is more common during the first pregnancy rather than later pregnancies,1 but other authors do not agree with this statement.6 In the present study, miscarriage did not occur more often in the first pregnancy. An association between smoking and miscarriage has also been reported by some authors,15,16 while other authors have not been able to confirm an increased risk for miscarriage among smokers.14,21 In the present study when smoking alone was studied, an increased occurrence of miscarriage was found among smokers compared with nonsmokers. However, this was not confirmed in the multiple regression analysis of factors influencing the occurrence of miscarriage. The relationship between exercise during pregnancy18 and increased body weight20,22 and the occurrence of clinical miscarriage has been studied previously. Exercise during pregnancy is usually recommended, but excessive physical strain around the time of implantation has been reported to be associated with later miscarriage.18 In the present study, the occurrence of miscarriage was not increased among those who exercised regularly compared with those who did not exercise, and there was no difference in BMI between women who had had a miscarriage and those who had not had a miscarriage. Obesity (BMI > 30 kg/m2) has been shown to be associated with an increased risk of first-trimester miscarriage and recurrent miscarriage. In a retrospective study, the difference in abortion rate between obese individuals was 38.1% compared with 13.3% in women with a normal BMI.20,22 In the present study, there were too few individuals with a BMI > 30 kg/m2 to perform a worthwhile statistical analysis.

The relationship between ever use of the COC pill and duration of COC use and the occurrence of miscarriage has been studied. Ford and MacCormac17 found that in women older than 30 years, those who used COCs for >9 years had a reduced risk of miscarriage. Risch et al.,13 however, found no association with the duration of oral contraceptive use. In the present study, COCs were not found to be associated with the miscarriage rate according to the multiple regression analysis of factors influencing the occurrence of miscarriage.

Conclusion

Clinical miscarriage constituted 12% of all pregnancies and one in four women had experienced a miscarriage by 39 years of age. Three or more miscarriages were reported by 7.4% of women. Miscarriage was not influenced by order of pregnancy.

Commentary on ‘A prospective longitudinal population-based study of clinical miscarriage in an urban Swedish population’

There are many snapshot views in the medical literature, but longitudinal views are hard to come by. The cohort data by Blohm et al. following women closely over several decades are thus impressive. The design, conduct, analysis and reporting are rigorous: (a) representative samples of women at a uniformly early time, at the age of 19 years, were assembled; such cohorts are called inception cohorts and are important to avoid missing outcomes by ‘starting the clock late’; (b) the follow up of the study was sufficiently long, with the longest follow-up period being 20 years (from 19 to 39 years) for the 1962 cohort; (c) outcome assessment is sound, relying on self-reported pregnancies with corroboration using hospital records, which clearly is important for ensuring accuracy as amply demonstrated by the authors; furthermore, self-reporting averts the common criticism of underestimation of miscarriage rates in studies that employ hospital data alone, as not all women with miscarriages get seen or admitted and (d) ascertainment bias is unlikely as pregnancy tests are impressively accurate, at least in recent decades. One obvious weakness in this study is the high attrition in follow ups, particularly with the later cohorts, but this is the norm in these days of highly mobile societies. Overall, this study is likely to have high internal validity, but what about ‘external validity’ (otherwise known as ‘generalisability’ or ‘applicability’)?

Whether data from urban Swedish women applies to your population is a judgement you would need to make; there is not a lot of science to judging external validity: the key question to ask is ‘Are my patients so different from those in the study that its results cannot apply?’ (Straus et al., EBM: How to Teach and Practice, 3rd edn, Elsevier). Take note that this is a different question to the commonly asked question: ‘Are my patients similar enough to those in the study that its results apply?’ Often, we are unreasonably stringent and reject evidence too quickly for a presumed reason of limited external validity. We, in BJOG, certainly judge the evidence from Blohm et al. to have wider applicability, for if we did not, we would have urged the study authors to publish the results in a Swedish journal!

Much of the findings agree with what we know already. The apparent surprise, however, is the 7.4% incidence quoted for women suffering three or more miscarriages; we are used to quoting a 1% incidence of recurrent miscarriage (Stirrat, Lancet 1990;336:673–5). Even taking into account that the quoted 7.4% may not all strictly be ‘recurrent miscarriers’ as the definition of recurrent miscarriage (RM) requires the miscarriages to be consecutive, the figure seems too high. Why should this be the case? The reason is due to the choice of denominators: the study authors report ‘3 or more miscarriages in those who miscarried’ (6/80 = 7.5%), while the commonly quoted RM risk of 1% refers to ‘3 or more consecutive miscarriages in all women’. This example illustrates the critical nature of the choice of the denominator in drawing inferences.

Arri Coomarasamy

Acknowledgements

Important contributions to this longitudinal study, which has now been in progress for over 20 years, have also been made by the following persons: Björn Andersch M.D., Ph.D., Anne Marie Kullendorff M.D., Gerd Larsson M.D., Ingela Lindh, Marianne Sahlén, Gunilla Sundell M.D and Anja Andersson. We also wish to thank Björn Areskoug for his valuable assistance regarding the statistical analysis. The study was supported by grants from the Göteborg Medical Society, Hjalmar Svensson’s Fund and from the University of Göteborg.

Ancillary