The effect of reproductive history on future pregnancy outcomes

Elise Whitley1,4, Pat Doyle2, Eve Roman3 and Bianca De Stavola2

1 Department of Social Medicine, University of Bristol, Bristol BS8 2PR, 2 London School of Hygiene and Tropical Medicine, London and 3 Leukaemia Research Fund, University of Leeds, Leeds, UK


    Abstract
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
The aim of this study was to examine the separate and joint effects of previous pregnancy history, year of pregnancy outcome, maternal age, height, smoking and fertility on risk of fetal death. Data were available from a study of female radiographers. Analyses were carried out on 3053 women with a total of 6993 pregnancies. Women reporting problems with conception or previous fetal losses had an increased risk of a pregnancy ending in a fetal death. In particular, women with primary or secondary infertility had an approximately fourfold increase in risk compared with women who reported no difficulties [odds ratio (OR): 3.92; 95% confidence interval (CI): (3.02, 5.07)]. This relationship was independent of pregnancy order and pregnancy history and was more marked in older maternal ages. The effect of pregnancy history was cumulative and possibly multiplicative in effect, with a threefold increase in the risk of losing a third pregnancy following two previous losses [OR: 3.19; 95% CI: (1.60, 6.35)]. There were no consistent patterns of risk associated with year of pregnancy outcome, maternal age, height or smoking status. These results suggest that previous pregnancy outcomes and problems with conception may be the strongest determinants of fetal loss in subsequent pregnancies.

Key words: epidemiology/fetal death/infertility/reproductive history


    Introduction
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
It is estimated that between 10 and 20% of clinical pregnancies may result in a fetal death (Kline et al., 1989Go; Regan et al., 1989Go; Everett, 1997Go). Identification of determinants of pregnancy outcome is complex and it is difficult to disentangle the effects of related factors such as parity and maternal age (Kline et al., 1989Go).

The objective of this study was to examine the effects on fetal death (defined as stillbirth or miscarriage at any gestational age) of reproductive history and other maternal risk factors, and to explore the relationships between them. Analysis of the effect of reproductive history on future pregnancy outcome is known to be prone to bias (Olsen, 1994Go; Weinberg et al., 1994Go) and the methodological approach adopted here attempts to address these potential problems.


    Materials and methods
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Data were available from a study of the health of children born to radiographers (Roman et al., 1996Go) recruited from membership files of the College of Radiographers during the period 1988–1990. As most of the risk factors considered here are maternal, analyses have been restricted to women only. In addition to general characteristics, information was collected on the date and outcome of each pregnancy. Possible problems with conception were classified according to responses to two questions: (i) `Did you have difficulty in starting any of your pregnancies? (i.e. a year or longer to conceive)' and (ii) `Have you ever tried unsuccessfully to start a pregnancy?'. Women were classed as fertility type A with no fertility problems [(i) no; (ii) no], type B with delayed conception in one or more pregnancies [(i) yes; (ii) no], or type C with primary or secondary infertility [(ii) yes]. A relatively large number of women failed to answer the questions and so an additional `unknown' category was also required.

Odds ratios (OR) and 95% confidence intervals (CI) were calculated using logistic regression. P values were based on appropriate likelihood ratio tests for trend or interaction unless otherwise specified.

Preliminary analyses concentrated on all pregnancies combined, exploring the effects on fetal death of year of pregnancy outcome, maternal age, height, smoking and fertility. It has been suggested (Hakim et al., 1995Go) that the adverse effect of infertility may be greater in older women and an appropriate interaction term was also included to investigate this potential synergism. Robust standard errors were calculated (Huber, 1967Go) to allow for the correlation between women's successive pregnancy outcomes.

To allow for the problem of women with a history of fetal death being over-represented, additional separate analyses were carried out for each of pregnancy orders 1, 2 and 3 (there were insufficient numbers to continue this approach for higher order pregnancies). These analyses considered those factors found to be significantly associated with pregnancy outcome in the analysis of all pregnancies combined. For pregnancy orders 2 and 3, additional terms were included to assess the impact of previous pregnancy outcomes. In second pregnancies, previous history was included as the outcome of the first. For third pregnancies, previous history was included as: (i) two terms representing the outcomes of each previous pregnancy; and (ii) total number of previous fetal deaths.

Analyses were carried out using data from women who had only ever had singleton live births or fetal deaths. The majority of excluded women had induced abortions. To assess the impact of these exclusions, analyses were repeated using data from all women in three ways: (i) using data from all women but dropping individual pregnancies with an outcome other than a live birth or fetal death; (ii) using data from all women, with induced abortions treated as live births and dropping pregnancies with other outcomes; (iii) the analysis of previous pregnancy outcomes was repeated with outcomes other than a fetal death or live birth included as a risk factor for subsequent fetal losses. In all analyses, multiple births were included as live births. Because some women might not have completed their families by the time of the survey, all analyses were repeated for those aged 40 years or over.


    Results
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
A total of 5838 women were approached and 4847 (83%) returned a completed questionnaire. Of these, 1453 (30%) had never been pregnant, 68 (1%) had missing information, and 273 (6%) had a pregnancy outcome other than a singleton live birth or fetal death (75% of which were induced abortions). The remaining 3053 women had a total of 6993 pregnancies, of which 973 (14%) ended in a fetal death. The average gestational age of the fetal deaths was 12.0 weeks (SD: 6.1) with only 66 (7%) occurring at more than 20 weeks.

The women were aged between 30 and 64 years with 1686 (55%) aged 40 years or over (Table IGo). Two thousand and three (66%) were lifelong non-smokers, 674 (22%) ex-smokers, and 376 (12%) current smokers. The majority (80%) reported no difficulties in conceiving any pregnancies (fertility type A). However, 276 (9%) were classified as fertility type B, 144 (5%) type C, and 182 (6%) unknown. The total number of live births ranged from 0 to 6, with 1701 women (56%) reporting two. The total number of fetal deaths ranged from 0 to 8 with 2310 women (76%) reporting none. Sixty women (2%) reported only having pregnancies ending in a fetal death.


View this table:
[in this window]
[in a new window]
 
Table I. Characteristics and pregnancies of the study population
 
All pregnancies combined
There was a significant trend of increasing odds of fetal death with increasing year of pregnancy outcome (Table IIGo) and a significant J-shaped association between pregnancy outcome and maternal age. A slight decrease in the odds of fetal death was observed in women who reported smoking at the time of their pregnancy with a particularly low OR in those who smoked 15 or more cigarettes per day. However, the 95% CI around this OR only just excluded 1.00 and an overall test for heterogeneity was not significant. There was no significant relationship between maternal height and pregnancy outcome.


View this table:
[in this window]
[in a new window]
 
Table II. Odds ratio (OR) (95% CIa) for fetal death versus live birth for all pregnancies combined
 
Fertility was strongly related to pregnancy outcome with significantly more fetal deaths in women reporting some difficulty in conception. This was particularly marked in the pregnancies of type C women [OR: 3.92; 95% CI: (3.02, 5.07)]. Women with unknown fertility also had significantly higher odds of fetal death. The effect of maternal age was examined separately for each of the four fertility types (Figure 1Go). The highest odds of fetal death occurred consistently in type C women, except in those aged <25 years at the time of their pregnancy, and this became increasingly more marked as maternal age increased (P value for interaction = 0.003).



View larger version (14K):
[in this window]
[in a new window]
 
Figure 1. Odds of fetal death by fertility type and maternal age. For definition of fertility types A, B and C, see Table IGo.

 
Individual pregnancy orders
The total number of pregnancies decreased with increasing pregnancy order (Table IIIGo) while the proportion ending in a fetal death increased from 13% in first and second pregnancies to 17% in third pregnancies, possibly reflecting some degree of reproductive compensation or poor recall of early fetal losses. There was some evidence of increasing odds of fetal death with increasing year of pregnancy outcome in first pregnancies but this was not apparent in higher orders. There was little sign of an association with maternal age and an apparent trend in third pregnancies did not reach statistical significance. Conversely, there was a highly significant association with fertility type with very similar ORs in all three pregnancy orders.


View this table:
[in this window]
[in a new window]
 
Table III. Adjusted odds ratio (OR)a (95% Confidence interval) for fetal death versus live birth for individual pregnancy orders
 
There was a consistent pattern of increasing odds of fetal death in women with previous fetal losses. The number of previous losses was also related to the third pregnancy outcome with the odds of fetal death in women with two previous losses more than three times greater than in women with none.

When women with other pregnancy outcomes were included, the results of the analyses were very similar (data not shown). There was some evidence of a raised risk of fetal death following one of these other outcomes (mostly induced abortions) but the numbers involved were small. Restricting attention to women over 40 years revealed that, although the older women had more live births in total, the proportion with at least one fetal death was almost identical to that in younger women and the results of the analyses were very similar (data not shown).


    Discussion
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
These analyses support the notion that women reporting difficulties with conception may have an increased risk of having a pregnancy ending in a fetal death (Templeton et al., 1991Go; Coulam, 1992Go; Joffe and Li, 1994Go; Hakim et al., 1995Go). One possible explanation for this association is that apparently infertile women actually do conceive but miscarry very early in the pregnancy (Hakim et al., 1995Go). There is also evidence in these data to suggest that the effect of subfertility may be particularly marked in older women (Hakim et al., 1995Go), although this could be due to older women having more opportunity or tendency to report difficulties with conception. The apparent significant increase in risk with unknown fertility status is perplexing but could be explained by women with more erratic reproductive patterns, i.e. several fetal losses, being less sure of dates and times to conception.

The pattern of increasing risk of fetal death following a previous loss has been widely reported elsewhere (Leridon, 1976Go; Roman et al., 1978Go; Risch et al., 1988Go; Kline et al., 1989Go; Regan et al., 1989Go; Coste et al., 1991Go; Parazzini et al., 1997Go). In the current analysis a similar association is seen, particularly in third pregnancies with a suggestion that it is the most recent event (i.e. the outcome of the second pregnancy) which has the greatest impact. However, it is difficult to attach much weight to this without similar evidence from higher order pregnancies. In these limited data it may be more informative to consider the number of previous fetal deaths. The increase in OR associated with one and two previous losses suggests that the effect may be cumulative and the magnitude of this change is consistent with a multiplicative effect.

These data came from a retrospective cohort study and the possibility of recall bias should be considered. It is possible, for example, that the absence of an association between fetal death and smoking status might be explained by poor recall or under-reporting of past smoking habits by women who had fetal deaths. A particular problem in this type of study is the self-reporting of pregnancy outcomes. If women with previous fetal losses were more likely to recognize a subsequent early loss, the risk of fetal death following a previous loss would be exaggerated. Similarly, women may have been more likely to recall more recent fetal losses which might, in part, explain the apparent trend of increasing fetal death with increasing year of pregnancy outcome.

There are also potential problems with the fertility measure used here, as women with a successful pregnancy history may be less likely to perceive any retrospective difficulty in conception, leading to an overestimate of the harmful effect of reduced fertility. In addition, assigning the same fertility type to all of a woman's pregnancies is not ideal as fecundity is known to decrease with age (Schwartz and Mayaux, 1982Go), making it potentially unrealistic to assume that the same difficulty applies equally to all pregnancies. Moreover, it is not possible to establish in these data whether the observed association suggests that subfertility leads to an increased risk of fetal death or vice versa.

Bias may also be introduced by omitting women with pregnancies not ending in a live birth or fetal death. However including these women, and in particular their induced abortions, made no perceptible difference to the results. The additional finding of a possible increase in risk of fetal death following an induced abortion is interesting, particularly in light of speculation that there may be damage caused to the cervix (Molin, 1993Go), but it is not possible to draw any conclusions from these restricted data.

There is still much to learn about factors that affect pregnancy outcome and the way in which they interact. In addition to recognised risk factors for fetal death, it is important to assess the impact of past reproductive experience. These data support the notion that having a previous fetal death may increase the risk of a subsequent loss. Analysis of the number of previous losses suggests that this increase in risk may be cumulative and possibly multiplicative in effect. There is also evidence of an association between subfertility and risk of fetal death, regardless of pregnancy order and previous pregnancy outcomes, and this is particularly marked in the pregnancies of older women.


    Notes
 
4 To whom correspondence should be addressed Back


    References
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Coste, J., Job-Spira, N. and Fernandez, H. (1991) Risk factors for spontaneous abortion: a case-control study in France. Hum. Reprod., 6, 1332–1337.[Abstract]

Coulam, C. (1992) Association between infertility and spontaneous abortion. Am. J. Reprod. Immunol., 27, 128–129.[ISI][Medline]

Everett, C. (1997) Incidence and outcome of bleeding before the 20th week of pregnancy: prospective study from general practice. Br. Med. J., 315, 32–34.[Abstract/Free Full Text]

Hakim, R., Gray, R. and Zacur, H. (1995) Infertility and early pregnancy loss. Am. J. Obstet. Gynecol., 172, 1510–1517.[ISI][Medline]

Huber, P. (1967) The behaviour of maximum likelihood estimates under non-standard conditions. Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, 1, 221–233.

Joffe, M. and Li, Z. (1994) Association of time to pregnancy and the outcome of pregnancy. Fertil. Steril., 62, 71–75.[ISI][Medline]

Kline, J., Stein, Z. and Susser, M. (1989) Conception to Birth: Epidemiology of Prenatal Development. Oxford University Press, New York.

Leridon, H. (1976) Facts and artefacts in the study of intra-uterine mortality: a reconsideration from pregnancy histories. Popul. Stud., 30, 319–335.[ISI]

Molin, A. (1993) Risk of damage to the cervix by dilation for first-trimester induced abortion by suction aspiration. Gynecol. Obstet. Invest., 35, 152–154.[ISI][Medline]

Olsen, J. (1994) Options in making use of pregnancy history in planning and analysing studies of reproductive failure. J. Epidemiol. Community Health, 48, 171–174.[Abstract]

Parazzini, F., Chatenoud, L., Tozzi, L. et al. (1997) Determinants of risk of spontaneous abortions in the first trimester of pregnancy. Epidemiology, 8, 681–683.[ISI][Medline]

Regan, L., Braude, P. and Trembath, P. (1989) Influence of past reproductive performance on risk of spontaneous abortion. Br. Med. J., 299, 541–545.[ISI][Medline]

Risch, H., Weiss, N., Clarke, A. et al. (1988) Risk factors for spontaneous abortion and its recurrence. Am. J. Epidemiol., 128, 420–430.[Abstract]

Roman, E., Doyle, P., Beral, V. et al. (1978) Fetal loss, gravidity, and pregnancy order. Early Hum. Dev., 2/2, 131–138.[ISI][Medline]

Roman, E., Doyle, P., Ansell, P. et al. (1996) Health of children born to radiographers. Occup. Environ. Med., 53, 73–79.[Abstract]

Schwartz, D. and Mayaux, M. (1982) Female fecundity as a function of age. N. Engl. J. Med., 306, 404–406.[ISI][Medline]

Templeton, A., Fraser, C. and Thompson, B. (1991) Infertility – epidemiology and referral practice. Hum. Reprod., 6, 1391–1394.[Abstract]

Weinberg, C., Baird, D. and Wilcox, A. (1994) Bias in retrospective studies of spontaneous abortion based on the outcome of the most recent pregnancy. Ann. N.Y. Acad. Sci., 709, 280–286.[Abstract]

Submitted on January 19, 1999; accepted on August 5, 1999.