Summary

Background Few studies have described the perinatal risks associated with infertility, other than for infertility treated by in-vitro fertilisation or gamete intrafallopian transfer. The aim of this analysis was to estimate the risks of perinatal death associated with treated and untreated infertility.

Methods A population-based case-control study of perinatal deaths was carried out in Leicestershire Health District over the period 1990–94, during which 60922 babies were delivered. Of these, 567 perinatal deaths were associated with 542 women. 972 mothers were randomly selected as controls. Medical, obstetric, and social data were collected for cases and controls from the medical notes and interviews with the women. The relative risks of perinatal death associated with treated and untreated infertility before the index pregnancy were estimated as odds ratios by means of unconditional logistic regression analysis.

Findings 65 (10%) of cases and 34 (3·5%) of the controls had infertility before the index pregnancy. History of infertility in the index pregnancy, irrespective of treatment, increased the risk of perinatal death (odds ratio 2·9 [95% CI 1·8–4·5]). The population attributable risk fraction for perinatal death related to infertility was 6·2% (3·4–9·0). 45 (54%) of the deaths, even in the untreated group, were associated with immaturity. Compared with women without infertility, women with untreated infertility were at increased risk of perinatal death (3·3 [1·6–6·8]). The risk of perinatal death associated with multiple births did not explain this finding. Similarly, treated infertility also increased the risk of perinatal death (2·7 [1·5–4·7]); the risks associated with multiple births explained some, but not all, of this excess. In Leicestershire, the overall underlying risk of a mother experiencing at least one perinatal death over the study was 9·0 per 1000 women. For women who experience infertility, this risk increases by about 18 per 1000 (6–30).

Interpretation Counselling for women before any form of infertility treatment should include discussion of the risks of perinatal death. Our results would benefit from confirmation. However, we advocate that at antenatal booking a history of infertility, irrespective of treatment, should be sought, because these women have a significantly increased risk of perinatal death, particularly associated with prematurity.

Introduction

Few population-based data have described the perinatal risks associated with infertility treated by in-vitro fertilisation (IVF) or gamete intrafallopian transfer (GIFT).¹ Women who become pregnant after IVF or GIFT are at increased risk of multiple pregnancy, preterm labour, low infant birthweight, and perinatal death.^2–4 Venn and Lumley⁵ showed that women with a history of infertility had a significantly increased risk of perinatal mortality. However, although they were able to distinguish between infertility treatment by IVF or GIFT they were unable to separate untreated infertility and infertility treated by other methods.

The aim of this analysis was to use data from a continuing case-control study of perinatal mortality to estimate the risk of perinatal mortality for women with infertility, and to distinguish between those who did and those who did not receive treatment.

Patients and methods

Since 1976 we have been undertaking a case-control study of perinatal death occurring to women living in Leicestershire.⁶ The analysis presented here relates to babies delivered in 1990–94. A perinatal death is defined as a stillbirth or death occurring within the first week of life and is classified by the modified Wigglesworth classification.^7,8 Infertility is defined as any mention in the medical records of delayed conception, or investigation or treatment for delayed conception, related to the index pregnancy. Medical, obstetric, and social data were collected for cases and controls from the medical notes and interviews with the women. The average time from delivery to interview was similar for cases and controls owing to the variation in discharge times between patients, with a proportion of both cases and controls being interviewed after discharge. The study was approved by the Leicestershire Health Authority Ethics Committee. Verbal informed consent was sought and obtained from all participating women.

Selection of mothers or infants as controls in perinatal epidemiology is complex. Use of controls from the same institution as the cases, although convenient, introduces temporal and institutional biases that tend to lead to an underestimation of risk.⁹ However, sampling frames from which controls might be selected are in many cases not established until after mothers have been discharged from hospital; hence, if interviews are used, additional recall bias may result. We therefore devised a control-sampling method to eliminate these potential biases. The method uses as its sampling frame the data collected on birth notification forms for the previous year's births in Leicestershire as a proxy for the births that will occur during the following year. Before each new year of data collection, a proxy random sample of 200 births is selected from the computerised database of births to women in the Leicestershire Health District during the previous year. Information relating to the time, date, and place of birth for each proxy birth is recorded. To allow for the calendar shift of weekdays from year to year and the effects of a leap year, the date of birth is adjusted by 1 day (2 in leap years) to ensure that the day of the week will be the same each year to avoid oversampling at weekends and holiday periods. The time, adjusted date, and place of birth of the proxy births are then used to define which births in the following year will constitute the controls. Each birth finally identified as a control is the baby born in the defined hospital, in the current year, whose date and time of birth is closest to that of the randomly selected proxy birth. Interviewing is thus facilitated because the researchers (although not the care providers) know the hospital, day, and time when they will need to undertake an interview, although the particular mother is not known until near to the time of interview.

The relative risks of perinatal death, associated with treated or untreated infertility during the index pregnancy were estimated as odds ratios by use of unconditional logistic regression analysis. Variables defined as potential confounders were those that had previously been shown to be predictive of perinatal mortality in this population:⁸ maternal age; race; height; late or no antenatal care; partner in social class III manual, IV, or V, or unemployed; maternal work status; pregnancy legitimacy; diabetes or gestational diabetes; smoking during pregnancy; having a general practitioner who was not on the obstetric list; urinary-tract infection; pre-eclampsia; and more than 7 days of antenatal inpatient care. The final regression model included only factors found to have a significant reduction in terms of the model log-likelihood by a forward selection procedure. No term for multiple pregnancy was included in the primary adjusted analysis because multiple pregnancies were predominantly associated with infertility treatment. Multiple pregnancy is likely to be on the causal pathway to perinatal death rather than being a confounder—for example, untreated or treated infertility may lead to an increased risk of perinatal death, and treated infertility may lead to an increased risk of multiple pregnancy, itself leading to an increased risk of perinatal death. Perinatal death was treated as a binary outcome (death, no death). The unit of analysis was the mother because all factors used in the risk adjustment were mother related. Mothers with a perinatal death in a multiple pregnancy may have had one or more deaths in the index pregnancy.

Results

Over the period 1990–94, there were 60922 births in Leicestershire Health District. This total included 567 perinatal deaths, associated with 542 women, which represented a perinatal mortality rate of 9·3 per 1000 births. A random sample of 972 women who gave birth to 983 infants was used as the control sample. Overall, 10% of the mothers of perinatal cases and 3·5% of the controls had experienced infertility before the index pregnancy. 34% of the case mothers and 35% of the controls had primary infertility. Table 1 shows the subdivision of these groups into two mutually exclusive categories: those with a history of infertility who did not receive infertility treatment to achieve the index pregnancy (untreated infertility), and those who received treatment for infertility to achieve the index pregnancy (treated infertility).

Click to see table 1

Nearly all the controls were singleton births. Most of the perinatal deaths of babies in multiple pregnancies were in twins. Perinatal death in infants of women with untreated infertility occurred predominantly in singletons, whereas most of the deaths in the treated group were in multiple pregnancies. 54% of the deaths in the infertility groups were due to immaturity, and 26% were macerated stillbirths. By contrast, in the overall perinatal death group, 30% of deaths were due to immaturity and 45% macerated stillbirths. The high proportion of deaths from immaturity in the infertility groups was also reflected in the mean duration of gestation at delivery (table 1). IVF or GIFT were the methods used for ten of the 43 treated cases and three of the 22 treated controls. The remaining treated cases and controls were conceived after drug-induced ovulation alone.

Table 2 shows the unadjusted odds ratios for risk factors found to be significantly associated with perinatal death. Maternal age, although not significant, was included because it is known to be associated with infertility.

Click to see table 2

We found that, after adjusting for the effects of confounders, women with infertility, whether or not they had received treatment, were 2·9 (95% CI 1·8–4·5) times more likely to have a perinatal death than women without infertility (table 3). The population attributable risk fraction for perinatal death related to infertility was 6·2% (3·4–9·0). Compared with women without infertility, those with untreated infertility were 3·3 (1·6–6·8) times more likely to have a perinatal death, and those with treated infertility were over 2·7 (1·5–4·7) times more likely. When, in a secondary analysis, a term to model the effects of multiple pregnancy (with mother as a unit of analysis) was included, the odds ratio associated with untreated infertility remained unchanged, whereas the odds ratio for treated infertility was 1·5 (0·8–2·9).

Click to see table 3

The group receiving infertility treatment to achieve the index pregnancy was subdivided by method of treatment. Although the numbers were small, infertile women treated by drugs alone had a 2·3 (1·2–4·4) excess risk of perinatal death. Women who had received IVF or GIFT had a 4·9 (1·2–19·1) excess risk. Secondary adjustment for the effects of multiple pregnancy (with mother as the unit of analysis) resulted in non-significant odds ratios of 1·6 (0·8–3·3) and 0·5 (0·1–3·5), respectively.

In Leicestershire, for women who experience infertility, our study suggests a risk of perinatal death of 27 per 1000 births (95% CI 15–39) and a risk of perinatal death after successful infertility treatment of 23 per 1000 births (10–37). This compares with a risk of perinatal deaths of 9·0 per 1000 births in the overall population (ie, normal population risk), indicating that for women who experience infertility this risk increases by about 18 per 1000 births (6–30).

Discussion

Our results confirm previous findings that women with infertility treated by IVF or GIFT are at increased risk of having a perinatal death.⁵ In addition, women receiving infertility treatments other than IVF or GIFT are also at increased risk of perinatal death. Furthermore, women with infertility who conceive spontaneously are at increased risk. Infertility was not the focus of our primary hypothesis, because only limited information about the nature of the infertility was available and the study had limited power to detect a small risk. Although the power of the study to detect an odds ratio of 2·0 at the 5% significance level was 94% (assuming a 3·5% prevalence of any infertility found in the controls), the power to detect a more modest risk of 1·5 was only 55%. Consequently, these findings must be confirmed in other studies. The results should be interpreted cautiously.

Among women with a history of infertility, untreated in the index pregnancy, deaths occurred predominantly in singletons (only one twin in this subpopulation), of which more than half the deaths were associated with immaturity. This finding supports the idea that these women had poor overall reproductive function. These women had difficulty in conceiving and, having conceived, they had difficulty in carrying the pregnancy to term and were at increased risk of an adverse outcome. Previous work from this study identified a group of women with a history of low fertility and a reduced ability to produce a well liveborn infant, which may have been due to an occupational exposure to fetotoxic agents.¹⁰ Williams and colleagues¹¹ have also found that untreated subfertility is a predictor for low birthweight.

Similarly, among women with treated infertility more than half the perinatal deaths were associated with immaturity. Although the numbers were small, there was an excess risk of death for women treated by druginduced ovulation only and for those treated by IVF or GIFT. Since in general only infertile couples receive infertility treatment, we cannot distinguish whether it is the disease or its treatment that leads to the increase in risk of perinatal death. However, our findings relating to untreated infertility indicate that the disease rather than the treatment seems to be a more plausible explanation.

Multiple pregnancy is a common consequence of infertility treatment.^12–14 Multiple pregnancies are at increased risk of perinatal death.^12,15 Compared with singletons, multiple births had a ten-fold increased risk of perinatal death in our study. In the relation between infertility and perinatal death, multiple pregnancy is on the causal pathway rather than being a confounder; therefore, we did not adjust for it in the primary anlysis. However, in a secondary analysis, we adjusted for the effects of multiple pregnancy (with mother as the unit of analysis) to estimate the excess risk of perinatal death associated with infertility. Not surprisingly, since untreated infertility was not strongly associated with multiple births, this adjustment had no effect for this group. After adjustment in the treated group, the relative risk was reduced, which suggested that some of the excess risk of perinatal death associated with infertility treatment is mediated through the risk of death associated with the excess occurrence of multiple pregnancies. We should emphasise, however, that our study had insufficient power to detect such a change with a high degree of certainty. Furthermore, the excess risk of death in the IVF and GIFT group seemed to be largely explained by the effect of multiple pregnancy, although given the small numbers and wide CIs there is much uncertainty.

The knowledge that infertility, treated or untreated, is a risk factor for perinatal death is important if there are any avoidable or treatable factors. Preterm delivery and associated mortality related to immaturity were features of both infertility groups. This knowledge may influence decisions about birth location because tertiary neonatal care is more likely to be required. It may also lead to decisions about the use of prophylactic antenatal steroids.

Counselling before any infertility treatment, including drug treatment to induce ovulation, should include a discussion of the potential risks of adverse outcome, including perinatal death. At antenatal booking, a history of infertility should be sought, irrespective of whether treatment was received, because babies born to these women have a significantly increased risk of perinatal death, particularly associated with prematurity. Decisions about the place of delivery for such women may be influenced by this information.



Contributors

E S Draper coordinated the Leicestershire perinatal mortality casecontrol study, carried out the data analysis, and wrote the early drafts of the paper. J J Kurinczuk, M Clarke, and E S Draper designed the control selection method. J J Kurinczuk assisted with the analysis and interpretation of the data. K R Abrams provided statistical advice and carried out the more complex analyses.

Acknowledgments

We thank the mothers who participated in this study; Lesley Parr and all the midwives who helped with the interviewing, and all the other staff in hospitals that provide maternity services to Leicestershire mothers; David Clayton for his original contribution to the control sampling idea and useful discussions on interpretation; and John Keogh for his clinical perspective. The Leicestershire perinatal mortality survey has been supported by a grant from Leicestershire Health Authority since 1976. J J Kurinczuk is funded by a programme grant (963209) from the Australian National Health and Medical Research Council.



References

^1. Human Fertilisation and Embryology Authority fifth annual report— July 1996. London: HFEA, 1996.

^2. Tan SL, Doyle P, Campbell S, et al. Obstetric outcome of in vitro fertilization pregnancies compared with normally conceived pregnancies. Am J Obstet Gynaecol 1992; 167: 778-784.

^3. Ezra Y, Schenker JG. Appraisal of in vitro fetilization. Eur J Obstet Gynecol Reprod Biol 1993; 48: 127-133.

^4. McFaul PB, Patel N, Mills J. An audit of the obstetric outcome of 148 consecutive pregnancies from assisted conception: implications for neonatal services. Br J Obstet Gynaecol 1993; 100: 820-825.

^5. Venn A, Lumley J. Births after a period of infertility in Victorian women 1982–1990. Aust NZ J Obstet Gynaecol 1993; 33: 379-384.

^6. Clarke M, Clayton DG, Mason ES, MacVicar J. Asian mothers' risk factors for perinatal death—the same or different? A 10 year review of Leicestershire perinatal deaths. BMJ 1988; 297: 384-387.

^7. Wigglesworth JS. Monitoring perinatal mortality. A pathophysiological approach. Lancet 1980; ii: 684-686.

^8. Clarke M, Mason ES, MacVicar J, Clayton DG. Evaluating perinatal mortality rates: effects of referral and case mix. BMJ 1993; 306: 824-827.

^9. Clarke M, Clayton DG. The design and interpretation of casecontrol studies of perinatal mortality. Am J Epidemiol 1981; 113: 636-645.

^10. Clarke M, Mason ES. Shoe manufacture and possible hazards to reproduction. BMJ 1988; 296: 466.

^11. Williams MA, Goldman MB, Mittendorf R, Monson RR. Subfertility and the risk of low birth weight. Fertil Steril 1991; 56: 668-671.

^12. Botting BJ, Macdonald DaviesI, Macfarlane AJ. Recent trends in the incidence of multiple births and associated mortality. Arch Dis Child 1987; 62: 941-950.

^13. Kurinczuk JJ, Pemberton RJ, Binns SC, Parsons DE, Stanley FJ. Singleton and twin confinements associated with infertility treatments. Aust NZ J Obstet Gynaecol 1995; 35: 27-31.

^14. Levene MI, Wild J, Steer P. Higher multiple births and the modern management of infertility in Britain. Br J Obstet Gynaecol 1992; 99: 607-613.

^15. Petterson B, Nelson KB, Watson L, Stanley F. Twins, triplets and cerebral palsy in births in Western Australia in the 1980s. BMJ 1993; 307: 1239-1243.