Preeclampsia is a major maternal pregnancy complication. Recent data indicate that almost 20% of pregnancy‐related deaths are from complications of preeclampsia or eclampsia.1 The risk of developing a hypertensive disorder in pregnancy is at least two times2–4 more likely among multiple births compared with singleton births. With the more widespread availability of assisted conception, there has been a dramatic increase in the number of multiple births.5 This rising trend has heightened public health concern regarding multiple birth‐related maternal morbidities, including preeclampsia. Furthermore, several reports have shown an increased prevalence of the hypertensive disorders of pregnancy among iatrogenic compared with spontaneously conceived births.6–11 However, one recent study has not supported these views.12
Accepting that there are other multiple gestation‐related maternal morbidities,13 we concentrated on preeclampsia as a primary outcome for this analysis. This retrospective study was initiated in a cohort of 528 multiple births to investigate the relationship of assisted reproductive technology and ovulation‐inducing drugs with preeclampsia. Using multivariable logistic regression, we determined if other established risk factors (maternal age, race, parity, fetal number, and preexisting maternal hypertension) and potentially new risk factors (ovulation induction cycles) had a confounding effect on the relationship between assisted reproductive interventions and preeclampsia.
MATERIALS AND METHODS
The study was approved by the Kaiser Foundation Institutional Review Board. The multiple birth cohort was drawn from a study population (n = 28,905) of all women who delivered between January 1994 and November 2000 at Colorado's Kaiser Permanente facilities. Data regarding the maternal medical and obstetric history and on intrapartum events were gathered by the attending physician on admission to the hospital and after delivery. This information was obtained from both patient interview and chart review. In this database, we identified 528 mothers who delivered multiple gestations during the study period. Two mothers delivered two sets of multiple births. To assure independence in the data analysis, the second multiple birth from each of these mothers was removed from the data set.
The perinatal database provided information on a number of risk factors. The categorization of these risk factors appears in parenthesis: race (black versus others), parity (nulliparity versus multiparity), and birth order (triplets and quadruplets versus twins). Maternal age was examined as a continuous and dichotomous (over 35 years versus under 35 years) variable. Descriptive data were collected on marital status. The perinatal database also provided information on maternal preexisting hypertension (yes/no). This was defined as hypertension (140/90 mmHg or greater) before conception or before 20 weeks' gestation.14,15 The primary exposure was treatment with assisted reproductive technologies (defined as procedures that involved handling of human oocytes or embryos).16 Secondary exposures were treatment with clomiphene citrate or human menopausal gonadotropins (HMG) in the absence of in vitro fertilization (IVF). Mothers with the primary and secondary exposures were compared with mothers without these exposures. Data on whether a pregnancy was assisted or spontaneous were obtained from this database if the birth occurred after October 1996. For multiple fetuses born before this date, this information was abstracted by chart review and then appended to the database. Chart review was conducted on all multiple births to ascertain if the mother had a history of previous ovulation induction (yes/no) with clomiphene citrate or HMG. To confirm inter‐rater reliability of the data, 10% of the charts were reviewed by a second member of the research team. There were very few discrepancies between the two reviewers. When any occurred, they were discussed and resolved.
We used a retrospective study design to answer the study's research questions regarding preeclampsia. The perinatal database was employed to identify mothers who had preeclampsia, categorized as mild or severe preeclampsia by the attending physician. We established that preeclampsia was primarily diagnosed using the definition described by the National High Blood Pressure Educational Program,14 that is, hypertension (increase of 30 mmHg or more systolic, or 15 mmHg diastolic from levels measured before 20 weeks' gestation or a reading of 140/90 mmHg or greater after 20 weeks if prior blood pressure [BP] was unknown) accompanied by proteinuria (qualitative measurement of 1+ (30 mg/dL) or 300 mg or more in a 24‐hour collection) and edema (clinically evident swelling or a rapid increase in weight without evident swelling). The diagnosis of pregnancy‐induced hypertension was confirmed when the BP changes were found on at least two occasions 6 or more hours apart. Severe preeclampsia was defined as BP 160 Hg or more systolic or 110 mmHg or more diastolic, proteinuria of 5 g or more in 24 hours (3+ or 4+ on qualitative examination), oliguria (less than 500 mL per 24 hours), elevated serum creatinine, thrombocytopenia, marked elevation of liver enzymes, cerebral or visual disturbances, epigastric pain, pulmonary edema or cyanosis, and intrauterine growth restriction or olighydramnios.15
We verified that there were no subjects with preeclampsia among the remainder of the multiple birth cohort by reviewing the medical records of all mothers. Cases of preeclampsia were then evaluated to confirm the definitions described above. The following information was recorded: the maternal BP readings (at initial prenatal visit and during the disorder), urinary protein by dipstick and 24‐hour collection, maternal edema and weight gain, maternal symptoms and signs of preeclampsia (epigastric pain, headache, pulmonary edema, cerebral, or visual disturbances), oliguria (less than 500 mL of urine in 24 hours), and fetal growth restriction. We also noted whether preeclampsia was an indication for early delivery. Information was also collected where possible on laboratory abnormalities associated with preeclampsia.15,17
Eighty‐six cases of preeclampsia fulfilled the definition described above. Eight mothers had, however, an atypical presentation. Six of these had pregnancy‐induced hypertension without significant proteinuria, but accompanied by a combination of abnormal laboratory tests and or preeclamptic symptoms. The remaining two mothers with an atypical presentation were normotensive. One woman had mild disease with proteinuria, moderately severe edema, elevated liver function tests, and an elevated serum urate. The other had severe disease characterized by a rapid weight gain (30 lb in 2 weeks), low levels of proteinuria, hyponatremia, very high liver function tests, and severe symptoms.
The data in this newly built data set were analyzed in two steps using SAS 8.2 (SAS Institute, Cary, NC). In step 1, the relative risk (RR) was used as a measure of association and was defined as the cumulative incidence of preeclampsia in women exposed to a risk factor divided by the cumulative incidence of preeclampsia in women not exposed to a risk factor. The study period was 6 years and 11 months. Measures of association were tested using the χ2 or Fisher exact test. Statistics are presented with 95% confidence intervals (CI) (P < .05).
Step 2 of the analysis involved multivariable logistic regression. This was used to estimate the adjusted odds ratios (OR) of association of the primary and secondary explanatory variables for preeclampsia and to test for any significant confounding. The OR was used as an approximation of the RR. The individual crude (unadjusted) ORs of all the risk factors for preeclampsia were calculated. A full model was then constructed with the primary, secondary explanatory variables, and other risk factors. Backward selection was performed on this model to remove variables that were not statistically significant. Confounding was then assessed by reintroducing each covariate back into the model along with the primary explanatory variable. A change in the parameter estimate of the primary explanatory variable of greater than 10% was considered evidence of confounding. Finally, a model was constructed containing the primary explanatory variable, covariates that remained in the reduced model along with significant confounders.
The clinical characteristics of the cohort's 528 multiple births, 502 (95%) twins, 25 (4.7% triplets and one set of quadruplets) are shown in Table 1. Between January 1994 and November 2000, there were 330 spontaneously (unassisted) conceived and 198 assisted multiple gestations. The mean maternal age was higher among mothers who became pregnant as a result of assisted reproductive technologies compared with the mean maternal age of mothers who had ovulation induction or unassisted pregnancies. Almost 75% of the mothers who received assisted reproductive technologies were over 35 years. There was a higher frequency of white, married, and nulliparous mothers among the assisted pregnancies, compared with the unassisted pregnancies. Lower numbers of higher‐order gestations (triplets or quadruplets) or of previous treatment with clomiphene citrate or HMG were seen among the unassisted group compared with the assisted group.
Ninety‐four (17.8%) of mothers developed preeclampsia. Among the women with preeclampsia, seventy (74%) had mild and 24 (26%) had severe preeclampsia. Five women developed preeclampsia superimposed on chronic hypertension (four had mild and one had severe preeclampsia). The mean gestational age ± standard deviation at onset of preeclampsia was 34 ± 3 weeks. The mean gestational age (± standard deviation) of women with preeclampsia at delivery was 35.6 ± 2 weeks. Preeclampsia was an indication for an induced delivery in 85% of the subjects.
The first measure of association was the RR. The RR of developing preeclampsia among mothers who received assisted reproductive technology, clomiphene citrate, or HMG compared with spontaneously conceived multiples is shown in Table 2. Although the RR was increased among mothers who received any assisted conception, it was only statistically significant among mothers who had received assisted reproductive technologies (RR = 3.2, CI 2.2, 4.9). Compared with spontaneously conceived multiples, the RR of developing mild or severe preeclampsia among mothers receiving assisted reproductive technologies was 2.7 (CI 1.7, 4.7) and 4.8 (CI 1.9, 11.6), respectively.
We then developed a full multivariable logistic regression model containing all forms of assistance along with other selected risk factors (Table 3, model A). Table 3 also shows the crude (unadjusted) OR of each risk factor for preeclampsia. In the full model, assisted reproductive technologies and nulliparity were the only risk factors significantly associated with preeclampsia. This full model also suggests that at older maternal ages, there was a slight increase in the risk of preeclampsia. After backward selection, the only variables that remained in the model were assisted reproductive technologies, maternal age, and nulliparity (Table 3, model B). In this latter model, after adjustment for maternal age and nulliparity, the odds of preeclampsia among women who received assisted reproductive technologies were 2.1 (CI 1.1, 4.1). Nulliparity and maternal age were also identified as significant confounders of the relationship between assisted reproductive technologies and preeclampsia. This effect can be seen in the change between the crude (3.8) and adjusted (2.1) OR of assisted reproductive technologies for preeclampsia. We present model B, Table 3, as the best current clinical model of the relationship between assisted reproductive technologies and preeclampsia.
In this cohort of 528 multiple gestations, mothers who received assisted reproductive technologies were two times more likely to develop preeclampsia. Nulliparity was also a significant independent risk factor for preeclampsia, as was advancing maternal age. In addition, there was an almost five‐fold increase in the risk of severe preeclampsia among women who received assisted reproductive technologies compared with spontaneously conceived multiple births.
Our findings support those of other authors.6–11 Tallo et al7 found that IVF (n = 101) mothers had more pregnancy‐induced hypertension (21%) compared with an equal number of non‐IVF controls (4%) who were matched to patients on maternal age, fetal number, and other risk factors. When adjusted for parity, patients remained at a greater risk for pregnancy‐induced hypertension than controls. Tanbo et al8 studied 140 assisted (IVF) singleton pregnancies and compared them with 643 controls matched for age and parity. A statistically higher frequency of pregnancy‐induced hypertension was found among the singleton‐assisted pregnancies compared with matched controls. Furthermore, Tan et al9 reported a higher risk of hypertension requiring hospitalization among IVF singleton pregnancies compared with singleton non‐IVF controls. Using multivariable logistic regression analysis, Maman et al reported similar results.10
In contrast, a Canadian study12 examined a number of birth outcomes, including pregnancy‐induced hypertension. Seventy‐two assisted multiple gestations (eight from ovulation induction, 64 from assisted reproductive technologies) were compared with 124 spontaneously conceived multiple gestation controls. Patients were matched to controls on maternal age, fetal number, and the presence of maternal medical problems to control for confounding. Contrary to our results, there was no significant difference in the incidence of pregnancy‐induced hypertension between the study and control groups. One reason for the difference in results between this Canadian study and our health maintenance organization cohort may be differences in the ages of the women selected for study. As previously reported, older women in our health maintenance organization are more likely to seek care involving assisted reproductive technologies.2 In the present study, the mean age of mothers receiving assisted reproductive technology was 37 years compared with 33 years in the Canadian assisted group.
Maternal age and parity were important variables in this multivariable logistic regression analysis. In agreement with the established literature,17 nulliparity and more advanced maternal age were independent risk factors for preeclampsia (Table 3, model B). In addition, both of these variables significantly confounded the relationship between assisted reproductive technologies and preeclampsia. In the final model, the OR for assisted reproductive technologies was reduced from 3.8 (unadjusted) to 2.1 when adjusted for maternal age and nulliparity. This effect underscores the importance of adjusting for these variables in studies of reproduction.
Based on historical17 and recent reports,1 we also included race (black versus others) as a risk factor for preeclampsia in this analysis. Black women (less than 10%) were poorly represented among mothers who received assisted conception. Race, therefore, became a less meaningful variable in our study. However, race deserves the careful attention of investigators who are studying a population with a broader demographic distribution. Similarly, although the crude OR suggested increased risk of preeclampsia among mothers with preexisting hypertension and with higher‐order gestations, these risk factors were present in a small number of women, making it difficult to see any effect in the fully adjusted logistic regression model.
The risk of developing preeclampsia was increased in mothers who became pregnant as a result of treatment with clomiphene citrate or HMG. This increase was not statistically significant. We also considered the maternal history of ovulation induction with clomiphene citrate or HMG as possible risk factors or confounders. Although the unadjusted risk of previous cycles of clomiphene citrate suggested some elevated risk (Table 3) for preeclampsia, this covariate became unimportant in the fully adjusted model (Table 3, model A) and was removed in the backward selection process (Table 3, model B).
There are inconsistencies in the definition of preeclampsia both in practice and as reported in the literature.14,15,17 In approaching this study, we established that in our practice, preeclampsia is primarily diagnosed on the appearance of de novo hypertension with proteinuria. However, we were careful to verify how each case was defined by chart review. We found 86 (91%) patients were diagnosed in accordance with this definition. There was also strict adherence to the criteria for severe preeclampsia. However, eight women with an atypical presentation were included among the 94 preeclampsia patients. This finding confirms that our physicians recognize the clinical variability of preeclampsia. It also underscores the difficulty that exists in creating strict research definitions of this outcome variable. There were several other limitations of this study. First, analysis of previous cycles of HMG was limited by small numbers, especially among mothers who received previous treatment with HMG and developed preeclampsia. However, we regard this as a necessary first probe of the significance of previous ovulation induction cycles with maternal pregnancy outcomes, which should be reexamined in a larger multiple birth cohort. Second, as discussed above, small numbers also restricted the analysis of several other covariates. Third, our denominator was limited to women who delivered at our hospitals. Currently, the perinatal database does not contain information on pregnancies that ended earlier or outside our facilities.
Our analyses suggest that assisted reproductive technology is an independent risk factor for preeclampsia. We recognize the complex nature of the decision to seek assisted conception with IVF or related procedures.18 It is possible that assisted reproductive technologies are a surrogate measure for other possible infertility factors (eg, cause of infertility, infertility treatments, ovulation induction, superovulation, donor eggs, donor sperm, etc), which may act alone or in combination to initiate the vascular events associated with preeclampsia. Some researchers have recently suggested that donor eggs may have some role in the development of preeclampsia.19 Given the advanced age of the assisted reproductive technology mothers (Table 1), it is possible that subclinical or clinical cardiovascular disease might provide another trigger necessary for the development of preeclampsia.
This study shows that women with multiple gestations who become pregnant as a result of assisted reproductive technologies are twice as likely to develop preeclampsia as women with spontaneously conceived multiple gestations. The public health implications for women are far reaching, as preeclampsia is associated with serious maternal morbidity and mortality. Preeclampsia and eclampsia now rank third, after embolism and hemorrhage, as the leading cause of pregnancy‐related deaths.1 Pregnancy‐induced hypertension also accounts for 15% of hospital admissions in the antenatal period.20 At this time, the overall public health impact of assisted conception in population‐based cohorts is unquestionably poorly understood. Studies are urgently needed in this area if women and their partners are to be fully informed of any risks associated with assisted conception and multiple birth. Intensive study of the longer‐term maternal21 and infant outcomes18 of infertility treatments is also required. It is our belief that with more longitudinal studies in this area, investigators will provide evidence‐based recommendations to those seeking assisted reproductive interventions and practice guidelines to those providing the care.
1. MacKay A, Berg C, Atrash H. Pregnancy-related mortality from preeclampsia and eclampsia. Obstet Gynecol 2001;97:533–8.
2. Lynch A, McDuffie R, Murphy J, Faber K, Leff M, Orleans M. Assisted reproductive interventions and multiple birth. Obstet Gynecol 2001;97:195–200.
3. Spellacy WN, Handler A, Ferre C. A case control study of 1253 twin pregnancies from a 1982–1987 perinatal data base. Obstet Gynecol 1990;75:168–71.
4. Sibai BM, Hauth J, Caritis S, Lindheimer MD, MacPherson C, Klebanoff M, et al. Hypertensive disorders in twin versus singleton gestations. Am J Obstet Gynecol 2000: 182;938–42.
5. Keith L, Oleszezuk JJ. Iatrogenic multiple birth, multiple pregnancy and assisted reproductive technologies. Int J Gynaecol Obstet 1999;64:11–25.
6. Frydman R, Belaisch-Allart J, Fries N, Hazout A, Glissant A, Testart J. An obstetric assessment of the first 100 births from in vitro fertilization program at Clamart, France. Am J Obstet Gynecol 1986;154:550–5.
7. Tallo CP, Vohr B, Oh W, Rubin LP, Seifer DB, Haning RV. Maternal and neonatal morbidity associated with in vitro fertilization. J Pediatr 1995;127:794–800.
8. Tanbo T, Dale PO, Lunde O, Moe N, Abyholm T. Obstetric outcome in singleton pregnancies after assisted reproduction. Obstet Gynecol 1995;86:188–92.
9. Tan S, Doyle P, Campbell S, Bernal V, Rizk B, Brinsden P, et al. Obstetric outcome of in vitro fertilization pregnancies compared with normally conceived pregnancies. Am J Obstet Gynecol 1992;167:778–84.
10. Maman E, Lunenfeld E, Levy A, Vardi H, Potashnik G. Obstetric outcome of singleton pregnancies conceived by in vitro fertilization and ovulation induction compared with those conceived spontaneously. Fertil Steril 1998;70:240–5.
11. Daniel Y, Ochshorn Y, Fait G, Geva E, Bar-Am A, Lessing JB. Analysis of 104 twin pregnancies conceived with assisted reproductive technologies and 193 spontaneously conceived twin pregnancies. Fertil Steril 2000;74;683–9.
12. Fitzsimmons B, Bebbington M, Fluker M. Perinatal and neonatal outcomes in multiple gestations: Assisted reproduction versus spontaneous conception. Am J Obstet Gynecol 1998;179:1162–7.
13. Senat M, Ancel P, Bouvier-Colle M, Breart G. How does multiple pregnancy affect maternal mortality and morbidity? Clin Obstet Gynecol 1998;41:79–83.
14. National High Blood Pressure Education Working Group. Report on high blood pressure in pregnancy. Am J Obstet Gynecol 1990;163:1689–712.
15. Lindheimer MD, Roberts JM, Cunningham FC, Chesley L. Introduction, history, controversies, and definitions. In: Lindheimer MD, Roberts JM, Cunningham FG, eds. Chesley's hypertensive disorders in pregnancy. 2nd ed. Stamford, CT: Appleton and Lange, 1998:3–36.
16. American Fertility Society. Revised minimum standards for in vitro fertilization, gamete intrafallopian transfer, and related procedures. Fertil Steril 1990;53:255–6.
17. Roberts JM. Pregnancy-related hypertension. In: Creasy RK, Resnik R, eds. Maternal fetal medicine. 4th ed. Philadelphia, PA: WB Saunders, 1999:833–72.
18. Buitendijk SE. Children after in vitro fertilization. An overview of the literature. Int J Technol Assess Health Care 1999;15:52–65.
19. Salha O, Sharma V, Dada T, Nugent D, Rutherford AJ, Tomlinson AJ, et al. The influence of donated gametes on the incidence of hypertensive disorders of pregnancy. Hum Reprod 1999;14:2268–73.
20. Scott CL, Chavez GF, Atrash HK, Taylor DJ, Shah RS, Rowley D. Hospitalizations for severe complications of pregnancy, 1987–1992. Obstet Gynecol 1997;90:225–9.
21. Venn A, Lumley J. Clomiphene citrate and pregnancy outcome. Aust N Z Obstet Gynaecol 1994;34:56–66.