The contribution of assisted conception, chorionicity and other risk factors to very low birthweight in a twin cohort
*Correspondence: Dr R. McDuffie Jr, Department of Obstetrics and Gynecology, 20th Street Medical Campus, 2045 Franklin Street, Denver, Colorado 80205, USA.
Objectives To investigate the contribution of assisted conception (assisted reproductive technology and ovulation induction), chorionicity and selected maternal risk factors for very low birthweight.
Design Retrospective twin cohort study.
Setting Staff model Colorado Health Maintenance Organization.
Sample Five hundred and sixty-two twin gestations [assisted = 193 (34%); unassisted = 369 (66%)].
Methods Data were collected from a perinatal database and medical record review. Data were analysed using univariate and multivariable logistic regression analysis.
Main outcome measure Very low birthweight.
Results Women with assisted twin gestation were more likely to be older, nulliparous, non-smokers, married, have a prior history of a miscarriage and a dichorionic placentation. There was no difference in the distribution of low and very low birthweight, discordant growth or preterm delivery between assisted and unassisted twin gestations. Significant risk factors for very low birthweight were: a prior preterm birth (odds ratio, OR, 3.8, 95% confidence interval, CI, 2, 7), monochorionicity (OR 3, 95% CI 2, 4.7), nulliparity (OR 2, 95% CI 1.3, 3), cigarette smoking (OR 1.8, 95% CI 1, 3) and prior miscarriage (OR 1.6, 95% CI 1, 2). Monochorionicity was significantly associated with adverse perinatal outcomes.
Conclusion Assisted conception did not play a significant role in the occurrence of very low birthweight in this cohort. A history of preterm birth and a monochorionic twin gestation were the leading risk factors for very low birthweight. Associated risk factors for very low birthweight were nulliparity, cigarette smoking and a prior miscarriage.
The recent epidemic of multiple births in the United States is one of the main contributory factors to the upward trend in the number of low birthweight (<2500 g) or very low birthweight (<1500 g) babies. Multiple births are associated with a disproportionate amount of perinatal morbidity and mortality1. One of the reasons for the increased frequency of multiple births is the more widespread availability of infertility treatments, which include assisted reproductive technologies and ovulation induction2. We have estimated, that the proportion of multiple births, in our Colorado Health Maintenance Organization due to one of these forms of assisted conception is 33%3.
The increase in multiple births has prompted investigation into the possible association between assisted conception and adverse perinatal outcomes in twin cohorts4–14. These studies have produced conflicting results, perhaps due to differences in study population characteristics, definition of assisted conception and study methods. Twin placentation15,16 and other events in the maternal social and prenatal history are known risk factors for adverse neonatal outcomes17,18. However, in many existing studies adjustment for these risk factors has not been included in the data analysis. The main purpose of this study was to examine the association between assisted conception (assisted reproductive technology and ovulation induction) and very low birthweight in twins, adjusting for chorionicity of the placenta and other maternal risk factors. Very low birthweight was chosen as the main outcome of this study because of its established importance as a predictor of not only adverse neonatal events19 but also of developmental problems later in life20.
The study was approved by the Kaiser Foundation Institutional Review Board. The cohort of twins was drawn from a study population of all women (n= 33,949) who delivered after 20 weeks of gestation, between January 1994 and December 2001 at Exempla Saint Joseph Hospital and Boulder Community Hospital in Colorado. During the study period of seven years, 564 sets of twins were delivered. Two mothers delivered two sets of twins. In order to assure independence in the data analysis, the second set of births from each of these mothers was removed from the data set.
We used a retrospective study design to answer the study's research questions. The primary exposure was assisted conception with either assisted reproductive technology (defined as procedures that involved handling of human oocytes or embryos) or ovulation induction (treatment with clomiphene citrate or human menopausal gonadotrophins) in the absence of assisted reproductive technology. The main outcome was very low birthweight (<1500 g). Data as to whether a pregnancy was assisted or spontaneous and on risk factors (Tables 1–3) were obtained from Colorado Kaiser Permanente Perinatal Database (described elsewhere3,21). Medical records were also reviewed to add new variables to the database and to establish that there was no misclassification of the category of conception (i.e. assisted vs unassisted). In addition, 20% of the medical records were reviewed by another member of the research team for quality assurance purposes. If discrepancies were found, they were discussed and resolved. Placentation was classified as monochorionic (monochorionic diamniotic or monochorionic monoamniotic) or dichorionic. We assumed that all twins who were of the opposite sex (n= 72) were dichorionic diamniotic. We reviewed postnatal placental pathology on the remaining 490 sets of twins to determine their chorionicity. Pathology on the placenta was missing for one set of twins.
Table 1. Comparison of maternal risk factors and outcomes among assisted (n= 193) and unassisted twin pregnancies (n= 369). Values are given as n (%) and RR [95% CI].
|Age ≥35 years||75 (39)||43 (12)||3.3 [2.4, 4.6]|
|African American||2 (1)||48 (13)||0.08 [0.02, 0.3]|
|Nulliparity||136 (70)||121 (33)||2.1 [1.8, 2.6]|
|Marital status (single)||6 (3)||88 (24)||0.1 [0.06, 0.3]|
|Cigarette smoking||10 (5)||62 (17)||0.3 [0.2, 0.6]|
|Maternal prenatal history|
|Prior preterm delivery||9 (5)||26 (7)||0.7 [0.3, 1.4]|
|Miscarriage||65 (34)||79 (21)||1.6 [1.2, 2.1]|
|Characteristics of this pregnancy|
|Monochorionic placentation*||8 (4)||112 (30)||0.1 [0.07, 0.3]|
|Selective fetal reduction||18 (9)||–|| |
|Low birthweight||113 (59)||218 (59)||1.0 [0.9, 1.1]|
|Very low birthweight||20 (10)||47 (13)||0.8 [0.5, 1.3]|
|Birth weight <5th centile for gestational age||11 (6)||18 (5)||1.2 [0.6, 2.4]|
|Discordant growth||17 (9)||33 (9)||1.0 [0.6, 1.7]|
| <37 weeks||104 (54)||206 (56)||1.0 [0.8, 1.1]|
| <32 weeks||15 (8)||43 (12)||0.7 [0.4, 1.2]|
| <28 weeks||8 (4.2)||19 (5.2)||0.8 [0.4, 1.8]|
Table 2. The relative risk of assisted conception and other selected risk factors for very low birthweight. Values are given as n (%) and RR [95% CI].
|Maternal age ≥35 years||30 (13)||95 (11)||1.2 [0.8, 1.7]|
|African American||16 (16)||109 (11)||1.5 [0.9, 2.4]|
|Marital status (single)||29 (15)||96 (10)||1.5 [1.0, 2.2]|
|Cigarette smoking||26 (18)||99 (10)||1.8 [1.2, 2.7]|
|Nulliparity||65 (13)||60 (10)||1.3 [0.9, 1.8]|
|Previous miscarriage||39 (14)||86 (10)||1.3 [0.9, 1.9]|
|Previous preterm delivery||17 (24)||108 (10)||2.4 [1.5, 3.7]|
|Any assisted conception||36 (9)||89 (12)||0.8 [0.5, 1.1]|
|Selective fetal reduction||2 (6)||123 (11)||0.5 [0.1, 1.9]|
|Monochorionicity*||48 (20)||77 (9)||2.3 [1.6, 3.2]|
Table 3. Multivariable logistic regression analysis showing the crude and adjusted odds ratios of assisted conception, placentation and other risk factors for very low birthweight. Values are given as OR [95% CI].
|Maternal age >35 years||1.2||1.3||[0.8, 2.2]|
|Marital status (single)||1.6||1.3||[0.8, 2.2]|
|Cigarette smoking||2.0||1.8||[1, 3]|
|African American||1.6||1.7||[0.9, 3]|
|Prior miscarriage||1.4||1.6||[1, 2]|
|Prior preterm birth||2.8||3.8||[2, 7]|
|Assisted conception||0.8||1.0||[0.6, 1.6]|
Information regarding birthweight and gestational age at delivery were obtained from the perinatal database. Gestational age was determined for all subjects relying primarily on first and second trimester ultrasound but also menstrual dates. From the birthweights, we distinguished low birthweight (<2500 g) from very low birthweight (<1500 g). Intrauterine twin growth curves described by Min et al.22. were used to classify the birthweight as less than or greater than the fifth centile for gestational age. We noted if there was evidence of discordant growth, here defined as greater than a 25% difference in weight between a twin pair. Three categories of preterm birth (less than 37, 32 and 28 weeks) were examined. Labour onset was either spontaneous, medically induced (for maternal or fetal complications) or none (an elective caesarean delivery).
The data were analysed in several stages using SAS 8.2 (SAS Institute, Cary, North Carolina). First, we determined the general clinical characteristics of the cohort. Second, we investigated if there was any statistically significant difference in the frequency of selected maternal characteristics, monochorionicity, selective fetal reduction and neonatal outcomes between mothers with assisted and unassisted conception. Third, using very low birthweight as the main outcome variable, we initially studied if as compared with spontaneously conceived twins there was an association between assisted reproductive technology, human menopausal gonadotrophin and clomiphene citrate with very low birthweight. We then determined the overall contribution of the grouped variable, assisted conception and other risk factors for very low birthweight.
The relative risk (RR) was used as a measure of association to test the relationship between the primary exposure and risk factors for very low birthweight. The relative risk was defined as the cumulative incidence of very low birthweight among mothers with the exposure divided by the cumulative incidence of very low birthweight among mothers without the exposure. The incidence of very low birthweight in the subjects exposed to a selected risk factor was compared with the incidence of very low birthweight among those not exposed. Measures of association between dichotomous variables were tested using the χ2 or Fisher's Exact Test. Differences in means of continuous variables were tested using the Student t test. Statistics are presented with 95% confidence intervals (CI) (P < 0.05).
In the final stage of the analysis, we conducted a multivariable logistic regression. This was used to estimate the odds ratios (OR) of assisted conception for very low birthweight adjusted for other covariates. The odds ratio was used as an approximation of the relative risk. Each explanatory variable was entered separately into the logistic model to determine its independent (unadjusted) effect on very low birthweight. A full model was then constructed with the combined variable (assisted conception), risk factors identified from the univariate analysis as associated with the outcome and potential confounders.
During the study period, 562 sets of twins were delivered after 20 weeks of gestation. There were 369 (66%) unassisted twin births. One hundred and ninety-three (34%) of the twin gestations received some form of assisted conception. Sixty-nine (12%) resulted from assisted reproductive technology, 94 (17%) from clomiphene citrate and 30 (5%) from human menopausal gonadotrophin. Four hundred and thirty-two women in the cohort were white (77%), 62 (11%) were Hispanic, 50 (9%) were African American and 18 (3.3%) were from other races. Stratified into assisted and unassisted groups, Table 1 demonstrates the distribution of risk factors and outcomes between the two groups. Compared with unassisted twin pregnancies, we found a higher frequency of older, nulliparous mothers among the assisted group. In contrast, fewer women with assisted conceptions were African American, single or smoked cigarettes. With regard to the maternal prenatal history, we found the incidence of prior preterm birth did not differ between the groups. However, a significantly higher incidence of prior miscarriage was found among mothers with assisted conceptions. No unassisted pregnancy was selectively reduced from a higher order gestation. In contrast, 18 (9%) of the assisted pregnancies had been selectively reduced. The incidence of monochorionicity was significantly lower among the assisted twin gestations. Using women with spontaneous onset of labour as the referent group, we found a significantly higher number of women with assisted pregnancies had a medically induced labour [64 (38%) vs 96 (29%), RR 1.3, 95% CI 1, 1.7]. We found no significant difference in the frequency of elective caesarean section delivery between the assisted and the unassisted group [23 (18%) vs 34 (12%), RR 1.4, 95% CI 0.9, 2.3]. The mean [SD] gestational age at delivery of twins who experienced spontaneous labour, medically induced labour or an elective caesarean section was 35  (range 20.2–40.6), 37  (range 27–40.4) and 37  (range 27.1–39.2) weeks, respectively. There was no significant difference (P= 0.6) in the mean birthweights between the assisted, 2326  (range 315–3775 g), and unassisted, 2313  (range 100–3930 g), births. Similarly, there was no significant (P= 0.4) difference in the mean gestational age at delivery between the assisted, 36  (range 20.2–40.2 weeks), and unassisted, 36  (range 20.2–40.6 weeks), groups. Table 1 shows no significant difference in the incidence of pregnancy outcomes between the assisted and unassisted twin pregnancies. Among the 345 women who had a spontaneous preterm (<37 weeks) birth, 70 (66%) had assisted conceptions and 156 (65%) had unassisted conceptions (RR 1, 95% CI 0.9, 1.2).
In comparison with spontaneously conceived twins, we found no association between assisted reproductive technology (RR 0.8, 95% CI 0.5, 1.4), human menopausal gonadotrophin (RR 0.7, 95% CI 0.3, 1.7) and clomiphene citrate (RR 0.8, 95% CI 0.5, 1.3) with very low birthweight. Because the trend and the association of these exposures with very low birthweight were similar, we grouped the methods of assisted conception together into one variable (i.e. assisted conception) for further analysis. Table 2 shows the relative risk of assisted conception and other maternal risk factors for very low birthweight. Assisted conception was not associated with very low birthweight. Further, no association was found between assisted conception and very low birthweight appropriate for the 5th centile for gestational age (RR 0.9, 95% CI 0.5,1.1) and very low birthweight less than the 5th centile for gestational age (RR 0.9, 95% CI 0.4, 1.9). A monochorionic placentation and a past history of a preterm delivery were the most significant risk factors for the delivery of a very low birthweight baby. Other significant risk factors associated with very low birthweight were maternal cigarette smoking and a single marital status. The relative risk of nulliparity, African American race, history of miscarriage and maternal age greater than 35 years for very low birthweight was increased. However, in the univariate analysis, these associations did not reach statistical significance. Eighteen (3%) of the twin pregnancies resulted from a selective fetal reduction from a higher order gestation. Thirteen selective fetal reductions occurred among conceptions following assisted reproductive technology. Selective fetal reduction was not associated with very low birthweight (Table 2).
Table 3 is the fully adjusted multivariable logistic regression model. Adjusted for the covariates in the model, assisted conception was not associated with very low birthweight. In contrast there was a significant association of monochorionicity, nulliparity, cigarette smoking, prior preterm birth and prior miscarriage for very low birthweight.
In this cohort of 562 twin births, we found no association between assisted reproductive technology, human menopausal gonadotrophin or clomiphene citrate and very low birthweight. A prior history of preterm birth, monochorionic placentation, nulliparity, cigarette smoking and a prior miscarriage were significant risk factors for very low birthweight (Table 3). There was no difference in the distribution of other neonatal outcomes (Table 1) between assisted and unassisted twin pregnancies.
In agreement with the results of this study, Fitzsimmons et al.8 found no difference in mean birthweights and gestational age at delivery among multiple births resulting from assisted reproductive therapy (assisted reproductive technology or clomiphene citrate) and spontaneously conceived multiple birth controls. This group of Canadian researchers did, however, report an increase in perinatal mortality among spontaneously conceived twins8. Agustsson et al.4 studied 69 assisted twin pregnancies and 453 natural conceptions from Iceland and Scotland. A similar gestational age and birthweight at delivery was found among the assisted and unassisted group. Another study from a twin cohort found no difference in duration of gestation, birthweight or other outcomes among 1241 twins from assistance (in vitro fertilisation and intracytoplasmic sperm injection) compared with natural twin gestations7. Using logistic regression models that controlled for maternal characteristics and placental membranes, a multicentred American, collaborative study of 2523 spontaneous and 415 assisted twin pregnancies suggests as association between assisted conception and low birthweight but not very low birthweight11. Another recent study from the United States found the risk of term and preterm low birthweight among twins conceived with assisted reproductive technology was similar to that of the general population of twins23.
Other authors have however, found an association between assisted conception and adverse pregnancy outcomes. Moise et al.13 included very low birthweight as an outcome in their study of the association between in vitro fertilisation and perinatal outcomes among dizygotic twins using univariate and multivariable analysis. Twenty in vitro fertilisation pregnancies were matched on maternal age, parity and ethnicity to 40 pairs of spontaneous twins. In this small study, there was a significantly higher frequency of prematurity, lower average birthweight and a higher very low birthweight rate among in vitro fertilisation compared with unassisted twins13. Other studies from Israel6 and Italy14 described similar results. Data from a multicentred Dutch study9 reported a lower mean birthweight, a higher low birthweight rate and more discordance in birthweights among 96 in vitro fertilisation twins compared with 96 matched controls. In agreement with this study, Lambalk and van Hooff10, also found less favourable obstetric outcomes among assisted (in vitro fertilisation or ovulation induction) compared with spontaneous twins. Bernasko et al.5 reported an increased frequency of discordant growth and low birthweight among assisted reproductive technology compared with spontaneously conceived twins.
The incidence of dichorionic twinning is related to factors that increase double ovulation such as older mothers, increasing parity, genetic factors and ovulation induction24. One of the major strengths of this study was the careful differentiation of placental chorionicity. Monochorionicity was found less frequently among assisted compared with unassisted twin gestations (Table 2). As shown by this and other studies16, monochorionicity was an important independent risk factor for adverse perinatal outcomes. As seen in Tables 2 and 3, compared with dichorionic twins, monochorionic twins were more than twice as likely to be born very low birthweight. As we continue to gather data and expand this cohort, we hope to further evaluate the role of monochorionicity as a confounder or effect modifying variable. Other authors have also recognised the important effect of this covariate by either restricting their twin studies to dizygotic twins10 or adjusting for this covariate using logistic regression analysis11. Maternal age, nulliparity, marital status, race, smoking status and a prior miscarriage were identified as significant confounding variables. Table 3 demonstrates the importance of inclusion of these covariates into the adjusted analysis. For example, nulliparity, which was a borderline significant risk factor for very low birthweight in the univariate analysis (Table 2), became a significant risk factor in the logistic regression analysis (Table 3).
Many authors have found higher elective caesarean delivery rates among iatrogenic compared with unassisted gestations. It has been suggested that babies from assisted conception may be delivered early because of overconcern by the obstetrician and parents for the outcome of these anxiously awaited births25. In this staff model Colorado Health Maintenance Organization, all mothers with multiple gestations receive a systematic level of organised prenatal care from a team of obstetricians and other maternal–fetal medicine specialists. This may explain the similar rates of elective caesarean section among the assisted and unassisted groups. The frequency of medically induced labours was higher among assisted compared with unassisted twin gestations. However, women with medically induced labours delivered at a later gestational age than women who had a spontaneous onset of labour.
We were unable to fully explore the clinical implication of nulliparity and other risk factors because of several limitations of the study. Nulliparity was significantly associated with very low birthweight and was found more frequently among mothers who had assisted pregnancies (Tables 1 and 2). Similarly, we found a higher frequency of a history of prior miscarriage among women with assisted conception. It is possible that some of the risk associated with these variables may be related to other factors in the woman's infertility history. These factors were not examined in this study. Analysis of selective fetal reduction was also limited by low numbers in the outcome group. Likewise, the limited representation of African American women among the assisted group of twin conceptions did not allow full evaluation of the contribution of this risk factor to very low birthweight in this cohort. We were also unable to investigate if the effect of assisted conception on very low birthweight was modified by chorionicity because of low numbers of very low birthweight babies among assisted twin gestations with a monochorionic placentation. Preliminary analysis of the possible association between assisted conception and birthweight less than the 5th centile for gestational age was limited by small numbers and also by the utilisation of intrauterine growth curves22 that were not adjusted for Denver's altitude (5280 ft above sea level).
Although this study did not find an association between assisted conception and very low birthweight, research efforts continue to follow this cohort to identify adverse sequelae related to not only the method of conception but also their plurality.
The authors thank Kimberly Bischoff MSHA for her careful management of the perinatal database and Michelle Perri for her help in obtaining pathology records. Grant support was from the Agency for Health Care Research and Quality (RO3 HS10700-01).