Twin pregnancy is a well-known and threatening complication of IVF cycles (Kulkarni et al., Reference Kulkarni, Jamieson, Jones, Kissin, Gallo, Macaluso and Adashi2013). During the past decade, intensive efforts have been made to reduce the rate of twin pregnancies worldwide by boosting policies of elective single embryo transfer. Even though additional efforts are needed, several countries have now shrunk the rate of twin pregnancies to less than 10% (Bhattacharya & Kamath, Reference Bhattacharya and Kamath2014).
In line with this current attempt of the scientific and medical communities to minimize the complications of IVF in general and the risk of twins in particular, there is now a growing interest in the association between IVF and the occurrence of monozygotic twins (Delrieu et al., Reference Delrieu, Himaya, Phillips and Kadoch2012; Franasiak et al., Reference Franasiak, Dondik, Molinaro, Hong, Forman, Werner and Scott2015; Gee et al., Reference Gee, Dickey, Xiong, Clark and Pridjian2014; Kanter et al., Reference Kanter, Boulet, Kawwass, Jamieson and Kissin2015; Knopman et al., Reference Knopman, Krey, Lee, Fino, Novetsky and Noyes2010; Reference Knopman, Krey, Oh, Lee, McCaffrey and Noyes2014; Luke et al., Reference Luke, Brown, Wantman and Stern2014; Nakasuji et al., Reference Nakasuji, Saito, Araki, Nakaza, Nakashima, Kuwahara and Sakumoto2014; Osianlis et al., Reference Osianlis, Rombauts, Gabbe, Motteram and Vollenhoven2014; Ren et al., Reference Ren, Liu, Chen, Zhu and Zhang2013; Tocino et al., Reference Tocino, Blasco, Prados, Vargas, Requena, Pellicer and Fernández-Sánchez2015; Vitthala et al., Reference Vitthala, Gelbaya, Brison, Fitzgerald and Nardo2009). This aspect deserves the utmost consideration because the clinical management of monozygotic twin pregnancies may be demanding (Corsello & Piro, Reference Corsello and Piro2010; Sperling et al., Reference Sperling, Kiil, Larsen, Qvist, Schwartz, Jørgensen and Tabor2006). Of note, 75% of these pregnancies are MCT (Shulman and van Vugt, Reference Shulman and van Vugt2006), a condition associated with major obstetrical complications.
Even if monozygotic twin pregnancies are generally deemed to be increased in IVF pregnancies (Delrieu et al., Reference Delrieu, Himaya, Phillips and Kadoch2012; Hall, Reference Hall2003; Vitthala et al., Reference Vitthala, Gelbaya, Brison, Fitzgerald and Nardo2009), the available evidence is still not fully consistent and estimates of the magnitude of this risk vary widely in the literature. While the rate of monozygotic twins in natural conceptions has been reported to be approximately 0.4% (Gee et al., Reference Gee, Dickey, Xiong, Clark and Pridjian2014; Hall, Reference Hall2003; Imaizumi & Nonaka, Reference Imaizumi and Nonaka1997), the rates reported in IVF pregnancies vary between 0.7 and 13% (Delrieu et al., Reference Delrieu, Himaya, Phillips and Kadoch2012; Gee et al., Reference Gee, Dickey, Xiong, Clark and Pridjian2014; Knopman et al., Reference Knopman, Krey, Lee, Fino, Novetsky and Noyes2010; Reference Knopman, Krey, Oh, Lee, McCaffrey and Noyes2014; Nakasuji et al., Reference Nakasuji, Saito, Araki, Nakaza, Nakashima, Kuwahara and Sakumoto2014; Osianlis et al., Reference Osianlis, Rombauts, Gabbe, Motteram and Vollenhoven2014; Ren et al., Reference Ren, Liu, Chen, Zhu and Zhang2013; Sobek et al., Reference Sobek, Zbořilová, Procházka, Šilhánová, Koutná, Klásková and Sobek2015; Tocino et al., Reference Tocino, Blasco, Prados, Vargas, Requena, Pellicer and Fernández-Sánchez2015; Vitthala et al., Reference Vitthala, Gelbaya, Brison, Fitzgerald and Nardo2009). These discrepancies may be explained by differences in characteristics of the studied populations and in methodological aspects. In fact, study designs used to investigate this issue mainly provide results that are exposed to significant confounders or diagnostic inaccuracies (Blickstein, Reference Blickstein2005; Osianlis et al., Reference Osianlis, Rombauts, Gabbe, Motteram and Vollenhoven2014). Few studies report DNA testing to confirm the diagnosis and all lack a control group. Of utmost relevance here is that the available studies on the risk of monozygotic twins in IVF pregnancies do not attempt to provide data on the local background rate of monozygosity in natural pregnancies and refer to previous out-of-date evidence (Gee et al., Reference Gee, Dickey, Xiong, Clark and Pridjian2014; Hall, Reference Hall2003; Imaizumi & Nonaka, Reference Imaizumi and Nonaka1997; Tong et al., Reference Tong, Caddy and Short1997). Even if the rate of monozygotic twins is believed to be constant all over the world, regardless of race and age, this assumption is actually speculative and may expose the results to significant inaccuracies.
In this study, we suggest approaching the issue from a different perspective; that is, using data from obstetric antenatal care units. In other words, we selected women with a diagnosis of MCT from a prenatal diagnosis unit and assessed whether the prevalence of pregnancies that were achieved with IVF differed from the local proportion of IVF pregnancies. This approach is expected to overcome the diagnostic confounders that typically affect case studies of children born from IVF since all included women were ascertained in the same manner by physicians who are experts in prenatal diagnosis. The exclusive inclusion of MCT pregnancies also protects our findings from the diagnostic inaccuracies that can occur when focusing on monozygotic pregnancies in general (i.e., the need for DNA testing for a definitive diagnosis). Moreover, this study design overcomes the above-mentioned limit of the unknown background rate of monozygosity. Finally, it also allows the recruitment of women early in pregnancy, prior to the occurrence of possible pregnancy complications (intrauterine demise) that may affect the birth of two viable children. This confounder may be significant if pregnancy outcome differs between natural and IVF MCT pregnancies.
Materials and Methods
This is a retrospective cross-sectional study reporting on women attending a specific service of the prenatal diagnosis unit of the Fondazione Ca’ Granda, Ospedale Maggiore Policlinico of Milan, Italy, which is exclusively dedicated to the monitoring and management of MCT pregnancies. Women were identified through the use of an electronic database. We included twin pregnancies with a sonographic diagnosis of monochorionicity that progressed beyond 16 weeks’ gestation. Higher order pregnancies (>2 fetal poles) were excluded. Pregnancies were diagnosed as monochorionic on the basis of the presence of a unique placenta and the absence of the twin peak sign (lambda sign) at the first sonographic assessment performed in our institution (between 8 and 16 weeks’ gestation; Sepulveda et al., Reference Sepulveda, Sebire, Hughes, Odibo and Nicolaides1996). All ultrasound assessments were performed by expert gynecologists with extensive and long-standing experience in obstetric sonography. Women who had been referred after 16 weeks’ gestation were excluded.
Data were retrospectively obtained from outpatient charts. They were completed using inpatient charts from the same institution. Women could be contacted by phone if relevant data were missing or if inconsistencies emerged. The data collected included baseline clinical characteristics, mode of conception, and pregnancy outcome. The study was accepted by the local Institutional Review Board.
A five-year period from 2007 to 2011 was used to achieve the scheduled sample size of at least 130 women. This sample size was calculated on the basis of an expected rate of IVF pregnancies during the study period of 2.2% (regional data extrapolated from 277 043 births; Parazzini et al., Reference Parazzini, Cipriani, Bulfoni, Bulfoni, Frigerio, Somigliana and Mosca2015), setting type 1 and 2 errors at 0.05 and 0.20, respectively and stating as clinically relevant a three-fold increase in the risk of MCT in IVF pregnancies. A binomial distribution model was used to calculate the 95% confidence interval (95% CI) of the proportions. The primary aim of the study was to determine the rate of IVF pregnancies among MCT pregnancies. The secondary aim was to compare the pregnancy outcome of IVF and natural pregnancies. Data were analyzed using the software SPSS 18.0 (Chicago, IL., USA). Data were compared using the Student's t test or Fisher's exact test, as appropriate; p values below .05 were considered statistically significant.
Results
One hundred and forty-five MCT pregnancies were selected. Ten of these were achieved with IVF, corresponding to a rate of 6.9% (95% CI: 3.5–11.8%), significantly higher than the natural background rate of 2.2% (Parazzini et al., Reference Parazzini, Cipriani, Bulfoni, Bulfoni, Frigerio, Somigliana and Mosca2015). The odds ratio (OR) of MCT in IVF pregnancies was 3.3 (95% CI: 1.6–5.9). When considering exclusively MCT pregnancies achieving delivery of two viable newborns (n = 132), the number of IVF pregnancies was nine (6.8%, 95% CI: 3.7–12.5%) and the corresponding OR was 3.3 (95% CI: 1.7–6.4).
A comparison of the baseline pre-pregnancy characteristics of women who had IVF pregnancies and those who conceived naturally is illustrated in Table 1. Women achieving pregnancy with IVF were older and the time to pregnancy was longer.
Baseline Characteristics of the Studied Women According to the Mode of Conception
Data is reported as number (%), mean ± SD or median (interquartile range).
Pregnancy and neonatal outcomes in women who become pregnant by IVF and in those who conceived naturally are shown in Table 2. The rates of delivery of two viable twins (90% and 91%), twin-to-twin transfusion syndrome (10% and 18%), delivery before 34 weeks’ gestation (22% and 25%), and small for gestational age (11% and 23%) were similar (p = 1.00, p = 1.00, p = 1.00 and p = .20, respectively). A statistically significant difference emerged for the proportion of newborns with a neonatal birth weight ≥2,500 g (0% vs. 20%, p = .03).
Pregnancy Outcome According to the Mode of Conception
PIH = pregnancy-induced hypertension. NICU = Neonatal Intensive Care Unit.
aBoth twins died at 21 weeks’ gestation in one IVF-ICSI pregnancy. In the remaining cases (all from natural conception), only one of the two twins died.
bRefers to 9 IVF-ICSI pregnancies and 135 natural pregnancies (the woman with the intrauterine demise of both twins is excluded).
cGestational diabetes mellitus (n = 8), placenta previa (n = 2), cholestasis (n = 2), cervical cerclage (n = 2), post-partum hemorrhage (n = 1), and varicella infection (n = 1).
dData refers to the 276 viable newborns (18 IVF-ICSI cases and 258 natural conceptions).
eBased on local data (Parazzini et al., Reference Parazzini, Cortinovis, Bortolus and Fedele1991).
fIntestinal obstruction (n = 1).
gRespiratory distress syndrome (n = 6), cardiovascular malformations (n = 2), intestinal obstruction (n = 2), clubfoot (n = 1), cystic lymphangioma (n = 1), sacral teratoma (n = 1), ovarian cyst (n = 1), aplasia cutis (n = 1), and severe hydronephrosis (n = 1).
The main cycle characteristics of the 10 women achieving pregnancy by IVF-ICSI are shown in Table 3. All women were transferred at the cleavage stage. No transfer at the blastocyst stage was recorded. Assisted hatching was never performed. All women transferred two embryos.
Treatment Cycle Characteristics of the 10 Women Conceiving With IVF
Discussion
Our results, based on the perspective of a prenatal diagnosis unit, indicate that the risk of MCT is three to four times higher in IVF pregnancies. To our knowledge, our study design has not been previously employed. Four studies reported the rate of MCT conceived by IVF but none related this finding to the local proportion of IVF pregnancies, thus impeding inferences with respect to the association (Chow et al., Reference Chow, Benson, Racowsky, Doubilet and Ginsburg2001; Ghalili et al., Reference Ghalili, McLennan, Pedersen, Kesby and Hyett2013; Ortibus et al., Reference Ortibus, Lopriore, Deprest, Vandenbussche, Walther, Diemert and Lewi2009; Sperling et al., Reference Sperling, Kiil, Larsen, Qvist, Schwartz, Jørgensen and Tabor2006). Results from these contributions are summarized in Table 4. Surprisingly, the proportions of IVF pregnancies in these studies were higher than the rate observed in our study and varied widely, from 8.1% (Ortibus et al., Reference Ortibus, Lopriore, Deprest, Vandenbussche, Walther, Diemert and Lewi2009) to 32.0% (Chow et al., Reference Chow, Benson, Racowsky, Doubilet and Ginsburg2001). The reasons for these discrepancies are obscure. Differences in the local rate of IVF pregnancies may play a role, but it is unlikely that they fully explain these findings. Three possible additional reasons may be proposed as follows:
Case Series on the Proportion of Women Achieving Pregnancy With IVF Among MZTs
Studies were identified by searching in PubMed for articles published in the English language between January 2000 and February 2015 and using the following MeSH search terms: ‘mzt’, ‘monozygotic’, ‘monochorionic’ combined with ‘ivf’, ‘in vitro fertilization’, ‘icsi’, ‘assisted reproductive’, and ‘ART’ with restriction to the human species. Data was extracted independently by two investigators (V.S. and E.S.) who performed an initial screening of the title and abstract of all articles to exclude citations deemed irrelevant. Manual search of cross references completed the search.
First, diagnostic accuracy may differ among studies. In our setting, the diagnosis of zygosity was made by highly expert physicians who are exclusively dedicated to prenatal diagnosis and we considered MCT pregnancies exclusively. However, it has to be acknowledged that we lack confirmation by DNA analyses and we cannot rule out some misdiagnoses. Noteworthy, the zygosity test was performed in only one out of the four previous studies (Sperling et al., Reference Sperling, Kiil, Larsen, Qvist, Schwartz, Jørgensen and Tabor2006), leading to the exclusion of four cases initially classified as MCT (4 out of 78, corresponding to 5%). On the other hand, it has to be pointed out that the sonographic diagnosis of MCT is nowadays considered highly reliable (Carroll et al., Reference Carroll, Soothill, Abdel-Fattah, Porter, Montague and Kyle2002; Sepulveda et al., Reference Sepulveda, Sebire, Hughes, Odibo and Nicolaides1996; Sperling et al., Reference Sperling, Kiil, Larsen, Qvist, Schwartz, Jørgensen and Tabor2006), such that the skill level of the physicians involved is unlikely to differ markedly among the five contributions (all contributions were published by researchers engaged in prenatal diagnosis) and that there is no rational reason to speculate that diagnostic inaccuracy may selectively affect IVF pregnancies. Therefore, it is implausible that this limitation may explain the extreme variability in the reported rates of IVF pregnancies among MCT pregnancies.
Second, one may speculate on a role for some confounders in the referral fluxes to the involved prenatal diagnosis units. In other words, in some settings, access to the prenatal diagnostic service may have been facilitated for women who became pregnant by IVF. We cannot speculate on the potential role of this confounder in the four previous studies because we do not know the local situation. Even if we cannot fully exclude a similar bias in our study, it has to be pointed out that, in our setting, the IVF and prenatal diagnosis units are two distinct and independent services that are run by different physicians and that refer to separate departments. Moreover, the prenatal diagnosis unit of our academic institution is a regional referral center for MCT pregnancies that converges cases from the whole region and this may further protect our data from confounders related to referral fluxes.
Third, the characteristics of the studied populations or the IVF protocols may differ among studies. For instance, the risk of MCT in IVF pregnancies has been related to a younger maternal age (Knopman et al., Reference Knopman, Krey, Lee, Fino, Novetsky and Noyes2010; Reference Knopman, Krey, Oh, Lee, McCaffrey and Noyes2014). In this regard, it is interesting to note that a recent study also suggested that the higher incidence of monozygotic twins after assisted reproduction techniques may be related to genetic information rather than to the IVF technology itself (Sobek et al., Reference Sobek, Zbořilová, Procházka, Šilhánová, Koutná, Klásková and Sobek2015). Moreover, there is some evidence, albeit debated, suggesting a role for ovarian hyperstimulation, prolonged culture up to the blastocysts stage, culture medium, number of embryos transferred, and breach of the zona pellucida with ICSI or assisted hatching (Chang et al., Reference Chang, Lee, Jee, Suh and Kim2009; Delrieu et al., Reference Delrieu, Himaya, Phillips and Kadoch2012; Franasiak et al., Reference Franasiak, Dondik, Molinaro, Hong, Forman, Werner and Scott2015; Kanter et al., Reference Kanter, Boulet, Kawwass, Jamieson and Kissin2015; Knopman et al., Reference Knopman, Krey, Lee, Fino, Novetsky and Noyes2010; Reference Knopman, Krey, Oh, Lee, McCaffrey and Noyes2014; Luke et al., Reference Luke, Brown, Wantman and Stern2014; Nakasuji et al., Reference Nakasuji, Saito, Araki, Nakaza, Nakashima, Kuwahara and Sakumoto2014; Ren et al., Reference Ren, Liu, Chen, Zhu and Zhang2013; Vitthala et al., Reference Vitthala, Gelbaya, Brison, Fitzgerald and Nardo2009). In other words, differences may be consequent to local differences in clinical and laboratory protocols (Vitthala et al., Reference Vitthala, Gelbaya, Brison, Fitzgerald and Nardo2009). Noteworthy, in the 10 IVF pregnancies included in our study, none of the embryos were transferred at the blastocyst stage and none underwent assisted hatching. Of the utmost interest here is that some authors reported a decrease in the rate of MCT in IVF pregnancies over time, and this has been related to improved expertise (Knopman et al., Reference Knopman, Krey, Oh, Lee, McCaffrey and Noyes2014; Moayeri et al., Reference Moayeri, Behr, Lathi, Westphal and Milki2007).
As a corollary of our study, we investigated pregnancy outcome according to the mode of conception. We observed a higher rate of newborns with a weight at birth below 2,500 g among IVF pregnancies. Conversely, the frequency of other pregnancy complications did not differ between the two groups. Our study is nonetheless underpowered for reliable conclusions. In this regard, it has to be noted that our data are mainly in agreement with the findings of Ghalili et al. (Reference Ghalili, McLennan, Pedersen, Kesby and Hyett2013), who also evaluated pregnancy outcome according to the mode of conception and who failed to document major differences. Prematurity, intrauterine growth restriction, and neonatal survival did not indeed differ in their study (Ghalili et al., Reference Ghalili, McLennan, Pedersen, Kesby and Hyett2013).
Some limitations of our study should be acknowledged. First, IVF cases may be more likely to be referred to the antenatal care unit compared with natural pregnancies. However, this bias is unlikely to play a critical role here because MCT is one of the most challenging clinical conditions in pregnancy, and affected women are systematically referred to highly specialized services. The decision to refer is thus unlikely to be influenced by the mode of conception. Second, it is plausible that women conceiving by IVF and those conceiving naturally may differ in some characteristics that can be related to the risk of monozygosity. In other words, we cannot exclude that women requiring IVF may be at increased risk per se. Noteworthy, women conceiving naturally were younger in our study. Because a younger maternal age has been associated with the risk of monozygosity in IVF (Knopman et al., Reference Knopman, Krey, Lee, Fino, Novetsky and Noyes2010; Reference Knopman, Krey, Oh, Lee, McCaffrey and Noyes2014), we cannot exclude the possibility that we have underestimated the risk. Overall, this is a limitation of all observational studies in any area of medicine. In this regard, however, it has to be pointed out that, even though some confounders cannot be excluded, the magnitude of the risk (OR >3) tends to support a causal effect. In a recent luminous editorial on the interpretation of observational studies, Grimes and Schulz (Reference Grimes and Schulz2012) emphasized that associations with an OR below three are more likely to be attributable to biases than to causal association. Third, we could not provide data on risk factors for MCT among the group of IVF pregnancies. In particular, we could not explore the independent effects of transfer at the blastocyst stage (all included women transferred at the cleavage stage), assisted hatching (this technique was never used), and ICSI (only two women underwent classical IVF). As alluded to above, there is currently a burning debate on this point that, to date, has not yet lead to consistent conclusions. This is an important point because the practice of IVF is rapidly changing and transfer at the blastocyst stage is gaining consent. In other words, if blastocyst transfer, for example, proves to be a risk factor for MZT, the figure emerging from our analysis may change in the near future (data here refer to the period 2007–2011 when the blastocyst strategy was uncommon in our region). Fourth, as controls we referred to a population-based rate as estimated in a recent study of our group in the same geographical area (Parazzini et al., Reference Parazzini, Cipriani, Bulfoni, Bulfoni, Frigerio, Somigliana and Mosca2015). This choice presents some limitations. The study periods of referral are similar but not identical (2010–2012 in the referral study and 2007–2011 in the index study) and data were extracted from birth registers on the basis of subjects’ declarations (and thus they are exposed to some inaccuracies). In conclusion, according to data obtained from a prenatal diagnosis perspective, women undergoing IVF face an increased risk of MCT pregnancy, thus confirming previous evidence obtained from IVF case studies. Evidence from other geographical areas using the same study design is needed to confirm our estimation of the risk. Moreover, large multi-center studies are required to assess whether the prognoses of IVF and natural MCT pregnancies differ.
