12.1 Introduction

This chapter is a narrative review of the epidemiology of IVF and ICSI as the dominant sub-set of all available assisted reproductive technologies (ART). It is designed to provide the reader with an overview of the principal questions concerning the safety and effectiveness of technologies that have emerged over the last 40 years, from the early years of obscure animal experimentation to become routine medical practice globally for the treatment of infertility. Innovation at the edge of knowledge necessarily involves many steps and occasional leaps based on judgement. The work discussed here focuses on a selected body of literature where the author has confidence in describing the strengths of the research. This is important as much of the relevant literature suffers from serious design imperfections, which in part reflects the limitations of an emerging technology where early available data are partial, fragmented, and related only to short-term and intermediate outcomes. Common weaknesses include small sample size [1,2], pooling of exposure groups [3], or, specifically for case-control studies, retrospective collection of data and questionable appropriateness of controls [1,4,5]. At present, ART is falling short of realising opportunities and obligations to improve the precision and speed of its innovation cycle by not engaging in detailed analyses of clinical data. Nevertheless, with the rapid expansion of ART, we are now in a position to make confident statements on a range of outcomes for both IVF and ICSI, albeit ones that are provisional in the context of future innovation and changing patient characteristics.

12.2 Recent History

The use of assisted reproductive technologies (ART) for the treatment of infertility is increasing dramatically. Globally, more than 7 million babies have been born from assisted conception and this population is now increasing by over 1 million per annum [6]. Australia reflects this international trend; 1 in 25 Australian births are now from ART treatment [7].

In Australia, between 2002 and 2012, the number of ART treatment cycles almost doubled from 35,000 to 65,000 per annum (Figure 12.1). This rapid increase in use of treatments shows limited signs of abating in Australia [7] or internationally [6]. Treatments for infertility range from the simple oral administration of drugs to stimulate ovulation, to more invasive treatments such as IVF and ICSI that involve in vitro manipulation of gametes (oocytes and sperm). Data on offspring is only available on births from ART. Since the first IVF birth in 1978, the field of ART has been characterised by continual and rapid innovation, and increasing success in pregnancy and live birth rates (Figure 12.1). However, as ART has expanded, there is also accumulating evidence of adverse health impacts.

Figure 12.1 Trends in ART treatments and outcomes in Australia and New Zealand, 1993–2012*.

*based on annual reports of the Australian and New Zealand Assisted Reproduction Database

Early clinical studies necessarily focused on the short-term outcomes of fertilisation and pregnancy rates. However, we, and others internationally, have demonstrated ART to be associated with increased risks of low birth weight, preterm birth, stillbirth, neonatal death, and major birth defects, and more recently cognitive deficits with the risk of each adverse outcome varying by laboratory and clinical techniques.

Each of these will be discussed in relation to IVF and ICSI based on Australian data, with reference to supporting international work where relevant.

12.3 The South Australian Birth Cohort

We have established a whole population birth cohort, which is a census of all births in South Australia (a State of the Commonwealth of Australia, population 1.6 million) for the period 1986–2002. This includes approximately 327,000 births, including those conceived with ART (~6,500), terminations, and birth defects ascertained to age 5 years (~17,000). These have been linked to all ART treatments, patient diagnoses, and laboratory procedures. This is comprehensive, detailed, representative, and generalisable to other research cites internationally. The cohort is undergoing expansion with plans to eventually become an historical repository covering the entire history of ART technologies and outcomes, including the dramatic expansion of ICSI, the shift to single embryo transfer (SET), routine use of cryopreservation, and changes in embryo culture technology (including extended culture).

12.4 Infertility Patient Records

Details of ART treatment, as defined by the National Health and Medical Research Council (NHMRC) [8], were provided by clinics registered to provide infertility treatment involving embryo manipulation. Greater than 99.99% of ART births were linked to the state birth registry as interstate movements for treatment were rare.

12.5 Perinatal Outcomes

Any ART birth is recorded within the State-wide perinatal collection, which by law requires notification of all live births and stillbirths of at least 20 weeks gestation or 400 g birth weight in South Australia using a standardised notification form (www.health.sa.gov.au/pehs/pregnancyoutcome.htm). Maternal pre-existing medical conditions and conditions in pregnancy are also recorded from the labour ward records onto the notification form. Approximately 20,000 births are recorded annually for South Australia. Notifications of all medical terminations of pregnancy are also required by State law, and those that are induced at 20 weeks gestation or more are included in the perinatal data collection. For completeness, we did not exclude births to women with unknown or out-of-state addresses (0.6% of entire sample, n = 1,916).

12.6 Birth Defects

Congenital abnormalities detected at birth or in the neonatal period (within 28 days of birth) are reported by doctors using a standardised Congenital Abnormality Form. The South Australian Birth Defects Register includes information on birth defects (including cerebral palsy) obtained from the perinatal and abortion statistics collections, as well as notifications from multiple sources up to the child’s fifth birthday. This is important as only half of major defects are identified at birth (communication, SA Birth Defects Register).

Post-neonatal or “acquired” cerebral palsy cases (i.e. attributed to events occurring after perinatal period) were not included. Birth defects were coded by registry staff independent of birth defect notifications, although blinding of mode of conception by clinical observers issuing notifications was not possible. Prior assessment of the same reporting method in an adjacent jurisdiction revealed no significant reporting bias [9].

Terminations of pregnancy for congenital abnormalities at gestations of less than 20 weeks are reported by law to the Department of Health and included in the State Birth Defects Register. Birth defect diagnoses are validated by cross-referencing of medical reports before being registered, and are coded according to the British Paediatric Association modification of the International Classification of Diseases, 9th Revision (ICD-9 BPA), including abnormalities that are structural, biochemical, chromosomal, or other genetic abnormalities. (www.wch.sa.gov.au/services/az/other/phru/birthdefect.html). Minor defects are generally excluded from the register, with the exception of those that require treatment or are disfiguring. Linkage of the ART patient record (used probabilistic matching software) and hand matching (using patient identifiers and birth outcome data) was undertaken. The birth defect data were linked to the perinatal outcomes collection and to the ART pregnancies by a unique accession number for each birth. Hand matching was used to resolve inconsistencies in the patient or birth data between the files, such as family name change for mothers.

Approval for the study was obtained from the ethics committees of the South Australian Department of Health, the University of Adelaide, and the Flinders University of South Australia. As patients were not identified, individual patient consent was not required by the ethics committees.

12.7 Exposures and Outcomes

12.8 Analysis

The available dataset contained a total of 327,420 births and terminations. After exclusion of births to mothers younger than 20 years (among whom there were only two ART births), there were 308,974 births for analysis. The prevalence of birth defects was compared between the following groups: (i) births as a result of each modality of infertility treatment, including spontaneous pregnancies while a patient was under care; (ii) births as a result of spontaneous pregnancies in women with previous ART birth; (iii) births to women with a history of infertility on their perinatal outcomes record and no history of ART treatment; and (iv) births to women in the general population with no recorded history of infertility or treatment.

Odds ratios (OR) were calculated comparing the prevalence of birth defects between groups using two-tailed p values with the aid of SAS statistical software. No adjustment for multiple births was made, except in sensitivity analyses to assess model robustness, as multiple gestation may be considered on the causal pathway between ART exposure and birth defects [9]. Information on the zygosity of twins was not available.

The “crude” estimates include minimal adjustment for the effect of clustering of births within women, using logistic Generalized Estimating Equations. The adjusted analyses included a-priori confounders of maternal age (categorized in 5-year age groups), parity, foetal sex, year of birth, maternal ethnicity, maternal country of birth, and maternal conditions in pregnancy (pre-existing hypertension, pregnancy-induced hypertension, pre-existing diabetes, gestational diabetes, anaemia, urinary tract infection, epilepsy, asthma), maternal smoking in pregnancy, postal code indicators of socioeconomic disadvantage from Socio-economic Indices for Areas [11], and maternal and paternal occupation, coded to the Australian Standard Classification of Occupations [10]. We also pre-specified subgroup analyses for singleton and multiple births, and used pre-specified contrasts to test treatment modality effects (including fresh vs frozen embryo cycles) using the same analytic strategy. There was no adjustment for multiple comparisons.

12.9 Outcomes

There is now a clear body of evidence demonstrating increased risks of poorer perinatal outcomes among ART children [12], with evidence first emerging in 1985 [13]. Historically, the risks have been attributed to an increased prevalence of multiple pregnancies arising in ART, largely due to multiple embryo transfer.

However, relative to natural conceptions, ART singletons have compromised health [12]. Evidence includes our own reports of increased risks of stillbirth (odds ratio [OR] 1.82, 95% CI 1.34–2.48), neonatal death (OR = 2.04 95% CI 1.27–3.26), preterm birth (OR = 1.64, 95% CI 1.46–1.84), low birth weight (OR = 1.98, 95% CI 1.77–2.20), and major birth defects (OR 1.30, 95% CI 1.16–1.45) among singletons conceived with any ART [14,15].

Consistent evidence from individual studies, including registry-based cohort studies [16,17] and meta-analyses has linked assisted conception using in vitro fertilisation (IVF) or intra-cytoplasmic sperm injection (ICSI) with an increased risk of birth defects [5, 9,18–21]. The associations between the use of these techniques and birth defects has appeared stronger for singleton than multiple births [4,22].

It is unclear whether the excess of birth defects following IVF and ICSI may be attributable to patient characteristics related to infertility [8], rather than to the treatments, and whether the risk is similar across ART and related therapies [5,23,24].

There is limited understanding of the causes of poor ART outcomes. Parental characteristics related to infertility are likely to contribute to the risk of adverse outcomes including preterm birth and birth defects [12]. Treatment-related factors, particularly mode of fertilisation, have also been proposed [12,25]. Recent meta-analyses confirm that pregnancies resulting from IVF/ICSI have worse perinatal outcomes than natural conceptions [12,26]. However, as both IVF and ICSI comprise a complex set of laboratory and clinical procedures, it is difficult to determine which particular aspects of these treatments are causing poor outcomes. For example, a typical cycle of IVF or ICSI now involves controlled ovulation induction, oocyte retrieval, fertilisation (either IVF or with ICSI), embryo maturation in culture, embryo transfer, and cryopreservation of excess embryos. As a result, there have been calls for research examining the specific factors that are modifiable, rather than implicating IVF or ICSI overall as harmful [27]. The elevated risks associated specifically with one treatment (e.g. ICSI) may also mask reductions in risk over time in other aspects of treatment.

Hence, the current literature is limited by a number of design issues, including the pooling of ART exposure groups, resulting in a lack of specificity in the type, and magnitude of effect, from specific treatment factors.

We have previously demonstrated that the risk of adverse outcomes varies across individual treatment groups. For example, the risk of neonatal death increased further for IVF singletons after fresh embryo transfer (OR = 4.92, 95% CI 2.65–9.11) compared with natural conceptions [7]. Furthermore, singleton ICSI pregnancies were more likely to be complicated by pregnancy-induced hypertension and an increased risk of macrosomia after embryo freezing (OR = 1.54, CI 1.0–2.28), which is consistent with previous studies [28] and may reflect an altered epigenetic signature in the embryo [29]. Risk of major birth defects also varied substantially across treatments, for example the risk was non-significant in IVF singletons but elevated for ICSI (OR = 1.55, 95% CI 1.24–1.94). We showed reduced risks of birth defects after frozen embryo cycles, particularly with ICSI (OR = 1.10, 95% CI 0.65–1.85) [6]. The defects were serious, including cardiac, urogenital, musculoskeletal, and neurological defects.

A limitation of our published work [14,15] is the age of the data, as the most recent births occurred in December 2002. As a result, the findings may not be a reliable guide for contemporary laboratory and clinical practices because the intervening years have seen a number of important changes in culture media, laboratory procedures, clinical practice, and patient infertility diagnosis.

Recent analysis of Nordic birth registry data for the years 1998–2007 [30] identified a steep decline over time in the prevalence of preterm birth among ART singletons (~14–8%), as well as smaller declines in small for gestational age, stillbirth, and infant death (in singletons and twins). The authors suggest that the improvements reflect changes in treatment mix, including greater use of SET, embryo cryopreservation, ICSI, as well as changes in the health profile of couples accessing treatment. However, the Nordic temporal trends were not analysed specifically by these factors, and so the authors could only speculate about their possible impact.

There is also some evidence of improved perinatal outcomes (preterm birth, low birth weight) in more recent ART cohorts examining singleton births from IVF and ICSI [12]. The latest of these cohorts includes ART births until 2006 [31]. Therefore, there is a major gap internationally in knowledge of the safety and effectiveness of ART treatments commonly used in the most recent decade. The extent to which specific changes in ART treatment mix contribute to variation in outcomes in Australia, taking into account patient profile, is unknown. This makes it imperative to create a more contemporary continuously updated dataset, in order to capture the key innovations and changes in practice.

12.10 Recent Innovations

The dramatic rise in the use of ICSI indicates that it is no longer used exclusively for severe semen defects (in ~15% of cases), and now accounts for more than 70% of all treatment cycles globally [32] (Figure 12.1). ICSI improves fertilisation rates, but bypasses a number of biological checkpoints. Recent evidence from AI Kissin’s group in Atlanta shows the “take home baby rate” is no better after ICSI, but has worse perinatal outcomes compared to IVF [33]. This confirms that there is a misinformed enthusiasm for ICSI internationally, with recent calls for quantification of the risks of ICSI when used for non-male factor infertility [27].

There are also calls for use of large datasets to examine specific patterns of birth defects associated with ICSI [2] to elucidate underlying mechanisms. Recent studies have indicated a predominance of defects of the cardiovascular, genitourinary, and gastrointestinal systems [2,34], although results have been inconsistent [20]. Existing studies lack statistical power to investigate associations, or to appropriately account for the underlying infertility of couples [26].

Within the last 10 years there has been a significant increase in elective SET, such that now it accounts for the majority (76%) of treatment cycles (Figure 12.1). This practice significantly reduces iatrogenic multiple birth and associated adverse outcomes [30,35,36]. It is important to examine whether the shift to use of SET in Australia has improved perinatal outcomes. First, even though SET is routine in Australia for younger women, multiple embryo transfer still occurs frequently, with twinning still increased fourfold over natural conceptions. In addition, singleton birth after a double embryo transfer (DET) is not benign, as there is evidence that singletons born after SET have higher birth weight than singletons born after DET, and fewer neonatal deaths [30,37].

Consistent with the literature on foetal loss and birth defects [38] we have recently shown that following DET, the presence of a non-progressing foetal co-twin at the 8-week ultrasound (e.g. an empty sac) is associated with a significantly increased risk of major birth defects in the survivor (OR = 2.78), and an overall 18% prevalence of major birth defects [39], which may be due to multiple embryos “cloaking” poor embryos from quality sensing by the endometrium [40]. Therefore, there is a need to further clarify the risks of multiple embryo transfer, particularly when this practice results in a singleton birth, as this is where we observe the defects occurring. We propose that there will be a reduction in the risk of birth defects after elective SET versus DET.

Increasing use of cryopreservation augments SET to reduce multiple pregnancy rates, while also reducing the risk of low birth weight and birth defects, particularly for ICSI [14]. Shifts away from a “slow” freezing method to vitrification whereby the embryo is plunged into liquid nitrogen are occurring, with recent data, particularly from “freeze only” protocols, indicating that children born following the transfer of vitrified embryos may have a higher birth weight when compared with those of fresh or slow frozen embryos [41]. The effect of this technique on perinatal outcomes and birth defects has not been evaluated in detail. The most recent meta-analysis [42] reported that only one study had included congenital abnormalities as an end-point. However, we can now show in our existing dataset (manuscript in preparation) that while cryopreservation is beneficial for defects overall, it does not reduce cardiac defects and may specifically increase the risk of circulatory defects after ICSI (OR = 4.7, CI = 1.89–11.77).

There has been an increased use of extended embryo culture and transfer of blastocysts (i.e. day 5 transfer) to select for longer surviving and potentially euploid embryos [43]. However, this may also alter the pattern of development, where blastocyst transfer is associated with increased monozygotic twinning, which is a risk factor for major adverse outcomes [44,45]. Blastocyst transfer is also associated with an increased risk of birth defects [46,47], which is proposed to occur due to non-physiologic oxygen exposure [48]. Adverse perinatal outcomes were not observed in a recent Australian study [49]; however, birth defects were not examined. Therefore, the impact of day of transfer on perinatal outcomes requires verification, while the risk for birth defects is unclear [12].

Factors in culture media influence fertilisation rates and clinical pregnancy rates, where even small alterations in laboratory parameters result in altered foetal growth in utero in animal models [50,51]. Assessment of birth outcomes in humans is scarce and inconsistent. The composition of the culture media for embryo development to the blastocyst stage may alter the birth weight of the babies [52,53], although this finding is not consistent across all studies [54,55]. We have identified that certain brands of culture media were associated with an increased risk of any cardiac defect (OR = 2.56, 1.05–6.23), and of Tetrology of Fallot (OR = 4.16, 1.16–14.9) compared to natural conceptions, controlling for mode of conception and other patient and treatment factors (manuscript in preparation). It is imperative that these and other adverse outcomes are investigated in a contemporary context, as there have been rapid developments in culture media content in the past decade.