The Alberta Health Technologies Decision Process (AHTDP) was established a decade ago to inform decisions on the public funding of nondrug health technologies by the provincial government. Since then, approximately fifty health technology assessments (HTAs) have been initiated, including one on in vitro fertilization (IVF).
In Alberta, IVF is not publicly funded. However, the ministry of health was recently asked to reconsider this position. As a result, an HTA of IVF was conducted under the auspices of the AHTDP. While the HTA was to address standard questions related to the safety, effectiveness and economic implications of IVF, its scope also included what is known about the effect of certain patient characteristics on clinical outcomes. Such information was needed to inform discussions around whether it may be clinically appropriate to consider incorporating patient eligibility criteria into funding decisions. This study reports on the systematic review undertaken to determine the state of the science around the potential impact of certain patient characteristics on the safety and effectiveness of IVF.
BACKGROUND
Infertility is clinically defined as the failure to achieve a clinical pregnancy following at least 12 months of unprotected sexual intercourse (Reference Zegers-Hochschild, Mansour and Ishihara1–Reference Gurunath, Pandian, Anderson and Bhattacharya3) (see Table 1 for descriptions of key terms in this study). Similar to prevalence rates reported in other western countries, approximately 10 to 15 percent of couples in Canada experience clinical infertility (Reference Evers4;Reference Bushnik, Cook, Yuzpe, Tough and Collins5).
Table 1. Key Terms and Definitions
It is well recognized that IVF is an effective treatment option for many couples with infertility. Demand for the procedure has grown, along with its use in broader patient populations, including postmenopausal women (Reference Ng and Ho6). However, neither its safety with respect to the procedure, pregnancy, and delivery in these broader populations, nor its effect on any offspring have been fully determined (Reference Caplan and Patrizio7–Reference Balasch10). Therefore, in an effort to balance patient demand, limited resources, and uncertainty over clinical benefit in such populations, as well as reduce the healthcare costs associated with complications of pregnancy, delivery, and preterm birth/delivery (PTB/PTD), many healthcare systems have considered or introduced funding policies that include patient eligibility criteria. These eligibility criteria commonly relate to maternal age, body mass index (BMI), and smoking status (Reference Homan, Davies and Norman11–Reference Dondorp, de Wert and Pennings14). However, the extent to which these policies reflect the current clinical evidence is unclear.
METHODS
A recent environmental scan of patient eligibility criteria contained in IVF funding policies around the world was used to identify patient characteristics to be assessed (Reference Dunn, Stafinski and Menon15). They were paternal age and maternal age, weight, and smoking. Relevant outcome measures (Table 2) were defined by an expert advisory committee, comprising general practitioners, reproductive endocrinologists, fertility specialists, gynecologists, neonatologists, and pediatricians.
Table 2. PICOS Elements of the Review Protocol
a Primary studies will be included if evidence gaps are identified after review of systematic reviews.
A comprehensive search for papers published between January 2007 and June 2014 that explored the impact of these characteristics on the safety and effectiveness of IVF was then conducted in two stages. The first stage considered systematic reviews only. The second stage considered primary studies of patient characteristics and outcomes that were not addressed in the systematic reviews.
Identification of Relevant Papers
Stage 1: Systematic Reviews. To identify reviews, a structured search strategy, which combined relevant controlled vocabulary terms, such as Medical Subject Headings, with additional nonindex keywords was developed and applied to key bibliographic databases (see Supplementary Table 1 for the complete details and outputs of the search). To identify unpublished reviews, Google, gray literature databases, and Web sites were searched, along with Web sites of guidelines, clinical trials, health technology assessment agencies, and key assisted reproductive technology-related international and national organizations. Furthermore, the reference lists of relevant papers were scrutinized and clinical experts in Alberta were contacted.
Stage 2: Primary Studies. Because no reviews assessing the impact of maternal characteristics on pregnancy/delivery or neonatal/infant complications, maternal smoking on the rates of ovarian hyperstimulation syndrome (OHSS), cycle cancellation, or multiple pregnancy/births, or paternal age on the chances of multiple pregnancy/birth were found, a separate search for primary studies was conducted. A structured search strategy similar to that used for the systematic reviews was developed and applied to PubMed (Supplementary Table 1).
The titles and abstracts from both stages were screened independently by two researchers (A.N.;K.K.) using the inclusion/exclusion criteria listed in Table 2. Full-text articles retrieved were reviewed independently by the same two researchers for final inclusion/exclusion using the same criteria. Agreement between researchers was assessed using the kappa statistic (Reference Cohen16). Disagreements were resolved through discussion and third-party review (TS).
Data Extraction and Critical Appraisal of Included Papers
Each researcher (A.N.;K.K.) independently extracted information from included reviews and studies using standard, pretested data abstraction forms and a set of decision rules (one for reviews and one for primary studies). The quality of each review was appraised using the Oxman and Guyatt index of scientific quality for systematic reviews (Reference Oxman and Guyatt17;Reference Oxman and Guyatt18). The quality of each primary study was assessed using the Oxford Levels of Evidence Scale (19). Where important details were missing from articles, the corresponding author was contacted for more information.
Synthesis of Extracted Information
Data extracted from reviews and primary studies were tabulated, to facilitate comparative analyses of the content and findings, and analyzed qualitatively. While a meta-analysis of reviews was planned, it could not be performed due to differences in inclusion/exclusion criteria across reviews and overlap between studies comprising reviews. Similarly, given the heterogeneity across included primary studies, a meta-analysis was not possible.
RESULTS
Overall Description of Included Reviews and Primary Studies
A total of 2,128 discrete citations were identified through the search for systematic reviews, of which, thirty-one potentially relevant reviews were selected for full assessment and ten for final inclusion (kappa = 0.92) (Figure 1). No reviews that assessed the impact of patient characteristic on neonatal/infant or pregnancy/delivery complications, paternal age on multiple pregnancy/birth, or maternal smoking on OHSS, cycle cancellation, or multiple pregnancy/birth were found (Table 3 and Supplementary Table 2). The search for primary studies yielded 7,181 discrete citations (Supplementary Figure 1), of which, seven were included (kappa = 1.00). Excluded studies and reasons for exclusion are listed in Supplementary Table 3. The outcomes of included studies are summarized in Supplementary Tables 4–12 (detailed methodological elements of each available upon request).
Table 3. Summary of Findings from Included Systematic Reviews and Primary Studies for Predefined Patient Characteristics and Outcomes of Interest
Note. Blanks indicate no reviews or primary studies were found.
aReduced pregnancy rates but not live birth rates.
Figure 1. PRISMA flowchart of literature search results and study selection for clinical effectiveness review: systematic reviews.
Maternal and Paternal Age
Parental age was assessed in four reviews of primarily retrospective cohort studies of fresh, autologous IVF/ICSI (intracytoplasmic sperm injection) with standard controlled ovarian stimulation (COS) protocols (Reference Watt, Elshaug, Willis and Hiller20–Reference van Loendersloot, van Wely and Limpens23). Two reviews assessed the effect of paternal age on pregnancy, miscarriage and live birth, one of which also explored maternal age and reported cycle cancellation, OHSS, and ectopic pregnancy rates (Reference Watt, Elshaug, Willis and Hiller20;Reference Dain, Auslander and Dirnfeld21). The third considered the impact of maternal age on pregnancy (Reference van Loendersloot, van Wely and Limpens23). The fourth comprised a meta-analysis of individual patient data from randomized clinical trials (RCTs) that assessed the effect of single versus double embryo transfer, including a subgroup analysis by maternal age, on live and multiple births (Reference McLernon, Harrild and Bergh22). No reviews evaluated the relationship between maternal or paternal age and obstetrical or neonatal/infant complications, or paternal age and multiple pregnancy/birth. Therefore, data were obtained from four additional primary studies (2012–2014): one retrospective cohort reporting birth defects in IVF/ICSI children among different maternal age groups and two retrospective cohorts and one prospective registry study assessing obstetrical complications after IVF/ICSI among women of different age groups (Reference Lawlor and Nelson24–Reference Ban Frangez, Korosec and Verdenik27). No additional primary studies assessing the effect of paternal age on the incidence of multiples, obstetrical complications, or neonatal complications were found.
Safety
One review assessing the effect of maternal and paternal age on OHSS and ectopic pregnancy rates following IVF/ICSI reported that, due to inadequate data, no conclusions could be made (Reference Watt, Elshaug, Willis and Hiller20). A retrospective analysis of primary data on singleton deliveries following IVF/ICSI showed no significant difference in PTD between women <33 years of age and those ≥33, or women <36 and those ≥36 (Reference Ban Frangez, Korosec and Verdenik27). However, a prospective IVF/ICSI registry study concluded that women under 40 were at a higher risk of PTB and delivering a low birth weight (LBW) baby than women 40 and older (Reference Lawlor and Nelson24). The authors attributed this finding to an increased likelihood of multiple births in women under 40, regardless of the number of embryos transferred. Finally, a third primary study comparing women >49 years of age with those <43 found no differences in birthweight, gestational age at delivery, or obstetrical complications, such as preeclampsia, gestational hypertension, gestational diabetes, and preterm prolonged rupture of membranes (Reference Kort, Gosselin and Choi26).
Evidence related to the risk of malformations among offspring conceived through IVF/ICSI is limited. Birth defect rates by maternal age were reported in one primary study, which considered all women undergoing IVF/ICSI at a clinic in China (Reference Yan, Wu, Tang, Ding and Chen25). This study found similar birth defect rates among the offspring of women aged 21–30, 30–35, 36–40, and 41 and over.
Efficacy/Effectiveness
Across the reviews, reduced pregnancy and live birth rates, as well as increased miscarriage rates after IVF/ICSI were associated with higher maternal age (Reference Watt, Elshaug, Willis and Hiller20;Reference McLernon, Harrild and Bergh22;Reference van Loendersloot, van Wely and Limpens23). Specifically, pooled data from one review of mainly retrospective cohort studies indicated that for each 1-year increase in maternal age, the odds of a clinical pregnancy after a fresh IVF/ICSI cycle were significantly decreased by 4–6 percent, while disaggregated data from another review of mainly retrospective studies showed reduced chances of pregnancy in women over 38, 35, and even 30 years of age (Reference Watt, Elshaug, Willis and Hiller20;Reference van Loendersloot, van Wely and Limpens23). In the most recent review, pregnancy rates were considerably lower in women >37 years of age (Reference Ban Frangez, Korosec and Verdenik27 percent) compared with those <37 (45 percent) (Reference Watt, Elshaug, Willis and Hiller20). They were also lower in women over 40–41 (3–9 percent), compared with those under 40–41 (17–28 percent), although statistical significance was not reported. A decline in the likelihood of live birth with increasing maternal age, particularly after age 40, was also seen (Reference Watt, Elshaug, Willis and Hiller20). Moreover, a meta-analysis of eight RCTs found that the odds of a live birth after IVF/ICSI were 40 percent greater in women <33 years of age compared with those ≥33 years of age (odds ratio [OR], 1.4; confidence interval [CI], 1.1–1.8) (Reference McLernon, Harrild and Bergh22).
In recent years, efforts to improve pregnancy and live birth rates in older women have included the use of IVF/ICSI with donor oocytes. A systematic review assessing donor IVF/ICSI suggested that pregnancy rates appeared to depend on the age of the oocyte donor, because rates were significantly lower in patients who received oocytes from donors 35–39 years old (32 percent) or >40 years old (23 percent) compared with those whose donors were 30–34 or <30 years of age (35–37 percent) (Reference Watt, Elshaug, Willis and Hiller20). In contrast, rates of miscarriage and the number of cycles required to achieve a live birth were found to depend on the age of recipients. Higher rates of miscarriage and a greater number of cycles required to achieve a live birth were observed in older recipients (>36–40), even when the donor was young.
One review addressing paternal age reported a significant linear decline in semen volume with increasing paternal age (Reference Dain, Auslander and Dirnfeld21). Most of the studies were retrospective cohorts or case-control studies reporting no significant age-effect on sperm concentration, motility, morphology, or implantation rates. However, some of these studies found that among men with oligozoospermia, the odds of an unsuccessful pregnancy increased by 5–11 percent for each 1-year increase in paternal age. In half of the included studies, live birth rates decreased with paternal age. One of those studies had adjusted for the type of infertility and the number of embryos transferred, reporting live birth rates of 38 percent, 17 percent, and 7 percent in women with male partners <35, 36–40, and >40 years of age, respectively. Furthermore, the proportion of women with a live birth after was significantly lower if the male partner was over 50 (41 percent) than if he was under 50 (56 percent; relative risk [RR], 1.5). Across most of the studies comprising the review, no significant variations in miscarriage rates with paternal age were found. In the single meta-analysis that assessed the effect of maternal age on multiple birth rates, no association between age and likelihood of a multiple birth was found (Reference McLernon, Harrild and Bergh22).
Maternal Weight
Five reviews, all of which included meta-analyses, evaluated the effect of maternal weight on the safety and effectiveness of IVF/ICSI (Reference Koning, Mutsaerts and Kuchenbecher28–Reference Jungheim, Schon and Schulte32). The reviews mainly comprised retrospective cohort studies of fresh, autologous IVF/ICSI cycles with a common COS protocol in unselected women or younger women (<35–44 years of age). One review also considered donor IVF/ICSI (Reference Jungheim, Schon and Schulte32). All five reviews reported at least one measure of the effectiveness of IVF/ICSI, including cycle cancellation (one review), pregnancy, miscarriage and/or live birth (five reviews), and multiple pregnancy (one review) (Reference Koning, Mutsaerts and Kuchenbecher28–Reference Jungheim, Schon and Schulte32). Safety data were limited to the incidence of OHSS (two reviews) and ectopic pregnancy (one review) and were, therefore, supplemented by four additional primary studies (retrospective cohorts published 2012–2014) reporting pregnancy/delivery and neonatal complications (Reference Ban Frangez, Korosec and Verdenik27;Reference Dickey, Xiong, Gee and Pridjian33–Reference Dickey, Xiong, Xie, Gee and Pridjian35).
In reviews and primary studies assessing the impact of BMI on the safety and effectiveness of IVF/ICSI, women with a BMI <20 were considered “underweight,” 20–24.9 “normal” weight, 25–29.9 “overweight,” 30–34.9 “obese,” and ≥35 “morbidly obese.” Many studies analyzed BMI at a cutoff of 25, comparing “underweight” or “normal” weight women (BMI <25) with “overweight,” “obese,” or “morbidly obese” women (BMI ≥25). Some studies used a BMI cutoff of 30 to compare “underweight,” “normal” weight, or “overweight” women (BMI <30) with “obese” and “morbidly obese” women (BMI ≥30).
Safety
The two reviews that addressed the effect of maternal weight on the safety of IVF/ICSI demonstrated no differences in OHSS or ectopic pregnancy rates between “normal” weight or “underweight” women and “overweight” or “obese” women (Reference Koning, Mutsaerts and Kuchenbecher28;Reference Maheshwari, Stofberg and Bhattacharya31).
Across the four primary studies that examined the relationship between BMI and PTB, the incidence of PTB increased with BMI. In the most recent of these studies, BMI >30 was identified as a significant risk factor for PTB among singleton births after IVF/ICSI (Reference Ban Frangez, Korosec and Verdenik27). Two of the studies presented retrospective analyses of a national IVF registry of over 80,000 IVF/ICSI births in the United States during different time periods. Both found that women whose BMI fell into one of three categories of increasing BMI cutoffs (“overweight,” “obese,” or “morbidly obese”) had a statistically significantly higher likelihood of PTB, early PTB, and very early PTB in singleton births (Reference Dickey, Xiong, Gee and Pridjian33;Reference Dickey, Xiong, Xie, Gee and Pridjian35).The fourth study, a prospective cohort study, also reported a positive correlation between BMI and singleton PTB rate. In addition, statistically significantly more obstetrical complications, including preeclampsia, hypertension, and gestational diabetes, were noted among “morbidly obese” women (20 percent) than “normal” weight women (12 percent; p < .05). The number of cesarean deliveries performed was also greater in “obese” women (51 percent) than in “normal” weight women (36 percent; p < .05) (Reference Zander-Fox, Henshaw, Hamilton and Lane34). However, no statistically significant association between BMI and stillbirths, neonatal deaths, or birth defects was found.
With respect to twin births, the findings appeared to depend on the BMI cutoff applied. Statistically significantly higher rates of PTB were only observed in women with a BMI defined as “morbidly obese” (Reference Dickey, Xiong, Gee and Pridjian33;Reference Dickey, Xiong, Xie, Gee and Pridjian35).
Efficacy/Effectiveness
Findings from the five reviews examining the relationship between maternal weight and pregnancy and live birth rates after IVF/ICSI varied (Reference Rittenberg, Seshadri and Sunkara29–Reference Jungheim, Schon and Schulte32;Reference Somigliana, Vercellini and Daguati36). In the first, neither clinical nor ongoing pregnancy rates differed with weight (Reference Koning, Mutsaerts and Kuchenbecher28). However, in the second, which comprised a meta-analysis of twenty-five studies (mostly retrospective cohort studies), the likelihood of pregnancy in women with a “normal” weight was greater compared with that in women who were “overweight” or “obese” (RR, 0.9; p < .05) (Reference Rittenberg, Seshadri and Sunkara29). In both of these reviews, women undergoing IVF/ICSI were significantly less likely to achieve a live birth if they were “overweight” or “obese” (OR, 0.8–0.9). In the third and earliest review (mostly retrospective cohort studies published 2000–2006), which included both biochemical and clinical pregnancies, pregnancy rates per woman were slightly lower in women with a BMI ≥25 than in women with a BMI <25 (OR, 0.81; CI, 0.67–0.98)) and were similar between women with a BMI ≥30 and <30 (Reference Maheshwari, Stofberg and Bhattacharya31). Cycle cancellation rates per woman were higher in women with a BMI ≥30 compared with women with a BMI <30 (p = .05), but no significant differences were seen at a BMI cutoff of 25. Furthermore, in women with a BMI ≥25, the numbers of oocytes retrieved were significantly lower compared with women with a BMI <25, even with higher doses and durations of gonadotropin stimulation (OR, 0.58) (Reference Maheshwari, Stofberg and Bhattacharya31). No significant differences in live birth rates were observed. One review specifically examined the effect of maternal BMI on the effectiveness of IVF/ICSI with donor embryos (Reference Jungheim, Schon and Schulte32). In this review, no significant differences in pregnancy or live birth rates between “underweight,” “normal” weight, “overweight,” or “obese” women were observed (Reference Jungheim, Schon and Schulte32).
In three reviews reporting miscarriage, “overweight” or “obese” women had a higher risk of miscarriage after IVF/ICSI compared with “normal” weight women (OR, 1.3–1.7) (Reference Rittenberg, Seshadri and Sunkara29;Reference Maheshwari, Stofberg and Bhattacharya31;Reference Somigliana, Vercellini and Daguati36). However, when studies where age may have been a confounder were excluded, differences were no longer significant (Reference Waylen, Metwally, Jones, Wilkinson and Ledger30). In the only review reporting multiple births, pooled data from five studies showed no significant association with BMI (Reference Koning, Mutsaerts and Kuchenbecher28).
Maternal Smoking Status
Pregnancy and live birth in women who were active smokers at the time of the IVF/ICSI procedure were compared with those in nonsmokers in a single review of prospective and retrospective cohort studies (Reference Waylen, Metwally and Jones37). Most studies were limited to autologous IVF/ICSI and used common COS protocols. Safety data in this review were limited to ectopic pregnancy rates and no reviews reporting the effect of smoking on cycle cancellation rates or the incidence of multiples after IVF/ICSI were found. Furthermore, no additional primary studies examining the effect of smoking on cycle cancellation, multiple pregnancy/birth, or complications in IVF/ICSI patients were found.
Safety
Based on the findings of a single meta-analysis, the risk of experiencing an ectopic pregnancy was greater in women who were active smokers at the time of IVF/ICSI than in nonsmokers, even after age was taken into account (OR, 15.7) (Reference Waylen, Metwally and Jones37). No reviews or additional primary studies examining the effect of smoking on pregnancy/delivery complications were identified.
Efficacy/Effectiveness
Findings from the single meta-analysis of maternal smoking status suggested that women who were active smokers at the time of IVF were almost half as likely to become pregnant or achieve a live birth and over 2 times as likely to experience a miscarriage than nonsmokers (p < .05), regardless of whether they used their own oocytes or donor oocytes (Reference Waylen, Metwally and Jones37). However, after excluding studies where age was considered to be a confounder, only differences in pregnancy rates remained significant.
DISCUSSION
This review presents the state of the science regarding the impact of different patient characteristics on the outcomes of IVF/ICSI to inform public funding deliberations, including consideration of possible patient eligibility criteria emerging from published clinical evidence.
Over the past 7 years, several systematic reviews assessing the impact of patient characteristics on IVF/ICSI have been published. Because they vary in quality and scope, a systematic review of reviews was performed. There can be several limitations to performing reviews of reviews. The initial review may not provide enough details on the primary studies. Where important details were missing, the corresponding author of the review was contacted for more information and the primary studies comprising the reviews were retrieved. Reviews of reviews can also be limited by the questions addressed in each review, which may not provide relevant or enough information. Where gaps in the evidence obtained from systematic reviews were identified, we searched for additional primary studies to fill them.
Although the systematic reviews were of high quality, they primarily comprised observational studies, most of which did not control for potential confounders. Consequently, the validity of results is limited and should be interpreted conservatively. Nonetheless, they suggest that in IVF/ICSI patients, increased maternal age, weight, and smoking during pregnancy constitute risk factors for certain pregnancy/delivery and infant complications. Advanced maternal age was associated with lower likelihoods of pregnancy and live birth and a greater risk of miscarriage. While semen volume seems to decrease with paternal age, whether or not this translates into reduced reproductive function or poorer success with IVF/ICSI remains unclear. Therefore, before eligibility criteria can be based on paternal age, more studies investigating the effect of paternal age on the safety and effectiveness of IVF/ICSI after controlling for maternal factors are needed. Evidence related to maternal smoking status at the time of IVF/ICSI treatment suggests that although smoking lowers the probability of becoming pregnant, it has little impact on the probability of achieving a live birth.
Given the differences in the impact of smoking status across outcome measures, discussions around the relative importance of each of these outcomes are needed before smoking status can be legitimately incorporated into eligibility criteria. Obesity appears to be negatively associated with both pregnancy and live birth rates. Thus, published evidence related to the safety and effectiveness of IVF supports consideration of eligibility criteria around maternal age and morbid obesity to optimize maternal and infant outcomes. Implementing such criteria may reduce resources spent on ineffective treatment cycles and healthcare costs associated with complications. What the exact cutoff values for maternal age and BMI should be warrants further investigation. While PTB and LBW were increased in women >40 years of age, pregnancy rates were shown to decline with age starting as young as 30 years. Pregnancy rates were lower in women with a BMI of ≥25. Risk of PTB was higher at a BMI of 25 in singleton pregnancies and 35 in twin pregnancies. Future research should focus on the impact of each of these factors together and in combination with procedural characteristics that may modify these effects, such as the number of embryos transferred and the use of adjunct procedures in combination with IVF (e.g., ICSI and preimplantation genetic diagnosis).
An in-depth discussion around what is an acceptable level of clinical effectiveness of IVF in order for it to be publicly funded is needed. There are few socially insured services with a clinical effectiveness as low as 5 percent. The exception may be life-threatening conditions with no treatment alternatives. Thus, it may be argued that the opportunity costs of funding IVF for patients in whom the probability of achieving a healthy live birth is below a certain level are simply too large. It then becomes inequitable to patients with other diseases for which interventions that offer similar or greater levels of effectiveness remain unfunded.
CONCLUSIONS
Based on the evidence reviewed, it may be appropriate to consider incorporating eligibility criteria around maternal age and obesity in public funding policies in Canada and internationally to optimize the safety and effectiveness of IVF and reduce costs associated with complications and ineffective treatment cycles. To determine appropriate cutoff points, further research and a discussion around acceptable levels of clinical effectiveness are needed.
SUPPLEMENTARY MATERIAL
Supplementary Tables 1–12
Supplementary Figure 1 http://dx.doi.org/10.1017/S0266462315000628
CONFLICTS OF INTEREST
The authors have no conflicts of interest to declare.
