- Split View
-
Views
-
Cite
Cite
Abha Maheshwari, Shilpi Pandey, Edwin Amalraj Raja, Ashalatha Shetty, Mark Hamilton, Siladitya Bhattacharya, Is frozen embryo transfer better for mothers and babies? Can cumulative meta-analysis provide a definitive answer?, Human Reproduction Update, Volume 24, Issue 1, January-February 2018, Pages 35–58, https://doi.org/10.1093/humupd/dmx031
- Share Icon Share
Abstract
Initial observational studies and a systematic review published 5 years ago have suggested that obstetric and perinatal outcomes are better in offspring conceived following frozen rather than fresh embryo transfers, with reduced risks of preterm birth, small for gestational age, low birth weight and pre-eclampsia. More recent primary studies are beginning to challenge some of these findings. We therefore conducted an updated systematic review and cumulative meta-analysis to examine if these results have remained consistent over time.
The aim of this study was to perform a systematic review and cumulative meta-analysis (trend with time) of obstetric and perinatal complications in singleton pregnancies following the transfer of frozen thawed and fresh embryos generated through in-vitro fertilisation.
Data Sources from Medline, EMBASE, Cochrane Central Register of Clinical Trials DARE and CINAHL (1984–2016) were searched using appropriate key words. Observational and randomised studies comparing obstetric and perinatal outcomes in singleton pregnancies conceived through IVF using either fresh or frozen thawed embryos. Two independent reviewers extracted data in 2 × 2 tables and assessed the methodological quality of the relevant studies using CASP scoring. Both aggregated as well as cumulative meta-analysis was done using STATA.
Twenty-six studies met the inclusion criteria. Singleton babies conceived from frozen thawed embryos were at lower relative risk (RR) of preterm delivery (0.90; 95% CI 0.84–0.97) low birth weight (0.72; 95% CI 0.67–0.77) and small for gestational age (0.61; 95% CI 0.56–0.67) compared to those conceived from fresh embryo transfers, but faced an increased risk (RR) of hypertensive disorders of pregnancy (1.29; 95% CI 1.07–1.56) large for gestational age (1.54; 95% CI 1.48–1.61) and high birth weight (1.85; 95% CI 1.46–2.33). There was no difference in the risk of congenital anomalies and perinatal mortality between the two groups. The direction and magnitude of effect for these outcomes have remained virtually unchanged over time while the degree of precision has improved with the addition of data from newer studies.
The results of this cumulative meta-analysis confirm that the decreased risks of small for gestational age, low birth weight and preterm delivery and increased risks of large for gestational age and high birth weight associated with pregnancies conceived from frozen embryos have been consistent in terms of direction and magnitude of effect over several years, with increasing precision around the point estimates. Replication in a number of different populations has provided external validity for the results, for outcomes of birth weight and preterm delivery. Meanwhile, caution should be exercised about embarking on a policy of electively freezing all embryos in IVF as there are increased risks for large for gestational age babies and hypertensive disorders of pregnancy. Therefore, elective freezing should ideally be undertaken in specific cases such as ovarian hyperstimulation syndrome, fertility preservation or in the context of randomised trials.
Introduction
In-vitro fertilisation involves hormonal stimulation of ovaries followed by surgical retrieval of oocytes and their insemination in the laboratory. Conventionally, embryos created by this process are transferred within the uterus after 2–5 days in culture, while any remaining embryos are frozen for subsequent use. Cryopreserved embryos are usually thawed and replaced in a natural or hormonally manipulated cycle in women in whom a fresh embryo transfer fails to result in a pregnancy or in those who return for a second baby.
The first live birth following the transfer of thawed cryopreserved embryos was reported in 1984. With refinement of technology over the last few decades, the number of frozen embryo transfers has increased as have pregnancy rates which, according to some authors, are better than those following the transfer of fresh IVF embryos (Chen et al., 2016).
Initial observational studies and a systematic review based on these which was published 5 years ago, have suggested that obstetric and perinatal outcomes are better in those conceived following frozen rather than fresh embryo transfers (Maheshwari et al., 2012), with reduced risks of preterm birth, small for gestational age babies, low birth weight babies and pre-eclampsia. Subsequent primary studies (Chen et al., 2016; Maheshwari et al., 2016) are beginning to challenge some of these initial findings. We therefore conducted a new systematic review incorporating all the published studies and including a cumulative meta-analysis to examine whether the results have remained consistent over time.
Materials and Methods
PRISMA guidelines for systematic reviews were followed (http://www.plosmedicine.org/article/info:doi%2F10.1371%2Fjournal.pmed.1000097). The protocol was registered at PROSPERO (CRD42016046131).
Data sources and searches
A literature search with no language restrictions was performed (1984–2016) on Medline, EMBASE, Cochrane Central Register of Clinical Trials, CINAHL and DARE (Supplementary Table S1). Relevant journals in the specialty (Human Reproduction, Human Reproduction Update, RBM online and Fertility and Sterility) were searched electronically. Cross references from the included studies were hand searched. Two review authors (AM, SP) independently conducted the searches and selected the studies to be included. Differences of opinion were resolved after team discussion. Contact with authors was attempted wherever additional information was needed. Data were extracted using pre designed 2 × 2 tables.
Quality assessment of included studies was performed independently by two authors (SP and AM). Any disagreement regarding type and quality of the study was resolved after discussion. Checklists from the critical appraisal skills programme (CASP) (http://www.phru.nhs.uk/pages/phd/resources.htm) were used to assess and assign a quality score. CASP is a critical appraisal tool consisting of 12 questions to appraise a cohort study systematically in three broad domains: Are the results of the study valid? (Section A); What are the results? (Section B); and Will the results help locally? (Section C). A score is then allocated out of 12 (Table I).
Study ID . | Design of study . | Population . | Method of data collection . | Risk of bias . | Scoring . |
---|---|---|---|---|---|
Aflatoonian et al. (2010a) | Unmatched Cohort study | 500 pregnancies obtained after the transfer of fresh ET and 200 pregnancies after FET from March 2006 to March 2008. | Questionnaires filled by gynaecologists, paediatrics and women regarding perinatal and obstetric outcomes |
| 9/12 |
Aflatoonian et al. (2016) | Unmatched Cohort study |
| Data were collected from the hospital records. In addition, a telephone questionnaire consists of data on maternal and neonatal factors was administered by a trained nurse based on patients and their husbands’ information. |
| 9/12 |
Belva et al. (2008) | Unmatched Cohort Study |
| Data on pregnancies, deliveries and neonatal history was obtained by gynaecologists, paediatricians and double checked with parents, when child was 2 months old. | Mode of delivery and Duration of infertility was significantly different in the two groups. | 10/12 |
Healy et al. (2010) |
|
| Data were collected using record linkage of national databases | Includes first singleton birth only, delivered after 20 weeks. No data on demographic profile of the women in FET vs. fresh group | 10/12 |
Imudia et al. (2013) | Retrospective cohort study | Twenty women who underwent elective cryopreservation of all embryos with subsequent cryothaw ET and 32 similar women with elevated peak E2 during controlled ovarian hyperstimulation for IVF who underwent a fresh ET. | Data were collected from Medical records | Study adjusted for confounders (body mass index, antral follicle count, peak serum E2 level) Excluded peak serum E2 > 4500 pg/ml | 9/12 |
Ishihara et al. (2014) | Cohort | Registered from 2008 through 2010 undergoing single embryo transfer cycles. Only singleton ongoing pregnancies >21 weeks of gestation were included. | Japanese nationwide registry of assisted reproductive technology (ART) with mandatory reporting for all ART clinics in Japan. |
| 11/12 |
Kato et al. (2012) | Retrospective Cohort | Single-centre retrospective cohort study of 6623 consecutive delivered singletons following 29 944 single-embryo transfers. January 2006 and December 2008 | Two-part questionnaire filled by patient at the 20th pregnancy week and after delivery. | There was no difference in baseline characteristics in both group. | 11/12 |
Li et al. (2014) | Retrospective Cohort study | Reterospective population-based cohort study from Jan 2009 to Dec 2011 of autologous fresh and frozen cycles in Australia and New Zealand | ART treatment information and perinatal outcomes were obtained from the Australian and New Zealand Assisted Reproduction Database (ANZARD). | 9/12 | |
Liu et al. (2013) | Retrospective single centre analysis | Retrospective, single-centre study of children born after Day 3 embryo transfer from fresh, slow frozen or vitrified embryos during the period January 2006 to May 2011 | Data obtained via patient filled questionnaire at 12 weeks |
| 8/12 |
Maheshwari et al. (2016) | Retrospective analysis | Retrospective analysis of annonymized HFEA data | Data taken from HFEA database (which gets reported to HFEA by clinics as part of regulatory requirement) |
| 11/12 |
Opdahl et al. (2015) | Unmatched cohort study | Nationwide data from registries of Denmark, Norway and Sweden | Data obtained from health registries | Baseline characteristics for women having fresh or frozen embryo transfer were not compared. As fresh and frozen embryo transfers were one of multiple comparisons | 11/12 |
Pelkonen et al. (2010) | Unmatched Cohort study (1995–2006) |
| Data taken from Finnish Medical Birth Register |
| 11.5/12 |
Pelkonen et al. (2014) | Register based cohort study |
| Linkage of fertility, birth and congenital anomalies registries | There was a higher proportion of nulliparous women in fresh ET group | 11/12 |
Pereira et al. (2016) | Retrospective review | Consecutive live deliveries from all patients who began IVF cycles at the single centre between 1 January 2010 and 30 September 2013. | Data collected by retrospective review of patients charts | Patients were of similar age, BMI, infertility diagnosis, endometrial thickness and there was no difference in the grading of blastocysts. | 11/12 |
Pinborg et al. (2010) | Matched Cohort study |
| Danish IVF and Danish Birth Register |
| 11/12 |
Pinborg et al. (2014) | The national register-based controlled cohort study |
| Registry data |
| 11/12 |
Rallis and Tremellen (2013) | Retrospective review |
| Clinic based data, case records, database | Basic demographic data other than age group was not available confounding factors for preterm birth such as previous pregnancy outcomes were not available. | 10/12 |
Roy et al. (2014) | Retrospective cohort | Single centre Assisted Reproduction clinic between March 2010 and November 2011 | Private IVF Clinic database | Data for the fresh group were restricted to the patients with three or fewer stimulation cycles who had single blastocyst transferred and one blastocyst cryopreserved. | 10/12 |
Shapiro et al. (2016) | Post hoc analysis of two RCT | Two RCTs from same centre one on hyper responders and one on normal responders | Birth weight outcome; post hoc analysis | Data obtained through personal communication | NA |
Shih et al. (2008) | Matched cohort with women acting as their own reference | Comparison groups: Frozen versus Fresh IVF/ICSI | Neonatal perinatal statistics unit Austraila | All pregnancies after 20 weeks were recorded. Fresh IVF/ICSI conception could be first/second one | 10/12 |
Shi et al. (2012) | Retrospective data | Single centre Assisted Reproduction clinic | The outcome data were obtained from a postal questionnaire of parents after delivery |
| 8/12 |
Wada et al. (1994) | Unmatched cohort |
| Data were collected from medical records | 7/12 | |
Wang et al. (2005) | Unmatched cohort study | Infants conceived through ART Procedures and born in Australia during 1996–2000 | The study used data from two national collections. Assisted conception data collection & Australian national perinatal data collection | Fresh and frozen pregnancies were subgroup analysis. Hence not matched for the confounders. | 9/12 |
Wennerholm et al. (1997) | Matched cohort |
| Data were collected after medical records review | Controls were matched for age and parity | 10.5/11 |
Wennerholm et al. (2013) | Reterospective Matched cohort study |
| Data on perinatal outcomes were obtained by linkage to the national Medical Birth Registries |
| 11/12 |
Wikland et al. (2010) | Unmatched Cohort Study |
| Data for obstetric and perinatal complications was collected from maternity records |
| 11/12 |
Study ID . | Design of study . | Population . | Method of data collection . | Risk of bias . | Scoring . |
---|---|---|---|---|---|
Aflatoonian et al. (2010a) | Unmatched Cohort study | 500 pregnancies obtained after the transfer of fresh ET and 200 pregnancies after FET from March 2006 to March 2008. | Questionnaires filled by gynaecologists, paediatrics and women regarding perinatal and obstetric outcomes |
| 9/12 |
Aflatoonian et al. (2016) | Unmatched Cohort study |
| Data were collected from the hospital records. In addition, a telephone questionnaire consists of data on maternal and neonatal factors was administered by a trained nurse based on patients and their husbands’ information. |
| 9/12 |
Belva et al. (2008) | Unmatched Cohort Study |
| Data on pregnancies, deliveries and neonatal history was obtained by gynaecologists, paediatricians and double checked with parents, when child was 2 months old. | Mode of delivery and Duration of infertility was significantly different in the two groups. | 10/12 |
Healy et al. (2010) |
|
| Data were collected using record linkage of national databases | Includes first singleton birth only, delivered after 20 weeks. No data on demographic profile of the women in FET vs. fresh group | 10/12 |
Imudia et al. (2013) | Retrospective cohort study | Twenty women who underwent elective cryopreservation of all embryos with subsequent cryothaw ET and 32 similar women with elevated peak E2 during controlled ovarian hyperstimulation for IVF who underwent a fresh ET. | Data were collected from Medical records | Study adjusted for confounders (body mass index, antral follicle count, peak serum E2 level) Excluded peak serum E2 > 4500 pg/ml | 9/12 |
Ishihara et al. (2014) | Cohort | Registered from 2008 through 2010 undergoing single embryo transfer cycles. Only singleton ongoing pregnancies >21 weeks of gestation were included. | Japanese nationwide registry of assisted reproductive technology (ART) with mandatory reporting for all ART clinics in Japan. |
| 11/12 |
Kato et al. (2012) | Retrospective Cohort | Single-centre retrospective cohort study of 6623 consecutive delivered singletons following 29 944 single-embryo transfers. January 2006 and December 2008 | Two-part questionnaire filled by patient at the 20th pregnancy week and after delivery. | There was no difference in baseline characteristics in both group. | 11/12 |
Li et al. (2014) | Retrospective Cohort study | Reterospective population-based cohort study from Jan 2009 to Dec 2011 of autologous fresh and frozen cycles in Australia and New Zealand | ART treatment information and perinatal outcomes were obtained from the Australian and New Zealand Assisted Reproduction Database (ANZARD). | 9/12 | |
Liu et al. (2013) | Retrospective single centre analysis | Retrospective, single-centre study of children born after Day 3 embryo transfer from fresh, slow frozen or vitrified embryos during the period January 2006 to May 2011 | Data obtained via patient filled questionnaire at 12 weeks |
| 8/12 |
Maheshwari et al. (2016) | Retrospective analysis | Retrospective analysis of annonymized HFEA data | Data taken from HFEA database (which gets reported to HFEA by clinics as part of regulatory requirement) |
| 11/12 |
Opdahl et al. (2015) | Unmatched cohort study | Nationwide data from registries of Denmark, Norway and Sweden | Data obtained from health registries | Baseline characteristics for women having fresh or frozen embryo transfer were not compared. As fresh and frozen embryo transfers were one of multiple comparisons | 11/12 |
Pelkonen et al. (2010) | Unmatched Cohort study (1995–2006) |
| Data taken from Finnish Medical Birth Register |
| 11.5/12 |
Pelkonen et al. (2014) | Register based cohort study |
| Linkage of fertility, birth and congenital anomalies registries | There was a higher proportion of nulliparous women in fresh ET group | 11/12 |
Pereira et al. (2016) | Retrospective review | Consecutive live deliveries from all patients who began IVF cycles at the single centre between 1 January 2010 and 30 September 2013. | Data collected by retrospective review of patients charts | Patients were of similar age, BMI, infertility diagnosis, endometrial thickness and there was no difference in the grading of blastocysts. | 11/12 |
Pinborg et al. (2010) | Matched Cohort study |
| Danish IVF and Danish Birth Register |
| 11/12 |
Pinborg et al. (2014) | The national register-based controlled cohort study |
| Registry data |
| 11/12 |
Rallis and Tremellen (2013) | Retrospective review |
| Clinic based data, case records, database | Basic demographic data other than age group was not available confounding factors for preterm birth such as previous pregnancy outcomes were not available. | 10/12 |
Roy et al. (2014) | Retrospective cohort | Single centre Assisted Reproduction clinic between March 2010 and November 2011 | Private IVF Clinic database | Data for the fresh group were restricted to the patients with three or fewer stimulation cycles who had single blastocyst transferred and one blastocyst cryopreserved. | 10/12 |
Shapiro et al. (2016) | Post hoc analysis of two RCT | Two RCTs from same centre one on hyper responders and one on normal responders | Birth weight outcome; post hoc analysis | Data obtained through personal communication | NA |
Shih et al. (2008) | Matched cohort with women acting as their own reference | Comparison groups: Frozen versus Fresh IVF/ICSI | Neonatal perinatal statistics unit Austraila | All pregnancies after 20 weeks were recorded. Fresh IVF/ICSI conception could be first/second one | 10/12 |
Shi et al. (2012) | Retrospective data | Single centre Assisted Reproduction clinic | The outcome data were obtained from a postal questionnaire of parents after delivery |
| 8/12 |
Wada et al. (1994) | Unmatched cohort |
| Data were collected from medical records | 7/12 | |
Wang et al. (2005) | Unmatched cohort study | Infants conceived through ART Procedures and born in Australia during 1996–2000 | The study used data from two national collections. Assisted conception data collection & Australian national perinatal data collection | Fresh and frozen pregnancies were subgroup analysis. Hence not matched for the confounders. | 9/12 |
Wennerholm et al. (1997) | Matched cohort |
| Data were collected after medical records review | Controls were matched for age and parity | 10.5/11 |
Wennerholm et al. (2013) | Reterospective Matched cohort study |
| Data on perinatal outcomes were obtained by linkage to the national Medical Birth Registries |
| 11/12 |
Wikland et al. (2010) | Unmatched Cohort Study |
| Data for obstetric and perinatal complications was collected from maternity records |
| 11/12 |
Study ID . | Design of study . | Population . | Method of data collection . | Risk of bias . | Scoring . |
---|---|---|---|---|---|
Aflatoonian et al. (2010a) | Unmatched Cohort study | 500 pregnancies obtained after the transfer of fresh ET and 200 pregnancies after FET from March 2006 to March 2008. | Questionnaires filled by gynaecologists, paediatrics and women regarding perinatal and obstetric outcomes |
| 9/12 |
Aflatoonian et al. (2016) | Unmatched Cohort study |
| Data were collected from the hospital records. In addition, a telephone questionnaire consists of data on maternal and neonatal factors was administered by a trained nurse based on patients and their husbands’ information. |
| 9/12 |
Belva et al. (2008) | Unmatched Cohort Study |
| Data on pregnancies, deliveries and neonatal history was obtained by gynaecologists, paediatricians and double checked with parents, when child was 2 months old. | Mode of delivery and Duration of infertility was significantly different in the two groups. | 10/12 |
Healy et al. (2010) |
|
| Data were collected using record linkage of national databases | Includes first singleton birth only, delivered after 20 weeks. No data on demographic profile of the women in FET vs. fresh group | 10/12 |
Imudia et al. (2013) | Retrospective cohort study | Twenty women who underwent elective cryopreservation of all embryos with subsequent cryothaw ET and 32 similar women with elevated peak E2 during controlled ovarian hyperstimulation for IVF who underwent a fresh ET. | Data were collected from Medical records | Study adjusted for confounders (body mass index, antral follicle count, peak serum E2 level) Excluded peak serum E2 > 4500 pg/ml | 9/12 |
Ishihara et al. (2014) | Cohort | Registered from 2008 through 2010 undergoing single embryo transfer cycles. Only singleton ongoing pregnancies >21 weeks of gestation were included. | Japanese nationwide registry of assisted reproductive technology (ART) with mandatory reporting for all ART clinics in Japan. |
| 11/12 |
Kato et al. (2012) | Retrospective Cohort | Single-centre retrospective cohort study of 6623 consecutive delivered singletons following 29 944 single-embryo transfers. January 2006 and December 2008 | Two-part questionnaire filled by patient at the 20th pregnancy week and after delivery. | There was no difference in baseline characteristics in both group. | 11/12 |
Li et al. (2014) | Retrospective Cohort study | Reterospective population-based cohort study from Jan 2009 to Dec 2011 of autologous fresh and frozen cycles in Australia and New Zealand | ART treatment information and perinatal outcomes were obtained from the Australian and New Zealand Assisted Reproduction Database (ANZARD). | 9/12 | |
Liu et al. (2013) | Retrospective single centre analysis | Retrospective, single-centre study of children born after Day 3 embryo transfer from fresh, slow frozen or vitrified embryos during the period January 2006 to May 2011 | Data obtained via patient filled questionnaire at 12 weeks |
| 8/12 |
Maheshwari et al. (2016) | Retrospective analysis | Retrospective analysis of annonymized HFEA data | Data taken from HFEA database (which gets reported to HFEA by clinics as part of regulatory requirement) |
| 11/12 |
Opdahl et al. (2015) | Unmatched cohort study | Nationwide data from registries of Denmark, Norway and Sweden | Data obtained from health registries | Baseline characteristics for women having fresh or frozen embryo transfer were not compared. As fresh and frozen embryo transfers were one of multiple comparisons | 11/12 |
Pelkonen et al. (2010) | Unmatched Cohort study (1995–2006) |
| Data taken from Finnish Medical Birth Register |
| 11.5/12 |
Pelkonen et al. (2014) | Register based cohort study |
| Linkage of fertility, birth and congenital anomalies registries | There was a higher proportion of nulliparous women in fresh ET group | 11/12 |
Pereira et al. (2016) | Retrospective review | Consecutive live deliveries from all patients who began IVF cycles at the single centre between 1 January 2010 and 30 September 2013. | Data collected by retrospective review of patients charts | Patients were of similar age, BMI, infertility diagnosis, endometrial thickness and there was no difference in the grading of blastocysts. | 11/12 |
Pinborg et al. (2010) | Matched Cohort study |
| Danish IVF and Danish Birth Register |
| 11/12 |
Pinborg et al. (2014) | The national register-based controlled cohort study |
| Registry data |
| 11/12 |
Rallis and Tremellen (2013) | Retrospective review |
| Clinic based data, case records, database | Basic demographic data other than age group was not available confounding factors for preterm birth such as previous pregnancy outcomes were not available. | 10/12 |
Roy et al. (2014) | Retrospective cohort | Single centre Assisted Reproduction clinic between March 2010 and November 2011 | Private IVF Clinic database | Data for the fresh group were restricted to the patients with three or fewer stimulation cycles who had single blastocyst transferred and one blastocyst cryopreserved. | 10/12 |
Shapiro et al. (2016) | Post hoc analysis of two RCT | Two RCTs from same centre one on hyper responders and one on normal responders | Birth weight outcome; post hoc analysis | Data obtained through personal communication | NA |
Shih et al. (2008) | Matched cohort with women acting as their own reference | Comparison groups: Frozen versus Fresh IVF/ICSI | Neonatal perinatal statistics unit Austraila | All pregnancies after 20 weeks were recorded. Fresh IVF/ICSI conception could be first/second one | 10/12 |
Shi et al. (2012) | Retrospective data | Single centre Assisted Reproduction clinic | The outcome data were obtained from a postal questionnaire of parents after delivery |
| 8/12 |
Wada et al. (1994) | Unmatched cohort |
| Data were collected from medical records | 7/12 | |
Wang et al. (2005) | Unmatched cohort study | Infants conceived through ART Procedures and born in Australia during 1996–2000 | The study used data from two national collections. Assisted conception data collection & Australian national perinatal data collection | Fresh and frozen pregnancies were subgroup analysis. Hence not matched for the confounders. | 9/12 |
Wennerholm et al. (1997) | Matched cohort |
| Data were collected after medical records review | Controls were matched for age and parity | 10.5/11 |
Wennerholm et al. (2013) | Reterospective Matched cohort study |
| Data on perinatal outcomes were obtained by linkage to the national Medical Birth Registries |
| 11/12 |
Wikland et al. (2010) | Unmatched Cohort Study |
| Data for obstetric and perinatal complications was collected from maternity records |
| 11/12 |
Study ID . | Design of study . | Population . | Method of data collection . | Risk of bias . | Scoring . |
---|---|---|---|---|---|
Aflatoonian et al. (2010a) | Unmatched Cohort study | 500 pregnancies obtained after the transfer of fresh ET and 200 pregnancies after FET from March 2006 to March 2008. | Questionnaires filled by gynaecologists, paediatrics and women regarding perinatal and obstetric outcomes |
| 9/12 |
Aflatoonian et al. (2016) | Unmatched Cohort study |
| Data were collected from the hospital records. In addition, a telephone questionnaire consists of data on maternal and neonatal factors was administered by a trained nurse based on patients and their husbands’ information. |
| 9/12 |
Belva et al. (2008) | Unmatched Cohort Study |
| Data on pregnancies, deliveries and neonatal history was obtained by gynaecologists, paediatricians and double checked with parents, when child was 2 months old. | Mode of delivery and Duration of infertility was significantly different in the two groups. | 10/12 |
Healy et al. (2010) |
|
| Data were collected using record linkage of national databases | Includes first singleton birth only, delivered after 20 weeks. No data on demographic profile of the women in FET vs. fresh group | 10/12 |
Imudia et al. (2013) | Retrospective cohort study | Twenty women who underwent elective cryopreservation of all embryos with subsequent cryothaw ET and 32 similar women with elevated peak E2 during controlled ovarian hyperstimulation for IVF who underwent a fresh ET. | Data were collected from Medical records | Study adjusted for confounders (body mass index, antral follicle count, peak serum E2 level) Excluded peak serum E2 > 4500 pg/ml | 9/12 |
Ishihara et al. (2014) | Cohort | Registered from 2008 through 2010 undergoing single embryo transfer cycles. Only singleton ongoing pregnancies >21 weeks of gestation were included. | Japanese nationwide registry of assisted reproductive technology (ART) with mandatory reporting for all ART clinics in Japan. |
| 11/12 |
Kato et al. (2012) | Retrospective Cohort | Single-centre retrospective cohort study of 6623 consecutive delivered singletons following 29 944 single-embryo transfers. January 2006 and December 2008 | Two-part questionnaire filled by patient at the 20th pregnancy week and after delivery. | There was no difference in baseline characteristics in both group. | 11/12 |
Li et al. (2014) | Retrospective Cohort study | Reterospective population-based cohort study from Jan 2009 to Dec 2011 of autologous fresh and frozen cycles in Australia and New Zealand | ART treatment information and perinatal outcomes were obtained from the Australian and New Zealand Assisted Reproduction Database (ANZARD). | 9/12 | |
Liu et al. (2013) | Retrospective single centre analysis | Retrospective, single-centre study of children born after Day 3 embryo transfer from fresh, slow frozen or vitrified embryos during the period January 2006 to May 2011 | Data obtained via patient filled questionnaire at 12 weeks |
| 8/12 |
Maheshwari et al. (2016) | Retrospective analysis | Retrospective analysis of annonymized HFEA data | Data taken from HFEA database (which gets reported to HFEA by clinics as part of regulatory requirement) |
| 11/12 |
Opdahl et al. (2015) | Unmatched cohort study | Nationwide data from registries of Denmark, Norway and Sweden | Data obtained from health registries | Baseline characteristics for women having fresh or frozen embryo transfer were not compared. As fresh and frozen embryo transfers were one of multiple comparisons | 11/12 |
Pelkonen et al. (2010) | Unmatched Cohort study (1995–2006) |
| Data taken from Finnish Medical Birth Register |
| 11.5/12 |
Pelkonen et al. (2014) | Register based cohort study |
| Linkage of fertility, birth and congenital anomalies registries | There was a higher proportion of nulliparous women in fresh ET group | 11/12 |
Pereira et al. (2016) | Retrospective review | Consecutive live deliveries from all patients who began IVF cycles at the single centre between 1 January 2010 and 30 September 2013. | Data collected by retrospective review of patients charts | Patients were of similar age, BMI, infertility diagnosis, endometrial thickness and there was no difference in the grading of blastocysts. | 11/12 |
Pinborg et al. (2010) | Matched Cohort study |
| Danish IVF and Danish Birth Register |
| 11/12 |
Pinborg et al. (2014) | The national register-based controlled cohort study |
| Registry data |
| 11/12 |
Rallis and Tremellen (2013) | Retrospective review |
| Clinic based data, case records, database | Basic demographic data other than age group was not available confounding factors for preterm birth such as previous pregnancy outcomes were not available. | 10/12 |
Roy et al. (2014) | Retrospective cohort | Single centre Assisted Reproduction clinic between March 2010 and November 2011 | Private IVF Clinic database | Data for the fresh group were restricted to the patients with three or fewer stimulation cycles who had single blastocyst transferred and one blastocyst cryopreserved. | 10/12 |
Shapiro et al. (2016) | Post hoc analysis of two RCT | Two RCTs from same centre one on hyper responders and one on normal responders | Birth weight outcome; post hoc analysis | Data obtained through personal communication | NA |
Shih et al. (2008) | Matched cohort with women acting as their own reference | Comparison groups: Frozen versus Fresh IVF/ICSI | Neonatal perinatal statistics unit Austraila | All pregnancies after 20 weeks were recorded. Fresh IVF/ICSI conception could be first/second one | 10/12 |
Shi et al. (2012) | Retrospective data | Single centre Assisted Reproduction clinic | The outcome data were obtained from a postal questionnaire of parents after delivery |
| 8/12 |
Wada et al. (1994) | Unmatched cohort |
| Data were collected from medical records | 7/12 | |
Wang et al. (2005) | Unmatched cohort study | Infants conceived through ART Procedures and born in Australia during 1996–2000 | The study used data from two national collections. Assisted conception data collection & Australian national perinatal data collection | Fresh and frozen pregnancies were subgroup analysis. Hence not matched for the confounders. | 9/12 |
Wennerholm et al. (1997) | Matched cohort |
| Data were collected after medical records review | Controls were matched for age and parity | 10.5/11 |
Wennerholm et al. (2013) | Reterospective Matched cohort study |
| Data on perinatal outcomes were obtained by linkage to the national Medical Birth Registries |
| 11/12 |
Wikland et al. (2010) | Unmatched Cohort Study |
| Data for obstetric and perinatal complications was collected from maternity records |
| 11/12 |
Study selection
Inclusion criteria included all published observational studies and randomised trials comparing obstetric and perinatal outcomes in singleton pregnancies following transfer of fresh and frozen embryos.
Exclusion criteria excluded studies if there was no comparator group, if obstetric and perinatal outcomes were not reported or if it was not possible to differentiate the outcomes for singletons. Case reports and case series were also excluded.
Outcome measures
The following outcome measures were included: small for gestational age (as defined by the authors of included studies), very preterm birth (delivery prior to 32 weeks), preterm birth (delivery prior to 37 weeks), low birth weight (<2500 g), very low birth weight (<1500 g), high birth weight (>4000 g), very high birth weight (>4500 g), large for gestational age (as defined by the authors of the included studies), antepartum haemorrhage (APH) (combination of placenta praevia, placental abruption and other bleeding), hypertensive disorders of pregnancy (including pregnancy induced hypertension, pre-eclampsia and eclampsia), congenital anomalies (major and minor), perinatal mortality (as defined by the authors of the included studies), and admission to a neonatal intensive care unit (NICU).
Assessment of heterogeneity
We assessed whether there was sufficient similarity between the eligible studies in their design and clinical characteristics to ensure that pooling was valid. I2 statistic was used to assess the impact of the heterogeneity on the meta-analysis. I2 > 50% was labelled as marked heterogeneity (Higgins et al., 2003).
Assessment of reporting biases
Funnel plots were constructed to test the small study effect where a statistically significant difference was obtained in outcome measure, if at least five studies reported that outcome. Egger’s regression test (Egger et al., 1997) was used to investigate whether the difference was due to publication or reporting bias.
Statistical analysis
For each outcome, data were extracted in 2 × 2 tables. When there was an outcome with no events in one of the groups, a correction factor of 0.5 was added to all cells in a 2 × 2 table in the calculation of risk ratio (Sweeting et al., 2004). The summary measures for each study were Risk Ratio/Relative Risk (RR) with 95% confidence intervals (CI). The ‘fresh embryo transfer’ group was considered as the reference group. The pooled estimates were obtained using both standard and cumulative meta-analysis. Although we analysed the data using both the fixed effect models and random effect models, results in the text are only reported from random effect models due to underlying heterogeneity in the studies. Cumulative meta-analyses (Lau et al., 1992) were carried to track the accumulation of evidence on the obstetrics and perinatal outcomes in singleton pregnancies subsequent to frozen embryo over the period of time. The statistical analyses were carried out using Stata MP version 14.
GRADE PRO software was used to generate the summary of finding table as well as quality of evidence.
Results
Results of the searches
The literature search yielded 126 citations. Of these, 106 were excluded after reading the title and the abstract. Full text versions of 20 articles were obtained, of which 16 were included, while another 10 publications were identified from hand searches of cross references and checking for advance access publications as well as articles in press. Hence, a total of 26 studies were included (Fig. 1). Studies from the same research group or region were carefully examined for any overlapping data. Authors were contacted if the information was unclear. Studies with overlapping data were excluded. Table I summarises details of all included studies, while excluded studies along with reasons for exclusion are listed in Table II.
Study . | Reason for exclusion . |
---|---|
Aytoz et al. (1999) | Data from Singleton and twins could not be separated |
Aflatoonian et al. (2010b) | No data on obstetric and perinatal outcomes |
Frydman et al. (1989) | There is no control group |
Henningsen et al (2011) | Overlapping data with Pinborg 2010 |
Kallen et al. (2005a) | 2 × 2 table cannot be made |
Kallen et al. (2005b) | Data for singleton cannot be separated |
Kansal Kalra et al. (2011) | Data on singletons cannot be separated |
Ku et al. (2012) | No obstetric and perinatal outcomes reported |
Wang et al. (2005) | Overlapping data with Shih 2008 |
Shapiro et al. (2011) | No data on obstetric and perinatal outcomes |
Chen et al. (2016) | Data from singleton and twins can’t be separated |
Takeshima et al. (2016) | Data for singletons cannot be separated to generate 2 × 2 table |
Wennerhol (2000) | Overlapping data from Wennerholm et al. (1997) |
Wennerholm et al. (2009) | Systematic review |
Study . | Reason for exclusion . |
---|---|
Aytoz et al. (1999) | Data from Singleton and twins could not be separated |
Aflatoonian et al. (2010b) | No data on obstetric and perinatal outcomes |
Frydman et al. (1989) | There is no control group |
Henningsen et al (2011) | Overlapping data with Pinborg 2010 |
Kallen et al. (2005a) | 2 × 2 table cannot be made |
Kallen et al. (2005b) | Data for singleton cannot be separated |
Kansal Kalra et al. (2011) | Data on singletons cannot be separated |
Ku et al. (2012) | No obstetric and perinatal outcomes reported |
Wang et al. (2005) | Overlapping data with Shih 2008 |
Shapiro et al. (2011) | No data on obstetric and perinatal outcomes |
Chen et al. (2016) | Data from singleton and twins can’t be separated |
Takeshima et al. (2016) | Data for singletons cannot be separated to generate 2 × 2 table |
Wennerhol (2000) | Overlapping data from Wennerholm et al. (1997) |
Wennerholm et al. (2009) | Systematic review |
Study . | Reason for exclusion . |
---|---|
Aytoz et al. (1999) | Data from Singleton and twins could not be separated |
Aflatoonian et al. (2010b) | No data on obstetric and perinatal outcomes |
Frydman et al. (1989) | There is no control group |
Henningsen et al (2011) | Overlapping data with Pinborg 2010 |
Kallen et al. (2005a) | 2 × 2 table cannot be made |
Kallen et al. (2005b) | Data for singleton cannot be separated |
Kansal Kalra et al. (2011) | Data on singletons cannot be separated |
Ku et al. (2012) | No obstetric and perinatal outcomes reported |
Wang et al. (2005) | Overlapping data with Shih 2008 |
Shapiro et al. (2011) | No data on obstetric and perinatal outcomes |
Chen et al. (2016) | Data from singleton and twins can’t be separated |
Takeshima et al. (2016) | Data for singletons cannot be separated to generate 2 × 2 table |
Wennerhol (2000) | Overlapping data from Wennerholm et al. (1997) |
Wennerholm et al. (2009) | Systematic review |
Study . | Reason for exclusion . |
---|---|
Aytoz et al. (1999) | Data from Singleton and twins could not be separated |
Aflatoonian et al. (2010b) | No data on obstetric and perinatal outcomes |
Frydman et al. (1989) | There is no control group |
Henningsen et al (2011) | Overlapping data with Pinborg 2010 |
Kallen et al. (2005a) | 2 × 2 table cannot be made |
Kallen et al. (2005b) | Data for singleton cannot be separated |
Kansal Kalra et al. (2011) | Data on singletons cannot be separated |
Ku et al. (2012) | No obstetric and perinatal outcomes reported |
Wang et al. (2005) | Overlapping data with Shih 2008 |
Shapiro et al. (2011) | No data on obstetric and perinatal outcomes |
Chen et al. (2016) | Data from singleton and twins can’t be separated |
Takeshima et al. (2016) | Data for singletons cannot be separated to generate 2 × 2 table |
Wennerhol (2000) | Overlapping data from Wennerholm et al. (1997) |
Wennerholm et al. (2009) | Systematic review |
Included studies
Methodology of included studies
Of the 26 included studies, one was a post hoc analysis of two RCTs (Shapiro et al., 2016), while the rest were cohort studies. Most (n = 21) were unmatched cohort studies. A high proportion of studies (n = 16) scored high (≥10) on the CASP scoring system. Data were obtained from databases and data linkage of routinely collected data and case notes except in three studies where clinical information was reported only by questionnaires filled by patients (Kato et al., 2012; Shi et al., 2012; Liu et al., 2013).
Population in the included studies
Although all studies were based on outcomes of pregnancies conceived through IVF/ICSI using fresh or frozen embryos, they varied in terms of the duration of pregnancy at which women were included: all clinical pregnancies (Belva et al., 2008; Imudia et al., 2013), all births beyond 20 weeks (Aflatoonian et al., 2010a, 2016; Wada et al., 1994; Wang et al., 2005; Shih et al., 2008; Healy et al., 2010; Rallis and Tremellen, 2013; Li et al., 2014), beyond 21 weeks (Ishihara et al., 2014); beyond 22 weeks (Pelkonen et al., 2010, 2014; Kato et al., 2012; Wennerholm et al., 2013; Opdahl et al., 2015) and beyond 28 weeks (Wennerholm et al., 1997; Wikland et al., 2010; Liu et al., 2013) and only live deliveries (Pereira et al., 2016).
Three studies (Wang et al., 2005; Healy et al., 2010; Opdahl et al., 2015) provided no information on the demographic profile of women who had fresh or frozen embryo transfer, as this comparison was part of a subgroup analysis. The characteristics in the two groups were similar except in: Pinborg et al., 2010 and 2014 (where data were adjusted at primary analysis), Pelkonen et al., 2010 and Belva et al., 2008 (where mothers in the frozen embryo transfer group were older), and Pelkonen et al., 2014 (where a higher proportion of nulliparous women were in the fresh embryo transfer group). No details on other confounders, such as parity, smoking, duration of infertility and pre-existing medical diseases, were available.
Exposure in the included studies
Studies varied in terms of when and how embryos were frozen and the methods used for endometrial preparation prior to embryo transfer after thawing. Methods of cryopreservation and the developmental stage at which embryos were frozen also varied within some studies especially in registry based datasets. Embryos were frozen either at Day 2/3, i.e. cleavage stage (Aflatoonian et al., 2010a,b, 2016; Wada et al., 1994; Pelkonen et al., 2010; Shi et al., 2012; Imudia et al., 2013; Liu et al., 2013) or Day 5/6, i.e. blastocyst stage (Li et al., 2014; Roy et al., 2014; Pereira et al., 2016) or both (Belva et al., 2008; Kato et al., 2012) using either vitrification (Aflatoonian et al., 2010a,b, 2016; Kato et al., 2012; Shi et al., 2012; Pereira et al., 2016) or slow freezing (Wada et al., 1994; Belva et al., 2008; Pelkonen et al., 2010; Imudia et al., 2013) or both techniques (Liu et al., 2013; Li et al., 2014).
Frozen thawed embryos were transferred in women following additional hormones to prepare the endometrium (Aflatoonian et al., 2010a,b, 2016; Imudia et al., 2013) or in natural unstimulated cycles (Belva et al., 2008; Rallis and Tremellen, 2013).
Outcomes
Pooled data for outcome measures were as follows.
Small for gestational age
Ten studies (n = 53 418 vs. 89 044 pregnancies following frozen vs. fresh cycles) have reported on the outcome of small for gestational age. This was defined as birth weight less than 2 standard deviations of the mean for that gestation (Pelkonen et al., 2010; Wennerholm et al., 2013; Ishihara et al., 2014; Pinborg et al., 2014) or less than the 10th centile (Kato et al., 2012; Li et al., 2014; Aflatoonian et al., 2016) or birth weight less than 22% of expected mean birth weight according to gestational age in a reference population (Wikland et al., 2010).
The risk of having a small for gestational age baby was significantly less in singleton pregnancies subsequent to frozen thawed embryo transfer as compared to those after fresh embryo transfer (RR 0.61; 95% CI 0.56–0.67) (Fig. 2a). There was minimal heterogeneity amongst the studies (I2 = 33.8%). The funnel plot did not suggest any publication bias (P = 0.77).
A statistically significant reduction in small for gestational age babies was first observed in 2010 after first publication (RR 0.49; 95% CI 0.33–0.75). Although subsequent studies have increased the precision of the point estimate, no substantive change has occurred in the direction or magnitude of the treatment effect (Fig. 2b).
Low birth weight (birth weight <2500 g)
Meta-analysis of the data based on twenty studies (n = 78 250 vs. 201 794 pregnancies following frozen vs. fresh cycles) showed that the risk of having a baby with birth weight less than 2500 g is significantly less (Fig. 3a) in singleton pregnancies following frozen thawed embryos, when compared to those following fresh embryos (RR 0.72; 95% CI 0.67–0.77). There was moderate heterogeneity (I2 = 55%) amongst the studies. Funnel plot did not reveal any publication bias (P = 0.15).
The evidence that frozen embryo transfer reduces the risk of low birth weight babies has been available since 1997 (Fig. 3b). Although subsequent studies have increased the precision of the point estimate, no substantive change has occurred in the direction or magnitude of the treatment effect.
Very low birth weight (birth weight <1500 g)
Thirteen studies (n = 71 218 vs. 189 008 pregnancies following frozen vs. fresh embryo transfer) have reported proportion of deliveries with birth weight less than 1500 g. The relative risk and 95% CI of having a baby with birth weight less than 1500g was lower (0.76; 0.69–0.82) following singleton pregnancies subsequent to frozen thawed embryo transfer as compared to those following fresh embryo transfer (Fig. 4a). There was no heterogeneity (I2 = 0%) amongst the studies. The funnel plot does not suggest publication bias (P = 0.16).
Cumulative meta-analysis shows (Fig. 4b) that this evidence has been available since 2012. Although subsequent studies have increased the precision of the point estimate, no substantive change has occurred in the direction or magnitude of the treatment effect.
Large for gestational age
Seven studies (n = 51 719 vs. 86 544 pregnancies following frozen vs. fresh cycles) have reported on outcome of large for gestational age. This was defined as birth weight greater than 2 standard deviations of the mean for that gestation (Pelkonen et al., 2010; Wennerholm et al., 2013; Ishihara et al., 2014; Pinborg et al., 2014) or more than the 90th centile (Kato et al., 2012; Li et al., 2014) or birth weight more than 22% of expected mean birth weight according to gestational age in a reference population (Wikland et al., 2010).
The relative risk and 95% CI of having a large for gestational age baby was higher (1.54; 1.48–1.61) in singleton pregnancies subsequent to frozen thawed embryo transfer as compared to those conceived following fresh embryo transfer (Fig. 5a). There was minimal heterogeneity amongst the studies (I2 = 11%). The funnel plot suggests no publication bias (P = 0.73).
Cumulative meta-analysis suggests that this evidence has been available since 2012 with further precision of point estimate provided by additional data, without changing the direction and magnitude of the effect (Fig. 5b).
High birth weight (birth weight >4000 g)
Three studies reported the outcome of birth weight more than 4000 g (n = 48 026 vs. 113 241 pregnancies following frozen vs. fresh embryo transfer). There was an increased risk (Fig. 6a) of having a baby with birth weight more than 4000 g in singleton pregnancies as a result of frozen embryo transfer when compared to those subsequent to fresh embryo transfer (RR 1.85; 95% CI 1.46–2.33).
A statistically significant effect was first observed in 2014 after first publication (RR 1.95; 95% CI 1.29–2.95). Additions of data from a subsequent large study has increased the precision of the point estimate, while no change has occurred in the direction or magnitude of the treatment effect (Fig. 6b).
Very high birth weight (birth weight >4500 g)
Four studies have reported the outcome of birth weight more than 4500 g (n = 55 313 vs. 164 542 pregnancies following frozen vs. fresh embryo transfer). There was an increased risk (Fig. 7a) of having a baby with birth weight more than 4500 g in singleton pregnancies as a result of frozen embryo transfer when compared to those subsequent to fresh embryo transfer (RR 1.86; 95% CI 1.58–2.19).
There was significant heterogeneity (I2 = 67%). Cumulative meta-analysis (Fig. 7b) suggests that significantly increased risk of very high birth weight babies was first reported in 2013 with no change in direction, estimate or precision by adding further data over the years.
Preterm delivery (delivery at less than 37 weeks)
Twenty studies (n = 78 386 vs. 202 236 pregnancies following frozen vs. fresh cycles) reported the proportion of deliveries occurring at less than 37 weeks of gestation. The definition of preterm labour or delivery was delivery prior to 37 weeks in all studies. There are no data on how many of them were spontaneous or induced preterm labour.
The relative risk of having a delivery at less than 37 weeks was reduced (RR 0.90; 95% CI 0.84–0.97) in singleton pregnancies following frozen thawed embryo transfer, when compared to those after fresh embryo transfers (Fig. 8a). There was marked heterogeneity (I2 = 65%) amongst the studies. The funnel plot did not reveal any publication bias (P = 0.73).
Cumulative meta-analysis (Fig. 8b) suggests that the evidence favouring frozen embryo transfer in terms of a reduced risk of preterm delivery was first available in 2005. In 2013, the addition of further data showed that there was no difference in the risk of preterm delivery between the two groups. However, new results from studies published after 2013 have re-confirmed the reduced risk of preterm delivery. Addition of several studies from 2014 to 2016 have increased the precision of our estimate without affecting the direction or magnitude of the treatment effect.
Very preterm birth (delivery at less than 32 weeks)
Twelve studies (n = 68 927 vs. 184 377 pregnancies following frozen vs. fresh embryo transfer) reported on deliveries prior to 32 weeks. The relative risk and 95% CI of a delivery at less than 32 weeks was lower (0.85; 0.74–0.97) in singleton pregnancies following frozen embryo thawed transfer when compared to those after fresh embryo transfer (Fig. 9a). There was moderate heterogeneity (I2 = 38.6%) amongst the studies. We could not differentiate between iatrogenic and spontaneous preterm delivery. The funnel plot was suggestive of a degree of publication bias (P = 0.04).
Cumulative meta-analysis suggests that the evidence in support of a reduced risk of very preterm delivery in singleton pregnancies after thawed frozen embryo transfer has only been available since 2016 (Fig. 9b).
Antepartum haemorrhage
Five studies were included in the meta-analysis (n = 36 911 vs. 26 244 pregnancies after frozen vs. fresh embryo transfer). Hayley et al. (2010) reported a comparison between fresh embryo transfer (stimulated) versus frozen embryo transfer (natural cycles only). They reported APH, postpartum haemorrhage and placenta praevia as well as accreta separately. Shi et al. (2012) reported all APH together, while Ishihara et al. (2014), Liu et al. (2013) and Pelkonen et al. (2010) reported placenta praevia, abruption and accreta separately.
There was no difference in risk of APH in singleton pregnancies following frozen thawed embryo transfer when compared to those after fresh embryos (RR 0.82; 95% CI 0.66–1.03). There was moderate heterogeneity (67.6%) amongst the studies (Fig. 10a).
Cumulative meta-analysis (Fig. 10b) suggest that data available by 2010–2013 suggested that the risk of APH was lower in singleton pregnancies in women who underwent frozen embryo transfer; however, by 2014 this outcome had changed to no difference following the accrual of fresh data. No studies after 2014 have reported this outcome.
Admission to neonatal intensive care unit
Five studies reported the outcome of admission to NICU (n = 3703 vs. 15 862 pregnancies after frozen vs. fresh embryo transfer). The length and the reasons for NICU admission were not specified. There was no increase in the risk of admission to NICU (RR 0.99; 95% CI 0.84–1.18) in pregnancies following frozen embryos (Fig. 11a). There was marked heterogeneity amongst the studies (I2 = 54%).
Cumulative meta-analysis for admission to neonatal unit showed no clear trend regarding the effect on singleton pregnancies as a result of frozen embryo transfer. This has not changed over the years with accrual of fresh data (Fig. 11b).
Congenital anomalies
Only six studies (n = 25 789 vs. 107 692 pregnancies following frozen vs. fresh embryo transfer) reported congenital anomalies (one matched cohort study). Both major and minor anomalies were pooled together. The relative risk of having a congenital anomaly was 1.01 (95% CI 0.87–1.16) in pregnancies following frozen thawed embryos as compared to fresh embryos (Fig. 12a). There was minimal heterogeneity (I2 = 28%) amongst the studies.
Cumulative meta-analysis for congenital anomalies showed no clear trend regarding effect on pregnancies as a result of frozen embryo transfer. This has been consistent over time despite accrual of fresh data (Fig. 12b).
Perinatal mortality
Twelve studies (n = 25 203 vs. 77 280 pregnancies following frozen vs. fresh embryo transfer) reported the outcome of perinatal mortality. Still birth and perinatal mortality are presented together in this report. Some studies reported only neonatal death (Shi et al., 2012; Roy et al., 2014). Of those who reported perinatal mortality, there was a variation in definition: death of child with a gestational age of more than 20 weeks or up to Day 28 of birth (Aflatoonian et al., 2010a,b, 2016; Pinborg et al., 2010); death occurring after the 24th week of gestation and during the first week of life; death after 22 weeks of gestation up to the first 7 days of life (Kato et al., 2012); death after 28 weeks of gestation up to the first 7 days of life (Pelkonen et al., 2010, Wikland et al., 2010, Liu et al., 2013, Wennerholm et al., 2013, Li et al., 2014); or death after 20 weeks, later terminations and all neonatal deaths (Shih et al., 2008).
There was no difference in perinatal mortality (RR 0.92; 95% CI 0.78–1.08) in singleton pregnancies after frozen thawed embryo transfers, when compared to those after fresh embryos (Fig. 13a). There was no heterogeneity amongst the studies (I2 = 0.80%). There was no publication bias (P = 0.41).
Cumulative meta-analysis for perinatal mortality showed no clear trend regarding effect on pregnancies as a result of frozen embryo transfer despite addition of fresh data over time (Fig. 13b).
Hypertensive disorders of pregnancy
Five studies reported the outcome of hypertensive disorders of pregnancy (n = 39 501 vs. 59 155 pregnancies following frozen vs. fresh embryo transfer). The relative risk of hypertensive disorders of pregnancy was higher in the frozen embryo transfer group (Fig. 14a) (RR 1.29; 95% CI 1.07–1.56). There was moderate heterogeneity (I2 = 66%).
Cumulative meta-analysis suggests that the evidence in support of an increased risk of hypertensive disorders in singleton pregnancies after thawed frozen embryo transfer has only become available since 2015 (Fig. 14b).
Discussion
Principal findings
Singleton pregnancies following frozen embryo transfer face a reduced risk of preterm birth, small for gestational age and low birth weight babies but a higher risk of large for gestational age babies as well as hypertensive disorders of pregnancies. Although more recent studies have increased the precision of the point estimate, no substantive change has occurred in the direction or magnitude of the treatment effect for these outcomes over time.
Strengths
This is a definitive, updated date systematic review on a key topic in assisted reproduction, at a time when frozen embryo transfer rates are rising sharply. In addition to conventional meta-analysis, we were also able to present a cumulative meta-analysis to assess temporal trends which might be influenced by improvements in freezing and thawing techniques over the years. The consistency in direction and magnitude of the treatment effect for the key outcomes confirms the validity of the published data.
Limitations
As there are no randomised controlled trials (except Shapiro et al., 2016, where birth weight was done as a post hoc analysis, with data that could be included obtained by personal communication) reporting perinatal outcomes in singleton pregnancies, this review is limited to data from observational studies. Hence the evidence was graded as low despite large numbers (Supplementary Table SII). There were variations in the studies whose data have been complied together not only in design but population, interventions (method of freezing and regimens in replacement cycles) and ascertainment of outcomes (Table I) We were also unable to adjust for confounders such as age, smoking, parity, duration of infertility and pre-existing medical illness. Without individual patient data, we were unable to determine if the risks are different for embryos after slow freezing or vitrification, whether embryos were frozen at cleavage or blastocyst stage of development and whether protocols were used for endometrial preparation. Although our cumulative meta-analysis was stratified by year of publication, a paper in 2016 contained data from 1997 (Maheshwari et al., 2016), hence the true effect of changes in freezing techniques over time cannot be fully captured.
We have combined both major and minor foetal abnormalities together because these were not available as separate data for most of the studies that did report of this. It is also acknowledged that authors might use different classification systems for foetal abnormalities, and that some studies may have included terminations in these abnormalities while others might not have. Again this data was not available in the studies included.
Comparison with other studies
The findings of low and high birth weight are consistent with the published literature and our previous systematic review (Maheshwari et al., 2012). The incidence of preterm delivery was reported to be lower after frozen embryo transfer in this as well as in the previous review. However, a recent randomised trial (Chen et al., 2016) did not find any difference in preterm birth rates, and neither did an analysis of a large national UK dataset (Maheshwari et al., 2016). Addition of results from this large dataset (Maheshwari et al., 2016) did not change the direction and magnitude of the effect for key outcomes in the cumulative meta-analysis. This provides a degree of confidence in the reliability of the existing data for the outcomes of birth weight and preterm delivery. The increased risk of hypertensive disorders of pregnancy in pregnancies following frozen embryo transfer in this report is similar to the findings in large randomised controlled trial (Chen et al., 2016).
Outcomes of APH, congenital anomalies, perinatal mortality, and admission to neonatal units are similar in pregnancies conceived from fresh and frozen embryos. As these outcomes have not been reported by all studies, the overall numbers are much lower. There is a possibility that addition of further data may change the current estimate of risk, especially for rarer outcomes such as perinatal mortality and congenital anomalies, where the number of observations is low.
Explanation of results
Hormonal stimulation of the ovaries in IVF causes a state of hyperestrogeneism at a time when fresh embryos are transferred. It has been hypothesised that this leads to abnormal endometrial angiogenesis leading to reduced implantation as well as abnormal placentation. This can account for the findings of small for gestational age babies, preterm deliveries and low birth weight babies. The uterine environment in a frozen replacement cycle is a more natural uterine environment as the effect of ovarian stimulation tends to worn off by the time point when embryos are replaced (Amor et al., 2009; Healy et al., 2010; Kansal Kalra et al., 2011). However, there is as yet no clear explanation for the increased chance of large for gestational age births. It is possible that higher implantation potential leads to better placentation and overgrowth of the foetus. Birth order, which is higher in babies, conceived from frozen thawed embryos, may play a role, but this has been challenged by the fact that the difference has persisted after adjustment for parity in various studies (Pinborg et al., 2014; Maheshwari et al., 2016). It has also been suggested that the freezing and thawing procedures may play an independent role in the growth potential of the foetus due to epigenetic alterations at the early embryonic stages (Pinborg et al., 2014). There was also no obvious biological explanation for the increase in hypertensive disorders.
Implications for clinical practice
Data from this review provides reassurance for embryo cryopreservation programmes in IVF, while, at the same time, suggesting a need for caution due to the higher risk of large for gestational age babies as well as the increased risk of hypertension in pregnancy. This is especially relevant as the threshold for freezing is falling and increasing numbers of embryos are being electively frozen and reserved for deferred transfer. In fact, in some centres, a ‘freeze all’ policy followed by thawed frozen embryo transfer has become the norm. It is to be remembered that both small for gestational age and large for gestational age has implications for health and diseases later in life. Hence, the routine use of a freeze all strategy may have long-term implications as well. Moreover all the evidence has been graded as low quality (Supplementary Table SII) as per GRADE matrix, primarily due to observational data.
We recommend that, on the basis of current evidence, elective freezing of all embryos should only be performed when there is a definite clinical indication or in the context of a clinical trial.
Implications for research
There have been a number of observational studies published over years to evaluate obstetric and perinatal outcomes in singleton pregnancies following thawed frozen embryo transfer. As is clear from the summary table (Table III), the data for birth weight and preterm delivery has reached saturation to the extent that even large datasets are not able to shift the magnitude and direction of the effect. Replication of data from different databases, geographical areas and populations, have proved the validity of the findings. Therefore, we do not feel that more data from observational studies are needed for the outcomes of preterm delivery and birth weight. Due to observational data, the quality of evidence has been graded as low, despite large numbers (Supplementary Table SII). This will not be altered by adding more observational data.
Risk of outcome . | Evidence . | Evidence available by year . | No further change in precision, magnitude or direction . | More observational data needed . |
---|---|---|---|---|
Small for gestational age | Lower in Frozen embryo transfer | 2010 | 2014 | No |
Low birth weight | Lower in Frozen embryo transfer | 1997 | 2014 | No |
Very low birth weight | Lower in Frozen embryo transfer | 2013 | 2016 | No |
Large for gestational age | Higher in Frozen embryo transfer | 2010 | 2014 | No |
High birth weight | Higher in Frozen embryo transfer | 2014 | 2016 | No |
Very high birth weight | Higher in Frozen embryo transfer | 2013 | 2014 | No |
Preterm delivery | Lower in Frozen embryo transfer | 2005 | 2014 | No |
Very preterm delivery | Lower in Frozen embryo transfer | 2016 | 2016 | No |
Antepartum haemorrhage | No difference | 2010 | 2014 | Yes |
Admission to NICU | No difference | 2012 | 2013 | Yes |
Congenital anomalies | No difference | 2014 | 2016 | Yes |
Perinatal mortality | No difference | 2014 | 2014 | Yes |
Hypertensive disorders of pregnancy | Higher in Frozen embryo transfer | 2015 | 2015 | Yes |
Risk of outcome . | Evidence . | Evidence available by year . | No further change in precision, magnitude or direction . | More observational data needed . |
---|---|---|---|---|
Small for gestational age | Lower in Frozen embryo transfer | 2010 | 2014 | No |
Low birth weight | Lower in Frozen embryo transfer | 1997 | 2014 | No |
Very low birth weight | Lower in Frozen embryo transfer | 2013 | 2016 | No |
Large for gestational age | Higher in Frozen embryo transfer | 2010 | 2014 | No |
High birth weight | Higher in Frozen embryo transfer | 2014 | 2016 | No |
Very high birth weight | Higher in Frozen embryo transfer | 2013 | 2014 | No |
Preterm delivery | Lower in Frozen embryo transfer | 2005 | 2014 | No |
Very preterm delivery | Lower in Frozen embryo transfer | 2016 | 2016 | No |
Antepartum haemorrhage | No difference | 2010 | 2014 | Yes |
Admission to NICU | No difference | 2012 | 2013 | Yes |
Congenital anomalies | No difference | 2014 | 2016 | Yes |
Perinatal mortality | No difference | 2014 | 2014 | Yes |
Hypertensive disorders of pregnancy | Higher in Frozen embryo transfer | 2015 | 2015 | Yes |
Risk of outcome . | Evidence . | Evidence available by year . | No further change in precision, magnitude or direction . | More observational data needed . |
---|---|---|---|---|
Small for gestational age | Lower in Frozen embryo transfer | 2010 | 2014 | No |
Low birth weight | Lower in Frozen embryo transfer | 1997 | 2014 | No |
Very low birth weight | Lower in Frozen embryo transfer | 2013 | 2016 | No |
Large for gestational age | Higher in Frozen embryo transfer | 2010 | 2014 | No |
High birth weight | Higher in Frozen embryo transfer | 2014 | 2016 | No |
Very high birth weight | Higher in Frozen embryo transfer | 2013 | 2014 | No |
Preterm delivery | Lower in Frozen embryo transfer | 2005 | 2014 | No |
Very preterm delivery | Lower in Frozen embryo transfer | 2016 | 2016 | No |
Antepartum haemorrhage | No difference | 2010 | 2014 | Yes |
Admission to NICU | No difference | 2012 | 2013 | Yes |
Congenital anomalies | No difference | 2014 | 2016 | Yes |
Perinatal mortality | No difference | 2014 | 2014 | Yes |
Hypertensive disorders of pregnancy | Higher in Frozen embryo transfer | 2015 | 2015 | Yes |
Risk of outcome . | Evidence . | Evidence available by year . | No further change in precision, magnitude or direction . | More observational data needed . |
---|---|---|---|---|
Small for gestational age | Lower in Frozen embryo transfer | 2010 | 2014 | No |
Low birth weight | Lower in Frozen embryo transfer | 1997 | 2014 | No |
Very low birth weight | Lower in Frozen embryo transfer | 2013 | 2016 | No |
Large for gestational age | Higher in Frozen embryo transfer | 2010 | 2014 | No |
High birth weight | Higher in Frozen embryo transfer | 2014 | 2016 | No |
Very high birth weight | Higher in Frozen embryo transfer | 2013 | 2014 | No |
Preterm delivery | Lower in Frozen embryo transfer | 2005 | 2014 | No |
Very preterm delivery | Lower in Frozen embryo transfer | 2016 | 2016 | No |
Antepartum haemorrhage | No difference | 2010 | 2014 | Yes |
Admission to NICU | No difference | 2012 | 2013 | Yes |
Congenital anomalies | No difference | 2014 | 2016 | Yes |
Perinatal mortality | No difference | 2014 | 2014 | Yes |
Hypertensive disorders of pregnancy | Higher in Frozen embryo transfer | 2015 | 2015 | Yes |
For other outcomes, especially rarer outcomes (neonatal death, congenital anomalies), it is important that an IPD-MA (individual patient data meta-analysis) is done from registries across the world. This will help for example in the analysis of major and minor congenital anomalies separately. It will also help in doing subgroup analysis for a specific group of patients, which is not possible in current report.
Although an IPDMA of registry data will be ideal this will not be without considerable investment and collaboration. There will be difficulties of data transfer due to local governances as well as the format (all data are in different format and collect different variables).
As the threshold for freezing has fallen, some clinics are choosing to opt for ‘freeze all’ programmes for an increasing number of IVF treatments in preference to the conventional policy of elective fresh embryo transfer. While the data generally provides reassurance for the safety of thawed frozen embryo transfers, there are some lingering concerns related to the risk of large for gestational age babies. This has created a state of equipoise which makes this an ideal time to conduct randomised controlled trials comparing an elective ‘freeze all’ policy with usual care, in terms of clinical and cost effectiveness.
Across the world, a number of trials with live birth as the primary outcome are either ongoing or have recently been completed (ACTRN 12616000643471; NCT01841528; NCT02746562; NCT02570386; NTR3187; ISCTRN- 61225414; NCT00963625; NCT00963079; NCT02471573; NCT01954758; ChiCTR-IOR-14005406). Follow up of offspring from these trials provides an opportunity to minimise bias in any future comparison of pregnancy outcomes such as preterm delivery, low and high birth weight, while an individual patient data meta-analysis approach permits outcomes in clinically relevant subgroups (e.g. older vs. younger women) to be compared.
Further mechanistic studies are needed to identify the biological reason for an increase in hypertensive disorders in pregnancies subsequent to frozen embryo transfer.
Conclusion
This systematic review confirms that singleton babies conceived by frozen embryo transfers are at lower risk of preterm delivery, small for gestational age and low birth weight. The direction and magnitude of effect for these outcomes have remained virtually unchanged over time while the degree of precision has improved with the addition of data from newer studies. Our results also show that frozen embryo transfer is associated with an increased risk of hypertensive disorders of pregnancy and large for gestational age in singleton babies.
Although replication of the research by several groups has added to the external validity of the results, the data from our cumulative meta-analyses suggest that further analyses of observational data from published studies are unlikely to change them. Given the current challenges around research funding, resources should be concentrated on following up pregnancy outcomes of relevant randomised trials and an IPD-MA of registry data.
Supplementary data
Supplementary data are available at Human Reproduction Update online.
Acknowledgements
The authors would like to thank authors of primary studies who responded to queries and provided additional data required to complete this review (Professors Anja Pinborg, Forest Gardner and Bruce Shapiro).
Authors’ roles
A.M. and S.B. conceived the manuscript. A.M. and S.P. did searches, quality assessment and data collation. E.A.R. did all the statistical analysis. M.H., S.B. and A.S. provided intellectual input from the protocol stage right through all versions of the manuscript. All authors contributed to the final version of the manuscript.
Funding
No external funding was sought in preparing this manuscript
Conflict of interest
A.M. and S.B. are co-applicants on the HTA/ NIHR grant, UK (ISCTRN-61225414) for E-Freeze Trial which is a randomised controlled trial comparing elective freezing of embryos with current policy of fresh embryo transfer. Otherwise the authors have no conflict of interest.