Approximately one million couples receive in vitro fertilisation (IVF) treatment every year worldwide [
]. Luteal phase support (LPS) has routinely been applied as part of this treatment. The use of agonistic or antagonistic gonadotropin-releasing hormone (GnRH) protocols in stimulated IVF/intracytoplasmic sperm injection (ICSI) cycles cause disruptions to the luteal phase, leading to inadequate development of the endometrium and asynchrony between endometrial receptiveness and embryo transfer. The most plausible cause of this condition is the development of multiple follicles upon ovarian stimulation, which results in superphysiological steroid concentrations and consequent inhibition of luteinising hormone (LH) secretion by the pituitary via negative feedback at the level of the hypothalamic-pituitary axis [
]. Despite the rapid recovery of the pituitary in GnRH-antagonist protocols, luteolysisis also prematurely induced after GnRH-antagonist co-treatment, resulting in a significant reduction in luteal phase length and a compromised reproductive outcome. For this reason, LPS remains mandatory in GnRH antagonist protocols used for IVF [
]. A large number of studies have shown that LPS improves the clinical pregnancy rate and thus the live birth rate, but the ideal LPS method remains unclear [
]. Although luteal human chorionic gonadotropin (hCG) supplementation has proven to be an effective way to overcome luteal phase defects, this treatment is frequently associated with an increased risk of ovarian hyperstimulation syndrome (OHSS) [
], so the current most widely used form of LPS is progesterone (P).
The use of P supplementation after oocyte retrieval (OR) is almost universal, but the optimal duration of P administration remains controversial. A recent large survey of 84 IVF centres in 35 countries, encompassing 51,155 cycles, found that P was continued until 10–12weeks of gestation in 67% of the cycles, whereas it was discontinued in 22% and 12% when foetal heart pulsations were recognised or when the β-hCG test was positive, respectively [
]. In the existing literature, P supplementation is variously terminated on or near the day of a positive β-hCG test [
] or extended to the day of the first ultrasound (5–7 weeks) [
], to the 8
], or as late as the 12
week of pregnancy [
]. Until recently, the available data have been insufficient to determine the optimal duration of therapy, and prolonged P protocols have been the rule, with most clinicians following the dictum, “better safe than sorry” [
]. A growing body of evidence, however, has challenged this concept and adds to the increasing concern that P supplementation of early pregnancy after IVF/ICSI might be unnecessary [
Four formulations of P are currently used for assisted reproduction, including vaginal, intramuscular (i.m.), oral and rectal preparations. Vaginal P was used for LPS as a single agent in 64% of cycles and in another 16% of cycles in combination with either i.m. (15%) or oral P (1%). As single agents, i.m. P was used in 13% of cycles, oral P in another 2% and hCG in 5% [
]. Vaginal P can result in similar pregnancy rates as i.m. P and is more comfortable and tolerable to patients [
], but it is more expensive. Conversely, i.m. P is often associated with a number of side effects, including painful injections, severe inflammatory reactions, and sterile abscesses [
]. Prolonged and repeated i.m. injections of P in oil may also lead to delayed forms of hypersensitivity reactions, with leukocytosis, marked eosinophilia and compromised pulmonary activity [
]. Orally administered P has a first-pass effect in which a high concentration is sent to the portal circulation, which, in turn, results in the production of many liver metabolites of P, some of which may be teratogenic. Despite the available literature on the teratogenic effects of prenatal oral P use [
], this agent is still used routinely by many practitioners. Therefore, taking into consideration the burden of LPS treatment, the adverse reactions to P and updated results suggesting that P supplementation during early pregnancy after IVF/ICSI might be unnecessary, we questioned whether the practice of early pregnancy P supplementation in IVF/ICSI patients should be discontinued.
The aim of this study was to perform a meta-analysis of all available randomised controlled trials (RCT) comparing early P cessation with P continuation after assisted conception in IVF/ICSI cycles to investigate potential differences in live birth, miscarriage and ongoing pregnancy rates. This review was performed in accordance with the preferred reporting items for systematic reviews and meta-analyses (PRISMA) statement principles [
Types of studies
The inclusion criteria for eligible studies were defined a priori during the design phase of this systematic review. Randomised controlled trials investigating the duration of P supplementation for luteal phase support in IVF/ICSI cycles were included. Trials using donor oocyte cycles or frozen transfers were excluded. No limitations were placed on language, date, or publication status.
Types of participants
Women undergoing IVF/ICSI who were evaluated for the effects of P supplementation duration on pregnancy outcomes were included.
Types of interventions
The interventions evaluated were early P cessation versus P continuation during the first trimester in pregnant women after IVF/ICSI.
Types of outcome measures
The primary outcome chosen for the meta-analysis was live birth rate (LBR, i.e.,a baby born alive after 24 weeks gestation). Secondary outcomes included ongoing pregnancy rate (OPR, pregnancy beyond 12 weeks of gestation, as confirmed by foetal heart activity on an ultrasound), and miscarriage rate (MR, the failure to achieve live birth after a positive β-hCG test).
Literature search and data collection
We performed an exhaustive electronic search in the following databases: MEDLINE (1946 to July 2012), EMBASE (1974 to July 2012), the Cochrane Central Register of Controlled Trials (CENTRAL), the Chinese biomedicine (CBM) literature database (1978 to July 2012), and the Wanfang database (1998 to July 2012). The search combined terms and descriptors related to IVF, ICSI, luteal phase support, and progesterone. To fit with the syntax used in each consulted database, the search strategy was modified with a series of terms suggestive of RCTs as set out by the Cochrane Handbook for Systematic Review of Intervention [
). No limit was placed on language. We also carefully browsed the references of relevant publications and added the related publications to the search. When questions related to the design or outcomes of the trials arose, we contacted the corresponding authors to confirm the information we extracted from their trials or to clarify any ambiguities.
Assessment of the risk of bias in the included studies
The risk of bias in the included studies was assessed independently by two reviewers (Xi-Ru Liu and Hua-Qiao Mu) according to the guidelines recommended in the Cochrane Handbook for Systematic Review of Intervention [
]. For each study, we assessed the risk of bias related to sequence generation, allocation, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting and other sources of bias. A judgment of ‘Yes’ meant a low risk of bias, a judgment of ‘No’ meant a high risk of bias, and ‘Unclear’ indicated an unclear risk of bias. Disagreements were discussed and resolved by consensus with a third reviewer (Qi Shi).
Data extraction and synthesis
Data extraction was performed independently by two reviewers (Xi-Ru Liu and Hua-Qiao Mu). Discrepancies were resolved by discussion with a third reviewer (Qi Shi).We extracted the following information from each eligible study: first author, year of publication, country of origin, sample size, and a number of patient characteristics, including the IVF protocol used, the exact dose of P, the route of administration, the timing of initiation and duration of luteal phase support with P, and IVF/ICSI outcomes.
Raw data were extracted from the eligible studies for each defined outcome and pooled using Review Manager 5.1 software. Dichotomous results from each study were expressed as relative risk (RR) with 95% confidence intervals (CI). These results were combined for the meta-analysis using a fixed-effects model in the absence of statistically significant heterogeneity or a random-effects model in the presence of statistically significant heterogeneity. The inter-study heterogeneity was evaluated using the
2 (Cochran's Q) statistic and the
2 value. Sensitivity analyses were performed for those studies that answered the research question of interest but used a quasi-randomised approach for patient allocation. Subgroup analyses were planned a priori based on: (1) the timing of randomisation, (2) the timing of initiation of P, (3) the GnRH analogue used for LH surge inhibition, and (4) the type and dose of P administration.
Although there is no firm evidence to support the continuation of LPS until the 10
week of gestation, this practice is used in the majority of IVF cycles worldwide [
]. This review compared the effects of early cessation with continuation of P supplementation for luteal phase support in pregnant women after IVF/ICSI, focusing on the live birth, ongoing pregnancy and miscarriage rates. The pooled results showed no significant differences in LBR between groups in which P supplementation was stopped on the day of a positive β-hCG test or for whom P supplementation was continued up to the 6
week of gestation. Similarly, the miscarriage and ongoing pregnancy rates were not affected by the duration of P administration. Because there was statistical heterogeneity in the studies analysed for OPR, we performed a subgroup analysis and a sensitivity analysis. The results of the subgroup analysis were in accordance with the above results. The findings were also stable after the sensitivity analysis, which excluded one study [
] in which odd or even patient birth years were used for patient allocation. Based on this analysis, we find no convincing evidence to support the routine use of P supplementation during early pregnancy in women undergoing IVF/ICSI. It is possible that the establishment of a pregnancy and rescue of the corpus luteum via trophoblastic hCG may make up for the possible luteal phase defect caused by the stimulated IVF cycles.
Most of the studies included in this review described their methods of sequence generation and allocation concealing. However, none of the studies mentioned blinding. Keeping trial participants, personnel, or assessors blinded to the assigned intervention might reduce the influence of subjective psychological factors on pregnancy outcomes, an important aspect of RCTs. However, owing to the nature of current LPS studies, absolute double blinding is often not practical, as it is not possible to blind the participants. None of the studies explicitly mentioned blinding of personnel or outcome assessors. Nevertheless, it is unlikely that pregnancy outcomes such as live birth, miscarriage or ongoing pregnancy can be affected by detection bias. In future studies, proper blinding protocols using a double-dummy design should be implemented, and a placebo control group should be established.
Due to the small number of studies that met the inclusion criteria and the different clinical characteristics of the participants, it was impossible to conduct meaningful subgroup analyses based on the initiation of P supplementation, the GnRH analogue used for luteinising hormone surge inhibition, or the type or dose of P administration. These analyses might become practical upon the accumulation of further studies. We were only able to analyse studies according to the different timing of randomisation, a potential source of clinical heterogeneity; here, we pooled the data from studies with similar enrolment designs. However, it should be noted that the aforementioned parameters do vary among the included studies, and we do not know whether these clinical variables might be associated with the effects of P supplementation on pregnancy outcomes. For example, P supplementation was initiated at different time-points in these studies. These were no significant differences between initiation on the day of OR or on the day of ET with regard to clinical pregnancy, ongoing pregnancy, or live birth rates, according to recent reports [
]. One study suggested that delaying the LPS until six days after OR can decrease the pregnancy rate [
]. Because P administration was initiated on the day of OR or ET in most of the eligible studies in our analysis, it is unlikely that these discrepancies affected our results. In a word, our results require confirmation in further studies in which the baseline of different groups is comparable to the greatest possible extent.
Two important limitations of our meta-analysis should be noted: (1) only six studies were included in this review, and the number of patients analysed is far below the sample size required to exclude a clinically important difference. A non-inferiority trial showing a difference of −4% or larger from a live birth rate of 80% would require a sample size of 3,140 women with a positive pregnancy test after IVF [
]. (2) The external validity of the study may be limited because existing studies excluded those patients with early bleeding, advanced age or polycystic ovary syndrome as well as those patients with an inadequate hCG rise or endometriosis, whose luteal phase may behave differently. Future trials should recruit such patients to better stratify the outcomes for these patient groups.
In conclusion, based on the currently available evidence, progesterone supplementation might be unnecessary beyond the first positive β-hCG test after IVF/ICSI. However, considering the large number of IVF cycles performed globally and the side effects and costs of progesterone treatment, additional well-designed RCTs are urgently needed to investigate the optimal duration of progesterone administration during early pregnancy in women undergoing IVF/ICSI.
The authors declare that they have no competing interests.
Hong-Bo Qi was responsible for designing and coordinating the study. All authors were responsible for data collection, data analysis, and data interpretation. Xi-Ru Liu, Hua-Qiao Mu, and Qi Shi were responsible for the statistical analysis and for writing the manuscript. Xiao-Qiu Xiao was responsible for reviewing the manuscript. All authors read and approved the final manuscript.