The impact of luteal serum progesterone levels on live birth rates—a prospective study of 602 IVF/ICSI cycles

Abstract

STUDY QUESTION

Is the chance of a live birth following IVF treatment and fresh embryo transfer affected by early and mid-luteal serum progesterone (P₄) levels?

SUMMARY ANSWER

Low as well as high serum P₄ levels in the early and mid-luteal phase reduce the chance of a live birth following IVF treatment with fresh embryo transfer.

WHAT IS KNOWN ALREADY

Data from non-human studies and studies of frozen–thawed embryo transfer cycles indicate that low as well as high P₄ levels during the mid-luteal phase decrease the chance of pregnancy. The altered P₄ pattern may disrupt the endometrial maturation leading to asynchrony between embryonic development and endometrial receptivity, thereby, compromising implantation and early development of pregnancy.

STUDY DESIGN, SIZE, DURATION

Prospective multicenter cohort study of 602 women undergoing IVF treatment. Patients were recruited from four Danish public Fertility Centers from May 2014 to June 2017. The study population was unselected, thus, representing a normal everyday patient cohort. Patients were treated in a long GnRH-agonist protocol or a GnRH-antagonist protocol and triggered for final oocyte maturation with either hCG or a GnRH-agonist. The same vaginal luteal support regimen was applied in all patients.

PARTICIPANTS/MATERIALS, SETTING, METHODS

Serum P₄ levels from the early or mid-luteal phase were correlated to positive hCG and live birth rates (delivery > gestational week 20). Patients were divided into four P₄ groups based on raw data of P₄ serum levels and reproductive outcomes during early luteal phase (P₄<60 nmol/l, P₄ 60–100 nmol/l, P₄ 101–400 nmol/l and P₄>400 nmol/l) and during mid-luteal phase (P₄<150 nmol/l, P₄ 150–250 nmol/l, P₄ 251–400 nmol/l and P₄>400 nmol/l).

MAIN RESULTS AND THE ROLE OF CHANCE

The optimal chance of pregnancy was achieved with serum P₄ levels of 60–100 nmol/l in the early luteal phase whereas the optimal P₄ level during the mid-luteal phase was 150–250 nmol/l. Below, but most distinctly above these levels, the chance of pregnancy was consistently reduced. With an early luteal P₄ level of 60–100 nmol/l, the chance of a positive hCG-test was 73%, 95% CI: [59, 84] following cleavage stage embryo transfer. In contrast, with P₄ levels >400 nmol/l, the chance of a positive hCG-test was significantly reduced to 35%, 95% CI: [17, 57], thus, an absolute risk difference of −38%, P = 0.01. A similar negative association between early luteal P₄ and live birth rate was found, although it did not reach statistical significance. During the mid-luteal phase, a P₄ level of 150–250 nmol/l resulted in an optimal chance of live birth: 54%, 95% CI: [37, 70] compared to 38%, 95% CI: [20, 60] with a P₄ level >400 nmol/l, thus, an absolute risk difference of −16%, P = 0.14. All estimates were adjusted for maternal age, maternal BMI, study site, final follicle count and late follicular P₄ levels.

LIMITATIONS, REASONS FOR CAUTION

This study is the first to explore the possible upper and lower thresholds for luteal P₄ following IVF treatment and fresh embryo transfer, and the optimal P₄ ranges found in this study should be corroborated in future clinical trials. Furthermore, the P₄ thresholds in this study only apply to fresh IVF cycles, using vaginal luteal phase support, as the optimal P₄ level in cycles using intramuscular P₄ may be different.

WIDER IMPLICATIONS OF THE FINDINGS

Future studies are necessary to explore whether additional exogenous luteal P₄ supplementation in the low P₄ group could increase the chance of a live birth following fresh embryo transfer, and whether patients with luteal P₄ levels >400 nmol/l would benefit from segmentation followed by subsequent transfer in frozen/thawed cycles.

TRIAL REGISTRATION NUMBER

NCT02129998 (Clinicaltrials.gov).

STUDY FUNDING/COMPETING INTEREST(S)

L.H.T. received an unrestricted grant from Ferring Pharmaceuticals, Denmark, to support this study. P.H. received unrestricted research grants from MSD, Merck, Gedeon Richter and Ferring Pharmaceuticals outside of this work as well as honoraria for lectures from MSD, Merck and Gedeon Richter outside of this work. U.K. received honoraria for lectures from MSD and Ferring Pharmaceuticals outside of this work. C.A. received unrestricted research grants from MSD, IBSA, and Ferring Pharmaceuticals outside of this work as well as honoraria for lectures from MSD and IBSA. H.O.E. and B.B.P. received an unrestricted research grant from Gedeon Richter outside of this work. K.E., L.B., D.P. and B.H. have no conflict of interest. Furthermore, grants from ‘The Health Research Fund of Central Denmark Region’, ‘The Research Foundation of the Hospital of Central Jutland’, ‘The Research Foundation of A.P. Møller’, ‘The Research Foundation of Aase & Ejnar Danielsen’, ‘The Research Foundation of Dagmar Marshall’, ‘The Research Foundation of Dir. Jacob Madsen & Hustru Olga Madsen’, ‘The Research Foundation of Fam. Hede Nielsen’ and ‘The Danish Medical Research Grant’ supported conducting this study. The providers of funding were neither involved in the conduction of the study nor in the writing of the scientific report.

serum progesterone, luteal phase, IVF, live birth rate, fresh embryo transfer

Introduction

Progesterone (P₄) is essential for the secretory transformation of the human endometrium and for the support of early pregnancy. Following ovarian stimulation for IVF treatment, the P₄ profile during the luteal phase is distinctly different from that of the natural cycle (Fauser et al., 2002).

As the optimal luteal P₄ level following IVF treatment is poorly understood, it is normal practice to supplement the luteal phase with exogenous P₄ after IVF treatment, using a standard regimen without taking ovarian response to stimulation or luteal steroid levels into account. From an early study by Humaidan et al. (2005), it was clear that a very low luteal P₄ level following fresh embryo transfer has a deleterious effect on the ongoing pregnancy rate. Patients in that study were triggered with a GnRH agonist for final oocyte maturation, followed by a standard vaginal P₄ support, only. This approach resulted in mid-luteal P₄ levels resembling what is seen during the natural cycle (39 nmol/l). The study was terminated prematurely due to a very low clinical pregnancy rate (6%) and an unexpected high early pregnancy loss rate (79%). In the subsequent randomized controlled trials, the luteal phase support after GnRH agonist triggering was supplemented with a small bolus of exogenous hCG to rescue the corpus luteum function (Humaidan et al., 2010, 2013). With this adaption of protocol, the subsequent mid-luteal P₄ levels increased to 77–409 nmol/l and ongoing pregnancy rates were comparable to those of the hCG-triggered control groups.

On this basis, the existence of a lower luteal P₄ threshold necessary to induce a normal secretory endometrial development and the establishment of a pregnancy following IVF treatment and fresh embryo transfer seems plausible. Whether an upper P₄ threshold exists remains unexplored. Interestingly, in frozen embryo transfer cycles (FET) recent studies reported a lower as well as a higher threshold for luteal P₄ in terms of reproductive outcomes (Labarta et al., 2017; Alsbjerg et al., 2018). Moreover, in a study by Yovich et al., looking at 529 single blastocyst transfers in FET/HRT cycles, an optimal range for mid-luteal P₄ of 70–99 nmol/l was reported. Below, but also above this range, the clinical pregnancy rate was significantly reduced by 20% (Yovich et al., 2015).

The aim of the present study was to explore the impact of early and mid-luteal P₄ levels on the reproductive outcomes following IVF/ICSI treatment in fresh embryo transfer cycles.

Materials and Methods

Study design

Prospective multicenter cohort study based on serum samples from the early and mid-luteal phases of women undergoing IVF treatment.

Study population

Patients were recruited from four different Danish public fertility centers (The Fertility Clinic Skive, The Fertility Clinic Odense University Hospital, The Fertility Clinic Horsens Region Hospital and The Fertility Clinic Herlev Hospital) from May 2014 to June 2017. The study population was unselected, representing the normal everyday patient cohort treated at the clinics. The inclusion criteria, therefore, allowed patients treated in a long GnRH agonist protocol and a GnRH antagonist protocol to participate as well as patients triggered with hCG or a GnRH agonist. All patients were under the age of 41 and with a BMI < 35 kg/m² as required by Danish national guidelines for public fertility treatment (Danish Fertility Society Guideline, 2016). All patients only participated once.

Written and oral information about the study was given to 1482 patients from which 609 persons (41%) declined to participate. The final study population included 602 patients who had an embryo transfer and relevant luteal phase blood samples taken (Fig. 1, flowchart).

Figure 1

Flowchart of study participation. ET, embryo transfer; OHSS, ovarian hyperstimulation syndrome.

Serum prolactin levels and thyroid-stimulating hormone (TSH) were within the normal range in all patients prior to treatment start. Clinical information regarding age, body mass index, smoking habits, primary diagnosis, antral follicle count and basal FSH/LH levels were obtained prior to treatment by clinical staff.

Ovarian stimulation

Due to the non-interventional nature of this study, treatment choices regarding protocol- and trigger type were made on a case-by-case basis according to patient characteristics and individual assessment.

Patients treated in a long GnRH-agonist protocol were down-regulated using daily SC injections of Buserelin (Suprefact^®, Sanofi, Denmark) starting in the mid-luteal phase of the preceding cycle. Ovarian stimulation commenced after 12–14 days of down-regulation in case of an endometrial thickness <4 mm. Final follicle maturation was induced with hCG 5000–10 000 IU (Pregnyl^®, MSD, Denmark or Ovitrelle, Merck Biopharma, Denmark) when two or more leading follicles reached a diameter of 17 mm.

If the GnRH antagonist protocol was used, ovarian stimulation commenced on cycle Day 2 or 3. Daily GnRH antagonist co-treatment was added at a follicle size of 12 mm or from cycle Day 6. Final oocyte maturation was induced with SC Buserelin 0.5 mg or hCG 5000–10 000 IU (Pregnyl^®, MSD, Denmark or Ovitrelle, Merck Biopharma, Denmark), when two or more leading follicles reached a diameter of 17 mm.

Ovarian stimulation was performed using either r-FSH (Gonal-f^®, Merck Biopharma, Denmark), rFSH/LH (Pergoveris, Merck Biopharma, Denmark) or hMG (Menopur^®, Ferring Pharmaceuticals, Denmark) alone or in combination with corifollitropin-alfa (Elonva, MSD, Denmark). Dose adjustments were performed according to ovarian response monitored by transvaginal ultrasound during treatment. Oocyte pick-up (OPU) was carried out 36 h after the ovulation trigger. IVF or ICSI was performed according to normal clinical practice. The quality of all available embryos was evaluated in accordance with the Consensus scoring system for cleavage-stage embryos described by the Alpha Scientist group (ALPHA Scientists In Reproductive Medicine and ESHRE Special Interest Group Embryology, 2011) or by use of the Gardner standard for blastocyst scoring (Gardner et al., 2000). A maximum of two embryos were transferred on either Day 2, 3 or 5 after OPU. The criteria for the day of transfer followed the local laboratory guidelines of the participating clinics.

Luteal phase support

All patients received vaginal luteal phase support in a standard regimen using 300 mg micronized P₄ daily (Lutinus^®, Ferring Pharmaceuticals), starting on the day after OPU. Patients triggered with the GnRH-agonist all had a bolus of 1500 IU hCG on the day of OPU. Based on the individual ovarian response to stimulation, some patients received an additional hCG bolus OPU + 5 based on a protocol previously described (Humaidan et al., 2013). A small fraction of patients (13%) had one bolus of GnRH agonist (Gonapeptyl^® 0.1 mg, Ferring Pharmaceuticals, Denmark) on OPU + 7 based on the individual clinical assessment. Vaginal P₄ administration continued until the day of pregnancy testing (hCG triggering) or until the seventh gestational week (GnRHa trigger).

Blood sampling

For each patient, blood sampling was performed on three occasions: on the day of triggering for final oocyte maturation (or the day before), on the day of embryo transfer (Day 2, 3 or 5) and on the day of pregnancy testing (OPU + 14). After coagulation at room temperature, blood samples were centrifuged, and serum was isolated and frozen immediately at −80°C until analysis.

Hormone assays

Serum P₄, estradiol and hCG concentrations were measured using commercially available automated electro chemiluminescent immunoassays (Immulite^® 2000XPi, Siemens Healthcare, Denmark and Architect^® i2000SR, Abbott Diagnostics, USA) routinely used for analysis at the Department of Biochemistry, Odense University Hospital, Denmark. All measurements were performed by experienced technicians according to manufacturer’s instructions.

The detection limit for P₄ was 0.6 nmol/l, and the in-house inter- and intra-assay coefficients of variation were 4.4 and 1.6%, respectively. The detection limit for estradiol was 0.05 nmol/l and the in-house inter- and intra-assay coefficients of variation were 2.8 and 1.6%, respectively. The detection limit for hCG was 1.2 IU/l, and the in-house inter- and intra-assay coefficients of variation were 3.4 and 1.7%, respectively.

Exposure

Due to a known steep increase in P₄ secretion from the early luteal phase towards the mid-luteal phase in the IVF cycle (Hassiakos et al., 1990), the outcome measures for patients with early luteal P₄ monitoring and patients with mid-luteal P₄ monitoring, are presented separately. Progesterone groups were defined based on raw data of reproductive outcomes and luteal P₄ levels during the early and mid-luteal phase separately (Supplementary Fig. S1). For objectivity and comparison, estimates were also calculated based on 10/50/90 and 25/50/75 percentiles.

Outcome

The serum β-hCG concentration was determined 14 days after OPU and was considered positive if β-hCG > 10 IU/l. In patients with β-hCG levels between 10 and 45 IU/l, a control β-hCG was performed after 48 h. Clinical pregnancy was defined as the presence of a live fetus within an intrauterine gestational sac at ultrasound examination in gestational weeks 7–8. Early pregnancy loss refers to (i) patients with an insufficient β-hCG value on the day of pregnancy testing (10–45 IU/l), and decreasing β-hCG values towards null in subsequent hCG-controls; (ii) patients with a positive hCG, but no intra- or extrauterine sac visualized on transvaginal ultrasound in gestational weeks 7–8; and (iii) patients with a fetus without visible heartbeat at ultrasound in gestational weeks 7–8. Clinical pregnancy loss was defined as the loss of a viable intrauterine pregnancy up to and including gestational weeks 20 + 0. A live birth was defined as the delivery of a live infant after gestational week 20 + 0. Clinical gestational dating was performed using the day of OPU as gestational week 2 + 0.

Confounding factors

The potential confounders included in the model were chosen a priori based on a Directed Acyclic Graph (DAG) (Supplementary Figs S2 and S3). DAGs are visual representations of causal assumptions and can help identifying confounding factors that obscure the real effect of the exposure on the outcome (Howards, 2018). Based on a structured analysis of the DAG, it is possible to identify a minimum, however, sufficient set of covariates to adjust for in the statistical analysis, which will cover all confounding elements. The web application DAGitty (www.dagitty.net) was used to draw and analyze the DAGs.

Statistical methods

Data are presented as percentages for categorical variables and as mean and standard deviation for continuous parametric variables and median and range for continuous non-parametric variables. Differences in categorical variables between P₄ groups were assessed with Pearson’s chi-square test or Fishers exact test when appropriate. Differences in continuous parametric data between the four P₄ groups were assessed using one-way analysis of variance followed by a post hoc pairwise comparison in case of a statistical difference between groups. Kruskal–Wallis test was used in case of non-parametric continuous data.

The correlation between luteal P₄ levels and the final follicle count was tested using a linear regression model. A multiple logistic regression model was used to assess the association between luteal P₄ levels and reproductive outcome. The model included the independent variables maternal age, maternal BMI, study site, final follicle count on the day of trigger (>12 mm) and late follicular P₄ level for estimates of positive hCG, clinical pregnancy and live birth (Supplementary Fig. S2). For estimates of early pregnancy loss adjustment was made for maternal age, maternal BMI, smoking, final follicle count and peak estradiol level on the day of trigger (Supplementary Fig. S3). The cut-off for late follicular P₄ (>4.77 nmol/l) was chosen based on the results of earlier studies (Bosch et al., 2010).

In case of missing data of covariates, patients were omitted from the final regression analysis (43 patients (10%) on Days 2–3 and 11 patients (3%) on Day 5). Patients with missing values were equally distributed across P₄ groups in both cohorts (P = 0.27 and P = 0.25, respectively). No patients were lost to follow-up.

All statistical analyses were performed using STATA version 13, StataCorp LLC, USA.

Ethical approval

All patients gave their written consent prior to study participation. The study was conducted according to the declaration of Helsinki for Medical Research and was approved by the local Ethics Committee of Central Denmark Region. ClinicalTrial.gov registration number NCT02129998.

Results

Demographic data

The population consisted of 602 women undergoing IVF treatment followed by fresh embryo transfer. The overall mean age and BMI of patients were 32.5 ± 4.5 years and 25.0 ± 4.2 kg/m², respectively. Baseline characteristics of participants are provided in Table I and cycle characteristics in Table II.

Table I

Baseline characteristics.

	Cohort for early luteal P4 monitoring								Cohort for mid-luteal P4 monitoring
Progesterone (nmol/l)	N	All	<60	60–100	101–400	>400	P	Progesterone (nmol/l)	N	All	<150	150–250	251–400	>400	P
Number of patients, n	432		17	70	316	29		Number of patients, n	170		34	44	39	53
Maternal age (years)	432	32.6 ± 4.5	34.1 ± 5.8	33.6 ± 5.0	32.4 ± 4.4	31.8 ± 3.5	0.091	Maternal age (years)	170	32.3 ± 4.6	32.0 ± 5.0	32.0 ± 4.9	32.3 ± 4.9	32.6 ± 4.5	0.920
Maternal BMI (kg/m2)	432	25.0 ± 4.2	26.5 ± 3.9	26.9 ± 4.3	24.7 ± 4.2	23.0 ± 3.5	<0.001	Maternal BMI(kg/m2)	170	25.1 ± 4.0	27.0 ± 3.7	26.5 ± 4.0	25.4 ± 3.8	22.6 ± 3.0	<0.001
Maternal smoking (%)	432	10	12	17	8	7	0.131	Maternal smoking (%)	170	5	6	7	3	4	0.775
Basal FSH (IU)	398*	6.2 (0.3–22.0)	8.6 (2.5–22.0)	5.5 (0.3–12.9)	6.3 (1.1–16.0)	6.1 (1.3–9.1)	0.063	Basal FSH (IU)	164*	6.3 (0.1–17.5)	6.2 (2.2–9.2)	6.3 (2.2–14.0)	5.8 (0.1–13.9)	6.5 (1.9–17.5)	0.795
Basal LH (IU)	385**	5.3 (0.2–19.1)	5.1 (3.6–19.1)	4.6 (0.2–19.1)	5.4 (0.4–17.0)	4.8 (2.0–12.2)	0.194	Basal LH (IU)	158**	5.8 (0.1–40.0)	5.9 (1.2–15)	7.3 (2.1–25.0)	5.2 (0.1–21)	5.4 (0.1–21.0)	0.197
Antral follicle count, n	432	12 (2–50)	7 (3–26)	10 (2–30)	12 (3–50)	15 (7–40)	<0.001	Antral follicle count, n	170	15 (4–50)	14 (6–28)	15 (4–30)	15 (6–32)	15 (5–50)	0.344
Primary diagnosis	432							Primary diagnosis	170
Unexplained (%)		27	17	27	28	21	0.760***	Unexplained (%)		21	14	18	21	28	0.438***
Tubal (%)		8	6	7	8	10	0.963	Tubal (%)		13	21	11	10	11	0.574
PCO/PCOS (%)		7	6	6	7	14	0.517	PCO/PCOS, % (%)		19	15	27	13	19	0.340
Endometriosis (%)		6	12	9	5	10	0.194	Endometriosis (%)		4	6	7	0	4	0.402
Male (%)		42	59	36	42	38	0.348	Male (%)		31	32	25	33	32	0.828
Single/female partner (%)		9	0	14	9	4	0.208	Single/female partner (%)		12	12	12	23	6	0.108
Other (%)		1	0	1	1	3	0.271	Other (%)		0	0	0	0	0	–

Baseline characteristics are presented as mean ± SD for continuous parametric data and as median (range) for continuous non-parametric data. Categorical data is presented as percentages (%).

*Data on basal FSH levels were missing in 40 patients (7%). Patients with missing data on basal FSH levels were equally distributed across P₄ groups on both Days 2–3 (P = 0.05) and Day 5 (P = 0.91).

**Data on basal LH levels were missing in 59 patients (10%). Patients with missing data on basal LH levels were equally distributed across P₄ groups on both Days 2–3 (P = 0.17) and Day 5 (P = 0.99).

***P-value describes the comparison between the chosen primary diagnosis category and the combined group of all other primary diagnosis categories.

SI conversion factor for P₄: nmol/l = 3.18 ng/ml.

Table II

Descriptive data of controlled ovarian stimulation, oocytes, embryo transfer and luteal phase support.

	Cohort for early luteal phase P4 monitoring								Cohort for mid-luteal phase P4 monitoring
Progesterone (nmol/l)	N	All	<60	60–100	101–400	>400	P	Progesterone (nmol/l)	N	All	<150	150–250	251–400	>400	P
Number of patients	432		17	70	316	29		Number of patients	170		34	44	39	53
Protocol	432							Protocol	170
Antagonist (%)		58	65	43	58	76	0.060	Antagonist (%)		75	88	75	82	68	0.190
Long GnRH agonist (%)		42	35	57	42	24		Long GnRH agonist (%)		25	12	25	28	32
Total FSH dose (IU)	432	2263 (750–7350)	2700 (1013–4500)	2338 (1050–7350)	2269 (750–6750)	1950 (900–3600)	0.022	Total FSH dose (IU)	170	2063 (500–4425)	2081 (900–4275)	2100 (900–4425)	2100 (875–3975)	1975 (500–3900)	0.646
Stim duration (days)	432	10.4 ± 2.0	10.7 ± 2.0	10.3 ± 2.1	10.4 ± 2.0	10.2 ± 1.5	0.717	Stim duration (days)	170	10.2 ± 2.0	9.4 ± 1.9	10.2 ± 2.1	10.4 ± 2.2	10.6 ± 1.8	0.041
Final follicle count > 12 mm on trigger day	432	9 (1–27)	3 (1–17)	5 (2–18)	9 (1–27)	13 (6–26)	<0.001	Final follicle count > 12 mm on trigger day	170	12 (2–29)	10 (2–18)	10 (5–29)	12 (6–19)	14 (6–23)	0.001
Mode of triggering for final oocyte maturation	432						<0.001	Mode of triggering for final oocyte maturation	170						<0.001
hCG (%)		61	82	79	59	31		hCG (%)		46	18	36	56	66
Agonist (%)		39	18	21	41	69		Agonist (%)		54	82	64	44	34
Number of oocytes retrieved (n)	432	7 (1–21)	2 (1–7)	5 (1–13)	8 (1–21)	11 (2–20)	<0.001	Number of oocytes retrieved (n)	170	12 (2–28)	11 (5–24)	11 (2–28)	12 (6–22)	13 (6–26)	0.032
Number of fertilized oocytes (n)	432	6 (1–21)	2 (1–6)	4 (1–12)	7 (1–21)	10 (2–20)	<0.001	Number of fertilized oocytes (n)	170	11 (2–27)	8 (4–24)	10 (2–27)	11 (6–22)	12 (6–24)	0.039
Single embryo transfer (%)	432	77	76	77	78	66	0.517	Single embryo transfer (%)	170	94	91	95	95	94	0.869
Double embryo transfer (%)		23	24	23	22	34		Double embryo transfer (%)		6	9	5	5	6
At least one top quality embryo for transfer (%)	432	62	53	59	62	69	0.682	At least one top quality embryo for transfer (%)	170	69	68	77	69	62	0.466
Mean embryo score	432							Mean embryo score	170
SET		1.5 ± 0.6	1.5 ± 0.8	1.4 ± 0.5	1.5 ± 0.6	1.5 ± 0.7	0.990	SET		1.4 ± 0.6	1.4 ± 0.6	1.3 ± 0.6	1.4 ± 0.7	1.5 ± 0.7	0.537
DET		1.7 ± 0.6	1.8 ± 0.5	1.5 ± 0.6	1.8 ± 0.6	1.6 ± 0.5	0.470	DET		1.4 ± 0.5	1.3 ± 0.6	1.7 ± 0.6	2.0 ± 0.0	1.0 ± 0.0	0.262
Luteal phase support	432							Luteal phase support	170
Vaginal progesterone only (%)		55	65	77	52	31	<0.001*	Vaginal progesterone only (%)		45	15	36	56	64	<0.001*
+ 1 Bolus of hCG (%)		13	6	0	13	45	<0.001	+ 1 Bolus of hCG (%)		27	29	34	18	26	0.416
+ 2 Boluses of hCG (%)		25	12	22	27	24	0.448	+ 2 Boluses of hCG (%)		26	53	30	23	8	<0.001
Vaginal P4 + Gonapeptyl (%)		7	17	1	8	0	0.026	Vaginal P4 + Gonapeptyl (%)		2	3	0	3	2	0.751

Descriptive data is presented as mean ± SD for continuous parametric data and as median (range) for continuous non-parametric data. Categorical data is presented as percentages (%).

*P-value describes the comparison between the chosen luteal phase support category and the combined group of all other luteal phase support categories.

SI conversion factor for P₄: nmol/l = 3.18 ng/ml.

Reproductive outcomes

The overall rate for positive hCG per transfer and live birth rate per transfer in the study was 48% (286/602) and 34% (203/602), respectively. The overall early pregnancy loss rate was 26% (75/286). There was a statistically significant correlation between the final number of follicles and the luteal P₄ level (P < 0.001).

Early luteal P₄ levels and reproductive outcome (n = 432)

As seen from Fig. 2, the relationship between luteal P₄ levels and reproductive outcomes seems to be non-linear. In the early luteal phase, the optimal P₄ level for a positive pregnancy outcome was between 60 and 100 nmol/l. Below, but most distinctly above this level, the chance of a positive reproductive outcome decreased (Table III). The clinical impact of different early luteal P₄ levels on the chance of a live birth is demonstrated in Fig. 3.

Figure 2

Reproductive outcome in different progesterone groups during the early and mid-luteal phase.

Table III

Reproductive outcomes in different luteal progesterone groups.

	Cohort for early luteal phase P4 monitoring
	OR for positive hCG				OR for clinical pregnancy				OR for live birth				OR for early pregnancy loss
	N	Crude OR [95% CI]	N	Adjusted OR [95% CI]	N	Crude OR [95% CI]	N	Adjusted OR [95% CI]	N	Crude OR [95% CI]	N	Adjusted OR [95% CI]	N	Crude OR [95% CI]	N	Adjusted OR [95% CI]
N	183/432	432*		389**	135/432	432*		389**	130/432			389**	48/183	432*		385***
P4 < 60 nmol/l	5/17	0.42 [0.13;1.31]	17	0.61 [0.18;2.14]	3/17	0.41 [0.11;1.60]	17	0.53 [0.13;2.21]	3/17	0.54 [0.14;2.10]	17	0.73 [0.17;3.05]	2/5	0.72 [0.14;3.62]	17	1.02 [0.18;5.78]
P4 60–100 nmol/l	35/70	1	70	1	24/70	1	62	1	20/70	1	62	1	11/35	1	62	1
P4 101–400 nmol/l	131/316	0.71 [0.42;1.19]	316	0.35 [0.18;0.68]	101/316	0.90 [0.52;1.56]	285	0.53 [0.26;1.04]	100/316	1.16 [0.65;2.05]	285	0.69 [0.34;1.38]	30/131	0.56 [0.27;1.19]	281	0.31 [0.12;0.85]
P4 > 400 nmol/l	12/29	0.71 [0.29;1.69]	29	0.19 [0.09;0.61]	7/29	0.61 [0.23;1.65]	25	0.21 [0.06;0.76]	7/29	0.80 [0.29;2.17]	25	0.28 [0.08;1.00]	5/12	1.12 [0.35;3.58]	25	0.44 [0.09;2.10]

	Cohort for mid-luteal phase P4 monitoring
	OR for positive hCG				OR for clinical pregnancy				OR for live birth				OR for early pregnancy loss
	N	Crude OR [95% CI]	N	Adjusted OR (95% CI)	N	Crude OR [95% CI]	N	Adjusted OR [95% CI]	N	Crude OR [95% CI]	N	Adjusted OR [95% CI]	N	Crude OR [95% CI]	N	Adjusted OR [95% CI]
N	103/170	170*		159**	76/170	170*		159**	73/170	170*		159**	27/103	170*		155***
P4 < 150 nmol/l	21/34	0.42 [0.15;1.17]	32	0.36 [0.13;1.04]	16/34	0.56 [0.22;1.41]	32	0.55 [0.21;1.42]	16/34	0.68 [0.27;1.68]	32	0.60 [0.23;1.55]	5/21	0.78 [0.23;2.65]	32	0.80 [0.23;2.80]
P4 150–250 nmol/l	35/44	1	42	1	27/44	1	42	1	25/44	1	42	1	8/35	1	42	1
P4 251–400 nmol/l	19/39	0.24 [0.09;0.68]	39	0.25 [0.09;0.68]	17/39	0.49 [0.20,1.19]	39	0.63 [0.25;1.54]	17/39	0.59 [0.24;1.42]	39	0.70 [0.28;1.71]	2/19	0.24 [0.05;1.28]	37	0.2 [0.04;1.06]
P4 >400 nmol/l	28/53	0.29 [0.11;0.75]	46	0.43 [0.14;1.27]	16/53	0.27 [0.11;0.66]	46	0.50 [0.18;1.36]	15/53	0.30 [0.12;0.73]	46	0.51 [0.19;1.39]	12/28	1.32 [0.48;3.60]	44	0.96 [0.27;3.38]

Early luteal phase P₄ quantifications were performed on OPU + 2 or OPU + 3. Mid-luteal phase P₄ quantifications were performed on OPU + 5.

*In the crude OR estimates, all 602 patient with embryo transfer were included.

**Due to missing data on the co-variate late follicular P₄ level in 43 patients on OPU + 2–3 and 11 patients on OPU + 5, the final adjusted regression model included 389 and 159 patients in the early luteal phase cohort and in the mid-luteal phase cohort, respectively. Patients with missing data were equally distributed across P₄ groups in both the early luteal phase (P = 0.47) and in the mid-luteal phase (P = 0.25).

***Due to missing data on the co-variate peak E₂ level in 47 patients on OPU + 2–3 and 15 patients on OPU + 5, the final adjusted regression model included 385 and 155 patients in the early luteal phase cohort and in the mid-luteal phase cohort, respectively. Patients with missing data were equally distributed across P₄ groups in both the early luteal phase (P = 0.50) and in the mid-luteal phase (P = 0.09).

SI conversion factor for P₄: nmol/l = 3.18 ng/ml. CI = confidence interval; OPU = day of oocyte pick-up.

Figure 3

Live birth rates in different early luteal P₄ groups following IVF treatment and cleavage-stage embryo transfer. All estimates are adjusted for maternal age, maternal BMI, final follicle count, study site and late follicular P₄ levels.

The risk of an early pregnancy loss decreased with P₄ >100 nmol/l compared to the reference group (Table III). Dividing the cohort into two groups with P₄ below or above 100 nmol/l, the OR for early pregnancy loss was significantly increased in the low P₄ group compared to the high P₄ group: 3.2, 95% CI: [1.21, 8.54].

Using 10/50/90 or 25/50/75 percentiles to define four P₄ groups revealed the same overall pattern for OR for live birth as seen with the a priori chosen P₄ groups presented above, but with less difference between groups (Supplementary Fig. S6). Adding smoking to the regression model did not change the estimates significantly (data not shown).

Mid-luteal P₄ levels and reproductive outcomes (n = 170)

In the mid-luteal phase, i.e. OPU + 5, the optimal P₄ range for a successful reproductive outcome was 150–250 nmol/l (Table III). As seen in Fig. 2, a similar non-linear pattern for both positive hCG, clinical pregnancy and live birth was found with a negative association between the reproductive outcomes and P₄ levels below and above the optimal range.

The chance of a live birth in different mid-luteal P₄ groups is shown in Fig. 4. The figure illustrates that for a reference person, the chance of a live birth following blastocyst transfer is 54%, 95% CI: [37, 70%] if the mid-luteal P₄ level is 150–250 nmol/l. In contrast, if the same reference person has a mid-luteal P₄ level >400 nmol/l, the chance of a live birth is reduced to 38%, 95% CI: [20, 60%], thus, an absolute risk difference of –16 percentage points, P = 0.14.

Figure 4

Live birth rates in different mid-luteal P₄ groups following IVF treatment and blastocyst transfer. All estimates are adjusted for maternal age, maternal BMI, final follicle count, study site and late follicular P₄ levels.

Using 10/50/90 or 25/50/75 percentiles to define four P₄ groups revealed the same overall pattern for live birth as seen with the a priori chosen P₄ groups (Supplementary Fig. S6). Adding smoking to the regression model did not change the estimates significantly (data not shown).

Confounding effects of protocol type, triggering type or luteal phase support

Based on the DAG neither protocol- nor trigger type was included in the final regression model as their possible confounding effects on the association between luteal P₄ and reproductive outcome were already accounted for by the covariates used in the statistical model. Adding trigger type or protocol type to the regression model did not change estimates significantly. Furthermore, the same reproductive pattern was seen when evaluating cycles with different protocol or trigger types separately, thus, precluding effect modification between protocol or trigger type and luteal P₄ levels (Supplementary Figs S4 and S5). The same non-linear pattern between luteal P₄ levels and reproductive outcomes was seen when examining Day 2, 3 or 5 separately (data not shown). A minor fraction of patients (13%) had one small bolus of GnRH agonist on OPU + 7. These patients were equally distributed across different P₄ groups (P > 0.05). Omitting these patients from the cohort did not change outcomes significantly (data not shown).

Prevalence of low and high luteal P₄ levels

In the early luteal phase, only 4% of patients had P₄ <60 nmol/l and 7% had P₄ >400 nmol/l. In contrast, in the mid-luteal phase, 20% of patients had serum P₄ levels <150 nmol/l and 31% had serum P₄ levels >400 nmol/l. Thus, measured on OPU + 5, 51% of all treated patients had P₄ levels consistent with a reduced chance of pregnancy compared to the optimal P₄ level.

Discussion

To the best of our knowledge, this study is the first in decades to explore the optimal luteal P₄ level following IVF and fresh embryo transfer.

The results of the present study suggest that low as well as high P₄ levels during the early and mid-luteal phases reduce the chance of a live birth during IVF treatment and fresh embryo transfer. In the early luteal phase (OPU + 2 or OPU + 3) the optimal range for P₄ was between 60 and 100 nmol/l, whereas the optimal P₄ level in the mid-luteal phase (OPU + 5) was 150–250 nmol/l.

Whether we looked at crude or adjusted OR, OR for positive hCG, clinical pregnancy or live birth or OR for reproductive outcomes in the early or mid-luteal phase, the pattern was distinctly the same. Thus, the probability of a positive reproductive outcome was reduced below and above the pre-defined optimal levels of P₄.

The importance of luteal P₄ for the establishment and maintenance of pregnancy is undebatable, and it is well accepted that the success of an IVF cycle is crucially dependent upon a sufficient luteal phase support (Humaidan et al., 2005). However, the optimal luteal P₄ level following IVF and fresh embryo transfer is poorly understood and did not receive much attention in recent years. Thus, there has been a lack of studies examining the optimal luteal P₄ levels in fresh embryo transfer cycles and therefore, a lack of defined P₄ thresholds to be used in the clinical setting. Consequently, luteal P₄ monitoring following fresh embryo transfer has not been implemented in standard IVF programs.

Several decades ago, numerous studies examined the early luteal P₄ profile following IVF treatment, correlating steroid levels to the chance of achieving a pregnancy (Balasch et al., 1995; Hassiakos et al., 1990; Howles et al., 1987; Liu et al., 1995). However, all the mentioned studies compared luteal P₄ levels in pregnant versus non-pregnant patients assuming by default that the pregnant group would display a higher P₄ level than the non-pregnant group. These studies showed no differences in P₄ levels between groups and, therefore, concluded that luteal serum P₄ did not impact reproductive outcomes. However, bearing in mind the results of the present study, the comparison between luteal P₄ in pregnant versus non-pregnant patients is sub-optimal, as both very low and very high P₄ levels seem to reduce the chance of pregnancy. Consequently, the P₄ level in the non-pregnant group will consist of a mean of both very high and very low P₄ concentrations and will, therefore, be comparable to the P₄ level of the pregnant group, thus masking the true effect of different P₄ levels on the reproductive outcome.

A single recent study examined the mid-luteal P₄ levels in 595 patients following blastocyst transfer (Petersen et al., 2018). As in the older studies, the P₄ levels in pregnant patients were compared to P₄ levels in non-pregnant patients, revealing comparable steroid levels. This led the authors to conclude that their data did not support the concept of an optimal P₄ range following IVF and fresh embryo transfer. However, with the present study we challenge this statement and emphasize that the true effect of high and low P₄ levels may be overlooked by the simple comparison between pregnant and non-pregnant patients as the association between luteal P₄ levels and pregnancy displays a non-linear pattern.

In the early Humaidan study, a very low mid-luteal P₄ level resulted in an early pregnancy loss of 79% (Humaidan et al., 2005). Likewise, in the present study, the early pregnancy loss was significantly higher in the lower P₄ groups (<100 nmol/l) compared to the higher P₄ groups (>100 nmol/l). This suggests, that within the optimal range for early luteal P₄ levels (60–100 nmol/l), the high end (close to 100 nmol/l) may be more appropriate with regard to live birth compared to the low end (close to 60 nmol/l). The same is seen when looking at raw data from the early luteal phase.

A similar pattern with a lower and higher P₄ threshold as seen in our study has been reported in both studies of non-human species (Nogueira et al., 2004) and following embryo transfer in human frozen–thawed cycles (Yovich et al., 2015). Thus, it seems biologically plausible that thresholds exist for luteal P₄ inducing optimal endometrial receptivity in the fresh transfer cycle as well. The clinical consequence of ‘too low’ luteal P₄ levels could be to supply patients with additional exogenous P₄ in the attempt to ‘rescue’ endometrial maturation in time for embryo implantation. In contrast, patients with ‘too high’ luteal P₄ levels might be better off with segmentation and subsequent embryo transfer in a frozen–thawed cycle.

The key strengths of the present study include its prospective design, the large cohort of patients, the P₄ quantification in two separate periods of the luteal phase and the systematic approach to the handling of confounding factors. Using observational data, it is crucial to adequately address confounding in order to describe a valid causal association between exposure and outcome. Using DAGs, we systematically examined factors, which might introduce bias in the estimates of reproductive outcomes. By choosing this approach over the more traditional definition of confounding, we avoided the risk of collider-stratification bias with certain adjustments and were able to define the minimum set of factors to adjust for to remove the unwanted confounding in the final regression analysis (Howards, 2018).

Important for the results, the same vaginal luteal support regimen was applied in all cycles. Thus, the serum P₄ level measured reflects the endogenous P₄ production in addition to the vaginal supplement in all participants. The observed optimal levels for P₄ were in line with the previously proposed lower threshold for P₄ (Yding Andersen and Vilbour Andersen, 2014), and the optimal P₄ level was shifted upwards going from the early to the mid-luteal phase as would be expected from physiology. The participants included in the study were unselected broadening the generalizability of the findings. Moreover, the results seem to be applicable for both the GnRH agonist and the GnRH antagonist protocol as well as different ovulation trigger concepts.

It is well known that mid-luteal P₄ monitoring is challenged by the presence of fluctuations (Filicori et al., 1984). However, in the early luteal phase, P₄ levels exhibit a non-pulsatile pattern in both the natural and stimulated cycle (OPU + 2) (Filicori et al., 1984; Tannus et al., 2017). Furthermore, as previously demonstrated by our group, the magnitude of P₄ peaks in the mid-luteal phase (OPU + 7) following IVF and fresh embryo transfer is significantly correlated to the median P₄ level (Thomsen et al., 2018). Thus, very large P₄ fluctuations are predominantly seen in patients with a mid-luteal P₄ concentration exceeding 250 nmol/l whereas patients with P₄ levels <60 nmol/l display clinically stable P₄ values throughout the day. Consequently, a low P₄ value can be regarded as a ‘true low value’—also when measured in the mid-luteal phase following IVF treatment. Thus, in the present study, the potential misclassification of mid-luteal P₄ measurements due to P₄ fluctuations would be a differentiated misclassification of exposure (Kesmodel, 2018). In the present work, this misclassification type would take the estimate towards the null-value (no difference between groups) and, thus, lead to an underestimation of the true association between luteal P₄ levels and reproductive outcomes. Consequently, the significant decrease in reproductive outcomes seen in this study cannot be explained by information bias due to misclassification of exposure status.

Luteal P₄ monitoring in IVF cycles is a balance between the inconvenience of an additional procedure for the patient and the gain in terms of increased success rates. In the present study the combined prevalence of patients with low and high P₄ levels and, thus, patients with a potential need of treatment adjustment, was only 11% in the early luteal phase compared to 51% in the mid-luteal phase. Because all patients only had one blood sample drawn during luteal phase, we were not able to make a firm conclusion on the individual serum P₄ profiles going from the early to the mid-luteal phase. However, it seems that approximately 40% of patients will have either an insufficient P₄ increase from the early to the mid-luteal phase, thereby, ending in the low P₄ category (<150 nmol/l) on OPU + 5, or a too steep P₄ increase taking them into the high P₄ group (>400 nmol/l) on OPU + 5. Thus, P₄ monitoring in the early luteal phase may imply a risk of overlooking a substantial number of patients compared to mid-luteal monitoring.

Taken together, the optimal time for luteal P₄ monitoring seems to be a balance between pros and cons. Early luteal monitoring implies a high accuracy of P₄ measurements, and a theoretical possibility of increasing endometrial receptivity in low P₄ patients by use of additional exogenous P₄, but also implies a low prevalence of patients in need of treatment adjustments. In contrast, mid-luteal monitoring detects a higher proportion of patients who require a change in treatment to succeed, but also implies a higher risk of P₄ fluctuations in the high P₄ group of patients. Furthermore, an intensified luteal phase support to low P₄ patients may be less efficient late in the cycle.

With this study, we hope to re-open the luteal phase discussion and to draw attention to the possibility of improving the reproductive outcomes of the fresh transfer IVF cycle. Thus, we suggest to move away from the ‘one model fits all’ and towards a more individualized luteal phase support policy. In recent years, much effort has been invested in the improvement of embryo culture techniques to obtain high blastulation rates. However, the present study suggests that in 51% of patients, high-quality blastocysts are transferred to a sub-optimal endometrial milieu. A disrupted luteal P₄ level seems to decrease the chance of live birth following blastocyst transfer with as much as 16 percentage points going from 54% in patients with optimal mid-luteal P₄ levels to 38% in patients with P₄ >400 nmol/l. Whether the 51% of patients with sub-optimal P₄ levels would be better off with an adjustment of treatment—P₄ supplementation in case of low P₄ and segmentation in case of high P₄ levels—must be investigated in future RCTs.

In conclusion, our data suggest that the association between luteal P₄ levels and reproductive outcomes is non-linear—thus, both low as well as high luteal P₄ levels reduce the chance of a positive reproductive outcome. We hope that this study, which is the first in decades to explore the optimal luteal P₄ levels following IVF and fresh embryo transfer, will spark the interest of the topic and lead to future clinical trials within the research area.

Authors’ roles

L.H.T., P.H., C.Y.A., K.E. and L.B. designed the study. All co-authors participated in the conduction of the study. L.H.T. drafted the article and U.K., C.Y.A., K.E. and P.H. all contributed to the interpretation of data and critically reviewed the article. All co-authors approved the final article.

Funding

An unrestricted grant from Ferring Pharmaceuticals, Denmark, supported this study. Furthermore, grants from ‘Health Research Fund of Central Denmark Region’, ‘Research Foundation of the Hospital of Central Jutland’, ‘Research Foundation of A.P. Møller’, ‘Research Foundation of Aase & Ejnar Danielsen’, ‘Research Foundation of Dagmar Marshall‘, ‘Research Foundation of Dir. Jacob Madsen & Hustru Olga Madsen’, ‘Research Foundation of Fam. Hede Nielsen’ and ‘The Danish Medical Research Grant’ supported conducting this study.

Conflict of interest

L.H.T. received an unrestricted grant from Ferring Pharmaceuticals, Denmark to support this study. P.H. received unrestricted research grants from MSD, Merck, Gedeon Richter and Ferring Pharmaceuticals outside of this work as well as honoraria for lectures from MSD, Merck and Gedeon Richter outside of this work. U.K. received honoraria for lectures from MSD and Ferring Pharmaceuticals outside of this work. C.A. received unrestricted research grants from MSD, IBSA, and Ferring Pharmaceuticals outside of this work as well as honoraria for lectures from MSD and IBSA. H.O.E. and B.B.P. received an unrestricted research grant from Gedeon Richter outside of this work. K.E., L.B., D.P. and B.H. have no conflict of interest. The providers of funding were neither involved in the conduction of the study nor in the writing of the scientific report.