For each woman an appointment was scheduled 7–14 days before the planned start of COS. All women had been using the OCP (ethinyl estradiol 30 mcg + levonorgestrel 150 mcg) since their last menstruation, for at least 10 days before the appointment. On the scheduled day, the participants were prepared for a gynecological examination and the uterine cervix was identified using a vaginal speculum. Only then was the sealed allocation envelope opened.
In the intervention group, endometrial biopsy was performed once with a pipelle de Cornier® (Laboratoires Prodimed, Neully-En-Thelle, France). The pipelle was introduced gently through the cervix up to the uterine fundus. The piston was then drawn back to the end of the biopsy cannula until it self-locked, creating a negative pressure. Aiming to cover the entire endometrium, the examiner applied regular back-and-forth movements (2–4 cm) while rotating the sampler, over the whole uterine cavity, during a period of 30 s. In those cases in which the pipelle suction orifice was clogged before the 30-s period had ended, we restarted the procedure using another pipelle and counted another 30 s. The obtained specimens were discarded. The control group was submitted to a sham procedure comprising drying of the cervix with gauze for 30 s. All procedures were performed by the same physician (W.P.M.). Thus, all women, regardless of intervention, underwent a gynecological examination comprising speculum examination and manipulation of the cervix for 30 s. The difference was only in the type of uterine manipulation: endometrial biopsy in the intervention group, and drying of the cervix with gauze in the control group. In both groups a visual analog scale (a 10-cm horizontal line anchored at one end with the words ‘no pain’ and at the other end with the words ‘worst pain imaginable’) was given to participants just after the procedure. They were asked to mark a point on this scale that would represent the pain associated with the procedure. The marked point was measured using a ruler by another researcher (C.O.N.).
On the planned COS start day (day 2–3 of the menstrual cycle), women underwent transvaginal ultrasound examination; COS was started only if no follicle ≥ 10 mm in diameter was observed or if estradiol levels were < 50 pg/mL in the presence of a follicle ≥ 10 mm. At the discretion of the physicians and participants, one of three different regimens was selected to induce COS: a) clomiphene citrate (CC) (Indux™, EMS, Hortolandia, Brazil, 100 mg/day for 5 days) + human menopausal gonadotropin (hMG) (Menopur™, Laboratórios Ferring Ltda, São Paulo, Brazil, 150 IU/day, every other day when using CC, and every day after CC was stopped) + gonadotropin-releasing hormone (GnRH) antagonists (Ganirelix (Orgalutran™), MSD Brasil, Sao Paulo, Brazil, 0.25 mg/day, started when a follicle ≥ 14 mm was observed); b) recombinant FSH (Puregon™, MSD Brasil, Sao Paulo, Brazil or Gonal™, Serono Produtos Farmaceuticos Ltda, São Paulo, Brazil, 150–300 IU/day) + GnRH antagonist (Ganirelix, 0.25 mg/day, started when a follicle ≥ 14 mm was observed); or c) recombinant FSH (150–300 IU/day) + GnRH agonist (leuprolide acetate (Lupron™), Abbott Laboratórios do Brasil Ltda, Rio de Janeiro, Brazil, 0.5 mg/day, started 5 days before stopping OCP).
Transvaginal ultrasound monitoring of follicle development was scheduled to start after 5 days of COS, repeating every other day. 3D-PD ultrasound examination was performed when we observed at least one follicle ≥ 17 mm. Final follicular maturation was triggered on this day or 1 day later, using recombinant human chorionic gonadotropin (hCG, 0.25 mg, single injection) between 22:00 and 23:00 hours. Oocyte retrieval was performed 34–36 h after the recombinant hCG injection.
All 3D-PD ultrasound examinations were performed using either a Voluson 730 or a Voluson E8 (GE Healthcare Austria GmbH & Co, Zipf, Austria) ultrasound machine equipped with a vaginal probe. They were conducted by a single observer (W.P.M.), using a standardized technique: briefly, the uterus was located and power Doppler ultrasound was applied using predetermined settings which were kept constant throughout the study (two-dimensional settings: THI, CRI and SRI, turned off; main power Doppler settings: Filter, low1; PRF, 0.6KHz; gain, 0.0; Quality, normal; Sub power Doppler settings: Freq, mid; Flow res, mid1; Smooth, 5/6; Ensemble, 12; Line Den, 7; PD map, 5 (Gently Color, on); Balance, G > 200; Artifact, on; L. Filter, 2). The resulting truncated sector was adjusted and moved to cover the endometrium in the longitudinal plane of the uterus. The woman was asked to breathe gently and to remain as still as possible and every effort was made by the observer to limit inappropriate movements of the transducer. A 3D-PD dataset of the uterus was then acquired using a sweep angle of 90° (‘high 1’ sweep mode) in order to acquire the entire endometrium. The 3D-PD datasets were saved to the hard drive of the ultrasound machine and were analyzed by the same observer (W.P.M.) using 4D View and Virtual Organ Computer-aided AnaLysis (VOCAL, GE Medical Systems, Zipf, Austria). Endometrial thickness was measured after the dataset had been manipulated to produce a standardized multiplanar view of the uterus. The endometrium was then defined using a 15° rotation step in this view, and its volume was calculated by the software. The volume was analyzed using the ‘color-off’ mode, which facilitates morphological visualization by removing Doppler signals from the display. 3D-PD data within the endometrium were quantified using the histogram facility to generate values for the vascularization index (VI), flow index (FI) and vascularization flow index (VFI).
One to three embryos were transferred 48–72 h after oocyte retrieval. The luteal phase was supported by the administration of micronized progesterone (600 mg/day), which was stopped if the pregnancy test (serum β-hCG test performed 14 days after embryo transfer) was negative or after week 12 of pregnancy.
We estimated a clinical pregnancy rate in the control group of 30%[30, 31]. To have an 80% power to detect a relative increase of 50% (intervention group ≥ 45%), with a significance level of 5%, we would need 162 subjects per group, i.e. a total of 324 participants. Initially we planned to include 400 participants, because we intended to exclude women who had their treatment cycle canceled for any reason (estimating that about 20% of participants would be excluded). However, after starting the study we realized that this would lead to an unacceptable risk of bias (e.g. if women submitted to endometrial scratching have more chance of cycle cancelation, this would go unnoticed), and we changed the analyses in order to respect the ITT principle. Thus, we reduced our total sample to 324, which was the planned number without exclusions.
One of the investigators (C.O.N.) performed the planned interim analyses after inclusion of 60 participants, and then after each inclusion of 30 participants. We planned these interim analyses in order to discontinue the study when a significant difference in clinical pregnancy rate between groups was observed; this was with the aim of reducing the number of women submitted to uncomfortable sham procedures. However, this strategy is acknowledged as being the main limitation of our study, since stopping a trial early causes an increased risk of overestimating the treatment effect.
Randomization and blinding
For allocation of the participants we used a computer-generated random sequence of numbers in blocks of 30 (each block having 15 numbers assigned to intervention and 15 to control). The allocation was sealed in consecutively numbered opaque envelopes and an envelope was assigned as the participant entered the study; however, sealed envelopes were only opened just before the procedure, to ensure allocation concealment.
One researcher (W.P.M.) was responsible for enrolling the participants, and for performing the endometrial scratching and sham procedures and 3D-PD evaluation. Another researcher (C.O.N.) made the random allocation list, and put the papers with the assigned group inside the opaque numbered envelopes, sealing them to conceal allocation. Participants, care providers and other outcome assessors were blinded; only one researcher (W.P.M.) was aware of the allocation.