Olesen 2019.

Study characteristics
Methods	RCT, 2 arms, 304 randomised Setting: Denmark, 4 public fertility clinics Recruitment period: February 2014 ‐ December 2017
Participants	Criteria related to previous IVF failure: yes, one or more prior implantation failures, Inclusion criteria: eligible patients were IVF or intracytoplasmic sperm injection patients with one or more prior implantation failures, despite top‐quality embryo or blastocyst (19) transfer(s). Further inclusion criteria were regular menstrual cycle (28–32 days), age 18–40 years, and a body mass index (BMI) 18–32 kg/m2 Exclusion criteria: women with congenital uterine abnormalities, fibroids, or polyps were excluded, as were women with suspected hydrosalpinges and adenomyosis.
Interventions	Study group: Scratching was performed, using a Pipelle de Cornier (Laboratoires Prodimed) in the luteal phase before ovarian stimulation at cycle day 18–22 for the intervention group. The scratching was carried out with the patient lying in a lithotomy position and was performed once in each quadrant of the endometrium. Control group: no procedure
Outcomes	Reported in paper: live birth, clinical pregnancy, miscarriage, multiple pregnancy. Obtained by author correspondence: authors confirmed the multiple pregnancy rates reported in the per‐protocol analysis also applied to the ITT analysis. The authors also confirmed that the paper reports a mistake in the miscarriage rates reported, which should be 8 in the scratch arm and 13 in the control arm. The paper reports that quote: "There were no uterine infections, bleeding, or adverse events reported, besides a short pain during the endometrial scratching procedure" however it is not clear how many women this occurred in.
Notes	Trial registration: NCT01963819 (registered Oct 2013, registered prospectively) Additional concerns and comments: none Funding: Health Research Fund of the Central Denmark Author correspondence: yes, undertaken with Mia Olesen miaolsen@rm.dk Publication: full‐text.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Paper states quote: "participants were randomized into blocks of 10 for each participating clinic in a ratio of 1:1, according to an Internet‐based randomization list"
Allocation concealment (selection bias)	Low risk	Paper states that the quote:"randomization list that was sealed in consecutively numbered opaque envelopes" and the authors confirmed that study staff did not know the block size which was used
Blinding of participants and personnel (performance bias) All outcomes	High risk	No blinding was used quote:"The study was nonblinded, and no sham procedure was carried out in the control group"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Although the authors reported that some women experienced a short pain during the scratch; this data was not available in a useable format for this review.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Appears to be no attrition, this was confirmed by the authors
Selective reporting (reporting bias)	Low risk	Trial was registered prospectively and all outcomes were reported
Other bias	Low risk	‐