Endometrial Scratching: What the 1,364-Woman Trial Found

Whether deliberately injuring the endometrium before IVF improves the chance of a live birth. The theory was that a small controlled injury triggers a repair response making the lining more receptive to an embryo.

Lensen and colleagues randomised 1,364 women undergoing IVF — 690 to endometrial scratch, 674 to no intervention — and published the result in the New England Journal of Medicine in 2019.

• 1,364 women undergoing IVF.
• 690 randomised to endometrial scratch, 674 to no intervention.
• Primary outcome: live birth.
• Published in the New England Journal of Medicine, 2019.

Trial size matters here. A study of a few dozen women can miss a real effect; one of this size is powered to find a clinically meaningful difference if there is one.

Nothing. Live birth occurred in 26.1% of women in the scratch group and 26.1% in the control group — an adjusted odds ratio of exactly 1.00, with a confidence interval from 0.78 to 1.27.

That is an unusually clean null result. Most trials produce a small difference in one direction that then has to be interpreted; this one produced the same figure twice.

• Live birth: 26.1% scratch, 26.1% control.
• Adjusted odds ratio 1.00 (95% CI 0.78–1.27).
• Ongoing pregnancy: no significant difference.
• Clinical pregnancy, miscarriage, ectopic: no significant differences.

The confidence interval is worth reading rather than skipping. Ranging from 0.78 to 1.27, it rules out both a large benefit and a large harm — the effect, if any, is too small to matter clinically.

Carefully, because null results are routinely dismissed as 'inconclusive' when they are nothing of the kind. A trial that finds no difference is not a trial that failed — it is a trial that answered the question.

The distinction lies in the confidence interval. A small, underpowered study produces a wide interval compatible with almost anything, and genuinely settles little. This trial produced an interval from 0.78 to 1.27 around an odds ratio of 1.00, which excludes any effect large enough to matter.

• Wide interval + no difference = we do not know.
• Narrow interval + no difference = there is no meaningful effect.
• This trial is the second kind.
• 'More research is needed' is not the honest reading here.

The same logic cuts the other way when a trial reports a difference. A large effect with an interval crossing 1.00 has not established anything either, however impressive the headline number looks.

Because early, smaller studies suggested a benefit, and the biological reasoning was plausible. That combination is how most add-ons enter practice — a mechanism that makes sense, plus preliminary data that later fails to replicate.

Smaller trials are more likely to produce extreme results in either direction, and the ones showing benefit are more likely to be published and cited. By the time an adequately powered trial arrives, the practice is already established.

• Plausible mechanism: injury triggers a repair response.
• Early small trials suggested benefit.
• Small trials produce unstable results.
• Practice became routine before the large trial reported.

This is the pattern across IVF add-ons rather than a peculiarity of scratching. Our article on what the trials show across add-ons sets out how consistently it repeats.

That being offered endometrial scratching is a reasonable moment to ask what evidence supports it. The procedure is invasive, usually charged for, and the largest trial found no benefit.

• It is an invasive procedure, not a passive add-on.
• It is typically charged separately.
• The largest trial found no live birth benefit.
• A clinic recommending it should be able to say why, given that trial.

None of this means a clinic offering it is acting in bad faith. Practice changes slowly, and a clinician who saw apparent benefit before 2019 may still believe in it. It does mean the conversation should include the trial.

Our article on what is worth investigating after a failed cycle covers what ESHRE recommends instead, and what it specifically advises against.

One further point about invasive add-ons specifically. A blood test that turns out not to help costs money; a procedure that turns out not to help costs money, discomfort and a small procedural risk. The threshold of evidence before agreeing should scale with what is being done to you, and scratching sits at the more invasive end of the add-on menu.

It is also worth separating the intervention from the clinic offering it. A clinic still recommending scratching may simply not have revisited its protocols since 2019 — which is worth asking about directly, because the answer tells you how the clinic handles evidence generally.