All audits

Fertility problems: assessment and treatment (NG257)

ForenPrompt forensic audit of the NICE guideline.

Summary

NG257 is the first full revision of NICE's fertility guideline since 2013, with new evidence reviews on male factor testing, ovarian reserve, treatment add-ons, IVF access and fertility preservation, and many older recommendations carried forward with amendments. The newly reviewed add-on recommendations (endometrial scratch, receptivity testing, immunological agents, PGT-A, IMSI, PICSI) are internally consistent with their rationale. The consequential defects sit in the amended counselling numbers, the figures that support them, and places where a recommendation drops a qualifier its own rationale states. The audit found no critical errors and 9 major errors; 9 minor findings are withheld.

Scope and coverage

Audited: the full 91-page guideline, including all recommendations, Tables 1 and 2, Figures 1 and 2 (values read from rendered figures and treated as approximate), terms, research recommendations, rationale and impact sections, and update information. Not audited: the evidence reviews (A to Y), the methods document and the economic model, which were not supplied; findings that would turn on them (for example, the multiple-pregnancy results behind stimulated IUI in 1.38.3, and the cost-effectiveness inputs behind 1.39) were not raised. The document carries few formal references. Load-bearing sources verified: Dunson 2004 and Steiner and Jukic 2016 (both VERIFIED), plus the external authorities needed to test findings (UK CMO 2016 materials, EAU 2025, cabergoline SmPC). Not verified: Schwartz 1982, Zaadstra 1991, the HFEA dashboard extraction, and the WHO semen reference values in 1.17.1. Source retrieval was available. Adversarial pass: 10 findings tested; 1 withdrawn from the table and moved to the Recommendation as a revision item (cabergoline, see below); 3 narrowed (items 4, 5 and 8); 6 survived unchanged.

Critical and major errors

No critical errors. Nine major errors, ordered by clinical consequence.

Discussion

1. Alcohol advice (Rec 1.6.1). The full passage reads: "Inform women, and trans men and non-binary people with female reproductive organs who are trying to become pregnant that drinking no more than 1 or 2 units of alcohol once or twice per week and avoiding episodes of intoxication reduces the risk of harming a developing fetus. Advise them that, according to the Chief Medical Officer's guidelines on alcohol, the safest approach is to avoid alcohol altogether." The UK Department of Health announced in January 2016 that the earlier advice to limit intake to 1 to 2 units once or twice a week had been removed for clarity (4). The Guidelines Development Group report found no scientific basis for a limit below which alcohol will not harm the fetus and identified exactly this two-part construction as able to be read as inconsistent (5). The recommendation was amended in 2026 and still carries the withdrawn number. Followed as written, the first sentence reads as a permitted intake for people trying to conceive. Remedy: delete the quantified sentence.

2. Embryo transfer at 40 to 41 (Rec 1.49.6 vs 1.49.9). For ages under 37 and 37 to 39, 1.49.6 makes double embryo transfer conditional on the absence of a top-quality embryo. For ages 40 to 41 it says only "consider double embryo transfer". Rec 1.49.9 says to use single embryo transfer whenever a top-quality blastocyst is available. A clinician reading the age table will not find the qualifier, and double transfer of a top-quality blastocyst raises the chance of a twin pregnancy in the age group with the highest baseline obstetric risk. Remedy: carry the top-quality qualifier into the 40 to 41 line.

3. Y chromosome microdeletions (Rec 1.26.1). The rationale says the evidence "showed that the rate of successful surgical sperm retrieval in people with Y chromosome AZFa or AZFb microdeletions was very low" and elsewhere that retrieval is "almost impossible". Neither supports an unconditional prohibition. The EAU 2025 update states that complete deletions of the AZFa and AZFb regions predict poor prognosis and that TESE should not be attempted in these patients (6). Partial AZFb deletions are a different entity, and successful micro-TESE has been reported after one (7). Followed as written, a man with a partial deletion is refused the only route to a genetically related child. Remedy: insert "complete", specify that AZFbc and AZFabc deletions are included, correct the region names, and use one cut-off (less than 1 million/ml) in both places.

4. Chance of conception by age (Rec 1.4.1, Table 1, Figure 1). Rec 1.4.1 tells every couple with a woman under 40 that more than 80% will conceive within a year. The document's own Figure 1 (Steiner and Jukic 2016, verified, a prospective cohort of women aged 30 to 44 trying for 3 months or less) places the 38 to 39 estimate at about 71% (figure reading; the gap exceeds the 10% gridline precision) (3). The cited Table 1 source reports that at ages 35 to 39 the proportion not conceiving in 12 cycles rises from 18% to 28% as male partner age increases from 35 to 40, a result derived from modelled fecundability in natural family planning users with recorded intercourse timing (2). The 82% in Table 1 therefore describes a 35 to 39-year-old woman with a younger partner, not every couple. Adversarial pass: the population average across all women under 40 may exceed 80%, so the finding is narrowed to the unreconciled sources and the undisclosed partner-age dependence. Because 1.4.4 and 1.18.2 direct age-specific counselling, the discrepancy reaches individual women. Remedy: present the age-specific values with partner age and source population stated.

5. IUI cumulative rates (Rec 1.4.2, Table 2). Arithmetic, shown for the six donor IUI rows: 1 minus 0.43 squared equals 0.815 (printed 81%); 1 minus 0.46 squared equals 0.788 (78%); 1 minus 0.53 squared equals 0.719 (72%); 1 minus 0.69 squared equals 0.524 (52%); 1 minus 0.83 squared equals 0.311 (31%); 1 minus 0.94 squared equals 0.116 (11%). The 12-cycle column is a constant per-cycle projection. In the same table, observed ICI data fall 1 to 8 percentage points below the equivalent projection in every row (for example Zaadstra over 35 years: observed 55% against a projected 63%). The phrase "about half will do so with a further 6 cycles" is the assumption restated as a finding. The footnote discloses that the values are theoretical and may be lower in health-related fertility problems, and that stimulated and unstimulated IUI are pooled while 1.37 offers unstimulated IUI; the adversarial pass narrows the finding to the recommendation text, which carries none of these qualifiers. Remedy: label the projections and move the footnote's limitations into the recommendation.

6. IVF success by age (Figure 2). The figure is titled per embryo transfer; the method divides live birth events by embryos transferred. With double transfer, 1 live birth from 2 embryos is 100% per transfer and 50% per embryo. Pooling 33 years of HFEA data also mixes eras of different transfer practice, so the curve cannot be read as current success per transfer, and the caption's "in that year" does not describe a pooled estimate. Rec 1.18.2 directs clinicians to this figure for counselling. Remedy: correct the label and state the pooling, or present recent-year per-transfer and per-embryo rates separately.

7. Endometriosis (Rec 1.36.2). The rationale for 1.38 and 1.36 both say the mild endometriosis participants were not a distinct group and that the committee therefore did not name them in the recommendations; 1.36.2 then names endometriosis without stage limits. Evidence from unexplained infertility, with at most rASRM stage 1 to 2 participants mixed in, is indirect for moderate, severe or deep disease. The committee's own qualification (no deep endometriosis, patent tubes) is absent from the recommendation, and stimulated IUI with occluded tubes cannot work. Remedy: carry the stage, deep-disease and tubal-patency conditions into the recommendation text.

8. Miscarriage during the waiting period (Rec 1.16.7). Bullet 1 protects the 1-year window before investigation for people conceiving through intercourse. Bullet 3 protects "the 12 cycles of artificial insemination", which corresponds to IVF access (1.39.3), not investigation. The 6-cycle window before investigation (1.16.6) is not named. The adversarial pass accepts that 12 cycles is a real window, so the finding is narrowed from a wrong number to an omission. It matters because the people affected (female same-sex couples, single women, people unable to have intercourse) are those for whom the rationale says the principle should apply. Remedy: name the 6-cycle window.

9. Viral testing and donor insemination (Rec 1.19.1). Rec 1.19.1 routes donor insemination to 1.51.2, which is a list of indications for choosing donor sperm and contains no testing instruction. The guideline has no recommendation on viral testing for donor insemination recipients. Donor screening falls under HFEA licensing, but the recipient pathway is left undefined by a pointer to unrelated text. Remedy: correct the pointer or add the missing recommendation.

Recommendation

A formal correction is warranted. Items 1, 2, 3, 6 and 9 are factual or drafting errors that NICE can correct without new evidence review and belong in a single erratum: remove the withdrawn alcohol limit, reconcile the 40 to 41 embryo transfer line with 1.49.9, qualify the AZF prohibition as complete deletions and fix the region names and cut-off, relabel Figure 2, and repair the 1.19.1 cross-reference. Item 3 warrants prompt notice to NICE rather than waiting for the correction cycle, because a man with a partial AZFb deletion can be refused sperm retrieval under the current wording. Items 7 and 8 are drafting corrections that restore conditions stated in the document's own rationale and also belong in the erratum. Items 4 and 5 concern how counselling numbers are presented and are well suited to correspondence, with a request that the next revision state partner age, source population and the projection assumption beside each figure.

Two evidence-to-recommendation matters for the next scheduled revision, outside the table. First, Rec 1.32.1 replaces bromocriptine with cabergoline for people whose aim is conception; the rationale cites fewer side effects, lower cost and weekly dosing, but does not address that the UK product information advises women seeking pregnancy to stop cabergoline one month before intended conception because of its long half-life and limited in utero data (8). The substitution may be sound, but the rationale should confront that labelled precaution. Second, Recs 1.19.4 and 1.19.5 [2013] still require intercourse to be limited to ovulation and offer sperm washing above 50 copies/ml; the 2026 BASHH/BHIVA guideline advises that additional measures such as partner PrEP are not indicated unless the partner with HIV is off treatment or has a detectable viral load (9). These recommendations were not reviewed in this update and are due for it.

All audits