NICE backs a saliva test for endometriosis in revised draft guidance. What does the evidence show?

NICE has released revised draft guidance on Endotest, a saliva-based microRNA test. It says the NHS can use the test during a four-year evidence-generation period.¹ Three other tests (DotEndo, Endomkit, and EndoSure) require further research before NHS funding can be approved. The guidance is not final. Consultation closes on 5 October 2026.

The draft positions the test late in the existing pathway. It applies only when endometriosis is suspected. The abdominal and pelvic exams must be normal. The transvaginal or transabdominal ultrasound should be negative, inconclusive, declined, or unsuitable. Results still need a further appointment to discuss, and the sample goes to the company’s central laboratory. The intended age range is 18 to 43 and the listed cost is £1,381 per person.¹

The accuracy data come from a multicentre external validation study of 971 symptomatic women across 17 French centres. It was published in NEJM Evidence in October 2025. Ziwig, the company that developed the test, funded it.² Endometriosis was confirmed through imaging, laparoscopy, or both. All control participants also underwent laparoscopy. Against that reference standard the 109-microRNA signature had a sensitivity of 97.3% (95% CI 96.4 to 98.0) and a specificity of 94.1% (95% CI 91.0 to 96.4). The negative likelihood ratio was 0.03. Among those with surgical confirmation, the saliva test misclassified 4.6% compared with 27.2% for imaging. That comparison covers the surgically confirmed subgroup only. It does not show how the test performs against ultrasound in an unselected primary care population. An earlier interim analysis of the first 200 participants from the same cohort reported similar figures.³

So why does NICE describe accuracy in primary care as unclear? The central question is whether results from a specialist population carry over to primary care. Prevalence changes predictive values. Differences in patient characteristics, disease spectrum and severity may also change sensitivity and specificity. Endometriosis prevalence in the validation cohort was 77%, which is what you get when you recruit from specialist centres. In that setting the positive predictive value was 98.2%. Apply the same sensitivity and specificity to a lower-prevalence population and the arithmetic shifts. At 30% prevalence the PPV falls to roughly 88%. At 10% it is closer to 65%, meaning around one in three positive results would be false. Negative predictive value stays above 98% in every scenario. Those figures are illustrative calculations, not study data,* and the true prevalence in the post-ultrasound primary care group NICE describes is unknown. The committee’s point is that nobody has measured it yet.¹

That is also what separates Endotest from the other three tests. Endomkit’s validation cohort of 77 people reported sensitivity of 46.2%. DotEndo rests on a single prospective study of 100 women. EndoSure showed a sensitivity of 91% to 96% in two centres. However, these estimates came from AI. The external assessment group rated all these studies as having a high risk of bias. Endotest was the only technology with a published external validation study.¹

Two further gaps deserve attention. No study for any of the four tests reported clinical outcomes. Diagnostic accuracy does not prove the test will shorten time to diagnosis, reduce unnecessary laparoscopy, improve quality of life or cut resource use. Those are the outcomes the four-year period is meant to capture. They are not benefits already shown. And the economic case rests on long-term modelling, not observed outcomes. The committee noted modelled QALY gains of 0.03 to 0.11 over a 60-year horizon. Those gains depend on assumptions about the quality-of-life benefit of receiving a diagnosis.¹

The clinical experts advising NICE said that a negative result shouldn’t prevent referrals if endometriosis is still suspected.¹ This aligns with the predictive values mentioned earlier. A negative result provides useful information, while a positive result depends more on the specific situation. Neither of these results changes the referral criteria in the NICE endometriosis guideline, which already states that normal examination and ultrasound do not exclude the condition.⁴

References

1.   National Institute for Health and Care Excellence. Technologies for diagnosing endometriosis in primary care: early-use assessment. Guidance assessment consultation document HTG10877. September 2026. https://www.nice.org.uk/guidance/indevelopment/gid-htg10877

2.   Bendifallah S, Roman H, Suisse S, et al. Validation of a saliva microRNA signature for endometriosis. NEJM Evid 2025;4(11):EVIDoa2400195. https://doi.org/10.1056/EVIDoa2400195

3.   Bendifallah S, Dabi Y, Suisse S, et al. Validation of a salivary miRNA signature of endometriosis: interim data. NEJM Evid 2023;2(2):EVIDoa2200282. https://doi.org/10.1056/EVIDoa2200282

4.   National Institute for Health and Care Excellence. Endometriosis: diagnosis and management. NICE guideline NG73. Last updated 11 November 2024. https://www.nice.org.uk/guidance/ng73

* Predictive values at 10% and 30% prevalence were calculated by Medicine Central using Bayes’ theorem, applying the sensitivity (97.3%) and specificity (94.1%) reported in the ENDOmiRNA external validation study² to hypothetical populations. PPV = (sensitivity × prevalence) ÷ [(sensitivity × prevalence) + ((1 − specificity) × (1 − prevalence))]. NPV is calculated analogously. At 10% prevalence the calculated PPV is 64.7% and NPV 99.7%. At 30%, PPV is 87.6% and NPV 98.8%. The 10% figure is NICE’s cited general population prevalence of endometriosis. It is not an estimate for symptomatic women who meet the proposed post-ultrasound testing criteria. The 30% figure is an arbitrary intermediate value. Neither represents the measured prevalence in the primary care population described in the draft guidance, which has not been reported. The 77% figure and its associated PPV of 98.2% and NPV of 91.3% are as published.²