Evidence and Limitations

What the existing literature reports

Several papers report high agreement between NLS assessment and a conventional reference standard, across different conditions and different groups. The figures are consistent enough to be interesting and weak enough to be unusable.

Why those numbers cannot be relied on

This is not an accusation of dishonesty. It is a description of what the papers did and did not do, judged against the reporting standards now expected for a diagnostic accuracy study.

• The samples are small. At 30 confirmed cases, an observed sensitivity of 93% carries a 95% confidence interval of roughly 78–99%. That width is wide enough to be compatible with a genuinely useful test and with a mediocre one.
• The operators were not blinded. Where the person running the index test knows the referral reason or the existing diagnosis, the result is not an independent assessment.
• Endpoints were not pre-registered. Without a protocol fixed before the data is examined, the reported endpoint can be chosen after the fact from several that were measured.
• The cases were already diagnosed. Confirming findings in a population known to have the condition is not the same as performance in a representative screening population, and the numbers are not transferable.
• Most work came from associated groups. Much of the literature originates with groups connected to the equipment’s development. That does not make it wrong; it does mean independent replication is missing.
• None of it tested a remote workflow. The studies describe in-person assessment with specific hardware. Nothing in them establishes anything about a remote service, a photograph, a sample or a written report.

The "93% accuracy" claim, specifically

A figure above 93% circulates widely in this industry. Its apparent source is a 2013 manufacturer-associated report stating that specified Metatron configurations confirmed 662 of 705 previously diagnosed pathologies.

QMH does not publish that figure as a claim, and the reason is worth stating in detail. The report was not an independent blinded validation. It concerns older hardware and software configurations. The cases had already been diagnosed conventionally. A separate annex in the same document reports 92.7% sensitivity against 37.6% specificity and 85.6% accuracy, which shows how much the headline depends on which endpoint is chosen. And none of it validates the current service, the practitioner interpretation, the written report, or any remote input.

What would actually settle the question

A study capable of answering this is not complicated, it is just rigorous. The design is standard for diagnostic accuracy work and the reason it has not been done is resourcing, not difficulty.

• A single, well-defined condition with an objective reference standard.
• The operator blinded to history, referral reason and laboratory result.
• The reference reader blinded to the index result.
• Primary endpoint — sensitivity and specificity — declared in writing before any data is examined.
• Analysis by a statistician independent of the sponsor.
• Pre-registration, reporting to STARD guidelines, and publication regardless of outcome.

What QMH is funding

QMH has proposed a blinded diagnostic-accuracy feasibility study to a university medical centre, fully funded by QMH, non-interventional, with the host institution holding the data, controlling the methodology, conducting the analysis and retaining publication rights with no sponsor veto.

The preferred first target is primary hypothyroidism, because TSH and free T4 give an unambiguous binary reference. At feasibility scale the study is not designed to resolve the question either. It is designed to establish whether the signal exceeds chance under blinding, and to size a definitive study of roughly 200 to 400 participants if it does.

A null result is a real and publishable outcome. That is the point of funding it this way.

What QMH does not claim

• No claim that the service diagnoses, detects, screens for or rules out any condition.
• No accuracy percentage attached to the QMH service.
• No claim of equivalence or superiority to any medical test.
• No claim that a photograph, hair, saliva or voice sample carries diagnostic information.
• No claim that the service prevents, treats, cures or reverses anything.
• No testimonial presented as evidence that the service caused a health outcome.

Questions people ask before booking