NIPT accuracy: What a ‘high-chance’ result really means

Posted in partnership with London Pregnancy Clinic

NHS screening rules require that a higher-chance NIPT result be explained as provisional and confirmed by diagnostic testing. They do not require anyone to say how provisional, and the honest answer varies from near-certain to close to a coin toss.

Non-invasive prenatal testing occupies an unusual position in antenatal care. It is accurate enough to have displaced a great deal of invasive testing, and not accurate enough to replace it. Everything difficult about how results are communicated sits in that gap.

A declaration of interest. London Pregnancy Clinic provides NIPT screening privately. The argument below is partly a criticism of how the private market sells it.

Within the NHS, the pathway is contingent rather than universal.

NIPT is offered after a higher-chance result from the combined or quadruple test, meaning a chance between 1 in 2 and 1 in 150 at term, and is available up to 21 weeks and 6 days.

The programme is explicit that it is a screening test.

A higher-chance result must be explained as not confirming the condition, an appointment must follow within three working days, and diagnostic testing by CVS or amniocentesis should be completed within three working days of the woman receiving her result.

That is a well-constructed pathway, and the requirement to state plainly that the result is provisional is the right one.

What the requirement does not cover

The rules require a category to be communicated. They do not require a probability.

A woman told she has a higher-chance NIPT result has been given a classification that covers an enormous range.

The probability that the result reflects a genuine finding depends on three things, and only one of them is the test.

The first is the condition. NIPT performs best for trisomy 21. Both the national programme and Antenatal Results and Choices, the UK charity that supports parents through exactly these results, note that it is less reliable for Edwards’ and Patau’s syndrome.

The programme goes further, warning that for women whose combined result was between 1 in 2 and 1 in 10, NIPT is less accurate for those two conditions specifically and an in-depth discussion is needed.

For the microdeletion panels sold in the private market, the picture is considerably weaker again.

The second is prior probability. This is the part that is routinely skipped, because it is counterintuitive.

The same test, performing identically, produces a different meaning depending on how common the condition is in the population being tested.

Where the condition is rarer, more of the positives are false, because the pool of true cases is small relative to the pool of false alarms.

Nothing about the laboratory changes. The arithmetic changes.

The third is what the woman was told to expect, which is frequently that NIPT is over 99 per cent accurate. That figure usually describes sensitivity, the proportion of affected pregnancies the test identifies. It is not the number she needs.

The number she needs is the probability that her result is correct, and those are different quantities that happen to be expressed in the same units.

Where this gets worse rather than better

In the NHS pathway, prior probability is already elevated by design, because NIPT is only reached after a higher-chance combined or quadruple result.

That makes the test perform relatively well in context.

The private market inverts this. NIPT is sold first-line, to anyone, at any age, with no preceding risk assessment.

A 25-year-old with no risk factors and no prior screening is exactly the situation in which a positive result is most likely to be wrong, and it is a situation the NHS pathway largely avoids and the commercial market actively creates.

Clinics selling first-line NIPT, this one included, therefore carry an obligation the NHS does not: to explain the consequence of testing a low-prevalence population before taking payment, rather than after a frightening result arrives.

The marketing in this sector does not generally do that.

Headline accuracy figures are quoted, prior probability is not mentioned, and the gap is left for the patient to discover at the worst possible moment.

Give the number

The practical fix is unglamorous. Alongside the category, give the probability.

The calculation is not difficult.

It requires the condition, the prevalence at that woman’s age, and the published performance characteristics of the specific test used, and it produces a figure that can be stated in a sentence.

A positive predictive value calculator illustrates the scale of the variation: modelled for trisomy 21, the probability that a higher-chance result is correct runs at roughly 97 per cent at age 40, around 91 per cent at 35, and about 72 per cent at 25.

Those are illustrative model outputs rather than measured rates, but the direction and magnitude are the point. At the younger end, roughly one in four positives is a false alarm.

A woman deciding whether to undergo an invasive procedure, and how to feel for the three days before it, is in a materially different position depending on which of those numbers applies to her.

Telling her only that her result is “higher chance” withholds the one piece of information that distinguishes her situation from everybody else’s in the same category.

It is worth noting that neither the national programme’s communication standards nor ARC’s own guidance quantifies how often a higher-chance result turns out to be incorrect.

This is not negligence. It reflects a genuine reluctance to hand someone a number that is model-derived, test-specific and easily misread.

That caution is understandable, but the alternative is not neutrality. A woman given no number does not hold no estimate.

She holds whichever one she inferred from the word “high” and from the accuracy figure she was quoted, and that estimate is almost always more pessimistic than the truth.