NIPT prenatal screening, by the numbers

Written and reviewed by a practicing emergency physician · Updated July 2026

A cell-free DNA test comes back "high probability" for trisomy 21. The assay behind it clears all but about 4 in every 10,000 unaffected pregnancies — a specificity near 99.96%, among the best of any screening test in medicine. Read on its own, that number makes a positive result feel close to settled. Work the arithmetic at the prevalence of a general screened population, though, and roughly 1 in 7 of those "positive" results belongs to a pregnancy that does not have the condition at all. Both things are true at once, and holding them together is the point of this page.

What cfDNA screening measures

Cell-free DNA screening — also called noninvasive prenatal testing, or NIPT — reads fragments of placental DNA that circulate in a pregnant person's blood, usually from about 10 weeks. By counting how much chromosome-21 material is present relative to the rest, it estimates the probability that the fetus has trisomy 21 (Down syndrome). It is a probability estimate, not a direct look at the fetal chromosomes — a distinction that turns out to matter enormously downstream.

Three numbers drive everything that follows:

• Sensitivity ≈ 99.7% — of pregnancies affected by trisomy 21, about 997 in 1,000 are flagged (Gil and colleagues' 2017 meta-analysis, pooled across cohort studies).
• Specificity ≈ 99.96% — of unaffected pregnancies, all but roughly 4 in 10,000 are correctly cleared.
• Prevalence ≈ 0.25% — in a broad screened population, about 1 pregnancy in 400 is affected (the CARE study found roughly 5 in 1,952). This is a pre-test probability averaged across maternal ages; it is not a constant, and it climbs steeply with age.

A test that catches 99.7% of cases and clears 99.96% of the unaffected sounds all but definitive. What a positive result means, though, depends on that third number — and 0.25% is very small.

The 2×2, worked by hand

Screen 10,000 pregnancies at those figures and every cell of the table is fixed. Split the group first: 10,000 × 0.0025 = 25 affected, and 9,975 unaffected.

Now apply the test to each group. Of the 25 affected, 25 × 0.997 = 24.9 — so essentially all 25 are flagged (true positives), with fewer than one missed. Of the 9,975 unaffected, 9,975 × 0.0004 = 3.99, so about 4 are flagged anyway (false positives) and the other 9,971 are correctly cleared (true negatives).

Follow the positive column — the same table the calculator's 2×2 view fills in live. About 29 pregnancies receive a high-probability result, and 25 of them are truly affected. The share of positives that are real is the positive predictive value (PPV):

PPV = 25 ÷ 29 ≈ 0.862 — about 86%.

Computed from the exact rates before anyone is rounded to a whole pregnancy — the way the calculator does it — the figure is 86.2%. Either way, roughly 1 in 7 positive screens (about 4 of the 29) is a false positive.

Now the negative column, which is the genuinely reassuring one. About 9,971 pregnancies get a low-probability result and very nearly all of them are truly unaffected:

NPV = 9,971 ÷ 9,971 ≈ 99.99%.

At this scale, fewer than one affected pregnancy hides in that column — the flip side of a 99.7% sensitivity. A low-probability cfDNA result is among the most reassuring numbers in prenatal screening; a high-probability result, at population prevalence, is better read as a strong reason to look closer than as an answer.

Why a 99.96%-specific test still raises false alarms

The reason is not a flaw in the assay; it is arithmetic. The unaffected group (9,975) dwarfs the affected group (25) by nearly 400 to 1. Multiply that large group by even a tiny false-positive rate — 0.04% — and you still get about 4 false positives, a real fraction of the roughly 25 true ones. When a condition is rare, the size gap between the two groups does more to shape a positive result than the test's accuracy does. That gap between how accurate a test is and what a positive result actually means is the base-rate fallacy, and it does not vanish just because a test is excellent.

It is worth naming what cfDNA gets right here. Older combined first-trimester screening flagged unaffected pregnancies far more often — on the order of a few percent — whereas cfDNA's false-positive rate is about 0.04%, so it sends far fewer people toward invasive testing and its predictive value at the same prevalence is much higher. In likelihood-ratio terms, a positive shifts the odds by a factor of roughly 2,500 (LR+ = 0.997 ÷ 0.0004). "Much better than the old screen," though, is not the same as "diagnostic."

Maternal age moves the pre-test probability

That 0.25% is an average, and the pre-test probability underneath it is not one number. The chance that a pregnancy is affected by trisomy 21 rises with maternal age — gradually through the twenties and thirties, then more sharply — so the same cfDNA result carries different weight for different people.

To see the mechanism, take an illustrative higher pre-test probability of 1% (used here only to show the effect — roughly the neighborhood associated with maternal age around 40). Screen 10,000 again: 100 affected, 9,900 unaffected. The test flags 100 × 0.997 ≈ 100 of the affected and 9,900 × 0.0004 ≈ 4 of the unaffected. Positives ≈ 104:

PPV = 100 ÷ 104 ≈ 96%.

The identical result that carried roughly 1-in-7 odds of a false positive at population prevalence now carries closer to 1 in 25. Nothing about the assay changed — same sensitivity, same specificity — only the population did. That single lever, pre-test probability in, post-test probability out, is what the calculator's Bayesian view makes visible: a positive cfDNA result multiplies the pre-test odds by that enormous likelihood ratio, but it starts from the odds you bring to it.

A screen, not a diagnosis

This is the practical reason professional guidance draws a hard line. The American College of Obstetricians and Gynecologists and the Society for Maternal-Fetal Medicine, in ACOG Practice Bulletin No. 226 (2020), are explicit that cfDNA is a screening test — not a diagnostic one — and that a high-probability result should be confirmed with diagnostic testing before it is treated as fact. Diagnostic testing here means chorionic villus sampling (CVS) or amniocentesis, which sample placental or fetal cells directly and examine the chromosomes themselves. The same guidance states plainly that screening results should not be the sole basis for an irreversible decision. The 1-in-7 false positives at population prevalence is the concrete reason that line exists.

The stakes on both sides

Confirmation is not free of cost, which is exactly why the decision deserves real numbers rather than a reflex. Amniocentesis carries a small procedure-related risk of miscarriage — on the order of 1 in 300 to 1 in 500; a 2019 meta-analysis by Salomon and colleagues placed the excess risk near 1 in 333. So a diagnostic test done to resolve an uncertain screen carries its own low but nonzero risk.

Set the two considerations side by side, which is all this site tries to do. On one side, a low-probability cfDNA result is extraordinarily reassuring, and cfDNA's high specificity means far fewer people face an invasive test than under older screening. On the other, at population prevalence about 1 in 7 high-probability results is a false alarm, so acting on an unconfirmed screen — in either direction — means acting on a result that has a meaningful chance of being wrong. The calculator's outcome panel places a benefit count next to a harm count on the same population so the trade-off is visible at once rather than one number at a time; the wider catalog of these trade-offs lives in screening harms and biases.

None of this points to a single right answer, and this page does not offer one. What the arithmetic offers is a clearer picture of what a cfDNA result is — a very good probability estimate, not a verdict — so that whatever comes next can be weighed on the real numbers, with a clinician or genetic counselor, rather than on a percentage read out of context.