If NIPT Detects Trisomy 21 About 99% of the Time, Why Is ‘High Risk’ Still Not a Diagnosis?

Geoffrey Chen

When Numbers Start Making Decisions · Season Five, “Things That Have Not Happened Yet” · Article 8

1. The easiest way to mishear 99%

A pregnant woman receives a high-risk result from non-invasive prenatal testing, or NIPT. The material says the detection rate for trisomy 21 is about 99%. Under stress, that can sound like “there is a 99% chance this fetus has Down syndrome”.

Detecting about 99% of affected pregnancies is sensitivity, not the posterior probability for this positive result. NSW Health policy describes a detection rate around 99% for trisomy 21 and a screen-positive rate below one per cent. Because trisomies 21, 18 and 13 are relatively uncommon, positive predictive values vary; the policy gives a broad range rather than 100%. NSW Health: Prenatal Screening and Diagnostic Testing

This is not a statistical technicality. It determines how much real-world authority a screen can carry. NSW policy requires a high-risk result to be confirmed using diagnostic testing such as chorionic villus sampling, or CVS, or amniocentesis. Termination should not be offered solely on the basis of an NIPT result.

2. Most analysed DNA is not taken directly from the fetus

NIPT analyses cell-free DNA fragments in maternal blood. NSW guidance explains that most fragments are maternal and a smaller proportion originates from the placenta, forming the fetal fraction used in analysis. The laboratory compares relative representation of chromosomal material to estimate whether a pregnancy is more likely to involve particular aneuploidies.

This is a highly effective screen without an invasive risk to the fetus, but it is not direct observation of every fetal cell. In uncommon cases placental and fetal chromosomal findings differ. Maternal biological factors, low fetal fraction and laboratory analysis may also affect the result.

NIPT is generally available from about ten weeks and commonly screens for trisomies 21, 18 and 13 and some sex-chromosome conditions. Commercial products differ. A low-risk result refers only to the conditions and limitations of the panel; it cannot guarantee the absence of other genetic, chromosomal or structural conditions.

A no-result outcome also contains information. A small percentage of tests fail because fetal fraction or laboratory quality requirements are not met. The next step may be repeat sampling, another screening approach or diagnosis. No result is not low risk and should not be administratively treated as reassurance.

3. Sensitivity, screen-positive rate and positive predictive value

A simplified example helps. Imagine a condition present in ten of 10,000 pregnancies. A test detects about 99% of affected cases and produces a small number of false positives among everyone else. Even a strong test can yield a positive group in which false positives are substantial because unaffected pregnancies greatly outnumber affected ones.

Sensitivity asks: if the condition is present, how often does the test detect it? Positive predictive value asks: given this positive result, how often is the condition actually present? The second depends on prevalence, age, prior screening, ultrasound and the exact test.

Advertising a 99% detection rate without explaining positive predictive value lets method performance acquire individual diagnostic authority. Emphasising false positives alone would also understate NIPT’s benefit in reducing unnecessary invasive testing. Accurate communication must keep both truths visible.

Baseline prevalence is not a minor adjustment made after the “real” test result. It is part of what an individual result means. The same analytical performance can produce different predictive values in populations with different ages or prior findings. Ultrasound and earlier screening may also alter the relevant starting risk. A report that supplies a category without helping the clinician place it in this context leaves the most consequential arithmetic to the anxious recipient. The laboratory should provide condition-specific performance and an appropriate predictive estimate, while the clinician should explain why that estimate is not a diagnosis.

4. What may “high risk” legitimately trigger?

A high-risk NIPT result can properly trigger prompt information, genetic counselling, specialist assessment and an autonomous decision about diagnostic testing. CVS or amniocentesis obtains placental tissue or amniotic fluid for diagnostic-level cytogenetic or molecular evidence. These procedures also involve timing, a small procedural risk and personal values requiring professional explanation.

NSW policy requires information and support so choices are informed. Victorian guidance likewise says prenatal screening is voluntary and distinguishes screening from diagnosis.

Offering diagnosis must not become coercion. Some people want certainty to make decisions or arrange medical and family support; others decline further testing according to their values. A genuine choice requires timely counselling, diagnostic access and accurate, non-discriminatory information about life with disability.

An irreversible decision needs a thicker evidence structure. High-risk screening can be powerful while still not being proportionate to diagnostic consequences on its own.

5. A number changes the experience of pregnancy

Before testing, pregnancy may be experienced through bodily change and ultrasound. A high-risk result reorganises the future into probability, waiting and choices. Even when diagnosis later excludes the condition, the intervening anxiety is real.

Success cannot therefore mean only more detection. Programs and clinicians should evaluate pre-test understanding, report language, time from positive result to counselling and diagnosis, handling of no results and equality of access.

In Australia, NIPT often involves out-of-pocket cost, and products differ across services. If only those able to pay obtain more accurate screening while public diagnostic services triage through results, the technology may amplify existing inequality.

Expanded panels raise another problem. The rarer the condition, the lower positive predictive value may become even when analytical performance appears strong. More reportable calculations can create more uncertain findings and heavier counselling requirements. The fact that a condition can be screened does not by itself show that it should be included.

6. Error and review occur at several levels

Prenatal screening errors are not all laboratory mistakes. Sample or identity errors concern data accuracy. Whether placental cell-free DNA measures the fetal condition concerns validity. Which probability receives the “high risk” label is a reporting threshold. What may happen after that label concerns decision proportionality.

Effective review checks gestational age, fetal fraction, test scope, laboratory quality, ultrasound findings and baseline risk and allows an independent clinician to interpret them. If diagnosis conflicts with NIPT, possible causes should enter the quality system rather than being dismissed with “screening can be wrong”.

Reports should avoid the omnibus claim “99% accurate”. Performance varies by condition, product and population. Sensitivity, specificity, positive predictive value and no-result rate answer different questions. A percentage without its denominator provides fear or reassurance without understanding.

7. A possible future life cannot be completed by one probability

The result forms among maternal biology, placental DNA, laboratory algorithm, population prevalence and a reporting threshold. It is not a complete description retrieved from a future child.

Clinical knowing is partly delegated to laboratories and statistical systems. Delegation expands visibility without eliminating human responsibility. A system can label risk without beliefs and without understanding a family’s values. Clinicians and pregnant people must still undertake explanation and choice.

The remaining human commitment is especially important here. A model cannot bear the meaning or consequences of continuing a pregnancy, ending it or preparing care. Irreversible decisions belong within a stronger structure of diagnosis, counselling and autonomous consent.

Conclusion: keep 99% attached to the question it actually answers

NIPT should continue as an effective, voluntary prenatal screen, with better equitable access. Detecting about 99% of trisomy 21 pregnancies is a major achievement and identifies people who may benefit from confirmation more effectively than older approaches.

Reports, advertising and clinical communication must not convert detection rate into an individual probability of being affected. High risk may initiate counselling and diagnostic choice; it cannot support an irreversible decision alone. No-result outcomes also require a clear pathway.

The closer a probability moves to 100%, the more easily it is heard as fact. That creates a greater duty to state the denominator, the object and the limitation. NIPT helps people face one possible future earlier. It cannot conclude that future before diagnosis exists.

Primary sources

Discover more from Geoffrey Chen

Subscribe to get the latest posts sent to your email.