NIPT for Sickle Cell Screening Would Need Hundreds of Thousands of Pregnancies, Model Finds

A decision tree model built around England’s sickle cell disease screening pathway has calculated what a rigorous diagnostic accuracy study of non-invasive prenatal testing (NIPT) would actually demand — and the numbers are sobering. Under one stakeholder-agreed scenario, a prospective cohort study meeting the minimally acceptable criteria would need to enrol 315,824 pregnancies; under a stricter scenario, 65,509. Both totals far exceed the roughly 6,200 pregnancies at risk of sickle cell disease identified in England each year, which is why the study’s authors conclude that a traditional cohort design is unachievable and propose alternative routes instead.

The analysis, published in Diagnostic and Prognostic Research, set out to answer a question that diagnostic studies often leave unaddressed: before testing a new screening tool, how accurate must it be, and how many participants would proving that accuracy require? The team, led by researchers working with experts from the UK National Screening Committee and the NHS Sickle Cell and Thalassaemia Screening Programme, applied a step-wise methodology published by Korevaar and colleagues in 2019, which quantifies the downstream consequences of test results and weights the harms of misclassification before any sample size is calculated.

The clinical context matters. Sickle cell disease is a genetic condition marked by chronic anaemia, acute painful episodes, organ infarction, chronic organ damage and a significantly reduced life expectancy, affecting an estimated 15,000 people in the United Kingdom, with higher prevalence among those of African and Caribbean backgrounds. England’s antenatal screening programme begins with a maternal carrier blood test. If that is positive, the biological father is offered the same test; only when both parents carry a sickle cell allele is invasive prenatal diagnosis — chorionic villus sampling or amniocentesis — offered to determine the fetus’s status.

That pathway has a well-documented weak point. When the father is unwilling or unavailable for testing, the pregnant woman is offered invasive prenatal diagnosis directly without completed risk information. This group accounted for 24% of all screen-positive women in the 2019 to 2020 screening year. Meanwhile, uptake of the paternal carrier blood test was just 70.4%, compared with 99.87% for the maternal test. Many women decline invasive prenatal diagnosis, particularly once pregnancy has progressed into the second trimester, and the outcome of those pregnancies remains uncertain until newborn dried bloodspot screening after birth.

NIPT, which analyses cell-free DNA circulating in the maternal bloodstream, presents a possible alternative to the paternal testing step. Small studies have suggested it may be effective in identifying pregnancies at risk of sickle cell disease, but it has not yet been proven accurate enough to serve as non-invasive prenatal diagnosis and should not currently be considered a replacement for invasive prenatal diagnosis. Its usefulness at this stage, the authors argue, may lie in replacing the paternal carrier blood test as a screening step — offered to all screen-positive women regardless of whether the father is available — with those receiving high-chance results still referred for invasive confirmation.

To define the minimally acceptable criteria, the researchers built a deterministic decision tree in Microsoft Excel, modelling the pathway from the maternal carrier blood test onwards. Model inputs were drawn from the latest NHS Sickle Cell and Thalassaemia Screening Programme data report, covering March 2019 to April 2020, supplemented by published literature and expert opinion validated in two stakeholder workshops. Two stakeholder-defined requirements anchored the analysis: the positive predictive value of NIPT should equal that of a completed paternal testing protocol — set at 25%, matching the proportion of pregnancies where both parents are carriers — and limits were placed on false negative diagnoses, since women receiving false negative NIPT results would be offered no further antenatal testing.

Two scenarios emerged. Scenario 1, tolerating a maximum of nine false negative diagnoses per year, required a sensitivity of 96.0% or greater and a specificity of 88.5% or greater. At those levels the model produced 4794 true negatives and 9 false negatives, giving a negative predictive value of 99.81%, alongside 213 true positives and 623 false positives, giving a positive predictive value of 25.51% and overall accuracy of 88.80%. Scenario 2, the more demanding threshold of at most two false negatives per year, required sensitivity of 99.0% and specificity of 88.0%, yielding a negative predictive value of 99.95%, a positive predictive value of 25.29% and accuracy of 88.43%. The high false positive counts, the authors note, would in practice be followed by offers of invasive prenatal diagnosis, mirroring the proportion of unaffected fetuses flagged by the current programme.

The sample size calculations then exposed the scale problem. For Scenario 1, with minimally acceptable sensitivity of 96% and an expected sensitivity of 96.5%, the study would need at least 12,448 fetuses with sickle cell disease; applying the assumed prevalence of 3.94% among screen-positive pregnancies translates to 315,824 total pregnancies. The specificity arm was similarly burdensome: an expected specificity of 89.0% against a minimum of 88.5% would require 34,279 fetuses without the disease, or 35,686 pregnancies. For Scenario 2, an expected sensitivity of 99.5% against a minimum of 99.0% demanded 2582 SCD-positive fetuses and 65,509 pregnancies, while the specificity arm required 35,597 SCD-negative fetuses and 37,058 pregnancies. Calculations assumed a 5% type 1 error rate and 10% type 2 error rate, in line with the Korevaar methodology.

The authors outline several ways to shrink these figures. If preliminary studies suggest higher accuracy is achievable, raising the expected values helps dramatically: an expected sensitivity of 99.0% in Scenario 1 would cut the requirement to 225 SCD-positive fetuses, or 5709 pregnant women, and an expected specificity of 90.0% would need 3826 women. Design mitigations suited to low-prevalance settings offer another route — using newborn screening as a proxy reference standard, a nested case-control design to reduce the number of index tests, or a two-phase design to limit invasive procedures. Even combined, the authors caution, the required case numbers remain challenging for a single study, and a pragmatic, stepwise approach — formal accuracy studies followed by clinical utility studies that also report accuracy, or eventual meta-analysis of comparable NIPT studies — may be needed.

The study carries clear limitations. Its assumptions, validated by stakeholders but untested in real-world conditions, include no time dependency, meaning NIPT uptake is likely modelled higher than would actually be observed, since women are less likely to accept screening as pregnancy progresses — though the non-invasive nature of NIPT may offset this. The sensitivity and specificity thresholds identified are lower than those achieved in other NIPT applications such as trisomy screening, because the criteria were deliberately set to at least match the performance of the current paternal testing stage rather than to outperform it. The authors also stress that implementation would require evaluation beyond accuracy, including cost and the number of invasive diagnoses compared with standard care, which they report in a parallel cost-effectiveness publication. Studies in UK populations are ongoing, they conclude, and decisions on study design should continue in light of those results.

Subject of Research: Decision tree modelling of minimally acceptable accuracy criteria and sample size requirements for NIPT in sickle cell disease antenatal screening in England

Article Title: Minimally acceptable criteria and required sample size for an accuracy study of non-invasive prenatal testing of sickle cell disease in screen positive women in England: results of a decision tree model

Article References: Vardanega, V., Bobrowska, A., Ruban-Fell, B., Doorbar, J. A., Lombardo, S., Seedat, F., Daniel, Y., Hogan, A., Leeflang, M. M. G., Mann, K., Schuh, A., & Marshall, J. (2026). Minimally acceptable criteria and required sample size for an accuracy study of non-invasive prenatal testing of sickle cell disease in screen positive women in England: results of a decision tree model. Diagnostic and Prognostic Research, 10(1), Article 18. https://doi.org/10.1186/s41512-025-00192-w

Image Credits: AI Generated

DOI: 10.1186/s41512-025-00192-w

Keywords: NIPT, sickle cell disease, prenatal screening, diagnostic accuracy, sample size, decision tree model, England, cell-free DNA, screening pathway, false negatives, predictive value, antenatal care