## Abstract

**BACKGROUND:** An estimate of fetal fraction (FF) is needed for DNA-based screening for trisomy 21 and other aneuploidies, but there is no gold standard to validate FF measurement methods. We specify a gold standard and use it to validate a method of measuring FF (SeqFF) in singleton pregnancies.

**METHODS:** The gold standard was a formula derived from 2 elements: (*a*) an estimate of the percentage of DNA fragments in maternal plasma from chromosome 21 (%Ch21) in pregnancies without trisomy 21, 18, or 13 (P_{U}) and (*b*) calculation of %Ch21 with increasing FF in trisomy 21 pregnancies (P_{21}). The SeqFF method was evaluated by plotting regression lines of %Ch21 and SeqFF estimates of FF in 31 singleton male and 31 female trisomy 21 pregnancies and comparing the regressions with the reference line derived from the gold standard formula.

**RESULTS:** The gold standard formula was P_{21} = (1/2)P_{U}FF + P_{U}, with FF expressed as a proportion, or converting %Ch21 to multiples of the median (MoM), P_{21}(MoM) = (1/2)FF + 1. Based on 3865 pregnancies, the P_{U} was 1.2935%. The regression lines for trisomy 21 pregnancies with male and female fetuses were almost identical to the gold standard reference line (regression slopes in MoMs 0.52 and 0.50, respectively, compared with 0.50 for the gold standard reference line).

**CONCLUSIONS:** The proposed gold standard can be used to validate different methods of estimating FF in singleton pregnancies. SeqFF is an accurate method of estimating FF.

Maternal plasma DNA analysis is an effective method of prenatal screening for Down syndrome (trisomy 21), Edwards syndrome (trisomy 18), and Patau syndrome (trisomy 13) (1). In massively parallel sequencing, the number of DNA fragments from chromosome 21, 18, or 13 is counted and expressed as a percentage of all the autosomal DNA fragments (2, 3). An increase of any one over the expected indicates an increased risk of that trisomy. The DNA test depends on how much of the maternal plasma DNA is derived from the fetus (more accurately, the placenta, but referred to as fetal), which is expressed quantitatively as the fetal fraction (FF); the higher the FF, the higher the screening performance of the DNA test (4). FF is the proportion of DNA fragments derived from the fetus expressed as a percentage. Discrimination between pregnancies with and without 1 of the 3 trisomies is greater with a higher FF and lower with a low FF. Therefore, the FF is an important factor in the interpretation of the DNA test.

FF can be estimated in several ways. Differences in the methylation of specified DNA sequences in fetal and maternal plasma DNA is a method (5), as is the quantification of paternally transmitted single-nucleotide polymorphisms (6), but these require a separate analysis from the screening analysis. Approaches that do not require such a separate testing procedure (i.e., intrinsic tests) have an advantage. One such approach is based on measuring the number of sequenced DNA fragments derived from the Y chromosome or measuring the underrepresentation (depletion) of X chromosome (3), which is useful if the fetus is male but useless if the fetus is female. Another intrinsic method makes use of the fact that in maternal plasma the DNA fragments of fetal origin are approximately 10 bp shorter than fragments of maternal origin with considerable overlap in the distributions (7). However, the method is an imprecise way of estimating FF (8). An intrinsic method that could overcome these limitations is one called SeqFF (9) that relies on the sequencing of DNA fragments that exhibit small quantitative differences in plasma DNA arising from the placenta (fetal) and from the mother.

Previous studies to validate methods of estimating FF have compared one method against another (8). Here, we specify a gold standard based on the relationship between FF and the percentage of DNA fragments from chromosome 21 (%Ch21) in trisomy 21 male and female pregnancies, and then use this gold standard to assess the accuracy of the SeqFF method to estimate FF in maternal plasma samples from 62 pregnancies with trisomy 21.

## Methods

FF was calculated using SeqFF (9) in 62 singleton pregnancies with trisomy 21 of known sex (31 males, 31 females) and 3785 unaffected singleton pregnancies. The pregnancies had been screened for trisomies 21, 18, and 13 at the Wolfson Institute of Preventive Medicine, London, UK, between December 2015 and September 2017. For this study, we defined an affected pregnancy as a pregnancy affected with trisomy 21, and an unaffected pregnancy as a pregnancy unaffected with trisomy 21, 18, or 13, or any known trisomy.

The expected %Ch21 in the plasma of a woman with an affected pregnancy (P_{21}) was the expected %Ch21 in an unaffected pregnancy (P_{U}) times the maternal fraction (1 − FF expressed as a proportion), plus the calculated %Ch21 in an affected pregnancy (3/2 × P_{U} because of the extra copy of chromosome 21 in an affected pregnancy) multiplied by FF expressed as a proportion. This gives rise to the following formula for the reference line defining %Ch21 and FF in affected pregnancies:

The %Ch21 was also expressed as a multiple of the median (MoM) for unaffected pregnancies, i.e., dividing both sides of the equation by P_{U}.

The formula specified a gold standard against which methods of determining FF could be validated. P_{U} was determined from the median %Ch21 in unaffected pregnancies. The %Ch21 was determined by “shotgun” DNA sequencing using an Ion Proton Thermo Fisher platform supplied by Premaitha, with associated reagents, chips, and assay software (10). For all affected pregnancies, a linear regression was performed of %Ch21 (P_{21}) on FF, as well as separately for affected pregnancies of each fetal sex, given that estimating FF could be more difficult if the fetus was female. The regression lines were constrained (forced) to pass through P_{U} at FF = 0 because if the FF was 0, the %Ch21 in affected pregnancies must be equal to P_{U}. The regression lines were compared with the gold standard reference line. In addition to excluding trisomy 21 pregnancies from the unaffected category, trisomy 18 and 13 pregnancies were also excluded to avoid the minimal but systematic underestimation of P_{U} that would occur given the extra DNA fragments from the extra copy of chromosome 18 or 13 in such pregnancies. The regression analyses were performed using Stata version 14. Institutional Review Board approval was not required because the project arose from and formed part of an audit of the Wolfson Institute of Preventive Medicine reflex DNA antenatal screening service (11).

## Results

The median %Ch21 in unaffected singleton pregnancies was 1.2935% (95% CI, 1.2933%–1.2938%). Therefore, the gold standard reference line was specified by the following equation, which took the form of y = mx + c: or, expressed in MoM values:

Fig. 1 shows the gold standard reference line and a line for unaffected singleton pregnancies at P_{U} (Fig. 1A) and the regression line for all 62 affected pregnancies (Fig. 1B). The regression line was close to the expected gold standard reference line (slope of regression line, 0.65707% compared with 0.64675% for the gold standard reference line; *P* value for the difference = 0.985), supporting the SeqFF method of determining FF as being accurate.

Fig. 2 shows, in a similar way to Fig. 1, the regression lines for affected singleton pregnancies with male (Fig. 2A) and female (Fig. 2B) fetuses, again with the reference lines. The regression lines were nearly identical to the gold standard reference line (slope of regression line, 0.66891% for pregnancies with male fetuses and 0.64748% for pregnancies with female fetuses; *P* = 0.662 and 0.756, respectively, compared with the gold standard reference line). Figs. 1 and 2 also express %Ch21 in MoMs; in our study, 1.2935% = 1 MoM. In MoMs, the slopes of the regression lines were 0.52 for pregnancies with male fetuses, 0.50 for pregnancies with female fetuses, and 0.51 for all pregnancies, compared with 0.50 for the gold standard reference line.

The median FF in the affected pregnancies was 8.5% (95% CI, 7.0%–10.2%) or 9.7% (95% CI, 8.0%–11.2%) if the 5 pregnancies with FF < 4% were excluded, as is often the case in prenatal screening programs (4). The median FF values in unaffected pregnancies were 9.6% (95% CI, 9.4%–9.7%) and 9.8% (95% CI, 9.7%–10.0%) respectively.

## Discussion

Our study shows that it is possible to specify a gold standard for determining FF in singleton pregnancies based only on fixing the intercept in Figs. 1 and 2 at the median %Ch21 value in unaffected pregnancies (P_{U}). The slope of the line is determined mathematically based on the mix of fetal and maternal DNA fragments in maternal plasma. We believe this is the first study to validate a method of estimating FF in pregnancies with Down syndrome. Previously validation was performed by comparing 2 methods of estimating FF (8). The gold standard reference line is based on only an estimate of P_{U}, which was here derived from 3865 pregnancies (1.2935%, which is, to 1 decimal place, the same as the value of 1.3% in nonpregnant individuals) (12). Given this “0 FF” %Ch21 value, the slope is determined mathematically by calculating the %Ch21 increment from an affected pregnancy that is added to 1.2935%. For example, if the FF were 10% (0.1 as a proportion) and the 0 FF %Ch21 value were 1.3%, then the expected %Ch21 in an affected pregnancy (P_{21}) would be (1/2) × 1.2935% × 0.1 + 1.2935%, or 1.365%. The close similarity between the gold standard reference line and the regression lines in Figs. 1 and 2 validates the SeqFF method in the estimation of FF.

Different methods of DNA analysis are likely to produce similar but systematically different estimates of the %Ch21, with some methods yielding higher values and some lower ones. However, this is not relevant to the efficacy of DNA analysis in prenatal screening, provided the overestimation or underestimation is proportionately the same in counting %Ch21 fragments and counting all DNA fragments.

The effect of systematic variations in estimating %Ch21 from one DNA method to another can be allowed for by expressing %Ch21 in MoMs, which is standard practice for conventional biochemical screening markers, such as human chorionic gonadotropin. Expressed in MoMs, the gold standard reference line is independent of the DNA testing method. If the method of determining FF systematically produced overestimates, the slope of the regression lines shown in Figs. 1 and 2 would be shallower and steeper if FF were underestimated. Any underestimation or overestimation of about ≥5% would be readily seen on the validation plots provided there were sufficient data points; ≥50 would usually be sufficient.

The use of MoMs means that data on pregnancies with trisomy 18 or 13 can be combined with data on trisomy 21 in validating assays of fetal fraction because the equation of the gold standard reference line P_{21,18, and 13} (MoM) = (1/2)FF + 1 applies to any of the 3 trisomic pregnancies, provided P_{U} is specific for the relevant trisomy. In our study, based on the 3865 unaffected pregnancies, P_{U} was 3.0980% (95% CI, 3.0973%–3.0989%) with respect to trisomy 18 and 3.7263% (95% CI, 3.7258%–3.7268%) with respect to trisomy 13. Fig. 3 shows the gold standard reference line and the regression line for the 62 pregnancies with trisomy 21, 34 pregnancies with trisomy 18, and 12 pregnancies with trisomy 13 identified during the screening period. An important advantage of combining data from the 3 trisomies in this way is that it increases the number of affected pregnancies available for fetal fraction assay validation plots.

The proposed gold standard for validating different methods of estimating FF in singleton pregnancies should be useful in prenatal screening based on cell-free DNA methods. The gold standard overcomes a limitation in previous attempts to validate different methods of FF estimation that were restricted to comparing one method of estimation against another. In addition, our results indicate that the SeqFF method is an accurate method of estimating FF.

## Footnotes

**Author Contributions:***All authors confirmed they have contributed to the intellectual content of this paper and have met the following 3 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; and (c) final approval of the published article.***Authors' Disclosures or Potential Conflicts of Interest:***Upon manuscript submission, all authors completed the author disclosure form. Disclosures and/or potential conflicts of interest:***Employment or Leadership:**None declared.**Consultant or Advisory Role:**None declared.**Stock Ownership:**None declared.**Honoraria:**None declared.**Research Funding:**None declared.**Expert Testimony:**None declared.**Patents:**N.J. Wald, GB1619812.9; R. Old, 1619812.9.**Role of Sponsor:**No sponsor was declared.

- Received for publication February 21, 2018.
- Accepted for publication June 7, 2018.

- © 2018 American Association for Clinical Chemistry