Noninvasive prenatal screening (NIPS) for fetal chromosome defects has high sensitivity and specificity but is not fully diagnostic. In response to a desire to provide more information to individual women with positive NIPS results, 2 online calculators have been developed to calculate posttest risk (PTR). Use of these calculators is critically reviewed. There is a mathematically dictated requirement for a precise estimate for the specificity to provide an accurate PTR. This is illustrated by showing that a 0.1% decrease in the value for specificities for trisomies 21, 18, and 13 can reduce the PTR from 79-64% for trisomy 21, 39-27% for trisomy 18, and 21-13% for trisomy 13, respectively. Use of the calculators assumes that sensitivity and specificity are constant for all women receiving the test but there is evidence that discordancy between screening results and true fetal karyotype is more common for older women. Use of an appropriate value for the prior risk is also important and for rare disorders there is considerable uncertainty regarding prevalence. For example, commonly used rates for trisomy 13, monosomy-X, triploidy, and 22q11.2 deletion syndrome can vary by >4-fold and this can translate into large differences in PTR. When screening for rare disorders, it may not be possible to provide a reliable PTR if there is uncertainty over the false-positive rate and/or prevalence. These limitations, per se, do not negate the value of screening for rare conditions. However, counselors need to carefully weigh the validity of PTR before presenting them to patients. Additional epidemiologic and NIPS outcome data are needed.
Introduction
When prenatal screening is provided using conventional screening tests (maternal serum and ultrasound markers), it has been common practice to report a patient-specific risk for each woman tested. Numerical estimates of risk have facilitated the combination of multiple risk factors such as maternal age and family history, as well the individual test components. Patients have been counseled according to risk and have made their decisions about whether or not to accept invasive testing on the basis of these risks.
The introduction of noninvasive prenatal screening (NIPS) using cell-free DNA (cfDNA) in maternal plasma changed the paradigm. These results are generally reported as test positive (increased, or high risk) or test negative (low risk). Some approaches present a risk score on the report but this only provides a measure of the likelihood that DNA from aneuploid cells is present in the maternal circulation; this is not a measure of the probability of true fetal chromosome imbalance. NIPS for various fetal chromosome abnormalities is associated with sensitivities and specificities that are much higher than conventional screening but the testing is not diagnostic. Moreover, it has been recognized that when the test is provided to women with low prior risks, the positive predictive value (PPV) of the testing is expected to be lower. Uncertainty about the true risk for fetal imbalance and the desire to counsel women more precisely has prompted the development of online calculators to assess individual women’s post-NIPS risk.
In this article I review these calculators. I explain why considerable caution is needed when using these calculators to compute a risk for individual patients.
General considerations
The online calculators use the formula:
(supplemental material, formula 5 ) where:
q i is the posttest risk (PTR) (usually expressed as percentage) carried out on the i th patient; p i is the pretest risk; D is the detection rate (sensitivity); and C is the specificity.
The PTR ( q i ) for each patient has been referred to as a PPV but this can be a source of confusion because there will be a PPV that reflects the overall test performance for the entire population of women screened. For the purposes of evaluating risk for individual women it would therefore be more accurate to refer to the number generated by the calculators as a personalized PPV (which would be equivalent to testing a population with the same prior risk, using a test with the same sensitivity and specificity) or, more simply, PTR. The distinction between PTR and PPV is discussed in more detail in the online supplementary material accompanying this article.
General considerations
The online calculators use the formula:
(supplemental material, formula 5 ) where:
q i is the posttest risk (PTR) (usually expressed as percentage) carried out on the i th patient; p i is the pretest risk; D is the detection rate (sensitivity); and C is the specificity.
The PTR ( q i ) for each patient has been referred to as a PPV but this can be a source of confusion because there will be a PPV that reflects the overall test performance for the entire population of women screened. For the purposes of evaluating risk for individual women it would therefore be more accurate to refer to the number generated by the calculators as a personalized PPV (which would be equivalent to testing a population with the same prior risk, using a test with the same sensitivity and specificity) or, more simply, PTR. The distinction between PTR and PPV is discussed in more detail in the online supplementary material accompanying this article.
The National Society of Genetic Counselors/Perinatal Quality Foundation calculator
This calculator was developed by members of the National Society of Genetic Counselors (NSGC) and the Perinatal Quality Foundation (PQF) and is designed for health care professionals with an understanding of predictive values; it is “intended to aid such health care professionals in counseling their patients.” PTR can be calculated for a variety of specific chromosome abnormalities. Users can enter sensitivity and specificity for NIPS or use default rates based on the metaanalyses of Gil et al . For some chromosome abnormalities default rates are unavailable. Prior risk (or prevalence) can be entered or, for some disorders, the prior risk can be based on maternal age at delivery. Prior risk is based on a fixed time point in pregnancy (gestational age 16 weeks). As well as PTR, the posttest level of reassurance provided by a negative result for an individual patient (referred to in the calculator as the negative predictive value) can be calculated.
The University of North Carolina calculator
The calculator developed by Grace et al from the University of North Carolina is designed “as a teaching tool to demonstrate the relationship between a priori risk, sensitivity, and specificity and to underline that cfDNA screening is not a diagnostic test.” The calculator is limited to trisomy 21, 18, and 13. Recognizing that the different NIPS tests available in the United States can be expected to have different performance characteristics, users can select the test (Verifi Illumina Inc, Redwood City, CA; Harmony Ariosa Diagnostics Inc, San Jose, CA; Materniti21 Sequenom Inc, San Diego, CA; or Panorama Natera Inc, San Carlos, CA). The option to use custom sensitivity and specificity rates is not available. Prior risk can be a user-specified value (gestational age for this is not needed), or it can be based on maternal age (range 20-44 years). When maternal age is used, the gestational age (between 10-20 weeks) also needs to be specified.
Limitations and implicit assumptions
Positive, negative, and intermediate results
Both calculators assume that the initial determination that a case is positive or negative will be independent of the prevalence. In fact, 2 of the commonly used NIPS methods compute a risk score that already incorporates maternal age in the algorithm and cases are only considered positive if this risk score is >1%. Use of the calculators is problematic in this situation because a case might have been classified differently if age had not been used. In effect, age is being used twice to compute the PTR in some cases.
Although age is a relatively weak contributor to risk (most cases have risk scores >99/100 and would likely test positive even maternal age were not included), there are some positive cases that fall close to the 1% cutoff. One of the test methodologies also presents some findings as having intermediate risk or “aneuploidy suspected” and for these cases there are insufficient data to assess a PTR.
Independence of sensitivity, specificity, and prevalence
A second assumption implicit in the calculation of patient-specific posterior risk is that sensitivity and specificity are independent of prior risk (prevalence); ie, the composite values of sensitivity and specificity that were derived from trials can be used for any woman. This assumption is widely accepted as valid for conventional screening and is the basis for computing patient-specific risks using single values for specificity and sensitivity. However, there are grounds to question the validity of the assumption for NIPS.
Reasons for a discordance between the NIPS result and the true fetal karyotype are becoming increasingly well understood and may be due to true fetal mosaicism, confined placental mosaicism, vanished twin, maternal mosaicism (including somatic cell gain or loss of X-chromosome and maternal cancer), maternal small copy number variants, and limitations of testing such as those associated with low amounts of conceptus-derived DNA and laboratory errors. Mosaicism due to a meiotic error (with subsequent correction to disomy), somatic cell gain or loss of an X-chromosome, maternal cancer, and vanished twin are each expected to be more common in older women and therefore more false-positive findings are expected. Indirect effects are also possible. For example, maternal weight gain (which positively correlates with maternal age), a small placenta (correlating with low maternal serum pregnancy-associated plasma protein-A and human chorionic gonadotropin), and earlier testing are 3 circumstances where the fetal fraction is expected to be lower and this might lead to more discordant results. Incorrect NIPS results due to maternal constitutional chromosome abnormalities and other test technical limitations presumably would be independent of prior risk. However, as discussed below, only a small change in the false-positive rate can materially alter posterior risks.
There are some observational data that support the view that discordancy rates may not be the same for all women. Bianchi et al observed that the mean maternal age was significantly higher in false-positive results for monosomy-X and also significantly lower for XXX false-positive findings. Dar et al reported the unexpected finding that the PPV for trisomies 21, 18, and 13, and monosomy-X was higher in a low-risk population compared to a high-risk group. On the other hand, 2 studies did show higher PPVs for trisomy 21 screening in older women compared to younger women. These studies did not assess in detail whether the PPV was proportionate to prior risk in multiple subgroups of women or consider aneuploidies other than trisomy 21.
Accuracy of sensitivity and specificity
The PTRs that are calculated are strongly dependent on accurate assessments of the specificity. This is illustrated for a theoretical test in Figure S1. Table 1 also illustrates this for some practical situations using the NSGC/PQF calculator. For the examples shown, an increase in the false-positive rate of just 1 in 1000 cases (0.1%) above that used in the default setting can change PTR from 79-64% for trisomy 21, 39-27% for trisomy 18, and 21-13% for trisomy 13, respectively. Table 1 also illustrates that PTR is only weakly dependent on the sensitivity; a change of 10% in the detection rate only alters the posterior risks by 2-4% for these 3 disorders.
| Disorder | Prior risk a | Sensitivity | Specificity | Posterior risk | Comment b |
|---|---|---|---|---|---|
| t21 | 1/296 | 99.2% | 99.91% | 78.8% | Default rates |
| 99.2% | 99.81% | 63.7% | 0.1% Higher FPR | ||
| 99.2% | 99.71% | 53.5% | 0.2% Higher FPR | ||
| 89.2% | 99.91% | 76.9% | 10% Lower DR | ||
| 79.2% | 99.91% | 74.8% | 20% Lower DR | ||
| t18 | 1/1152 | 96.3% | 99.87% | 39.1% | Default rates |
| 96.3% | 99.77% | 26.6% | 0.1% Higher FPR | ||
| 96.3% | 99.67% | 20.2% | 0.2% Higher FPR | ||
| 86.3% | 99.87% | 36.5% | 10% Lower DR | ||
| 76.3% | 99.87% | 33.7% | 20% Lower DR | ||
| t13 | 1/2576 | 91.0% | 99.87% | 21.4% | Default rates |
| 91.0% | 99.77% | 13.3% | 0.1% Higher FPR | ||
| 91.0% | 99.67% | 9.7% | 0.2% Higher FPR | ||
| 81.0% | 99.87% | 19.5% | 10% Lower DR | ||
| 71.0% | 99.87% | 17.5% | 20% Lower DR |
a Based on woman age 35 y (at delivery) and 16 wk’ gestation
b Default rates are sensitivity and specificity used in National Society of Genetic Counselors/Perinatal Quality Foundation calculator.
It should be remembered that the trials that established NIPS metaanalysis performance were based on selected cases. There was heterogeneity in trial entry criteria and often cases with mosaicism, complex karyotypes, or low fetal fraction were excluded. Moreover, performance is expected to be different for the various NIPS laboratories, and test protocols have evolved since the trials were conducted. For any one laboratory, the number of cases included in their trials is such that the 95% confidence intervals around the specificity estimates may be relatively wide. There is therefore some degree of uncertainty as to the true false-positive rates in actual clinical practice.
Accuracy of prior risk
The PTR is dependent on the prior risk. In many instances the prior risk will be based on maternal age. For trisomy 21 and 18, the differences in the various available maternal age-specific prevalence curves and the correction factors to generate gestational age-specific prior risks are rather small and the corresponding differences in the PTRs are minor. This is illustrated in Table 2 for 2 illustrative women aged 21 and 45 years. However, for other chromosome abnormalities there can be relatively large differences in prior risk, depending on the choice of reference source, and there is therefore a corresponding uncertainty about the PTR. For monosomy-X, the NSGC/PQF calculator is based on prevalence that includes mosaic cases (not necessarily detectable by NIPS) and abnormal ultrasound findings. Other much lower rates are available that are based on apparently nonmosaic 45,X. For triploidy and 22q11.2 deletion syndrome there is as much as 4-fold differences in published prevalence rates.
| Disorder | Age 21 y a | Age 45 y a | Prior risk source | ||
|---|---|---|---|---|---|
| Prior risk | Posterior risk, % | Prior risk | Posterior risk, % | ||
| t21 | 1/1160 | 48.7 | 1/22 | 98.1 | Default b |
| 1/1041 | 51.5 | >1/22 | >98.1 | UNC calculator c | |
| 1/1226 | 47.4 | 1/25 | 97.9 | Morris et al, 2002; Savva et al, 2006 d | |
| t18 | 1/4521 | 14.1 | 1/84 | 89.9 | Default |
| 1/3545 | 17.3 | >1/73 | >91.1 | UNC calculator | |
| 1/3174 | 18.9 | 1/77 | 90.7 | Savva et al, 2010; Morris and Savva, 2008 e | |
| t13 | 1/1795 | 27.0 | n/a | n/a | Default |
| 1/10,901 | 6.0 | >1/221 | >76.1 | UNC calculator | |
| 1/7998 | 8.0 | 1/404 | 63.5 | Savva et al, 2010; Morris and Savva, 2008 | |
| MX | 1/568 | 40.9 | 1/568 | 40.9 | Default |
| 1/995 | 28.3 | 1/2094 | 15.8 | Ferguson-Smith and Yates, 1984; Forabosco et al, 2009 f | |
| XXY | 1/1613 | 29.2 | 1/162 | 80.5 | Default |
| 1/2815 | 19.1 | 1/196 | 77.3 | Carothers et al, 1978; Hook, 1981 g | |
| XXX | 1/1795 | 27.0 | 1/253 | 72.5 | Default |
| n/a | n/a | 1/294 | 69.4 | Hook et al, 1983 | |
| XYY | 1/2000 | 24.9 | 1/2000 | 24.9 | Default |
| 1/3101 | 17.6 | 1/1362 | 32.8 | Forabosco et al, 2009 | |
| Triploidy | 1/30,000 | n/a | 1/30,000 | n/a | Default |
| >1/7937 | n/a | >1/7937 | n/a | Wellesley et al, 2012 | |
| 22q11.2del | 1/4000 | n/a | 1/4000 | n/a | Default |
| 1/992 | n/a | 1/992 | n/a | Grati et al, 2015 h | |
b Value in National Society of Genetic Counselors/Perinatal Quality Foundation calculator
d Term risk (Morris et al, 2002) adjusted by using maternal age-specific regressed loss rates for affected pregnancies (Savva et al, 2006) and 1.5% loss rate for unaffected pregnancies
e Term risk (Savva et al, 2010), adjusted by using loss rates for affected pregnancies (Morris and Savva, 2008) and 1.5% loss rate for unaffected pregnancies
f For age 21 y rate based on all women age <35 y (Forabosco et al, 2009), and for age 45 y rate based on all women age ≥35 y (Forabosco et al, 2009 plus Ferguson-Smith and Yates, 1984)–apparently nonmosaic cases
g Based on term rate from regression curve (Carothers et al, 1978) with rate of 5/10,000 at age 30 y (Hook, 1981) and no excess loss for affected pregnancies
Prior risk for trisomy 21, 18, and 13 based on conventional screening incorporate maternal age and is therefore subject to the same considerations. In addition, there will be additional sources of uncertainty due to the compounding of conventional test measurement errors.
Comment
NIPS is a screening test. As such, it is appropriate to consider the PPV for a total population or a subpopulation. However, extending the concept to calculate the PTR for any individual patient positive for a particular chromosome abnormality can be problematic because of uncertainty in the false-positive rate. Additionally, there is a need for an accurate value for the prior risk and that may not be available for rare disorders. These considerations, per se, do not negate the value of the screening. In fact, for the reasons discussed above, NIPS sensitivity and specificity for the autosomal trisomies may be higher for women with low prior risks. The difficulty in accurately assessing specificity and prior risk constitute basic practical and statistical limitations that can be expected to arise whenever specificity is very high and/or prevalence is low. However, these requirements do place limitations on the information that that can be provided following a positive test.
The offer of NIPS when there is a low prior risk is already standard practice; even when women are high risk for one particular autosomal trisomy they are often at very low risk for other chromosome abnormalities but it is accepted practice to examine the cfDNA for the full set of conditions. Counselors therefore need to carefully weigh the validity of a patient-specific PTR, taking into consideration a broad set of factors ( Box 1 ). For an advanced maternal age woman who received a positive trisomy 21 or 18 NIPS result from a laboratory with well-established test performance, the calculated PTR is likely to provide a reasonable estimate of her risk. For younger women, currently, these risks may be an underestimate. In some other situations, calculations based on upper and lower boundary values for prior risk and specificity could be considered. If a meaningful patient-specific PTR cannot be provided, counseling could be based on an overall estimate of the PPV derived from laboratory experience recognizing that there are patient-specific considerations. It should also be recognized that there are numerous screening tests where patient-specific PTR are not routinely provided; for example, prenatal screening for open neural tube defects using the maternal serum alpha-fetoprotein test.
There is a need for more information on the prevalence of the disorders and the performance of the screening tests for various groups of women in actual clinical practice. NIPS reference laboratories should collect pregnancy outcome information and evaluate expected vs observed true-positive rates for specific subpopulations. However, the success of this approach is highly dependent on referring physicians comprehensively providing pregnancy outcome data. In the past, state and national registries have recorded rates of abnormality in prenatal diagnosis specimens and newborns and thereby provided invaluable epidemiologic data. The introduction of NIPS adds a new impetus for this important public health activity. Data collection for rare disorders is a long-term undertaking and it is likely that there will be ever-increasing requirements for greater granularity in the available information. In the short-term, it should be recognized that analysis of cfDNA has significantly improved the performance and broadened the scope of screening for fetal chromosome abnormalities and this can already be highly beneficial using currently available information for counseling.
Supplemental Appendix
Definitions and mathematical background
- (a)
Calculation of posttest risk (PTR)
Let D i be detection rate (sensitivity); F i be false-positive rate (1-specficity, C i ); and p i be prior risk of disorder (expressed as proportion) of i th case undergoing screening.
p i can also be as odds ratio, 1:n i
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
