Noninvasive Prenatal Testing and Detection of Occult Maternal Malignancies

Since the introduction of noninvasive prenatal testing (NIPT)7 in prenatal diagnostics in 2011, incidental findings of an occult maternal malignancy following a “false-positive” NIPT test have been reported repeatedly. In these cases, NIPT, originally designed to interrogate placenta-derived “fetal” cell-free DNA (cfDNA) in the maternal circulation, also detected the presence of tumor-derived cfDNA. This freely circulating tumor DNA is characterized by multiple chromosomal imbalances—a hallmark of human neoplasms—that skew the signals and give rise to an aberrant NIPT result (Fig. 1). Worldwide controversy exists about disclosing such NIPT results as suggestive of a maternal cancer. At the same time, these incidental findings prompted retrospective investigations into the association between a deviating NIPT outcome and an occult maternal malignancy. However, because these studies were not designed to evaluate the ability of NIPT to detect cancer, deductions about the test's specificity and sensitivity for cancer detection could not be made. In a recent publication, Ji and colleagues collected over 1.9 million NIPT data and retrospectively evaluated the performance of a novel bioinformatics algorithm to detect occult maternal malignancies in pregnant women possessing multiple chromosomal aneuploidies in NIPT results (1). Their work confirmed that the presence of multiple reproducible chromosomal aneuploidies in plasma cfDNA of asymptomatic pregnant women was indeed an indicator of cancer risk. In their hands, the positive predictive value (PPV) of detecting multiple chromosomal aneuploidies for the identification of a maternal malignancy was 7.6%. To increase performance, Ji and colleagues further developed their cancer detection algorithm by introducing a new parameter, that is, the “fraction of significant (sub)chromosomal copy number alterations (CNAs), detected across the entire genome” (called the “FCNV score”). FCNV scores were compared among the purported “cancer group,” the “noncancer group,” and the group of control participants for whom no multiple chromosomal aneuploidies were detected. Implementation of this FCNV score resulted in a 4-fold improvement of the pipeline's PPV (30%) compared to the diagnostic performance of multiple chromosomal aneuploidies on their own. Combining the cancer detection model with plasma tumor marker testing further improved the analytical performance of the pipeline, reaching a PPV of 75%, while at the same time providing some information that might facilitate tumor typing. It seems likely that the approach presented by Ji and colleagues could be further optimized. First, the data on the presence or absence of a cancer diagnosis in the patient cohort were retrieved retrospectively and were based on interrogation of the study participants. Therefore, it is likely that occult malignancies might have been present in some of the negatives in what was assigned to be the “noncancer group” by the time of the last follow-up. As a result, statistics might be skewed and the power of the training set be reduced. Furthermore, their definitions of “multiple CNAs” and “significant CNAs” remain unclear. Every whole-genome sequencing approach and algorithm in use would define those differently. Moreover, by not only using the “fraction of significant CNA scores” but additionally including the types of these “significant CNAs” known to be associated with cancer, the performance of the pipeline may well increase. In our practice, 10 of 11 persons (out of a total of about 75000 pregnant women undergoing routine NIPT testing) who had detectable (sub)chromosomal aberrations in their NIPT profile and who were subsequently referred to the whole-body MRI unit to screen them for the presence of malignant-like lesions did present with a tumor. Furthermore, the aberrant CNA profiles were confirmed to be derived from the malignancy. This result amounts to a PPV of 90% (2), underscoring the potential of this technology for highly specific presymptomatic cancer detection during pregnancy. During pregnancy, fetal DNA fragments are released from the placenta into the maternal blood. When a tumor is present, cell-free circulating tumor DNA that represents the genetic makeup of the malignancy and hence might be characterized by chromosomal imbalances may also be shed in the bloodstream. On genomewide NIPT assessment for the presence of chromosomal imbalances in the fetus, these tumor-derived aneuploidies will skew the signals, giving rise to an aberrant NIPT result. When the observed imbalances are incompatible with fetal development, a maternal malignancy might be invoked. In most papers that have so far reported on the association between an aberrant NIPT and a maternal malignancy, including the study by Ji and colleagues, the ratios of incidentally detected cancers on routine NIPT testing have ranged from 2 to 10 cases in 100000 pregnancies, which is about 5–50 times lower than the reported incidence rates of cancer in pregnancy (which are about 1 in 1000–2000 cancer diagnoses during pregnancy). This difference suggests that a substantial number of cancer cases are missed. The present algorithms are restricted to detect copy variable tumors and will miss copy neutral and single chromosomal aneuploid tumors. Additionally, other tumor characteristics, such as tumor type, size, and proliferation status, might determine the degree of tumor cfDNA shedding into the circulation and hence have an effect on the sensitivity of tumor detection. Because these factors are not yet fully understood, more fundamental investigations and much larger systematic evaluations are needed to determine which tumor types and stages lend themselves to assessments using cfDNA analysis. Nevertheless, current results suggest a great opportunity for NIPT-based technologies for disease detection and monitoring in pregnancy. Indeed, cancer during pregnancy is a delicate condition, demanding minimal invasiveness of diagnostic procedures. However, current tumor markers that can be tested noninvasively (like CA 125, squamous cell carcinoma antigen, and CA 15.3) may be unreliable because of pregnancy-specific variations (3). Also, symptoms of cancer (such as fatigue, nausea, abdominal discomfort, and vaginal blood loss) are easily misinterpreted as physiologic gestational symptoms. As a consequence, pregnant women are more often diagnosed at advanced stages of cancer than their nonpregnant peers, with a consequent negative effect on prognosis. Accumulating evidence indicates that early chemotherapeutic treatment during pregnancy has no adverse effect on pediatric outcome (4). Instead, an early start of chemotherapy has been shown to allow for a reduction of the number of pregnancy terminations and number of premature births and may add to a prognosis of the mother similar to that of nonpregnant women. Hence, proper identification of cancer through routine NIPT screening would have a major effect on the management of these pregnant women. Reporting the presence of a potential cancer during pregnancy is inevitably a balance between the value of diagnosing a true cancer and the risk of a false positive causing unnecessary stress and invasive follow-up testing. With PPVs reaching 85% and higher, the value of reporting outweighs the associated risks. We suggest that a stepwise workup could be applied, implying the involvement of a multidisciplinary team of specialists. First, NIPT testing should be preceded by counseling in which pregnant women are informed about potential incidental findings. Depending on local guidelines and the cultural and social acceptances, families could be offered the option to opt out of receiving results beyond the trisomy report. In case of an aberrant NIPT outcome, a second NIPT should be performed on an independent blood sample; when positive, the chromosomal cfDNA profile should be compared with the (constitutional) chromosomal pattern in maternal tissue. In parallel, genetic investigations into the fetal karyotype (amniocentesis) could be offered by a clinical geneticist. When the results are suggestive for an underlying maternal disorder, the patient should be referred to a multidisciplinary team consisting of a medical oncologist, hematologist, gynecological oncologist, and a perinatologist-obstetrician with experience in invasive prenatal diagnostics. Proper posttest counseling by a clinical geneticist should also be in place. The test report should note that it is suggestive of a maternal malignancy and provide the pregnant woman and her physician an estimate of the detection rate and the possibility of a false-positive result. In a next stage, identification of the primary tumor site should be done. In the approach of Ji and colleagues, a plasma protein marker assay was used for identification of the primary tumor origin. However, as stated above, these plasma tumor markers on their own show moderate diagnostic sensitivity and are less reliable in pregnancy. As an alternative, a whole-body (diffusion-weighted) MRI that allows for radiation-free imaging of the entire body during 1 examination or sequential organ-specific examinations may be a more diagnostically sensitive approach. Because lymphomas are the most frequently detected tumor type before breast cancer following an aberrant NIPT outcome, whole-body MRI examination should be complemented with the analysis of hematological parameters (such as cell counts and cytology and clinical biochemistry) to search for the presence of a hematological malignancy that might not be detectable via MRI scanning. With future novel developments in which supramolecular information contained in the cfDNA is exploited, it may well become possible to identify the tumor tissue of origin and guide oncological investigations based on cfDNA analyses. Whenever the prenatal clinical investigations do not provide an explanation for the observed deviating NIPT result, postpartum analysis of placental tissue would allow examining the possibility of placental mosaicism. If negative, a follow-up NIPT in the postpartum period could be performed to evaluate the evolution of the cfDNA profile. It should be noted that aberrant NIPT profiles have also been related to the presence of benign tumors such as uterine leiomyomas (5). Finally, vitamin B12 or folate deficiency or the presence of autoimmune diseases such as systemic lupus erythematosus might also lead to an abnormal NIPT result, although the profile of chromosomal imbalances found in these conditions might be discriminated from typical cancer-like CNAs. We would strongly advocate the installation of institutional or national guidelines, incorporating the above-mentioned aspects and allowing for an optimal clinical management of aberrant NIPT outcomes. Such guidelines have been implemented in Belgium (Belgian Guidelines for Managing Incidental Findings Detected by NIPT). From the above, it can also be appreciated that pregnant women with a confirmed neoplastic disease should be excluded from NIPT testing since NIPT assessment of fetal chromosomal constitution is not reliable in this condition (Fig. 1). In conclusion, to allow diagnostically sensitive and specific presymptomatic cancer detection in pregnant women during routine NIPT testing, further technical and bioinformatics optimizations are warranted. Given the added value of this noninvasive detection method for (the prognosis of) pregnant cancer patients, women and society would benefit from implementing prospective reporting. noninvasive prenatal testing cell-free DNA positive predictive value copy number alteration.