Correspondence

ment practices. The observations from our pilot

study of imetelstat support a unique mechanism
of action that deserves further laboratory-based
investigation.
Ayalew Tefferi, M.D.
Mayo Clinic
Rochester, MN
Since publication of his article, the author reports no further
potential conflict of interest.

1. Cassinat B, Verger E, Kiladjian JJ. Interferon alfa therapy in

CALR-mutated essential thrombocythemia. N Engl J Med 2014;
371:188-9.
2. Kuriakose ET, Gjoni S, Wang YL, et al. JAK2V617F allele burden is reduced by busulfan therapy: a new observation using an
old drug. Haematologica 2013;98(11):e135-7.
3. Kiladjian JJ, Mass A, Cassinat B, et al. Clonal analysis of
erythroid progenitors suggests that pegylated interferon alpha2a treatment targets JAK2V617F clones without affecting TET2
mutant cells. Leukemia 2010;24:1519-23.
DOI: 10.1056/NEJMc1512663

Cell-free DNA Analysis for Noninvasive Examination of Trisomy

To the Editor: Norton et al. (April 23 issue)1
report near-perfect accuracy of detection for trisomy 21 (Downs syndrome) with the use of cellfree DNA (cfDNA) (sensitivity, 100% [38 of 38
cases of trisomy 21]; false positive rate, 0.06%
[9 false positives among 15,841 women]) in the
Noninvasive Examination of Trisomy (NEXT)
study. These seemingly promising results may be
misleading because they excluded 488 patients
(3% of their sample) with indeterminate cfDNA
results. The prevalence of aneuploidy was higher
among these patients than in the overall cohort
(2.7% vs. 0.4%); thus, their exclusion may introduce bias.2 Estimates of accuracy should consider
indeterminate results to be either positives or
negatives according to how they would be handled in clinical practice.2
Given their increased risk of aneuploidy, patients with indeterminate results would probably
undergo additional testing. Thus, it may be appropriate to classify their cfDNA results as positives. This classification would result in a false
positive rate of 3.0% and a positive predictive
value of 7.6%, much lower than the reported
positive predictive value of 80.9%. Alternatively,
if indeterminate results were classified as negatives, sensitivity would be reduced to 38 of 41 cases
(93%) (95% confidence interval [CI], 80 to 98).
Assuming that no patients with indeterminate
results on standard screening had trisomy 21,
the sensitivity of cfDNA testing and standard
screening (33 of 41 cases [81%]; 95% CI, 66 to
90) would not be significantly different (P=0.22
by McNemars test).
Rebecca SmithBindman, M.D.
University of California, San Francisco
San Francisco, CA
rebecca.smith-bindman@ucsf.edu

Diana Miglioretti, Ph.D.

University of California, Davis
Davis, CA
No potential conflict of interest relevant to this letter was reported.
1. Norton ME, Jacobsson B, Swamy GK, et al. Cell-free DNA
analysis for noninvasive examination of trisomy. N Engl J Med
2015;372:1589-97.
2. Bossuyt PM, Reitsma JB, Bruns DE, et al. The STARD statement for reporting studies of diagnostic accuracy: explanation
and elaboration. Ann Intern Med 2003;138:W1-12.
DOI: 10.1056/NEJMc1509344

To the Editor: Norton and colleagues found

that cfDNA testing for trisomy 21, as compared
with standard screening, had a better global performance during the first trimester of pregnancy. However, they did not provide information
about the 14 fetal chromosomal abnormalities in
the 15,841 screened pregnancies, other than for
trisomies 13, 18, and 21.
Were these 14 aneuploidies diagnosed prenatally because of abnormal features on follow-up
ultrasonography or because of stillbirths or miscarriages? Or were they detected by standard
screening or postnatally? The answers to these
questions may help to determine whether a routine policy of general screening for aneuploidy
with the use of ultrasonography and cfDNA testing rather than standard screening is the best
strategy.
Loc Sentilhes, M.D., Ph.D.
Bordeaux University Hospital
Bordeaux, France
loicsentilhes@hotmail.com

LaurentJ. Salomon, M.D., Ph.D.

University of Paris
Paris, France

n engl j med 373;26nejm.org December 24, 2015

The New England Journal of Medicine

Downloaded from nejm.org on January 18, 2016. For personal use only. No other uses without permission.
Copyright 2015 Massachusetts Medical Society. All rights reserved.

2581

The

n e w e ng l a n d j o u r na l

Christophe Vayssire, M.D., Ph.D.

University Toulouse III
Toulouse, France
No potential conflict of interest relevant to this letter was reported.
DOI: 10.1056/NEJMc1509344

of

m e dic i n e

and maximizing detection of all aneuploidies in

this group.
Currently, screening by means of cfDNA is
not designed to detect all aneuploidies, and trisomy 21 composed just over 50% of chromosomal abnormalities in the NEXT cohort. As we
noted, Women who desire a comprehensive assessment may prefer diagnostic testing with
karyotype or chromosomal microarray analysis.
Although the introduction of cfDNA testing has
rapidly reduced the use of chorionic-villus sampling and amniocentesis, the value to many patients of comprehensive fetal genetic analysis
and the increasing safety of diagnostic procedures are key to informed decision making. A
recent meta-analysis calculated the rate of pregnancy loss from amniocentesis at 1 loss in 1000
procedures; this rate was little different from the
background risk.5 To our knowledge, from a
population standpoint, no studies have evaluated the appropriate balance between cfDNA
screening and diagnostic testing. As in all prenatal diagnosis, pretest counseling is key to ensuring that couples choose the appropriate test
that is suited to their preferences.

The authors reply: Our study was designed

tocompare cfDNA testing with standard firsttrimester screening for trisomy 21 only in patients who had results from both tests and not to
compare performance in actual clinical practice.
The issue of no call results could be addressed
by a comparative-effectiveness trial following an
entire enrolled cohort, including patients who do
not complete first-trimester screening or have
no-call results on cfDNA screening.
On secondary analysis of results from the
NEXT cohort, the no-call group was found to
have an increased risk of aneuploidy (odds ratio,
6.35; 95% CI, 3.48 to 11.57), particularly among
samples with a low fraction of fetal cfDNA (odds
ratio, 11.4; 95% CI, 5.6 to 23.2). This increased
risk was present among patients with a no-call
result after the first attempted analysis; a second
sample was not obtained, as occurs in clinical
MaryE. Norton, M.D.
practice. Data from other publications and laboUniversity of California, San Francisco
ratories support this finding, with an odds ratio San Francisco, CA
from 4.2 to 9.2.1-4 To our knowledge, this group mary.norton@ucsf.edu
has not been studied directly, and data are lackRonaldJ. Wapner, M.D.
ing to clarify risks.
University Medical Center
A low fraction of fetal cfDNA is also associ- Columbia
New York, NY
ated with maternal weight; in the NEXT study,
publication of their article, the authors report no furthe median maternal weight in women with a therSince
potential conflict of interest.
low fraction of fetal cfDNA was 93.7 kg, as compared with 65.8 kg in women with a successful 1. Palomaki GE, Kloza EM, Lambert-Messerlian GM, et al.
sequencing of maternal plasma to detect Down syndrome:
result (P<0.001). Ideally, a correction factor for DNA
an international clinical validation study. Genet Med 2011;13:
maternal weight might be applied to determine 913-20.
how much a low fraction of fetal cfDNA in- 2. Palomaki GE, Kloza EM, Lambert-Messerlian GM, et al. Circulating cell free DNA testing: are some test failures informacreases the risk of aneuploidy.
tive? Prenat Diagn 2015;35:289-93.
Clearly, patients with assay failure for any 3. Pergament E, Cuckle H, Zimmermann B, et al. Single-nuclereason require follow-up with repeat cfDNA otide polymorphism-based noninvasive prenatal screening in a
and low-risk cohort. Obstet Gynecol 2014;124:210-8.
screening, traditional biochemical and ultraso- high-risk
4. Turocy JF, Norem C, Blumberg B, Norton ME. Chromosomal
nographic screening, or invasive diagnostic test- abnormalities detected in patients with failure to obtain test
ing. To wit, the status of Downs syndrome in results using non-invasive prenatal testing. Am J Obstet Gynecol
212:S45. abstract.
the three fetuses in the low fetal fraction group 2015;
5. Akolekar R, Beta J, Picciarelli G, Ogilvie C, DAntonio F.
was detected by means of standard first-trimes- Procedure-related risk of miscarriage following amniocentesis
ter screening (with measurement of serum ana- and chorionic villus sampling: a systematic review and metalytes and nuchal translucency). Further studies analysis. Ultrasound Obstet Gynecol 2015;45:16-26.
may evaluate methods of combining approaches DOI: 10.1056/NEJMc1509344

2582

n engl j med 373;26nejm.org December 24, 2015

The New England Journal of Medicine

Downloaded from nejm.org on January 18, 2016. For personal use only. No other uses without permission.
Copyright 2015 Massachusetts Medical Society. All rights reserved.