Discussion
In this comprehensive analysis of CF NBS algorithms in the USA, we found
substantial variations in IRT cutoffs and CFTR variant analysis.
These results have important clinical implications for the effectiveness
of different states’ CF NBS programs in identifying newborns with CF,
and they provide compelling evidence of the need for a nationwide QI
effort to make CF NBS more consistent across the country and aligned
with best practices.
Although the Centers for Disease Control and Prevention and the
Association of Public Health Laboratories (APHL) are two
federally-supported organizations with national responsibilities for
NBS, they have never shared regional NBS methodologies and outcomes. The
only other study that presented data on state CF NBS algorithms was
published by Pique, et al in 2015[12]. However, data for that study
was acquired at a time when many states were using a CFTR variant
analysis kit (Hologic, Marlborough, MA) that was subsequently withdrawn
from the market due to manufacturing defects. Furthermore, at the time
of their study, many states had not yet progressed to a DNA-based
algorithm. Our study presents comprehensive, up-to-date data on CF NBS
in the USA at a time when all states are using IRT/DNA algorithms.
Our results have important implications for equity in CF NBS. For
example, the two states analyzing only for F508del in the high IRT
specimens cannot exceed 85% sensitivity in a typical population of
racially and ethnically diverse American infants, However, it has been
known for decades[11, 13] that African American and Hispanic infants
have variants other than those in commonly used CFTR panels. In
fact, the American College of Medical Genetics[14, 15] variant lists
were developed primarily for the white population. Although some
biotechnology companies have expanded their CFTR panel kits to
provide broader racial and ethnic coverage, even with the use of larger
CFTR variant panels, e.g. 139 variants, more infants from racial and
minority groups will be missed compared to white infants (M McGarry,
unpublished observations). Ultimately CFTR sequencing
methods[16, 17] are needed to better address the equity challenges.
Our results show that an infant’s probability of being diagnosed early
through CF NBS depends on where the birth occurs and also potentially on
when a baby is born, i.e., the season of the year. These differences are
attributable to both the DNA/CFTR tier and IRT cutoff values and
whether they are fixed or floating. Assessment of IRT has demonstrated
conclusively that it is a heat-labile biomarker[3, 7] that is also
affected by kit-related variations[3, 18]. Every state now shows
ambient temperature swings that are being increased by climate change.
Although the impact of high temperatures in lowering IRT levels can
potentially be mitigated by expedited courier delivery of NBS specimens,
the kit-to-kit variations persist. In fact, IRT levels as low as 40
ng/ml have been recorded in states with a 95thpercentile cutoff value [3]. As Martiniano et al reported from
analysis of 14 years of monthly IRT data, these variations have been
evident since at least 2006 and IRT levels show a downward drift in
recent years [18]. That study also revealed the clinical impact of
IRT levels slightly below fixed IRT cutoff values as they cause false
negative results and “missed cases.” Consequently, while our survey
was underway, Colorado, after critically reviewing a large database of
over 800,000 babies screened, changed from a fixed threshold at 60 ng/ml
to a floating cutoff at the 96th percentile. This is
an example of the type of large database driven QI effort needed in many
states to reduce the number of false negative results rather than
relying on short term, small data sets for evaluations and cutoff
conclusions—the apparent modus operandi of most states. Our
results demonstrate that many states are not utilizing the optimal
method of determining IRT cutoff by continuing to use fixed cutoffs.
It should be emphasized that none of the NBS tests for other genetic
conditions show such great variation in cutoff values. Even though the
tests used in screening for congenital hypothyroidism have shown
seasonal and kit-related variations[19], their impact has apparently
not altered sensitivity but does lower the positive predictive value in
colder months. The issue of fixed and floating cutoffs has been
described in detail in a document published by APHL[20]. In a
section on CF, it is stated that “The IRT cutoff is floating and/or
fixed. A floating cutoff is recommended because IRT is subject to
seasonal variations and lot-to-lot variability of the reagents.” In
addition to a higher likelihood of achieving equity, the floating cutoff
provides the advantage of a predictable number of samples for DNA/CFTR
analyses in the second tier.
It has become increasingly clear that the various CF NBS algorithms are
not equivalent in sensitivity and efficiency/timeliness. False negative
results are more likely with higher IRT cutoff values[3, 18] and
fewer CFTR variants[12]. This raises the question of how such
wide variations in CF NBS algorithms arose. A review of the historical
evolution of CF NBS tests provides possible answers. In the case ofCFTR variant analysis, the likely explanation is that DNA
biotechnology has evolved faster than NBS labs have been able to keep up
with opportunities to expand their panels, while adding costs as the new
options were marketed in association with greater knowledge ofCFTR pathogenic variants of CF patients[10, 21].
As for the variations in IRT cutoff values, and whether they are fixed
or floating, it seems likely that the answer also lies in an historical
perspective. Originally, before the discovery of the CFTR gene in
1989, all CF NBS algorithms were IRT/IRT and required what we now regard
as relatively high cutoff values to ensure that screening was practical.
The CFF raised questions about this and other aspects of IRT-based
screening in an influential 1983 position paper[22]. However even
after the IRT/DNA or IRT/IRT/DNA algorithms were implemented, the
2-sample states continued with fixed cutoffs through 2020 with the
exceptions of Texas and more recently Colorado.
The limitations of this study include our focus on a narrow window of
time, i.e., the second half of 2021, during a period in which algorithm
changes were occurring. For example, we learned during the analysis of
data that Oregon is changing to a floating IRT cutoff and the following
states are transforming to next generation sequencing: Florida,
Kentucky, and Utah. CF NBS algorithms are constantly changing in every
state, and it is likely that regular surveys like our will need to be
conducted in order to maintain accurate and current information on CF
NBS practices in the USA. In addition, on a national basis we only
evaluated the initial IRT cutoff value in the 2-specimen states that
employ IRT/IRT/DNA. However, very little research has been done on the
optimal IRT cutoff value for the second specimen, and this can be a
source of false negative results also[23].
The wide variations in CF NBS algorithms are unique among newborn
screening protocols, and involve both IRT cutoffs and CFTR variant
analysis. The only consistency is that all states now use a 2-tier
strategy beginning with IRT and then, if it is out-of-range, progressing
to CFTR variant analysis. Although CF NBS has been offered in the
USA now for over a decade, our results demonstrate the need for
continued improvement and modification of CF NBS algorithms in order to
optimize detection of CF newborns and achieve equity and inclusion in CF
NBS.