We created a Minnesota-specific pathogenic variant panel by identifying CYP21A2 pathogenic variants specific to the Minnesota population to add to the panel of 12 common CYP21A2 pathogenic variants used clinically for diagnosis. Blood samples were collected from Minnesota children with CAH and their parents at participating sites (University of Minnesota Masonic Children’s Hospital, the Mayo Clinic, and Children’s Hospitals of Minnesota). The Centers for Disease Control and Prevention (CDC) Newborn Screening and Molecular Biology Branch processed the blood samples to extract DNA, cryopreserve white blood cells, and create DBS. The study was approved by the institutional review board at all clinical sites. Informed consents from parents and assents from minors were obtained from all study participants. The CDC’s National Center for Environmental Health Office of Science determined that this work was not considered human subject research under 45 CFR 66.012(d) as all specimens were deidentified to the CDC.

Disease-causing variants were detected using a long-range amplification-based method developed at the CDC (15) that amplified the CYP21A2 gene as well as any recombinant 30-kb deletion or gene conversion alleles for each blood sample, and these amplicons were then detected by the Qiagen QIAxcel Capillary electrophoresis instrument and/or agarose gel electrophoresis (Fig. 1). All reactions contained an endogenous positive control to detect general assay failure. CYP21A2 copy number was confirmed by a multiple ligation probe amplification (MLPA) SALSA MLPA CAH P050 kit from MRC Holland on all samples including samples negative for CYP21A2 amplification and samples with apparent homozygous CYP21A2 sequence to identify any potential hemizygosity. Amplicons were then Sanger-sequenced to detect variants in CYP21A2 and the deletion breakpoints in any 30-kb deletion and gene conversion alleles, using the SeqScape software. Each allele’s predicted clinical phenotypes were determined by previous reports and publications, and phasing of multiple pathogenic variants in patient samples, cis vs trans, were inferred through inheritance patterns from parental samples. The novel variants’ potential pathogenicity was modeled using the amino acid substitution algorithms SIFT and PolyPhen2 (16).

A high-throughput allele-specific primer extension (ASPE) assay was developed at the CDC to detect the CYP21A2 pathogenic variants identified in the Minnesota CAH population in addition to the core 12 CYP21A2 pathogenic variants common to clinical testing panels. Following DNA extraction, the assay was accomplished in 2 parts: 1) long-range polymerase chain reaction (PCR) amplification of the CYP21A2 gene and detection of the 30-kb deletion alleles, and 2) Luminex-based ASPE assay for the expanded Minnesota-specific panel of pathogenic variants. The ASPE assay was modeled after the assay developed by Oh et al (17) but modified to capture both large-scale gene conversions as well as small microgene conversions (intragenic conversions) in a single amplicon. Both assays were transferred from the CDC to the Minnesota Department of Health (MDH) where they were validated for in-house use.

Validation of the ASPE assay at CDC was performed using DBS samples from Minnesota families. The 30 kb deletion PCR and ASPE assays were validated through repeat testing from samples with previously known clinical genotypes. Allelic ratios for the ASPE genotypes were set through least 20 intrarun test results and 20 between-run tests, each repeated by 2 analysts, for each pathogenic and reference allele. ASPE primers, amplification primers, and sequence primers are listed in Supplementary Tables 1, 2, and 3 respectively (18).

We performed a retrospective analysis of 1 year of NBS results generated by the MDH NBS program, comparing the metrics of the 2-tier steroid-based NBS protocol used by MDH that year vs a potential molecular-based NBS protocol using a lowered 17OHP cutoff. Both the molecular-based NBS protocols (Fig. 2) and steroid-based (Fig. 3) NBS protocols were assessed by examining DBS results from specimens received by MDH from January 1, 2015 to December 31, 2015. In all protocols first-tier 17OHP level, as measured by a PerkinElmer FIA assay performed on AutoDELFIA or Genetic Screening Processor (GSP) instrumentation, was assessed to determine whether second-tier testing was warranted. The steroid-based protocol used different weight-based 17OHP cutoff levels than the molecular-based protocols to trigger second-tier testing (Table 1).

For all protocols, the final diagnostic result after confirmatory testing was used to classify screening results for the infant as true positive (TP), true-negative (TN), or FP; in addition, those infants not referred for further testing were classified as TN. However, some infants classified as TN were subsequently reclassified as FN based on later, clinical diagnoses.

If the first-tier screening result was out of range, the original DBS was automatically sent for a second-tier diethyl ether extraction-based assay without contacting the newborn’s family (see Fig. 3). If the second-tier 17OHP level was normal, the result was reported as screen-negative. If the second-tier 17OHP level was elevated (≥19 ng/mL), the result was considered screen-positive and the newborn was referred to a pediatric endocrinologist for serum 17OHP and/or adrenocorticotropin stimulation confirmation testing to determine if the newborn was TP or FP.

In Minnesota, per the MDH 3-screen low birth weight steroid protocol, infants with a birth weight of less than 1800 grams had 3 17OHP FIA routine screens, the first collected at 24 to 48 hours, a second at 14 days, and a third at 30 days or at discharge. If any screen had elevated first-tier 17OHP, it was automatically sent for second-tier 17OHP testing. Confirmatory diagnostic testing was performed only if the third screen had elevated second-tier 17OHP or if the infant was symptomatic any time during the 3 screens.

Deidentified specimens received by the MDH from January 1, 2015 to December 31, 2015, were pulled for retrospective molecular analysis if the first-tier 17OHP FIA level was above the lowered 17OHP cutoff levels shown in Table 1. The lowered cutoff levels were based on the 17OHP levels of the known FN cases from 1999 through 2015, which resulted in the first-tier 17OHP cutoff levels being 33% to 47% lower than the steroid-based NBS protocol, depending on the weight category (3, 4). If the first-tier screening result was above the lowered 17OHP cutoff level, second-tier CYP21A2 genotyping was performed on the original DBS (see Fig. 2) using the Minnesota-CDC panel of pathogenic variants. If no CYP21A2 pathogenic variants were found, the result was considered screen-negative. If 1 or more pathogenic variants were found, the sample was sent for third-tier CYP21A2 sequencing to validate the genotypes from the second-tier ASPE-genotyping assay. Samples confirmed to have 1 pathogenic variant were deemed TN. Samples confirmed to have 2 or more pathogenic variants were considered screen-positive, and the newborn would be referred to a pediatric endocrinologist for serum 17OHP and/or adrenocorticotropin stimulation confirmation testing to determine if the newborn was TP or FP (see Fig. 2).

The final diagnostic result was used to classify the steroid-based NBS protocol and molecular-based NBS protocol result as TP, TN, FP, or FN. All screening metrics were calculated per each infant rather than per sample.

Separately for each birth-weight class and for all children combined, we compared the steroid-based and molecular-based NBS protocols according to these screening metrics:

The protocols were compared for FNR and FPR using a 2-sided exact McNemar test. For PPV, we used a permutation test because it is effectively exact and accounts for having applied the 3 NBS protocols to the same children. The permutation test was 2-sided, as follows: The test statistic was PPV for the molecular protocol minus PPV for the 2-tier steroid assay protocol; 10 000 artificial (ie, randomly permuted) data sets; for each such data set, the 2 protocol labels (steroid assay and molecular) were randomly permuted separately for each child. The P value is the fraction of the 10 000 artificial data sets in which the absolute value of the test statistic was greater than or equal to the absolute value of the test statistic in the actual data.

A secondary analysis examined the association between first-tier 17OHP concentrations (measured by FIA) and CYP21A2 pathogenic variants grouped as “no pathogenic variant,” “1 pathogenic variant,” and “≥2 pathogenic variants.” Because the distribution of 17OHP concentration had an extremely long right tail, we considered the following tests: one-way analysis of variance (ANOVA) of the first-tier 17OHP concentrations, one-way ANOVA of the logarithm of the first-tier 17OHP concentrations (addressing the long tail), and the Wilcoxon rank-sum test (addressing the long tail and outliers).

In recent years, reported FNR rates in CAH newborn screening vary among state NBS programs in the United States. Although some programs report no FNs, the accuracy of reported rates is susceptible to unreported cases of classic CAH that get diagnosed months or years after the original screen and depend on the diagnosing physician reporting the missed case to their state NBS program.

Biological factors can lead to FNs in steroid-based assays applied to specimens collected soon after birth. Transiently low 17OHP levels in a CAH-affected infant can be due to suppression of the fetal HPA axis activity in utero either by increased sensitivity of the fetal HPA axis to maternal cortisol or by increased maternal cortisol levels in fetal circulation due to decreased inactivation of maternal cortisol to cortisone by the placental enzyme 11β-hydroxysteroid dehydrogenase type II (11β-HSD-2).

During pregnancy most maternal cortisol cannot normally reach the fetus because it is oxidized by the placental enzyme 11β-HSD-2 to cortisone, the inactive form of cortisol (20, 21). Downregulation of the 11β-HSD-2 gene may alter the enzyme activity, allowing a greater proportion of maternal cortisol to cross into the placenta and then enter fetal circulation. Also, by 34 to 35 weeks’ gestation, activity of placental 11β-HSD2 decreases, which in turn allows a surge of maternal cortisol into fetal compartments (22, 23). In a fetus affected with CAH, a decrease in 11β-HSD-2 activity would allow higher amounts of maternal cortisol in fetal circulation, leading to decreased activity of the fetal HPA axis and transiently low 17OHP levels at 24 to 48 hours after birth when the DBS sample is collected. The later the timing of the collection of an NBS or serum confirmation sample from affected infants, the higher the 17OHP concentration (3–5). Another potential cause of transiently low 17OHP is increased glucocorticoid sensitivity to maternal cortisol that crosses into fetal circulation. Specific glucocorticoid receptor haplotypes have been associated with varying glucocorticoid sensitivity, including those that cause individuals to be more sensitive to glucocorticoids (24, 25).

In the present study, both cases of CAH-affected newborns missed under the steroid-based NBS protocol illuminate the shortcomings of steroid-based first-tier and second-tier assays. Steroid-based assays, even the most sensitive, cannot account for the influence of factors during intrauterine life, fetal HPA axis glucocorticoid sensitivity, or the state of the HPA axis at the time of the NBS collection. In case 1, the first-tier 17OHP by FIA level was below the steroid-based NBS protocol cutoff level so that second-tier steroid testing was not triggered and it was reported as a TN (see Fig. 3). In contrast, the 17OHP level was above the molecular NBS protocol’s cutoff level and would have triggered second-tier molecular testing that identified 2 CYP21A2 pathogenic variants. In case 2, the first-tier 17OHP level was elevated (screen-positive) and automatically triggered second-tier testing. However, the second-tier diethyl ether extraction assay reported the 17OHP level as normal; therefore, the result was deemed screen-negative and the newborn was not referred for confirmatory testing. Specimens of newborns affected by CAH correctly identified to be out of range by first-tier FIA but ultimately negated by normal second-tier results have been reported (3, 4). In our previous study of NBS for CAH in Minnesota, second-tier testing by LC-MS/MS overturned 4 cases with out-of-range first-tier results, leading to 4 FN cases (3). Under the molecular NBS protocol, case 2’s elevated first-tier 17OHP level would have triggered second-tier molecular testing, which identified 2 CYP21A2 pathogenic variants.

In addition to the risk of adrenal and/or SW crisis, late diagnoses due to missed cases of CAH can harm children’s health outcomes, including growth and pubertal development. As we previously reported, 5 male children with SV-CAH missed by Minnesota NBS in the years 1999 to 2010 and not diagnosed until ages 2.3 to 5.5 years had significant bone age advancement (median 7 years difference between bone age and chronological age) and required multiple costly interventions to address precocious puberty, advanced bone age, and reduced predicted final adult height (4). All 5 had increased emergency room visits or hospitalizations before diagnosis due to protracted symptoms with acute illness. During the same period, 3 female children with SV-CAH missed by NBS had atypical genitalia at birth, but were not diagnosed until months or years later (3 months, 3.4 years, and 6.5 years) (3). Two of these female patients also had evidence of SW and were started on fludrocortisone.

A potential solution to address FNs in steroid-based protocols would be to lower the 17OHP cutoff levels of first-tier FIA to correspond to the lowest levels of known FN cases within a state’s NBS program as was done in our study for the molecular protocol. Hypothetically, if the MDH lowered the steroid-based protocol’s 17OHP cutoff levels to the same as the molecular protocol, it would have generated 4018 first-tier screen-positives that would be sent to a second-tier diethyl ether extraction method rather than the 621 first-tier samples that were actually sent that year. Using the same percentage of FPs that occurred under the steroid-based protocol would have led to approximately 595 FPs in this hypothetical example, a significantly higher number than would be acceptable.

FPs have long presented a challenge to NBS for CAH. Levels of 17OHP can be transiently high at birth, particularly in premature newborns (5, 26). During the first 2 weeks of life in preterm infants, 17OHP levels have been shown to be inversely correlated with postconceptional age, with the normalization of 17OHP concentrations being slower in infants with birth weight of less than 1000 g and gestational age between 26 and 28 weeks (27). Several confounding biological factors associated with prematurity such as physiologically lower enzymatic activity of the 21-hydroxylase and/or 11β-hydroxylase enzymes due to immaturity of the adrenal glands, birth stress, illness, and poor kidney function can cause transiently elevated 17OHP levels and high FPRs in NBS for CAH in premature babies (28–30). We have previously reported that later sample collection in premature babies (<1800 g) plays a more important role in reducing FP results than the use of a second-tier steroid-assay (5). In the present report (see Tables 3 and 4), most FPs for the steroid-based NBS protocol were seen in newborns weighing less than 2500 (76/91) FPs. In contrast, the molecular protocol resulted in 52 FPs with only 5 in newborns weighing less than 2500 g. The reason for the significantly higher number of FPs in newborns weighing less than 2500 g is that first- and second-tier steroid-based assays are time dependent in that they capture 17OHP concentrations at the time of sample collection, which can be transiently high thus leading to more FPs. Steroid-based assay’s time dependency was the impetus for the 3-screen low birth weight protocol (collections at 24–48 hours, at 2 weeks, and at 30 days) as it allows for the HPA axis to recover. The reason for fewer FPs in less than 2500-g newborns in the molecular protocol is because second-tier molecular screening results are not affected by the immaturity of the HPA axis. As a result NBS samples with transiently elevated 17OHP levels are less likely to have multiple pathogenic variants and therefore second-tier molecular testing will lead to fewer FPs in newborns weighing less than 2500 g compared to those weighing more than 2500 g.

Families and individuals commonly have multiple CYP21A2 variants in cis configuration (ie, on the same parental chromosome), which presents a challenge for interpreting and reporting test results. As part of testing the 103 families, we found that 13% of the CAH families in this study had 2 or more CYP21A2 variants on the same parental chromosome as determined by the inheritance patterns of the parents to the affected probands.

Since parental blood samples were not available for the 1-year retrospective analysis of NBS specimens, it could not be determined if multiple pathogenic CYP21A2 variants were in cis, indicating a carrier state, or in trans, indicating a possible CAH disease state. Future research on improving the identification of the phase of detected pathogenic variant without the need for parental blood samples may help elucidate a molecular protocol that could eliminate both FPs and FNs while maintaining specificity. Using the Minnesota family samples with 2 or more pathogenic CYP21A2 variants and known phasing, Greene et al (31) showed that manual determination of phasing by using an allele-specific PCR and a total of 5 variant sites can improve assessment of CAH disease risk in a newborn without the requirement of family studies for genetic phasing. Until new sequencing techniques allow phasing of the pathogenic variants in the NBS samples newborn without the requirement of family studies, it would be timelier to refer all newborns with 2 or more pathogenic variants found by second-tier directly for confirmatory testing and skip third-tier sequencing.

Lowering the first-tier 17OHP cutoff level could present another challenge. When the first-tier 17OHP cutoff levels in the molecular NBS protocols were set to the lowest known 17OHP levels of known FN cases in Minnesota from 1999 through 2015, 4018 specimens were reflexed to second-tier CYP21A2 genotyping whereas 621 specimens were reflexed to second-tier using the steroid-based protocol’s higher first-tier 17OHP cutoff levels (see Table 1). If the 17OHP cutoff level of the molecular protocol was set higher at 18 ng/mL instead of 16 ng/mL, the number of samples reflexed to second-tier CYP21A2 genotyping would have dropped to 3191 and we still would have identified the 2 FN samples missed by the steroid-based protocol.

Another challenge in implementing a molecular NBS protocol is the time it takes to report the final result. In most cases the timeline for an NBS program to complete the 3 tiers of the molecular NBS protocols would be 5 days. However, various factors could affect the timeline (eg, batch analyzing, number of personnel, equipment) and the process could take up to 7 days, which would still be within a comparable timeline seen in states requiring 2 screens. Currently 13 states require a routine second screen from all infants between 8 and 14 days of life, which improves NBS metrics because it allows the transient levels to reflect disease state. A recent report from the Northwest Regional Newborn Screening Program found that 25% of affected infants were identified on the second screen, of whom 39% were diagnosed with SW-CAH (32).

One limitation of this study is that retrospectively performing the second-tier CYP21A2 genotyping and third-tier CYP21A2 sequencing of 1 year’s NBS specimens does not allow reporting of potential confounding factors that may be identified when performing the second-tier and third-tier tests live. Also, while the study’s retrospective nature allowed identification of cases of CAH missed by NBS, it is still possible that other children with CAH were not yet diagnosed or had moved to another state. A prospective, multi-state study incorporating molecular testing into current protocols would be needed to assess metrics, costs, and confounding factors.

Another study limitation was the use of a single year of NBS data. The numbers of FP and FN results for CAH fluctuates from year to year, as does the PPV, which in surrounding years varied in a range from 1% to 4%. That range is low compared with the roughly 8% PPV previously reported for 2-tier screening for CAH in Minnesota using a second-tier LC-MS/MS assay (12).

In conclusion, although introducing molecular testing into NBS protocols significantly improved the overall NBS for CAH metrics, a state NBS program would need to consider many factors as molecular testing in the NBS for CAH poses its own unique challenges.