Abstract
Objectives. To evaluate neonatal screening for cystic fibrosis (CF), including study of the screening procedures and characteristics of false-positive infants, over the past 10 years in Wisconsin. An important objective evolving from the original design has been to compare use of a single-tier immunoreactive trypsinogen (IRT) screening method with that of a two-tier method using IRT and analyses of samples for the most common cystic fibrosis transmembrane regulator (CFTR) (ΔF508) mutation. We also examined the benefit of including up to 10 additional CFTR mutations in the screening protocol.
Methods. From 1985 to 1994, using either the IRT or IRT/DNA protocol, 220 862 and 104 308 neonates, respectively, were screened for CF. For the IRT protocol, neonates with an IRT ≥180 ng/mL were considered positive, and the standard sweat chloride test was administered to determine CF status. For the IRT/DNA protocol, samples from the original dried-blood specimen on the Guthrie card of neonates with an IRT ≥110 ng/mL were tested for the presence of the ΔF508 CFTR allele, and if the DNA test revealed one or two ΔF508 alleles, a sweat test was obtained.
Results. Both screening procedures had very high specificity. The sensitivity tended to be higher with the IRT/DNA protocol, but the differences were not statistically significant. The positive predictive value of the IRT/DNA screening protocol was 15.2% compared with 6.4% if the same samples had been screened by the IRT method. Assessment of the false-positive IRT/DNA population revealed that the two-tier method eliminates the disproportionate number of infants with low Apgar scores and also the high prevalence of African-Americans identified previously in our study of newborns with high IRT levels. We found that 55% of DNA-positive CF infants were homozygous for ΔF508 and 40% had one ΔF508 allele. Adding analyses for 10 more CFTR mutations has only a small effect on the sensitivity but is likely to add significantly to the cost of screening.
Conclusions. Advantages of the IRT/DNA protocol over IRT analysis include improved positive predictive value, reduction of false-positive infants, and more rapid diagnosis with elimination of recall specimens.
Cystic fibrosis (CF) is one of the most common autosomal recessive diseases in the Caucasian population, with incidence estimates ranging from 1 in 2000 to 1 in 4000.1 A diagnosis of CF is typically considered because of either a positive family history, neonatal meconium ileus, or the subsequent development of other disease manifestations, eg, failure to thrive, steatorrhea, and/or chronic lung disease. Routinely, the diagnosis is confirmed with a sweat test to demonstrate characteristically high chloride concentration. Unfortunately, there is often a delay in diagnosis.1Newborn screening for CF became feasible in 1979 when Crossley and co-workers2 showed that neonates with the disease have an elevated immunoreactive trypsinogen (IRT) level. Subsequent studies showed that the IRT assay, when used as either a single test method or with recall (repeat specimen on infants above a designated cutoff level), allowed identification of most asymptomatic infants with CF.3-5 Early detection is potentially advantageous because of the common delay in diagnosis and the opportunities screening creates to initiate therapeutic interventions and genetic counseling before the overt presentation of symptoms.6-9 However, the IRT test has several problems particularly in relation to its low positive predictive value. This can be improved by analyzing a second specimen,1,5 but the recall strategy has the drawback that for many infants in the United States obtaining a follow-up blood sample is not possible. Although only limited information is available, recall specimen loss as high as 22% has been reported for the Colorado study,5 however, in an Australian study, loss attributable to recall was only 2%.10
In 1985, comprehensive evaluation of CF newborn screening in Wisconsin was implemented, both to determine an optimal screening strategy and to determine the benefits, risks, and costs of early detection of individuals with CF. The evaluation of the clinical implications for early detection is ongoing and is not included in the current report. Assessment of screening procedures in the Wisconsin CF Neonatal Screening Project has been completed in a two-stage process. First, after four years of screening using a single IRT test, we reported a sensitivity of only 85%,11 a value similar to that reported by others.5 Lowering the IRT cutoff level to improve sensitivity was explored, but the corresponding increase in false-positive infants, who require sweat tests makes this approach undesirable, especially because the false-positive population includes disproportionately high representation of lower risk neonates.12 Consequently, we concluded that “the strategy for cystic fibrosis newborn screening will need to evolve into a true two-tier screening test”.11 With identification in 1989 of the gene responsible for CF (CFTR) and a mutation, ΔF508, that accounted for approximately 70% of all mutant CF chromosomes,13 it became possible to add DNA testing as a second tier to the IRT screen. Therefore, after developing molecular genetics procedures for analysis of ΔF508, using DNA extracts from Guthrie card specimens, we implemented an IRT/DNA testing protocol on a pilot basis with a lower IRT cutoff.14 Similar protocols have been studied sequentially in Australia.10,15 We now report a concurrent comparison between IRT and IRT/DNA testing protocols and other data from nine years of investigating CF neonatal screening procedures and false-positive infants. In addition, we demonstrate that multimutation testing adds very little when screening typical US populations that have a high frequency of ΔF508 CFTR and no other mutation at high frequency.
MATERIALS AND METHODS
The overall design of the Wisconsin CF neonatal screening project has been described previously.1,11 Approval for screening protocols was obtained from the Human Subjects Committee of the University of Wisconsin and the Research and Publications Committee/Human Rights Board of the Children's Hospital of Wisconsin. The investigation was designed to include extensive statewide surveillance for new CF patients that might have gone undetected. Two screening strategies, referred to as IRT and IRT/DNA testing have been compared; in both, the IRT level was determined by radioimmunoassay16 (Cis Sorin; Sylmar, CA) using samples from Guthrie cards typically obtained in the first four days of life (median, 2 days). For the IRT method, implemented on April 15, 1985 and terminated June 30, 1991, those infants with levels ≥180 ng/mL were considered positive. This IRT cutoff level was selected after reviewing of results from the Colorado program5 and analyzing approximately 10 000 consecutive dried-blood specimens to estimate the 99.8 centile. The IRT/DNA protocol was implemented on July 1, 1991 and ended June 30, 1994. Samples from neonates with IRT levels ≥110 ng/mL were analyzed for the presence of the ΔF508 CFTR mutation using samples obtained from the original Guthrie card. Those neonates with at least one ΔF508 allele were subsequently contacted and CF status determined using the quantitative pilocarpine sweat chloride test as the gold standard diagnostic procedure (≥60 mEq/L being positive). For both studies the IRT levels were blinded for half of the newborn population and this group became a standard diagnosis or control group; data from this group are not included in the current analyses because the unblinding process will continue until 1998.
Our molecular genetics methods have been summarized previously.14 We obtained denatured DNA from dried-blood specimens on Guthrie cards by first soaking the punched out filter paper circle in methanol and then performing a hot-water extraction. Polyacrylamide gel electrophoresis of PCR-amplified DNA served to identify the ΔF508 mutation.14 Mutations S549N, R553X, and G551D were screened by PCR amplification of exon 11, followed by restriction enzyme digests that are diagnostic of each mutation. PCR reactions contained Cetus buffer (10 mM Tris-HCl, pH 8.3, 50 mM KCl, 1.5 mM MgCl2), 200 mM each dNTP, 1.0 μm each primer (11i-3, 11i-5,17 and .4U Taq polymerase) (Perkin-Elmer Cetus), and either 1-μg genomic DNA or a sample of DNA prepared from a Guthrie card.14 Reactions were cycled (MJ Research PTC-100) with an initial denaturation step of 2 min at 95°C, followed by 30 cycles of amplification consisting of 1 min at 94°C, 1 min at 55°C, and 1 min at 72°C, ending with a 5-min soak at 72°C. 10 μl of PCR product used in separate restriction digests containing either MboI, HincII or DdeI. The digested DNA was analyzed on a 10% acrylamide gel, and the bands visualized by silver staining. The gel was soaked in .1% AgNO3 (v/v) for 20 min, rinsed in distilled water, and the bands developed in a solution containing 1.5% NaOH, .15% formaldehyde, and .01% sodium borohydride. Mutations, G542X, W1282X, R117H, R553X, N1303K, 1717–1G→A, R560T, and 621+1G→T were analyzed by the ARMS procedure using published primers and conditions.18
A total of 360 patients were studied by multimutation analysis (80% of the Wisconsin CF population). All patients had positive sweat tests and were being followed regularly at one of the state's two CF centers. DNA was obtained from either Guthrie cards, blood, or cheekbrush samples. DNA was isolated from whole blood using the Puregene DNA isolation Kit (Gentra Systems, Inc). Cheekbrush samples were processed as described.19 DNA samples with known mutations were obtained from Coriell Cell Repository and used as positive controls.
RESULTS
Table 1 provides summary information for the early diagnosis screened group, generated from April 15, 1985 to June 30, 1994. These data represent only half of all births in Wisconsin with the remainder of the neonates randomized to the standard diagnosis group. For the IRT method, .17% of all newborns were above the cutoff, whereas .13% were positive for the IRT/DNA method, and were referred for a sweat test. In the hypertrypsinogenic group of infants the frequency of the ΔF508 mutation was twice that found in the general population confirming an observation reported previously.14 The two screening protocols identified 67 CF patients (Table 1). In addition, 5 patients were identified because of meconium ileus and in the IRT group there were 5 false-negative patients, 3 of whom had CF siblings (Table 1). These data also give an overall estimate of the CF incidence (1:4223) in Wisconsin. Consistent with our previous report,14 CF incidence in Wisconsin is significantly less than the 1 in 2500 frequently quoted.20 Although it is remotely possible that a small number of CF patients have not been ascertained, our surveillance methods lead us to believe that an approximate incidence of 1 in 4000 is more accurate for the total populations of Wisconsin and US newborns.21 This incidence also is similar to the population incidence in the Colorado study.5
Summary of the IRT and IRT/DNA Screening Methods
Table 2 summarizes the sensitivity, specificity, and positive predictive values achieved with the two screening protocols. The specificity of the two protocols is very high, as with most neonatal screening tests,22 and would not be affected appreciably by typical variations in the IRT cutoff level.11 Sensitivity, on the other hand, showed considerable variation as in other CF screening studies.5,23 The IRT/DNA protocol tended to yield higher sensitivity because of the lower IRT cutoff value, but statistical analyses did not reveal significant differences compared to the IRT method. Nevertheless, during the 3-year IRT/DNA screening phase, the sensitivity was 100% excluding patients with meconium ileus (Table 2).
Sensitivity, Specificity, and Positive Predictive Value of the Two Screening Methods
Based on initial data,14 we had expected the positive predictive value of the IRT/DNA procedure to be greater than the IRT method. However, the data from the two different time periods show that the positive predictive values for the IRT and IRT/DNA methods are 12.5% and 15.2%, respectively, which are not significantly different. One possible reason for this is that the patient populations are different. To examine this possibility, we extracted those patients that would have been detected by the IRT screen from the IRT/DNA data set. These data are presented in the last row of Table 2. This allows direct comparison of the IRT and IRT/DNA methods using the same neonate population and shows that the positive predictive value of the IRT/DNA method was significantly greater than if the IRT only protocol had been used (15.2% vs 6.4%). The improvement in positive predictive value using the IRT/DNA method occurs in part because of eliminating false-positive infants. We have shown that African-American infants and also neonates with low Apgar scores often have an elevated IRT even though they have a relatively low risk of CF.12 Table3 shows that by comparison with the Wisconsin newborn population in general, the number of individuals with a normal Apgar score in the 7 to 10 range is significantly reduced in the false-positive group detected with the IRT screening methods, ie, the IRT test tends to identify stressed neonates with low Apgar scores.12 In contrast, individuals with a false-positive IRT/DNA screen have a similar Apgar distribution as the general population.
Apgar Score and Race Breakdown for False-Positive Patients
If the IRT false-positive infants are grouped by race, it can be seen that there is an overrepresentation of African-American infants with respect to their proportion in the general population of Wisconsin newborns. This is of concern because of the relatively lower incidence of CF in the African-American population (1 in 11 000−17 000),21 compared to the incidence of CF in the Caucasian population (1 in 2500 −4000).24 Breakdown of the false-positive infants for the two screening methods by race is shown in Table 3. African-Americans represent 27% of the false-positive infants in the IRT screen compared to 5.5% in the IRT/DNA population. The inclusion of the DNA tier to the screen therefore has the added benefit of excluding the overrepresentation of this population of infants, who are at lower risk for CF, from those requiring a sweat chloride test. This advantage of the IRT/DNA screening protocol would be magnified in areas in which a large proportion of screened infants are from African-American populations at lower risk for CF.
When a positive IRT/DNA screening result is communicated to the primary care physicians, they frequently wish to know if the IRT level has any predictive value in terms of the likelihood that a particular infant will have CF. Table 4 presents data indicating that neonates with very high IRT levels are considerably more likely to have CF compared to those with low values. Whereas there is a strong correlation, great care and adequate counseling should accompany use of such data in providing risk estimates. It also may be useful in deciding the urgency with which the follow-up sweat test, that always is recommended, is obtained. Our protocol required the use of the sweat chloride test to confirm a diagnosis of CF. In infants with two known CFTR mutations a sweat test may not be necessary; however, if this approach is used the DNA test should be repeated on a new sample before indicating the infant has CF, to eliminate any possible laboratory errors.
IRT Level as a Predictor of CF in Infants With Positive IRT/DNA (ΔF508) Screen*
Our current implementation of the IRT/DNA screening protocol only includes screening for the most common CFTR mutation, ΔF508, as a one mutation, second tier test. Our assessment of the 77 CF patients identified in the Wisconsin CF Neonatal Screening Project (Table 1) revealed that 55% had two ΔF508 alleles and 40% were compound heterozygotes. One way to improve the screening protocol further would be to increase the percentage of CF chromosomes detected. To examine this possibility we screened samples from 360 patients followed at Wisconsin CF centers for several of the more frequent CFTR mutations. Table 5 shows the frequencies of these mutations in the Wisconsin population. These frequencies are similar to previously published reports25 and reflect the distribution of CFTR mutations in the US population of CF patients (Table 6). We also were interested in the impact on the detection percentage of adding either all or some of these mutations to the DNA protocol. The frequency of ΔF508 is .7125 and therefore, assuming Hardy-Weinberg equilibrium, 91.74% of our CF patients will be expected to carry at least one copy of the ΔF508 mutation. Table 5 shows that in theory, there is only a 5.5% increase in the detection rate even when all 10 mutations are added to the DNA screen. This would result in detection of one extra CF patient per year in Wisconsin. In view of the increased cost that would be incurred with current methods to add the 10 mutations to the screen it is unlikely that this would be warranted for typical populations in the US. A decision with respect to the mutations used in the second tier would need to consider some potentially hidden costs, such as additional medical costs and legal costs that may be incurred because infants were missed by the screen. It should be noted that short of completely sequencing the CFTR gene in every newborn, some patients will be missed by screening. Further, in populations that have large subpopulations with high frequencies of a particular mutation, for example, Ashkenazi Jews (ΔF508 (27%) and W1282X (51%)),9 inclusion of other mutations to the second tier screen may be warranted. The selection of mutations to use in the screen, based on frequency in the CF population also should recognize that sometimes the frequency of a mutation in the general population is different from that in the CF population.9
Frequency of CFTR Mutations and Effect of Adding Each to CF Detection Using IRT/DNA Test
DNA Analysis of Genotyped CF Patients in the US*
DISCUSSION
Traditionally a diagnosis of CF was made because of either: 1) a positive family history, which occurs in about 20% of cases26,27; 2) the occurrence of meconium ileus; or 3) as a result of either intestinal malabsorption or chronic pulmonary obstructive disease. Once the characteristic signs and symptoms become evident, the disease can be diagnosed by performing a sweat test using quantitative pilocarpine iontophoresis. Unfortunately, the diagnosis of CF is often delayed. Data available from the US Cystic Fibrosis Foundations's Patient Registry indicate that the mean age in 1992 was 4.0 years and the median 1.1 years.1 Some studies suggest that delay in diagnosis is associated with worse disease, including severe malnutrition and irreversible pulmonary infections.4,28,29 Other observations, however, do not support a relationship between age of diagnosis and prognosis.22,30 On the basis of retrospective analysis of outcomes related to age of diagnosis, Shwachman et al31recommended newborn screening for CF in 1970. Newborn screening for CF screening became feasible in 1979 when Crossley et al2applied measurement of IRT to neonatal dried-blood specimens collected on Guthrie cards in New Zealand. They demonstrated that the IRT test could be applied successfully to such specimens, thereby allowing screening in centralized laboratories. Subsequent experience in other countries was encouraging. In the US, however, the Cystic Fibrosis Foundation recommended caution and more research, including critical assessment of the screening procedures and study of the value of early treatment related to prognosis; these recommendations were published in 1983 after a Task Force review.32
As a result of nine years of investigation in Wisconsin and elsewhere,10,15 an attractive test has been developed and its validity assessed. The method involves an initial measurement of IRT as a prescreen with a cutoff low enough to detect almost all CF infants without meconium ileus. Advantages of the IRT/DNA test compared with the IRT method include the following. The positive predictive value of the test is increased considerably. Populations at lower risk for CF are preferentially excluded from the screen. Both tiers of the screen use samples from the initial Guthrie card, eliminating the need to collect a second sample, which can result in 2 to 20% of the infants being lost to follow-up.5,10
Concurrent with our study, Ranieri et al15 and Wilcken et al10 have employed a similar strategy. They report results of adding DNA testing for the ΔF508 mutation, and Ranieri et al15 also investigated adding three other CFTR mutations (G542X, R553X, and G551D) to the screen. Although in theory the addition of these extra mutations should increase the sensitivity of the screen, in practice all the individuals detected with CF in our study had at least one copy of the ΔF508 mutation. The positive predictive value for the South Australian study of Ranieri et al15 was 31.5%, twice that for the Wisconsin study. Wilcken et al10 reported positive predictive values of 13% when one ΔF508 allele was present and 37.4% overall. The differences are attributable to different IRT cutoff values used in the various studies. In addition, the predictive value of a screening test is dependent not only on sensitivity and specificity but also on characteristics of the population screened, particularly the prevalence of preclinical disease in those tested. Furthermore, the incidence of CF appears to be higher in Australian newborns.10,15,22 In both Australian studies,10,15 only infants with an IRT level above the 99th centile were screened for mutations compared to the 98th centile (2036/104 308) in our study.
Table 4 shows that 1404 of the 2036 samples positive in the IRT prescreen (≥110 mg/mL) had IRT levels below ≥180 ng/mL; but, there were no CF patients in this group. These data indicate that one can potentially choose a higher IRT level for the first tier cutoff, without decreasing the sensitivity of the screen. It is now clear, however, that there are considerable problems with absolute measures of IRT over extended periods of time.16,33 Additionally, the IRT assay seems to exhibit a cyclical season-dependent change with a higher number of samples that need to be screened for CFTR mutations in winter compared to summer. Given the temperature extremes in Wisconsin, this is likely to be attributable to the lability of the IRT on the Guthrie card during normal postal transfer to the centralized IRT testing laboratory.33 To avoid these problems, we recommend a protocol which targets a predetermined number of high IRT specimens to be processed for DNA analysis thereby facilitating use of an optimum batch size to reduce costs, as we are currently doing in Wisconsin.34 This means that the absolute IRT value chosen as the cutoff changes slightly from day to day, but ensures that those samples most likely to be from CF patients are included in the DNA testing.
Based on the Wisconsin and Australian experience, it appears that processing the 1% highest IRT specimens (99th centile) is most appropriate. As described in detail elsewhere,34 for the routine IRT/DNA method now being used in Wisconsin, we have lowered the IRT cutoff further (56 ng/mL by the fluoroimmunometric method) and also alert physicians when the level is quite high and the DNA (ΔF508) test is negative. This maximizes the sensitivity, whereas the IRT/DNA method decreases the overall number of false-positive individuals.
Given that our group, Ranieri et al,15 and Wilcken et al10 have shown that IRT/DNA testing is superior to IRT alone, a decision as to the number of mutations to be screened for in the second tier must be considered. To obtain complete ascertainment, it would be necessary to include many more than the 10 mutations analyzed in this study. On the other hand, it is unlikely that any nonΔF508 mutations would have a frequency high enough to yield many CF patients in Wisconsin or in other typical US populations. Therefore, if multiple mutations are included in the screening protocol, careful consideration should be given to those that are added, depending on the population characteristics of the region being screened. For example, South Australia15 has a very small Jewish population and, consequently, analysis for the W1282X mutation was not included in the DNA screen. By contrast, this mutation was the third most frequent in Wisconsin, and in areas with even larger Jewish populations the omission of this particular mutation would result in decreased sensitivity of the screen.
We have shown that neonatal screening for CF using a two-tiered IRT/DNA screening protocol has high specificity and sensitivity. In addition, other advantages of the IRT/DNA protocol over IRT analysis include improved positive predictive value, reduction of known false-positive infants and more rapid diagnosis with elimination of recall specimens.
ACKNOWLEDGMENTS
This research has been supported by grant A001 5–01 from the Cystic Fibrosis Foundation and grants DK34108 and RR03186 from the National Institutes of Health.
Footnotes
- Received August 6, 1996.
- Accepted October 15, 1996.
Reprint requests to (P.M.F.) Professor of Pediatrics and Dean, University of Wisconsin Medical School, 1300 University Ave, Madison, WI 53706.
- CF =
- cystic fibrosis •
- IRT =
- immunoreactive trypsinogen •
- DNA =
- deoxyribonucleic acid
REFERENCES
- Copyright © 1997 American Academy of Pediatrics
In this issue
Jump to section
Related Articles
Cited By...
- Outcomes of Infants With Indeterminate Diagnosis Detected by Cystic Fibrosis Newborn Screening
- Diagnosis of cystic fibrosis in London and South East England before and after the introduction of newborn screening
- A Decision-Tree Approach to Cost Comparison of Newborn Screening Strategies for Cystic Fibrosis
- The Need for Vigilance: The Case of a False-Negative Newborn Screen for Cystic Fibrosis
- Growth and pulmonary outcomes during the first 2 y of life of breastfed and formula-fed infants diagnosed with cystic fibrosis through the Wisconsin Routine Newborn Screening Program
- Clarification of Laboratory and Clinical Variables That Influence Cystic Fibrosis Newborn Screening With Initial Analysis of Immunoreactive Trypsinogen
- Guidelines for Implementation of Cystic Fibrosis Newborn Screening Programs: Cystic Fibrosis Foundation Workshop Report
- Cost-effectiveness of 4 Neonatal Screening Strategies for Cystic Fibrosis
- Markedly elevated neonatal immunoreactive trypsinogen levels in the absence of cystic fibrosis gene mutations is not an indication for further testing
- Initially Misleading Communication of Carrier Results After Newborn Genetic Screening
- Content of Communication by Pediatric Residents After Newborn Genetic Screening
- Implementation of Newborn Screening for Cystic Fibrosis Varies Widely Between States
- Population-Based Newborn Screening for Genetic Disorders When Multiple Mutation DNA Testing Is Incorporated: A Cystic Fibrosis Newborn Screening Model Demonstrating Increased Sensitivity but More Carrier Detections
- Fatty Acids, Alpha-Fetoprotein, and Cystic Fibrosis
- Genetic Counseling and Neonatal Screening for Cystic Fibrosis: An Assessment of the Communication Process
- Early Diagnosis of Cystic Fibrosis Through Neonatal Screening Prevents Severe Malnutrition and Improves Long-Term Growth
- Blood concentrations of pancreatitis associated protein in neonates: relevance to neonatal screening for cystic fibrosis
- Cystic Fibrosis Newborn Screening: Impact on Reproductive Behavior and Implications for Genetic Counseling