Efficacy of Statewide Neonatal Screening for Cystic Fibrosis by Assay of Trypsinogen Concentrations

Feasibility and Validity of Screening by Trypsinogen Assay

Pilot Studies and Choice of Cutoff Levels

We performed pilot studies in three hospitals before implementing screening statewide, to gain experience in measuring trypsinogen and to determine a cutoff level for achieving an acceptable recall rate (i.e., to identify infants who required repeat measurements).¹⁶ Immunoreactive trypsinogen was measured in dried blood on 3.2-mm disks according to a modification of the radioimmunoassay described by King et al.¹⁷ In pilot studies, we attained a recall rate of 2 per 1000 by selecting an initial cutoff level of 140 μg per liter. The level remained at 140 μg per liter throughout the study except for a seven-week period (October 1 to November 20, 1985), when it was changed to 160 μg per liter to increase the specificity of testing. The cutoff level was changed back to 140 μg per liter because of our experience with one infant (initial trypsinogen concentration, 140 μg per liter), who presented with severe malnutrition and was found to have cystic fibrosis at 46 days of age. All infants with levels ≥140 μg per liter who were born during this seven-week period were subsequently retested according to the original protocol.

To demonstrate persistent elevation of the trypsinogen concentrations, a repeat sample was taken if the initial level was elevated.¹⁸ We chose to use a cutoff level for repeat testing that was below the initial cutoff level of 140 μg per liter to account for the expected decline in trypsinogen levels with age. Our first choice of a cutoff level for repeat testing, 120 μg per liter, was later changed to 80 μg per liter (after December 31, 1983) because an infant with cystic fibrosis was found to have an elevated level on initial testing but a level of 87 μg per liter on repeat testing. This infant was identified through a concomitant pilot program of screening with immunoreactive lipase.¹⁹ Thus, the use of two cutoff levels (120 and 80 μg per liter) on retesting was evaluated during the study period.

Screening of Newborns and Physician Notification

We screened all infants born in Colorado between April 1, 1982, and March 31, 1984, and between October 1, 1984, and September 30, 1987. The study protocol was approved by the Human Subjects Committee of the University of Colorado Health Sciences Center. The brochure given to mothers that explained neonatal screening included a section concerning cystic fibrosis and the voluntary nature of the study. Portions of the dried blood specimens collected when the infants were one to four days of age for routine genetic screening were assayed for immunoreactive trypsinogen. All specimens with levels above 100 μg per liter were reassayed in duplicate. A specimen was considered positive if the average of the two values was ≥140 μg per liter. If the two values differed by more than 15 percent, the specimen was assayed in duplicate again. With the cutoff level of 140 μg per liter, the intraassay coefficient of variation was 11.2 percent and the interassay coefficient of variation was 13.3 percent.

The physician of record was notified by telephone and by mail when a positive result was obtained. Information about the published predictive value of the test was given, and a repeat blood specimen was requested. The physician also received a consent form for the parents; it further explained the voluntary nature of the study. Repeat blood specimens were assayed in duplicate. If the average of the values of these specimens exceeded the cutoff level for repeat testing, the physician was notified by telephone and a sweat test was scheduled.

Outcome Variables

To determine the feasibility of screening statewide, we evaluated the recall rate and the infant's age at the time of repeat trypsinogen assay and sweat testing. To determine the efficacy and validity of screening statewide, we determined the incidence of cystic fibrosis as well as the false positive and false negative rates. Since infants with meconium ileus have clinical symptoms that allow a diagnosis to be made in the neonatal period before completion of the screening process, they were excluded from the computation of the false positive and false negative rates.¹⁵

Determination of the Age-Related Decline in Trypsinogen Levels

The rates of decline in trypsinogen concentrations during the first few months of life in healthy infants and infants with cystic fibrosis were determined from measurements obtained on initial and repeat testing.

Evaluation of Sweat Testing in Infants Less Than Three Months of Age

Figure 1. Figure 1. Study Design for Screening for Cystic Fibrosis (CF) by Trypsinogen Measurements in Dried Blood Spots.

Sweat-testing protocols A and B are described in the Methods section.

The diagnosis of cystic fibrosis was based on the elevation of sweat chloride concentrations (>60 mmol per liter) according to Gibson—Cooke quantitative pilocarpine iontophoresis, with a test yield of at least 50 mg.²⁰ Figure 1 outlines the study design for trypsinogen screening and the sweat-test protocols A and B.

Protocol A: Infants with Positive Initial and Positive Repeat Tests

If the infant resided in the Denver area, the parents were asked to bring the child to our laboratory for sweat testing. If the family resided elsewhere in Colorado, a technologist traveled to the family's hometown to perform the test. If the test performed at the laboratory was positive, the family was seen the same day by one or more staff members of the Cystic Fibrosis Center. If a test performed elsewhere was positive, the family's physician was notified by telephone and follow-up through the Cystic Fibrosis Center was arranged.

Protocol B: Infants with Positive Initial but Negative Repeat Tests

To evaluate our choice of cutoff levels for repeat testing, we performed sweat tests from October 1, 1984, to September 30, 1985, in 168 consecutive infants with a positive initial trypsinogen value (≥140 μg per liter) but a negative repeat value (<80 μg per liter).

Statistical Analysis

Statistical analysis was performed with a commercially available statistics package, SAS,²¹ with a VAX 1150 minicomputer. Confidence intervals for the incidence of cystic fibrosis were determined through normal approximation to binomial distribution.²² Exact confidence intervals for false negative rates were determined with the use of published tables.²³ Trypsinogen levels on initial and repeat testing and sweat sodium and chloride values at diagnosis were compared by paired t-tests, since paired values were available for each infant. All values are expressed as means ±SE.

Feasibility and Validity of Screening

Recall Rate and Age at Testing

We screened 279,399 newborns during the study period. Of the 96,434 infants born during the first part of the study (April 1, 1982, to December 31, 1983), when the repeat cutoff trypsinogen level was 120 μg per liter, 241 had a positive initial value (≥140 μg per liter), for a recall rate of 2.5 per 1000. Two infants had meconium ileus; 183 infants underwent repeat testing (compliance rate, 76.6 percent). Thirty-one of these 183 (16.9 percent) had positive repeat values (≥120 μg per liter). Sweat testing was accomplished in all these infants, and the diagnosis of cystic fibrosis was made in 23.

During the second part of the study (after December 31, 1983), when the cutoff level for repeat testing was 80 μg per liter, 182,965 infants were screened; 643 had positive initial values (≥140 μg per liter), for a recall rate of 3.5 per 1000. Five infants had meconium ileus; 510 infants underwent repeat testing (compliance rate, 79.9 percent). Ninety-seven of these 510 infants (19.0 percent) had positive values (≥80 μg per liter) on retesting. Sweat testing was accomplished in 96 of these infants, and the diagnosis of cystic fibrosis was made in 31.

Figure 2. Figure 2. Trypsinogen Values on Initial (□) and Repeat ( + ) Testing.

The panel at left shows the results in normal infants with values ≥140 μg per liter on initial testing and their values on repeat testing at ≤90 days of age. The panel at right shows the initial values in 59 of 61 infants with cystic fibrosis who did not have meconium ileus, as well as values in the 53 infants with cystic fibrosis who had repeat testing at ≤90 days of age. See the Methods section for details. The upper and lower dashed lines represent levels of 120 and 80 μg per liter, respectively. There is some overlap between the initial and repeat values in each panel.

The recall rate for the entire study period was 3.2 per 1000. Repeat trypsinogen testing was performed at a mean age of 38 days, and sweat testing at a mean age of 49 days. Figure 2 shows the results of initial and repeat testing in all normal infants with initial values ≥140 μg per liter and repeat tests at ≤90 days of age. Figure 2 also shows the results of initial testing in 59 of 61 infants with cystic fibrosis who did not have meconium ileus (2 were excluded because the exact date of initial testing was uncertain) and repeat testing in the 53 who were tested when ≤90 days of age.

Identification of Infants with Cystic Fibrosis

Seventy-three infants were identified during the study period as having cystic fibrosis. Twelve infants (16.4 percent; 95 percent confidence interval, 9 to 27 percent) presented with meconium ileus. Five of these 12 infants had negative initial trypsinogen values. Of the 61 remaining infants, 54 were identified through the screening process. Seven infants without meconium ileus who were born during the study period were not identified through screening and are described below.

The incidence of cystic fibrosis during the study period was 1 in 3827 (95 percent confidence interval, 1 in 3113 to 1 in 4967). Seventy of the 73 affected infants (95.9 percent) were born to parents who were white. The incidence among white infants (77.0 percent of newborns studied) was 1 in 3073, before adjustment for infants who may have not been identified because of noncompliance. When adjusted for compliance with repeat testing (79.1 percent), the estimated incidence among white infants was 1 in 2521 (95 percent confidence interval, 1 in 2035 to 1 in 3310).

False Positive Rates

The false positive rate for initial testing was 92.2 percent (639 of 693). The false positive rate for repeat testing was 25.8 percent (8 of 31; 95 percent confidence interval, 12 to 45 percent) during the first part of the study (cutoff level, 120 μg per liter) and 67.7 percent (65 of 96; 95 percent confidence interval, 57 to 77 percent) during the second part (cutoff, 80 μg per liter).

False Negative Rates

The false negative rate, which reflected false results due to laboratory errors, changes in procedure, or biologic factors (initial level < 140 μg per liter or repeat level < 120 μg per liter), was 11.5 percent (7 of 61; 95 percent confidence interval, 5 to 22 percent). The rate of false negative results due only to laboratory error or changes in procedure was 4.9 percent (3 of 61; 95 percent confidence interval, 1 to 14 percent). Each of the three laboratory errors involved an initial blood sample. In all three cases (including the case occurring during the seven weeks when a different initial cutoff level was used, as described in the Methods section), the level in the initial sample was greater than or equal to the initial cutoff level (140 μg per liter) but a repeat specimen was not obtained because the investigator failed to notify the physician.

Table 1. Table 1. Clinical Presentation of 73 Infants with Cystic Fibrosis Diagnosed during the Study Period.

The rate of false negative results due to biologic factors on initial testing was 4.9 percent (3 of 61; 95 percent confidence interval, 1 to 14 percent). Trypsinogen levels in two infants (130 and 123 μg per liter) among the three with initial levels below 140 μg per liter were close to the initial cutoff level. These infants presented with clinical signs that led to sweat testing. Only one infant (also described in the Methods section) had a false negative result on repeat testing (<120 μg per liter) during the first part of the study. Table 1 shows the diagnostic profiles of infants identified as having cystic fibrosis during the study period. Twelve infants had meconium ileus, and five of these had a negative initial test. Since they would have been recognized as having the disease on clinical grounds, they were not included in the calculation of the false negative rate. Since other infants with cystic fibrosis may have had no symptoms that led to sweat testing, the calculated percentage of false negative results should be regarded as a minimum.

Age-Related Decline in Trypsinogen and Comparison of Cutoff Levels for Repeat Testing

The mean trypsinogen concentration in infants positive for cystic fibrosis on initial testing (296±14 μg per liter) was significantly higher than that on repeat testing (258± 17 μg per liter) (P<0.02). Thus, the expected age-related decline in trypsinogen concentrations in cystic fibrosis, though small, was detectable during the first three months of life.

Figure 3. Figure 3. Percentage of Normal Infants (Stippled Bars) and Infants with Cystic Fibrosis (Solid Bars) with Trypsinogen Values ≥80 μg per Liter on Repeat Testing.

Trypsinogen values were persistently elevated in the infants with cystic fibrosis, whereas they fell rapidly in the normal infants. There was a significant difference (P<0.05) between the percentage in the normal group at 10 to 19 days and the percentages for all other periods.

To determine the optimal period for obtaining repeat blood specimens, we compared trypsinogen levels in infants with cystic fibrosis with those in normal infants. In the normal infants, the proportion of repeat values ≥80 μg per liter was significantly higher when specimens were collected at 10 to 19 days of life than at 20 to 29 days (P<0.01) (Fig. 3). There were no significant differences in the proportion of trypsinogen levels ≥80 μg per liter between samples from normal infants that were collected from 20 to 29 days of life, 30 to 39 days, 40 to 49 days, and 50 to 59 days.

Values obtained during the second part of the study, when the cutoff level for repeat testing was 80 μg per liter, were analyzed to evaluate the effect of lowering the cutoff level for repeat testing from 120 to 80 μg per liter. Three infants with cystic fibrosis (including one previously described as having a false negative result due to laboratory error; concentration, 140 μg per liter) had values on repeat testing between 80 and 120 μg per liter during this period. Thus, the calculated marginal benefit from lowering the cutoff level for repeat testing was the identification of 9.4 percent (3 of 32; 95 percent confidence interval, 2 to 25 percent) of the infants with cystic fibrosis who did not have meconium ileus. The marginal loss from lowering the cutoff level was an increase in the percentage of infants who had positive results on repeat testing, and who therefore required sweat testing, from 10.0 percent (cutoff, 120 μg per liter) to 19.2 percent (cutoff, 80 μg per liter).

Sweat Testing

Figure 4. Figure 4. Sweat Chloride Concentrations in Infants Less Than Three Months of Age.

Only one infant with cystic fibrosis had a concentration below 70 mmol per liter. Only one normal infant had a concentration above 40 mmol per liter.

We performed 353 sweat tests in infants 1 to 11 weeks of age (mean, 6.3), including 49 infants referred during the study period because of clinical symptoms or a family history of cystic fibrosis. Sweat yields were inadequate in three patients (0.8 percent; 95 percent confidence interval, 0 to 3 percent). Figure 4 shows sweat chloride values in the infants with cystic fibrosis and the normal infants.

Protocol A: Infants with Positive Initial and Positive Repeat Tests

Of the 54 infants in whom cystic fibrosis was diagnosed through screening, 53 had initial sweat chloride levels above 70 mmol per liter. One infant had a sweat chloride level of 46 mmol per liter on initial testing, but a level of 74 mmol per liter on subsequent testing after electrolyte replacement for hypochloremia. These infants had sweat chloride values (101.7±2.0 mmol per liter) that were higher than their sweat sodium values (80.2±1.9 mmol per liter) (P<0.001); four infants had sweat sodium values below 60 mmol per liter. Among the infants without cystic fibrosis, three had sweat sodium levels between 40 and 50 mmol per liter; only one infant in this group had a sweat chloride level between 40 and 50 mmol per liter.

Protocol B: Infants with Positive Initial but Negative Repeat Tests

To determine the validity of our choice of a threshold value for the results of repeat testing, we performed sweat tests in 168 infants with a positive initial test (≥140 μg per liter) and a negative repeat test (<80 μg per liter). All the results in this group were normal, leading to an estimated incidence of false negative results of less than 2 percent (95 percent confidence interval, 0 to 2 percent) with the use of such a cutoff level (<80 μg per liter).

We found that neonatal screening for cystic fibrosis through serial measurement of immunoreactive trypsinogen in dried blood spots is a feasible and valid tool for identifying affected infants. The application of this technique to a large population over a period of several years, in conjunction with the identification of infants with meconium ileus, resulted in early identification of affected infants less than two months of age, with an incidence close to that expected for the disease.^{24 , 25} The use of an initial cutoff level of 140 μg per liter resulted in an acceptable recall rate (3.2 per 1000) and a false positive rate (92.2 percent) similar to rates in neonatal screening programs for other diseases (Cunningham GC [Council of Regional Networks for Genetics Services]: personal communication). The number of false negative results could be reduced by the choice of an appropriate cutoff level for repeat testing and more attention to laboratory reporting. A repeat cutoff level of 80 μg per liter allowed adequately for the age-related decline in trypsinogen concentrations in infants who were less than three months of age when repeat specimens were obtained. Sweat testing was highly specific in infants in this age group.

Unlike most neonatal screening programs for other diseases, in which the diagnosis is confirmed by testing repeat specimens, trypsinogen screening for cystic fibrosis has an appreciable false positive rate even for repeat testing. Screening newborns throughout Colorado, however, required fewer than 200 repeat trypsinogen assays and fewer than 40 sweat tests per year. We view this degree of laboratory support as appropriate for this screening program and as clearly feasible. In addition, the false positive rate for retesting could be reduced further if the specimens for retesting were obtained after three weeks of age. However, delays in repeat testing may put stress on the infants' families.²⁶ The specificity and sensitivity of screening may be improved by performing mutation analysis after the discovery of an elevated initial trypsinogen level.¹⁵ The clinical importance of elevated trypsinogen concentrations in normal infants is not clear, although this finding has been associated with depressed Apgar scores in a small fraction of infants.²⁷

A recent report concluded that a cutoff level of 75 μg per liter for repeat testing would provide maximal specificity and sensitivity for trypsinogen screening.¹⁵ It cautioned against the use of trypsinogen screening, however, in part because of an unacceptably high false negative rate. The initial cutoff level of 180 μg per liter in that study led to a recall rate below 1.5 per 1000, the lowest reported among screening programs based on trypsinogen assay.^{11 , 12} The low recall rate probably contributed to the high false negative rate reported by that group.¹⁵ After excluding laboratory error, we found that 95 percent of the infants with cystic fibrosis who did not have meconium ileus could be identified with use of an initial cutoff level of 140 μg per liter and a cutoff level of 80 μg per liter on retesting.

Initial trypsinogen concentrations were not elevated in five infants with meconium ileus, confirming that levels may be normal in infants presenting with this disorder.^{15 , 28} In infants with meconium ileus or other conditions involving intestinal obstruction, such as meconium plug syndrome or jejunal and ileal stenoses, or in any patient with clinical evidence suggesting cystic fibrosis, the sweat test should be performed without initial trypsinogen measurement.

The only means of confirming the diagnosis of cystic fibrosis in infants with unidentified genetic variants of this disease remains the sweat test. During our study, sweat testing proved feasible and specific during early infancy. Borderline or ambiguous values on sweat testing are therefore not a major impediment to the establishment of screening programs. It is also clear, however, that for screening programs based on trypsinogen assay to succeed, the quality of sweat testing must be high. Our data indicate that during infancy, measurement of sweat chloride distinguishes infants with cystic fibrosis from normal infants better than does measurement of sweat sodium.²⁹

The cost of screening averaged $1.71 per infant. This figure includes all direct costs associated with initial and repeat assays, follow-up sweat testing, and reporting of results, but not the costs of blood sampling and processing or any indirect costs. The direct cost of identifying a single infant with cystic fibrosis through screening averaged $8,877. The cost per infant screened is similar to the incremental cost of $2.52 (direct plus indirect costs) at the Mountain States Regional Newborn Screening Laboratory of the Colorado Department of Health in 1990.

In conclusion, neonatal screening for cystic fibrosis can be efficiently performed by serial measurement of immunoreactive trypsinogen in dried blood spots and follow-up sweat testing. It is still unclear whether mass screening of newborns for cystic fibrosis should be adopted, since an overall benefit of early diagnosis has not been demonstrated. Potential benefits include fewer hospitalizations in the first years of life,^{2 , 3} diagnosis and treatment of unsuspected protein malnutrition and deficiencies of fat-soluble vitamins,^{4 , 5} earlier genetic counseling, and a reduction in the anxiety associated with delayed diagnosis.³⁰ Potential problems of screening include the psychological effects on families of infants with false positive results²⁷ and infants with cystic fibrosis.