Thirty-years of screening for cystic fibrosis in East Anglia
- Jacqui Calvin1,
- Sarah L Hogg1,
- Donna McShane2,
- Sharon Anne McAuley2,
- Richard Iles2,
- Robert Ross-Russell2,
- Fiona M MacLean3,
- Mary E Heeley4,
- Anthony F Heeley4,
- Written on behalf of the Norfolk, Suffolk and Cambridgeshire Paediatric Cystic Fibrosis Network
- 1Biochemical Genetics Unit, Cambridge University Hospitals NHS Foundation Trust, Cambridge, UK
- 2Department of Paediatrics, Cambridge University Hospitals NHS Foundation Trust, Cambridge, UK
- 3Medical School, St George's University of London, London UK
- 4Formerly Biochemical Genetics Diagnostic Unit, Peterborough District Hospital, Peterborough, UK
- Correspondence to Dr Jacqui Calvin, Biochemical Genetics Unit, Cambridge University Hospitals NHS Foundation Trust, Box 247, Hills Road, Cambridge CB2 0QQ, UK;.
- Received 23 March 2012
- Accepted 4 September 2012
- Published Online First 16 October 2012
Background Newborn screening for cystic fibrosis (CF) relies on the measurement of immunoreactive trypsinogen (IRT) originating from the pancreas. The Norfolk, Suffolk and Cambridgeshire screening programme initially exploited the persistent increase in IRT seen in CF (IRT-IRT protocol) and later changed to include mutation analysis as a second tier test (IRT-DNA-IRT protocol).
Results During a 30 year period 582 966 babies have been screened by IRT-IRT and 147 764 by IRT-DNA-IRT (total 730730), resulting in 296 screen positive cases of CF and 29 false negatives (including 10 false negatives with meconium ileus). Ten missed CF cases were pancreatic insufficient, however all were diagnosed before their first birthday, suggesting that a false negative result did not forestall appropriate clinical investigation. The IRT-DNA-IRT protocol had a much improved positive predictive value (PPV) of 85.9% compared to 67.3% for IRT-IRT, excluding CF babies with meconium ileus. The PPVs increased to 82.2% and 98.2% respectively if only well, term babies were considered. The main factor to account for this improvement in PPV has probably been the incorporation of DNA analysis in the second tier testing.
Conclusions The diagnosis of screen-positive babies proved difficult in a minority of cases with the classification of some patients changing with evolving phenotype. Our results illustrate the importance of collecting outcome data over a long time period for accurate assessment of the screening programme. This study provides evidence that newborn screening for CF is a valid undertaking that detects 95% of unsuspected CF cases presenting before 3 years of age.
What is already known on this topic
Newborn screening programmes detect the majority of unsuspected cases of cystic fibrosis (CF).
A normal screening test does not preclude a diagnosis of CF.
IRT alone lacks specificity in screening for CF and needs to be combined with second line tests.
What this study adds
This paper summarises the outcome data, in terms of true positives, false positives and false negatives for 30 years of newborn screening for CF.
Newborn screening for cystic fibrosis (CF) was introduced in Norfolk, Suffolk and Cambridgeshire, UK in January 1980, following a pilot scheme to assess cut-offs.1 The screening programme utilised the measurement of pancreatic immunoreactive trypsinogen (IRT) in dried blood spots, based on the observation by Crossley et al2 that IRT is increased in babies with CF. Throughout the 30 year period both the test protocol and reagents used have changed. Initially a persistent increase in IRT on a repeat sample constituted a positive screen. The protocol was changed to include mutation analysis in May 2004, similar to that described by Pollitt et al.3 This paper summarises the outcome data, in terms of true positives, false positives and false negatives.
Table 1 shows the methods and protocols used over the 30 year period together with the action limits for IRT. Initially the first samples were collected between days 6 and 10 of life, changing to days 5–8 (ideally day 5) from April 2000 onwards. The table does not take into account temporary changes in the cut-offs adjusted in response to new reagent kit lots.
From January 1980 to April 2004 (inclusive) a two stage IRT-IRT protocol was used, taking advantage of the prolonged hypertrypsinaemia seen in the majority of babies with CF. A raised IRT prompted a repeat sample taken 2–3 weeks later. The cut-off was set at the 99.5th centile such that only 1 in 200 babies required a repeat sample. Persistent hypertrypsinaemia constituted a positive screening test; followed up with clinical referral, sweat testing and mutation analysis when this became available. This was changed to an IRT-DNA-IRT protocol in May 2004; samples with a raised IRT (>99.5th centile) were tested for p.Phe508del (ΔF508) and those with one or two mutations detected underwent a 29 mutation panel (ELUCIGENE CF29, Tepnel Molecular Diagnostics, Abingdon, UK). Those with two mutations went on to the second panel to confirm the result and to differentiate p.Phe508del and p.Ile507del. The initial panel was changed to a four mutation panel (p.Phe508del, p.Gly542X, p.Gly551Asp, 621+1G>T, ELUCIGENE CF4) in January 2009, in line with the UK national protocol,4 again followed by the 29 mutation panel. Repeat samples for IRT were obtained at day 21–28 where mutation analysis was negative and the initial IRT was above a specified cut-off (100 μg/l up to December 2008, then the 99.9th centile (120 μg/l) thereafter) or the DNA results were inconclusive with only one mutation detected. The IRT-DNA-IRT protocol requires a very small number of second samples (less than 0.05% of babies screened).
Data on false positive and false negative results have been collated. All false negatives are included irrespective of their age of presentation as the UK screening programme does not define an upper age limit for false negative results. The false negatives have been divided into pancreatic sufficient and insufficient based on clinical assessment, need for enzyme supplementation and, in later years, faecal elastase results. False positives have been defined as babies reported as ‘CF suspected’ but subsequently not diagnosed clinically.
Incidence, sensitivity and positive predictive value
The number of cases detected, incidence, sensitivity and positive predictive values (PPV) for the 30 year period together with data for IRT-IRT (24 years) and IRT-DNA-IRT (6 years) are given in table 2. Thirteen screen positive babies were siblings of known CF cases and would have been investigated accordingly. The ‘unexpected, early onset’ sensitivities refer to the detection of unexpected cases presenting before the age of 3 years (ie, late-presenting cases, siblings, meconium ileus (MI) and a baby with echogenic bowel have been excluded). The PPVs with and without the inclusion of sick neonates (CF unrelated) are shown for comparison. The 95% CIs were calculated according to the efficient-score method, described by Newcombe.5 The sensitivity of the IRT-DNA-IRT protocol appears lower than IRT-IRT (despite no obvious change in incidence) however this observation is not statistically significant (Fisher Exact Probability Test, p=0.23). It is possible that increasing clinical awareness of atypical mild CF has contributed to the false negative rate in recent years.
Table 1 shows the total number of babies screened using the different protocols and the number of false positive results encountered. Using the IRT-DNA-IRT protocol the number of unexplained false positives in apparently well, term babies was 1 per 147 764 compared with 1 in 13 557 for the IRT-IRT protocol. The number of false positives in sick neonates also decreased, from 1 in 10 999 to 1 in 18 470. The two stage IRT-IRT assay resulted in a total of 96 false positives of which 43 occurred in well, term babies with a variety of disorders in the remaining 53 (listed in table 3). The nine false positives generated by the IRT-DNA-IRT protocol were all negative for the p.Phe508.del mutation and detected on the basis of prolonged hypertrypsinaemia. Only one of the nine false positive babies detected by the IRT-DNA-IRT protocol was a well, term baby as shown in table 3. The false positive rate for the 4 years pre and post the change to earlier sample collection (day 5–8) remained stable at 0.02% (IRT-IRT protocol).
True positives on IRT-IRT protocol clinically misidentified as probable carriers
Five babies born between 1989 and 2000 had persistent hypertrypsinaemia and were investigated clinically with sweat tests and p.Phe508del mutation analysis. All had one copy of the common mutation but, based on the reference ranges in use at the time, had normal or equivocal sweat test results (shown in table 4) hence were thought to be CF carriers. Cases 1–4 later presented at ages ranging from 5 weeks to 2 years with gastrointestinal and respiratory problems whereas the fifth child was diagnosed at the age of 14 years having had nasal polyps for many years. Further DNA analysis at clinical presentation revealed a second mutation in all cases. Their results illustrate the now well-described phenomenon of normal or equivocal sweat test results, which may change with time, with certain genotypes.6 ,7 It is now accepted that a reduced upper limit of 29 mmol/l for sweat chloride should be used for the screen positive neonate,8 but even with this lower cut-off one child would have had a normal chloride result in the first few months of life. Had the IRT-DNA-IRT protocol been in use at the time the four cases with compound heterozygosity for p.Phe508del/p.Arg117His would have been identified at the mutation stage, therefore misclassification is less of an issue with the current protocol. This very small group illustrates that in some infants neonatal hypertrypsinaemia may be a more effective indicator of CF than the results of sweat testing at an early age.
Twenty-nine CF cases were not detected by screening; of these 10 had MI, accounting for 19% of the 53 MI cases documented. The remaining 19 missed cases are shown in table 5. Cases 2, 8 and 11 screened using the IRT-IRT protocol all had raised initial IRTs. Case 2 appears to have been missed on the second IRT however the diagnosis of CF in this baby is not secure as it was based on post mortem findings alone with no confirmatory sweat test or genetic studies. The specificity of pathological findings for the diagnosis of CF has since been questioned, particularly in severely ill neonates.9 Case 8 would have been detected by the current IRT-DNA-IRT protocol but this is not clear for case 11, as the patient's genotype is not known. Case 16, screened using the IRT-DNA-IRT protocol, was also missed on the second IRT giving a combination of results consistent with unaffected carrier status. This result was supported by a normal sweat test but the child was subsequently identified as a compound heterozygote for p.Phe508del/p.Pro750Leu.
Fifteen out of 19 non-MI CF cases had an initial IRT below the cut-off. In case 5 there was a technical problem with elution of a poor quality (clotted) blood spot with a repeat sample (collected as there was a clinical suspicion of CF) giving a raised result. This underscores the importance of paying meticulous attention to the quality of the dried blood specimen.10
The exact numerical value for IRT has not been documented for three of the false negative results. In five cases the IRT concentration was close to the action limit of the 99.5th centile whereas in four cases the initial IRT was less than the 96th centile for the neonatal population. Lowering the cut-off to the 99th centile would have detected at least three extra cases but this would have caused a potential increase in the false positive rate, an increase in the retest rate (IRT-IRT) and increased identification of unaffected carriers (IRT-DNA-IRT).
Ascertainment of CF cases in this region is believed to be close to 100% as the clinical teams are highly supportive of the programme and notify the laboratory of any newly diagnosed case. However this data may not be complete as families may have moved out of the area and there may be false negatives in the population that are not yet recognised.
True negatives clinically misidentified as false negatives
One child labelled as a false negative (IRT less than 25 μg/l) with rectal prolapse and two sweat chloride results in the equivocal range (45 and 47 mmol/l) was later thought not to have CF based on clinical findings and failure to identify any mutations in the CFTR gene despite extensive sequencing. Twins presenting at 8 years of age with transient steatorrhoea and respiratory problems were initially thought to have CF following an abnormal sweat test in one child. After thorough investigation including mutation analysis, multiple sweat chlorides, nasal potential difference and pancreatic function tests which were all normal, the boys were deemed not to have CF. At age 10 years the symptoms had entirely resolved with the exception of mild asthma.
This programme was one of the earliest to be established worldwide and may have identified the first infant with CF through screening. The first two CF infants detected by the programme were born in June and July 1980, pre-dating the infant reported by the New Zealand programme by some months.11
Overall the screening programme has proved effective in detecting 95% of unexpected, early presenting CF cases, in accordance with the experience of other established programmes.3 ,12 ,13 The new ‘national’ protocol with four mutations in the initial panel is likely to be slightly more sensitive than the 2004 IRT-DNA-IRT protocol based on a single mutation. However, the majority (79%) of undetected CF cases were missed at the initial IRT stage and would not have progressed to the second stage of either protocol.
The occurrence of false negative results in CF infants presenting with MI was reported by several authors in the 1980s and in our experience seems to be more prevalent among those whose enteral feeding is most severely curtailed.10 However, not all screening programmes seem to have encountered such problems.14 In any case, false negatives in babies presenting with MI are not generally of concern, since these cases should always be thoroughly investigated for CF. Of the remaining 19 false negatives 15 (including all pancreatic insufficient cases) were diagnosed within the first year of life, suggesting that the screening result did not forestall appropriate clinical investigation.
As expected the increased specificity of the DNA step has reduced the impact of ‘non-CF’ increases in IRT both in sick and well, term babies. The pathophysiology responsible for the hypertrypsinaemia observed in some ‘non-CF’ sick babies is likely to be complex. However, it is reasonable to suppose that hypertrypsinaemia will occur when dysmorphogenic or inflammatory processes affect the pancreatic exocrine system, when renal clearance of circulating IRT is impaired or when disturbances in cellular metabolism affect the normal transport of electrolytes and water in ductular epithelia, thus mimicking the pathology resulting from mutations in the CFTR gene. Some causes are readily identified including chromosomal abnormalities,15 ,16 renal insufficiency,17 and galactosaemia;18 however several factors may be contributing in sick neonates.19 In the overall setting of neonatal population screening, an earlier test is likely to have a negative impact on specificity. However, our data does not demonstrate a difference in the false positive rate pre and post the change to testing at day 5–8 (IRT—IRT protocol). The kits utilised at various stages may have affected the results, as there are various forms of IRT in blood that may react differently in the assays; however their relative merits are the subject of continuing debate.14 ,20 Between 1980 and 1990, as other centres adopted this screening strategy, poor interlaboratory performance in a quality assurance scheme highlighted problems related to the adaptation of the first generation of IRT radioimmunoassays for use with dried blood spots.21 The issues involved have been extensively reviewed.10 The changing IRT methodologies reported here reflect the subsequent availability of commercial assays more suitable for dried blood spots and more amenable to laboratory automation.
Undoubtedly the major advantage of second tier mutation analysis is the resulting improvement in the PPV of the screening test, as excluding CF in a screen (false) positive baby can require prolonged follow-up with ensuing anxiety for the family. This is coupled with an equally important reduction (10-fold) in the requirement to obtain a second blood specimen to complete the screening process.
Our results highlight the importance of collecting outcome data over a long time period in order to accurately assess the screening programme. Some of the categories initially assigned on clinical evaluation in the newborn period changed with evolving clinical phenotype. The initial diagnosis reflected the extent of mutation analysis, the weight given to a normal sweat test result together with the reference ranges in use and, of course, the clinical picture in early life. The five babies mistakenly thought to be unaffected carriers and the three children initially reported as false negatives illustrate the need for long-term follow-up and update of screening outcomes. The problems associated with identifying atypical and/or mild cases of CF on screening are well recognised such that there is a European consensus for the evaluation and management of infants with an equivocal diagnosis following newborn screening.22
Overall the findings show that newborn screening for CF is a valid undertaking that detects the majority of unsuspected CF cases presenting early but clinicians need to bear in mind that approximately 5% may be missed and a negative CF screen should not preclude further investigation if suggestive symptoms are present. Over the 30 year period the performance of the East Anglian screening programme has not only been sustained but has also improved.
The authors wish to thank the CF Research Trust and the MRC for providing funding in the 1980s. We are also grateful to the Molecular Genetics Department, Cambridge University Hospitals NHS Foundation Trust and to all the CF clinical teams in Norfolk, Suffolk and Cambridgeshire for their continuous and much appreciated support of the screening programme.
Contributors All authors contributed to the collection and analysis of the screening outcome data and to the writing of the report.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.