Abstract
Importance Genetic disorders, including congenital anomalies, are a leading cause of morbidity and mortality in infants, especially in neonatal and pediatric intensive care units (NICU and PICU). While genomic sequencing is useful for diagnosis of genetic diseases, results are usually reported too late to guide inpatient management.
Objective To test the hypothesis that rapid whole genome sequencing (rWGS) increases the proportion of infants in NICUs and PICUs receiving a genetic diagnosis within 28 days.
Design An investigator-initiated, partially blinded, pragmatic, randomized controlled study with enrollment from October 2014 - June 2016, and follow up until December 2016.
Setting A regional neonatal and pediatric intensive care unit in a tertiary referral childrens hospital.
Participants Sixty five of 129 screened families with infants aged less than four months, in neonatal and pediatric intensive care units, and with illnesses of unknown etiology, completed the study.
Intervention Parent and infant trio rWGS.
Main Outcome and Measure The hypothesis and end-points were formulated a priori. The primary end-point was rate of genetic diagnosis within 28 days of enrollment or first standard test order.
Results Twenty six female proband infants, 37 male infants, and two infants of undetermined sex were randomized to receive rWGS plus standard tests (n=32, cases) or standard tests alone (n=33, controls). The study was terminated early due to loss of equipoise: 63% (21) controls received genomic sequencing as standard tests. Nevertheless, intention to treat analysis showed the rate of genetic diagnosis within 28 days to be higher in cases (31%, ten of 32) than controls (3%, one of 33; difference, 28% [95% CI, 10% to 46%]; p=0.003). Among infants enrolled in the first 25 days of life, the rate of neonatal diagnosis was higher in cases (32%, seven of 22) than controls (0%, zero of 23; difference, 32% [95% CI, 11% to 53%]; p=0.004). Age at diagnosis (median in cases 25 days, range 14-90 days vs median in controls 130 days, range 37-451) and time to diagnosis (median in cases thirteen days, range 1-84 days vs median in controls 107 days, range 21-429 days) were significantly less in cases than controls (p=0.04).
CONCLUSIONS rWGS increased the proportion of infants in a regional NICU and PICU who received a timely diagnosis of a genetic disease. Additional, adequately powered studies are needed to determine whether accelerated diagnosis is associated with improved outcomes in this setting. ClinicalTrials.gov Identifier: NCT02225522.
INTRODUCTION
A premise of pediatric precision medicine is that outcomes are improved by replacement of clinical diagnosis and empiric management with genetic diagnosis and genotype-differentiated treatment1-9. The evidence base for pediatric precision medicine is still underdeveloped10,11. Ill infants are especially in need of precision medicine since genetic diseases are a leading cause of mortality, particularly in neonatal intensive care units (NICU) and pediatric intensive care units (PICU)5-7,12-16. Amongst high-cost health care, NICU treatment is one of the most cost-effective17-19. Since disease progression can be very rapid in infants, genetic diagnoses must be made quickly to permit consideration of precision interventions in time to decrease morbidity and mortality5,6,20-23. For a few genetic diseases, newborn screening has shown early neonatal diagnosis and rapid, precise intervention to dramatically improve outcomes24,25. The potential expansion to newborn diagnosis for symptomatic infants for all 5000 genetic diseases26 has been made technically possible by the advent clinical genomic sequencing (whole genome sequencing (WGS) or whole exome sequencing (WES), and next-generation sequencing gene panel tests (NGS)). In particular, rapid WGS (rWGS) can allow genetic diagnosis in two days20,27.
There is substantial evidence that a higher proportion of symptomatic children with likely genetic disease receive etiologic diagnoses by WGS and WES than other genetic tests3-7,6,28-35. Published NICU or PICU experience with rWGS, however, is limited to case reports and one retrospective study5,6,20-23. In the latter, 57% of infants received genetic diagnoses in a median of 23 days (day of life 49)6. However, it has not yet been unequivocally demonstrated whether rWGS improves timeliness of genetic diagnosis relative to standard genetic tests. Here we report results of Newborn Sequencing In Genomic medicine and public HealTh Randomized Controlled Trial (RCT) 1 (NSIGHT1), the first RCT of genomic testing in patients24. Specifically, NSIGHT1 compared rates of genetic diagnosis in NICU and PICU infants with possible genetic diseases at 28 days from enrollment by standard tests alone vs standard tests plus trio rWGS.
METHODS
Full details of methods are reported in Supplementary Material.
Trial Design
NSIGHT1 tested the a priori hypothesis that rWGS increases the proportion of infants receiving a genetic diagnosis within 28 days in a partially blinded, randomized controlled study in a regional NICU and PICU in a tertiary referral children’s hospital (Children’s Mercy – Kansas City). Enrollment was from October 2014 - June 2016, and follow up until November 2016. Inclusion criteria were infants in the NICU or PICU of age less than four months with illnesses of unknown etiology and one of the following: 1. A genetic test order or genetic consult; 2. A major structural congenital anomaly or at least three minor anomalies; 3. A laboratory test suggested a genetic disease; or 4. An abnormal response to therapy. Exclusion criteria were an existing genetic diagnosis, or features pathognomonic for a chromosomal aberration. The NICU census was reviewed daily for eligible infants by enrollment coordinators. NICU clinicians were notified of eligible infants, who were nominated through a standard form. NICU and PICU clinicians notified families of eligible infants about the study, and enrollment coordinators then approached parents for informed consent. Enrolled infants were randomly assigned to receive standard clinical tests (controls) or standard clinical tests plus trio (infants and parents where available) rWGS (cases; Figure S1). Parents and clinicians were blinded until by day ten, when they were notified of randomization assignment, to allow consideration of crossover to rWGS.
Rapid Genome Sequencing
rWGS was performed using previously described methods that yielded variant calls within two – seven days of blood draw5,6,20,27. Variants were interpreted by board certified molecular geneticists using American College of Medical Genetics guidelines for pathogenic and likely pathogenic classifications36. Genotypes were confirmed clinically by Sanger sequencing. Secondary and incidental findings were not reported.
Standard Genetic Testing
Standard clinical testing for genetic diseases was performed based on clinician judgment, assisted by subspecialist recommendations. The set of genetic tests considered to be standard was developed by two molecular genetics laboratory directors (Supplementary Methods).
Trial End Points
The primary end point was the diagnostic sensitivity within 28 days of enrollment or first standard test order. Secondary end points were the diagnostic sensitivity by day of life (DOL) 28, total diagnostic sensitivity, time-to-diagnosis, rate of clinical utility (proportion of patients with a change in management related to test results), length of hospitalization, and 6 month mortality rate. Clinical utility was determined by clinician surveys and reviews of the electronic health record by at least two pediatric subspecialist experts in genomic medicine to identify changes in treatments, procedures, consultations, testing, genetic or reproductive counseling, and recommendations for specific follow up related to the diagnosis37. A modified Delphi method was used to determine inclusion of change in management where there was disagreement.
Statistical Analysis
Statistical analyses were based on the intention-to-treat principle. Fisher’s exact test was used to compare 28-day diagnostic rates, total diagnoses, clinical utility of diagnoses, and diagnoses before discharge. A two-sample t-test was performed to compare age at hospital discharge. Kaplan-Meier analyses were used to compare time to diagnosis, which was measured from the date of first standard test order for controls or date of enrollment for cases, and age at diagnosis. Age at death was compared with the log-rank test38. When there was evidence of a non-constant hazard ratio, between-group differences were evaluated with the Peto-Peto test39,40.
RESULTS
Patients
65 (50%) of 129 nominated infants completed the study (Figure 1). 32 infants randomized to rWGS plus standard genetic tests (cases) and 33 to standard tests alone (controls, Figures 1, S1). Phenotypes were highly diverse and typically present at birth (Tables 1, S1). Fewer control infants had cardiovascular findings (6% vs 28%; difference, −22% [95% CI, −40% to −4%]; p=0.02) than cases, which may have affected likelihood for genetic disease (Table 1).
Standard Diagnostic Tests
The proportion of infants receiving standard genetic tests and age at first standard test order were similar in both arms (Table 1). Infants received an average of 3.1 (range 0-10) standard genetic tests (Table 1, S3). 21 (64%) of 33 control infants received non-expedited NGS, WES or WGS standard tests, compared with fourteen (44%) of 32 cases (Table 1, S3). The average age at first standard test order was 14 days (range 0–120 days). Standard tests yielded fifteen (43%) genetic diagnoses in the 35 subjects tested, seven (50%) in 14 cases, and eight (38%) in 21 controls (Table 2, S4). Of note, five (8%) diagnoses by standard tests were not detected by rWGS at the time of study: Four (6%) were copy number or structural variants and one (2%) was a change in DNA methylation. The median time from first standard test order to diagnosis was 64 days (range 16-450 days). The average age at diagnosis by standard genetic tests was 113 days (range 16– 451 days). Six (9.4%) of 64 infants received a diagnosis by standard tests prior to hospital discharge (Table S5).
Rapid Whole Genome Sequencing
Ten of 32 cases (31%) received diagnoses by rWGS (Table 2, Table S4). Including five crossovers, 12 (32%) of 37 infants received rWGS diagnoses (Table 2, S5). On average, enrollment occurred on DOL 22 (range one −101; Table 1), an average of eight days later than standard tests. The median time to rWGS diagnosis, including clinical confirmatory testing, was fourteen days (range eight – 35 days; Table S5). The median age at WGS diagnosis in patients randomized to rWGS was 28.5 days (range 14 – 90 days). Among crossovers, the median age at WGS diagnosis was 94.5 days.
Diagnoses
Twenty-two genetic diagnoses were reported in 21 (32%) of 65 infants (Table 2). The most common mechanism was de novo variant occurrence (eleven of eighteen (61%) diagnoses; Table 2). The most common inheritance pattern was autosomal dominant (thirteen of eighteen (72%) diagnoses). Cross-over to rWGS was requested for seven (21%) of the 33 controls. Five were granted, yielding two diagnoses. In both, diagnosis by rWGS occurred first but was recapitulated by standard tests (Table 2). Twenty (31%) of the 65 infants (91% of those with a diagnoses) had attendant changes in management (Table S4).
Early study termination
The study was terminated after 21 months due to growing availability of NGS panels, WES and WGS as standard tests, which shifted the baseline of comparison over the course of the study. These were associated with high rates of cross-over requests and higher utilization of NGS panel, WES or WGS standard genetic tests among controls (64% including cross overs) than cases (44%; Table S3).
End-Point Testing
End-points were analyzed on the basis of intention to treat (Figures 1, S1). The primary end point, rate of genetic diagnosis within 28 days of enrollment, was higher in cases (31%, ten of 32) than controls (3%, one of 33; difference, 28% [95% CI, 10% to 46%]; p=0.003 Table 3, Figure 2). For neonates enrolled within the first 25 days of life, the rate of diagnosis by DOL 28 was higher in cases (32%, seven of 22) than controls (0%, zero of 33; difference, 32% [95% CI, 11% to 53%]; p<0.01; Table 3). Age at diagnosis and time to diagnosis differed significantly between arms, after accounting for non-proportional rates of diagnosis (Table 4, Table S5): The median age at diagnosis in cases was 25 days (range 14-90 days) vs median in controls was 130 days (range 37-451). The median time to diagnosis in cases was 13 days (range 1-84 days) vs median in controls 107 days (range 21-429 days).
Five secondary end-points did not differ significantly between arms (Table 3, 4, S4). They were the proportion of infants in whom diagnoses had clinical utility (41% of cases vs 21% of controls; difference, 19% [95% CI, −3% to 42%]), proportion of infants with a change in medical management (clinical utility, 22% of cases vs 9% of controls; difference, 13% [95% CI, −5% to 30%]), proportion of patients who received diagnoses prior to hospital discharge (28% of cases vs 9% of controls; difference, 19% [95% CI, 0% to 38%]), average length of NICU/PICU stay, 6-month mortality, and age at death.
Discussion
NICU and PICU infants receiving trio rWGS plus standard clinical testing had a higher rate of genetic diagnosis and shorter time to diagnosis than infants receiving standard tests alone. In intention to treat analysis, rWGS was associated with significantly more genetic diagnoses within 28 days of enrollment (31%, 10 of 32) than standard tests alone (3%, 1 of 33; difference, 28% [95% CI, 10% to 46%]; p=0.003). The rate of neonatal (DOL 28) diagnosis was higher in cases (32%, 7 of 22) than controls (0%, 0 of 23; difference, 32% [95% CI, 11% to 53%]; p=0.004). Of note, standard genetic testing was ordered an average of 8 days before enrollment, which benefitted the control arm over rWGS cases for these analyses. Nevertheless, age at diagnosis and time to diagnosis were significantly shorter in rWGS cases, after accounting for non-proportional rates of diagnosis.
The rate of genetic diagnosis by rWGS in a NICU or PICU was reported previously in one cohort6. Enrollment in that study was at average DOL 26 (vs DOL 22 herein). The rate of diagnosis by rWGS therein was 14% (5 of 35) by DOL 28, and 34% (12 of 35) within 28 days of enrollment, which were similar to herein (32% and 31%, respectively). The total rate of genetic diagnosis by rWGS herein (32%) was within the range reported for WGS and WES studies3-7,6,28-35.
Timely return of rWGS diagnoses was limited by two research factors that may not be part of routine clinical practice: firstly, confirmatory testing by “the clinically accepted standard” was required for research rWGS diagnoses – but is not necessarily required for laboratory developed NGS tests – which lengthened the time to rWGS diagnosis by 7 – 10 days. Indeed, all diagnostic rWGS findings in the current study were concordant with orthologous methods. For well covered, pathogenic and likely pathogenic, single nucleotide variants in regions of high WGS quality, a median time-to-result of five days is anticipated6,20,27.
Secondly, enrollment occurred relatively late during the NICU or PICU stay (DOL 22). While parents are interested in receipt of genomic sequencing at birth, an enrollment rate of 6% was reported for WES in NICU infants in another cohort41,42. Delay in enrollment herein reflected two logistical factors. First, since a criterion for enrollment was suspicion by the provider of an underlying genetic disease, nomination was often delayed until a genetic test or consult had been ordered. In such cases, the time of enrollment delayed the study test, rWGS, compared to standard testing; nevertheless, there was still a decreased time to diagnosis with rWGS. Secondly, NSIGHT1 required informed consent from both parents; the logistics and complexity of obtaining informed consent in a NICU or PICU setting are arduous. In future studies, it will be important to seek enrollment close to day of admission. This would be facilitated by simpler enrollment criteria, requirement of informed consent from a single parent, and limiting eligibility for enrollment to within several days of admission.
Clinical WGS continues to improve with respect to rate of genetic diagnosis and time to diagnosis27. In particular, the diagnostic rate is increasing through ongoing identification of novel disease genes, improved reference genome sequences, and better identification of disease-causing copy number, repeat expansion, regulatory, splicing and structural variations32,43-50. These recent advances were not reflected in the current study. WES and WGS have similar analytic performance for exonic and splicing variants, which comprised seventeen of twenty two diagnoses. However, four diagnoses were associated with copy number or structural variants, for which WGS has superior analytic performance to WES. rWGS is methodologically simpler than WES, and thus two days faster than possible with rapid WES.
NSIGHT1 was terminated early, primarily due to loss of equipoise noted by some nominating clinicians during the study. Some practitioners grew to regard randomization to standard tests alone to be an inferior intervention than standard tests plus trio rWGS. This was associated with seven (21% of controls) requests to cross-over control infants to the rWGS arm following clinician un-blinding, five of which were granted. It was also associated with a higher rate of order of NGS panel, WES or WGS standard genetic tests in controls (64%) than cases (44%). Standard genomic sequencing tests accounted for 63% (5) of the 8 genetic diagnoses in controls. As a result, there was not a significant difference between arms in the total number of genetic diagnoses, a secondary end-point (41% [13] diagnoses among 32 infants in the rWGS arm, 24% [8] of 33 in controls; difference, 16% [95% CI, −6% to 39%]; p=0.19). Future pragmatic RCT designs in genomic medicine will require careful attention to the principle of equipoise and to the rapid evolution of clinical NGS-based testing51-52. The more widespread use of gene panel testing in the NICU during the course of this study was a significant departure from our experience at study conception. Our study was not intended to evaluate the relative diagnostic yield of panel testing over rWGS. Consequently, the study was not powered to evaluate the non-inferiority of panels over rWGS.
The rationale for rWGS in NICU infants is to enable consideration of acute precision interventions in time to decrease morbidity and mortality5,6,21-24. In two prior studies of genomic sequencing in infants, genetic diagnoses led to precision medicine that was considered life-saving in 5%, and that avoided major morbidity in 6% 6,7. In those studies, early diagnosis (DOL 49) led to greater implementation of precision medicine (65%) than later diagnosis (DOL 374, 39%), particularly with regard to palliative care guidance. As in the current study, assessments of clinical utility were based on actual changes in management, which were limited by clinician experience with genomic medicine and rare genetic diseases. This is a major challenge for NICU and PICU implementation of genomic medicine for rare genetic diseases53. Unfortunately, early termination of the current study resulted in loss in power for the secondary endpoints: There were not significant differences in the overall rate of clinical utility of diagnoses, length of admission, rate of diagnosis before discharge, mortality and age at death. The clinical utility of diagnoses and rate of diagnosis before hospital discharge trended towards being higher in the rWGS arm (difference, 19% [95% CI, −3% to 42%], p=0.11, and 19% [95% CI, 0% to 38%], p=0.06, respectively). Additional studies are clarify whether shorter time to diagnosis is associated with changes in clinical utility of diagnoses, outcomes, or healthcare utilization.
Conclusions
Among infants with suspected genetic diseases in a regional NICU or PICU, the addition of rWGS decreased the time to diagnosis. We suggest that rWGS should be considered as a first-tier genetic test in NICU and PICU infants with suspected genetic diseases7. Since genetic diseases are among the leading cause of death in the NICU and PICU, as well as overall infant mortality, implementation of rWGS is likely to have broad implications for the practice of neonatalology.
Article Information
Author contributions
Concept and design: SFK, LKW, JEP.
Acquisition, analysis or interpretation of data: All authors.
Drafting of manuscript: SFK, JEP, MMC, NS, JAC.
Critical Revision of the Manuscript for Important Intellectual Content: DPD, PMA, SLC, MAC, RTF, JAH, HK, RJM, JLR, SLT.
Statistical analysis: Clark.
Obtained Funding: Kingsmore, Willig, Leeder.
Administrative, Technical or Material Support: SFK, JAC.
Study Supervision: JEP, SFK, JAC.
Competing interests
None.
Funding/Support
Grant U19HD077693 from NICHD and NHGRI.
Role of the Funder/Sponsor
NICHD and NHGRI programmatic staff assisted investigators during performance of the research activities. Funders had no role in the collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.
Additional Contributions
We thank Drs. John Lantos, Sarah Soden and Howard Kilbride for advice and assistance. A Deo lumen, ab amicis auxilium.
Data and material availability
Data are available at LPDR (https://www.nbstrn.org/research-tools/longitudinal-pediatric-data-resource).
Footnotes
KEY POINTS
QUESTION: What proportion of acutely ill inpatient infants receive a diagnosis of a genetic disease within 28 days with rapid whole genome sequencing?
FINDINGS: In a randomized controlled trial of 65 infants, the diagnostic sensitivity of rapid whole genome sequencing within 28 days was 31% vs 3% with standard genetic testing, a significant difference.
MEANING: In NICU and PICU infants with diseases of unknown etiology, rapid whole genome sequencing may be warranted as a first-line diagnostic test.