Article Text

Download PDFPDF

Original research
Genetic aetiology of early infant deaths in a neonatal intensive care unit
  1. Lin Yang1,2,
  2. Xu Liu1,3,
  3. Zixiu Li2,
  4. Peng Zhang3,
  5. Bingbing Wu2,
  6. Huijun Wang2,
  7. Liyuan Hu3,
  8. Guoqiang Cheng3,
  9. Laishuan Wang3,
  10. Wenhao Zhou1,2,3,4
  1. 1 Clinical Genetic Center, Children's Hospital of Fudan University, Shanghai, China
  2. 2 Key Laboratory of Birth Defects, Children’s Hospital of Fudan University, Shanghai, China
  3. 3 Department of Neonatology, Children's Hospital of Fudan University, Shanghai, China
  4. 4 Key Laboratory of Neonatal Diseases, Children's Hospital of Fudan University, Shanghai, China
  1. Correspondence to Professor Wenhao Zhou, Department of Neonatology, Children's Hospital of Fudan University, Shanghai 201102, China; zhouwenhao{at}


Background Congenital anomalies are the leading cause of early neonatal death in neonatal intensive care units (NICUs), but the genetic causes are unclear. This study aims to investigate the genetic causes of infant deaths in a NICU in China.

Methods Newborns who died in the hospital or died within 1 week of discharge were enrolled from Children’s Hospital of Fudan University between January 1, 2015 and December 31, 2017. Whole exome sequencing was performed in all patients after death.

Results There were 223 deceased newborns with a median age at death of 13 days. In total, 44 (19.7%) infants were identified with a genetic finding, including 40 with single nucleotide variants (SNVs), two with CNVs and two with both SNVs and CNVs. Thirteen (31%, 13/42) patients with SNVs had medically actionable disorders based on genetic diagnosis, which included 10 genes. Multiple congenital malformation was identified as the leading genetic cause of death in NICUs with 13 newborns identified with variants in genes related to multiple congenital malformations. For newborns who died on the first day, the most common genetic cause of death was major heart defects, while metabolic disorders and respiratory failure were more common for newborns who died in the first 2 weeks.

Conclusion Our study shows genetic findings among early infant deaths in NICUs and provides critical genetic information for precise genetic counselling for the families. Effective therapies enable the improvement of more than a quarter of newborns with molecular diagnoses if diagnosed in time.

  • early infant death
  • neonatal intensive care units
  • whole exome sequencing
  • genetic counseling
  • precise treatment

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Congenital anomalies (ie, congenital malformations, deformations and chromosomal abnormalities) are one of the leading causes of infant death, accounting for 20% of all infant deaths.1–3 Congenital anomalies were the second main cause (21.3%) of neonatal mortality in Shanghai, while complications of preterm birth (33.6%) were the top cause.4 In addition, metabolic, immunologic and other systemic disorders are also severe causes of death in neonatal intensive care units (NICUs). Understanding the genetic causes of infant deaths in NICUs is important and helps improve potentially preventable therapy and genetic counselling for the family. It may also be helpful for research on developing management guidelines that optimise outcomes. However, little is known about the genetic causes of death in NICUs, except for chromosome abnormalities.

The reason for the unclear understanding of genetic causes in early infant deaths that occur in NICUs is that the ascertainment rates are low. On the one hand, the full clinical phenotype is not manifested.5 On the other hand, genetic heterogeneity in the death cohort is immense.6 7 Both of these lead to difficulties in identifying the genetic causes of newborn deaths. Whole exome sequencing (WES) is advantageous because it requires no prior assumption about the disease to be diagnosed.8 With the continuing reduction of sequencing cost, WES is now becoming increasingly widely used in clinical diagnosis.9 10 Therefore, WES is an effective way to explore the genetic causes of infant deaths occurring in NICUs.

In this study, we used WES data to investigate single nucleotide variants (SNVs) and CNVs simultaneously in a 3-year death cohort, from the NICU of Children’s Hospital of Fudan University. We aimed to provide an overview of the genetic causes of death in the NICU and the proportion of diseases with interventions.


NICU death cohort

All patients in NICU were transferred to the Children’s Hospital of Fudan University by our transfer team from other hospitals including Obstetrics & Gynaecology Hospitals and Children’s Hospitals. Patients were recruited from the Children’s Hospital of Fudan University with the following inclusion criteria: (1) newborns with an age at admission of <28 days after birth from the NICU between January 1, 2015 and December 31, 2017; and (2) newborns that died while in the hospital or died within 1 week of discharge. Patients were excluded if they had chromosome abnormalities identified using karyotype. Genetic testing for all patients was performed after death for accurate diagnosis.

Clinical manifestations and laboratory results of each proband were ascertained comprehensively by a physician and through the review of medical records. All samples used in this study were collected with appropriate informed consent.

Whole exome sequencing

Genomic DNA fragments of patients were enriched for WES using the Agilent (Santa Clara, California, USA) SureSelectXT Human All Exon 50 Mb kit. DNA libraries were sequenced on the HiSeq2000/2500 sequencer according to the manufacturer’s instructions (Illumina, San Diego, California, USA). Clean reads were aligned to the reference human genome (UCSC hg19) by Burrows-Wheeler Aligner (v.0.5.9-r16). Subsequent processing was performed by using SAMtools and Picard ( The variant calling process followed the GATK best practice (V.3.2).

Annotation of SNVs

SNVs were annotated by ANNOVAR11 and VEP12 software and compared computationally with the list of reported variants from the Human Gene Mutation Database (HGMD, professional version). Synonymous variants, intronic variants that were more than 15 bp from exon boundaries (which were unlikely to affect mRNA splicing) and common variants (minor allele frequency,>1%), were discarded. Missense variants were assessed with SIFT,13 PolyPhen-214 and MutationTaster.15

Pathogenicity of SNVs was defined based on American College of Medical Genetics and Genomics criteria.16 Specifically, a case was classified as molecularly diagnosed when pathogenic or likely pathogenic variant(s) (P/LP) were detected in a disease gene that sufficiently explained the phenotypes of the studied individual; additionally, the zygosity of the mutant allele(s) was compatible with the disease gene (when parental data were available, family cosegregation analysis was required). The variants with unknown significance (VUSs) were variants in gene(s) associated with one or more clinical phenotypes of the patients, and absent or present in GnomAD/ExAC databases with low frequencies. All the selected variants were then validated by Sanger sequencing.

Detection of CNVs based on WES data

CNVs were analysed with the use of a home-modified CANOES17 using a read depth calculated by BEDtools (V.2.17). Annotation at the gene level was based on in-house databases and four public databases: OMIM, HGMD, Swiss-Prot and RefSeq. Region-level annotation was performed by the Database of Genomic Variants, and Database of Genomic Variation and Phenotype in Humans using Ensembl Resources. CNVs that passed computational phenotype filtration, overlapped with the reported pathogenic region, encompassed disease genes and had a variant range larger than 1 Mb were considered candidate CNVs.


Demographics and clinical characteristics

The total number of inpatients with an age at admission of <28 days after birth was 3989 between 2015 and 2017. During this period, the total number of deaths was 243 (6.1%), including in-hospital deaths and deaths within 1 week of discharge. We enrolled 223 newborns in this study (figure 1). There were 134 males (60.1%) and 89 females (39.9%), with a median birth weight of 2640 g and a median gestational age of 35 weeks. The median age at death was 13 days. The distribution of gestational age, age at admission and age at death in all deceased infants is shown in figure 2A and B. A total of 136 (60.9%) newborns were hospitalised on the first day after birth. Most of the 223 patients who died in the NICU were born at 37–39 weeks and died in the first 2 days after birth. The detailed demographic and basic clinical characteristics of the death cohort are shown in table 1.

Figure 1

Flow diagram of genetic testing for infant deaths in the NICU in the study. exome sequencing was performed for all patients, and SNVs and CNVs were analysed. GA, gestational age; NICU, neonatal intensive care unit; P/LP, pathogenic or likely pathogenic; SNVs, single nucleotide variants; VUSs, variants with unknown significance.

Figure 2

Percentage of all deaths by gestational age, age at admission, age at death, and time distribution of death in patients of different genetic groups. (A) percentage of all deaths by gestational age; four patients had no record of gestational age. (B) percentage of all deaths by age at admission and age at death. (C) the time of death for 24 infants with underlying genetic conditions in seven disease groups; treatment was rejected or stopped for 20 patients and these patients died at home.

Table 1

Demographic and clinical characteristics of study population

Whole exome sequencing

In this cohort, on average, a total of 83.22 million sequencing reads were generated, and the average on-target sequencing depth was 111.69×. At least 97.11% of the target regions were covered, and 90.3% of them were covered by 20×. These results suggested that the data were sufficient to yield high-quality data for further analysis of both SNVs and CNVs.

Overview of the spectrum of genetic disorders

We identified 42 SNVs and 4 CNVs in 44 (223, 19.7%) patients in this study. Two patients carried both SNVs and CNVs. SNVs included 28 P/LP variants in 21 patients and 28 VUSs in 21 patients. When including the four patients with chromosome abnormalities identified using karyotype, the three patients with a trisomy 21 and the one patient with a 4q32 deletion and an 8p23 duplication, there were 48 (21.1%, 48/227) patients with genetic findings. There were 56 infants born at a gestational age <32 weeks, and 6 (10.7%) of them had genetic findings (five with SNVs and one with CNVs). The other 167 patients were born at a gestational age ≥32 weeks, and 38 (22.8%) were identified with genetic variants (35 with SNVs, 1 with CNVs, and two with both SNVs and CNVs). Differences in diagnosis rates between the two gestation groups were marginally significant (p=0.0775). The genetic diagnosis rate was higher in patients at ≥32 weeks gestation than in patients at <32 weeks gestation. Preterm, especially earlier gestational ages, mortality was most commonly caused by extreme prematurity and complications of premature birth. With increasing gestational age, the genetic aetiology of mortality increased. Among the 221 families included in this study, 131 enrolled patients were first-born children, 17.6% of which were identified with genetic variants, and 73 were the second child, of which 24.7% were identified with genetic variants.

SNV identification and singe gene disorder categories

In total, 56 phenotype-related variants spanning 38 genes were identified in 42 patients (17.9%). Molecular diagnosis with 28 P/LP variants was made in 21 patients, and 28 VUSs in 21 patients were identified as being associated with one or more clinical features of the patients. The origin of VUSs in 18 patients could not be established without the parents’ DNA. Some of these VUSs could potentially be classified as likely pathogenic if parental samples were available and de novo occurrence (dominant disorder) or phase information (recessive disorder) is confirmed by co-segregation analysis. Four variants in three patients were classified as VUS, because these VUSs in the genes which were not completely consistent with the clinical manifestations of the newborns even though confirmed by parental samples. Of all 56 variants, 16 P/LP and 6 VUS variants were previously reported, and the other 34 variants were novel. The detailed information of all variants is shown in table 2, figure 3 and online supplementary table S1.

Supplemental material

Figure 3

Fifty-six variants were identified in 42 patients. Genes were classified into nine categories. The inner circle indicates disease categories, and the outside circle shows the identified genes. The area size indicates the patient number. (A) pathogenic or likely pathogenic variants identified in 21 patients; (B)) variants with unknown significance identified in 21 patients.

Table 2

Clinical phenotype and genetic variants in 42 patients

Diseases were classified into nine categories according to 38 phenotype-related genes that were diagnosed (figure 3). Multiple congenital malformation, metabolic/biochemical and haematological disorders were the three most common diseases identified in 13, eight and six patients, respectively. Neurological, allergic/immunological/infectious, skeletal, cardiovascular, dermatological and digestive disorders were found in less than five patients.

Consistency between clinical and genetic diagnosis in nine categories among patients with SNVs

Multiple congenital malformation was identified as the leading genetic cause of death in the NICU. However, the clinical manifestations were not typical and most patients died from withdrawal of life-sustaining treatment. The mean age at death for these patients was 14 days. Thirteen newborns (13/42, 31%) were identified with 13 variants in genes related to multiple congenital malformations including 4 P/LPs or 9 VUSs. The four patients with P/LP variants received a diagnosis of atypical or had unrecognised features of genetic disorders. P148 with oesophageal atresia and deformity of the external ear was identified with a de novo and reported pathogenic variant in the CHD7 gene18 ; P051 with hypotonia and neonatal encephalopathy was identified with a paternal and reported pathogenic nonsense variant in the MAGEL2 gene19 ; P006 with hypoplastic left heart was identified with a novel splice acceptor variant in the NOTCH1 gene; and P137 with multiple organ dysfunction syndrome was identified with a de novo and reported pathogenic variant in SHOC2 gene.20 Of these, 3 patients (75%) received diagnoses of novel genetic disorders that were not suspected or recognised by WES, except P148, which was suspected initially with CHARGE syndrome. As multiple congenital malformation was the leading genetic cause of death in the NICU, it should be considered in newborns who do not fulfil the specific clinical criteria.

Eight newborns (8/42, 19%) were identified with 16 variants in genes related to metabolic disorders, including 14 P/LPs or 2 VUSs. For these newborns, the mean age at death was 3 days. Of the seven patients with P/LP variants, 6 were consistent with their clinical phenotypes, except patient P164. P164 was born at 39 weeks and died from cardiogenic shock, metabolic acidosis on the first day of life, and gastrointestinal symptoms include vomiting. We identified novel compound heterozygous variants in the OPLAH gene (c.3303+1G>C; c.3853C>T, p.Q1285X). Larsson et al reported that recurrent vomiting and diarrhoea were observed in patients with OPLAH gene mutations.21 However, mutations in the OPLAH gene were not the main cause of death for this newborn.

For 10 patients with P/LP variants in neurologic disorder genes (three newborns), allergic/immunological/infectious disorder genes (two newborns), haematological disorder genes (two newborns), skeletal disorder genes (two newborns), and dermatological disorder genes (one newborn), the molecular diagnoses were consistent with their clinical manifestations. The information is listed in table 2 and online supplementary table S1.

The proportion of genetic disorders with specific treatment

Among the newborns with SNVs in disease causal genes, we observed that 13 of 42 (31%) patients with medically actionable disorders based on molecular diagnosis included CYBB, IL10RA, SLC25A20, CPT2, MUT, ACY1, SERPINC1, ADAMTS13, VWF and F12. The implications for treatment decisions included avoiding BCG vaccination (CYBB),22 hematopoietic stem cell transplantation (CYBB, IL10RA),23 feeding with a special formula diet (medium-chain triglyceride low-fat diet and carnitine supplementation for SLC25A20 24, and vitamin B12 with a low protein diet and carnitine for MUT 25), and specific medicine (bezafibrate for CPT2 26, biotin for ACY1 27, antithrombin therapies for SERPINC1 28, plasma infusion and immunomodulatory therapies for ADAMTS13 29, 1-desamino-8D-arginine vasopressin for VWF 30, and anticoagulant therapies for F12 31). Therefore, if genetic testing and molecular aetiology were performed and assessed for all newborns after birth, 31% of patients could have had a chance to be cured or at least relieved.

In addition to the potential effect on the clinical management of the proband, genetic counselling was provided for all 44 families who received a molecular diagnosis or had uncertain WES results. Three of the 21 families with a definitive molecular diagnosis underwent prenatal diagnosis based on the probands’ genetic findings (P076, P118 and P182). Moreover, another point worth mentioning was that medical disputes in two families had been resolved due to the molecular diagnosis.

CNV identification

Four patients were identified with CNVs (1.8%, 4/223). Deletions in chromosome 1 were identified in two patients, P136 with a 1.4 Mb deletion at 1p36 and P169 with a 2.2 Mb deletion at 1q21. Two patients (P172 and P132) carried deletions at 10q11 (3.2 Mb) and 15q11-q13 (5.3 Mb). Detailed information on all CNVs is shown in table 3.

Table 3

CNVs identified in four patients

Patients with digenic phenotype-related variants

Three patients were identified with two phenotype-related variants. One patient (P185) with coagulation disorder carried one frameshift VUS in the F11 gene and two missense VUSs in the VWF gene. Two patients were identified with both SNVs and CNVs. P169 had a 2.2 Mb deletion at 1q21, and also carried two VUS variants in the ADAMTS13 gene. The other patient, P132, had a 15q11-q13 deletion and a reported pathogenic variant in the OCA2 gene in this deleted region.

Primary causes of death for different ages among 44 patients

The median age at death of newborns with underlying genetic conditions was 8.5 days. The distribution of age at death in these patients is shown in figure 2C. Among them, 20 infants were withdrawn from life-sustaining treatment, as demanded by the parents. The rest of the patients received symptomatic supportive treatment. For these deceased newborns with underlying genetic conditions, the death age distribution was similar to that of the entire cohort.

Among the deceased newborns with definitive or potential molecular diagnoses, 6 newborns died on the first day of life, and the most common cause of death was major heart defects or cardiac failure (figure 2C). One newborn was withdrawn from life-sustaining treatment, as demanded by the parents. Of the six patients, three patients (P192, P219, P132) had severe heart defects. P192 died of heart failure, tricuspid prolapse with severe tricuspid regurgitation, respiratory failure and severe neonatal asphyxia, with an identified VUS in the SCN5A gene. Mutations in the SCN5A gene are associated with arrhythmia, cardiomyopathy and sudden infant death syndrome. The variant in the SCN5A gene has been reported to be pathogenic for long QT syndrome.32 However, this newborn lacked the typical clinical evidence, so the variant was annotated as a VUS. P219 died of a heart defect, pulmonary artery atresia, a ventricular septal defect and transposition of the aorta and carried a VUS in the CREBBP gene. Congenital heart defects have been observed in Rubinstein-Taybi syndrome caused by CREBBP gene mutations. P132 died of a heart defect and complete pulmonary vein ectopic drainage, with an identified 15q11-q13 deletion. However, severe cardiovascular defects are uncommon among features of both Rubinstein-Taybi syndrome and 15q11-q13 deletion syndrome. P164 died from cardiogenic shock as described above. One patient (P116) died from acute digestive tract perforation, and a reported missense variant in the CHD7 gene was identified.33 One patient (P153) had epidermolysis bullosa, metabolic acidosis and disseminated intravascular coagulation, and the parents rejected the treatment. Compound heterozygous P/LP variants in the ITGB4 gene were identified in this newborn.

For the first 2 weeks of life, the most common genetic causes of death were metabolic disorders and respiratory failure (figure 2C). Compound heterozygous MUT variants were identified in two patients (P168 and P182) with methylmalonic acidemia and hyperammonemia and who died from multiple organ dysfunction on the sixth and 11th day, respectively. SLC25A20 variants were identified in twin brothers (P080 and P081) with carnitine palmitoyl transferase II deficiency. They shared the same pathogenic compound heterozygous variants and died from circulatory failure, respiratory failure, hyperammonemia and arrhythmia 3 to 4 days after birth.

For the postneonatal period, the most common genetic causes of death included immune deficiency disorders and haematologic disorders (figure 2C). CYBB was identified in two male patients (P076 and P179) with sepsis, pneumonia, thrombocytopenia and anaemia. STXBP2 compound heterozygous variants were identified in twin brothers (P118 and P119) born at 35+6 weeks. They were diagnosed with haemophagocytic syndrome and died on the 70th and 88th day after birth.


The cause of death in infants and children has been reported in several studies. Primary Children’s Medical Centre reported that 34.4% (180/523) deaths (ranging from <1 year to >4 years) were associated with malformations/genetic disorders.34 Another study enrolled 35 infants younger than 4 months of age with an acute illness of suspected genetic cause in NICUs or paediatric intensive care units (PICUs), and identified 21 (60%) infants with a genetic disorder.35 Several studies concerned with death in NICUs have demonstrated the causes and clinical characteristics of death in neonates,36 end-of-life care practices and decisions,37 as well as the value of postmortem autopsy.38 However, genetic causes of death in NICUs are less frequently reported. Jacob et al 39 have investigated the causes of death in 641 NICU infants and found 5% of patients with genetic anomalies. In a retrospective study reported by Meng et al, 278 infants from NICUs, PICUs and cardiovascular ICUs with an age of <100 days were referred to clinical exome sequencing, and the genetic disorder diagnosis rate was 48.1% for 81 deceased infants.40 In this study, 6 (10.7%, 6/56) infants at <32 weeks gestation and 38 (22.8%, 38/167) infants at ≥32 weeks gestation were identified with genetic factors. The mortality in preterm newborns, especially at earlier gestational ages, was most commonly caused by extreme prematurity and the complications of premature birth. Genetic factors showed greater effect on death in NICUs while gestational age increased.

To the best of our knowledge, our study is the first to investigate the genetic spectrum and diagnosis yield in early infant deaths in a NICU using WES. The data presented herein suggest that disorders caused by genetic factors are a significant cause of mortality in NICUs, representing 19.7% (44/223) of all deaths within a 3-year period in the Children’s Hospital of Fudan University. The CNV diagnosis rate was relatively low (1.8%, 4/223) in our study, which might be influenced by the patient bias. Newborns with multiple severe malformations have less opportunity to be transferred to our hospital. Our findings show concrete genetic findings among early infant deaths in a NICU, expanding the mutation spectrum of the newborn death cohort.

In our cohort of 223 newborn deaths, 42 patients were identified with single genetic variants. Multiple congenital malformations were the leading genetic causes of deaths in the NICU, but the clinical manifestations for the patients were not typical, and most of them died from rejecting treatment. For the other eight categories of diseases, the concordance rate between genetic diagnoses and clinical phenotypes was higher (16 in 17 patients with P/LP variants). The constellation of symptoms that constitute the clinical diagnosis of a genetic disorder may not be entirely present or may be difficult to recognise in the newborn period. Our findings emphasise the importance of genetic testing in NICUs to gain evidence for early accurate diagnosis and intervention for symptoms that may appear later.

Two patients were identified with both SNVs and CNVs. P169 carried a 2.2 Mb deletion at 1q21 and two VUS variants in the ADAMTS13 gene. The main features of the 1q21.1 microdeletion are neurological problems and various physical anomalies. However, no such phenotypes were present in this patient except liver failure since the patient died from diffuse intravascular coagulation, pulmonary haemorrhage, thrombocytopenia, metabolic acidosis and hyperammonemia on the 18th day after birth. This patient also carried two missense variants (NM_139025: c.10C>T, p.R4C, NM_139025: c.3619G>A, p.G1207S) in the ADAMTS13 gene, associated with thrombotic thrombocytopenic purpura/Upshaw-Schulman syndrome (MIM#274150), which is rare in newborns but matched some features of this patient. The ADAMTS13 gene was reported in two newborns with Upshaw-Schulman syndrome characterised by thrombocytopenia and haemolytic anaemia,41 which were also displayed in our patient. The other patient, P132, was identified with a 15q11-q13 deletion, and a reported pathogenic variant in OCA2, which was described above. Prader-Willi syndrome (PWS) is characterised by hypotonia and can be accompanied by various complications in the cardiovascular system including venous thromboembolisms, myocardial infarction and pulmonary hypertension.42 The variant in OCA2 is associated with oculocutaneous type II albinism, which has a highly variable phenotype. P132 had hypotonia, feeding difficulty and cryptorchidism, and these features were associated with PWS. Furthermore, total anomalous pulmonary venous drainage and imperforate anus were observed in this patient, which are rare in patients with PWS. For the OCA2 gene, a reported variant in one allele and the whole gene deletion (15q12-13.1) in another allele matched the autosomal recessive inheritance model.

Genetic information is important for developing healthcare strategies at NICUs for the treatments and prevention of morbidity and mortality due to congenital malformations. Among the 44 patients with positive genetic findings, 13 patients were diagnosed with medically actionable disorders based on molecular diagnosis. Specifically, among 21 patients with P/LP SNVs, 7 were with medically actionable genetic disorders. Ultimately, early diagnosis of medically actionable disorders is crucial to prevent further episodes and improve the overall outcome for affected newborns. Additionally, further research into the biological mechanisms of these disorders will lead to new developments in their prevention and treatment.

Among the 221 families included in this study, 131 (59.3%) enrolled patients were first-born children. Clinical diagnosis in newborns is complex due to the incomplete phenotypic characterisation, and genetic screening can help correct or provide additional information for a probability of clinical misdiagnosis.43 44 Genetic screening for newborn deaths in NICUs can effectively identify genetic aetiologies and provide clinical information for accurate and precise genetic counselling and prenatal screening of the families, which will be an effective method to help bear a healthy baby without the same genetic variant(s) carried by the former sibling, especially for families who lost their first child. In this study, five families benefited from the identification of genetic aetiology and avoided the recurrence of the same disease in their family.

WES has advantages over individual gene and small targeted gene panels for mutation detection. Both single gene-based and small panel-based sequencing using genomic locus-specific probes are applicable only for patients with a strong clinical suspicion of a specific genetic defect. This is often challenging in newborn deaths because their clinical phenotypes are hard to fully investigate, and they may have non-specific phenotypic features shared by several different genetic disorders. In contrast, WES is advantageous in that it is not confined to a predetermined set of associated genes or confined to the time of the investigation.

In summary, the large number of early deaths in the NICU in this study provided a more comprehensive perspective on the genetic causes of neonatal mortality. In total, 42 patients were identified with SNVs, including 21 with P/LP variants, and four patients with CNVs (two patients with both SNVs and CNVs). Furthermore, we observed that if genetic testing and molecular aetiology were performed and assessed for all newborns after birth, 13 patients could have had a chance to be cured or at least ameliorated. Genetic analysis for newborns in NICUs can effectively identify genetic etiologies and provide early precise treatment and genetic counselling for families.


We express our deep gratitude to the patients and the families for their willingness and cooperation in the study.



  • LY and XL contributed equally.

  • Correction notice This article has been corrected since it was published Online First. Authors LY and XL have contributed equally.

  • Contributors WZ and LY designed the project. ZL, PZ, LH, GC and LW collected the data. LY, BW and HW did the data curation. XL and ZL analysed the data. XL, ZL, LY and WZ drafted the manuscript. All authors revised and approved the manuscript.

  • Funding This work was supported by grants from the National Key Research and Development Program of China (2016YFC0905100), the Shanghai Municipal Commission of Health and Family Planning (GDEK201701), the Shanghai Shen Kang Hospital Development Center (SHDC12017110), Science and Technology Commission of Shanghai Municipality (16ZR1446500), and Shanghai Sailing Program (16YF1401000).

  • Competing interests None declared.

  • Patient consent for publication Obtained.

  • Ethics approval This study was approved by the ethics committees of Children’s Hospital of Fudan University (2015 (No. 169)).

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data availability statement Data are available upon reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information.