Development and external validation of machine learning algorithms for postnatal gestational age estimation using clinical data and metabolomic markers.

BACKGROUND: Accurate estimates of gestational age (GA) at birth are important for preterm birth surveillance but can be challenging to obtain in low income countries. Our objective was to develop machine learning models to accurately estimate GA shortly after birth using clinical and metabolomic data. METHODS: We derived three GA estimation models using ELASTIC NET multivariable linear regression using metabolomic markers from heel-prick blood samples and clinical data from a retrospective cohort of newborns from Ontario, Canada. We conducted internal model validation in an independent cohort of Ontario newborns, and external validation in heel prick and cord blood sample data collected from newborns from prospective birth cohorts in Lusaka, Zambia and Matlab, Bangladesh. Model performance was measured by comparing model-derived estimates of GA to reference estimates from early pregnancy ultrasound. RESULTS: Samples were collected from 311 newborns from Zambia and 1176 from Bangladesh. The best-performing model accurately estimated GA within about 6 days of ultrasound estimates in both cohorts when applied to heel prick data (MAE 0.79 weeks (95% CI 0.69, 0.90) for Zambia; 0.81 weeks (0.75, 0.86) for Bangladesh), and within about 7 days when applied to cord blood data (1.02 weeks (0.90, 1.15) for Zambia; 0.95 weeks (0.90, 0.99) for Bangladesh). CONCLUSIONS: Algorithms developed in Canada provided accurate estimates of GA when applied to external cohorts from Zambia and Bangladesh. Model performance was superior in heel prick data as compared to cord blood data.

Families' healthcare experiences for children with inherited metabolic diseases: protocol for a mixed methods cohort study.

INTRODUCTION: Children with inherited metabolic diseases (IMDs) often have complex and intensive healthcare needs and their families face challenges in receiving high-quality, family centred health services. Improvement in care requires complex interventions involving multiple components and stakeholders, customised to specific care contexts. This study aims to comprehensively understand the healthcare experiences of children with IMDs and their families across Canada. METHODS AND ANALYSIS: A two-stage explanatory sequential mixed methods design will be used. Stage 1: quantitative data on healthcare networks and encounter experiences will be collected from 100 parent/guardians through a care map, 2 baseline questionnaires and 17 weekly diaries over 5-7 months. Care networks will be analysed using social network analysis. Relationships between demographic or clinical variables and ratings of healthcare experiences across a range of family centred care dimensions will be analysed using generalised linear regression. Other quantitative data related to family experiences and healthcare experiences will be summarised descriptively. Ongoing analysis of quantitative data and purposive, maximum variation sampling will inform sample selection for stage 2: a subset of stage 1 participants will participate in one-on-one videoconference interviews to elaborate on the quantitative data regarding care networks and healthcare experiences. Interview data will be analysed thematically. Qualitative and quantitative data will be merged during analysis to arrive at an enhanced understanding of care experiences. Quantitative and qualitative data will be combined and presented narratively using a weaving approach (jointly on a theme-by-theme basis) and visually in a side-by-side joint display. ETHICS AND DISSEMINATION: The study protocol and procedures were approved by the Children's Hospital of Eastern Ontario's Research Ethics Board, the University of Ottawa Research Ethics Board and the research ethics boards of each participating study centre. Findings will be published in peer-reviewed journals and presented at scientific conferences.

Real world external validation of metabolic gestational age assessment in Kenya.

Using data from Ontario Canada, we previously developed machine learning-based algorithms incorporating newborn screening metabolites to estimate gestational age (GA). The objective of this study was to evaluate the use of these algorithms in a population of infants born in Siaya county, Kenya. Cord and heel prick samples were collected from newborns in Kenya and metabolic analysis was carried out by Newborn Screening Ontario in Ottawa, Canada. Postnatal GA estimation models were developed with data from Ontario with multivariable linear regression using ELASTIC NET regularization. Model performance was evaluated by applying the models to the data collected from Kenya and comparing model-derived estimates of GA to reference estimates from early pregnancy ultrasound. Heel prick samples were collected from 1,039 newborns from Kenya. Of these, 8.9% were born preterm and 8.5% were small for GA. Cord blood samples were also collected from 1,012 newborns. In data from heel prick samples, our best-performing model estimated GA within 9.5 days overall of reference GA [mean absolute error (MAE) 1.35 (95% CI 1.27, 1.43)]. In preterm infants and those small for GA, MAE was 2.62 (2.28, 2.99) and 1.81 (1.57, 2.07) weeks, respectively. In data from cord blood, model accuracy slightly decreased overall (MAE 1.44 (95% CI 1.36, 1.53)). Accuracy was not impacted by maternal HIV status and improved when the dating ultrasound occurred between 9 and 13 weeks of gestation, in both heel prick and cord blood data (overall MAE 1.04 (95% CI 0.87, 1.22) and 1.08 (95% CI 0.90, 1.27), respectively). The accuracy of metabolic model based GA estimates in the Kenya cohort was lower compared to our previously published validation studies, however inconsistency in the timing of reference dating ultrasounds appears to have been a contributing factor to diminished model performance.

Core Outcome Sets for Medium-Chain Acyl-CoA Dehydrogenase Deficiency and Phenylketonuria.

BACKGROUND: Evidence to guide treatment of pediatric medium-chain acyl-coenzyme A dehydrogenase (MCAD) deficiency and phenylketonuria (PKU) is fragmented because of large variability in outcome selection and measurement. Our goal was to develop core outcome sets (COSs) for these diseases to facilitate meaningful future evidence generation and enhance the capacity to compare and synthesize findings across studies. METHODS: Parents and/or caregivers, health professionals, and health policy advisors completed a Delphi survey and participated in a consensus workshop to select core outcomes from candidate lists of outcomes for MCAD deficiency and PKU. Delphi participants rated the importance of outcomes on a nine-point scale (1-3: not important, 4-6: important but not critical, 7-9: critical). Candidate outcomes were progressively narrowed down over 3 survey rounds. At the workshop, participants evaluated the remaining candidate outcomes using an adapted nominal technique, open discussion, and voting. After the workshop, we finalized the COSs and recommended measurement instruments for each outcome. RESULTS: There were 85, 61, and 53 participants across 3 Delphi rounds, respectively. The candidate core outcome lists were narrowed down to 20 outcomes per disease to be discussed at the consensus workshop. Voting by 18 workshop participants led to COSs composed of 8 and 9 outcomes for MCAD deficiency and PKU, respectively, with measurement recommendations. CONCLUSIONS: These are the first known pediatric COSs for MCAD deficiency and PKU. Adoption in future studies will help to ensure best use of limited research resources to ultimately improve care for children with these rare diseases.

Evaluation of the quality of clinical data collection for a pan-Canadian cohort of children affected by inherited metabolic diseases: lessons learned from the Canadian Inherited Metabolic Diseases Research Network.

BACKGROUND: The Canadian Inherited Metabolic Diseases Research Network (CIMDRN) is a pan-Canadian practice-based research network of 14 Hereditary Metabolic Disease Treatment Centres and over 50 investigators. CIMDRN aims to develop evidence to improve health outcomes for children with inherited metabolic diseases (IMD). We describe the development of our clinical data collection platform, discuss our data quality management plan, and present the findings to date from our data quality assessment, highlighting key lessons that can serve as a resource for future clinical research initiatives relating to rare diseases. METHODS: At participating centres, children born from 2006 to 2015 who were diagnosed with one of 31 targeted IMD were eligible to participate in CIMDRN's clinical research stream. For all participants, we collected a minimum data set that includes information about demographics and diagnosis. For children with five prioritized IMD, we collected longitudinal data including interventions, clinical outcomes, and indicators of disease management. The data quality management plan included: design of user-friendly and intuitive clinical data collection forms; validation measures at point of data entry, designed to minimize data entry errors; regular communications with each CIMDRN site; and routine review of aggregate data. RESULTS: As of June 2019, CIMDRN has enrolled 798 participants of whom 764 (96%) have complete minimum data set information. Results from our data quality assessment revealed that potential data quality issues were related to interpretation of definitions of some variables, participants who transferred care across institutions, and the organization of information within the patient charts (e.g., neuropsychological test results). Little information was missing regarding disease ascertainment and diagnosis (e.g., ascertainment method - 0% missing). DISCUSSION: Using several data quality management strategies, we have established a comprehensive clinical database that provides information about care and outcomes for Canadian children affected by IMD. We describe quality issues and lessons for consideration in future clinical research initiatives for rare diseases, including accurately accommodating different clinic workflows and balancing comprehensiveness of data collection with available resources. Integrating data collection within clinical care, leveraging electronic medical records, and implementing core outcome sets will be essential for achieving sustainability.

Metabolic gestational age assessment in low resource settings: a validation protocol.

Preterm birth is the leading global cause of neonatal morbidity and mortality. Reliable gestational age estimates are useful for quantifying population burdens of preterm birth and informing allocation of resources to address the problem. However, evaluating gestational age in low-resource settings can be challenging, particularly in places where access to ultrasound is limited. Our group has developed an algorithm using newborn screening analyte values derived from dried blood spots from newborns born in Ontario, Canada for estimating gestational age within one to two weeks. The primary objective of this study is to validate a program that derives gestational age estimates from dried blood spot samples (heel-prick or cord blood) collected from health and demographic surveillance sites and population representative health facilities in low-resource settings in Zambia, Kenya, Bangladesh and Zimbabwe. We will also pilot the use of an algorithm to identify birth percentiles based on gestational age estimates and weight to identify small for gestational age infants. Once collected from local sites, samples will be tested by the Newborn Screening Ontario laboratory at the Children's Hospital of Eastern Ontario (CHEO) in Ottawa, Canada. Analyte values will be obtained through laboratory analysis for estimation of gestational age as well as screening for other diseases routinely conducted at Ontario's newborn screening program. For select conditions, abnormal screening results will be reported back to the sites in real time to facilitate counseling and future clinical management. We will determine the accuracy of our existing algorithm for estimation of gestational age in these newborn samples. Results from this research hold the potential to create a feasible method to assess gestational age at birth in low- and middle-income countries where reliable estimation may be otherwise unavailable.

What is in a Name? Parent, Professional and Policy-Maker Conceptions of Consent-Related Language in the Context of Newborn Screening.

Newborn bloodspot screening programs are some of the longest running population screening programs internationally. Debate continues regarding the need for parents to give consent to having their child screened. Little attention has been paid to how meanings of consent-related terminology vary among stakeholders and the implications of this for practice. We undertook semi-structured interviews with parents (n = 32), healthcare professionals (n = 19) and policy decision makers (n = 17) in two Canadian provinces. Conceptions of consent-related terms revolved around seven factors within two broad domains, decision-making and information attainment. Decision-making comprised: parent decision authority; voluntariness; parent engagement with decision-making; and the process of enacting choice. Information ascertainment comprised: professional responsibilities (including disclosure of information and time to review); parent responsibilities; and the need for discussion and understanding prior to a decision. Our findings indicate that consent-related terms are variously understood, with substantive implications for practice. We suggest that consent procedures should be explained descriptively, regardless of approach, so there are clear indications of what is expected of parents and healthcare professionals. Support systems are required both to meet the educational needs of parents and families and to support healthcare professionals in delivering information in a manner in keeping with parent needs.

External validation of postnatal gestational age estimation using newborn metabolic profiles in Matlab, Bangladesh.

This study sought to evaluate the performance of metabolic gestational age estimation models developed in Ontario, Canada in infants born in Bangladesh. Cord and heel prick blood spots were collected in Bangladesh and analyzed at a newborn screening facility in Ottawa, Canada. Algorithm-derived estimates of gestational age and preterm birth were compared to ultrasound-validated estimates. 1036 cord blood and 487 heel prick samples were collected from 1069 unique newborns. The majority of samples (93.2% of heel prick and 89.9% of cord blood) were collected from term infants. When applied to heel prick data, algorithms correctly estimated gestational age to within an average deviation of 1 week overall (root mean square error = 1.07 weeks). Metabolic gestational age estimation provides accurate population-level estimates of gestational age in this data set. Models were effective on data obtained from both heel prick and cord blood, the latter being a more feasible option in low-resource settings.

Incidental screen positive findings in a prospective cohort study in Matlab, Bangladesh: insights into expanded newborn screening for low-resource settings.

BACKGROUND: Newborn screening programs are essential preventative public health initiatives but are not widely available in low-resource settings. The objective of this study was to describe the frequency and nature of screen positive determinations as made by a Canadian newborn screening program in a cohort of infants born in Matlab, Bangladesh. Dried newborn cord and heel-prick blood spot samples collected as part of a validation study nested within a preterm birth research cohort were collected between January 2017 and July 2018 and analyzed in a Canadian newborn screening laboratory where the laboratory's disease panel and screening thresholds were applied. RESULTS: A total of 1661 newborn samples (520 heel-prick and 1141 cord blood samples) were available for analysis. Based on the applied screening thresholds, 61 samples (22 by heel-prick and 39 by cord blood) were screen positive for conditions included in the Canadian disease panel. Congenital hypothyroidism was the most common determination for heel-prick (n = 17) and cord blood (n = 12) samples. Carriers of hemoglobinopathy variants were identified in 6.9% of both tested heel-prick and cord blood samples. CONCLUSIONS: This study provides insight into the nature and frequency of treatable congenital conditions in a rural Bangladesh community where such data were previously unavailable. As comment to the feasibility of newborn screening in the region we confirm that screening based on cord blood sampling continues to be the most acceptable modality to parents in such settings. Acknowledged barriers include early infant discharge, which may affect the reliability of initial screening thresholds to determine disease risk. We further highlight the importance of continuing efforts in the country to identify infants with congenital hypothyroidism.

Metabolic Signatures of Cystic Fibrosis Identified in Dried Blood Spots For Newborn Screening Without Carrier Identification.

Cystic fibrosis (CF) is a complex multiorgan disorder that is among the most common fatal genetic diseases benefiting from therapeutic interventions early in life. Newborn screening (NBS) for presymptomatic detection of CF currently relies on a two-stage immunoreactive trypsinogen (IRT) and cystic fibrosis transmembrane conductance regulator (CFTR) mutation panel algorithm that is sensitive but not specific for identifying affected neonates with a low positive predictive value. For the first time, we report the discovery of a panel of CF-specific metabolites from a single 3.2 mm diameter dried blood spot (DBS) punch when using multisegment injection-capillary electrophoresis-mass spectrometry (MS) as a high-throughput platform for nontargeted metabolite profiling from volume-restricted/biobanked specimens with quality control. This retrospective case-control study design identified 32 metabolites, including a series of N-glycated amino acids, oxidized glutathione disulfide, and nicotinamide that were differentially expressed in normal birth weight CF neonates without meconium ileus ( n = 36) as compared to gestational age/sex-matched screen-negative controls ( n = 44) after a false discovery rate adjustment ( q < 0.05). Also, 16 metabolites from DBS extracts allowed for discrimination of true CF cases from presumptive screen-positive carriers with one identified CFTR mutation and transient neonatal hypertrypsinogenemic neonates ( n = 72), who were later confirmed as unaffected due to a low sweat chloride (<29 mM) test result. Importantly, six CF-specific biomarker candidates satisfying a Bonferroni adjustment ( p < 7.25 × 10-5) from three independent batches of DBS specimens included several amino acids depleted in circulation (Tyr, Ser, Thr, Pro, Gly) likely reflecting protein maldigestion/malabsorption. Additionally, CF neonates had lower ophthalmic acid as an indicator of oxidative stress due to impaired glutathione efflux from exocrine/epithelial tissue and elevation of an unknown trivalent peptide that was directly correlated with IRT (ρ = 0.332, p = 4.55 × 10-4). Structural elucidation of unknown metabolites was performed by high-resolution MS/MS, whereas biomarker validation was realized when comparing a subset of metabolites from matching neonatal DBS specimens independently analyzed by direct infusion-MS/MS at an accredited NBS facility. This work sheds new light into the metabolic phenotype of CF early in life, which is required for better functional understanding of CFTR mutations of unknown clinical consequence and the development of more accurate yet cost-effective strategies for CF screening.

Metabolic profiles derived from residual blood spot samples: A longitudinal analysis.

Background: Secondary use of newborn screening dried blood spot samples include use for biomedical or epidemiological research. However, the effects of storage conditions on archival samples requires further examination. The objective of this study was to determine the utility of residual newborn samples for deriving reliable metabolic gestational age estimates. Methods: Residual newborn dried blood spot samples that had been stored for 2-, 4-, 6-, or 12-months in temperature controlled (21°C) conditions were re-analyzed for the full panel of newborn screening analytes offered by a provincial newborn screening lab in Ottawa, Canada. Data from re-analyzed samples were compared to corresponding baseline newborn screening values for absolute agreement, and Pearson and intraclass correlation. Performance of a gestational age estimation algorithm originally developed from baseline newborn screening values was then validated on data derived from stored samples. Results: A total of 307 samples were used for this study. 17-hydroxyprogesterone and newborn hemoglobin profiles measured by immunoassay and high-performance liquid chromatography, respectively, were among the most stable markers across all time points of analysis. Acylcarnitines exhibited the greatest degree of variation in stability upon repeat measurement. The largest shifts in newborn analyte profiles and the poorest performance of metabolic gestational age algorithms were observed when samples were analyzed 12-months after sample collection. Conclusions: Duration of sample storage, independent of temperature and humidity, affects newborn screening profiles and gestational age estimates derived from metabolic gestational dating algorithms. When considering use of dried blood spot samples either for clinical or research purposes, care should be taken when interpreting data stemming from secondary use.

Temporal Signal Pattern Recognition in Mass Spectrometry: A Method for Rapid Identification and Accurate Quantification of Biomarkers for Inborn Errors of Metabolism with Quality Assurance.

Mass spectrometry (MS)-based metabolomic initiatives that use conventional separation techniques are limited by low sample throughput and complicated data processing that contribute to false discoveries. Herein, we introduce a new strategy for unambiguous identification and accurate quantification of biomarkers for inborn errors of metabolism (IEM) from dried blood spots (DBS) with quality assurance. A multiplexed separation platform based on multisegment injection-capillary electrophoresis-mass spectrometry (MSI-CE-MS) was developed to provide comparable sample throughput to flow injection analysis-tandem MS (FIA-MS/MS) but with greater selectivity as required for confirmatory testing and discovery-based metabolite profiling of volume-restricted biospecimens. Mass spectral information is encoded temporally within a separation by serial injection of three or more sample pairs, each having a unique dilution pattern, alongside a quality control (QC) that serves as a reference in every run to facilitate between-sample comparisons and/or batch correction due to system drift. Optimization of whole blood extraction conditions on DBS filter paper cut-outs was first achieved to maximize recovery of a wide range of polar metabolites from DBS extracts. An interlaboratory comparison study was also conducted using a proficiency test and retrospective neonatal DBS that demonstrated good agreement between MSI-CE-MS and validated FIA-MS/MS methods within an accredited facility. Our work demonstrated accurate identification of various IEM based on reliable measurement of a panel of primary or secondary biomarkers above an upper cutoff concentration limit for presumptive screen-positive cases without stable isotope-labeled reagents. Additionally, nontargeted metabolite profiling by MSI-CE-MS with temporal signal pattern recognition revealed new biomarkers for early detection of galactosemia, such as N-galactated amino acids, that are a novel class of pathognomonic marker due to galactose stress in affected neonates.

Experiences of caregivers of children with inherited metabolic diseases: a qualitative study.

BACKGROUND: We sought to understand the experiences of parents/caregivers of children with inherited metabolic diseases (IMD) in order to inform strategies for supporting patients and their families. We investigated their experiences regarding the management of disease, its impact on child and family life, and interactions with the health care system. METHODS: From four Canadian centres, we conducted semi-structured telephone interviews with parents/caregivers of children with an IMD who were born between 2006 and 2015 and who were participating in a larger cohort study. Participants were selected with the aim of achieving a diverse sample with respect to treatment centre, IMD, and age of the child. Interviews emphasized the impacts of the disease and its treatment on the child and family and explicitly queried perceptions of interactions with the health care system. We identified emergent themes from the interview data. RESULTS: We completed interviews with 21 parents/caregivers. The 21 children were aged <1 to 7 years old with IMD that included amino acid disorders, urea cycle disorders, fatty acid oxidation disorders, and organic acid disorders or 'other' IMD. Most parents reported that they and their families had adapted well to their child's diagnosis. Parents used proactive coping strategies to integrate complex disease management protocols into routine family life. An important source of stress was concern about the social challenges faced by their children. Participants reported positive interactions with their most involved health care providers within the metabolic clinic. However, they reported challenges associated with the health care system outside of disease-specific metabolic care, when encountering systems and providers unfamiliar with the child's disease. CONCLUSIONS: The successful use of proactive coping strategies among parents of children with IMD in this study suggests the potential value of promoting positive coping and is an important direction for future study. Parents' social concerns for their children were important stressors that warrant consideration by health care providers positioned to support families. Our results with respect to experiences with care highlight the important role of specialized metabolic clinics and point to a need for better coordination of the care that takes place outside the disease-specific management of IMD.

Consent for newborn screening: parents' and health-care professionals' experiences of consent in practice.

Consent processes for newborn bloodspot screening (NBS) are variable, with a lack of descriptive research that depicts how the offer of NBS is made to parents. We explored the experience, in practice, of consent for NBS. Semistructured interviews in two Canadian provinces were held with: (1) parents of children offered NBS (n=32); and (2) health-care professionals involved in the NBS process (n=19). Data on recollections of NBS, including consent processes, were utilized to identify emerging themes using the method of constant comparison. Three themes were relevant to NBS consent: (1) The 'offer' of NBS; (2) content and timing of information provision; and (3) the importance of parental experiences for consent decisions. Recollections of consent for NBS were similar between jurisdictions. Excepting midwives and their patients, NBS was viewed as a routine part of giving birth, with little evidence of an informed consent process. Although most parents were satisfied, all respondents suggested information about NBS be provided long before the birth. Accounts of parents who declined screening highlight the influence of parental experiences with the heel prick process in screening decisions. Findings further our understanding of consent in practice and highlight areas for improvement in parent-provider interactions.

Metabolic Clinic Atlas: Organization of Care for Children with Inherited Metabolic Disease in Canada.

INTRODUCTION: Nearly all children in Canada with an inherited metabolic disease (IMD) are treated at one of the country's Hereditary Metabolic Disease Treatment Centres. We sought to understand the system of care for paediatric IMD patients in Canada in order to identify sources of variation and inform future research priorities. METHODS: Treatment centres were contacted by email and invited to complete a web-based survey. The questionnaire addressed, for each centre, the population size served and scope of practice, available human resources and clinic services and research capacity. Survey responses were analyzed descriptively. RESULTS: We received responses from 13 of the 14 treatment centres invited to participate. These centres represent at least 85% of the Canadian population, with over half of the centres located in southern Ontario and Quebec. All centres reported paediatric patients with IMDs as their main patient population. A variety of dedicated staff was identified; every centre reported having at least one physician and one dietician. The most common ancillary services available included telehealth (11/12 respondents) and biochemical genetic laboratory testing (10/12), with a high variability of access to on-site laboratory tests. A majority of centres indicated access to additional off-site services, but barriers to these were reported. All but one centre indicated previous experience with research. CONCLUSIONS: The variation we identified in the organization of care highlights the need to investigate the association between practice differences and health outcomes for paediatric IMD patients to inform policies that establish equitable access to services that are beneficial.