Paediatric Reference Intervals: Current Status, Gaps, Challenges and Future Considerations

Abstract

Establishing paediatric reference intervals (RIs) is a challenging task due to difficulties in subject recruitment, collection of adequate blood volume, and the inherent physiological changes of many biomarkers with age. Despite these challenges, several national and international initiatives have demonstrated: (a) the feasibility of prospectively designed paediatric RI studies; (b) the development of continuous RIs; and (c) the comparison of reference values across analyser types to harmonise paediatric RIs. Whilst these studies have improved the interpretation of paediatric test results and compliance with international accreditation (ISO15189) requirements, several gaps and challenges in translating current paediatric RIs into routine laboratory practice remain. Future priorities for paediatric RI studies include: (a) determination of the impact of discrete versus continuous RIs, analyser-specific versus harmonised RIs, and prospective collection versus data mining on the proportion of results outside the RIs; (b) understanding the clinical implications of analyser-to-analyser variation in reference values and use of evidence-based paediatric harmonised RIs where applicable; (c) adaptation of laboratory information systems to incorporate continuous RIs; (d) further understanding of the biological variation in paediatric biomarkers; (e) studies to address the paucity of accurate data for neonatal RI development; (f) periodic demonstration of RIs being clinically ‘fit-for purpose’; and (g) agreement and policy updates for use of modern, best practice statistical methods in estimation of paediatric RIs. Furthermore, in vitro diagnostic manufacturers may require incentivised paediatric RI studies and publications through co-ordinated grants and collaboration at end-user sites to reduce the burden on sole users.

Introduction

Laboratory test results are critical to modern clinical medicine, with claims that nearly 70% of physicians’ medical decisions are based on information provided by laboratory test reports.¹ The importance of laboratory test results is further evident from the sheer number of pathology tests performed. According to The Royal College of Pathologists of Australasia (RCPA), approximately 500 million laboratory tests are performed each year in Australia and more than 11 million Australians have at least one test per year.²

A test result by itself is of little value unless it is reported with a reliable and appropriate clinically-related interpretation. A reference interval (RI) is one such clinical assessment tool frequently used to interpret test results. Statistically, a RI is defined as a range that includes α% (0<α<100) of the reference population, where lower and upper limits are (100 − α)/2 and (100 + α)/2 centiles respectively.³ The interval between the two extreme percentiles represents the central part of the population, and is frequently estimated by performing the relevant test(s) on healthy individuals or a defined reference population. The basic concept of a RI is that observations that are outside the interval are considered unusual or abnormal, with the understanding that this will falsely classify (100 − α)% of the reference population as ‘unusual’.

The appropriateness of any RI is a critical factor in the clinical interpretation of a test result and subsequent patient diagnosis and treatment. For this reason, the International Organization for Standardization (ISO) have specified requirements for all aspects of laboratory testing, including pre-analytical, analytical and post-analytical (e.g. RIs) phases.^4,5 According to these requirements, clinical laboratories need to periodically demonstrate the appropriateness of the RIs for the population served and the measurement system in use.^4,5

The establishment of a RI is a complex and time-consuming process, especially for children.⁶ The RI establishment process is broadly classified into five steps: (i) definition of a reference population; (ii) selection of the reference individuals; (iii) collection of reference samples; (iv) laboratory testing of the reference samples according to routine, standardised operating procedures; and (v) application of statistical methods to the laboratory results generated in step (iv).⁶ For children, the determination of RIs is difficult because it requires recruitment of adequate numbers of healthy children, collection of relatively large blood volumes proportional to the patient’s blood reserve, and an understanding of the physiological changes in biomarkers associated with normal growth and development.^7–9

One of the main steps in establishing paediatric RIs is recruiting healthy children representative of the community that a specific laboratory or hospital serves.¹⁰ Ideally, each child in the community should have the same probability of being selected as a representative of the reference population. Hence, recruitment of children should occur prospectively from a community having no current interaction with a health service. However, this is extremely difficult, both ethically and pragmatically. A reasonable compromise is to actively recruit otherwise healthy children who present to healthcare services for minor surgical procedures, who would usually not be having clinically investigative blood tests because they are assumed to be normal. From a neonatal perspective, healthy term newborns on the postnatal wards of hospitals represent an ideal sampling population. However, very low consent rates, restrictions in the amount of blood that can be safely collected from children with respect to body weight, a requirement for specialist phlebotomy procedures and staff combined with other ethical (e.g. single collection attempt) and logistical challenges (e.g. storage, transportation and cost) limit the ease with which many laboratories worldwide can attain prospective collection of paediatric blood samples.^11,12 This is particularly true for smaller, non-specialised laboratories that service regional health centres. Alternatively, conducting retrospective studies with data extraction, often referred to as ‘data mining’, from existing laboratory information systems (LIS) or testing leftover samples from laboratory storage have been used in the estimation of paediatric RIs.^13–16 Data mining has the advantage of generating large datasets compared to the direct approach of obtaining and testing samples for the single purpose of establishing RIs.¹⁶ However, despite the large sample size and rigorous statistical analysis, the representativeness of the findings, and thus RIs generated, is influenced by the fact that the majority of the blood tests were initially conducted for investigation of possible disease states, an issue that may affect paediatric more than adult testing.^15,16

In children, changes in specific biomarkers are associated with normal, physiological growth and development (e.g. creatinine and alkaline phosphatase (ALP)) and hence appropriate paediatric RIs must also reflect these changes from birth through adolescence.^13,14,17 Traditionally, RIs were estimated for different discrete age-groups to reflect these age-specific changes in biomarkers.^6,18 However, several studies have identified that discrete age-group RIs are misleading, specifically when testing children whose age is close to the defined age-group cut-off values.^13,14 For example, the continuous increase in creatinine concentration with increased muscle mass (and therefore age) or increase in ALP associated with the osteoblastic activity during the pubertal growth spurt make it evident that discrete age-group RIs alone are not sufficient to capture these physiologically-based changes.^14,19

Despite recent activities to achieve standardisation of laboratory results, between-method differences for the same tests remain.^20–22 As sequential paediatric specimens on the same patient are often tested by different laboratories, the application of appropriate reagent- and analyser-specific RIs for reliable clinical decision making is important. Most importantly, the clinical implication of any inter-analyser variation must be understood when making a decision based on analyser-specific RIs.¹⁷ While there are initiatives to harmonise paediatric RIs used between different laboratories and/or different analysers, the decision to harmonise is often based on clinical consensus between pathologists in the UK or using Bhattacharya analysis in Australia, rather than clinical evaluation of test results that were outside the RIs.^22,23

Considering the challenges in establishing paediatric RIs, the process is likely to be practically limited to large-scale, paediatric specialist centres with dedicated research resources or multinational diagnostic manufacturers. Laboratories that analyse small numbers of paediatric specimens amongst predominantly adult cohorts and/or smaller regional laboratories may be able to adopt only previously published paediatric RIs. Hence the verification and validation of RIs in paediatric populations will play an important role in ensuring appropriate RIs are in use at local laboratory sites.

In this context, the aim of this review is to briefly outline the current status of paediatric RI studies, describe some of the existing gaps and challenges, and highlight important future priorities in the practical implementation of paediatric RIs, in order to improve clinical diagnosis and reduce patient risk.

Current Status

Population-Based, Prospective Paediatric Studies

The quality of RIs depends on the selection and recruitment of the individuals chosen to provide a specific RI, the control and quality of the laboratory’s pre-analytical, analytical and post-analytical process and the statistical methods applied. While controlling all these aspects is challenging, several national and international initiatives have demonstrated that establishing paediatric RIs using samples collected prospectively from population-based studies is feasible. According to the International Federation of Clinical Chemistry and Laboratory Medicine (IFCC), several countries and regions have taken initiatives to establish paediatric RIs.²⁴ In Germany, the German Health Interview and Examination Survey for Children and Adolescents (KiGGS) presented the age- and sex-specific percentile distribution of key laboratory parameters based on more than 14,000 blood and serum samples.²⁵ In the US, the Children’s Health Improvement through Laboratory Diagnostics (CHILDx) study is focusing on determining paediatric RIs for a number of clinical laboratory assays by collecting both blood and urine specimens from healthy subjects.²⁶ In Scandinavian countries, the Scandinavian Initiative for Establishment of Pediatric Reference Intervals (NORICHILD) has reported discrete age-group paediatric RIs based on blood collected from nationwide community surveys.²⁷

The Canadian Laboratory Initiative on Pediatric Reference Intervals (CALIPER) is one of the largest paediatric RI projects completed to date and its numerous publications of paediatric RIs have contributed significantly to improving the clinical diagnosis and interpretation of paediatric test results.¹⁸ CALIPER has collected thousands of samples from healthy children in Canada across several ethnic groups and has established paediatric RIs on Abbott Architect analysers.²⁸ CALIPER has also mathematically transferred the established paediatric RIs to other analytical platforms, and published a systematic review of the statistical methods applied in the estimation of RIs and a simulation study comparing statistical methods for estimation of age-group RIs.^29–31

In Australia, the Lifestyle Of Our Kids (LOOK) study, which is a community-based longitudinal study, followed a cohort of 852 healthy eight-year-old children and reported RIs for 37 analytes at age 8, 10 and 12 y.³² Most recently, the Australian Harmonising Age Pathology Parameters in Kids (HAPPI Kids) study also reported continuous age- and sex-specific RIs for common biochemistry analytes across five commercially available analysers.^19,33 The HAPPI Kids study was the first to directly compare test results across analysers and provide scientific evidence in support of harmonising RIs for the majority of biochemistry analytes.^19,33 It has collected and tested samples from more than 5000 children with the aim to publish age- and sex-specific RIs for common haematology and immunology analytes, as well as to investigate the impact of blood group on RIs.^19,33 Furthermore, the HAPPI Kids team has conducted a systematic review of statistical methods, focusing on the choice between age as a discrete or continuous variable, and a simulation study comparing the statistical methods applied in the estimation of continuous RIs.^34,35

Addressing the lack of accurate paediatric RIs in China, the Pediatric Reference Intervals in China (PRINCE) project aims to establish and verify paediatric RIs for 31 analytes using data collected from approximately 15,000 healthy children from 10 children’s hospitals.³⁶ To date the project has evaluated the study feasibility and efficiency of the method using data from 602 children.³⁶

Discrete Age-Group or Continuous Age-Specific RIs

Considering the challenges in clinical diagnosis based on the discrete age-group RIs for biomarkers that change with age, several studies have established continuous age- and sex-specific paediatric RIs.^{13,14,33,37,38} Bussler et al. established continuous RIs for alanine aminotransferase (ALT), aspartate aminotransferase (AST) and gamma-glutamyl transferase (GGT) as 3^rd and 95^th percentiles using a widely used semi-parametric statistical method (lambda-mu-sigma (LMS)) based on data from 1746 healthy children (3131 cases) aged 11 m to 16 y.³⁷ Zierk et al. reported continuous age- and sex-specific RIs for 22 haematological and biochemical analytes also using an LMS-type method based on available data from a hospital laboratory database for children aged 6 m to 18 y.¹⁴ Recently, CALIPER reported continuous RIs for 38 biochemical markers using non-parametric quantile regression based on the data from the CALIPER database for children aged 6 m to <19 y.³⁸ As mentioned, the HAPPI Kids study also established continuous age-specific RIs for 30 biochemistry analytes across five analysers using quantile regression where (fractional) power variables of age and sex were used as the covariates, using prospectively collected data from 360 healthy children aged 1 m to <18 y.³³ With this method, the HAPPI Kids study reported continuous age-specific RIs as a function of age and sex, enabling a RI to be calculated for any age.³³

Analyser-Specific or Common Paediatric RIs

Addressing inter-analyser variation, the CALIPER study mathematically transferred and validated discrete age-group paediatric RIs from Abbott Architect analysers to other systems including the Beckman Coulter DxC800, Ortho Vitros 5600, Roche Cobas 6000 and Siemens Vista 1500 analysers for common biochemical and immunological analytes.^28,31,39 The authors reported the transferred RIs were, for most analytes, similar to the RIs established on the Abbott Architect c800, while in some cases (e.g. lipase on Siemens Vista and apolipoprotein B on Ortho Vitros) the reference limits were higher than those for the Abbott Architect.³¹ The Australasian Association of Clinical Biochemists (AACB) and RCPA endorsed harmonised paediatric RIs for nine biochemical analytes (i.e. sodium, serum and plasma potassium, chloride, bicarbonate, creatinine, calcium, phosphate, magnesium and ALP for children 0 w to <18 y), established using paediatric data from 15 laboratories (over 200,000 test results) using Bhattacharya analysis.²² However, the comparison of analytical methods to estimate variability between manufacturers was performed using adult samples and may not be representative of the paediatric population, particularly for neonates.²² Most recently, the HAPPI Kids study reported common paediatric RIs for 24 biochemical analytes in addition to analyser-specific RIs for 30 biochemical analytes on Ortho Vitros 5600, Abbott Architect c8000, Roche Cobas c701, Siemens ADVIA 1800 and Beckman Coulter AU5800.^19,33 The HAPPI Kids study is the first study to provide scientific evidence of statistical and clinical significance of differences in age-specific mean reference values across analysers for the paediatric population.¹⁹ This head-to-head comparison of reference values was possible as each blood sample was divided into five aliquots and tested directly on the five different analysers.³³

Gaps and Challenges

Recruitment of Healthy Neonates and Infants

Accurate and reliable paediatric RIs that reflect dynamic changes during the first few days, weeks and months of life are still lacking. Difficulties in recruiting adequate numbers of healthy term neonates and infants were acknowledged by all the major population-based paediatric RI studies.^18,19,40 The CALIPER paediatric RIs for neonates and infants (0–<1 y) are limited as children were recruited from outpatient clinics.¹⁸ Similarly, studies from Denmark and Sweden addressed the limitation of small sample sizes (6 m to 18 y) by merging datasets to establish paediatric RIs.⁴⁰ The HAPPI Kids study successfully collected blood samples from term neonates within 72 h of birth, albeit with low consent rates (16%), but was unsuccessful in collecting adequate samples from pre-term neonates born at 32–36 w gestation (unpublished data). However, the HAPPI Kids study had no samples collected from children aged between 72 h and 1 m of age.¹⁹

Pre-Analytical and Analytical Processes for Testing Paediatric Samples

Routine pre-analytical and analytical processes are important factors that require control when establishing paediatric RIs.⁹ Following safe clinical practice, blood collection for the HAPPI Kids study was limited to <3% of the total blood volume for a child.³³ In addition, ethical consideration of younger children limited multiple or repeated blood collection attempts in this study.¹⁹ Post-collection, standard operating procedures defined HAPPI Kids specimen processing, but nevertheless authors reported data exclusion due to pre-analytical interferences in certain subsets of samples (e.g. sodium, total protein due to evaporation, and total bilirubin due to light exposure).¹⁹ Adeli et al. also raised concerns regarding the ability of automated laboratory equipment to handle small volume specimens, leading to frequent manual processing and difficulties in standardising specimen processing.¹⁷ This problem was also observed in the analysis of the HAPPI Kids specimens, where automated processing was overridden to prevent the analyser from aborting automated specimen analysis due to inadequate sample volumes.¹⁹ In addition to operational pre-analytical and analytical concerns, intra-individual biological variation or total allowable error for different analytes based on adult samples may not be applicable for the paediatric population, especially during the first few years of life.¹⁷ Addressing the lack of data on biological variation specific to paediatric populations, the CALIPER study first published paediatric within-day biological variation and quality specifications for 38 biomarkers using nested analysis of variation techniques.⁴¹ Considering the challenges in recruiting healthy children using direct sampling approach, Loh et al. proposed an indirect method for estimating within-individual biological variation and estimated between-individual biological variation for 22 biomarkers which were comparable to those reported in the CALIPER study.^42,43

Validation and Assimilation of Continuous Paediatric RIs in a Routine Laboratory

Whilst continuous age-specific RIs improve our understanding of the changes in biomarker concentrations with age, there is no current recommendation or endorsed guideline for the validation of continuous age-specific RIs for use by a routine laboratory. In contrast, the Clinical and Laboratory Standards Institute (CLSI) published a guideline for the validation of age-group RIs.⁶ It is uncertain to what extent this guideline is applicable to continuous age-specific RIs. As such, the HAPPI Kids study developed a method to validate its published continuous age-specific RIs with some modification of the current guideline.⁴⁴ Publication and endorsement of a standard protocol for continuous RI validation would provide direction and support for the adoption of continuous paediatric RIs in local laboratories.

Even so, incorporating continuous RIs into existing LIS designed for discrete intervals remains challenging.⁴⁵ At present, LIS have limited capacity to visually display continuous reference curves or calculate continuous RIs from equations based on age and/or sex.¹⁹ Moreover, some laboratories still report test results in a portable document format (PDF) or hard copy, making it difficult to track sequential changes in test results appropriately over time. In addition to the challenges associated with the translation of continuous RIs into LIS, many pathology results are also viewed in other online systems, including electronic medical records (EMR) and Medical Practitioner desktop or mobile software packages. In these situations, developments are required not only to the LIS, but also to prepare the RIs, display them if this is required in a rendered format (e.g. PDF or directly on-screen) and also systematically transfer the results to receiving systems that have the capacity to accurately and safely render the results. This poses additional complexity for the electronic management of patient results and applicable RIs, but as the diversity in receiving systems and software expands, standardisation is vital. Recent studies endorsing the use of continuous paediatric RIs provide impetus for the development of improved modern LIS functionality and associated integrated EMR and other mobile and desktop software used to view and interpret patient results at the medical practitioner interface.^13,19,37,38

Future Considerations

Further Research Addressing the Gaps and Challenges

Target Studies to Establish Neonatal and Infant RIs

Physiological change in neonates is very dynamic, with the first few days of life being described as ‘the most dramatic period of change that occurs in human life’.⁴⁶ As such, concentrations of respiratory and cardiovascular biomarkers can change within minutes after birth, whilst other biomarkers change more slowly as organ systems evolve to transition the neonate from the intrauterine environment to adult physiology.⁴⁶ These dramatic changes in biomarkers make the interpretation of pathological test results challenging, especially in the context of currently available paediatric RIs. As discussed previously, there is a further lack of reliable and accurate neonatal and infantile RIs. Targeted studies of children aged 0, 1, 2, 3, 7, 14 and 28 days, and during the 2^nd, 3^rd, 4^th, 5^th, 6^th, 9^th and 12^th months of life are required. In an effort to address the existing limitations of available data, the CALIPER study has undertaken the recruitment of expectant mothers to facilitate the monitoring of circulating biomarkers at birth and throughout the first year of life.¹⁸ Hence, priority should be given to analytes with proven clinical utility during the neonatal and infancy periods.

Statistical Methods in the Estimation of RIs

There is a wide variety of statistical methods that have been applied in the estimation of age-specific RIs. These include Royston’s parametric curves, Cole’s LMS, generalised additive models for location, scale and shape (GAMLSS) and quantile regression.^3,47–49 Underlying each method is a set of defined assumptions which the method will provide optimum results. While there have been a few studies comparing the performance of different statistical methods based on available data, there is a lack of well-designed statistical simulation studies comparing the performance of these methods for different scenarios i.e. complexity of relationship between age and biomarkers, distribution of samples across age and sample size.^48,50 The minimum sample size required to estimate age-specific paediatric RIs is also frequently not agreed upon. Royston proposed a formula for sample size calculation for estimating age-specific RIs using samples distributed uniformly across age and requiring the ratio of the standard error (SE) of the estimated limits to be no more than 10% of the standard deviation of the variation in the population.³ A modification to the formula for various age distributions has also been proposed.^3,51 However, this approach based on the SE of the limit is not applicable for LMS or GAMLSS. There is also no recommendation for sample size requirement when utilising these statistical methods. Hence, statistical simulation studies may provide guidance for the minimum sample size required to minimise bias for different scenarios mentioned earlier.

Comparison of Flagging Rates – Discrete vs Continuous RIs

One justification for the use of continuous age-specific RIs is the improvement in clinical decision making, particularly in children with ages close to the age cut-off values used in discrete age-group RIs. As shown in the Figure, a girl aged 7 y with a creatinine result of 55 μmol/L is within the discrete age-group RI and outside the continuous RI. In contrast, a girl aged 12 y with a creatinine result of 65 μmol/L is outside the discrete age-group RI and within the continuous RI. Zierk et al. compared the flagging rate of test results as lying outside the RIs in sick children and a reference population between discrete age-group RIs used in laboratories and published continuous age-specific RIs. They reported that, using continuous age-specific RIs, the flagging rate reduced substantially for most blood count analytes and biochemical analytes, except for red cell distribution width, AST, ALT, potassium and lactate dehydrogenase (LDH).¹⁴ However, the proportion of samples and patients considered pathologic based on the continuous RIs was >5% for all analytes.¹⁴ Contrary to Zierk et al., the recent CALIPER study reported that the flagging rate of test results outside the continuous RIs was approximately 2.5% below or above the lower or upper limit respectively for most analytes.³⁸ The flagging rate did not reduce substantially when discrete age-group and continuous age-specific RIs were compared using the same sample.³⁸ However, the authors discussed that, for continuous age-specific RIs, the flagging rates of test results outside the RIs were consistent across age while discrete age-group RIs performed better near the centre of the age-group.³⁸ The difference in flagging rates between the study of Zierk et al. and the CALIPER study is expected as one is an indirect study based on patient samples and the other is based on presumed healthy children.^14,38 While the overall flagging rate can be made the same with both types of intervals, the continuous RI is more likely to correctly classify each patient and should be used as a gold standard to compare the discrete age-group RIs.

Figure.

Visual comparison of trends in reference values over time between hypothetical individuals. Creatinine results (μmol/L) for hypothetical patients Eden and Ella are within the Harmonising Age Pathology Parameters in Kids (HAPPI Kids) and the Australasian Harmonised Reference Intervals for Paediatrics (AHRIP) reference intervals (RIs). However, longitudinal changes (e.g. increasing for Eden and decreasing for Ella) in their creatinine in comparison with the trend observed for the HAPPI Kids continuous RI may be of clinical concern and would allow the physician to better monitor their physiological condition.

Comparison of Paediatric RIs Between Countries

Several countries have established age-specific continuous RIs and have harmonised RIs across laboratories or methods.^18,19,22 However, to date there has been no comparison of these RIs between countries. Considering CALIPER, Zierk et al. and HAPPI Kids studies have published RIs for common biochemistry analytes, a unique opportunity now exists to compare the documented paediatric RIs between countries using an indirect approach.^14,19,38 Whilst this type of comparative study will be confounded by pre- and post-analytical differences and statistical methods applied, it may still assist in understanding ethnic and geographic differences in certain biomarkers after adjusting for confounding factors. The findings would significantly contribute to the establishment of global harmonisation guidelines and standards for paediatric RIs.

Translation of Research Findings into Practice

Interpretation of Pathology Results From Multiple Laboratories

Due to potential variations in pathology reference values derived from different analytical principles and/or analyser types (from the same or multiple laboratories), laboratories have a responsibility to report patient results with valid RIs for the method in use and the population they serve. Laboratories also play an important role in the education of physicians for the interpretation of different pathology test results obtained from multiple laboratories.^17,19 At present, it remains extremely difficult for physicians to identify from the patient pathology report the biomarkers with adopted harmonised or common RIs (e.g. results comparable between laboratories), and those with laboratory-specific RIs that cannot be compared. This issue is of particular importance in health system models in which patients are diagnosed, managed and treated within both core and peripheral services and clinics. CALIPER has taken the initiative in addressing the assessment of paediatric laboratory test results for physicians, healthcare workers, parents and patients worldwide by developing and launching a mobile application. The CALIPER App is an easy and transportable tool for the assessment of the latest reference value database (CALIPER) based on healthy children and adolescents.⁵² The development of mobile applications and complex LIS that enable the end-user to access all RIs relevant to the patient’s assessment would enable better clinical use of our understanding of inter-analyser differences.

Incorporating Results into Electronic Databases

The inability of existing LIS to incorporate advanced mathematical functions or graphical representation of patient pathology data has been identified and remains a key challenge to incorporate continuous age-specific RIs. Overcoming these post-analytical reporting issues may enable easier visual recognition of biomarker trends over time and thus improve longitudinal follow-up of paediatric patients. An example of a visual comparison of trends in reference values over time for hypothetical individuals is shown in the Figure.

Beyond the LIS, the need for universal standards for patient data and RI data transmission becomes even more apparent when the wide variety of medical interfaces used for this purpose is considered. This extends to both larger EMR systems and also desktop and mobile software applications used by medical practitioners to promptly view, interpret and action patient results. In these situations, further technological advancement is required, not only to the LIS, but also to prepare the RIs, display them in a rendered format (e.g. PDF or directly on-screen) and also systematically transfer the results to receiving systems. All of these processes require a capacity to accurately and safely render the results.

Multinational LIS and/or EMR developers have an active role to play in the future development and delivery of end-user requested functionality to improve clinical care in paediatric patient cohorts. From a commercial perspective, these types of LIS and EMR advancements would be directly transferrable to a wide variety of different patient cohort subsets, where physiological or treatment-based changes in biomarkers are evident (e.g. obstetric patients, therapeutic drug monitoring regimes, oncology patients etc.).

Updating Guidelines for Newer Statistical Methods

As outlined, several studies recently published continuous age-specific RIs for the interpretation of pathological data in children.^13,14,19,38 The statistical methods applied in the estimation of continuous age-specific RIs are very different from those utilised in the derivation of discrete age-group RIs. Additionally, there is a variety of suitable methods for estimating continuous age-specific RIs, without any available consensus, guideline or standards from peak professional bodies (e.g. IFCC Reference Intervals and Decision Limits Committee) to assist medical professionals and biostatisticians in this task. Hence, further tasks that lie ahead for relevant professional bodies are the discussion of, consensus agreement on and guideline development of statistical best practice for the estimation of continuous RIs (or revision of guidelines for generation of discrete age-group RIs). It is likely that statistical simulation and comparative studies will be of methodological significance for guideline development.

Sustainability of Paediatric RIs and Clinical Interpretation

Collaboration with Diagnostic Industry Partners

International accreditation standards for medical laboratories (e.g. ISO15189) mandate that individual laboratories be responsible for the periodic review of biological RIs and, where applicable, the further investigation and corrective action of RIs (a) that are no longer appropriate for the reference population served, or (b) when pre-analytical or analytical laboratory procedures have changed.^4,5 This requirement is aimed at maintaining the analytical and clinical interpretative quality of pathology results and acknowledges the relevance of changes in the ethnic distribution of local populations and advancements in automated analyser technologies, analytical methods, reagents and calibrators (as previously discussed). Depending on the origin of the RIs, the commutability of the RI values and the traceability of available metrological and pre-metrological information, the task of individual laboratories maintaining ‘accurate’ RIs can range from onerous (e.g. production of reference values and estimation of the RI) to minimal (e.g. adoption of a published RI without validation). For paediatric patients, where reference values and published RIs have been historically sparse and laboratory testing is commonly performed by non-paediatric specialist laboratories, establishing or adopting paediatric RIs can be particularly difficult to attain. Even in laboratories with access to large numbers of paediatric specimens or major studies (e.g. CALIPER or HAPPI Kids), the cost and time associated with the continual review of and/or re-establishment of appropriate paediatric RIs for changed laboratory conditions remain extremely challenging. The ethical and resource burden on sole laboratories to complete paediatric RI studies and publications may be improved if there were incentivised grants to facilitate further collaboration between vendors and end-user groups, potentially coordinated by vendor agents. A further step would be if existing regulatory frameworks for in-vitro diagnostics (IVDs) were extended to specifically mandate IVD manufacturers to provide paediatric RIs for products developed, marketed and sold as ‘fit for purpose’ for intended paediatric use. The ability of IVD manufacturers to meet such paediatric RI regulatory obligations would require continued collaborations and future development of collective commercial partnerships with major paediatric research teams. Once established, the ongoing verification of RIs can be achieved by the use of external quality assurance material to ensure analytical stability (with a focus on analyte concentrations found in paediatric samples), the application of an ‘average of normal’ or alternative indirect techniques.

Test Results Presented as a Percentile

The representation of an individual’s pathology test result outcome with appropriate RIs at discrete time points (e.g. elevated biomarker at day x) can complicate clinical interpretation by masking a trend in longitudinal test results that is not associated with intra-individual variation. This has been depicted in the Figure, where creatinine results for hypothetical patients Eden and Ella at 5, 10 and 15 y are shown. Their results fell within the HAPPI Kids and Australasian Harmonised Reference Intervals for Paediatrics (AHRIP) RIs, but longitudinal trends in their respective creatinine levels (increasing for Eden and decreasing for Ella) in comparison with the trend observed for the HAPPI Kids continuous RI may be of clinical concern. In contrast, Eleanor’s creatinine results of 30, 45 and 60 mmol/L (at 5, 10 and 15 y respectively) demonstrate her creatinine value remained stable on the 50^th percentile. Future consideration should be given to the representation of paediatric pathology test results as percentiles, as commonly used for other child growth parameters (e.g. weight, height and head circumference) and the potential advantages that this may have in assisting physicians in the clinical interpretation of individual data and thus monitoring of physiological conditions.

Conclusion

Paediatric RIs appropriate for the population being served by a laboratory and the pre-analytical, analytical and post-analytical procedures utilised in generation of test results are essential for improving clinical interpretation and subsequent clinical decisions. Despite the past and present challenges of establishing paediatric reference values and estimating paediatric RIs, there have been new and significant developments published that offer exciting improvements to pathology result interpretation, patient safety and ease of clinical management. However, with rapid advancement in technologies, both the testing processes of the laboratory and the ways of reporting patient results are changing rapidly. Hence, long-term planning to direct future research, software development and best practice clinical and statistical guidelines for paediatric RI development, implementation, review and accreditation will be critical for sustainability of quality and acceptance into routine laboratory practise.