Pediatric and adult reference interval harmonization in Canada: an update

Introduction

Reference intervals (RIs) are one of the most widely used decision-making tools in laboratory medicine and are employed primarily as standards of health in assessing patients for potential disease states. Patients and physicians often assume that test results from different laboratories are equivalent or at least comparable and will be interpreted with the same evidence-based RIs. This assumption is part of a growing expectation for standardized patient care across testing sites, driven in part by integrated health networks with cross-appointed healthcare professionals and multi-institution health information databases. However, it is underappreciated that laboratories operating within the same healthcare system and even using the same instruments may report different RIs for a given analyte. There is a high risk of test result misinterpretation when RIs are not evidence-based or appropriately harmonized; this can potentially lead to unnecessary and even invasive investigations, misdiagnosis or missed diagnosis. The implementation of evidence-based harmonized RIs for laboratory tests is a critical step towards accurate result interpretation and optimized patient care.

There are two key barriers to the determination and implementation of harmonized RIs. The first barrier is a lack of method standardization between manufacturers and even across different instruments from the same manufacturer. Method bias exists for many analytes that are not yet standardized and lack a reference method and/or traceability to a primary or secondary reference material [1]. If these calibration and method differences generate analytical variation beyond a tolerable limit, a harmonized RI could lead to clinical errors; unaccounted for method bias generates inappropriate flagging of results, while a widened interval that accommodates method bias reduces clinical sensitivity. Thus, widespread RI harmonization inherently depends on higher level agreement across different manufacturers and instruments. The second barrier is inadequate data from healthy, representative populations from which to derive robust RIs. Current RIs are often based on out-of-date and inappropriately designed or underpowered studies, particularly for analytes that vary significantly by sex, age or other factors necessitating multiple partitions for accurate interpretation. Furthermore, the identification of healthy individuals may be complicated by undiagnosed or subclinical disease, and obtaining specimens from healthy volunteers may be difficult, especially in the case of pediatrics. Together, these two barriers have hindered the development of accurate RIs with widespread harmonization.

In Canada, the lack of RI harmonization is a major challenge in the provision of quality clinical laboratory services and requires urgent action. To address this challenge, we have developed a strategy to initiate nationwide RI harmonization for the benefit of laboratory professionals, physicians, and most importantly patients (Figure 1).

Figure 1:

The proposed Canadian approach to developing nationwide harmonized reference intervals (hRIs).

The Canadian approach to reference interval harmonization

In 2015, members of the Canadian Society of Clinical Chemists established a working group to drive the determination and implementation of nationwide evidence-based harmonized RIs. As recently outlined [2], the main objectives of this group are to (a) review pediatric and adult RIs currently in use in clinical laboratories across Canada, (b) assess the available evidence on RIs obtained in a priori studies of healthy populations, and (c) develop appropriate guidelines on harmonized RIs and support their implementation across Canada.

To address our first objective, we issued a national survey in April of 2016 and collected the reported RIs for seven common laboratory tests from 37 Canadian laboratories [2]; tests included six biochemical markers and one hematological marker: alanine aminotransferase (ALT), alkaline phopshatase (ALP), calcium, creatinine, free thyroxine (FT4), sodium and hemoglobin. There was extensive variation in reported upper and lower reference limits across and within manufacturer groups (Figure 2A for ALP). Forty laboratories also measured the six biochemical markers in a commutable reference sample to establish a baseline of analytical variation between testing sites and manufacturers. When compared to the assigned value of the reference sample or the all results median, it was not uncommon to find that the bias in reported reference limits was larger than the analytical bias and even in the opposite direction (Figure 2B for ALP). The variation observed in reported reference limits was often greater than the variation in reference sample results for the assessed analytes (Figure 2C), which is consistent with previous findings by the Australasian Association of Clinical Biochemists Harmonization Group [3]. We also observed greater variation in reported pediatric reference intervals compared to adults (Figure 2C). The results of this national survey confirmed a clear opportunity for harmonization, as there is substantial variability in reported RIs that is often not explained by manufacturer-specific bias.

Figure 2:

Results from a Canadian clinical laboratory survey [2] on the current state of reference interval and measurement variation show clear opportunities for harmonization.

(A) The range of reported upper reference limits (URL; mean, dark blue diamonds) and lower reference limits (LRL; mean, light blue circles) is extensive both within and between manufacturer groups, as shown for alkaline phosphatase (ALP) reported for a 14-year-old female. Bars indicate range of reference limit values reported for each manufacturer group, with instances of overlap between the lowest URL and highest LRL (red arrows). Dotted lines indicated the mean URL (dark blue) and LRL (light blue) across all manufacturer groups. (B) Reference limit variation does not reflect compensation for manufacturer bias. For example, measurements of ALP in a reference sample (blue squares, mean; bars, range) were all lower than the reference-assigned target value (red dashed line), and in three cases (red arrows) a negative bias was associated with higher than average mean reference limits (panel A; i.e. the analytical bias and reference limit bias were in opposite directions). (C) The variation of reference limits was often greater than the average between-laboratory coefficient of variation (CV_BL, light blue bars) across all manufacturers determined from the reference sample measurements, as defined previously [2]. Variation in the reported URLs is shown for a 14-year-old female (dark blue diamonds) and 50-year-old male (green diamonds). ALT, alanine aminotransferase; FT4, free thyroxine.

The task of RI harmonization requires a systematic, staged approach. By consensus, we identified 24 common biochemical markers in five test groups for the first round of harmonization in Canada: (1) Electrolytes: sodium, potassium, chloride, total carbon dioxide, magnesium; (2) Renal function markers: creatinine, calcium, phosphate; (3) Hepatic markers: ALT, ALP, albumin, total protein, total bilirubin, lactate dehydrogenase (LDH); (4) Endocrine markers: glucose, hemoglobin A_1c (HbA_1c), FT4, free triiodothyronine (FT3), thyroid stimulating hormone (TSH); and (5) Lipids: cholesterol, high density lipoprotein cholesterol (HDL-C), triglycerides, low density lipoprotein cholesterol (LDL-C, calculated), non-HDL-C (calculated). Information was gathered, where available, on each analyte including reporting units, specimen types, pediatric and adult biological variation, assay manufacturers, routine and reference methods, calibration traceability, manufacturer-provided RIs, and evidence-based RIs from major national studies [4], [5], [6], [7]. One of the central factors to consider for each analyte is method bias in the context of total allowable error defined based on biological variation. The current extent of method biases will be determined by analysis of proficiency testing surveys using commutable materials and additional split-sample comparison studies, combined with RI mid-point comparisons from transference studies. Based on our nationwide survey, we anticipate that all analytes will benefit from RI harmonization [2], with a few amenable to broad harmonization and most requiring manufacturer/instrument-specific harmonization.

The Canadian approach overcomes one of the key barriers to harmonized RI development with our robust data from healthy pediatric and adult populations. Since 2009, the Canadian Laboratory Initiative on Pediatric Reference Intervals (CALIPER) has analyzed specimens from nearly 10,000 healthy community children and adolescents, and published recommended RIs for many biomarkers, with age and sex partitions as required [7]. While the original CALIPER studies were performed on the Abbott platform, transference studies have been completed that enable widespread implementation of these evidence-based RIs on other major biochemical analytical platforms (Beckman, Ortho, Roche and Siemens) [7]. In addition, the Canadian Health Measures Survey (CHMS) involved approximately 12,000 healthy Canadians aged 3–79 years, with participants selected to be representative of 96.3% of the ethnically diverse Canadian population [8], [9]. The collected blood samples and subsequent laboratory analyses allowed determination of appropriately age and sex partitioned RIs for over 35 biomarkers on the Ortho Vitros platform [8], [9]. Consensus preliminary harmonized RIs based on these national databases will be assessed for statistical and clinical significance, and further partitioned where necessary by method manufacturer and instrument type.

Our extensive experience with the CALIPER initiative indicates that RIs generated on large numbers of healthy individuals may in some cases be impractically tight for application within hospital settings. To reduce barriers of harmonized RI implementation and improve clinical utility, CALIPER and CHMS data will be complemented by data from predominately healthy individuals and outpatients provided by Canadian community laboratories. We will verify and refine our preliminary harmonized RIs using statistical analyses of these high-volume data, such as the Bhattacharya approach, which enable identification of the underlying analyte concentration distribution that is representative of the healthy population. The rate at which results are flagged as abnormal due to falling outside of the proposed interval will also be reviewed to ensure suitability for the tested population and practicality in a clinical environment. As a final step, we will take advantage of the significant work published by national harmonization initiatives around the world, and evaluate any discrepancies between our preliminary harmonized RIs and those implemented in other countries [4], [5], [6].

The culmination of this work will be consensus guidelines containing evidence-based harmonized RIs for Canadian laboratories. A key component of our overall approach is to not only produce guidelines, but also to advocate for widespread implementation and provide support for laboratories to determine if harmonized RIs are appropriate for their population and available biochemical analytical platforms. Furthermore, a continuous monitoring/improvement cycle will be critical, as there are several ongoing initiatives to standardize the measurement of individual analytes, such as the International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) Working Group for Standardization of Thyroid Function Tests (C-STFT) for FT4 and TSH. Regular re-evaluations of analytical bias and variation should enable development and implementation of improved evidence-based and cross-manufacturer harmonized RIs.

Overall, our national survey results indicate that most Canadian laboratory professionals are aware of the gaps and inconsistencies in both pediatric and adult RIs [2]. Encouragingly, 92% of survey respondents across Canada either agreed or strongly agreed that there is a need for harmonized RIs and decision limits in clinical laboratories [2]. However, it is important to consider that nationwide implementation and support for harmonized RIs will require the ongoing engagement of stakeholders beyond clinical biochemists and other laboratory professionals, including clinicians and other healthcare professionals, accreditation programs, quality assurance programs, in vitro diagnostics manufacturers and information technologists. Together, we anticipate that the nationwide harmonization of RIs will lead to a significant improvement in laboratory testing and a reduction in medical errors resulting in enhanced patient safety and a higher quality of healthcare for all Canadians.