Harmonising Adult Reference Intervals in Australia and New Zealand – the Continuing Story

Abstract

Reference intervals (RIs) are used to help clinicians determine if a patient can be classified as being in a diseased or healthy state and there are often sound scientific and clinical reasons for differences in RIs. One of the current strategic priorities for the Australasian Association of Clinical Biochemists is to encourage and assist laboratories to achieve harmonisation of RIs for common clinical chemistry analytes where sound calibration and traceability are in place. This need is based on good laboratory practice, providing the clinician with results that allow appropriate and reliable clinical interpretation and progression further toward the national e-health framework and a single electronic health record. After reviewing and considering studies related to bias as well as both a priori and a posteriori RI studies nationally and internationally and the consideration of flagging rates and clinical relevance, an initial group of 12 harmonised RIs were endorsed by the Royal College of Pathologists of Australasia in 2014. In 2015, after further stakeholder consultation, a second group of six harmonised RIs for common chemistry analytes has been proposed for adults which includes ALT and AST where methods do not use pyridoxal-5′-phosphate as an activator and lipase excluding the Ortho Clinical Diagnostics and Siemens Dimension assays.

Introduction

There are often sound scientific reasons for differences in RIs; however both a survey of laboratory RIs by the Australasian Association of Clinical Biochemists (AACB) and evidence from the UK Pathology Harmony study found that laboratories using the same analytical platform and reagents often had different RIs with no sound basis for these differences.^1,2

A decade on since Graham Jones and Tony Barker spoke at length around Australia and New Zealand on the benefits of harmonising RIs and six years since the AACB Committee for Common Reference Intervals was re-formed, the AACB and Royal College of Pathologists of Australasia (RCPA) endorsed harmonised RIs for 12 common chemistry analytes.³ The harmonised adult RIs are shown in Table 1. The AACB has recently proposed a further six analytes as candidates for harmonised RIs: ALT, AST, GGT, CK, lipase and total bilirubin.

Table 1.

Approved and recommended adult harmonised reference intervals.^a

Analyte	Male	Female
Sodium	135 – 145 mmol/L
Potassium b	3.5 – 5.2 mmol/L
Chloride	95 – 110 mmol/L
Bicarbonate	22 – 32 mmol/L
Creatinine c	60 – 110 μmol/L	45 – 90 μmol/L
Calcium	2.10 – 2.60 mmol/L
Calcium (albumin adjusted)	2.10 – 2.60 mmol/L
Phosphate d	0.75 – 1.50 mmol/L
Magnesium	0.70 – 1.10 mmol/L
Lactate Dehydrogenase	120 – 250 U/L
[L to P] (IFCC) e
Alkaline Phosphatase f	30 – 110 U/L
Total Protein	60 – 80 g/L
Total Bilirubin	1 – 20 µmol/L
Creatine Kinase	<60 y: 45 – 250 U/L	30 – 150 U/L
	60+ y: 40 – 200 U/L
Alanine Transaminase (no pyridoxal 5-phosphate)	5 – 40 U/L	5 – 35 U/L
Aspartate Transaminase (no pyridoxal 5-phosphate)	5 – 35 U/L	5 – 30 U/L
Gamma Glutamyl Transferase	5 – 50 U/L	5 – 35 U/L
Serum Lipase g	10 – 60 U/L

^aUnless otherwise specified, the intervals are for serum or plasma for adults (18 y and older). The bias study indicates that these analytes are largely harmonised. Available at: http://www.aacb.asn.au/documents/item/351 (Accessed on April 29, 2016).

^bThis range is proposed for use for both serum and plasma. Laboratories testing only heparin plasma may choose to use a lower interval.

^cCreatinine has harmonised reference intervals for adults up to the age of 60 y. For older ages laboratories may elect to maintain these.

^dStarting at age 20 y to align with paediatric intervals.

^e[L to P] (IFCC), lactate to pyruvate method (IFCC method).

^fStarting at age 22 y to align with paediatric intervals.

^gThe reference interval for adult serum lipase excludes Siemens Dimension and Ortho Clinical Vitros. There are linear relationships between the ‘harmonised’ assay group and the Dimension and Vitros: ‘Harmonised’ = Dimension × 0.21 − 0.6; ‘Harmonised’ = Vitros × 0.27 +12.

This review is undertaken to discuss the steps and considerations within the AACB RI harmonisation working group when determining those recommended RIs and seeking endorsement by the RCPA for those recently published and future recommendations. The AACB and RCPA invited pathologists and medical scientists to harmonise RIs at the same time as other RCPA initiatives for standardisa tion of pathology units, terminology and report formatting and flagging were being undertaken.⁴ The AACB Committee for Common Reference Intervals identified the difficulties associated with a classical or direct (a priori) approach to harmonisation such as addressing assay traceability, bias and imprecision and proposed that a harmonisation approach needed to consider between-method bias. RIs are used to help clinicians determine if a patient can be classified as being in either a healthy or diseased state so any method bias may result in misclassification of patients. In the case of the recommended and proposed AACB harmonised RIs, specific performance limits based on biological variation have been applied to determine whether bias would prevent harmonisation across platforms.⁵ An assumption was made that the populations served by the different laboratories in Australia and New Zealand are similar. The committee has been cognisant when developing harmonised RIs that where a RI is harmonised, the resulting RI will be wider than that derived from a single laboratory because the analytical variation (CVa) includes a greater number of analytical platforms in a large number of laboratories using more than one method.⁶

Bias and Commutability

The presence and magnitude of any bias can be determined by comparing results obtained by the various analytical methods using shared patient samples, samples from external quality assurance (EQA) programs or inter-laboratory internal quality control programs.¹ Results tend to show greater variability and method dependence when using ‘artificial’ material rather than fresh or frozen patient samples. The method differences seen using these artificial materials such as EQA material are commonly due to matrix effects.⁷ To assess whether these artificial materials behave in the same manner as authentic clinical samples and are suitable for analysis of bias, commutability should be determined. Without assessment of commutability it is not possible to determine whether any biases observed are artifactual or genuine.⁸ Commutability is a property of reference materials where those materials have the same inter-assay relationships as those of clinical samples or the mathematical relationships between the results of different measurement procedures for that reference material and for representative samples from healthy and diseased persons are the same. The accuracy of results is not implied by commutability, only that results for that reference material have the same mathematical relationship between methods that was observed for clinical samples measured by those methods.⁹

Because of method differences seen using artificial material in the RCPA Quality Assurance Programs (RCPAQAP) and the potential for bias, a study was undertaken by the AACB to determine bias using unadulterated commutable human serum.¹ In that study 27 commonly performed analytes were assayed on 33 serum samples using the eight major chemistry platforms as determined by reviewing the RCPAQAP enrolments in 24 public and private laboratories throughout Australia and New Zealand. The serum samples used were largely from single volunteers (24/33); however 9/33 were from pooled serum samples ensuring that the entire generally accepted RI concentrations for each analyte were covered. Of the 27 tested analytes, 19 demonstrated that bias would not prevent harmonisation.¹

Reference Interval Studies

The a priori approach to determining a RI is a time-consuming process requiring considerable professional judgement and significant amounts of data on which to base decisions.

There have been some very successful multicentre studies that involved different regions and countries in the formulation of harmonised RIs. In European regions and countries, Scandinavia and Spain have been most notable in their pursuit of harmonisation. The Nordic NORIP (Nordic Reference Interval Project) study is one of the most successful and widely reported of these studies with numerous publications since 2004.^10,11

Australian Reference Interval Study

The Aussie Normals study was a formal a priori RI study of 1876 male and female healthy adult Australians in the age group 18–95 y. Up to 91 biochemistry analytes were measured on Abbott Architect analysers. The partitioning undertaken in this study was by age, gender and body mass index (BMI) where appropriate.¹² The values obtained from this formal RI study, along with professional judgement comparison and other national and international studies such as NORIP and data from the a posteriori analysis undertaken by Sonic Health, have been used to confirm the harmonised RIs recommended for use in Australia and New Zealand.

Data Mining

An alternative approach for establishing RIs is to do an indirect or a posteriori study. Results stored in laboratory databases have the ability to serve as ideal repositories of information regarding the general population. Expert groups can provide RI information through mining of millions of data points from primary care patients. This method has advantages over the direct RI validation process by providing large amounts of data on the local population being tested and reflects the actual analytical and pre-analytical conditions for the tested population.

A data mining a posteriori approach that has been used in Australia and has provided useful information is the use of Bhattacharya analysis.^13,14 This method assumes a Gaussian or log Gaussian distribution of results. This approach is valid only if there is a majority of results from the primary care population such that the healthy distribution of values can be clearly identified in the midst of a smaller number of non-healthy values. This is the method used by Sonic Health and was used as part of the AACB/RCPA paediatric RI harmonisation project with over 200,000 paediatric data points provided by 15 laboratories for the main general chemistry analytes from birth to 18 years of age. These data were mined and, after consideration of paediatric physiology, used in the establishment of the recommended partitioned paediatric harmonised RIs.³

Clinical Considerations of the Reference Interval and Flagging Rates

The proposed reference limits for adults have been supported by flagging rates which provide an indication of the clinical considerations of a RI. Excess flagging of results can lead to inappropriate testing due to decreased specificity of the RI. The flagging rates provided in this review are those when community samples are considered.

Horowitz has suggested that laboratories should be mindful of excessive partitioning that is due to minor changes in physiology rather than pathology.¹⁵ It is also important to emphasise that RIs should ordinarily consider the normal physiology of the population and not be adjusted to take into account what may be considered pre-analytical complications. The impact of the pre-analytical effect of delayed sample transport on potassium levels and flagging rates is an example that was considered by stakeholders before final agreement that a higher upper reference limit (URL) of 5.2 mmol/L, rather than 5.0 mmol/L, be selected for the harmonised RI. The sample transport delay should ideally be eliminated. However, this may be difficult to achieve given the vast geography of Australia and New Zealand. The professional judgement consensus process for deriving common RIs is not perfect and there are limitations. As noted with potassium, harmonised intervals are usually wider than those for a singly-derived RI obtained on the same platform.

Traceability of Assays

Traceable analytes with Joint Committee for Traceability of Laboratory Medicine (JCTLM)-listed reference materials and ref erence measurement procedures are more likely to share common RIs. The reference method for bilirubin is the American Association of Clinical Chemistry (AACC) absorption spectrometry assay that is traceable to a certified reference material (Table 2).¹⁶ Assay of various enzymes is traceable to International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) reference measurement procedures based on kinetic spectrophotometry at 37 °C (Table 2). However, not all countries or laboratories may be using IFCC recommended methods as is the case in Australia where non-pyridoxal-5′-phosphate (P-5-P) aspartate transaminase (AST) and alanine transaminase (ALT) methods predominate. The AACB bias study using commutable samples demonstrated that for ALT the median bias in samples with concentrations <20 U/L was 44% (interquartile range (IQR) 39%, 59%), 20–50 U/L 32% (IQR 31%, 33%), and >50 U/L, 7% (IQR 5%, 11%) between methods employing P-5-P and those not using P-5-P. For AST the bias was less pronounced; the median bias in samples with concentrations <30 U/L was 19% (IQR 17%, 20%) and those samples with concentrations >30 U/L 4% (IQR 1%, 15%). Results of a brief comparison of two laboratories employing Abbott Diagnostics and Ortho Clinical Diagnostics (OCD) ALT P-5-P assays were variable but showed differences of up to 60% in concentrations <50 U/L, but comparable results when the concentrations were >50 U/L. This bias difference between P-5-P methods was also seen in a sub-study of three OCD and three Abbott laboratories using RCPAQAP material. The AACB has deemed that a harmonised RI with non-P-5-P methods is better than no harmonised RI. A future goal may be for Australian laboratories to use IFCC methods for transaminase analysis.^3,4

Table 2.

Traceability of enzymes and bilirubin according to JCTLM listings.

Analyte	Reference material	Reference measurement method/procedure
Alanine aminotransferase (ALT)	ERM-AD454/IFCC	Kinetic spectrophotometry; IFCC reference measurement procedure (37 °C) for ALT
Aspartate aminotransferase (AST)	ERM-AD457/IFCC	Kinetic spectrophotometry; IFCC reference measurement procedure (37 °C) for AST
Bilirubin	SRM 916a	Absorption spectrometry; AACC reference method for bilirubin 14
Creatine kinase (CK)	ERM-AD455/IFCC	Kinetic spectrophotometry; IFCC reference measurement procedure (37 °C) for CK
γ-glutamyl transferase (GGT)	ERM-AD452/IFCC	Kinetic spectrophotometry; IFCC reference measurement procedure (37 °C) for GGT
Lipase	N/A	N/A

N/A, not available; ERM, European Reference Material; IFCC, International Federation of Clinical Chemistry and Laboratory Medicine; JCTLM, Joint Committee for Traceability in Laboratory Medicine; SRM, National Institute of Standards and Technology Standard Reference Material

Recommended Harmonised Reference Intervals

An initial group of 12 harmonised RIs was endorsed by the AACB and RCPA in 2014. They were: sodium, potassium, chloride, bicarbonate, creatinine, calcium and adjusted calcium, phosphate, magnesium, lactate dehydrogenase (L-P, IFCC), alkaline phosphatase and total protein. From harmonisation meetings held in 2015 and national polling, where stakeholders were asked questions such as ‘Do they agree with the recommendations?’ and ‘Was further information required?’, a second group of six harmonised RIs has been proposed for endorsement. They are: total bilirubin, creatine kinase (CK), ALT (no P-5-P), AST (no P-5-P), gamma glutamyl transferase (GGT) and lipase. A commentary on these RI follows and the agreed RIs may be seen in Table 1.

Total Bilirubin

There are method differences and sex-related differences for total bilirubin but these are unlikely to be of clinical significance. The bias analysis demonstrated median bias of 1.2% between methods at a concentration of 15 µmol/L (IQR −19%, 12%). These bias results are similar to those obtained using the RCPAQAP material which showed a median bias of 3% at 28 µmol/L (IQR −8%, 6%). Due to the varying limits of detection (LoD) between assays, some laboratories do not report results to small concentrations. Stakeholder consensus was that there should be a lower limit concentration rather than ‘less than’. The recommended harmonised RI is 1–20 µmol/L. The difference between method LoD and the lack of evidence associated with the clinical significance of results below the lower reference limit (LRL) supported the LRL consensus agreement.

Creatine Kinase

The harmonised CK RIs are based on the current Sonic Health RIs and are consistent with the Aussie Normals study. In females the proposed RI is lower than for the Aussie Normals and may reflect the number of females in the younger age groups who identified as being more fitness conscious. The numbers in the Aussie Normals were too small to analyse this hypothesis. Data from both Sonic Health and the Aussie Normals show gender differences. These differences are seen in Figure 1. The AACB bias study and the RCPAQAP liquid serum chemistry program demonstrate that there is no impediment to harmonised RIs due to bias. The AACB/RCPAQAP data are supported by published studies for methods traceable to the IFCC reference method.¹⁷ All current major suppliers claim traceability to the IFCC reference method. Figure 2A shows the RCPAQAP liquid serum chemistry profile and that all methods have acceptable bias for a normal CK concentration. In Figure 3, showing the median and IQR for the Aussie Normals by gender and three age ranges, we can see that the upper IQR for males <60 y is much higher than that for subjects >60 y.¹² This difference is not seen in females. The difference in males is also highlighted by the flagging rates which are much higher in the younger age groups, particularly the <30 y group. The data are shown in Table 3. The recommended harmonised RIs are: males: 18–60 y: 45–250 U/L; >60 y: 40–200 U/L; and females: 30–150 U/L.

Figure 1.

Sonic Health age and gender reference intervals (RI) for creatine kinase (CK in U/L). The horizontal bar in the figure shows the median concentration for each RI. (courtesy K Sikaris, presented by G Jones as part of the 2015 AACB Harmonisation workshop update, 20/8/2015)

Figure 2.

RCPAQAP liquid serum chemistry profile by analyser type. 2A: creatine kinase; 2B: GGT; 2C: lipase; 2D: lipase profile by harmonised analyser type (excluding Ortho Vitros and Siemens Dimension). The x-axis represents the analyser type and the y-axis the assay concentration. The outer unbroken black lines represent the Allowable Limits of Error (ALE) as determined by the RCPAQAP; the inner dotted lines represent 1/10^th of the accepted reference interval around that target concentration; the grey unbroken line represents the imprecision around the average analyser concentration. (courtesy G Jones as part of the 2015 AACB Harmonisation workshop update, 20/8/2015)

Figure 3.

Creatine kinase (CK) median and interquartile concentration ranges of Aussie Normals subjects for three age ranges.

Table 3.

Flagging rates for creatine kinase (CK) in males (courtesy K. Sikaris).

Age (y)	<200 U/L	<250 U/L
20–29	37%	31%
4–49	27%	17%
60–69	19%	13%

Alanine Transaminase

A greater number of laboratories around Australia and New Zealand are currently using ALT methods that are not P-5-P activated. The recommendations by the stakeholders were:

• use individuals with a healthy BMI to set RIs

• a non-P-5-P recommended RI (which can be used as an anchor for a P-5-P method RI)

• consensus is for a lower numerical limit (i.e. ‘5–x’) rather than < (i.e. ‘<40’); 5 is proposed and this is supported by data from the Aussie Normals, with only one patient flagging

• median bias between methods not employing P-5-P for concentrations <40 U/L was 1.0% (IQR −6%, 6%).

The recommended gender-specific ALT RIs are: 5–40 U/L for males; 5–35 U/L for females, in normal BMI patients. Like bilirubin, the lack of evidence associated with the clinical significance of results below the LRL and the very low flagging rate supported the LRL consensus agreement.

Aspartate Transaminase

Like ALT, a greater number of laboratories around Australia and New Zealand are currently using AST methods that are not P-5-P activated. The recommendations by the stakeholders were:

• a non–P-5-P recommended RI (which can be used as an anchor for a P-5-P method RI)

• consensus for lower limit as a number rather than ‘<’

• median bias between methods not employing P-5-P for concentrations <40 U/L was <1.0% (IQR −3%, 4%).

The recommended gender-specific AST RIs are: 5–35 U/L for males; 5–30 U/L for females, in normal BMI patients. The rationale for the consensus recommendation of the LRL was the same as for ALT.

Gamma Glutamyl Transferase

The bias study conducted by the AACB shows a spread of GGT results between the different analyser method groups; however only the Beckman DxC shows bias that exceeds the RCPAQAP allowable limits of error (ALE) (Figure 4). The study showed that the bias shown by the Beckman DxC exceeded the lower limit of the RCPAQAP ALE at concentrations >20 U/L. This is also demonstrated in the RCPA liquid serum chemistry QAP (Figure 2B). External studies show acceptable between-method performance provided there is traceability to the IFCC method which is JCTLM-listed.¹⁷ The data suggest gender-specific RIs and, ideally, health-associated limits. The Aussie Normals study demonstrated BMI differences;¹² however it is difficult to adopt RIs in association with BMI at this stage as BMI is not routinely provided to the laboratory meaning only gender-associated limits are currently feasible. Most data sets show male activities are greater than for females and these GGT values increase with age in the adult population. Age and gender differences can be seen in the Aussie Normals data shown in Figure 5.

Figure 4.

Gamma glutamyl transferase (GGT) bias analysis for 8 analyser types. The RCPAQAP allowable limits of error (ALE) are indicated by the dotted lines.

Figure 5.

Aussie Normals gamma glutamyl transferase (GGT) distribution by age and gender. Regression analysis trend lines are shown for males and females. Females: open circles, trend line dashed; males: diamonds, trend line solid.

Based on the RCPAQAP 2014 survey, a GGT of 60 U/L is the most common URL for males (age 40 y), which is similar to the Aussie Normals data for males <35 y. When a BMI of <25 kg/m² is considered in a small cohort of subjects aged <45 y in the Aussie Normals study, the male URL was around 40 U/L (n=44) and for females around 35 U/L (n=46).¹² The Sonic Heath data provides a URL of 50 U/L for males and 35 U/L for females, as does the New Zealand South Island Quality Assurance Group (SIQAG). Based on these data, the recommended harmonised RIs for GGT are: males: 5–50 U/L; and females: 5–35 U/L. The lack of clear clinical evidence of the significance of GGT concentrations below 5 U/L supported the recommendation for the LRL. Flagging rates at these gender-specific URLs are seen in Figure 6, which shows that the flagging rates are >20% for women and >15% for men over the age of 50 y. These flagging rates are consistent with data from the Aussie Normals study which showed age-specific RIs with an increase of between 14% for males >35 y and 19% for males ≥75 y when compared with men <35 y.¹² Similar differences were seen with women. GGT concentrations were 15% higher for males and 25% for females when comparing subjects with a BMI of <25 kg/m² and those subjects with a BMI of >25 kg/m². Because of the difficulties with health-associated and age-related limits, the inclusion of the Beckman DxC assays into harmonised RIs seems a reasonable proposal.

Figure 6.

Flagging rates by age and gender for gamma glutamyl transferase (GGT) for females >35 U/L (n=20,765) and males >50 U/L (n=8,895). (courtesy G Jones as part of the 2015 AACB Harmonisation workshop update, 20/8/2015)

Lipase

Both the bias study and the RCPAQAP liquid serum chemistry program showed three discrete method groups: Siemens Dimension, Ortho Vitros and a ‘harmonised’ group. The data from the RCPAQAP liquid serum chemistry program are shown in Figure 2C. Assays in the harmonised group do not show any significant bias in either the AACB bias study or the RCPAQAP liquid serum chemistry program that would prevent adoption of common RIs. Figure 2D shows the RCPAQAP liquid serum chemistry analysis from the harmonised group.

Most manufacturers’ lipase assays use coupled enzyme spectrophotometric methods that use as substrate either 1,2-diglyceride, 1-oleyl-2,3-diacetoyl glycerol, or 1,2-O-dilaurylrac-glycero-3-glutaric acid-[4-methylo-resorufin]-ester. At this time there is no lipase reference method listed on the JCTLM database or recommended by the IFCC. Whilst specific post-analytical correction factors are not provided, it is noted that there are linear relationships between the harmonised assay group and the Vitros and Dimension. Approximate relationships are as follows:

• Harmonised = Dimension × 0.21 − 0.6

• Harmonised = Vitros × 0.27 +12

No significant age and gender differences are seen in either the Aussie Normals or Sonic Health data albeit that the URL for the <70 y age subgroup in the Aussie Normals is slightly lower. The Aussie Normals overall RI (<55 U/L) is also slightly lower than that of the Sonic Health group (<60 U/L) and may reflect the difference in this <70 y age group. Based on RCPAQAP Reference Intervals survey 2014, 60 U/L is the most common URL once Vitros and Dimension are excluded. The RI recommended for adult serum lipase (both sexes), excluding Siemens Dimension and Ortho Vitros, is 10–60 U/L. Flagging rates at 60 U/L are seen in Figure 7. It is suggested, as a patient safety measure, that laboratories using the Siemens Dimension or OCD assays should note on their reports that the RI is not commutable with alternate assays.

Figure 7.

Flagging rates by age and gender for lipase at >60 U/L. (courtesy G Jones as part of the 2015 AACB Harmonisation workshop update, 20/8/2015)

Conclusions

The AACB Committee for Common Reference Intervals has now recommended harmonised RIs for 18 common chemistry analytes which are performed in most routine clinical chemistry laboratories. These RIs are recommended based on professional opinion and consensus with RI limits compared from both a priori and a posteriori analysis and clinical relevance and flagging rates taken into consideration. Future harmonisation workshops have been scheduled where discussions relating to future analytes as candidates for harmonisation will be discussed as well as those currently under consideration such as urea, albumin, globulin, ferritin, TSH, fT4, hGH, IGF1 and others. Also being discussed in 2016 are pregnancy-related RIs and their flagging rates.