Article Text
Abstract
Aim: Setting up reference intervals from the local service populations is one of the major responsibilities of clinical laboratories. Yet, this task is difficult to achieve because it is costly and time consuming when compared with validating reference intervals from assay manufacturers.
Methods: Following the recommendations of the International Federation of Clinical Chemistry, healthy local Chinese adults were recruited to set up reference intervals for common serum analytes. Statistical methods recommended by the National Committee for Clinical Laboratory Standards were used for defining the reference limits.
Results: Data from 335 subjects were analysed. The reference intervals set up were broadly similar to those provided by the assay manufacturer, except for sodium and potassium. Glomerular filtration rate was estimated by the modification of diet in renal disease equation, with or without modification for Chinese. Body mass index had a significant impact on serum urate and alanine aminotransferase levels.
Conclusion: Reference intervals of common serum analytes have been set up for the local Chinese population. A good example of quality laboratory service has also been set up to provide clinicians with reliable reference intervals that they can confidently rely on for the diagnosis and management of patients.
Statistics from Altmetric.com
The International Federation of Clinical Chemistry (IFCC) defines “reference value” (RV) as the value obtained by observation or measurement of a particular type of quantity on a reference individual in which the individual is selected using defined criteria.1 The distribution of RVs contributes the “reference distribution” in which the reference limits (RLs) are derived. Between and including the two RLs is the reference interval (RI).1 While setting up the RI is one of the major responsibilities of clinical laboratories, it is a time consuming and costly exercise. Validating RIs provided by the assay manufacturers is relatively simple and economical, and is the most commonly adopted alternative. Nevertheless, some assay manufacturers may be quoting RIs reported in the literature using different measuring principles. Although they may have tried to set up their own assay-specific RIs, manufacturers usually provide little information on the details of their reference subjects and their sampling procedures.
As with many other clinical laboratories, our laboratory has validated and adopted the RIs of common serum analytes provided by the assay manufacturer. From time to time, we receive consultations from clinicians concerning patients with borderline high or low laboratory results. Whether these observations are related to the choice of RIs or whether they are genuine pathological findings can be difficult to differentiate. It has been almost two decades since the Expert Panel on Theory of Reference Values of the IFCC published its series of recommendations on the procedures for developing RIs.1–6 Despite these comprehensive recommendations being available, medical literature reporting RIs of common serum analytes for the Chinese population is still lacking. Since appropriate RIs are essential for quality patient care, and renal and liver function tests, bone profile and urate are the most commonly requested serum tests in our laboratory, we felt an urge to set up RIs of these analytes for our local Hong Kong Chinese population.
METHODS
Selection of subjects
Staff of the Department of Pathology, Queen Elizabeth Hospital, Hong Kong, and their families, relatives and friends, were invited to participate by completing a health questionnaire that was designed based on IFCC and Scandinavian Committee on Reference Values recommendations.2 7 The selection criteria were blind to all participants except for the authors, and the returned questionnaires were screened by two of the authors. Inclusion and exclusion criteria are listed in table 1.
Subjects with fasting glucose (FG) ⩾7.0 mmol/l were excluded from result analyses because of the possibility that they maybe suffering from diabetes mellitus. Similarly, subjects with total cholesterol (TC) ⩾6.2 mmol/l, and/or fasting triglyceride (TG) ⩾2.3 mmol/l, were excluded from result analyses based on the cut-offs for high TC and TG defined in the Third Report of the Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults.9
Blood sampling procedures
Written informed consent was obtained from all selected subjects. They were instructed to consume their usual diet and to avoid strenuous physical exercise such as hiking and marathons for at least 3 days, and refrain from alcoholic beverages for at least 1 day before blood taking. All subjects were fasted for at least 8 h. Before blood was taken, all subjects were requested to sit for 30 min at 23°C. Blood was taken between 08:30 and 11:30 in March, 2005, while the subjects were in sitting position. Venous blood was collected from their median cubital veins into vacutainers containing clot activator and sodium fluoride (Vacuette; Greiner Bio-one GmbH, Kremsmuenster, Austria) via 21 gauge needles without application of tourniquet to avoid venous stasis. Forearm exercise or fist clenching was prohibited.
Biochemical analyses
All specimens were centrifuged at 3500 g at 18°C for 7 min within 2 h of collection (Heraeus Megafuge 1.0R; DJB Labcare, Newport Pagnell, Buckinghamshire, UK). Serum was analysed immediately for sodium, potassium, urea, creatinine, total protein, albumin, total bilirubin, alkaline phosphatase (ALP), alanine aminotransferase (ALT), total calcium, phosphate, urate, total cholesterol and triglyceride. The globulin level was calculated by subtracting albumin from total protein concentration. Estimated glomerular filtration rate (eGFR) was calculated by using the four-variable modification of diet in renal disease (MDRD) equation10 as well as the recently published modified MDRD equation for Chinese.11 Adjusted calcium was calculated by using the correction formula published by Payne et al.12 Sodium fluoride plasma was used for measuring the FG level. All analyses were performed using Modular Analytics (Roche Diagnostics, Mannheim, Germany) with the manufacturer’s reagents (table 2). Commercial quality control materials at concentrations within and outside the RI were analysed at least once every 4 h for each analyte (Lyphochek, Unassayed Chemistry Control (Human) Levels 1&2; Bio-Rad Laboratories, Hercules, CA, USA). The performance of the assays was reflected by the standard deviation and coefficient of variation derived from these observed means, and it was interpreted based on Westgard rules.
Statistical analyses
Statistical methods recommended by the National Committee for Clinical Laboratory Standards (NCCLS) were used to detect outliers and to define RIs.13 Briefly, the presence of outliers was tested by using the Dixon’s test14:
QexpL = (X2−X1)/(XN−X1)
and QexpU = (XN−XN−1)/(XN−X1)
where QexpL and QexpU are the experimental critical values at the lower and upper limits, respectively, N is the number of results arranged in ascending order, X1 is the smallest value and XN is the largest value among all the results of a particular analyte. The result was rejected when the corresponding Qexp was greater than the critical value; the critical value was chosen as 1/3(XN−X1). After excluding outliers, the lower and upper RLs were defined as the concentrations of analytes at the 2.5th and 97.5th percentiles, respectively. The 90% confidence intervals of the RLs were also calculated.
Necessity to partition the data into sex-specific RVs was tested by the standard normal deviate test:
z = (X1−X2)/[(S12/n1)+(S22/n2)]½
where z is the calculated critical value, X1 and X2 are the observed means of the male and female RVs, respectively, S12 and S22 are the observed variances, and n1 and n2 are the number of RVs in each group, respectively. The calculated z was compared with the critical value z*, which was calculated as
z* = 3[(n1+n2)/240]½
Partitioning the RVs into sex-specific RIs was necessary when z was greater than z*. Pearson correlation, which was calculated by SPSS for Windows 11.0.1, was used to test for the correlation between BMI and each analyte. All other calculations were computed using Microsoft Excel 2002.
RESULTS
A total of 166 male and 226 female subjects were recruited for participation in the project. Fifty-seven subjects (28 males) with FG, TG and/or TC results above the cut-off values as stated above were excluded and were referred for medical advice by the authors. As a result, 138 male and 197 female subjects were included in the final result analyses, with median ages of 40 (range 21–75 years) and 41 years (range 21–81 years), respectively. Twenty-two subjects (eight males) were ⩾60 years old. Twenty-eight female subjects were post-menopausal, defined as the absence of menstruation for at least 1 year.15 The median body mass index (BMI) was 22.9 kg/m2 in male and 21.3 kg/m2 in female. One serum sample exceeded the haemolysis index of potassium and the test was not performed. An outlier was detected in the ALP and ALT tests and the results were excluded. The 2.5th and 97.5th percentiles of all the serum analytes tested and their 90% confidence intervals are shown in table 2. The critical value z* for all analytes was 3.54. Necessity to partition RVs into sex-specific subclasses was demonstrated statistically in all the analytes except for total protein and adjusted calcium.
eGFR values at 2.5th and 97.5th percentiles were defined and are shown in table 3.
The difference in the number of subjects in each class of eGFR as calculated by the MDRD equation with or without modification for Chinese is shown in table 4.
The degree of correlation between BMI and each of the tested serum analyte is shown in table 5.
Significant correlations were demonstrated between BMI and urate and ALT in both sexes, compatible with those reported in other ethnic groups.16–19
Take-home messages
Setting up reference intervals from the local service populations is one of the major responsibilities of clinical laboratories.
Appropriate reference intervals are essential for quality patient care.
Serum creatinine remains a valuable renal marker in our locality until we have set up our own or validated any equation for the estimation of glomerular filtration rate.
Around 7% of the local population might have dyslipidaemia. A regular cholesterol surveillance programme should be considered by the government of Hong Kong.
DISCUSSION
According to the statistics from the Census and Statistics Department of the Hong Kong Special Administrative Region Government in 2001, nearly 95% of the population (equivalent to 6.3 million) in Hong Kong was of Chinese origin.20 We believe this exercise is worthwhile because it is one of the ways to provide quality laboratory service, and appropriate RIs have an ultimate impact on patient management.
Most of the RIs we set up were similar to those provided by the assay manufacturer. Yet, we observed notable differences in our lower RL of sodium and upper RL of potassium when compared with those provided by the assay manufacturer, who quoted the RIs from reference 21. Observations similar to ours have been reported previously using two other analytical principles,22 23 suggesting that these phenomena are genuine rather than method specific. The relatively high-sodium and low-potassium Chinese diet,24–26 as well as ethnic-related differences in the renal handling of these two electrolytes,27 28 may account for these differences. Sodium and potassium are analytes with little individuality.29 Correctly established population-based RIs are valuable references guiding interpretation of individual measurements. Our findings illustrated that adopting RIs from the literature or from the manufacturers without considering the ethnicity or other biological variables of the reference population might result in under- or overdiagnosis of certain conditions.
The RIs of creatinine of both sexes found in our study were much lower compared with those quoted in other references.21 30 This observation again supports our notion of setting up RIs for our own population. Serum creatinine is undoubtedly far from ideal as a marker of GFR31 and that eGFR is recommended by various authorities.32–34 When eGFR was derived from the original four-variable MDRD equation, 69% of the subjects had eGFR ⩾90 ml/min/1.73 m2. The proportion of subjects with this range of renal function increased to 94% when the equation with modification for Chinese was used. Whether this observed discrepancy was due to underestimation of renal function by the original equation which was corrected by the modified equation was to be validated in our local population. In the meantime, it appears premature for us to rely solely on eGFR derived from any of these equations to reflect renal function. Serum creatinine remains a valuable renal marker in our locality until we have set up our own or validated any equation for the estimation of GFR.
In addition to setting up RIs of our own population, our findings also gave us a crude estimate of the prevalence of abnormal lipid and glycaemic status in our reference population. From our raw data (not shown), the prevalences of hypercholesterolaemia, hypertriglyceridaemia and FG ⩾7.0 mmol/l in our reference population were 7.7%, 3.5%, and 1.3%, respectively. Dyslipidaemia and diabetes mellitus are major risk factors of cardiovascular diseases, and these diseases accounted for 27.2% of all deaths in Hong Kong in 2004.20 The Third Report of the National Cholesterol Education Program Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults recommends that all adults should have a fasting lipoprotein profile performed once every 5 years.9 Similar recommendations are not available locally. The relatively high prevalence of hypercholesterolaemia in our apparently healthy non-obese reference individuals is positive evidence to support the promotion of regular cholesterol surveillance programme in Hong Kong.
The limitation of our study was the lack of reference subjects ⩾60 years old and post-menopausal women. As a result, we could not collect enough data to test for the effects of age and menopausal status on the RIs. Moreover, due to the limitation of our laboratory information system, which could only provide integers for sodium and figures up to one decimal place for potassium, the effect of these significant figures on the final RIs of these electrolytes might not be negligible. Furthermore, our subjects were not tested for their hepatitis B status. The prevalence of hepatitis B surface antigen carrier in our population ranges from 0.0 to 11.5%,35 and only 44% of the patients in primary care are aware of their hepatitis B status.36 Although known hepatitis B carriers were excluded from participation, those who did not know their hepatitis B status may be recruited and may lead to bias in the liver function test results. Moreover, since our assays were not calibrated with international standard materials, the RIs we set up are only applicable when identical analyser and reagents are used.
In conclusion, we have set up RIs of common serum analytes of our own population. We have also set up a good foundation for quality laboratory service where clinicians can confidently rely on for the diagnosis and management of patients.
Acknowledgments
We would like to thank all the subjects and all the staff members of the Department of Pathology, Queen Elizabeth Hospital, Hong Kong, for their participation in this project. Special thanks to Ms Heidi Iu, Ms Ala Lee, Ms Angela Kwan, Mr WY Ho, Ms Helen Wong and Ms FY Chan for their assistance in organising and conducting this project.
REFERENCES
Footnotes
Competing interests: None.