Nationwide Multicenter Reference Interval Study for 28 Common Biochemical Analytes in China (2024)

A nationwide multicenter study was conducted in the China to explore sources of variation of reference values and establish reference intervals for 28 common biochemical analytes, as a part of the International Federation of Clinical Chemistry and Laboratory Medicine, Committee on Reference Intervals and Decision Limits (IFCC/C-RIDL) global study on reference values.

A total of 3148 apparently healthy volunteers were recruited in 6 cities covering a wide area in China. Blood samples were tested in 2 central laboratories using Beckman Coulter AU5800 chemistry analyzers. Certified reference materials and value-assigned serum panel were used for standardization of test results. Multiple regression analysis was performed to explore sources of variation. Need for partition of reference intervals was evaluated based on 3-level nested ANOVA. After secondary exclusion using the latent abnormal values exclusion method, reference intervals were derived by a parametric method using the modified Box–Cox formula.

Test results of 20 analytes were made traceable to reference measurement procedures. By the ANOVA, significant sex-related and age-related differences were observed in 12 and 12 analytes, respectively. A small regional difference was observed in the results for albumin, glucose, and sodium. Multiple regression analysis revealed BMI-related changes in results of 9 analytes for man and 6 for woman. Reference intervals of 28 analytes were computed with 17 analytes partitioned by sex and/or age.

In conclusion, reference intervals of 28 common chemistry analytes applicable to Chinese Han population were established by use of the latest methodology. Reference intervals of 20 analytes traceable to reference measurement procedures can be used as common reference intervals, whereas others can be used as the assay system-specific reference intervals in China.

Ethics

This study was approved by the Ethics Committee of Peking Union Medical College Hospital, Peking Union Medical College & Chinese Academy of Medical Science. Written informed consents were obtained from all participants prior to the study.

Selection of Reference Population

Laboratory Analysis

Data Analyses

Multiple regression analysis (MRA) was performed to identify sources of variation (SVs) possibly influencing the test results,¹⁰ including region, sex, age, BMI, levels of cigarette smoking, daily alcohol consumption, and regular physical exercise with respective levels categorized into 4, 5, and 8 grades as follows: none, ≤250, 250 to 500, 500 < pack × year; none, ≤12.5, 12.5 to 25, 25 to 50, 50< g ethanol/day; none, 1 to 7 days/week. In the analysis of regional differences by use of dummy variables, Beijing was set as the reference region. A given explanatory variable was considered to be of practical importance when its standardized partial regression coefficient, which corresponds to the partial correlation coefficient (r_p), was >0.15, which corresponds to P ≈ 0.0 with the large data size around 1500.

Standard deviations (SD) attributable to 3 major SVs¹⁰ of test results: between-region (SD-reg), between-sex (SD-sex), between-age (SD-age), together with net between-individual (SD-btw-ind) were computed by 3-level nested ANOVA. The relative magnitude of each SD was expressed as the ratio of SD (SDR) over SD-btw-ind: SDR_sex, SDR_age, and SDR_reg for SD due to sex, age, and region, respectively. An SDR > 0.3 was regarded as a guide to consider partitioning reference values by the SV.^3,10

The latent abnormal values exclusion (LAVE)¹⁰ method was applied as a method for secondary exclusion with Alb, UA, Glu, TG, LDL-C, HDL-C, AST, ALT, GGT, LDH, CK, and CRP set as the reference test items for judging the need for exclusion. Modified Box–Cox power transformation formula was used to transform reference values into the Gaussian distribution.¹⁰ The RI was first computed as the mean ± 1.96 × SD of the transformed reference values selected by the LAVE method, and then reverse transformed to the original scale. The 90% confidence intervals (CI) of the lower and upper limits of the RIs (LLs and ULs) were calculated by use of the bootstrap method through random resampling of the same dataset for 100 times. Making use of the procedure, both LLs and ULs were optimized by taking their averages of the repeatedly derived values.

Profile of the Population

Age and sex distributions of the participants in the 6 cities are shown in Table ​Table1.1. The average age and SD of participants were 41.8 ± 14.1 (n = 1509) years for man and 42.2 ± 13.8 (n = 1639) years for woman. As a whole, male female ratio was almost equal (52.1% vs 47.9%). The BMI was 23.6 ± 2.9 (n = 1507) for man and 22.2 ± 2.8 (1639) for woman. 31.8% of the male participants were smokers whereas females seldom smoke. About 35% of participants had different levels of exercise and the ratios of those who do regular exercise were almost equal for both sexes. A total of 26.2% of the male participants had different levels of drinking whereas females seldom drink (Figure ​(Figure11).

Traceability

The serum panel results of Beijing were compared to the assigned values and the results of Chongqing. The results of the comparison is shown in Supplemental Figure 1; as correlation matrix graphs made from assigned values, and test results from Beijing and Chongqing for each analyte. Urea, UA, TBil, CRE, Glu, TG, Na, K, Cl, Ca, IP, CK, Tf, and TTR were regarded as traceable to the RMP based on the comparison of Beijing's results with the assigned values.

The results of Chongqing were converted to those of Beijing by using the major axis regression line when the coefficient of variation (CV) of the regression line slope, CV(b), was below the critical value of 5.5%.⁴ The results for the analyte were judged convertible and merged except for those of Na, Cl, Mg, and TBil. The failure of the latter 4 analytes was attributed to the narrow coverage of the test ranges by the serum panel. However, the SD ratio (SDR) representing between-laboratory differences computed by use of the 1-way analysis of variance³ were 0.24, 0.09, 0.44, 0.14, respectively for the analytes, and thus their results except Mg were merged without conversion (Supplemental Table 2; Mg test results from Chongqing were not included in derivation of the RI.

The major axis linear regression between the assigned values and merged values of Chongqing to Beijing was used to assess the need for recalibration of test results.⁴ When difference in the lower or upper limit of the RI before and after recalibration was >1/4 of SD of the original RI, recalibration of test results was performed by using regression coefficients (intercepts and slopes). This criterion corresponds to Frasers’ allowable limit of bias. By this criterion, the recalibration based on the assigned values was performed for TP, GGT, TC, HDL-C, and LDL-C.

Regional Variations of Reference Values

The MRA results in Table ​Table22 showed that test results for Glu from Beijing were higher than those from other cities except Chongqing with r_p < −0.15. Dalian showed a largest negative shift. We could not identify the reason, but the time between sampling and serum separation was carefully controlled to be identical. Both males and females of Hangzhou had a significantly high level of Na, and Guangzhou had a high level of UA, compared to Beijing. Males of Urumqi had a significantly high level of ALT than other 5 cities. Males of Chongqing and Urumqi both had much higher GGT than other cities. Though MRA results (r_p) of LIP, Mg, K, and Cl were above 0.15, we considered the actual difference are very small and not of practical importance (Table ​(Table33).

Age- and Sex-Related Changes of Results

Age-related changes were analyzed by combination of MRA and 3-level nested ANOVA (Table ​(Table4),4), and 1-dimensional scattergrams (Figure ​(Figure3).3). We observed that 5 analytes changed significantly with age in both sexes: Alb, LDL-C, TC, TG, and Urea. Alb steadily decreased with age, especially in males, and others increased. ALP, AMY, CK, Fe, GGT, Glu, LDH, and TG increased with age only in females. In males, the level of lipids increased mainly until the age of 40, whereas in female, the increase occurred more prominently after 40 years of age. The level of SDRs for sex (SDR_sex) was significantly high (≥0.30) for 13 analytes: ALT, AST, CK, CRE, Fe, GGT, HDL-C, IP, TBil, TG, TTR, UA, and Urea.

Other Sources of Variations of the Results

MRA results (Table ​(Table2)2) revealed that test results for 7 analytes (UA, TG, HDL, ALT, GGT, CK, and AMY) in females and 10 analytes (UA, TC, TG, HDL-C, LDL-C, ALT, GGT, CK, AMY, TTR) in males were related to BMI. Of all the related analytes, only HDL-C and AMY decreased with the increase of BMI. Alcohol-related changes for GGT and TTR, and smoking-related change in TP were observed in males. No appreciable changes in test results in association with exercise level were noted in any analyte in both sexes.

Reference Intervals

RIs were derived both parametrically (P) and nonparametrically (NP) with LAVE (−), LAVE (allowing 1 abnormal value), and LAVE (allowing no abnormal value). Most analytes had generally identical RIs determined by both P and NP methods, except for TBil, Glu, TG, AST, ALT, GGT, CK, and LIP, the ULs of which were higher derived by NP than by the P method. We observed that the ULs of some analytes decreased significantly when stricter secondary exclusion criteria were applied. It was seen for Glu, TG, TC, HDL-C, LDL-C, AST, ALT, LDH, ALP, GGT, CK, and LIP in female. And in males, those analytes were UA, TG, TC, AST, ALT, LDH, CK, GGT, and LIP. Examples of the comparison of RIs for both males and females by P and NP methods with or without LAVE method are shown in Figure ​Figure2.2. Those for other analytes, see Supplemental Figure 2 and Supplemental Figure 3. As a general rule, we derived the RIs by use of the P method without applying the LAVE procedure for analytes whose RIs did not change much by the procedure, such as TP, TBil, Na, K, and Ca. On the other hand, for those analytes which are sensitive to the LAVE procedure, such as TG, AST, ALT, GGT, and CK, we adopted the RIs by use of P method with LAVE (abn = 1). We did not adopt RIs by LAVE (abn = 0), because it led to narrow RIs by too strict exclusion of individuals with latent diseases. The final RIs of this study and comparison with other studies are listed in Table ​Table55.

Traceability

Judging from test results of the serum panel, Urea, TBil, UA, CRE, Glu, TG, Na, K, Cl, Ca, IP, CK, Tf, and TTR were regarded as traceable to the reference materials. The results of TP, GGT, TC, HDL-C, and LDL-C were recalibrated by use of the major axis linear regression⁴ based on test results for the serum panel which has assigned values for those analytes.³ Therefore, the RIs derived for them in this study are applicable to any Chinese laboratory as long as the assays are standardized and regional differences in test results do not exist. The serum panel results for ALT and AST were lower than the assigned values, it is considered to be lack of Pyridoxal-5-phosphate (P5P) in the reagent for the 2 items. As the values of Alb, TBil, Mg, Fe, LDH, ALP, AMY, and LIP of the serum panel were not assigned by IFCC/C-RIDL in this study, the results of these analytes should be regarded as Beckman AU-specific RIs.

Exclusion of Latent Diseases

When determining the RI, the first thing is to define reference individual as appropriately as possible in consideration of many properties which constitute the status of healthiness by consensus among the investigators. We asked a series of questions and measured the anthropometric parameters including BMI and blood pressure to exclude those who possibly possessed pathological conditions. However, we may still include a lot of individuals who have latent diseases, especially those of high prevalence among apparently healthy individuals in China such as metabolic syndrome and diabetes mellitus.^11–13 The LAVE method, which depends on association of abnormal test results among test items affected by those diseases, is considered a good way to deal with such situation.^10,14,15 By application of the LAVE method with the stricter criteria,³ ULs derived were lowered in both P and NP methods. At the same time, the gaps of the ULs between P and NP methods became narrower. This is because more and more abnormal results from individuals with latent diseases were removed by the stricter LAVE criteria.

Sources of Variations

In conducting the study on RIs, another important aspect is to explore SVs of each test item. Age, sex, BMI, and region are considered the most important SVs. With the use of MRA and nested-ANOVA, we reconfirmed age- and sex-related variations in many of the common analytes. Though our multicenter study covered a wide area of China, no significant regional differences were discovered in most analytes, expect Na, UA, ALT, GGT, and Glu. Test results of Na were higher in Hangzhou, which may reflect higher consumption of salt in foods. The ULs of UA for both male and female from Guangzhou were much higher than other cities. It is consistent with the report that Guangzhou had a higher prevalence of hyperuricemia than other cities.¹⁶ Higher UL of ALT for Urumqi may be due to higher BMI compared to other areas. ULs of GGT for males in both Chongqing and Urumqi are high. In interpreting the deviation of their ULs, we have to note that GGT has a very skewed distribution and thus its 90% CI of the UL is much wider than other analytes when RIs were determined province by province using a small sample. Anyway, a higher daily consumption of alcohol may partially explain the high GGT in Chongqing. However, after adjustment of alcohol consumption by use of MRA, GGT is not so high (compared with Beijing) in Chongqing and Urumqi (Table ​(Table22).

The life habit such as smoking may be another important SV to affect the RIs. The relation of smoking, drinking, and exercise with RIs were analyzed. We confirmed alcohol-related changes for GGT, TTR, and smoking-related changes for TP in males.^17,18 No such relations were seen in females, which is obviously due to very low proportion of females with habits of smoking and drinking.

Once notable SVs were observed, it is important to determine its relative magnitude of variation associated with the SV in order to evaluate the necessity of developing separate RIs according to the source. Several methods, such as Harris–Boyd method,¹⁹ Lahti method,²⁰ and Ichihara method¹⁸ have been proposed for use. We adopted the Ichihara method in this study, which depends on 1-way ANOVA or 2 or multilevel nested ANOVA. It computes the SD ratio (SDR), which is the SD due to a given factor divided by the SD_BI, or SD due to between-individual variation. The guide value for considering the partition is SDR = 0.3, which corresponds to the allowable limit of bias (B_A) in analytical system:³ that is, the ratio of allowable bias is B_A/SD_TB ≤ 0.25, where SD_TB represents total biological SD (= SD_BI unadjusted for SD due to sex and age) and is generally larger than our SD_BI. Thus, the ratio was raised to 0.30 in consideration of the general difference between SD_TB and SD_BI. The ratio was regarded appropriate for most cases. However, actual scatter plots for comparing test results across provinces or for sex-specific age profiling of test results are of relevance in interpreting SDR and for deciding whether or not to partition reference values. For example, we observed a steady decline of Alb by increasing age with SDR_age = 0.55. However, actual difference of the RI between age-subgroups is slight. Similarly, the SDR_reg for K is 0.39, but actual between province differences are rather small and of no clinical significance. The same thing is true to TP, Na, and Cl. The common features of these special analytes are the narrow range of the RI for these test results, and thus SDR tends to be exaggerated by placing relatively smaller SD_BI (≈1/4th of the RI) in the denominator. Another example of need for caution in using SDR is the case of LDH. Although SDR_sex is 0.0, when SDR_age was computed for each sex separately, SDR_age is 0.51 for males and 0.11 for females. This implies that between-sex difference was masked due to the interaction of age on the analysis of sex-related change. Therefore, SDR_sex was computed after subgrouped by age at 50 years. Without partition at age of 50, 14 analytes (50%) have SDR_sex > 0.3, whereas 18 analytes (64%) have SDR_sex > 0.3 with age <50 and only 8 analytes (29%) have SDR_sex > 0.3 with age ≥50 (Table ​(Table44).

In fact, there are multiple test items which showed differential effect of sex on age-related change, as shown in Figure ​Figure3.3. In males, nutritional markers (TG, TC, LDL-C, ALT, AST, and GGT) increased from 20 to 40 years of age and unchanged or decreased thereafter. This mode of change reflects life style change with age, whereas in females, they increased from 30 to 60 with steeper changes between 45 and 55. The steeper part is interpreted as reflecting hormonal changes pecular to females. These observations point to the importance of computing SDR_sex and SDR_age in multiple ways and interpreting the values in reference to the sex- and age-related profiles.

In this study, the ULs of the RIs for TG, TC LDL-C, and UA are all higher than the clinical decision limits (CDLs),^21,22 although they became appreciably lowered by application of the LAVE method, which effectively reduced the influence of metabolic conditions. Further attempt of applying stricter exclusion criteria in the recruitment might match the RIs a little closer to the CDLs, but it makes a prior study of the RI impossible with difficulty of recruitment. Therefore, it is important to reduce the influence of latent disease, but we cannot make the ULs identical with the CDLs (Figure ​(Figure33).

In fact, CDLs were determined by consensus among experts of the diseases targeted by the CDLs for early detection and intervention (based on clinical experiences and epidemiological study), whereas the RIs are determined from individuals with as many healthy attributes as possible so that they serve as references for the healthy status, but they are not meant for distinguishing the healthy from the disease. Therefore, the RIs of these items with known CDLs should not be confused with the RIs. The CDLs should be primarily used for judging the need for preventing or diagnosing the diseases, but the RIs are still necessary for use in other clinical situations which are unrelated to the CDL-associated diseases.

Comparison of RIs With Other Studies

The RIs of CRE, TBil, K, AST, LDH, ALP, GGT, CK, and AMY showed great difference compared with those from the reagent inserts (Table ​(Table5).5). So, laboratories should evaluate the clinical applicability of the RIs from reagent inserts before use.

The RIs of TP, Alb, Urea, CRE, TBil, Na, K, Cl, Ca, IP, MG, Fe, AST, ALT, LDH, ALP, and GGT are close to those from Health occupation standards of China,^23–29 which were approved by the Ministry of Health P. R. China, whereas the RI of AMY is much lower than that from the standard, which may because that the results of AMY were not standardized in our study.

Compared to another multicenter study by Mu et al⁹ in China, which aimed at liver function tests, RIs of ALT for females, TBil, AST, and LDH are very close. ULs of ALT and GGT for males are a little higher in our study. This may be attributable to the different exclusion criteria. In Mu's study, those who consumed alcohol >30 g/day and/or with fatty liver disease had been excluded, whereas in our study, alcohol consumption and fatty liver disease were not considered in the recruitment because of planned use of the LAVE method. There were ∼2.3% of the participants who consumed >25 g/day in our study. But the UL of GGT for females in Mu's study (43U/L) is about 39% higher than that in our study, which is 31 U/L with LAVE (abn = 1). We get 43U/L by LAVE(−) method, which is the same as the one by Mu's study. From this finding, it can be said that the LAVE method is more appropriate than Mu's method when applied to those apparently healthy individuals.

The RI of TP differed much between Mu's and our studies. We found that the RIs from Turkey (66 − 82 g/L with Abbott Architect 8000),³⁰ Saudi (62−77 g/L with Abbott Architect 16000), Hongkong (66−80 g/L with Roche Modular)⁸ are close to ours (65−79 g/L) whereas that from Shuo Yang's study(68−86 g/L with Roche Modular)⁷ is close to Mu's(67.9–85.1 g/L for 4 cities and 70.1−90.1 g/L for another 2 cities).

The RI of Alb in our study is lower than Mu's study. They used 3 different assay systems and the RIs differed with each other. The final RI was adjusted according to the results from the 3 assay systems. Among the 3 systems, RI from AU system (described as Olympus in the article) is ∼42−52 g/L, which is very close to ours (41−52 g/L). So it is important to have specific RI for each assay system.

As compared to Turkey's and Saudi Arabia's³¹ multicenter study which used the same protocol with us, the RIs of TP, Alb, CRE, Glu, Na, K, Cl, Ca, IP, and Mg, Fe, LDH, ALP, GGT are close with each other. Among the 3 countries, Turkey has a highest lipids level, and Saudi Arabia has the lowest ALT level, which appear to be mainly due to difference in ethnicity and life style, and partly due to diffences in assay systems.

In conclusion, this is the largest multicenter RI study in China involving 28 commonly measured tests aimed at adult population, which also is a part of global multicenter collaborative project for derivation of RIs. Age and sex differences were observed in many items and it was important to decide the need for partitioning RIs based on combined analysis of nested-ANOVA and sex and age profile of reference values. No regional differences of practical importance were observed in any of the test items. The LAVE method was indispensable to derive the RIs more appropriately by minimizing the influence of metabolic disorders.

Supplemental Digital Content:

Supplemental Digital Content:

Supplemental Digital Content:

The authors express their sincere gratitude to Beckman Coulter (China) Co., Ltd, which generously provided all reagents required for testing. They are very grateful to Mr. Shogo Kimura and Mr. Reo Kawano of Yamaguchi University Graduate School of Medicine for their dedicated contribution for the data analyses and preparation of all the figures for this study.

Author contributions: Experiments conceiving and design: LQ, KI, MC, ML, ZT, S L, LW. Recruitment of volunteers, sample collection and processing: JH, MC, ML, ZT, SL, LW. Laboratory analysis: LX, XC, XQ, PL, LH, SY, MC. Data analysis: KI, LX, LQ. Manuscript preparation and editing: LX, KI. Project organization: LQ.

Abbreviations: Alb = albumin, ALP = alkaline phosphatase, ALT = alanine aminotransferase, AMY = amylase, AST = aspartate aminotransferase, BMI = body mass index, Ca = calcium, CDL = clinical decision limit, CI = confidence intervals, CK = creatine kinase, Cl = chloride, CLSI = Clinical and Laboratory Standards Institute, CRE = creatinine, C-RIDL = Committee on Reference Intervals and Decision Limits, CV(b) = CV of the regression slope b, CV = coefficient of variation, Fe = ion, GGT = γ-glutamyltransferase, Glu = glucose, HDL-C = high-density lipoprotein-cholesterol, IFCC = International Federation of Clinical Chemistry, IP = inorganic phosphate, K = potassium, LAVE = latent abnormal values exclusion, LDH = lactate dehydrogenase, LDL-C = low-density lipoprotein-cholesterol, LIP = lipase, LL = lower limit, Me = median, Mg = magnesium, MRA = multiple regression analysis, Na = sodium, NP = nonparametric, P = parametric, P5P = pyridoxal-5-phosphate, RI = reference interval, RMP = reference measurement procedure, r_p= partial correlation coefficient, SD = standard deviation, SDR = standard deviations ratio, SOP = standard operation procedure, SV = sources of variation, TBil = total bilirubin, TC = total cholesterol, Tf = transferrin, TG = triglycerides, TP = total protein, TTR = transthyretin, UA = uric acid, UL = upper limit.

LX, MC, ML, ZT, SL, and LW contributed equally to this study.

The authors have no funding and conflicts of interest to disclose.

1. CLSI and IFCC, C28-A3 Document. Defining, Establishing and Verifying Reference Intervals in the Clinical Laboratory: Approved Guideline. 3rd ed.2008; 28:1–76. [Google Scholar]

2. Ozarda Y, Ichihara K, Barth JH, et al.Protocol and standard operating procedures for common use in a worldwide multicenter study on reference values. Clin Chem Lab Med2013; 51:1027–1040. [PubMed] [Google Scholar]

3. Ichihara K.Statistical considerations for harmonization of the global multicenter study on reference values. Clin Chim Acta2014; 432:108–118. [PubMed] [Google Scholar]

4. Ichihara K, Ozarda Y, Klee G, et al.Utility of a panel of sera for the alignment of test results in the worldwide multicenter study on reference values. Clin Chem Lab Med2013; 51:1007–1025. [PubMed] [Google Scholar]

5. Wang W, Zhong K, Bai Y, et al.Investigation and analysis of reference intervals of routine clinical chemistry assays in china. Chin J Lab Med2011; 34:1139–1143. [Google Scholar]

6. Chan AO, Lee KC, Leung JN, et al.Reference intervals of common serum analytes of Hong Kong Chinese. J Clin Pathol2008; 61:632–636. [PubMed] [Google Scholar]

7. Yang S, Qiao R, Li Z, et al.Establishment of reference intervals of 24 chemistries in apparently healthy adult Han population of Northern China. Clin Biochem2012; 45:1213–1218. [PubMed] [Google Scholar]

8. Lo YC, Armbruster DA.Reference intervals of common clinical chemistry analytes for adults in Hong Kong. eJIFCC2012; 23:5–10. [PMC free article] [PubMed] [Google Scholar]

9. Mu R, Chen W, Pan B, et al.First definition of reference intervals of liver function tests in China: a large-population-based multi-center study about healthy adults. PloS One2013; 8:e72916. [PMC free article] [PubMed] [Google Scholar]

10. Ichihara K, Boyd JC.An appraisal of statistical procedures used in derivation of reference intervals. Clin Chem Lab Med2010; 48:1537–1551. [PubMed] [Google Scholar]

11. Yang W, Lu J, Weng J, et al.Prevalence of diabetes among men and women in China. N Engl J Med2010; 362:1090–1101. [PubMed] [Google Scholar]

12. Xi B, He D, Hu Y, et al.Prevalence of metabolic syndrome and its influencing factors among the Chinese adults: the China Health and Nutrition Survey in 2009. Prev Med2013; 57:867–871. [PMC free article] [PubMed] [Google Scholar]

13. Song QB, Zhao Y, Liu YQ, et al.Sex difference in the prevalence of metabolic syndrome and cardiovascular-related risk factors in urban adults from 33 communities of China: The CHPSNE study. Diab Vasc Dis Res2015; 12:189–198. [PubMed] [Google Scholar]

14. Ichihara K, Ceriotti F, Tam TH, et al.The Asian project for collaborative derivation of reference intervals: (1) strategy and major results of standardized analytes. Clin Chem Lab Med2013; 51:1429–1442. [PubMed] [Google Scholar]

15. Ichihara K, Ceriotti F, Kazuo M, et al.The Asian project for collaborative derivation of reference intervals: (2) results of non-standardized analytes and transference of reference intervals to the participating laboratories on the basis of cross-comparison of test results. Clin Chem Lab Med2013; 51:1443–1457. [PubMed] [Google Scholar]

16. Chinese Society of Endocrinology. Concensus on treatment of hyperuricemia and gout in China. Chin J Endocrinol Metab2013; 29:913–920. [Google Scholar]

17. Ichihara K, Kawai T.Determination of reference intervals for 13 plasma proteins based on ifcc international reference preparation (CRM470) and NCCLS proposed guideline (C28-p, 1992): a strategy for partitioning reference individuals with validation based on multivariate analysis. J Clin Lab Anal1997; 11:117–124. [PMC free article] [PubMed] [Google Scholar]

18. Ichihara K, Itoh Y, Lam CW, et al.Sources of variation of commonly measured serum analytes in 6 Asian cities and consideration of common reference intervals. Clin Chem2008; 54:356–365. [PubMed] [Google Scholar]

19. Harris EK, Boyd JC.On dividing reference data into subgroups to produce separate reference ranges. Clin Chem1990; 36:265–270. [PubMed] [Google Scholar]

20. Lahti A.Are the common reference intervals truly common? Case studies on stratifying biochemical reference data by countries using two partitioning methods. Scand J Clin Lab Invest2004; 64:407–430. [PubMed] [Google Scholar]

21. Smith HS, Bracken D, Smith JM.Gout: current insights and future perspectives. J Pain2011; 12:1113–1129. [PubMed] [Google Scholar]

22. Executive summary of the third report of the National Cholesterol Education Program (NCEP) expert panel on detection, evaluation, and treatment of high blood cholesterol in adults (adult treatment panel, iii). JAMA2001; 285:2486–2497. [PubMed] [Google Scholar]

23. Ministry of Health PRC. Reference intervals for common clinical biochemistry tests—part 1: Serum alanin aminotransferase, aspartate aminotransferase, alkaline phosphatase and gama-glutamyltransferase. WS/T 404.1-2012. [Google Scholar]

24. Ministry of Health PRC. Reference intervals for common clinical biochemistry tests—part 2: Serum total protein and albumin. WS/T 404.2-2012. [Google Scholar]

25. Ministry of Health PRC. Reference intervals for common clinical biochemistry tests—part 3: Serum potassium, sodium and chloride. WS/T 404.3-2012. [Google Scholar]

26. Ministry of Health PRC. Reference intervals for common clinical biochemistry tests—part 5: Serum urea and creatinine. WS/T 404.5-2015. [Google Scholar]

27. Ministry of Health PRC. Reference intervals for common clinical biochemistry tests—part 6: Serum calcium, phosphorus, magnesium, iron. WS/T 404.6-2015. [Google Scholar]

28. Ministry of Health PRC. Reference intervals for common clinical biochemistry tests—part 7: Serum lactate dehydrogenase, creatine kinase. WS/T 404.7-2015. [Google Scholar]

29. Ministry of Health PRC. Reference intervals for common clinical biochemistry tests—part 8: Serum amylase. WS/T 404.8-2015. [Google Scholar]

30. Ozarda Y, Ichihara K, Aslan D, et al.A multicenter nationwide reference intervals study for common biochemical analytes in Turkey using Abbott analyzers. Clin Chem Lab Med2014; 52:1823–1833. [PubMed] [Google Scholar]

31. Borai A, Ichihara K, Masoud A, et al.Establishment of reference intervals of clinical chemistry analytes for adult population in Saudi Arabia: an study conducted as a part of the IFCC global study on reference values. Clin Chem Lab Med2015; [Epub ahead of print]. [PubMed] [Google Scholar]