A Nationwide Nepalese Study to Establish Reference Intervals for Major Biochemical Analytes with Elucidation of Nepalese Features of Reference Values

Abstract

Despite immense importance of reference intervals (RIs) for clinical diagnosis, there have been no reliable RIs available for Nepalese. Hence, this nationwide study was organized to establish RIs for 30 common biochemical parameters. This study was conducted following the harmonized protocol provided by IFCC Committee on Reference Interval and Decision Limits (C-RIDL) with recruitment of 617 apparently healthy volunteers (18 − 65 years) by near-equal gender balance from 5 major cities. Fasting blood were collected, serum was separated and measured collectively using Beckman-Coulter/Olympus AU480 chemistry analyzer. The sources of variations of reference values (RVs) were evaluated by multiple regression analysis and nested ANOVA. Latent abnormal values exclusion (LAVE) method was applied to reduce influence of latent diseases. RIs were standardized based on a value-assigned serum panel provided by C-RIDL. By ANOVA, no between-city differences were observed, while sex-related changes were typically noted for urate, creatinine, iron, γ-glutamyl transferase (GGT), immunoglobulin M, and transferrin, but not for high-density lipoprotein cholesterol. Age-related changes were observed for total cholesterol (TC), triglyceride, low-density lipoprotein cholesterol, and C-reactive protein (CRP). RIs were successfully derived all parametrically. The LAVE procedure was effective in lowering upper limits for aspartate aminotransferase, alanine aminotransferase (ALT), and CRP. Compared to other collaborating countries, Nepalese RIs were low for urea, cholesterols, ALT, and high for triglyceride, GGT, CRP, immunoglobulin G, and complements. The RIs for major chemistry analytes were derived and standardized for nationwide use in Nepal. This study distinctly elucidated sources of variation and international features of Nepalese RIs.

Supplementary Information

The online version supplementary material available at 10.1007/s12291-023-01123-6.

Introduction

The reference interval (RI) is used for the interpretation of patients test results, which defines the common range of values from apparently healthy subjects. Therefore, RI must be derived carefully for use in the patient population to be served by the clinical laboratory [1]. The International Federation of Clinical Chemistry (IFCC) published a series of six papers which recommended that each laboratory should produce its own reference intervals and verify them regularly [2, 3]. The C28-A3 guideline was issued by Clinical Laboratory Standards Institute (CLSI) in collaboration with IFCC [4]. The guideline states necessary steps for the selection of reference individuals, pre-analytical and analytical considerations, for derivation of RI. The International Organization for Standardization (ISO) 15189 also states that each laboratory should periodically re-evaluate its own RIs [5, 6].

However, the derivation of reference intervals by each laboratory is not feasible without collaborative efforts among laboratories [7]. Therefore, most laboratories utilize RIs supplied by in vitro diagnostic manufacturers, but these intervals may be dated, method specific, or not suitable for a given patient population. The situation is certainly true to laboratories in developing countries including Nepal. With this background, IFCC Committee on Reference Intervals and Decision Limits (C-RIDL) initiated a worldwide multicenter study on references values (RVs) in 2011 to facilitate derivation of country-specific RI, by providing a harmonized protocol [8], and value-assigned serum panel to facilitate standardized derivation of RIs [9].

Since there are no reliable RIs available for Nepalese population, we joined the C-RIDL project in 2016 with the following objectives.

1. To determine RIs for 30 most-commonly measured biochemical analytes specific for Nepalese population.

2. To standardize the RIs so that they are used in common nationwide.

3. To elucidate physiological sources of variations of RVs not only for judging the need of partitioning or excluding RVs secondarily, but also for facilitating appropriate clinical interpretation of laboratory test results.

4. To explore Nepalese specific features of standardized RIs by comparison to those of collaborating countries toward the globalization of medical practice.

Materials and Methods

Study Population and Recruitment

The study design, contents of the informed consent form and health questionnaire was approved by Nepal Health Research Council (NHRC), Ethical Review Committee Reg. no. 57/2014. Age, sex, height, weight, abdominal circumference, smoking habit, alcohol consumption and level of exercise were included in the general health questionnaire. A total of 617 individuals were selected out of apparently healthy volunteers from the five major cities of Nepal, following the C-RIDL protocol [8]. Health information was gathered through questionnaire comprising health and lifestyle questions.

Inclusion/Exclusion Criteria

Inclusion criteria: Healthy individuals who (1) regarded themselves healthy, (2) with age 18–65 years, (3) understood the objective of the study, (4) ideally not taking regular medication. Exclusion criteria: Individuals (1) with diabetes on oral therapy or insulin (diet alone is acceptable), (2) with history of chronic liver or kidney disease, (3) with blood test results that clearly point to a severe disease, (4) with history of being a hospitalized or otherwise seriously ill during the previous 4 weeks, (5) donated blood in the previous 3 months, (6) in carrier state for HBV, HCV, or HIV, (7) with a history of participation in other research study involving an investigational product in the past 12 weeks, and female participants with pregnancy or within 1 year after child birth.

Sample Collection and Handling

Volunteers were requested to avoid excessive physical exertion/exercise for 3 days prior to the sampling, excessive eating and drinking the night before, and to fast overnight for 10–12 h. Blood samples (2 mL of whole blood in sodium fluoride and 4 ml in yellow top vacutainer (BD, USA) were collected by trained phlebotomists at five centers of the country between 7:00 and 10:00 AM. To avoid variation due to postural influence and physical stress, the volunteers were requested to rest for 20 or more mins prior to collection [7]. Serum and plasma samples requiring centrifugation were spun 2–4 h after collection at 2000 g, for 10 min in a non-refrigerated centrifuge, aliquoted in cryo vials and stored at − 80˚C at the National Ayurveda Research and Training Center (NARTC).

Target Analytes and Measurements

After completion of the sampling, the sera stored at − 80˚C were tested for the following analytes collectively in four baches by use of Beckman-Coulter/Olympus AU480 biochemistry auto-analyzer: total protein (TP), albumin (Alb), total bilirubin (TBil), urea, uric acid (UA), creatinine (Cre), sodium (Na), potassium (K), chloride (Cl), calcium (Ca), iron (Fe), glucose (Glu), total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase (ALP), γ-glutamyltransferase (GGT), lactate dehydrogenase (LDH), amylase (AMY), creatinine kinase (CK), immunoglobulin G, A, and M (IgG, IgA, IgM), complement component 3 and 4 (C3, C4), C-reactive protein (CRP), transferrin (TF). The reagents, calibrators and quality control sera were proprietary to the manufacturer. Computed parameters were globulin (Glb) as (TP–Alb), and non-high density lipoprotein cholesterol (nonHDL) as nonHDL = TC–HDL-C. The assay principles, reagents, and analytical precision of each analyte are listed in Suppl. Table 1.

Quality Control (QC) and Traceability

Internal quality control specimens at both normal and pathological concentration were run for each analyte before sample testing. Further, to determine the traceability and to make the test results comparable across participating countries to the IFCC global study, a half of the panel of 50 sera with the assigned values were measured in batches along with the samples [9]. The major axis linear regression between the assigned values and the measured values was used to assess the need for recalibration of the test results. When difference in the lower or upper limit (LL, UL) of the RI before and after recalibration was more than 1/4 of standard deviation (SD) of the original RI, recalibration of test results was performed by using the major-axis linear regression coefficients (intercepts and slopes). This criterion corresponds to Fraser’s allowable limit of bias [10].

Source of Variations of Reference Values

Multiple regression analysis (MRA) was performed for all analytes after partitioning the dataset by sex. RVs of each analyte were set as objective variable and explanatory variables were set fixed to age, body mass index (BMI), the levels of cigarette smoking, alcohol consumption and regular physical exercises. Levels of smoking, alcohol consumption and physical exercise were classified into three, five and eight categories, respectively, using the following criteria: none, < 20, ≥ 20 cigarettes/day; none, < 12.5, 12.5 − 25, 25 − 50, > 50 g ethanol/day; none, 1 to 7 days/week. The standardized partial regression coefficient (r_p) calculated by MRA indicates the level of association of a given factor with RVs. It corresponds to partial correlation coefficient taking values between −1.0 and 1.0. The practical significance (effect size) was set to |r_p|≥ 0.2 as a midpoint effect size of Cohen’s small (0.1) and medium (0.3) correlation coefficient [11].

In judging the need for partitioning RVs by sex, and age, we primarily used standard deviation ratio (SDR) SDRsex and SDRage. They were calculated by nested ANOVA as a ratio of between-sex, between-age SD (SDsex, SDage) to between-individual SD, which corresponds to SD_RI described below. We adopted ≥ 0.4 as a threshold SDR for judging the need for partitioning RVs by sex and/or age [12].

However, SDR represents between-subgroup difference at the center of the RV distribution, which may not reflect between-subgroup difference (bias: Δ) at lower or upper limits (LL or UL) of the RI, namely, ΔLL or ΔUL. Therefore, we also calculated a bias ratio (BR), or a ratio of ΔLL or ΔUL to SD comprising RI (SD_RI) [ =|UL − LL|/3.92]. The formula for a case of partitioning by sex is: BR_LL =|LL_M− LL_F|/(UL_MF− LL_MF)/3.92, BR_UL =|UL_M− UL_F|/(UL_MF− LL_MF)/3.92, where subscripts of MF, M, and F attached to LL or UL indicate the RI derived without partition by sex (MF) and the RIs after partition by male (M) and female (F), respectively. The same calculation was used for determining the requirement for age-specific RIs by setting LL and UL derived with/without partitioning by age [12].

Derivation of Reference Intervals

Abnormal samples were excluded to cope with results attributable to highly prevalent diseases like metabolic syndrome and diabetes. In addition to excluding those individuals with BMI > 26 kg/m², as a method of secondary exclusion, the multivariate iterative method called latent abnormal values exclusion (LAVE) was applied for simultaneous derivation of RIs for multiple test items [13–15]. In its first iteration, the RI is derived test item by item independently disregarding latent abnormal values. From the second iteration on, the RI for each item is computed from data after excluding those from individuals with two or more values outside the RIs which were derived in the previous iteration for a list of predefined reference test items. This procedure results in the adjustment of the RIs when abnormal test results are associated with each other. Usually the RIs of all test items stabilize after 3–5 times iteration. In this study, the LAVE method was employed in reference to the values of Alb, UA, Glu, TC, TG, HDL-C, LDL-C, AST, ALT, GGT, CK, and CRP with iteration = 6 and allowance of up to one result outside the RI and up to one missing result in the reference test items. For the analysis of reference values for lipids, UA, Fe and Ca, those who were taking drugs or supplements for dyslipidemia and hyperuricemia, and for raising serum iron and calcium, respectively, were not included. Parametric (P) method was used for computing RIs after transforming the distribution of reference values into Gaussian form by use of two-parameter Box-Cox transformation [13]. RIs by the non-parametric (NP) method were also derived for comparative purpose. In order to improve the precision of the RI, the bootstrap method through 50-time resampling of the final dataset chosen by the LAVE method was applied to obtain smoothed lower and upper limits (LL, UL), and median (Me) of the RI both for the P and NP methods. This resampling procedure also served for predicting 90% confidence intervals (CI) for the limits of the RI.

Results

Demographic Profiles of the Participants

All individuals with overtly abnormal values were excluded from the study based on the possession of the following extreme values: Alb ≤ 32 g/L, Na ≤ 125 mmol/L, Glu ≥ 13 mmol/L, TG ≥ 7 mmol/L, ALT ≥ 200 U/L, GGT ≥ 300 U/L, and CK ≥ 800 U/L. As a result, 555 samples were used for the subsequent analyses. The ratio of female: male participants were 46:54%. Age distribution for the range of 18‒65 years was balanced. The numbers of female: male participants from the five major cities were 29:50 in Biratnagar, 39:61 in Dharan, 55:69 in Janakpur, 99:92 in Kathmandu, 33:28 in Pokhara. The breakdown of volunteers by city, sex and age, profile of the cohort in terms of BMI, ABO blood group, alcohol consumption, smoking, exercise and veg/non veg diet are shown in Suppl. Table 2.

Standardization of Reference Values

A half of 50 panel sera provided by C-RIDL were measured for all analytes except for TP, TBil, IP, Mg, and Glu. Then, the comparison of measured values with assigned values was performed by use of the major-axis regression line as shown in Suppl Fig. 1. By applying reference limits (LL, and UL) of each analyte which were derived without standardization, if a difference of predicted LL (or UL) from LL (or UL) exceeded 1/4 of the width of RI, we chose to recalibrate RVs of the analyte by using the regression line. According to this scheme, we performed recalibration of RVs for Cre, HDL-C, AST, ALT, and LDH, before performing the subsequent data analyses.

Multiple Regression Analysis

To explore sources of variation (SVs) for proper interpretation of derived RI, MRA was performed separately for each sex by setting RVs of each analyte as an objective variable and a fixed set of following factors as explanatory variables: age, BMI, regular exercise, habits of smoking, hours of sitting per day, vegetarian food habits. The values shown in Table 1 summarizing the results are standard partial regression coefficients (r_p) with significant values highlighted by a background color of yellow (|r_p|= 0.2–0.39) and orange (|r_p|≥ 0.4). Guided by the threshold of |r_p|≥ 0.2, age related changes were observed for Alb, urea, Glu, and TC in both sexes, for Glb, and Ca in males, and for UA, TG, LDL-C, non-HDL, ALP, CRP, IgM, and C3 in females. BMI related changes were observed for Glu, GGT, and CRP in both sexes, for UA, and TG in male and for AMY in female. Exercise habits showed positive relation with HDL-C, AST, ALT, and CK, in contrast, hours of sitting per day showed negative relation with those analytes. Whereas no notable changes were shown in any analyte in association with alcohol consumption, smoking habits, and vegetarian meal habits regardless of sex.

Table 1.

Multiple regression analysis for sources of variation of RVs

The values listed are standard partial regression coefficients (r_p). |r_p|≥ 0.2 was considered practically significant and shown in bold font with light orange color (0.2 ≤|r_p|< 0.4) and yellow (|r_p|≥ 0.4) background

N number of subjects, R multiple regression coefficient, BMI body mass index, Drk alcohol drinking level, Smk smoking level, Exer exercise level, Sit sitting/standing hours/day, Veg vegetarian diet

Need for partitioning of reference interval was analyzed by ANOVA. SD ratio based on two-level nested ANOVA (2N-ANOVA) was calculated for gender and age subgroups. SDR ≥ 0.4 was adopted as threshold for partitioning by gender and age [16]. The SDRs between-sex (SDRsex) and between-age (SDRage) differences were considered as shown in Table 2. SDRage was calculated for each sex separately as (SDRageM, and SDRageF) because age-related change differs by gender, and thus we performed the two stage calculation and adopted the latter SDRage in general. From SDRsex, gender-differences were prominent in RVs in five analytes UA, Cre, Fe, GGT, IgM. Unmatched SDRage between two genders were observed for eleven analytes Alb, ALP, Cre, GGT, Glu, LD, LDL-C, TC, TG, Urea and UA. It was apparent that age related changes are predominant in females. Sex and age-related changes are graphically shown as one-dimensional scattergram for 12 representative analytes in Fig. 1 and for all analytes in Suppl. Figure 2.

Table 2.

SD ratio (SDRs) for sex and age-related changes in RVs

SDRage and SDRsex were calculated for both gender by 2N-ANOVA. SDRageF and SDRageM were also computed separately for male and female by one-way ANOVA. SDR ≥ 0.3 is indicated by bold font, and SDR ≥ 0.4 was marked by background color in two grades (< 0.6: light orange; ≥ 0.6: orange)

Fig. 1.

Sex and age related changes in RVs for 12 representative analytes

Derivation of Reference Intervals

The RIs were derived for 30 biochemical analytes using both parametric and non-parametric methods after applying secondary exclusion based on the LAVE method. RIs derived by parametric method and their 90% confidence intervals for RI limits were narrower than those derived by non-parametric method as shown for 6 representative analytes in Fig. 2. Therefore, we adopted RIs by parametric method for all anlaytes. The figure also shows that LAVE effect in modifying RI limits was noted for AST and ALT in both sexes, CRP in males, LDL-C in females. Table 3 presents the complete list of RIs and 90% CIs derived by parametric method. The choice for partitioning RVs by sex, age, and the need for LAVE method were based on Suppl. Tables 3 and 4. The decision to adopt RIs derived by applying LAVE or BMI restriction was shown in Suppl. Table 5.

Fig. 2.

Comparison of RIs computed in 4 ways

Table 3.

List of RIs determined considering sex, age, BMI, and latent abnormal values

Discussion

We applied two-parameter Box-Cox formula for parametric method. It almost invariably succeeded in achieving Gaussian transformation for precise prediction of central 95% intervals as reported elsewhere, while we confirmed raised upper limit and its wider 90% CI predicted by nonparametric method especially for analytes with skewed distributions like AST, ALT, and CRP possessing frequent extreme values in the upper tail. The better performance of the parametric method is attributable to the inclusion of a step of truncating RVs outside the mean ± 2.81 SD after transformation [17].

LAVE method was adopted to reduce the influence of latent pathological and/or inappropriate sampling conditions [13, 14]. This study showed less impressive effect of LAVE, with notable lowering of UL found only in 4 analytes: AST, ALT in female, and GGT, CRP in male, which contradicts from those reported from Turkey, Saudi Arabia, and China [18–20]. However, when BMI restriction to ≤ 26 kg/m² was applied, ULs for TG, Glu, LDL-C, GGT, and CRP were lowered. This finding reflects existence of substantial fraction of subjects with metabolic syndrome, and points to the need for a stricter application of LAVE, which may have improved its efficacy though not possible due to substantial reduction in data size from our sample size.

Regarding sources of variations of RVs explored by multiple regression analysis (MRA) revealed age related changes in both genders of Alb, Urea, Glu, TC, and CRP, and female predominant age-related changes in UA, TG, LDL-C, non-HDL, ALP, and IgM as shown in Table 1. However, for the latter analytes, actual partition of RIs by age were done for both genders together only for LDL-C and non-HDL due to insufficient data size. The MRA also ascertained BMI-related increase in liver enzymes (ALT, AST, and GGT) mostly in male, inflammatory markers (CRP, C3), and nutritional markers (uric acid, Glu, and TG), all well documented in the literature [21–24]. When BMI increases, adipocytes inflammation and predispose towards insulin resistance and promote more production of uric acid. Some of these findings of BMI-related changes led us to decide on restricting BMI level before computing RIs for Glu, GGT, and CRP [25].

Other notable finding from MRA was positive association of regular exercise, but negative association of daily siting hours, for HDL-C, AST, and CK, especially in females. Whereas no significant association to RVs of any analyte were noted for alcohol consumption, smoking, and vegetarian meal habits with lack of enough number of subjects with those habits.

RIs derived in a standardized manner in this study were compared with those of other studies from the nine collaborating countries which were also standardized by common measurement of value assigned panel of sera provided by C-RIDL and summarized in Suppl. Table 6. Prominent features of Nepalese RIs among nutritional markers were picked out and compared by bar-chart as shown in Fig. 3. Lowered levels of RIs for cholesterol (TC, HDL-C, and LDL-C) were apparent. Besides, as shown in Suppl. Figure 3, the finding was the same by comparison to the 8-country Asian study [26] that ensured standardized derivation of RIs. By contrast, Nepalese RIs for TG and GGT were rather raised in both figures of between-country comparisons. Therefore, the lowered cholesterol RIs is not attributable to insufficient nutritional intake since the raised TG implies higher consumption of carbohydrate in Nepal.

Fig. 3.

Comparison of RIs across 10 countries for eight analytes with features specific to Nepalese

Whereas another feature of Nepalese RIs was significantly lowered RI for ALT (UL of 46 U/L) in males. The lowered ALT RI was evident by comparison to the Asian study (Suppl. Figure 3) and two Indian studies (UL: 74 U/L) [27, 28]. Meanwhile, such lowered RI of ALT was also shown among black African population in a recent report from South Africa that adopted the same C-RIDL protocol [29]. Therefore, the lowered ALT level is certainly not caused by insufficient nutrition considering raised levels of TG and GGT. These findings implicate a genetic factor is more presumable cause of lowered cholesterols and ALT RIs in Nepalese population.

Another finding from Fig. 3 was a lowered RI for urea in Nepal. It may be attributable to a low intake of protein by contrast to excess intake of carbohydrate. A higher level of CK in Nepalese males is attributable to a longer daily hours of standing or a higher physical activity among Nepalese as are true to males in Ghana and Kenya.

In Suppl Table 6, a Nepalese feature of RIs for inflammatory markers (CRP, immunoglobulins, and complements) was not apparent due to missing RIs from China, Saudi, and Turkey, and due to the presence of African countries that exhibited high levels of RIs for the inflammatory markers. While when turned to the standardized RIs reported in Asian shown in Suppl Fig. 3, it became apparent that Nepalese RIs for CRP, IgG, C3, and C4 were obviously raised, and even higher than those countries in Southeast Asia: Malaysia, Vietnam, and Indonesia, where exposure to microbial agents is high due to tropical climate and unsanitary conditions [26]. While, with lack of climate problem, the raised Nepalese RIs for the inflammatory markers are most likely attributable to the unsanitary social conditions.

Conclusion

Standardized RIs for 30 common biochemical analytes were derived in careful considerations of sources of variations of each analyte. They can be used in common in all clinical laboratories of Nepal, if assays are well standardized, for unbiased interpretation of patients test results. The characterization of Nepalese specific features of RIs, namely, lowered RIs for cholesterols, ALT, and urea, and raised RIs for TG, GGT, and inflammatory markers, will facilitate laboratory test reading towards the globalization of medical practice.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

This research was supported financially by the Japan Society for the Promotion of Science (JSPS) [No. 24256003: 2012–2014].

Declarations

Conflict of interest

Nothing to declare by any author. This prospective observational study was carried out in compliance with the Declaration of Helsinki for recruitment of healthy volunteers, blood sampling, testing, and reporting the results.

Ethical Approval

The study design, contents of the informed consent form and health questionnaire were approved by Nepal Health Research Council (NHRC), Ethical Review   Board Reg. no. 57/2014.