Establishing Ghanaian adult reference intervals for hematological parameters controlling for latent anemia and inflammation

Abstract Background In Ghana, diagnostic laboratories rely on reference intervals (RIs) provided by manufacturers of laboratory analyzers which may not be appropriate. This study aimed to establish RIs for hematological parameters in adult Ghanaian population. Methods This cross‐sectional study recruited 501 apparently healthy adults from two major urban areas in Ghana based on the protocol by IFCC Committee for Reference Intervals and Decision Limits. Whole blood was tested for complete blood count (CBC) by Sysmex XN‐1000 analyzer, sera were tested for iron and ferritin by Beckman‐Coulter/AU480, for transferrin, vitamin‐B12, and folate was measured by Centaur‐XP/Siemen. Partitioning of reference values by sex and age was guided by “effect size” of between‐subgroup differences defined as standard deviation ratio (SDR) based on ANOVA. RIs were derived using parametric method with application of latent abnormal values exclusion method (LAVE), a multifaceted method of detecting subjects with abnormal results in related parameters. Results Using SDR ≥ 0.4 as a threshold, RIs were partitioned by sex for platelet, erythrocyte parameters except mean corpuscular constants, and iron markers. Application of LAVE had prominent effect on RIs for majority of erythrocyte and iron parameters. Global comparison of Ghanaian RIs revealed lower‐side shift of RIs for leukocyte and neutrophil counts, female hemoglobin and male platelet count, especially compared to non‐African countries. Conclusion The LAVE effect on many hematological RIs indicates the need for deliberate secondary exclusion for proper derivation of RIs. Obvious differences in Ghanaian RIs compared to other countries underscore the importance of country‐specific RIs for improved clinical decision‐making.


| INTRODUC TI ON
Laboratory tests serve many purposes, including disease diagnosis and monitoring, predicting risk, and managing therapy. The critical contribution of laboratory tests in healthcare delivery is largely dependent on availability of appropriate information supporting proper interpretation. 1,2 Basically, this information is provided in the form of a reference interval (RI) [2][3][4] or the central 95% range of the distribution of reference values (RVs) from well-defined healthy individuals. 2,3 Considering the relevance of RIs, the International Federation of Clinical Chemistry (IFCC) has recommended that each laboratory establishes its own reference values and estimates the corresponding RIs using well-defined procedures. Based on the IFCC's recommendation, the Clinical and Laboratory Standard Institute (CLSI) published the C28-A3 guideline in determining RI for quantitative clinical laboratory tests. 5 Following the guideline, several RI studies involving hematological parameters have been conducted in recent years in African countries. [6][7][8][9][10] However, these reported RIs were discordant. 11 This is probably attributable to either small sample size, loose criteria for volunteer recruitment, not accounting for latent diseases, or use of nonparametric method for analysis that is susceptible to extreme values. [12][13][14] For improved derivation of RIs, country-based multicenter study was recommended in 2012 by the IFCC Committee on Reference Intervals and Decision Limits (C-RIDL). 15 A harmonized protocol was issued by C-RIDL, which optimized the methods for recruitment, sampling, measurement, secondary exclusion, and statistical analyses to allow for reproducible and standardized derivation of RIs. 12,15 Studies following the C-RIDL protocol were recently conducted in Turkey 16 and Kenya 11 which established RIs from well-defined healthy volunteers.
In Ghana, a few studies have been conducted to establish hematological RIs, 13,[17][18][19] but none employed a prescribed well-designed protocols as in other African studies, and these RIs may not be suitably applied to general use. As a matter of fact, most clinical laboratories continue to depend on hematological RIs provided by manufacturers. Therefore, it is crucial to establish RIs specific to the Ghanaian population by applying improved methodologies, and to serve as the RIs for unbiased clinical decision-making. As a part of the global multicenter study coordinated by C-RIDL, we set out to establish a population-based RIs for hematological parameters, including serum iron markers, and related vitamins for healthy Ghanaian adults based on the international harmonized protocol and utilizing advanced statistical methods.

| Volunteer recruitment
Apparent healthy volunteers were recruited in accordance with the protocol published by the IFCC, Committee on Reference Intervals and Decision Limits (C-RIDL), for multicenter studies. 12 The reference individuals were recruited from public institutions, schools, health facilities, social centers, churches, and mosques from June to December 2017. A total of 501 study participants aged 18-70 years from the Greater Accra and Northern regions in Ghana representing different ethnic groups were recruited. The sample size was appropriately computed as to be adequate and to guarantee reproducible test results for making between-country comparison. 12 Structured questionnaires were administered to volunteers to collect data on demographics, lifestyles, nutrition pattern, and health-status. The selection of eligible participants was based on well-defined inclusion and exclusion criteria, in accordance with the IFCC/C-RIDL protocol. 12

| Blood sampling and storage
Blood drawing was done under basal conditions as recommended in the IFCC/C-RIDL protocol between 7:00-10:00 am after overnight fasting for 10-14 hours, avoidance of strenuous muscular exertion for three prior days, sitting still for at least 20 minutes prior to venipuncture to avoid postural changes. 20 A fasting blood sample of 2 mL was drawn into two plastic vacutainer tubes containing ethylenediaminetetraacetic acid (EDTA) (BD Vacutainer® Blood Collection Tube, West Africa) for complete blood count (CBC). Subsequently, to ensure uniformity and prevent coagulation, each sample tube was immediately inverted gently 3 ~ 4 times. Also, 9 mL fasting blood samples were drawn into BD Vacutainer® SST tubes (Becton-Dickinson Corp) containing a clot-activator. Analytes tested in serum were the non-CBC parameters listed below. Anthropometric measurements (weight, height, waist circumference, blood pressure) were taken on the day of sample collection.

| Analytical procedure
Whole blood samples were analyzed for CBC within 6 hours after sample collection using a Sysmex XN-1000 analyzer (Sysmex Corp).
The serum specimens for non-hematological parameters (IgG, IgA, IgM, C3, C4, CRP) were centrifuged after sampling to separate the serum daily. The sera were stored at −80°C until the time of collective measurements. Those analytes were measured together with other analytes for a parallel study for establishing chemistry/immunology RIs. Test results for a part of chemistry analytes were used in this study as a part of reference tests in the latent abnormal values exclusion (LAVE) procedure (see below) for reducing the influence of individuals with latent anemia or inflammation.

| Quality control
All laboratory investigations were carried out in accordance with the laboratory's Quality Manual and standard operating procedures (SOPs). We adhered to instructions described in each analyser's manual and reagent/kit package inserts. During the period of collective measurements of non-hematological parameters in batches of 80 ~ 100 specimens, between-day quality control was performed by testing mini-panel of sera as specified in IFCC/C-RIDL protocol. 12

| Ethical approval
The study protocol was approved by the Ethical Review Committee

| Sources of variation (SV) of reference values
Multiple regression analysis (MRA) was performed separately for males and females by setting RVs of each analyte as an objective variable. Explanatory variables, age, BMI, ethnicity (Akan = 1; others = 0), Systolic BP, and Diastolic BP were set constant. The degree of association of each explanatory variable with the objective variable was expressed as a standardized partial regression coefficient (rp), which takes a value between −1.0 and 1.0. We considered |rp| ≥ 0.2 as an appreciable effect size between small (0.1) and medium (0.3) correlation specified by Cohen. 21

| Partitioning criteria for reference values
The need for partitioning RVs by sex and age was judged by standard deviation ratio (SDR). The SDR represents a ratio of betweensubgroup SD (variation of the subgroup means from grand mean) to between-individual SD (approximately 1/4 the width of RI, representing between-individual SD). Two-level nested ANOVA was performed to compute between-sex SD and between-age group SD after partitioning age as 18-29, 30-39, 40-49, and 50+ years. The SDR for between-sex SD (SDR sex ) and for between-age SD (SDR age ) was computed as a ratio to residual SD (or between-individual SD).
Because between-age variation changes by sex, one-way ANOVA was also performed to compute SDR age for each sex. The SDR ≥ 0.40 was considered as practically significant between-subgroup difference to derive sex-or age-specific RIs. 13,22 However, SDR is sometimes insensitive to actual difference (bias) at LL or UL after partitioning because SDR represents central tendency of variation by the SV, not necessarily represents variation in the periphery. Therefore, we set up an additional index called bias ratio (BR) at LL or UL, BR LL , or BR UL , to cope with the problem.
where subscript M, F, and MF represent male, female, and male + female, respectively.
In accordance with the convention of allowable bias 23 specification of a minimum level: 0.375 × SD G (=SD RI ), we regard BR UL > 0.375 as an auxiliary criterion for partitioning RVs when SDR does not match to actual between-subgroup difference at ULs (or LLs).

| Derivation of reference intervals
Parametric and nonparametric methods were used in deriving the RIs. For the parametric method, the reference values were first transformed into Gaussian distribution by use of the modified Box-Cox power transformation formula 13,24 , in determining the mean and SD, the final RI was calculated as the mean ± 1.96 SD (after truncating the values outside mean ± 2.81 SD once), which corresponds to the lower and higher limits (LL and UL) of the RI under the transformed scale. Then, the limits were reverse transformed to get the LL and UL in the original scale.
The 90% confidence intervals (CIs) for both LL, UL of the RI were estimated by use of the bootstrap method through iterative resampling 50 times. Making use of this procedure, the final RIs were set to averages of LL and UL computed repeatedly.
One important aspect of the data analysis was to detect and remove inappropriate values that represent latent diseases of common occurrence. Therefore, prior to derivation of RIs, the LAVE method was used to exclude subjects with latent disease conditions/abnormal values. The LAVE method is an iterative optimization method for filtering reference individuals by excluding subjects with abnormal values in related analytes. 13,24 In the initial calculation, RIs were calculated test by test independently without any exclusion, but in the subsequent iterative calculation, any individual who had two or more results outside the previously derived RIs among the reference tests (described below) were excluded.
The RIs were computed in two parts: one for markers of erythropoiesis and iron metabolism (folate, VitB12, Fe, TF, ferritin, RBC, Hb, Ht, MCV, MCH, MCHC, and RDW), the other for leukocytes and platelets (WBC, differential counts, PLT, and MPV). Reference tests used for the erythrocytes group were Fe, TF, ferritin, Hb, Ht, and MCV, and for the leukocytes group were IgG, IgA, IgM, C3, CRP, and ferritin.
The BR LL or BR UL was also used for judging the need for the LAVE method by the following equation.

| Demographic profile of the participants
The total sample size of reference individuals included in data analysis was 501, comprising 54% (n = 270) males and 46% (n = 231) fe-

| Sources of variation and the scheme for partitioning reference values
The MRA results are shown in Table S1. By regarding the effect size (practical significance) of rp as 0.20, blood pressure had no association  (Table 1). Note: Two-level nested ANOVA was used to compute between-sex SD (SD sex ) and between-age SD (SD age ) and between-individual SD (SD bi ). SDR sex was derived as SD sex /SD bi . For deriving sex-specific SDR age (SDR age M, and SDR age F), one-way ANOVA was performed separately for each sex. Values of SDRs ≥ 0.3 are shown in bold and SDRs ≥ 0.5 were highlighted by gray background.
The graphical representation of sex-and age-related changes for all the hematological parameters are shown in Figure S1. In Figure 1, the graphs are shown for 8 or 12 selected analytes with conspicuous sex and/or age-related variations.

| Derivation of reference intervals
We derived the RIs in four ways by parametric (P) and nonparametric (NP) method with/without LAVE. The comparisons of RIs for twelve selected analytes are shown in Figure 2.  With regards to Eos#, the Ghanaian RIs were appreciably lower, especially in females, compared to all the African countries, reflecting very low prevalence of allergic diathesis in the Ghanaian population recruited for the study (Section 3.1).
For platelets, the RI for Ghanaian males was shifted to a lower side compared to most of other countries. Also, the UL of Ghanaian females for platelet is higher than most of the non-African countries.
Further, we compared our CBC RIs with three other studies conducted in Ghana, all of which used the nonparametric method without any procedure for secondary exclusion. As shown in Figure   S2, with regard to erythrocyte parameters, LLs for RBC, Hb, and Ht of our study were conspicuously higher than other studies both in males and females. For the leukocyte parameters, there was a noticeable trend of lower levels of our ULs for WBC, Neu#, Lym#, Mon#, and Eos# than other studies, with much narrower width of the RIs. On the other hand, the LLs of platelets in our study were generally higher in both males and females.

| D ISCUSS I ON
Proper interpretation of RVs requires understanding the sources of variation that could influence laboratory tests such as sex, age, ethnicity, and BMI. Detailed evaluation of these factors was an important part of this study, which provided a clue for the need for partitioning or exclusion of RVs. We used SDR as a measure to quantify the magnitude of between-sex and age differences. By setting SDR = 0.4 as its threshold, we observed significant between-sex differences in RVs for many erythrocyte and iron parameters: RBC, Hb, Ht, RDW, Fe. We compared our results with Kenya, 11 Japan, [25][26][27] and Turkey, 16 and found that our SDR sex values are consistent with these countries.
Besides the parameters of erythrocyte and iron metabolism, between-sex difference was noted only for PLT (SDR sex = 0.46).
This finding is consistent with Kenyan's result (SDR sex = 0.41), 11 but contradicts what was reported in Japan (SDR sex = 0.13 ~ 0.23), 27 and Turkey (SDR sex = 0.23). 16 This gender difference of RIs for PLT, females: 157-402 × 10 9 /L vs males: 115-339 × 10 9 /L has been attributed to the effects of endogenous female sex hormones, mainly the estrogen, on the activation of platelet formation. 28  With respect to the age-related changes, RBC showed age-related decrease with SDR age = 0.42 among males, which was in contrast with other studies conducted in Kenya, 11 Japan, 26,27 and Turkey. 16 This finding in the Ghanaian population may be due to dietary factors, since Ghanaian meals are high in carbohydrates and low in protein or due to unknown genetic predisposition which could TA B L E 2 Reference intervals for hematological analytes derived by parametric method  38 China, 39 and Australia. 40 The effect of chronic antigenic stimulation may account for this observation and probably the higher prevalence of parasitic infections among Ghanaians. 11,19,41 We also observed that in our population, Eos# levels were higher compared with Spain, Australia, and Malaysia 38,40,42 but were relatively lower compared with other African countries. 10

F I G U R E 3 (Continued)
It is also noteworthy that neutrophil counts are much higher than lymphocyte counts in Turkey, Malaysia, Australia, and UK 16,38,40,43 while both counts are approximately the same in Ghana and other African countries. 10,11 No clear explanation is available, but we assume that immunological surveillance system depends on environmental and/or genetic factors. Our study also showed that folate levels were lower compared with those reported in Asian study 22 and in the interim report of the IFCC global study. 15 This observation may be due to nutritional factors such as low intake of green vegetables among Ghanaians, since most of Ghanaian meals are carbohydrate and cereal based.
Interestingly, comparison of our RIs with three other studies conducted in Ghana revealed significant differences between them as shown in Figure S2.

| CON CLUS ION
This study confirmed the advantage of the parametric method over the nonparametric method and the need for application of the LAVE method for analytes, which are easily influenced by high prevalence of conditions such as latent anemia and inflammations. Based on the objective criteria (SDR and BR), partitioning RVs by sex was shown to be necessary not only for erythrocyte parameters, but also for iron markers, PLT, and Eos#. It is of note that the Ghanaian RIs for WBC and Neu# were markedly low compared to those of non-African countries. The low RI for Hb for the female Ghanaian even after the LAVE procedure is attributable to generally low iron reserves.
The robust statistical techniques used in this study made the RIs well matched to Ghanaian population. We hope the hematological RIs from this study will be adopted in regional laboratories and thereby lead to improved clinical decision-making and healthcare in Ghana and the surrounding countries.

ACK N OWLED G EM ENTS
The primary financial support for this research was provided by

CO N FLI C T O F I NTE R E S T
The authors have no competing interests.