The red cell reference intervals for male and female Emiratis are noticeably broader (Table 1) than in other Caucasian populations in which α+ thalassemia is rare [15]. Similar observations were reported on young adults in Saudi Arabia, in whom the frequency of α+ thalassemia allele varies between 0.07 and 0.5, as well as in Palestinians [10, 11]. This finding is expected in any population with considerable variations in the number and size of red cells, i.e., α+ thalassemia heterozygotes and homozygotes and normal homozygotes. In general, the effect of α+ thalassemia allele frequency on the three genotypes and their aggregate effects on the mean values of red cell parameters are shown in Figure 4. This analysis shows that the standards developed in populations with frequent α+ thalassemia are shifted to one side and wider, less precise, and critically depend on the frequency of α+ thalassemia allele. Therefore, such populations require separate reference intervals, one for subjects with phenotypically normal red cells and another for those with small red cells.
We separated two cell populations based on their size (MCV) rather than MCH, despite earlier reports that MCH is more useful in separating α+ thalassemia homozygotes from normal homozygotes [8]. The main reason for preferring MCV over MCH was that in our sample the frequency distribution of MCV was more clearly bimodal than that of MCH (Figures 1 and 2). As we were using mixture analysis (which breaks down a population into its constituent subpopulations by decomposing a frequency distribution into a mix of two normal distributions; here belonging to the normal and α+ thalassemia homozygote phenotype subpopulations, respectively), MCV rather than MCH appeared a better parameter to accomplish this task. In the study that found MCH a more useful than MCV in separating normal from α+ thalassemia homozygotes, the subjects' genotypes were known and α+ thalassemia heterozygotes were excluded from analysis [8], which we could not do. Nonetheless, the distributions of normal and small red cell populations clearly overlap (Figure 1). In our study, a value of 78.0 fl seems to best separate subjects with normal red cells from those with small red cells. The validity of this finding is supported by the finding in another study in which the same value of MCV best predicted α+ thalassemia homozygote defined by genotyping [18].
When tribal population stratification was taken into account (Figure 3) and adjusted for inbreeding, phenotype-derived α+ thalassemia allele frequency was estimated at 0.34, and. the prevalence of phenotype-derived α+ thalassemia homozygotes at 0.12. These results are similar to reports of homozygote frequencies obtained using genotyping (0.11) and other phenotyping (0.14) methods on the same population [1, 2]. Heterogeneity in allele frequency among tribes may well be due to founder effects, with random numbers of α+ thalassemia alleles segregated into subpopulations at the time of the foundation of current tribes. These differences were preserved by the practice of endogamy, which limits gene exchanges between the tribes. This substructure in the Emirati population (the consequence of tribal history) is also present in other Gulf Arab societies, and may explain reported variations in α+ thalassemia frequency among different Arab populations.
Remarkably, nearly half of the studied population is deduced to be α+ thalassemia heterozygous. Although in clinical practice these individuals are indistinguishable from normal, their erythroid indices are between the normal and α+ thalassemia homozygotes [7, 8]. Yet, contrary to expectations, the high prevalence of these phenotype-derived heterozygotes did not "blur" the bimodality of the distribution of MCV (Figure 1), suggesting that MCV values of most phenotype-derived α+ thalassemia heterozygotes are well within the normal range. Indeed, in another study of red cell sizes in known genotypes, 64% of the α+ thalassemia heterozygotes had MCV >78.0 fl [18].
As expected, the erythroid parameters for phenotype-derived normal and phenotype-derived α+ thalassemia homozygotes are significantly different (Table 3). For phenotypically normal subjects, the reference intervals closely overlap with those for Caucasians in which α+ thalassemia homozygosis is rare (Figure 5) [19–24]. Additionally, reference intervals markedly or completely overlap with the intervals published for adults genotyped as αα/αα (Figure 6) [7]. The results show that phenotypically normal Arabs have the same erythroid parameters as people of European origin. For phenotype-derived α+ thalassemia homozygotes, the reference intervals overlap with those for adults with -α/-α genotype (Figure 7) [7]. These comparisons validate our results obtained with phenotyping and mixture analysis of phenotypes. The observed variations of reference intervals in Figures 5, 6, 7 and in other studies would seem attributable to sample size, subject selection, and sample handling and processing.
A possibly contentious issue is the use of RDW ≥14.0% to identify iron deficiency. This unsatisfactory test may have introduced errors in estimating the prevalence of phenotype-derived α+ thalassemia homozygotes. This bias however is likely to be small, as nine times more women than men are excluded (Table 2), and the prevalence of phenotype-derived α+ thalassemia homozygotes in two genders is not significantly different (p = 0.34). In general, in the absence of significant iron deficiency, which is more prevalent in women than in men, there is no evidence that MCV of men and women are different. Thus, the use of the upper limit of normal RDW in males (Table 1A) to exclude iron deficiency in the females seems reasonable. A similar value for the upper limit of normality of RDW is found in Caucasian males of comparable age [15].