A Predictive Statistical Model To Identify Thalassemic Phenotype (β⁰ vs β⁺) in Heterozygous β-Thalassemia.

β thalassemias are a heterogeneous group of genetic alterations characterized by a decreasing (β⁺) or eliminating (β⁰) expression of b globin gene. Previous studies have reported that mean corpuscular volume (MCV) and mean corpuscular haemoglobin (MCH) values would be useful to differentiate between β⁺ and β⁰ thalassemia phenotypes in heterozygous carriers. An accurate and valid risk model for belonging to each of these groups may be valuable for prioritizing identification of genetic alterations in some populations. The aim of this study was to develop a theoretic model that combine information from the haematological indices for phenotypic prediction (β⁺ vs. β⁰) in heterozygous β thalassemia. The study was conducted on 238 unrelated β thalassemia carriers. Hematological parameters were obtained using analyzer Coulter ®GEN-S™. HbA₂ and HbF were analysed by high performance liquid chromatography. Molecular analysis for β globin and α-globin genes was performed by real-time PCR, direct sequencing techniques and standard PCR techniques. β-thalassemic subjects who also exhibited either αthalassemia or extra α globin genes were excluded. Statistical significance was calculated by using the χ2 or Fisher exact test for qualitative variables and the t test for continuous variables. A multivariate logistic regression model, with stepwise selection, was fitted from a case group of 64 β⁺thalassemia subjects and a reference group of 174 β⁰carriers. Odds ratios and their 95% confidence intervals (CIs) were computed. Statistical analysis was performed with the SPSS 12.0 statistical software package. The mean level of HbA₂ (4.5 ± 0.7% vs. 5.1 ± 0.7%, [p<0.0001]), MCV (66.9 ± 4.6 fL vs. 62.9 ± 3.1 fL, [p<0.0001]) and MCH (21.6 ± 1.8 pg vs. 20.1 ± 1.1 pg, [p<0.0001]) were significantly different between the β⁺ and β⁰ groups. By using an multivariate analysis, the three significant variables that finally entered the logistic regression model were MCH, HbA₂ and HbF. These variables were dichotomized as HbA₂> 4 vs. HbA2< 4, MCH> 20.5 vs. MCH< 20.5 and HbF> 1.5 vs. HbF< 1.5. The adjusted model implies that with a HbA2< 4, the chance of having β⁺thalassemia increase by a factor 8.3 (CI 1.6–43.1) compared with a HbA₂> 4. Analogously, at a MCH > 20.5, the chance of having β ⁺thalassemia increase by a factor 9.4 (CI 4.3–20.2) compared with a MCH< 20.5. Regarding to HbF, it is 1.3 times more likely (CI 1.1–1.9) to be included in the β⁺thalassemia group with HbF< 1.5 than with HbF> 1.5. For this adjusted model, with a 26,8% heterozygous β⁺thalassemia prevalence, model test performance characteristics included: sensitivity 61.4% and specificity 89.1% (p=0.000). The logistic multivariate model to allow accurate prediction of the phenotype of β thalassemia trait (β⁺ vs β⁰) in the 84.1% of the carriers. The Hosmer-Lemeshow statistics indicated fine goodness of fit of the logistic regression equation (P=0.739). In summary, this logistic regression analysis using a combination of hematological indices such as MCH, HbA₂ and HbF has showed an acceptable value in predicting phenotype of heterozygous b thalassemia (β⁺ vs β⁰) in a predominant Spanish population. The present model should be validated by independent data from different populations.