The Thal-index with the BTT prediction . exe to discriminate β-thalassaemia traits from other microcytic anaemias

Several attempts have been made previously to differentiate β-thalassaemia trait (BTT) from other microcytic anaemias using formulae with red cell (RC) parameters. Presently available formulae have low sensitivity and specificity. We wanted to develop a more precise algorithm, which could be used in situations where the gold-standard test for thalassaemia diagnosis: the high performance liquid chromatography (HPLC) is not available. The study was carried out prospectively from November 2008 to March 2010 from randomly collected blood samples with a mean cell volume (MCV) of less than 80 fL. HbA2 measured by HPLC was used to diagnose BTT. We used Fishers stepwise linear discriminant function analysis to develop an algorithm with RC parameters. Calculated new index Thal-index was then subjected to receiver operating characteristic curve analysis to identify best cutoff to discriminate BTT from other microcytic blood films. Software was developed to predict the BTT status (BTT prediction.exe). New index, referred to as the Thal-index, was calculated using discriminant function analysis and is given as Thal-index=[(0.615¥MCV) +(0.518¥mean corpuscular hemoglobin)+ (0.446¥red cell distribution width)]. A value of 59 for Thal-index has 90% sensitivity and 85% specificity for differentiating BTT from other microcytic anaemias. This showed better sensitivity and specificity compared to other formulae presently used (i.e., Mentzer in Eshani, et al.). Our study gives a better answer to set-up where HPLC is not available. Although this cannot replace HPLC, BTT prediction.exe is useful to predict instantly and is the first ever computer program available for this function. Introduction β-thalassemia trait (BTT) is one of the common causes of hypo-chromic microcytic anemias and differentiating it from the more common iron deficiency anaemia and other causes such as alpaha thalassaemia and sideroblastic anaemia without using sophisticated equipment has always been a challenge in low resource settings. Several formulae using red cell characteristics are already available which attempts to differentiate these anaemias with varying successes.1-4 We wanted to improve the rate of differentiation by using discriminant function analysis which is a technique used to build a predictive model of group membership based on observed characteristics of each case.5 A technique based on discriminant function analysis is not currently in use, nor is there easy to use computer software, which helps to differentiate BTT from other types of hypo chromic microcytic anaemias. We developed a new discriminant function using the red blood cell (RBC) parameters, which we would like to introduce as Thal-index as well as a computer program (BTT prediction.exe). Materials and Methods The study was carried out prospectively from November 2008 to March 2010 at Thalassaemia Unit at Ragama, Sri Lanka. All blood samples, which were referred for evaluation to the Thalassaemia Unit with a mean cell volume (MCV) less than 80 fL irrespective of haemoglobin level, were included consecutively to the sampling frame and 326 were selected randomly for the analysis. Complete blood counts were performed using a Beckman coulter AC.T diff.2TM analyzer 2002 (Block Scientific Incorporation, Bohemia, New York, USA) in all 326. Blood levels of HbA2 were determined using Chromsystems by high-performance liquid chromatography variant TTM analyzer 2000 (Bio-Rad Laboratory, Hercules, California, USA) in all 326. HbA2 levels equal to or more than 3.5 was considered as BTT but the higher values were compared with other parameters such as HbF and other haemoglobinopathies were excluded from the analysis. RBC indices were compared irrespective of HbA2 levels and significances were calculated between the BTT and other microcytic blood films [RBC count, hemoglobin level, haematocrit, MCV, mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration, and red cell distribution width (RDW)]. Stepwise multivariate discriminant analysis determined those indices that best differentiated the above 2 groups after HbA2 assay. The new index was calculated and its sensitivity and specificity was compared with the following 4 indices: England-Fraser=MCVRBC-(5¥Hb)-k, Mentzer=MCV/RBC, Sriva stava=MCH/RBC, Green and King (G&K) index=MCV2¥RDW/ 100Hb.3 Sensitivity and specificity were calculated. The new index cut off point was determined by plotting a receiver operating characteristics curve and the best values were taken into consideration and a computer program was developed to get the thalassaemia status directly from the three RBC indices.


Introduction
β-thalassemia trait (BTT) is one of the common causes of hypo-chromic microcytic anemias and differentiating it from the more common iron deficiency anaemia and other causes such as alpaha thalassaemia and sideroblastic anaemia without using sophisticated equipment has always been a challenge in low resource settings.
2][3][4] We wanted to improve the rate of differentiation by using discriminant function analysis which is a technique used to build a predictive model of group membership based on observed characteristics of each case. 5 technique based on discriminant function analysis is not currently in use, nor is there easy to use computer software, which helps to differentiate BTT from other types of hypochromic microcytic anaemias.
We developed a new discriminant function using the red blood cell (RBC) parameters, which we would like to introduce as Thal-index as well as a computer program (BTT prediction.exe).

Materials and Methods
The study was carried out prospectively from November 2008 to March 2010 at Thalassaemia Unit at Ragama, Sri Lanka.All blood samples, which were referred for evaluation to the Thalassaemia Unit with a mean cell volume (MCV) less than 80 fL irrespective of haemoglobin level, were included consecutively to the sampling frame and 326 were selected randomly for the analysis.
Complete blood counts were performed using a Beckman coulter A C .T diff.2TM analyzer 2002 (Block Scientific Incorporation, Bohemia, New York, USA) in all 326.Blood levels of HbA2 were determined using Chromsystems by high-performance liquid chromatography variant T TM analyzer 2000 (Bio-Rad Laboratory, Hercules, California, USA) in all 326.HbA2 levels equal to or more than 3.5 was considered as BTT but the higher values were compared with other parameters such as HbF and other haemoglobinopathies were excluded from the analysis.RBC indices were compared irrespective of HbA2 levels and significances were calculated between the BTT and other microcytic blood films [RBC count, hemoglobin level, haematocrit, MCV, mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration, and red cell distribution width (RDW)].Stepwise multivariate discriminant analysis determined those indices that best differentiated the above 2 groups after HbA2 assay.The new index was calculated and its sensitivity and specificity was compared with the following 4 indices: England-Fraser=MCV-RBC-(5¥Hb)-k, Mentzer=MCV/RBC, Sriva -stava=MCH/RBC, Green and King (G&K) index=MCV2¥RDW/ 100Hb. 3Sensitivity and specificity were calculated.
The new index cut off point was determined by plotting a receiver operating characteristics curve and the best values were taken into consideration and a computer program was developed to get the thalassaemia status directly from the three RBC indices.

Results
There were 326 samples of which 232 were diagnosed as having β-thalassaemia trait and the remainder as other anaemias.BTT group had mean HbA2 levels 4.7 (SD 0.54), other cohort had 2.53 (SD 0.32).Table 1 illustrates the significances of RBC parameters between the two cohorts.
Discriminant analysis identified MCV, MCH and RDW as the best set of indices for differentiating the 2 diagnoses.The new index, referred to as the Thal-index, was calculated using discriminant function analysis and formula is given as: The Thal-index gave 59.0 as the cutoff where 59 or below given as BTT.It has a sensitivity of 90% and a specificity of 85%.

Table 2 . Comparison of different indices used in the study.
*Index developed by authors of the current study.