New Bioinformatics-Based Discrimination Formulas for Differentiation of Thalassemia Traits From Iron Deficiency Anemia

Transcription

1 New Bioinformatics-Based Discrimination Formulas for Differentiation of Thalassemia Traits From Iron Deficiency Anemia Abdul Hafeez Kandhro, MSc, 1,2 Watshara Shoombuatong, PhD, 1 Virapong Prachayasittikul, PhD, 3 Pornlada Nuchnoi, PhD 2,4, * Laboratory Medicine 48:3: ABSTRACT Thalassemia traits (TTs) and iron deficiency anemia (IDA) are the most common disorders of hypochromic microcytic anemia (HMA). The present study aimed to differentiate TTs from IDA by analyzing discrimination formulas and provides comprehensive data of hemoglobin disorders prevalent in Pakistan. Among 12 published discrimination formulas, 6 formulas MI, EF, G&K, RDWI, R, and HHI were the most reliable to discriminate TTs from IDA. The failure cutoff values were improved by the random forest (RF) decision-tree approach. Moreover, the Shine and Lal (S&L) formula, which completely failed to discriminate In clinical practice, HMA is a widespread hematological abnormality that is frequently caused by IDA and hemoglobinopathies (TTs). 1,2 The causes of IDA include inadequate intake of iron, chronic blood loss in the gastrointestinal tract, and, in women, prolonged menstrual bleeding. TTs may result from hereditary defects of one or more globin chains. 2,3 Depending on the type of globin chain defects present, the clinical phenotypes may present as Abbreviations TP, True positive; FP, False positive; FN, False negative; TN, True negative; SN, Sensitivity; SP, Specificity; AUC, Area under curve; ATT, alpha-thalassemia trait; BTT, beta-thalassemia trait; HBE, hemoglobin E; HBS, hemoglobin S; IDA, iron deficiency anemia; TT, thalassemia trait; MI, Mentzer Index; E&F, England & Fraser Index; S&L, Shine and Lal formula; EF, Ehsani formula; SF, Srivastava formula; PP, Palestinian population; G&K, Green & King Index; RDWI, RDW Index; R, Ricerca; HHI, Huber-Herklotz Index. 1 Center of Data Mining & Biomedical Informatics, Faculty of Medical Technology, Mahidol University, Bangkok, Thailand, 2 Center for Research & Innovation Faculty of Medical Technology, Mahidol University, Bangkok, 3 Department of Clinical Microbiology & Applied Technology, Faculty of Medical Technology, Mahidol University, Bangkok, 4 Department of Clinical Microscopy, Faculty of Medical Technology, Mahidol University, Bangkok IDA from TTs with original cutoff value (<1530), improved with the use of new proposed cutoff value (<1016) and was found to successfully discriminate all cases of TTs from those with IDA. In addition, 2 newly proposed formulas discriminated TTs from IDA more reliably than the original 12 formulas assessed. The proposed formulas could play a crucial role for clinicians to discriminate between TTs and IDA. Keywords: thalassemia trait, iron deficiency anemia, discrimination formula, Pakistan heterozygosity, homozygosity, or double heterozygosity. In addition, co-inheritance of structural hemoglobin variants occurs, such as hemoglobin C, E, or S with TTs. The common co-inheritance forms (hemoglobin E/β-thalassemia and hemoglobin S/β-thalassemia) are present in many countries, including Pakistan, 4-10 India, 11,12 Iran, 13 Turkey 14 and Saudi Arabia. 15,16 Most individuals with TTs, such as α-tt and β-tt, are asymptomatic and have a normal life expectancy. The identification of iron deficiency prior to the development of anemia is difficult. Low mean corpuscular volume (MCV) or low mean cell hemoglobin (MCH) are both indicators for thalassemia. However, these same indicators are found in IDA as well. 17 Hemoglobin electrophoresis used in the identification of α-tt does not provide possible diagnosis, but it is helpful to exclude other microcytic anemias. 18,19 The diagnosis of IDA depends on the serum ferritin level and transferrin saturation level rather than iron staining of bone marrow biopsies. The diagnosis of thalassemia depends on hemoglobin typing and DNA testing. All of these methods require significant expenditures of time and resources to confirm the diagnosis. *To whom correspondence should be addressed. pornlada.nuc@mahidol.ac.th To discriminate between these etiologies of HMA, researchers have proposed formulas based on RBC indices to American Society for Clinical Pathology, All rights reserved. For permissions, please journals.permissions@oup.com 230

2 Table 1. Evaluation of RBC Indices and Standard Discrimination Formula in Differentiation of Thalassemia Traits From IDA According to the Proposed Cutoff Values Along With New Proposed Formulas Cutoff Values RBCs indices / Formulas Standard Cutoff Proposed Cutoff TP FP FN TN SN SP AUC RBC < HB < HCT < MCV < MCH < MCHC < RDW < MI < E&F <0 < S&L <1530 < EF < SF <3.8 < PP <27 < G&K < RDWI < R < HHI < Kerman1 <300 <64.0, <298.0, > Kerman2 <85 >37.0, < Our formula (Kandhro1) < Our formula (Kandhro2) < Formulas (Standard discrimination formulas), Cutoff (Standard / previously published cutoff), TP, True positive; FP, False positive; FN, False negative; TN, True negative; SN, Sensitivity; SP, Specificity; AUC, Area under curve; ATT, alpha-thalassemia trait; BTT, beta-thalassemia trait; HBE, hemoglobin E; HBS, hemoglobin S; IDA, iron deficiency anemia; TT, thalassemia trait; MI, Mentzer Index; E&F, England & Fraser Index; S&L, Shine and Lal formula; EF, Ehsani formula; SF, Srivastava formula; PP, Palestinian population; G&K, Green & King Index; RDWI, RDW Index; R, Ricerca; HHI, Huber-Herklotz Index. differentiate between IDA and TTs. Eight common discrimination formulas are used, such as Mentzer Index (MI), 20 England & Fraser Index (E&F), 21 Shine and Lal formula (S&L), 22 Ehsani formula (EF), 23 Srivastava formula, 24 Palestinian population (PP), 25 Green & King Index (G&K), 26 and RDW Index (RDWI). 27 Existing discrimination formulas vary in sensitivity and specificity, and no single index exists that is suitable for every person. Because these indices were developed in the context of different populations, generations and genders, they may be neither suitable nor easily applied to daily clinical use. Despite multiple mutations in the β-globin gene, a limited number of different mutations account for most of the thalassemia patients in each area. Therefore, the current study aimed to generate a new mathematical formula using a new bioinformatics-based approach, called Random Forest (RF), for accurate discrimination of TTs from IDA in the Pakistani population. RF is nonparametric, interpretable, and efficient and has high prediction accuracy for many types of data. 28 RF-based, new formula cutoff values could be used (a) to reduce the burden of expensive testing for thalassemia, (b) in areas where confirmation by advanced laboratory facilities is lacking, and (c) in high-risk thalassemia areas. Consequently, all the efforts would reduce the carrier rates of α- or /β-tts in the future and with the help of genetic counseling reduce the birth rate of α- or /β-thalassaemia major. Materials and Methods Study Design The retrospective study focused on 3 years ( ) of encoded and unlinked clinical laboratory data obtained from the Healthcare Molecular and Diagnostic Laboratory, Hyderabad, Pakistan. The data consisted of 3030 HMA Lab Medicine 2017;48;

3 Figure 1 Work flow diagram of the study. cases defined as having an MCV <75 fl and an Hb <13 g/dl. The data set included complete blood counts (CBCs) with morphological examination, serum ferritin levels, and Hb electrophoresis. We reviewed the results of the data set and established diagnoses. We classified HMA cases (n=610) into six groups: α-tt (n=7), β-tt (n=300), HBE/β-TT (n=14), HBS/β-TT (n = 15), and IDA/TT (n = 18), and IDA (n = 256) (Table 1). The other 2420 cases had hypothyroidism, anemia of chronic disorders, hepatitis B, C, or D infection, or H. pylori infection and were excluded from the study. The study was conducted under the approval of Mahidol University Central Institutional Review Board (MU-CIRB, COE No. MU-CIRB 2016/ ). Twelve discrimination formulas were evaluated: Mentzer Index (MI), 20 (MCV/RBC); England and Fraser Index (E&F), 21 (MCV RBC- (5 Hb) k); Shine and Lal formula (S&L), 22 (MCV2 MCH/100); Ehsani formula (EF), 23 (MCV 10 RBC); Srivastava formula, 24 (MCH/RBC); Palestinian population (PP), 25 (MCV RBC 3 Hb); Green & King Index (G&K), 26 (MCV2 RDW /(Hb 100)); RDW Index (RDWI), 27 (MCV RDW/RBC); Ricerca (R), 27 (RDW/RBC); Huber-Herklotz Index (HH), 29 ((MCH RDW 0.1/RBC) + RDW); Kerman1, 30 (MCV MCH/RBC); and Kerman2, 30 (KERMAN1 10/MCHC). Figure 1 shows the workflow diagram of present study. Computational Approach Random Forest (RF), based on an ensemble-based decision tree, 28,31 is a widely used ensemble learning method. Breiman and Cutler introduced the RF method to improve prediction performances of classification and regression trees (CART) by growing many weak CART trees. 31 Out-ofbag (OOB) data are used for evaluating feature importance as follows: (a) two-thirds of a training data set is used to construct the predictive classifier and the remainder is used for evaluating the performance of such classifier; and (b) the importance of each feature can be evaluated by measuring the decrease in prediction performance. The performance evaluation can be reported in terms of either accuracy or the Gini index.the proposed cutoffs of our formulas as well as proposed cutoffs of standard formulas were obtained by Random Forest decision tree approach. In addition, the 10-fold cross-validation was applied to training data sets for performance testing. Generation of Proposed Formulas Our data consists of confirmed cases of TT and IDA cases. In Table 1 and Table 2, RBC, MCV, MCH, RDW, and HCT provide higher accuracies. Based on the above parameters, we then manually made various formulas and analyzed 232 Lab Medicine 2017;48; DOI: /labmed/lmx029

4 Table 2. Evaluation of Standard Discrimination Formula in Differentiation of β-thalassemia Trait From IDA According to the Proposed Cutoff Values Along With the New Proposed Formulas' Cutoff Values RBCs indices / Formulas Standard Cutoff Proposed Cutoff TP FP FN TN SN SP AUC RBC < HB < HCT < MCV < MCH < MCHC < RDW < MI < E&F <0 < S&L <1530 < EF < SF <3.8 < PP <27 < G&K < RDWI < R < HHI < Kerman1 <300 <63.7, <298.9, > Kerman2 <85 >37.97, < Kandhro1 < Kandhro2 < Formulas (Standard discrimination formulas), Cutoff (Standard / previously published cutoff), TP, True positive; FP, False positive; FN, False negative; TN, True negative; SN, Sensitivity; SP, Specificity; AUC, Area under curve; ATT, alpha-thalassemia trait; BTT, beta-thalassemia trait; HBE, hemoglobin E; HBS, hemoglobin S; IDA, iron deficiency anemia; TT, thalassemia trait; MI, Mentzer Index; E&F, England & Fraser Index; S&L, Shine and Lal formula; EF, Ehsani formula; SF, Srivastava formula; PP, Palestinian population; G&K, Green & King Index; RDWI, RDW Index; R, Ricerca; HHI, Huber-Herklotz Index. on two types of data; in the first data set, formulas were computed with RBCs indices, while in the second data set, formulas were computed with Hb-electrophoresis, ferritin and confirmed result (Remarks). After more than 10 different possible formulas were tested, 2 were found to provide the best results: (RBC/HCT + 0.5*RDW) and (RDW*5)/RBC. Both those formulas were then analyzed by means of the bioinformatics-based RF decision-tree approach to set the cutoff value for discrimination of TT and IDA. Two sets of cutoff values were obtained for Kandhro-1 and Kandhro-2 formulas: <8.2 and <16.8, respectively. We tested Kandhro-1 and 2 formulas using two different sets of cutoff values with standard formulas. Statistical Analysis We analyzed the data using the Statistical Package for Social Sciences (SPSS) v 18.0 (SPSS Inc, Chicago, IL, USA). One-way analysis of variance (ANOVA) was applied to compare data among the six groups. The sensitivity, specificity, positive and negative predictive values, Mathew correlation coefficient, and area under receiver operative curve (AUROC) were calculated for each formula for comparison of the discriminating formula performance among TT/IDA and β-tt/ida group. A P value of <.05 was defined as statistically significant. Results The baseline characteristics and laboratory data and 12 discriminating formula are summarized in online supplementary Table 1. By using the ANOVA test, significant differences (P <.0001) were observed among the 6 groups in all parameters except age (P =.097). Significant differences also were found in serum ferritin levels among all 6 groups (P <.0001). The accuracies of RBC indices and formulas for discrimination of TTs and IDA with number of cases are illustrated in Table 1. Higher sensitivity and specificity (>95%) and Lab Medicine 2017;48;

5 area under curve (AUC) (1.0) were obtained for both RBC parameters and discrimination formulas. RBC indices (such as RBC, MCV, MCH, and RDW) and discrimination formulas (eg MI, EF, G&K, RDWI, R, and HHI) could distinguish TTs from IDA. However, HB, HCT, and MCHC showed lower specificity (<90%). Their accuracy ranged from 93.22% to for SN and from 12.1% to 76.56% for SP. The E&F, Kerman1, and Kerman2 missed 43, 166, and 166 IDA cases, respectively. In addition, SF, PP, and Kerman2 missed 5, 1, and 131 TT cases, respectively. Thus, SN, SP, and AUC varied greatly, although E&F sensitivity remained 100% (as all TT cases were correctly discriminated from IDA). Very low levels of accuracies were observed in MCHC (12.1% SP and 0.42 AUC). Likewise, very low AUC were observed (Kerman2, 0.38 followed by Kerman1 0.76, respectively). However, the S&L formula completely failed to discriminate IDA cases since all calculated values were below the standard cutoff values (data not shown). RBC indices results are comparable to those of other studies. The mean results of RBC indices in a small study of β-tt and IDA were 5.04 for RBC, HB (10.4), HCT (33.34), MCV (66.82), MCH (20.76), and MCHC (31.08), which is comparable with our results. 32 Ali, et al 33 reported RBC, HB, and MCH values of 5.1, 11.0, and 22.7, respectively, in a small study from Karachi, Pakistan. Another study of 521 subjects with β-tt and IDA reported RBC, MCV, MCH, and RDW values of 5.1, 72.8, 19.9, and 13.9 respectively; only the RBC (5.1) value is comparable to our results. 34 Proposed Formulas In the present study we applied RF approach to generate and test 2 new mathematical formulas for use in the population of Sindh Province, Pakistan. These new formulas were Kandhro1 (RBC/HCT + 0.5*RDW) and Kandhro2 (RDW*5)/ RBC). Analysis of these formulas by the RF, decision tree approach revealed cutoff values with good reliability for discrimination of TTs from IDA. Figure 2 shows the feature importance of the 12 standard formulas and the 2 new proposed formulas. The highest Gini index was found among 11 discrimination formulas, including the 2 proposed formulas, while a lower Gini index was noted in 3 formulas, namely K1, K2, and E&F. From the original data set (n=610), we partitioned data for the RF model as the training data set (n=428). An external data set (n=182) was used for validating the RF model. Analysis of these formulas by the RF decision-tree Figure 2 Standard discrimination formulas (with improved cutoff values) and proposed discrimination formulas. The discrimination formulas feature importance plot shows mean decrease Gini index. The plot shows each discrimination formula on the y-axis, and their importance on the x-axis. The most- to least-important discrimination formulas ordered as top-to-bottom and their importance is given by the position of the dot on the x-axis. Based on mean decrease Gini index, from top-to-bottom as most- to least-important discrimination formulas are SL, HHI, GK, R, RDWI, MI, Kan1, Kan2, E, S, and PP. approach revealed cutoff values that provide good reliability for discrimination of TTs from IDA. The cutoff values of these formulas such as, Kandhro1 < 8.2 and Kandhro2 < 16.8, sample data less than cutoff values indicated TTs, while more than cutoff values indicated IDA The proposed and standard cutoff values of 14 discrimination formulas and the associated parameters are illustrated in Table 1. Eleven discrimination formulas (MI, E&F, S&L, EF, SF, PP, G&K, RDWI, R, HHI, and Kerman1) were found to be the most reliable for differentiating thalassemia traits from IDA, with 100% SN, 100% SP, and 1.0 AUC. Six of the original 12 formulas (MI, EF, PP, G&K, RDWI, and R) exhibited the same SN, SP, and AUC after minor adjusting of cutoff values. Four formulas (E&F, SF, PP, and Kerman1) had improved sensitivities, specificities and AUC. The Kerman2 formula, even with the proposed cutoff values, still did not discriminate TT from IDA. However, these proposed cutoff values did improve the SN, SP, and AUC as compared to standard cutoff values. Interestingly, the 234 Lab Medicine 2017;48; DOI: /labmed/lmx029

6 S&L formula, which failed to discriminate TT from IDA with standard cutoff value (data not shown), successfully discriminated all TT cases from IDA cases with the new proposed cutoff value. The newly proposed formulas, Kandhro1 and Kandhro2, showed 100% accuracy with proposed cutoff values. Similarly, the proposed cutoff values of 12 discrimination formula as compared to standard cutoff values and 2 newly proposed formulas for discrimination between β-tt and IDA are illustrated in Table 2. Comparing Proposed Formulas With Standard Formulas We performed 1-way ANOVA for comparing each proposed formula against competitors and shown in online supplementary Table 2. We found that among 12 discrimination formulas, only 10 (when compared with Kandhro 1) or 11 (when compared with Kandhro 2) formulas are significant in comparison with proposed formulas. This suggested that proposed formulas could be applied in future studies. Cross-Validation The main data set partitioned 2/3 data as a training data set, which is used for analysis of 10-fold cross-validation procedure. We found AUCROC equal 1.0 in 10-fold cross-validation analysis, indicating that Random Forest decision tree approach described here could better classify and discriminate thalassemia traits from IDA Discussion The present study provides comprehensive data on hemoglobin disorders prevalent of Sindh province, Pakistan. A total of 610 HMA cases were enrolled, of whom 256 cases were confirmed to have IDA, 07 to have α-tt, 300 to have β-tt, 14 to have HBE/β-TT, 15 to have HBS/β-TT, and 18 to have IDA/TT. Most of the TTs cases are asymptomatic and, without specialized tests, may be missed or sometimes misdiagnosed as iron deficiency anemia. Since 1970, several researchers have proposed discriminating formulas to differentiate between IDA and TTs. Some common discrimination formulas are MI, 20 E&F, 21 S&L, 22 EF, 23 SF, 24 PP, 25 G&K, 26 and RDWI. 27 Unfortunately, discrimination formulas have varied in sensitivity and specificity, and no single index exists that is suitable for every person. Most of the discrimination formulas were used to differentiation of β-tt and IDA with 100% sensitivity. Therefore, the present study aimed to determine which existing formula could best distinguish TTs from IDA. In addition, we generated new mathematical formulas with a bioinformatics-based approach (ie, RF) for accurate discrimination of TTs from IDA in HMA cases of different ethnic populations of Sindh Province in Pakistan. This study showed that only 6 of the 12 formulas evaluated were reliable in terms of high SN, SP and AUC values. However, the other 6 formulas had statistical values, that varied greatly. The reliability of using these indexes was controversial and varied in many previously published studies from different regions of the world Because these indices developed in the context of different populations, generations and genders, they may be neither suitable nor easily applied to general clinical use. More than 200 mutations have been identified that are involved in impairing transcription, processing or translation of α- or β-globin genes. 38,39 In Pakistan, 16 mutations are commonly recognized, 40 and very recently 5 more have been reported. 40 Niazi et al 41 evaluated 7 discrimination formulas (MI, S&L, SF, E&F, R, G&K, and RDWI) for distinguishing β-tt from IDA. RDWI, MI, and S&L have good discriminative function compared with E&F, G&K, and R. However, none of the formulas are 100% SN and SP. Shen et al 42 evaluated the validity of 12 discrimination indices for differentiation between IDA and β-tt in Chinese children. G&K was the best for differentiation between IDA and β-tt, followed by R, E&F, RDWI, PP, HHI, MI, E, SF, and S&L in decreasing order of performance. In Pakistan, multiple β-globin gene mutations have been identified, although a small number of mutations usually account for most of the thalassemia patients in an area. These mutations affect a certain amount of globin chain synthesis and cause variability in RBC indices. 30 The present study data contained a high number of β-tt cases which are best discriminated from IDA by only 6 discrimination formulas. In one cross-sectional prospective study, different ethnic groups were enrolled over a period of 5 years and their blood samples were analyzed at the DNA level for the 12 most common β-thalassemia mutations in Pakistan. All 12 mutations were found among the different ethnic groups, but prevalence was highest in the Sindhi ethnic group. 43 Although many authors have reported that certain discrimination formulas were reliable for differentiation of IDA from TTs in their Lab Medicine 2017;48;

7 own population, no study has shown whether all these discrimination formulas are valid in other populations. Similarly, remarkable inconsistencies among the 12 discrimination formulas were found in the present study population, and it was not possible to select an ideal discrimination formula with the published standard cutoff values. Therefore, the present study sought to improve the statistical values by applying an RF classification and regression trees (CART) approach and analyzing data for proposed discrimination formula cutoff values that could fit best for our different ethnicities. The proposed cutoff values were obtained by using the R package program and RF decision tree approach. Tables 1 and Table 2 show proposed cutoff values of 12 discrimination formulas as compared to standard cutoff values, along with 2 new proposed formulas and their cutoff values. Nearly all discriminating formulas improved in terms of SN, SP, and AUC when the proposed cutoff values were used. Kerman1 and Kerman2 had showed lower accuracies in standard cutoff values, however, improved with proposed cutoff values. Interestingly, the S&L formula, which completely failed to discriminate IDA with the standard cutoff value (<1530), improved with the proposed cutoff value (<1068), and successfully discriminated all TT cases from IDA. In Pakistan, the remarkable racial mix (due to migrations, invasions, and commercial interactions) represents significant genetic diversity. Consanguineous and close cousin marriage trends, high birth and fertility rates, low educational level, unawareness of trait status before marriages, and lack of public health facilities in remote areas have led Pakistan towards high rates of β-thalassemia carriage. In addition, the scarcity of screening facilities that can perform amniocentesis for DNA analysis during pregnancy and the lack of genetic counseling of positive cases for possible termination of pregnancy have led to a high risk of congenital transmission of β-thalassemia. Therefore, the current study provides new formulas to improve the diagnosis of TT independent of access to high tech methods. Analysis of these proposed formulas by a RF decision tree approach revealed that they could be used with good reliability for discrimination of thalassemia traits from IDA. Further study is necessary to confirm using spectrum of similar diseases and larger sample size. Conclusion Our proposed formulas, with their high sensitivity and specificity, could play a crucial role for clinicians who rely on laboratory results to discriminate between β-tt and other causes of HMA, specifically in premarital screening. Due to the high prevalence of IDA and β-thalassemia traits in many regions of Pakistan, the patient's history of iron intake and a history of thalassemia in a patient s family, along with a CBC, may not be adequate for a firm diagnosis. The proposed formulas provide an additional method to help differentiate IDA from TT. Therefore, the proposed formulas need to be evaluated in future studies of larger population samples. Recommendations There is variation of genetic basis of TTs in each population. The new proposed formulas (Kandhro1 and Kandhro2) along with standard formulas should be further investigated for screening TTs and IDA in high-risk endemic areas using larger population size. Base on our study, MI, EF, G&K, RDWI, R, and HHI are reliable discrimination formulas for ease of use in general population. Red cell indices and discrimination formula, along with clinical findings, should be carefully interpreted to increase the reliable diagnosis of suspected cases. LM Acknowledgments We are grateful for subjects participating in the study. This study was supported by the Faculty of Medical Technology, Mahidol University, Bangkok, and Thailand Research Fund to PN. We sincerely thank Dr. Arthur Brown for language editing and valuable comments on the manuscript. We also acknowledge the following people for their contributions: AHK for collecting and analyzing data as well as the draft of the manuscript; WS and VP for analyzing the data; and PN for formulating research questions, study design, critical comments on the data processing and results as well as revising the manuscript. Our proposed formulas could help to simplify the screening for TTs among high-risk groups in remote areas. They could also be used as a tool in premarital screening of close cousins and in large district or provincial surveys. References 1. Killip S, Bennett JM, Chambers MD. Iron deficiency anemia. Am Fam Physician. 2007;75(5): Lab Medicine 2017;48; DOI: /labmed/lmx029

8 2. Urrechaga E. Red blood cell microcytosis and hypochromia in the differential diagnosis of iron deficiency and beta-thalassaemia trait. Int J Lab Hematol. 2009;31(5): Martin A, Thompson AA. Thalassemias. Pediatr Clin North Am. 2013;60(6): Steinberg MH, Forget BG, Higgs DR, Weatherall DJ. Disorders of hemoglobin: genetics, pathophysiology, and clinical management. Cambridge University Press; Population Reference Bureau Lodhi Y. Economics of thalassemia management in Pakistan. Thalassemia awareness week Ansari SH, Shamsi TS, Ashraf M, et al. Molecular epidemiology of β-thalassemia in Pakistan: far reaching implications. Int J Mol Epidemiol Genet. 2011;2(4): Ghani R, Manji MA, Ahmed N. Hemoglobinopathies among five major ethnic groups in Karachi, Pakistan. Southeast Asian J Trop Med Public Health. 2002;33(4): Lorey FW, Arnopp J, Cunningham GC. Distribution of hemoglobinopathy variants by ethnicity in a multiethnic state. Genet Epidemiol. 1996;13(5): Vetter B, Schwarz C, Kohne E, Kulozik AE. Beta-thalassaemia in the immigrant and non-immigrant German populations. Br J Haematol. 1997;97(2): Gribble JN. World Population Data Sheet Washington, DC: Population Reference Bureau Madan N, Sharma S, Sood SK, Colah R, Bhatia LH. Frequency of β-thalassemia trait and other hemoglobinopathies in northern and western India. Indian J Hum Genet. 2010;16(1): Khorasani G, Kosaryan M, Vahidshahi K, Shakeri S, Nasehi MM. Results of the national program for prevention of beta-thalassemia major in the Iranian Province of Mazandaran. Hemoglobin. 2008;32(3): Keskin A, Türk T, Polat A, Koyuncu H, Saracoglu B. Premarital screening of beta-thalassemia trait in the province of Denizli, Turkey. Acta Haematol. 2000;104(1): Alhamdan NA, Almazrou YY, Alswaidi FM, Choudhry AJ. Premarital screening for thalassemia and sickle cell disease in Saudi Arabia. Genet Med. 2007;9(6): Wong LP, George E, Tan J-AMA. Public perceptions and attitudes toward thalassaemia: influencing factors in a multi-racial population. BMC Public Health. 2011;11: Rahim F. Microcytic hypochromic anemia patients with thalassemia: genotyping approach. Indian J Med Sci. 2009;63(3): Matos JF, Borges KB, Fernandes APS, Faria JR, Carvalho MDG. RDW as differential parameter between microcytic anemias in pure and concomitant forms. Jornal Brasileiro de Patologia e Medicina Laboratorial. 2015;51(1): Muncie HL Jr, Campbell J. Alpha and beta thalassemia. Am Fam Physician. 2009;80(4): Mentzer WC Jr. Differentiation of iron deficiency from thalassaemia trait. Lancet. 1973;1(7808): Hamblin TJ. Differentiation of iron deficiency from thalassemia trait by routine blood-count. Lancet. 1973;1(7804): Shine I, Lal S. A strategy to detect beta-thalassaemia minor. Lancet. 1977;1(8013): Ehsani MA, Shahgholi E, Rahiminejad MS, Seighali F, Rashidi A. A new index for discrimination between iron deficiency anemia and beta-thalassemia minor: results in 284 patients. Pak J Biol Sci. 2009;12(5): Srivastava PC, Bevington JM. Iron deficiency and-or thalassaemia trait. Lancet. 1973;1(7807): Sirdah M, Tarazi I, Al Najjar E, Al Haddad R. Evaluation of the diagnostic reliability of different RBC indices and formulas in the differentiation of the beta-thalassaemia minor from iron deficiency in Palestinian population. Int J Lab Hematol. 2008;30(4): Green R, King R. A new red cell discriminant incorporating volume dispersion for differentiating iron deficiency anemia from thalassemia minor. Blood Cells. 1988;15(3): ; discussion Ricerca BM, Storti S, d Onofrio G, et al. Differentiation of iron deficiency from thalassaemia trait: a new approach. Haematologica. 1987;72(5): Breiman L. Random forests. Mach Learn. 2001;45(1): Huber AR, Ottiger C, Risch L, Regenass S, Hergersberg M, Herklotz R. Thalassämie-Syndrome: Klinik und Diagnose. Syndromes thalassémiques: clinique et diagnostic. Paper presented at: Schweiz Med Forum; Miri-Moghaddam E, Sargolzaie N. Cut off determination of discrimination indices in differential diagnosis between iron deficiency anemia and β- thalassemia minor. Int J Hematol Oncol Stem Cell Res. 2014;8(2): Brieman L, Friedman J, Olshen R, Stone C. Classification and regression trees. Wadsworth & Brooks. Cole Advanced Books & Software; Hussain Z, Malik N, Chughtai A. Diagnostic significance of red cell indices in beta-thalassemia trait. Biomedica. 2005;21: Ali N, Moiz B, Bin Azhar W, Zaidi N, Memon R. Carrier detection for beta-thalassemia trait in general Pakistani population: a way forward. Hematology. 2012;17(4): Qazi RA, Shams R, Hassan H, Asif N. Screening for beta thalassemia trait. Journal of Rawalpindi Medical College (JRMC). 2014;18(1): Nalbantoğlu B, Güzel S, Büyükyalçın V, et al. Indices used in differentiation of thalassemia trait from iron deficiency anemia in pediatric population: are they reliable? Pediatr Hematol Oncol. 2012;29(5): Demir A, Yarali N, Fisgin T, Duru F, Kara A. Most reliable indices in differentiation between thalassemia trait and iron deficiency anemia. Pediatr Int. 2002;44(6): Pornprasert S, Panya A, Punyamung M, Yanola J, Kongpan C. Red cell indices and formulas used in differentiation of β-thalassemia trait from iron deficiency in Thai school children. Hemoglobin. 2014;38(4): Jouini L, Sahli CA, Laaouini N, et al. Association between clinical expression and molecular heterogeneity in β-thalassemia Tunisian patients. Mol Biol Rep. 2013;40(11): Sharma N, Das R, Kaur J, et al. Evaluation of the genetic basis of phenotypic heterogeneity in north Indian patients with thalassemia major. Eur J Haematol. 2010;84(6): Yasmeen H, Toma S, Killeen N, Hasnain S, Foroni L. The molecular characterization of beta globin gene in thalassemia patients reveals rare and a novel mutations in Pakistani population. Eur J Med Genet. 2016;59(8): Niazi M, Tahir M, e Raziq F, Hameed A. Usefulness of redcell indices in differentiating microcytic hypochromic anemias. Gomal Journal of Medical Sciences. 2010;8(2) Shen C, Jiang YM, Shi H, et al. Evaluation of indices in differentiation between iron deficiency anemia and beta-thalassemia trait for Chinese children. J Pediatr Hematol Oncol. 2010;32(6):e218 e Ansari SH, Baig N, Shamsi TS, et al. Screening immediate family members for carrier identification and counseling: a cost-effective and practical approach. J Pak Med Assoc. 2012;62(12): Lab Medicine 2017;48;