Sysmex Journal International Vol.8 No.2 (1998)
A Study of β Thalassemia Screening using an Automated Hematology Analyzer Yasumasa AKAI*, Fumio KUBOTA*, Sunil K BICHILE*2, Niriksha B MEHTA*2, Prakash R PANGAM*2, Vijay PAREKH*3, Hiroyuki NAKAMOTO* and Tokuhiro OKADA* * Research Division, Sysmex Corporation, 4-4-4 Takatsukadai, Nishi-ku, Kobe 651-2271, Japan. *2 Department of Haematology, T. N. Medical College & B. Y. L. Nair Hospital, Mumbai, India. *3 Sysmex Singapore Pte Ltd, Singapore.
Screening for β thalassemia, based on analytical parameters obtained from a hematology analyzer, such as RBC indices, was investigated in Mumbai, India. A total of 186 samples, from 28 patients with β thalassemia, 29 patients with iron deficiency anemia (IDA), 70 normal healthy individuals, and 59 others, were analyzed, using the Sysmex K-4500ª automated hematology analyzer. Assessment of the analytical parameters determined by the K-4500, such as RBC indices, and other previously reported discriminant functions (DF) from other hematology analyzers, suggested RDW-SD was most useful. When the criteria of MCV ² 80 and RDW-SD < 32 was judged to indicate positivity for β thalassemia, the sensitivity and specificity were 88.5% and 93.5%, respectively. Essentially, thalassemia occurs in different forms among different localities, including different genotypes, so that different tendencies for RBC indices are suggested. In applying this screening method to the other localities (geographic areas), the cut-off value may be set differently, in consideration of demographics and other factors. (Sysmex J Int 8 : 110-114, 1998) Key Words
β Thalassemia, Screening, Red Cell Distribution, Discriminant Function
Note : The corporate name of TOA Medical Electronics Co., Ltd. has been changed to Sysmex Corporation since October 1, 1998.
INTRODUCTION Thalassemia, a form of congenital hemolytic anemia, is characterized by hypochromic microcytic anemia. It is also accompanied by ineffective hematopoiesis, splenohepatomegaly, etc. It is not a single disease but involves multiple subtypes. The majority of cases found in the northwestern region of India are of the β thalassemia subtype. Although the differential diagnosis of homozygous β thalassemia (thalassemia major) is relatively easy on the basis of medical interviews and physical findings, the heterozygous β thalassemia (thalassemia minor) may be difficult to diagnose solely on the basis of manifestation. Thalassemia minor does not always require treatment; however, its diagnosis must be made for medical counseling regarding potential homozygous birth resulting from marriage of a couple of heterozygous parents. The diagnosis is generally made by screening patients with hypochromic microcytic anemia based on HGB concentration, MCV, and the other parameters determined by hematology cell analysis, then conducting a series of specific tests, including HGB electrophoresis. Traditionally, discrimination of thalassemia from the other diseases has been attempted using hematology analyzers by many investigators. England, et al.1, 2) studied the discrimination between thalassemia and iron deficiency anemia (IDA) using a discriminant function (DF) based on MCV data, HGB concentration and RBC count from a hematology analyzer. Other reports are also available from similar studies using DF based on RDW-CV, MCH, etc.3-7). In India, there are relatively high numbers of
patients with IDA showing hypochromic microcytic anemia, features common with β thalassemia, and discrimination between β thalassemia and IDA is a daily clinical issue from the viewpoint of treatment. In the present study, we investigated β thalassemia screening based on the above-mentioned methods and analytical parameters, such as RBC indices, as determined by a hematology analyzer, at Nair Hospital in India. As a result, RDW-SD was shown to be comparable with various linear discriminant functions, demonstrating its utility. This paper reports on the study in detail.
MATERIALS AND METHODS The samples assayed were obtained from 31 male medical students, 73 female medical students, and 82 patients at the hematology department. The study population breakdown by diagnosis was; normal in 70 patients, β thalassemia in 28, IDA in 29, and leukemia and the other diseases in 59 (Table 1). References to specific diagnoses, as noted in Table 1, have been established as final diagnoses after considering all diagnostic criteria. All samples were collected in K2EDTA as an anticoagulant. Each sample was analyzed within 8 hours after sample collection by use of Sysmex K-4500™ a hematology analyzer. Additional data analysis was accomplished by a connected external personal computer. Parameters determined by the K-4500, such as RBC indices, and previously reported DF were used for the analysis (Table 2).
− 110 −
Sysmex Journal International Vol.8 No.2 (1998)
Table 1 Samples and diagnose Sample
Medical students Male Female Patient*
Diagnosis**
ß thalassemia IDA Normal Others***
Table 2 Parameters used for analysis
case 104 31 73 82
RBC indices MCV RDW-CV* RDW-SD*
Discriminant function (DF) 28 29 70 59
Green’s DF Shine’s DF Bruno’s DF England’s DF Mentzer’s DF Srivastava’s DF
* Routine patient at hematology department ** Final diagnosis *** AML, CML, sickle cell, ITP, etc.
= = = = = =
Formulae MCV2 × RDW-CV / 100 / HGB MCV2 × MCH / 1000 0.096MCV + 0.415RDW-CV − 0.139RBC MCV − 5 × HGB − RBC MCV / RBC MCH / RBC
Reference Green, et al.3) Shine, et al.4) Bruno, et al.5) England, et al.1,2) Mentzer 6) Srivastava, et al.7)
* The Sysmex K-4500 has the function to provide RDW-CV and RDW-SD.
RESULTS MCV 100
Discrimination between β thalassemia, IDA and normal individuals
(fL)
80
60
40
β Thalassemia
IDA
Normal
RDW-CV 30
26
22
(%)
Fig. 1 shows scatter plot diagrams of various parameters and DF by subject group (β thalassemia, IDA, normal), demonstrating that there is a major overlap of the ranges of distribution of the clinical groups of interest. This means that differential diagnosis of β thalassemia based on a single parameter is difficult because many of the members of the IDA and normal groups show false positive response. When the scatter plot diagrams are examined, focusing on discrimination between the normal and patient groups (combined β thalassemia and IDA), the two groups show different distributions of MCV, demonstrating that discrimination between the normal and combined patient groups is relatively easy. For RDW-SD andthe method described by Shine’s DF (MCV 2 × MCH/1000)4) as well, discrimination between the normal and patient groups is relatively easy. These findings suggest that sensitivity and specificity can be increased by selecting particular patients from the patient group on the basis of these parameters for the first step, then using another parameter for further discrimination between β thalassemia and IDA.
18
14
Screening based on MCV (MCV cut-off value setting)
10
IDA
Normal
RDW-SD 80 70 60 (fL)
Of the parameters enabling easy discrimination between the normal and diseased groups, MCV (a parameter commonly determined by hematology analyzers) was used for screening. From the receive operating characteristic (ROC) curve, MCV ≤ 80 was taken as the cut-off value for the patient group (data not shown). Positive response was obtained in 28 out of the 28 β thalassemia patients, 23 out of the 29 IDA patients, and 7 out of the 70 normal individuals.
β Thalassemia
50 40
Parameters in combination with MCV ≤ 80
30
For the samples selected with the cut-off value MCV ≤ 80, the possibility of discrimination between thalassemia and IDA with the use of other parameters was assessed. Each cut-off value was set from the ROC curve − 111 −
20
β Thalassemia
IDA
Normal
Fig. 1-1 Scattered plots of various parameters by subject group
Sysmex Journal International Vol.8 No.2 (1998)
Green’s DF
Shine’s DF
300
300
250
250
200
200
150
150
100
100
50
50
0
β Thalassemia
IDA
0
Normal
β Thalassemia
IDA
Normal
Mentzer’s DF
England’s DF 40
60 50
30
40 30
20
20 10
10
0 -10 -20
β Thalassemia
IDA
0
Normal
β Thalassemia
Srivastava’s DF
IDA
Normal
Bruno’s DF
12
30
10
26
8 22
6 18
4 14
2 0
β Thalassemia
IDA
10
Normal
β Thalassemia
IDA
Normal
Fig. 1-2 Scattered plots of various parameters by subject group
so that the highest sensitivity was obtained (data not shown). Table 3 shows the agreement rates, sensitivities, and specificities of individual parameters. For all additional parameters, the sensitivity ranged from 78.6 to 89.3%, showing no major difference except for the discriminant functions for the methods of Srivastava’s DF (MCH/RBC)7) and Mentzer’s DF6). The specificity ranged from 30.4 to 91.3%, including considerably low specificity for RDW-CV, etc. The method of Mentzer showed highest specificity and lowest sensitivity, while RDW-CV and methods of England’s DF (MCV-5 × HGB-RBC) 1, 2) showed highest sensitivity and lower specificity. In short, sensitivity and specificity have trade-off relationship, and there is no parameter having high values in both sensitivity and specificity. Out of the K-4500 parameters, RDW-SD (sensitivity 85.7%, specificity 78.3%) was chosen for further investi-
gation as a parameter with a relatively good balance of sensitivity and specificity.
Application to all samples RDW-SD was applied to all the samples (i.e. normal, β thalassemia, IDA, and others) with MCV ≤ 80, and the results were assessed. The RDW-SD cut-off value was changed, as appropriate, and an ROC curve was drawn for sensitivity and specificity (Fig. 2). When the sensitivity was set at 100%, the specificity was 76.0%, and when the specificity was set at 100%, sensitivity was 26.9%. Using the ROC curve, MCV ≤ 80 and RDW-SD < 32 (sensitivity 88.5%, specificity 93.5%) were determined as the cut-off values expected to be most distant from the y = x line and to ensure the highest accuracy.
− 112 −
Sysmex Journal International Vol.8 No.2 (1998)
Table 3 Value agreement rates for individual parameters MCV 80
β Thalassemia IDA
RDW-SD <32
RDW-CV <22
N
P
N
P
N
P
N
P
N
P
N
P
N
P
N
P
4
24
3
25
4
24
5
23
6
22
6
22
3
25
4
24
18
5
7
16
16
7
14
9
21
2
19
4
12
11
16
7
Sensitivity (%) Specificity (%) FP (%) FN (%)
85.7 78.3 21.7 14.3
Green’s DF <80
89.3 30.4 39.0 10.7
Shine’s DF <110
85.7 69.6 22.6 14.3
82.1 60.9 39.1 17.9
Mentzer’s DF <13
78.6 91.3 8.7 21.4
Srivastava’s DF <4
78.6 82.6 17.4 21.4
England’s DF <15
89.3 52.2 30.6 10.7
Bruno’s DF <13
85.7 69.6 30.4 14.3
N = Negative, P = Positive Sensitivity (%) = TP / (TP + FP) Specificity (%) = TN / (TN + FN) FP% = FP / (FP + TP) FN% = FN / (FN + TP)
ROC-curve (MCV & RDW-SD) 100 90
Sencitivity (%)
80 70 60 50 40 30 20 10 0 100
80
60
40
20
0
1-specificity (%)
Fig. 2 ROC curves obtained with the use of MCV & RDW-SD
DISCUSSION The purpose of thalassemia screening is to provide information for counseling regarding thalassemia major in babies born to parents with heterozygous thalassemia, as well as its treatment. The present study was conducted to propose an efficient method of screening for such purposes in our patient demographic population. If thalassemia screening is to be conducted for screening of the general population, rather than patients, a large number of samples must be assayed. In the light of the potential high incidence of β thalassemia (as high as 20% in our area), it is important to increase screening efficiency. Increasing the sensitivity for reduction of overlooked cases would decrease the specificity, resulting in a large number of false positive cases which, in turn, increases the number of additional tests to establish diagnosis, such as HGB electrophoresis testing and, hence, an increase in examination cost and decrease in work efficiency. In the present study, discrimination between normal and patient groups was relatively easy, on condition that the patient group comprised samples with microcytic anemia
as diagnosed with MCV ≤ 80 (MCV is known to serve as an index for microcytic RBC). The validity of MCV ≤ 80 was thus confirmed. In other words, it can be concluded that MCV ≤ 80 is of high specificity for elimination of the normal group. On the other hand, RDW-SD < 32 alone is ineffective because it does not enable discrimination of normal or normocytic IDA, etc., although it is effective in the β thalassemia / IDA group with MCV ≤ 80. Regarding RDW, the CV method (RDW-CV) is commonly known and the SD method is especially used in Japan. The Sysmex K-4500 has the function to provide the both parameters. As a result of this study, the SD method was found to be more effective for β thalassemia screening. The principles of derivation of RDW-CV and RDW-SD can be compared as follows: RDW-CV is obtained as the value of 1 SD on the assumption of a normal distribution of RBC histograms, while RDW-SD is obtained as the distribution width of the most frequent value of RBC histograms at 20% frequency. In other words, RDW-CV reflects the distribution width for a relatively upper portion of the RBC distribution curve, and RDW-SD for a relatively lower portion. In general, IDA is often accompanied by anisocytosis, and RDW-SD can be used as an index for this abnormality. It is, therefore, reasonable to use RDW-SD for discrimination between IDA and β thalassemia, which usually does not show such heterogeneity in cell size. If patients with anisocytosis are present in the β thalassemia patient group, however, discrimination from IDA is impossible so that the sensitivity is slightly decreased. On the other hand, because RDW-CV tends to increase in IDA, IDA and thalassemia can be discriminated at relatively high sensitivity. In IDA, however, specificity lessens because of the presence of patients with low RDW-CV values despite irregular RBC size. Paterakis, et al. investigated thalassemia discrimination using RDW and MCV8). They used the Coulter S-Plus II analyzer in combination with the RDWCV method. Their subject patients were diverse, including α thalassemia, β thalassemia and IDA, and the necessity of two linear discriminant functions, including MCV and RDW, respectively, complicated the discrimination.
− 113 −
Sysmex Journal International Vol.8 No.2 (1998)
They suggested discrimination between IDA, α/β thalassemia group and normal group by their method. In the present study as well, the RDW-CV method was also assessed, but improved results could not be obtained due to the low specificity. In the study by Bessman9), the distributions of heterozygous β thalassemia, heterozygous α thalassemia, and IDA were examined with the RDW data (from the Coulter S-Plus analyzer) and the RDW-CV% data (obtained with a further size distribution analysis) were plotted on the ordinate and abscissa, respectively. The term RDW-CV, as used herein, is defined as half the particle size distribution width providing a percentile from 16 to 84% for the particle size distribution frequency determined by Gaussian distribution approximation. It is closer to the RDW-SD from K-4500 than the RDW from the S-Plus. He also demonstrated that the use of RDW resulted in an overlap of the distribution of β thalassemia and IDA, while the use of RDW-CV did not. This finding agrees with our findings in that good results were obtained with RDW-SD in the present study. We also obtained data from the Coulter JT hematology analyzer on some of the samples analyzed in the present study. The results obtained using the RDW from the Coulter JT analyzer were comparable to those obtained using the RDW-CV data from K-4500 (data not shown). The coefficients and cut-off values used in previously reported discriminant function methods have been determined on the basis of samples used in the respective studies; some involved the use of β thalassemia alone, and others included α thalassemia. Such values can not directly serve well for investigation. Although we set our own cut-off values, we could not reconsider the reported coefficients. Since thalassemia occurs in different forms among different localities, including genotypes, we do not think that the findings from a particular locality is always applicable to other localities. It is very difficult to set DF, parameters, and cut-off values for screening that are universal all over the world. It is desirable that cutoff values and DF coefficients be changed by demographics or by laboratory patient population to determine a DF that can serve as a diagnostic guideline, whenever possible.
CONCLUSION 1) The validity of β thalassemia screening, based on information from a hematology analyzer, was assessed at Nair Hospital in India. On the assumption that when the combined requirement of MCV ≤ 80 and RDW-SD < 32 was satisfied, the patient was positive for β thalassemia, the sensitivity and specificity were 88.5% and 93.5%, respectively, suggesting the potential of the present method for screening. However, the cutoff values remain to reconsidered in view of local variation. 2) The results suggest that RDW-CV, provided by both Coulter S-Plus and Sysmex K-4500 with the same calculation method, was inappropriate to discriminate β thalassemia with the others due to its low specificity. 3) RDW-SD from Sysmex K-4500 is calculated similar to the Bessman’s RDW-CV and found to be potentially more appropriate them the other for the β thalassemia discrimination.
ACKNOWLEDGEMENTS Mr. Vazirani, Transasia Bio-Medicals Ltd., is gratefully acknowledged for providing the instrument for the study, and Dr. R.M. Rowan for providing valuable information. References 1 ) England JM, Fraser P : Discrimination between iron-deficiency and heterozygous-thalassaemia syndromes in differential diagnosis of microcytosis. Lancet, 1 : 145-148, 1979. 2 ) England JM, Fraser P : Differentiation of iron deficiency from thalassaemia trait by routine blood-count. Lancet, 1 : 449-452, 1973. 3 ) Green R, et al. : A new red cell discriminant function incorporating RDW to differentiate iron deficiency anemia from thalassemia minor. Am J Clin Pathol, 90 : 507,1988. 4 ) Shine I, Lal S : A strategy to detect beta-thalassaemia minor. Lancet, 1 : 692-694.1977. 5 ) Bruno M, et al. : Relevance of red cell distribution width (RDW) in the differential diagnosis of microcytic anaemias. Clin lab Haematol, 13 : 141-151, 1991. 6 ) Mentzer WG : Differentiation of iron deficiency from thalassaemia trait. Lancet, 1: 882, 1973. 7 ) Srivastava PC, Bevington JM : Iron deficiency and-or thalassaemia trait. Lancet, 1 : 832, 1973. 8 ) Paterakis G, et al. : The performance characteristics of an expert system for the “On-Line” assesment of thalassemia trait and iron deficiency -micro hema screen. Blood Cells, 15 : 541-561,1989. 9 ) Bessman JD : Evaluation of automated whole-blood platelet counts and particle sizing. Am J Clin Pathol, 74 (2) : 157162,1980.
− 114 −