Professional Documents
Culture Documents
Validity Assessment of Nine Discriminant Functions Used
Validity Assessment of Nine Discriminant Functions Used
Summary
Iron deficiency anemia (IDA) and thalassemia minor are two of the most common causes of microcytic
anemias worldwide. Because of similar red blood cell count parameters and blood picture, it was
imperative to develop other measures that would differentially and correctly diagnose these two
anemias. Several mathematical formulas and simple RBC indices have been introduced as simple, fast
and inexpensive means of providing differential diagnosis for IDA and thalassemia minor. The Objective
of this study was to apply and compare nine well-documented discriminant functions on a population of
153 confirmed cases of microcytic anemias (IDA n ¼ 56, b-thalassemia minor n ¼ 47 and a-thalassemia
n ¼ 50) and to measure validity using Youden’s Index. The results show that England and Fraser
(E & F) Index had the highest Youden’s Index value (98.2) in correctly differentiating between IDA and
a- and b-thalassemia minor, while Shine and Lal Index was found ineffective in differentiating between
microcytic anemias in our population. E & F Index showed with great sensitivity and specificity to be the
best discriminant function to differentiate between IDA and thalassemia minor cases.
ß The Author [2006]. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oxfordjournals.org 93
doi:10.1093/tropej/fml070 Advance Access Published on 13 December 2006
S. M. ALFADHLI ET AL.
TABLE 1
Mathematical discriminant functions used to differentiate between iron deficiency anemia and thalassemia minor
diagnosed to have IDA. Forty-seven patients In all cases studied RBC count was found to be lower
were diagnosed to have b-thalassemia minor based than normal range in IDA (4.32 0.42 1012 l1),
on low Hb and MCV levels, normal serum iron while values were within normal range in b- and
and elevated levels of Hb A2 (>3.3%). The remaining a-thalassemia (5.39 0.89 and 4.98 0.82, respec-
50 patients were diagnosed to have a-thalassemia tively). Hemoglobin level was decreased in the three
based on the detection of Hb H inclusion bodies anemias studied, however, the level of decrease in IDA
in peripheral blood smear stained with Brilliant was greater than that in thalassemia. MCV was also
Crystal Blue stain. low in all cases although the level of decrement in
CBC results were obtained using Beckman b-thalassemia (63.4 7.12 fL) cases was greater than
Coulter HMX Automated Hematology Analyzer, that found in IDA (67 5.9 fL) and a-thalassemia
USA. Serum iron, TIBC and transferring saturation (66.5 6.45 fL). As expected serum iron levels were low
were determined using BN ProSpecÕ System from in IDA (3.4 1.2 mmol l1), while Hb A2 levels were
Dade Behring, Inc, Germany, while serum ferritin high in b-thalassemia minor (5.97 1.32%). All cases
was measured using Elecsys System from Roche of a-thalassemia showed Hb H inclusions confirming
Diagnistics, Germany. Hb A2 levels were determined the proposed diagnosis. IDA cases were also confirmed
using b-thal Hb A2 Quick Column kit from Helena by high TIBC levels with an average of 3.8 0.4 g l1
Biosciences, UK. (normal range 2.1–3.5 g l1) and low transferring
Patients, results were analyzed individually using saturation levels with an average of 3.7 1.3%
the nine discrimination functions in this study. (normal range 20–40%). Ferritin (measured only in 28
Sensitivity, specificity and Youden’s Index (a mea- cases of IDA) was low with an average of
sure of validity) [(sensitivity þ specificity) 100] were 4.8 1.8 ng ml1 (normal range 13–400 ng ml1).
calculated for each function. Data were analyzed The nine investigated discriminant functions were
using Microsoft Office Excel program 2003. applied on all microcytic anemia cases in this study,
In addition to RBC count and red blood cell including a-thalassemia cases (Table 3). E & F Index
distribution width (RDW), the following mathema- identified 100% of b- and a-thalassemia cases and
tical functions were also included in our analysis: almost 100% (98.2%) of IDA cases. Although, RBC
Mentzer Index (MI), Shine and Lal (S & L) Index, count perfectly identified (100%) cases of IDA, only
England and Fraser (E & F) Index, Green and King 78.7 and 58% of b- and a-thalassemia cases
(G & K) Index, RDW Index (RDWI), Srivastava respectively, correctly fit the cut off value of this
Index (SI) and Ricerca Index (RI) (Table 1). index.
S & L Index had the lowest number of correctly
identified cases of IDA as only one case was correctly
Results identified from 56 diagnosed cases (1.7%), while
The means and standard deviations of CBC the RDW had the lowest number of correctly
parameters, serum iron and Hb A2 in the 153 identified cases of b- and a-thalassemia (6.3 and
patients investigated in this study are shown in 12%, respectively). Almost all discriminant functions
(Table 2). (except for RDW) had higher percentages of
5.97 1.32
2.53 0.59
correctly identified cases of b-thalassemia compared
2 0.4
2–3.3%
Hb A2
with a-thalassemia.
It is worth noting that thalassemia cases were
compared separately to IDA cases. The discriminant
functions are not intended to be used to differentiate
36.54 12.65
33.76 14.55
between b- and a-thalassemia.
6–37 mmol/ l
3.4 1.2
In order to determine, the best function to be used
S.Fe
in differentiating IDA from b- and a-thalassemia,
sensitivity, specificity and Youden’s Index were
calculated (Table 4). In differentiating IDA from
b-thalassemia, the Youden’s Index from highest
17.41 2.36
17.54 4.13
11.5–14.5%
17.06 2 SI > RI > RDWI > G & K > RDW > S & L. In differ-
RDW
Discussion
Several studies have derived discriminant functions
based on RBC indices that can be used to differ-
entiate IDA from thalassemia minor. The differential
Results are expressed as mean SD
0.29 0.03
0.34 0.05
0.33 0.06
4.32 0.42
5.39 0.89
4.98 0.82
TABLE 3
Cut-off values and number of cases correctly identified by the nine discriminant functions for the
three anemias investigated
Discriminant function Cut–off values IDA (n ¼ 56) b-thals (n ¼ 47) a-thals (n ¼ 50)
Except for S & L Index and RDW, all other of various laboratory parameters. Differentiation
functions had a lower Youden’s Index when IDA and between IDA and ACD. Neth J Med 2001;59:270–9.
a-thalassemia were compared than when IDA and 7. Mentzer WC. Differentiation of iron deficiency from
b-thalassemia were compared (Youden’s Index for thalassemia trait. Lancet 1973;1:882–4.
8. Shine I, Lal S. A strategy to detect beta-thalassemia
E & F Index was the same in both comparisons).
minor. Lancet 1977;1:692–4.
This suggests, that it is probably more difficult to 9. England JM, Fraser P. Discrimination between
differentiate between IDA and a-thalassemia than it iron-deficiency and heterozygous thalassemia
is to differentiate between IDA and b-thalassemia syndromes in differential diagnosis of microcytosis.
using discriminant functions alone (also note close Lancet 1979;1:145–8.
MCV, MCH and MCHC values; Table 2). Therefore, 10. Green R, King R. A new red blood cell discriminant
iron studies and Hb H detection must be incorpo- incorporating volume dispersion for differentiating iron