Tropical Medicine and International Health volume 11 no 8 pp 12951302 august 2006

doi:10.1111/j.1365-3156.2006.01679.x

Relationship between haemoglobin and haematocrit in the denition of anaemia

1,2, John J. Aponte1,2, Clara Mene ndez1,2, Jahit Sacarlal2,3, Pedro Aide2,4, Mateu Espasa1,2, Inacio Llorenc Quinto 2,4 1,2 Mandomando , Caterina Guinovart , Eusebio Macete2,5, Rosmarie Hirt6, Honorathy Urassa6, Margarita M. Navia1,2, Ricardo Thompson2,4 and Pedro L. Alonso1,2
1 2 3 4 5 6 Centre de Salut Internacional (CSI), Hospital Clinic/IDIBAPS, Universitat de Barcelona, Barcelona, Spain Centro de Investigac a a, Mozambique o em Sau de da Manhic Faculdade de Medicina, Universidade Eduardo Mondlane, Maputo, Mozambique Instituto Nacional de Sau de, Maputo, Mozambique Direcc a o Nacional de Sau de, Ministerio de Sau de, Maputo, Mozambique Ifakara Health Research and Development Centre, Ifakara, Tanzania

Summary

introduction Anaemia is the most frequent haematological disorder in childhood. The notion that denes naemia does not change throughout life, although parameters used for its evaluation show signicant variations during childhood. Haematocrit (Hct) (%) is usually dened as three times the value of haemoglobin (Hgb) (g/dl), while the clinical denition of anaemia is related to either an abnormal Hct or Hgb value. objective To evaluate the agreement between Hgb and Hct values in the denition of anaemia, the relationship between these two parameters and their age-dependence. methods The Hct and Hgb paired values from children aged 218 months from Ifakara (Tanzania) and children aged 14 years from Manhic a (Mozambique) were analysed. Haematological determinations of the Manhic a samples were done using a KX-21N cell counter (Kobe, Japan) and Ifakara samples were analysed in a semiautomatic cell counter (Sysmex F800 microcell counter, TOA Medical Electronics, Kobe, Japan). The j-statistic was used to calculate the agreement between anaemia denitions in each group. Crude and multivariate relationship between Hct and Hgb levels were analysed by linear regression model estimation. The age-dependence of the crude ratio (Hct/Hgb) was analysed using linear regression models and fractional polynomials. results The prevalences of mild and moderate anaemia as dened by Hgb levels in the Manhic a group were 61% and 6%, respectively, and 41% and 2% by Hct. In the Ifakara group these were 74% and 10%, respectively, by Hgb and 42% and 3% by Hct, respectively. Agreement between mild and moderate anaemia denitions made up from Hgb or from Hct levels were from fair to moderate. Hct levels decreased with age for high Hgb levels, whereas they increased for low Hgb levels. The classication of cases is improved when higher age-related cut-off values for Hct are used. The crude relationship between Hct and Hgb levels was signicantly different from 3, and this was modied by age. The evaluation of the age-dependence ratio (Hct/Hgb) showed a non-linear relationship with an asymptotic trend to 3. conclusions Measurement of haematocrit count is easy and can be performed in most rural health care centres. However, the corresponding Hgb levels cannot be derived with an acceptable accuracy using the value 3 as a conversion factor. Furthermore, the commonly assumed equivalent cut-off points for anaemia denitions need to be re-evaluated. keywords anaemia, agreement, haemoglobin, haematocrit

Introduction Anaemia is the most frequent haematological disorder during childhood. While the absolute number of anaemia associated deaths during the rst years of life is not sufciently documented, severe anaemia is unquestionably

associated with an increased risk of infant morbidity and mortality in sub-Saharan Africa (Brabin et al. 2001). Although the notion that denes anaemia (a decrease of circulating erythrocitic mass) does not change throughout life, the parameters used for its evaluation show signicant variations during childhood. 1295

2006 Blackwell Publishing Ltd

Tropical Medicine and International Health L. Quinto et al. Relationship between haemoglobin and haematocrit

volume 11 no 8 pp 12951302 august 2006

According to WHO estimations, nearly 1300 million people, that is about 30% of the worlds population, were anaemic in the 1980s (DeMaeyer & Adiels-Tegman 1985). It is estimated that, at the start of this century, this number has increased up to two billion, or 40% of the worlds population (ACC/SCN 2000; Stoltzfus & Dreyfuss 2000). The geographical distribution of disease shows areas of high prevalence, so that the burden of anaemia is very disproportionate, with 42% of pre-school children affected in non-industrialised countries, compared with 17% in industrialised ones (ACC/SCN 2000). In sub-Saharan Africa, between one and two-thirds of the children younger than 5 years suffer from anaemia (DeMaeyer & AdielsTegman 1985; ACC/SCN 2000). Furthermore, anaemia in childhood accounts for more than one half of hospital paediatric mortality in areas with intense malaria transmission (Lackritz et al. 1992; Schellenberg et al. 1999). The WHO denition of anaemia in children between 6 and 60 months of age is a haemoglobin (Hgb) level lower than 11 g/dl, which is equivalent to a haematocrit (Hct) lower than 33% (Stoltzfus & Dreyfuss 2000; Bain & Bates 2001). The Hgb provides a direct measure of the oxygen carrying capacity of the blood, whereas the Hct provides an indirect one. Besides, their calculation depends on the method used for their determination. Both parameters can be assessed either with an automated blood-counter or by manual methods such as microhaematocrit readings for Hct, or colorimetric methods. The Hgb estimates the erythrocitic function and is more stable to plasma volume changes such as dehydration, which makes it somehow more reliable for the assessment of anaemia. Unfortunately, in many settings automated methods for Hgb determinations are not available and rough values are estimated using observed Hct levels, which is a simpler and cheaper approach (specially in studies carried out away from western hospitals/centres, where it is often difcult to have complete haemograms). To conrm whether the commonly used relationship between Hgb and Hct is really 3 [Hct (%) Hgb (g/dl) 3] (Bain & Bates 2001), in this study we have evaluated the agreement between these two parameters in the denition of anaemia using a series of samples of children living in two malaria-endemic areas in subSaharan Africa.

a total of 5397 samples, 2474 corresponded to children from Manhic a District (Maputo province) in southern Mozambique who were taking part in the RTS,S/AS02A malaria candidate vaccine trial (Alonso et al. 2004). The remaining samples were taken at ve cross-sectional surveys, done during the follow-up of 847 infants at 2, 5, 8, 12 and 18 months of age as part of a study to evaluate the effect of chemoprophylaxis and iron supplementation on malaria and anaemia during the rst year of life, carried out in Ifakara town (Tanzania) (Menendez et al. 1997). The Manhic a study was carried out at the Centro de Investigac a a (CISM) between April de da Manhic o em Sau 2003 and May 2004. The CISM runs a demographic surveillance system in the study area (Manhic a DSS 2002). Lists of potentially eligible resident children were produced from this census. All of them were visited at home; parents or guardians were read information sheets and criteria for recruitment was checked, including conrmed residency in the study area and full immunisation with EPI vaccines. Interested parents or guardians were invited to the Manhic a Health Centre or the Ilha Josina Health Post. Screening included a brief medical history, physical examination and nger-prick blood sampling for haematology and biochemistry tests. The results of these blood tests were the ones used in the present study, independently of the participation of the child in the malaria candidate vaccine trial. The Ifakara study was carried out at the St. Francis Designated District Hospital (SFDDH) between January 1995 and October 1996. Women who gave birth at SFDDH and who reported being permanent residents of the Ifakara town were invited to participate in the study. On admission to the hospital for delivery, an eligible woman was given a copy of the consent letter (in Kiswahili), which included detailed information on procedures and the potential risks and benets of the study. Laboratory methods Haematological determinations of the Manhic a samples were done using a KX-21N cell counter (Kobe, Japan), which allows the determination of blood parameters from total blood in EDTA, including among others Hgb and Hct. Blood parameters of the Ifakara samples were analysed in a semiautomatic cell counter (Sysmex F800 microcell counter, TOA Medical Electronics, Kobe, Japan). Statistical methods Mild anaemia was dened as an Hgb level lower than 11 g/ dl, while moderate anaemia was dened as an Hgb level below 8 g/dl. The cut-off points for Hct levels were 33%

Materials and methods Subjects and sampling Blood samples from children from two malaria endemic areas in sub-Saharan Africa were included in the study. From 1296

2006 Blackwell Publishing Ltd

Tropical Medicine and International Health L. Quinto et al. Relationship between haemoglobin and haematocrit

volume 11 no 8 pp 12951302 august 2006

for mild and 25% for moderate anaemia, respectively. The Ifakara and Manhic a samples were analysed separately. Agreement between anaemia denitions was evaluated using the j-statistic and the results interpreted using the classication proposed by Landis and Koch (1977). Linear regression models were estimated in order to evaluate the relationship between Hgb and the Hct values. Robust variance estimates were calculated for the Ifakara samples, to take into account that there was more than one sample from the same child. For each group, the appropriate linear regression models were estimated to evaluate a crude relationship between Hgb and Hct or a multivariate relationship between these two parameters, adjusted by age, gender and Hgb per age interaction. Model estimation was performed without constant term (intercept), and nested models (multivariate models vs. crude models) were compared by the decrease in the residual sum of squares, using an F-test. Differences between the estimated Hgb coefcients and the theoretical value of 3 were performed using the Wald test. The change of the Hct (%)/Hgb (g/dl) ratio with age, was evaluated by a linear regression model for the Manhic a samples. Deviances from linearity in the estimated models were evaluated by fractional polynomials of rst or second degree (Royston 1991;Royston & Altman 1994),andthe best powers for each curve were estimated by maximum likelihood. All analyses were done using Stata Version 8.2 (2003).

Results The 2474 samples from Manhic a came from children with a mean age of 34.1 months (95% CI: 33.6, 34.7), of whom 52% were boys. The remaining 2923 blood samples were from 847 infants from Ifakara, 50% boys, collected during cross-sectional surveys at 2, 5, 8, 12 and 18 months of age with a median of four samples per child. In Manhic a, using the Hgb denition, 1497 (61%) of the children were classied as having mild, and 137 (6%) as having moderate anaemia, whereas using the Hct denition, 1002 (41%) were classied as having mild and 57 (2%) as having moderate anaemia. For mild anaemia there were 34% (505/1497) of children classied as anaemic by Hgb that were not considered anaemic using the Hct level. For moderate anaemia, the agreement was similar as 59% (81/137) of the children classied as having moderate anaemia by Hgb did not classify as anaemic when using the Hct level (Table 1). Results from the Ifakara samples were similar, although the agreement tended to be lower than with the Manhic a ones. Prevalences of mild and moderate anaemia dened by Hgb levels were 74% and 10%, respectively, by Hgb and 42% and 3% by Hct, respectively. In this group, 45% (979/2172) and 72% (206/287) of samples classied by Hgb levels with mild

Table 1 Agreement between haemoglobin and haematocrit in the denition of anaemia

Observed Agreement (%) Hgb < 11 g/dL Hct < 33% MANHIC A No Yes IFAKARA No Yes No 967 10 709 42 Yes 505 992 979 1193

Expected Agreement by chance (%)

Kappa

Landis & Koch interpretation

79

48

0.60

Moderate

65

46

0.35

Fair

Hgb < 8 g/dL Hct < 25% MANHIC A No Yes IFAKARA No Yes No 2336 1 2635 1 Yes 81 56 206 81

97

92

0.56

Moderate

93

88

0.41

Moderate

2006 Blackwell Publishing Ltd

1297

Tropical Medicine and International Health L. Quinto et al. Relationship between haemoglobin and haematocrit

volume 11 no 8 pp 12951302 august 2006

and moderate anaemia, respectively, were not detected using Hct levels (Table 1). Figure 1 shows the relationship between Hgb and Hct levels for the Manhic a and Ifakara samples, respectively, using different regression models. In both cases, the model including an interaction between the age and Hgb was the one with the best t. Table 2 shows the estimated parameters and 95% CI for these models. The Wald test showed that the coefcient for Hgb was signicantly different from 3, P < 0.0001 both for Manhic a and Ifakara. According to the estimated models, Hct decreased with age for high Hgb levels, whereas it increased for low levels (Figure 1). Figure 2 shows the Hct for Hgb equal to 11 and 8 g/dl (mild and moderate anaemia using Hgb test) in the Manhic a group. The Hct values at 2, 5, 8, 12 and 18 months for the Ifakara group were 36.7, 36.6, 36.6, 36.5 and 36.4%, respectively, when Hgb was 11 g/dl and 27.1, 27.8, 28.4, 29.3 and 30.6% when Hgb was 8 g/dl. Using these Hct values as an age-related cut-off, we obtained a new classication of cases whose sensitivity,

specicity, positive predictive and negative predictive values, using the Hgb test as a Gold Standard, are shown in Table 3. Percentages of cases classied as anaemic by Hgb but not by Hct were lower than those described before. Thus, when Hct levels were used with the estimated age-related cut-off, 8% (118/1497) and 21% (29/137) of the samples identied as anaemia cases (mild and moderate, respectively) by Hgb were not in the Manhic a group and 8% (182/2172) and 36% (104/287) in the Ifakara group. The crude ratio age-dependency, Hct (%)/Hgb (g/dl), showed a non-linear relationship with an asymptotic trend to 3. This ratio ranged from 3.34 at 10 months and decreased with age until it became stable at around 3.13 (see Figure 3). The best t for this ratio was a grade two model with powers equal to 2 and 3. Discussion Both Hct and Hgb levels could be affected by factors such as the method and equipment used for its

70

MANHIA 1 year 2 years 3 years 4 years

10 2

20

Haematocrit (%) 30 40 50 60

10 12 14 Haemoglobin (g/dL)

16

18

20

70

IFAKARA

Haematocrit (%)

2 months 5 months 8 months 12 months 18 months

10 2

20

30

40

50

60

10 12 14 Haemoglobin (g/dL)

16

18

20

Observed values Crude Model

Hct = 3 x Hgb Multivariate Model

Figure 1 Relationship between the Hgb and Hct levels.

1298

2006 Blackwell Publishing Ltd

Tropical Medicine and International Health L. Quinto et al. Relationship between haemoglobin and haematocrit

volume 11 no 8 pp 12951302 august 2006

Table 2 Coefcients and 95% condence intervals (CI) of the linear regression models estimated from each group MANHIC A Coefcients Haemoglobin (g/dl) Age (months) Haemoglobin Age 3.27 0.22 )0.02 P <0.001 <0.001 <0.001 95% CI 3.26 to 3.29 0.20 to 0.23 )0.02 to )0.02 IFAKARA Coefcients 3.34 0.83 )0.08 P <0.001 <0.001 <0.001 95% CI 3.31, 3.36 0.73, 0.93 )0.09, )0.07

40

Haematocrit (%) 30 35

Hct cut-off (mild anaemia) = 36.0

0 - 0.03 *Age (months)

20

25

Hct cut-off (moderate ana

emia) = 26.18 + 0.03 *Ag

e (months)

12
Figure 2 The Hct values equivalent to Hgb levels of 8 and 11 g/dl by age.

18

24

30

36 42 Age (months)

48

54

60

Haemoglobin = 8 g/dL

Haemoglobin = 11 g/dL

Table 3 Classication of cases by age-related cut-off for Hct using Hgb tests as gold standard Hct < age-related cut-off (mild) MANHIC A No Yes IFAKARA No Yes Hgb < 11 g/dL No 794 183 493 258 Yes 118 1379 182 1990

Sensitivity Specicity (%) (%) PPV (%) NPV (%)

92

81

88

87

92

66

89

73

Hgb < 8 g/dL Hct < age-related cut-off (moderate) No Yes MANHIC A No Yes IFAKARA No Yes 2300 37 2500 136 29 108 104 183

79

98

74

99

64

95

57

96

PPV: Positive Predictive Value NPV: Negative Predictive Value.

2006 Blackwell Publishing Ltd

1299

Tropical Medicine and International Health L. Quinto et al. Relationship between haemoglobin and haematocrit

volume 11 no 8 pp 12951302 august 2006

Fractional Polynomial of 2nd degree and powers 2 and 3 Haematocrit (%) / Haemoglobin (g/dL) 4.0 2.0 12 2.5 3.0 3.5

18

24

30

36 42 Age (months)

48

54

60

Observed values 95% Confidence Interval

Fitted values

Figure 3 Evolution of the Hct (%)/Hgb (g/dl) ratio.

determination, environment or subjects differences that may cause a spurious change in the measured value and lead to inaccuracies (Keen 1998). In this study, we have evaluated this relationship in two groups of samples, from different countries and measured with different equipments (although both automated). Results obtained in both are quite similar. Differences observed between the two studies might be due to several causes, mainly to in the age of the children studied (in Ifakara, children were younger than in Manhic a), the laboratory equipments, malaria endemicity and aetiologies of anaemia. In accordance with what has been previously shown, in our study Hgb and Hct tests at commonly used cut-offs, detected different prevalences of anaemia in the same population (Graitcer et al. 1981). In a previous study, with more than 13.000 samples, it was reported that in general, anaemia prevalence is higher when calculated using Hgb levels than when using Hct levels (Graitcer et al. 1981). Our results are in accordance with these ndings, specially with respect to mild anaemia, for which differences were around 20% in the Manhic a group (61% vs. 41%) and 32% in the Ifakara group (74% vs. 42%). Determining Hgb seems to be a more sensitive way to determine mild and moderate anaemia, bearing in mind that there is not a gold standard. The signicant number of anaemic children diagnosed using Hgb levels, not detected using Hct levels, might be due to the spurious elevation of Hct caused by poorly packed iron decient cells (Graitcer et al. 1981). Nevertheless, as shown in Figure 1, Hgb and Hct are closely related, although the 1300

usual transformation three times the Hgb (g/dl) equals the haematocrit (%) is not accurate. Thus, the correspondence between the cut-off points is not adequate, requiring higher values for the Hct cut-off point to obtain the same sensitivity than for Hgb. Regression models show that the association between the cut-offs of Hgb and Hct not only depends on the age but also on the Hgb level. In the Manhic a group, the 11 g/dl cut-off for Hgb in relation to Hct decreases with age (from around 36% before 12 months of age to 34% at 60 months of age). Whereas, as shown in Figure 2, the 8 g/dl Hgb cut-off in relation to the Hct increases with age from 26% to 28% for the same ages. For both Manhic a and Ifakara, the predicted Hct levels for 8 and 11 g/dl of Hgb are far from the commonly used equivalent Hct levels of 24% and 33%, respectively. The decrease in the percentage of undetected cases is notorious when using the estimated Hct values according to age. In both groups we observed an 8% in mild anaemia with the estimated values (Table 3) vs. a 34% and 45% observed when using the standard cut-off values (Table 1). With moderate anaemia, a 59% in the Manhic a group is reduced to 21%, while in the Ifakara group it goes down from 72% to 36%. The reason why Hct values are commonly used within the clinical practice in rural African settings has to do with the fact that they are easy and cheap to do using manual techniques. In this study we obtained all values using a Sysmex, which is an expensive system. In general, in a rural setting, the running costs for Hct are very low (around 0.07 USD) compared to what a hemocue would cost

2006 Blackwell Publishing Ltd

Tropical Medicine and International Health L. Quinto et al. Relationship between haemoglobin and haematocrit

volume 11 no 8 pp 12951302 august 2006

(around 2 USD). However, it does require more infrastructure including a centrifuge (around 3000 USD), a Hawskley Micro-Hct reader (around 200 USD) and electrical power supply. A hemocue only requires the apparatus in itself (around 1000 USD) as it can be powered with a battery. These are approximate numbers and can vary within countries, but in general in studies involving large populations it is cheaper to measure Hct. The relationship between Hct and Hgb is expressed with the Mean Corpuscular Hgb Concentration (MCHC). The MCHC varies depending on the type of anaemia. An increased MCHC is seen in spherocytosis but not in pernicious anaemia, whereas decreased levels may indicate iron deciency, blood loss, B6 deciency or thalassemia. It could be the case that obtaining a single conversion factor is not feasible, as the relationship depends on the prevalence of anaemia in each population and on the type of anaemia pre-dominating within it. These data show that Hgb levels cannot be derived from the Hct values with an acceptable accuracy using the general rule of dividing by 3. The relationship between Hgb and Hct is not exactly 3 and it changes with age during the rst years of life. Due to the lack of agreement, the commonly assumed equivalent cutoff points for anaemia denitions need to be re-evaluated. More information is needed for other age groups (adults) and different aetiologies of anaemia. This information is of relevance for both clinical diagnosis and management of anaemia cases, as well as for descriptive and intervention studies on anaemia. It will also help to plan more efciently anaemia control measures in the community. References
ACC/SCN (2000) Fourth Report on the World Nutrition Situation. ACC/SCN in collaboration with IFPRI, Geneva. Alonso PL, Sacarlal J, Aponte JJ et al. (2004) Efcacy of the RTS,S/AS02A vaccine against Plasmodium falciparum infection and disease in young African children: randomised controlled trial. Lancet 364, 14111420.

Bain BJ & Bates I (2001) Basic haematological techniques. In: Practical Haematology, 9 Edn (eds SM Lewis, BJ Bain & I Bates) Churchill Livingstone, Edinburgh, pp. 1946. Brabin BJ, Premji Z & Verhoeff F (2001) An analysis of anemia and child mortality. The Journal of Nutrition 131, 636S645S. DeMaeyer E & Adiels-Tegman M (1985) The prevalence of anaemia in the world. World Health Statistics Quarterly 38, 302316. Graitcer PL, Goldsby JB & Nichaman MZ (1981) Hemoglobins and hematocrits: are they equally sensitive in detecting anemias? The American Journal of Clinical Nutrition 34, 6164. Keen ML (1998) Hemoglobin and hematocrit: an analysis of clinical accuracy. Case study of the anemic patient. American Nephrology Nurses Association Journal 25, 8386. Lackritz EM, Campbell CC, Ruebush TK et al. (1992) Effect of blood transfusion on survival among children in a Kenyan hospital. Lancet 340, 524528. Landis JR & Koch GG (1977) The measurement of observer agreement for categorical data. Biometrics 33, 159174. Manhic a DSS (2002) Mozambique. In: Population and Health in Developing Countries. Vol. 1: Population, Health, and Survival at INDEPTH Sites. International Development Research Centre, Canada, pp. 189195. Menendez C, Kahigwa E, Hirt R et al. (1997) Randomised placebo-controlled trial of iron supplementation and malaria chemoprophylaxis for prevention of severe anaemia and malaria in Tanzanian infants. Lancet 350, 844850. Royston P (1991) Constructing time-specic reference ranges. Statistics in Medicine 10, 675690. Royston P & Altman DG (1994) Regression using fractional polynomials of continuous covariates: parsimonious parametric modelling. Applied Statistics 43, 429467. Schellenberg D, Menendez C, Kahigwa E et al. (1999) African children with malaria in an area of intense Plasmodium falciparum transmission: features on admission to the hospital and risk factors for death. The American Journal of Tropical Medicine and Hygiene 61, 431438. Stata Corporation (2003) Stata Statistical Software: Release 8.0. Stata Corporation, College Station TX, USA. Stoltzfus RJ & Dreyfuss ML (2000) Guidelines for the Use of Iron Supplements to Prevent and Treat Iron Deciency Anemia. INACG/WHO/UNICEF. Internation Life Sciences Institute, Washington DC, USA.

Corresponding Author Llorenc Quinto , CSI, Hospital Cl nic de Barcelona, C/Rossello 132 2n 2a. 08036 Barcelona, Spain. Tel.: +34 93 227 5706; Fax: +34 93 227 9853; E-mail: lquinto@clinic.ub.es

2006 Blackwell Publishing Ltd

1301

Tropical Medicine and International Health L. Quinto et al. Relationship between haemoglobin and haematocrit

volume 11 no 8 pp 12951302 august 2006

Relation entre he moglobine et he matocrite dans la de nition de lane mie objectifs Evaluer la concordance entre les valeurs dhe moglobine et dhe matocrite dans la de nition de lane mie, la relation entre ces deux parame ` tres et leur association avec la ge. thodes Des valeurs paires dhe me a au moglobine et dhe matocrite denfants de 2 a ` 18 mois dIfakara en Tanzanie et denfants de 1 a ` 4 ans de Manhic Mozambique ont e a ont e te analyse es. Les de terminations he matologiques des e chantillons de Manhic te re alise es en utilisant un compteur de cellules KX-21N (Kobe, Japon) et ceux de Ifakara par un compteur semi automatique (Sysmex F800 microcell counter, TOA Medical Electronics, Kobe, Japon). La valeur statistique Kappa a e te utilise e pour calculer les concordances entre les de nitions dane mie dans les deux groupes. Les associations brutes et multivarie es entre les taux dhe matocrite et dhe moglobine ont e te analyse es en utilisant un mode ` le destimation par re gression line aire et des polynomiales fractionnelles. sultats Les pre re valences dane mie faible et mode re e e taient de 61%et 6%respectivement de nies par les taux dhe moglobine dans le groupe de Manhic a et de 42%et 3%respectivement par de nition base e sur lhe matocrite. Dans le groupe dIfakara, elles e taient de 74%et 10%respectivement par de nition base e sur lhe moglobine et 42%et 3%respectivement par de nition base e sur lhe matocrite. Les concordances entre les de nitions base es sur les taux dhe moglobine ou dhe matocrite pour lane mie faible et mode re e e taient de faibles a ` mode re es. Les taux dhe matocrite diminuaient avec la ge pour les taux e leve s dhe moglobine, mais ils augmentaient pour des taux e leve s dhe moglobine. La classication des cas est ame liore e lorsque des valeurs limites, plus e leve es et lie es a ` la ge e taient utilise es. La relation brute entre taux dhe matocrite et dhe moglobine e tait signicativement diffe rente de 3 et cela e tait modie par la ge. Le valuation du rapport he matocrite/he moglobine en fonction de la ge a re ve le une relation non line aire avec une tendance asymptotique vers 3. conclusions La mesure du taux dhe matocrite est facile et peut e tre re alise e dans la plupart des centres de soin de sante ruraux. Cependant, les niveaux correspondants dhe moglobine ne peuvent pas e tre de duits avec une pre cision acceptable en utilisant la valeur 3 comme facteur de conversion. De plus, les valeurs limites ge ne ralement assume es comme e quivalents pour la de nition de lane mie devraient e tre re e value es. s ane mots cle mie, concordance, he moglobine, he matocrite

Relacio n entre hemoglobina y hematocrito en la denicio n de anemia objetivo Evaluar la concordancia entre los valores de hemoglobina y hematocrito en la denicio n de anemia, la relacio n entre estos dos para metros y su dependencia de la edad. todos Se analizaron los valores pareados de hematocrito y hemoglobina de nin me os de 2 a 18 meses provenientes de Ifakara (Tanzania) y nin os entre 1 y 4 an a (Mozambique). Las determinaciones hematolo a se hicieron utilizando un contador gicas de las muestras de Manhic os provenientes de Manhic hematolo gico KX-21N y las muestras de Ifakara se analizaron en un contador semiautoma tico Sysmex F800 microcell counter. Se utilizo el ndice Kappa para calcular la concordancia entre las deniciones de anemia en cada grupo. La relacio n cruda y multivariada entre niveles de hematocrito y hemoglobina se analizo mediante estimacio n del modelo de regresio n linear. La dependencia por edad de la proporcio n cruda (hematocrito/hemoglobina) se calculo mediante modelos de regresio n linear y polinomios fraccionados. resultados La prevalencia de anemia leve y moderada denida por los niveles de hemoglobina en el grupo de Manhic a fue del 61% y 6%, respectivamente y 41% y 2% por hematocrito. En el grupo de Ifakara estas fueron respectivamente de 74% y 10% por hemoglobina y 42% y 3% por hematocrito. La concordancia entre las deniciones de anemia leve y moderada construidas a partir de los niveles de hemoglobina o hematocrito fueron de aceptable a moderada. Los niveles de Hct disminuyeron con la edad para niveles altos de Hgb, mientras que aumentaron para niveles bajos de Hgb. La clasicacio n de los casos mejora cuando se utilizan valores de corte relacionados con la edad ma s altos. La relacio n cruda entre niveles de hematocrito y hemoglobina fue signicativamente diferente a 3 y esto fue modicado por edad. La evaluacio n de la proporcio n dependiente de edad (hematocrito/hemoglobina) mostro una relacio n no linear con una tendencia asinto tica a 3. conclusiones Medir el hematocrito es fa cil y puede hacerse en la mayor a de centros sanitarios rurales. Sin embargo los niveles de hemoglobina correspondientes no pueden derivarse con una precisio n aceptable utilizando el valor de 3 como factor de conversio n. Ma s au n, los puntos de corte para anemia, comu nmente asumidos como equivalentes, necesitan ser reevaluados. palabras clave anemia, concordancia, hemoglobina, hematocrito

1302

2006 Blackwell Publishing Ltd