Estimating the prevalence of anaemia: a comparison of three methods.

Estimating the prevalence of anaemia: a comparison of three methods Mayang Sari,1 Saskia de Pee,2 Elviyanti Martini,3 Susilowati Herman,4 Sugiatmi,5 Martin W. Bloem,6 & Ray Yip7 Objective To determine the most effective method for analysing haemoglobin concentrations in large surveys in remote areas, and to compare two methods (indirect cyanmethaemoglobin and HemoCue) with the conventional method (direct cyanmethaemoglobin). Methods Samples of venous and capillary blood from 121 mothers in Indonesia were compared using all three methods. Findings When the indirect cyanmethaemoglobin method was used the prevalence of anaemia was 31–38%. When the direct cyanmethaemoglobin or HemoCue method was used the prevalence was 14–18%. Indirect measurement of cyanmethaemoglobin had the highest coefficient of variation and the largest standard deviation of the difference between the first and second assessment of the same blood sample (10–12 g/l indirect measurement vs 4 g/l direct measurement). In comparison with direct cyanmethaemoglobin measurement of venous blood, HemoCue had the highest sensitivity (82.4%) and specificity (94.2%) when used for venous blood. Conclusions Where field conditions and local resources allow it, haemoglobin concentration should be assessed with the direct cyanmethaemoglobin method, the gold standard. However, the HemoCue method can be used for surveys involving different laboratories or which are conducted in relatively remote areas. In very hot and humid climates, HemoCue microcuvettes should be discarded if not used within a few days of opening the container containing the cuvettes. Keywords Hemoglobinometry/methods; Anemia/epidemiology; Prevalence; Comparative study; Indonesia (source: MeSH ). Mots cle´s He´moglobinome´trie/me´thodes; Ane´mie/e´pide´miologie; Pre´valence; Etude comparative; Indone´sie (source: INSERM). Palabras clave Hemoglobinometrı´a/me´todos; Anemia/epidemiologı´a; Prevalencia; Estudio comparativo; Indonesia (fuente: BIREME ) Bulletin of the World Health Organization, 2001, 79: 506–511. Voir page 510 le re´sume´ en franc¸ais. En la pa´gina 510 figura un resumen en espan˜ol. Introduction Anaemia, defined as a reduced haemoglobin con- centration, is associated with increased perinatal mortality, increased child morbidity and mortality, impaired mental development, impaired immune competence, increased susceptibility to lead poison- ing, and decreased performance at work (1–4). To identify the most likely cause of anaemia, such as low iron intake or infection, distribution curves of the haemoglobin concentration of different groups within a population can be compared (5). In the developing countries of south-east Asia the prevalence of anaemia among pregnant women is as high as 60–70% (5). In Indonesia, a national survey in 1992 found that 64% of pregnant women and 56% of children under 5 years old were anaemic (6); a household survey in 1995 found that 51% of pregnant women and 41% of children under 5 years had anaemia (6). About 30% of female workers and 24–35% of schoolchildren also were anaemic (7). Estimates of the prevalence of anaemia depend on the methods used for assessing haemoglobin concentration and on the cut-off point applied: the cut-off point is different for different groups in a population (8, 9). For a prevalence survey, the choice ofmethod formeasuring haemoglobin concentration 1 Head of Nutrition, Helen Keller International Indonesia, PO Box 4338 Jakarta, Indonesia (email: msari@hki-indonesia.org). Correspondence should be addressed to this author. 2 Regional Nutrition Research Advisor, Helen Keller International Asia-Pacific Regional Office, Jakarta, Indonesia. 3 Senior Program Officer, Helen Keller International Indonesia, Jakarta, Indonesia. 4 Head of Biochemical Research Team, Nutrition Research and Development Center, Bogor, Indonesia. 5 Program Officer, Helen Keller International Indonesia, Jakarta, Indonesia. 6 Regional Director Asia-Pacific Regional Office, Helen Keller International Asia-Pacific Regional Office, Jakarta, Indonesia. 7 Head of Health and Nutrition, UNICEF, Beijing, China. Ref. No. 99-0222 Research 506 # World Health Organization 2001 Bulletin of the World Health Organization, 2001, 79 (6) and the decision whether to test venous or capillary blood depend not only on the performance of the method but also on the conditions under which the blood will be collected, such as the remoteness of the location, whether laboratory support is available, whether staff are experienced, and whether the study population is cooperative. Direct cyanmethaemoglobin measurement: the gold standard The gold standard for assessing haemoglobin concen- tration is the direct cyanmethaemoglobin method. However, this method requires that a laboratory with a spectrophotometer is available within a few hours’ travelling time from where the blood is collected and, to avoid inter-laboratory variability, it is necessary that all measurements are conducted by the same labora- tory. Thus, a different method is usually required for surveys that are conducted over a comparatively large area and/or in remote conditions. One possible method is the indirect measurement of cyanmethaem- oglobin, for which blood is dried on filter-paper for transport and then redissolved at the laboratory for measurement. Another method uses a new generation of haemoglobin photometer, the HemoCue (Hemo- Cue, Angelhom, Sweden). The HemoCue method can be used in the field to analyse blood collected in a microcuvette. The photometer is easy to transport because it is small and light; it is battery operated and gives consistent results (10). The aim of this study was to determine which of these two alternative methods for determining haemoglobin concentration (indirect measurement of cyanmethaemoglobin orHemoCue) ismore useful in large surveys in remote areas and which method performs better when compared with the direct measurement of cyanmethaemoglobin. Each meth- od was used for both venous and capillary blood collected from the same participants. Methods Participants This study was part of an end-line survey for monitoring the impact of a social marketing project in the area of the city of Banjarmasin that is under the governance of South Kalimantan. The study was carried out byHelenKeller International Indonesia in collaboration with the provincial health office in South Kalimantan in March and April 1998. The participants were 121mothers who had children aged under 5 years of age and who had participated in the end-line survey. Of these 121 mothers, 62 were in Banjarmasin and 59 were in Martapura (another city in South Kalimantan), the control area for the assessment of the social marketing campaign. Written informed consent was obtained from the participants after explaining the aim of the blood collection. The procedure for blood collection was approved by the Medical Ethical Committee of the Indonesian Ministry of Health. Study design Haemoglobin concentration was determined in six different ways for each participant. Both venous and capillary blood samples were collected, and the haemoglobin concentration of each sample was determined using direct measurement of cyan- methaemoglobin (the conventional spectrophoto- metry method), the indirect method (blood was dried on a paper-filter to preserve it for analysis in the laboratory 2 weeks after collection), and the HemoCue method. Blood collection Dry syringes were used to collect 3 ml of venous blood from the forearm. The blood was transferred to a glass tube and divided for assessment by the different methods. For capillary blood, the partici- pant’s fingertip was warmed, cleaned with alcohol, and punctured with a needle using an Autoclix apparatus. The first drop of blood was discarded; blood was then collected for the three methods using a Sahli pipette twice and a HemoCue microcuvette. For both the direct and indirect cyanmethaem- oglobin methods, duplicate measurements were made from the same blood sample. For the HemoCue method, only one measurement was carried out for capillary blood; for venous blood, two assessments were made, using an old and new microcuvette. Analysis of blood samples Anaemia was defined as haemoglobin concentration <120 g/l, as assessed by the direct method on venous blood. Direct cyanmethaemoglobin measurement of venous blood was used as the gold standard for determining the sensitivity and specificity of the other methods. Variability within samples was calculated from duplicate measurements (direct and indirect cyanmethaemoglobin methods only). Sys- tematic differences between the methods were assessed using samples from the same participant. For each participant, the difference between two methods was calculated and those differences were averaged at group level to assess whether they were different from zero. Direct and indirect cyanmethaemoglobin methods. For the direct method, 20 ml blood were mixed with 5 ml Drabkin’s solution in the field. After 2–4 hours the blood was analysed at the local laboratory by experienced technicians using a spectrophotometer. For the indirect cyanmethae- moglobin method, 20 ml blood was transferred onto Whatman filter-paper and dried at room temperature. After the blood had dried, the filter- paper was put into a small plastic bag (one sample per bag), which was placed in an envelope and sealed. The filter-papers were taken to the laboratory at the Nutrition Research and Develop- ment Center in Bogor, where they were analysed 2 weeks after being collected. The dried blood was then diluted in 5 ml Drabkin’s solution, and the 507Bulletin of the World Health Organization, 2001, 79 (6) Estimating the prevalence of anaemia haemoglobin concentration was determined using a spectrophotometer. Both direct and indirect de- terminations used the Merck-test 3317 (Merck, Darmstadt, Germany) for determining haemoglo- bin concentration. HemoCue method. For the HemoCue meth- od, one drop of blood was collected in a HemoCue microcuvette and the haemoglobin concentration was read directly in the field. As the HemoCue cuvettes are very hygroscopic, we compared the results when using cuvettes from a container that had been opened on the day the sample was taken with those obtained using cuvettes from a container that had been opened 2–25 days before use. Statistical analysis Data were entered using FoxPro 2.6 for Windows and were converted to SPSS 7.5 for Windows for analyses. Results from duplicate measurements for the direct and indirect method were averaged and the haemoglobin concentration was reported as the mean along with the standard deviation. Results The average age of the 121 participants was 29.2 years. Almost 40% of the mothers had attended secondary school or had had higher education. The mean body mass index was 21.9 kg/m2; however, in 15.7% of participants the body mass index was <18.5 kg/m2. The prevalence of anaemia using direct assessment of venous blood was 14%. Intra-sample variability The standard deviation (SD) for the difference between the first and the second measurement of the same sample using the direct cyanmethaemoglo- bin method was 4.1 g/l for venous and 3.9 g/l for capillary blood, while for the indirect measurement the SD was 11.6 g/l for venous and 10.3 g/l for capillary blood (data not shown). Comparison of haemoglobin concentration using the three methods A summary of the results is shown in Table 1. The highest mean haemoglobin concentration was found using the HemoCue. The coefficient of variation was lowest for the direct method using venous blood and highest for the indirect assessment of venous blood. The prevalence of anaemia ranged from 14.0% (when the direct method was used to assess venous blood) to 38.0% (indirect assessment of capillary blood); the prevalence of anaemia was significantly higher when using the indirect method for both type of blood samples. The prevalence of anaemia was comparable when assessed using the HemoCue and the direct method for both types of blood samples. Sensitivity and specificity. Table 1 shows the sensitivity and specificity of the different methods in detecting anaemic participants. The sensitivity of direct assessment of capillary blood (94.1%) and the HemoCue assessment of venous blood (82.4%) was higher than that the other methods (70.6% for HemoCue for capillary blood and 76.5% for indirect assessment of either venous or capillary blood). Specificity was highest for the HemoCue assessment Table 1. Comparison of different methods for assessing haemoglobin (Hb) concentration in blood samples (n = 121) Parameter Direct Indirect HemoCue cyanmethaemoglobin cyanmethaemoglobin Venous Capillary Venous Capillary Venous Capillary Meana Hb concentration g/l 130.1 128.6 127.4 123.4 131.9 131.9 ({) ({) ({) ({) (§) ({) ({) SDb of the mean g/l 11.4 12.0 14.5 12.7 13.7 12.0 % of participants with Hb 4.9 6.6 9.9 12.4 6.6 5.8 concentration: <110 g/l 110–119 g/l 9.1 11.6 21.5 25.6 9.9 8.2 5120 g/l 86.0 81.8 68.6 62.0 83.5 86.0 Coefficient of variation (%) 8.76 9.33 11.38 10.29 10.39 9.10 Sensitivityc (%) NAe 94.1 76.5 76.5 82.4 70.6 Specificityd (%) NAe 94.2 76.0 68.3 94.2 95.2 a Mean values with a different symbol in parentheses ({, {, §) were significantly different from each other, P < 0.05 (ANOVA with post-hoc test for least significant differences). b SD = standard deviation. c Proportion of anaemic participants detected by a particular method. d Proportion of non-anaemic participants detected by a particular method. e NA = not applicable. Direct cyanmethaemoglobin measurement of venous blood was the standard against which other methods were judged. 508 Bulletin of the World Health Organization, 2001, 79 (6) Research of capillary (95.2%) or venous (94.2%) blood, and for direct assessment of capillary blood (94.2%). The specificity of the indirect method was relatively low (68.3% for capillary and 76% for venous blood). Table 2 shows the correlation coefficients of the results of comparisons between direct assessment of venous blood and other methods. The correlation between the direct assessment of venous blood and the indirect assessment of venous or capillary blood was lower than that between direct assessment of venous blood and the HemoCue method for both types of blood samples. Table 2 also shows the percentage of participants for whom the difference between two methods (the direct cyanmethaemoglo- bin assessment of venous blood and another specified method) divided by the mean of twomethods was less than 5%. The proportion with less than 5% difference was higher for comparisons with direct cyanmethae- moglobin assessment of capillary blood and the HemoCue assessment of venous and capillary blood. HemoCue: comparison of new and old microcuvettes The HemoCue assessment of venous blood was done twice: once using a microcuvette from a container whose seal had been broken on the day of the measurement and once with a microcuvette from a container whose seal had been broken 2–25 days before the measurement. Table 3 shows the results of this comparison. The mean haemoglobin concentra- tion as assessed with old cuvettes was 2.7 g/l (2.3%) higher than that assessed with new cuvettes. A closer analysis of the data showed that there was no difference between the old and new cuvettes when the old cuvettes came from a container whose seal had been broken no more than 12 days before the measurement. Cuvettes opened more than 12 days before the measurement detected significantly higher haemoglobin concentrations than new cuvettes (data not shown). The prevalence of anaemia as assessed with new and old (2–25 days old) cuvettes was significantly different (P< 0.05). The sensitivity of the old cuvettes (2–25 days old) for detecting anaemia in participants was 60.0% and the specificity was 96.0%. Discussion We found that the prevalence of anaemiawas twice as high when the indirect method was used (31–38%) and compared with the results of the direct method or the HemoCue (14–18%). This may have been caused by the blood on the filter-paper being incompletely dissolved, although the technicians soaked the paper in Drabkin’s solution until no stains were visible on it. The large difference in the estimated prevalence of anaemia determined using two different methods means that methodological differences should be examined critically before the results of different surveys are compared. The haemoglobin concentration assessed in capillary bloodwas slightly lower than that assessed in venous blood, except when the HemoCue was used: when using HemoCue the results were the same. Because the capillary blood vessel is small, the red- cell volume of capillary blood is 1–3% lower than that of venous blood, and therefore its haemoglobin concentration is lower (10). TheHemoCue, however, cannot detect such a small difference. The method of choice for evaluating anaemia in remote areas would be the HemoCue method to assess venous blood; our second choice would be the HemoCue assessment of capillary blood; our third choice would be to the direct cyanmethaemoglobin method of assessing capillary blood. These choices were made using the comparisons of the sensitivity Table 2. Comparison of haemoglobin concentration in samples from the same participant (n = 121) as assessed by direct cyanmethaem- oglobin measurement of venous blood and other methods Method Correlation Mean paired % of participants coefficienta differenceb g/l for whom the difference between 2 methodsc divided by mean of 2 methods < 5% Direct cyanmethaemoglobin 0.946 –1.5 (–2.2 to –0.8) 65.3 measurement of capillary blood Indirect cyanmethaemoglobin 0.625 –2.7 (–4.8 to –0.6) 36.4 measurement of venous blood Indirect cyanmethaemoglobin 0.635 –6.8 (–8.6 to –4.9) 29.8 measurement of capillary blood HemoCue measurement 0.917 1.8 (0.8 to 2.8) 60.3 of venous blood HemoCue measurement 0.848 1.8 (0.6 to 3.0) 52.1 of capillary blood a Pearson’s correlation coefficient; P < 0.05 for all comparisons. b Calculated as [haemoglobin concentration of sample calculated by method specified] minus [haemoglobin concentration of venous blood calculated by direct cyanmethaemoglobin method]. The figures in parentheses are 95% confidence intervals. c For each participant, the results from each method were compared to those obtained by direct cyanmethemoglobin measurement of venous blood. Table 3. Results using HemoCuea with old and new microcuvettes (n = 121) Parameter Oldb cuvette Newc cuvette Mean haemoglobin 134.6 (13.7) concentrationd g/l (SDe) 131.9 (13.7) Anaemia prevalence (%) 13.2 16.5 Sensitivity (%) 60.0 82.4 Specificity (%) 96.0 94.2 a HemoCue assessment of venous blood. b Cuvette from a container opened 2–25 days before use. c Cuvette from a container opened on the day of use. d Paired difference = 2.7 + 8.6 g/l (P < 0.05). e SD = standard deviation. 509Bulletin of the World Health Organization, 2001, 79 (6) Estimating the prevalence of anaemia and specificity of the different methods (Table 1). Given our experience of using these methods in the field, we would not recommend the use of the indirect cyanmethaemoglobin method because it identified a much higher prevalence of anaemia. A comparison of the results obtained with the HemoCue usingmicrocuvettes taken from a contain- er opened on the day of the test and using those taken from a container that had been opened 2–25 days earlier showed that 1–2 weeks after breaking the seal the haemoglobin concentration was assessed as too high. Two studies from Indonesia have reported that unreliable data were obtained when cuvettes were used from a container on which the seal had been broken more than a couple of days before the measurement (M. Gliwitzki, personal communica- tion, 1996) (11). In both of those studies, as well as in the study presented here, the containers were closed immediately after taking out a few cuvettes because of their hygroscopic nature. Thus, themanufacturer’s statement that cuvettes should be used within 2 months of breaking the seal does not seem to hold for humid climates. In areas where there is high humidity, the 50 cuvettes contained in one container should be used within a few days of breaking the seal. This study found that the prevalence of anaemia was overestimated by the indirect assess- ment method. Thus, where field conditions and local resources allow it, haemoglobin concentration should be assessed using the gold standard, the direct cyanmethaemoglobin method. However, the HemoCue is appropriate for surveys that are conducted over comparatively large areas, for those that require the use of different laboratories, or for those that take place in remote areas where adequate laboratory facilities are not available. The HemoCue is easy to use in the field: it is battery operated and there is little inter-observer error. However, the HemoCue costs a considerable amount of money, and this should be considered when planning and budgeting for data collection. n Acknowledgements We thank the field supervisors; field workers from the Nutrition Academy in Banjarmasin, South Kaliman- tan; the participants; local government staff; staff of the Health Laboratory at the municipal level (Balai Laboratorium Kesehatan, Banjarmasin, South Kali- mantan) for haemoglobin assessment (HemoCue and direct cyanmethaemoglobin); and the Nutrition Re- search andDevelopment Center in Bogor, in particular Ms Yetty Yuniar, for conducting the indirect cyan- methaemoglobin assessment of haemoglobin. Conflicts of interest: none declared. Re´sume´ Estimation de la pre´valence de l’ane´mie : comparaison entre trois me´thodes Objectif De´terminer quelle est la me´thode la plus efficace pour mesurer la concentration en he´moglobine lors d’enqueˆtes a` grande e´chelle re´alise´es dans des re´gions recule´es, et comparer deux me´thodes (mesure indirecte de la cyanme´the´moglobine et HemoCue) avec la me´thode classique (mesure directe de la cyanme´- the´moglobine). Me´thodes Des pre´le`vements de sang veineux et capillaire ont e´te´ re´alise´s en Indone´sie chez 121 me`res et les re´sultats obtenus par les trois me´thodes ont e´te´ compare´s. Re´sultats Lorsqu’elle e´tait de´termine´e par mesure indirecte de la cyanme´the´moglobine, la pre´valence de l’ane´mie e´tait de 31-38 %. Par mesure directe de la cyanme´the´moglobine ou par la me´thode HemoCue, elle e´tait de 14-18 %. La mesure indirecte de la cyanme´- the´moglobine donnait le plus fort coefficient de variation et le plus grand e´cart type de la diffe´rence entre la premie`re et la deuxie`me mesure du meˆme e´chantillon de sang (10-12 g/l avec la mesure indirecte contre 4 g/l avec la mesure directe). Par comparaison avec la mesure directe de la cyanme´the´moglobine sur le sang veineux, la meilleure sensibilite´ (82,4 %) et la meilleure spe´cificite´ (94,2 %) revenaient a` la me´thode HemoCue applique´e au sang veineux. Conclusion Lorsque les conditions de travail sur le terrain et les ressources locales le permettent, la concentration en he´moglobine doit eˆtre estime´e par la me´thode de mesure directe de la cyanme´the´moglobine, qui est la me´thode de re´fe´rence. La me´thode HemoCue peut toutefois eˆtre utilise´e dans les enqueˆtes faisant appel a` diffe´rents laboratoires ou re´alise´es dans des re´gions relativement recule´es. En climat particulie`rement chaud et humide, les microcuves HemoCue doivent eˆtre jete´es si elles ne sont pas utilise´es dans les quelques jours qui suivent l’ouverture du re´cipient. Resumen Estimacio´n de la prevalencia de anemia: comparacio´n de tres me´todos Objetivo Determinar el me´todo ma´s eficaz para analizar las concentraciones de hemoglobina en grandes encuestas llevadas a cabo en zonas remotas, y comparar dos me´todos (cianometahemoglobina indirecta y Hemo- Cue) con el me´todo convencional (cianometahemoglo- bina directa). Me´todos Se compararon los resultados obtenidos con los tres me´todos en muestras de sangre venosa y capilar de 121 madres de Indonesia. Resultados El me´todo de la cianometahemoglobina indirecta arrojo´ una prevalencia de anemia del 31%- 38%. Con el me´todo de la cianometahemoglobina 510 Bulletin of the World Health Organization, 2001, 79 (6) Research directa y con HemoCue, la prevalencia fue del 14%- 18%. La medicio´n indirecta de la cianometahemoglo- bina fue el me´todo que presento´ el coeficiente de variacio´n ma´s alto y la mayor desviacio´n esta´ndar de la diferencia entre la primera y la segunda evaluacio´n de la misma muestra de sangre (10-12 g/l con el me´todo indirecto, frente a 4 g/l con el me´todo directo). Comparado con la determinacio´n directa de la ciano- metahemoglobina en sangre venosa, HemoCue presento´ la mayor sensibilidad (82,4%) y especificidad (94,2%) cuando se aplico´ a la sangre venosa. Conclusio´n Si las condiciones sobre el terreno y los recursos locales lo permiten, la concentracio´n de hemoglobina deberı´a determinarse con el me´todo de la cianometahemoglobina directa, que serı´a la prueba de referencia. Sin embargo, el me´todo HemoCue puede emplearse en las encuestas realizadas con la participa- cio´n de diferentes laboratorios o en zonas relativamente remotas. En los climas muy ca´lidos y hu´medos, una vez abierto el recipiente en que se suministran, las microcubetas de HemoCue no utilizadas al cabo de unos pocos dı´as deben ser desechadas. References 1. Dallman PR. Iron. In: Brown M, ed. Present knowledge in nutrition. Washington, DC, Nutrition Foundation, 1990: 241–250. 2. Basta SS et al. Iron deficiency anemia and the productivity of adult males in Indonesia. American Journal of Clinical Nutrition, 1979, 55: 985–988. 3. Pollit E et al. Iron deficiency and educational achievement in Thailand. American Journal of Clinical Nutrition, 1989, 50: 698–702. 4. Soewondo S, Hussaini M, Pollitt E. Effects of iron deficiency on attention and learning process in preschool children: Bandung, Indonesia. American Journal of Clinical Nutrition, 1989, 50: 667–674. 5. Second report on the world nutrition situation. Vol. 1: Global and regional results. Geneva, United Nations ACC Sub-committee on Nutrition, 1992–93. 6. Directorate of Community Nutrition. Nutrition in Indonesia: problems, trends, strategies and programs. Jakarta, Indonesia Ministry of Health, 1993. 7. WHO/UNICEF/UNU. Iron deficiency anaemia: prevention, assessment and control. Report of a joint WHO/UNICEF/UNU consultation. Geneva, World Health Organization, 1998 (unpublished report). 8. Stoltzfus RJ. Rethinking anaemia surveillance. Lancet, 1997, 349: 1764–1766. 9. Yip R. Iron deficiency: contemporary scientific issues and international programmatic issues. Journal of Nutrition, 1994, 124: 1479S–1490S. 10. Anemia detection in health services: guidelines for program managers. Seattle, WA, Program for Appropriate Technology in Health,1996. 11. De Pee S et al. Lack of improvement in vitamin A status with increased consumption of dark-green leafy vegetables. Lancet, 1995, 346: 75–81. 511Bulletin of the World Health Organization, 2001, 79 (6) Estimating the prevalence of anaemia