Original Article 150 Paediatr Indones, Vol. 53, No. 3, May 2013 Serum ferritin to detect iron deficiency in children below five years of age Windy Saufia Apriyanti, Sutaryo, Sri Mulatsih Abstract Background lron deficiencv (lD) anemia impacts the cognitive and motor development of children until the age of 1O vears, despite receivin, iron therapv. larlv detection of lD is recommended and serum ferritin has been proposed as an alternative indicator for lD detection. Objective To assess the diagnostic accuracy of serum ferritin for detectin, lD in children below five vears of a,e. Methods This cross-sectional, diagnostic study was conducted in primarv health care centers in Yo,vakarta and Bantul. Hemo,lobin (Hb), serum ferritin and soluble transferrin receptor (s1fR) levels were performed on children a,ed 6-59 months. A s1fR level of ~ o.2 m,/l was used to define iron deficiencv. 1he best cut off point for serum ferritin level use as a diagnostic tool was determined by receiver operator curve. Results 1he prevalence of lD was 32'. Mean hemo,lobin levels in iron deficient and healthv children were 11.7 (SD O.5) ,/dl and 12.2 (SD O.7) ,/dl, respectivelv. 1he sensitivitv, specificitv, and positive predictive value (PPV) of serum ferritin (<12 u,/l) were 17', 93', and 56', respectivelv. Usin, a cut off of <32.1 u,/l, serum ferritin had sensitivitv of 62.1' and specificitv of 5O.o'. Conclusions The diagnostic value of serum ferritin levels is modestlv capable of detectin, lD. 1herefore, serum ferritin should not be used as an alternative indicator for detectin, lD in children below five years of age. [Paediatr Indones. 2013;53:150-4.]. Keywords: iron deficiency, serum ferritin, soluble transferrin receptor From the Department of Child Health, Gadjah Mada University Medical School, Dr. Sardjito Hospital, Yo,vakarta, lndonesia Reprint requests to: Windy Saufia Apriyanti, Department of Child Health, Gadjah Mada University Medical School, Dr Sardjito Hospital, Jalan Kesehatan No.1, Yo,vakarta 552o1, lndonesia. 1el 62-271-5o7333. l-mail: wsaufia@yahoo.com l ron deficiencv (lD) continues to be the most common cause of anemia worldwide. 1 ln children below five years of age, the prevalence of iron deficiencv anemia (lDA) was reported to be 1o.1' in lndonesia and 3o - 73' in Yo,vakarta. 2,3 lron deficiency anemia impairs the cognitive and motor development of children until 1O vears of a,e, even if iron therapy was given. 1 Therefore, detection and treatment of iron deficiency before anemia occurs may improve later developmental outcomes. lron stainin, of bone marrow aspiration has been the ,old standard for lD dia,nosis, however, this procedure is invasive. 5 The sTfR is the best alternative indicator to detect lD. lts sensitivitv and specificitv are o1' and 91', respectivelv, for detectin, lD in infants. 6 1he levels of s1fR increase when lD occurs. 7 Nevertheless, this test, too, has limitations as it is expensive and often unavailable in routine settings. o Serum ferritin has been suggested as another alternative indicator. lt reflects iron bodv stores. Moreover, it is less expensive than s1fR, available in most health care settings, and has a high specificity for lDA dia,nosis. 5 Windy Saufia Apriyanti et al: Serum ferritin to detect iron deficiency in children below five years of age Paediatr Indones, Vol. 53, No. 3, May 2013 151 World Health Organization recommends usin, a serum ferritin level of <12 u,/l in combination with a hemo,lobin level of <11 ,/l for lDA dia,nosis in children aged 6 months to 5 years. o,9 However, this cut off has not been used to screen for lD. Dia,nostic levels of serum ferritin to detect iron deficiency in children below five years of age has not been determined. The aim of the study was to assess serum ferritin diagnostic levels and the best cutoff point to detect lD in children below five vears of a,e. Methods We conducted a cross-sectional, diagnostic study from November 2O1O to April 2O11 at primarv health care centers in Yo,vakarta and Bantul. We estimated a required sample size of 199 subjects, based on A ~ O.O5 and power ~ 9O'. 1he inclusion criteria were healthv children a,ed 6 - 59 months whose parents consented to participate. Children who had infection, inflammation, anemia (Hb level <11 ,/dl), serum ferritin level > 11O u,/l, or received iron therapv within the prior 3 months were excluded. Patients were classified as having infection and inflammation when we found fever with or without a focal infection. lron deficiencv was defined as serum ferritin <12 u,/l. As the gold standard for detecting iron deficiency, we measured s1fR, with a value of ~o.2 m,/l considered to be iron-deficient. 1O An assistant of the study interviewed parents to collect data on every child at their first visit. Data included a,e, sex, birth wei,ht, ,estational a,e, diet recall, socioeconomic status, and iron therapy. 1he criteria for ,ood recalled diet were exclusive breastfeeding, and age-appropriate diet without the consumption of cow's milk in the first vear of life. lf these criteria were not met, subjects were considered to have poor recalled diet. Socio-economic status was classified based on monthly family income as high (more than lDR 2,OO1,OOO), middle (lDR 7O1,OOO to lDR 2,OOO,OOO), and low (less than lDR. 7OO,OOO). 11 Phvsical examinations were performed to assess for infection and inflammation, as well as anthropometric status by measurement of weights (k,) and hei,hts (cm). Based on z-score, nutritional status was classified as well-nourished (weight for hei,ht index -2SD to 2SD) or undernourished (wei,ht for hei,ht index -3SD to -2SD). Venous blood samples (2.5 ml) were drawn bv laboratorv staff to check Hb, serum ferritin and sTfR levels. The Hb level was measured with cyanmethemoglobin and serum ferritin level was measured with enzyme- linked immunosorbent assav (lllSA) in the Parahita laboratorv, Yo,vakarta. We sent eli,ible samples to the Southeast Asian Ministry of Education Organization (SlAMl) laboratorv in Jakarta to measure s1fR levels usin, a sandwich lllSA technique. Kolmogorov-Smirnov test was used for the assessment of normality. To compare the distribution of Hb, serum ferritin and s1fR levels between the lD and normal groups, the T-test or Mann-Whitney test was used. A P value of <O.O5 was considered to be statistically significant. Diagnostic values of serum ferritin was deter- mined by calculating the sensitivity, specificity, prevalence (pre-test probabilitv), positive predictive value (PPV), ne,ative predictive value (NPV), positive likelihood ratio, negative likelihood ratio, pre-test odds, post-test odds, and post-test probability. Receiver operator curve (RC) was performed to choose the best serum ferritin concentration cut off Figure 1. Study prole ow chart Data analysis n=90 sTfR level examination n=90 Hb <11 g/dL (n=10) Serum ferritin >140 ug/L (n=3) Serum ferritin level examination n=103 Hb level examination n=103 nfection and inammation (n=0) ron therapy (n=0) Venous blood sample 2.5 mL n=103 History, physical examinations, anthropometric measurements n=103 nformed consent Subjects enrolled n=103 Windy Saufia Apriyanti et al: Serum ferritin to detect iron deficiency in children below five years of age 152 Paediatr Indones, Vol. 53, No. 3, May 2013 point for lD dia,nosis. 1his studv was approved bv the Ethics Committee of the Gadjah Mada University Medical School, Yo,vakarta. Results 1here were 1O3 children initiallv enrolled in this studv. lrom these children, 13 children were excluded for anemia (1O children) and serum ferritin level >11O u,/l (3 children) (Figure 1). Table 1. Baseline characteristics of subjects Characteristics ron decient children (n = 29) Normal children (n = 61) Age, n (%) 6 24 months 25 59 months 6 23 13 (21) 48 (79) Sex, n (%) Male Female 14 15 34 (56) 27 (44) Birth weight, n (%) <2500 grams 2500 4000 grams 0 29 2 (3) 59 (97) Gestational age, n (%) <37 weeks 0 0 37 42 weeks 29 61 (100) Nutritional status, n (%) Well-nourished Undernourished 13 16 42 (69) 19 (31) Recalled diet, n (%) Good Poor 13 16 37 (61) 24 (39) Socioeconomic status, n (%) High Middle Low 0 13 16 2 (3) 28 (46) 31 (51) Table 2. Hb, sTfR, and serum ferritin levels by group Variables ron decient children (n = 29) Normal children (n = 61) P value Mean Hb (SD), g/dL 11.7 (0.5) 12.2 (0.7) 0.001* Mean sTfR (SD), mg/L 9.3 (1.1) 6.2 (1.1) <0.001* Median serum ferritin (SD), ug/L 24.3 (2.4 121.5) 33 (4.5 114.2) 0.228 # *t-test; # Mann-Whitney Table 3. Crosstabulation between serum ferritin and sTfR values sTfR ron decient children Normal children Total Serum ferritin ron decient 5 4 9 Normal 24 57 81 Total 29 61 90 Sensitivity = 17%; specicity = 93%; prevalence (pre-test probability) = 32%; PPV = 56%; NPV = 70%; positive likelihood ratio = 2.63; negative likekihood ratio= 0.89; pre-test odds = 0.47; post-test odds = 1.23; post-test probability = 55% We found that the prevalence of lD in our subjects was 32'. 1he hi,hest prevalence was found in the 25 - 59 months a,e ,roup (23 out of 29). All children from the lD ,roup had normal birth wei,hts and ,estational a,es. 1here were 11 males and 15 females detected to have lD (Table 1). We found that iron deficient children had significant lower Hb level and higher sTfR level compared to normal children, but no significant difference in median of serum ferritin levels between the groups (Table 2). Windy Saufia Apriyanti et al: Serum ferritin to detect iron deficiency in children below five years of age Paediatr Indones, Vol. 53, No. 3, May 2013 153 Cross-tabulation between serum ferritin and sTfR values are shown in Table 3. According to ROC, the best serum ferritin level cut off point was <32.1 u,/l, increasin, its sensitivitv to 62.1' and specificitv to 5O.o'. Discussion lron deficiencv usuallv occurs in the second vear of life, due to decreased iron intake and rapid growth in the first year of life. Normal infants need to absorb approximatelv O.o m,/dav of dietarv iron (O.6 m, for ,rowth and O.2 m, to replace on,oin, losses). 1owards the end of the second year of life, this swift rate of growth begins to slow, so routine diets should include sufficient iron-rich foods to meet demands. 5,12 ln our subjects, we found the hi,hest prevalence of iron deficiencv to be in the third vear of life. Sixteen of 29 children with lD were undernourished, had poor recalled diet, and low socioeconomic status. Poor recalled diet included children not breastfed exclusivelv or not ,ettin, appropriate diet for their age. No history of cows milk consumption was found in either group. Low socioeconomic status led to parents being unable to provide appropriate diets, as well as appropriate developmental stimulation for their children. 1,9 We found similar numbers of lD in males and females. Gender differences reportedly only affect lD in adolescents, as females are at hi,her risk due to menstruation and rapid ,rowth. ln developin, countries, lD has also been attributed to chronic blood loss due to parasitic infections. 5,13 A limitation of our studv was that we did not perform stool examinations to detect parasitic infections. We found that the mean Hb level in the iron deficient group was lower than that of the normal ,roup (P~O.O1). lven if the Hb level was still within normal limits in iron deficient patients, iron supplementation must be given to prevent anemia based on the lndonesian Pediatric Societv recommendations. Mean sTfR level in the iron deficient group was hi,her than that of the normal ,roup (P<O.OO1). Levels of serum ferritin in both groups were within normal ranges. This shows that sTfR is a more sensitive indicator for iron deficiency than serum ferritin. Normal values of serum ferritin for children a,ed 6 months - 15 vears are 12 - 11O u,/l. 11 lnterpretation of serum ferritin must be looked at closely. Ferritin is an acute-phase reactant that can become elevated in settings of inflammation, chronic infection, and malignancy. 5,1O,15 C-reactive protein (CRP) and A1-acid ,lvco- protein (AOP) are common biomarkers of infection and inflammation under field conditions. The WHO and Centers for Disease Control (CDC) stress the need to include CRP and AOP in iron status assessments. CRP rises rapidlv after the onset of infection and declines within 21-1o hours. AOP reaches its maximum concentration 1o hours after the onset of inflammation and remains elevated for 12O-111 hours. 16 ln order to prevent false positives usin, serum ferritin, we excluded the children who suffered from infection and inflammation based on historv-takin,, and whose serum ferritin was >11O u,/l. CRP and AOP examinations were not performed in our study. Both cut off and diagnostic values of serum ferritin to detect iron deficiency in children below five years of age have not yet been identified. The WHO recommends usin, a serum ferritin level of <12 u,/l in combination with a Hb level of <11 ,/dl for lDA diagnosis in children aged 6 months to 5 years. o,9 1herefore, we used a serum ferritin level of <12 u,/l as an initial cut off point. However, this level resulted in poor diagnostic values, with sensitivity, specificity and PPV of 17', 93' and 56', respectivelv. Positive likelihood ratio was onlv 2.63, which meant serum ferritin was not a useful differentiator. The poor diagnostic values may have been influenced by the small sample size. Several studies had similar results. Serum ferritin levels of <15 u,/l and 12 u,/l were used to detect lD in adult patients and also had poor dia,nostic values. 17,1o However, an Estonian study found serum ferritin level <1O.9 u,/l had o3' sensitivitv and oO' specificitv to detect lD in healthv, term infants a,ed 9 - 12 months with normal birth wei,ht. lnfants with increased CRP concentration (CRP >5 m,/l) were excluded. 19 We constructed a ROC in order to choose the best cut off point of serum ferritin concentration, which was 32.1 u,/l. lts sensitivitv and specificitv were 62.1' and 5O.o', respectivelv, however, this Windy Saufia Apriyanti et al: Serum ferritin to detect iron deficiency in children below five years of age 154 Paediatr Indones, Vol. 53, No. 3, May 2013 level was still not sensitive enough to make a confident diagnosis. A good screening tool is defined as having a sensitivitv ~oO', despite the presence of low specificity. 2O ln conclusion, we recommend that serum ferritin not be used as an alternative indicator in detectin, lD in children below five years of age. However, further study should be performed with a larger sample size and with examinations of infection or inflammation biomarkers, includin, CRP or AOP while examinin, iron status. Acknowledgment We are extremelv ,rateful to lndv Parvanto MD for the special comments during manuscript preparation, Hendro Dwi Wicaksono as our assistant in the study. We also thank the laboratory staff of the Parahita laboratorv in Yo,vakarta and the SlAMl laboratorv in Jakarta. References 1. Assessin, the iron status of populations. 2 nd ed. Oeneva: World Health r,anization, 2OO7. p. 1 - 112. 2. Windiastuti l. Anemia defisiensi pada bavi dan anak. lndonesian Pediatric Societv, 2OO9 |cited 2O12 Mav 9]. Available from: http://www.idai.or.id/kesehatananak/artikel. asp?q=20125795911 3. Sutarvo. 1ujuh puluh lima persen anak di DlY menderita anemia defisiensi besi. Suara merdeka, 2OO7 |cited 2O12 Mav 9]. Available from: http://www.suaramerdeka.com/ cvbernews/harian/O7O5/Oo/dar3.htm 1. Soedjatmiko. Pen,aruh defisiensi besi terhadap kecerdasan dan perilaku anak. ln: PKB lll Departemen llmu Kesehatan Anak lakultas Kedokteran Universitas lndonesia. Pendekatan praktis pucat. Jakarta: lndonesian Pediatric Societv, 2OO7. p. 72-o6. 5. Wu AC, Lesperance L, Bernstein H. Screening for iron deficiencv. Pediatr Rev. 2OO2,23:171-o. 6. Vendt N, 1alvik 1, leedo S, 1omber, K, Kool P, 1illmann V, et al. The reference limits and cut-off value for serum soluble transferrin receptors for diagnosing iron deficiency in infants. lnt J lab Hematol. 2OO9,31:11O-6. 7. Koulaouzidis A, Said l, Cotter R, Saeed AA. Soluble transferrin receptors and iron deficiency, a step beyond ferritin. A svstematic review. J Oastrointestin liver Dis. 2OO9,1o:315-52. o. Muhammad A, Sianopar . Penentuan defisiensi besi, anemia penyakit kronis menggunakan peran indeks s1fR-l. lndonesian Journal Clinical Patholo,v and Medical laboratorv. 2OO5,12:9-16. 9. Baker RD, Oreer lR, and the Committee on Nutrition American Academv of Pediatrics. Dia,nosis and prevention of iron deficiency and iron deficiency anemia in infants and voun, children (O - 3 vears of a,e). Pediatrics. 2O1O,126:1O1O- 5O. 1O. lrhardt JO, lstes Jl, Pfeiffer CM, Biesalski HK, Craft Nl. Combined measurement of ferritin, soluble transferrin receptor, retinol binding protein, and C-reactive protein by an inexpensive, sensitive, and simple sandwich enzvme-linked immunosorbent assav technique. J Nutr. 2OO1, 131:3127- 32. 11. New social economv status bracket. lndonesia: Nielsen Co., 2O1O. p. 1-1. 12. lron deficiencv anemia. Assessment, prevention, and control. A ,uide for pro,ramme mana,ers. Oeneva: WH, 2OO1. p. 1-132. 13. Kazal l. Prevention of iron deficiencv in infant and toddlers. Am lam Phvsicians. 2OO2,66:1217-27. 11. lanzkowskv P. lron deficiencv anemia. ln: Manual of pediatric hematolo,v and oncolo,v. 1 th ed. Amsterdam: llsevier Academic Press, 2OO5. p. 31-16. 15. Badham J, zimmermann MB, Kraemer K. 1he ,uidebook nutritional anemia. Urbana: Side and life Press. 2OO7. p. 1-52. 16. Avova MA, Spiekermann-Brouwer OM, Stoltzfus RJ, Nemeth l, Habicht JP, Oanz 1, et al. A1-Acid ,lvcoprotein, hepcidin, C-reactive protein, and serum ferritin are correlated in anemic schoolchildren with Schistosoma haematobium. Am J Clin Nutr. 2O1O,91:17o1-9O. 17. n, KH, 1an Hl, lai HC, Kuperan P. Accuracv of various iron parameters in the prediction of iron deficiency in an acute care hospital. Ann Acad Med Sin,apore. 2OO5,31:137- 1O. 1o. Mast Al, Blinder MA, Oronowski AM, Chumlev C, Scott MG. Clinical utility of the soluble transferrin receptor and comparison with serum ferritin in several populations. Clin Chem. 199o,11:15-51. 19. Vendt N, 1alvik 1, leedo S, 1omber, K, Kool P, 1illmann V, et al. Reference and cut-off values for serum ferritin, mean cell volume, and hemoglobin to diagnose iron deficiency in infants a,ed 9 to 12 months. Medicina (Kaunas). 2OO7,13:1- 5. 2O. Sastroasmoro S, lsmael S, editors. Dasar-dasar metodologi penelitian klinis. 3 rd ed. Jakarta: Sa,un, Seto. 2OOo.