THE JOURNAL OF PEDIATRICS • www.jpeds.

com CLINICAL AND LABORATORY

OBSERVATIONS
Re-Evaluation of Serum Ferritin Cut-Off Values for the Diagnosis of Iron
Deficiency in Children Aged 12-36 Months
Kawsari Abdullah, PhD, MBBS1, Catherine S. Birken, MD, MSc, FRCPC1,2,3,4,5, Jonathon L. Maguire, MD, MSc, FRCPC2,3,4,5,6,7,
Darcy Fehlings, MD, MSc, FRCPC2,3,5,8, Anthony J. Hanley, PhD3,9,10, Kevin E. Thorpe, MMath3,6, and
Patricia C. Parkin, MD, FRCPC1,2,3,4,5

An ongoing challenge has been determining clinically relevant serum ferritin cut-offs in the diagnosis of iron defi-
ciency in children aged 1-3 years. We identified 2 potential clinically relevant serum ferritin cut-off values through
their association with clinically important cut-off of hemoglobin as the indicator of anemia. (J Pediatr 2017;■■:■■-■■).

See editorial, p •••

I
ron deficiency is the most common nutritional defi-
ciency found in young children, peaking in prevalence
between 1 and 3 years of age. It may lead to irreversible
neurocognitive impairment.1,2 The iron status of children can
be assessed with hematologic tests (ie, hemoglobin [Hb] con-
centration, hematocrit, mean cell volume, and red blood cell
distribution width) and iron-specific biomarkers (ie, serum fer-
ritin concentration, serum iron, erythrocyte protoporphyrin
concentration, total iron-binding capacity, transferrin satu-
ration, and serum transferrin receptors).3,4
Because changes in Hb concentration occur only at the late
stage of iron deficiency, it is considered a late indicator of iron
deficiency. The American Academy of Pediatrics recom-
mends universal screening for anemia with determination of
Hb concentration with a cut-off of <110 g/L for children aged
1-3 years.2 Strong evidence has been found showing this
Hb cut-off for anemia to be associated with delayed
neurodevelopment in young children.5,6 In addition, the Ameri-
can Academy of Pediatrics, Centers for Disease Control and
Prevention, and a Cochrane Systematic Review recommend
clinical management in young children whose Hb level is less
than this level.2,4,7 Hence, this cut-off (<110 g/L) has clinical
importance for both practitioners and iron deficiency
researchers.
Serum ferritin is one of the most widely used and specific
biomarkers of iron status in young children,3,4,8 reflecting body
iron stores.9 Currently recommended serum ferritin cut-off
values for identifying iron deficiency in children are between
<10 and 12 µg/L.2,3 This cut-off is recommended in a review
by Dallman et al published in 1980, which cites an original study
by Siimes et al published in 1974.10,11 In this study, children
aged 0-15 years with iron deficiency anemia (n = 13) were iden-
tified to have a serum ferritin range of 1.5-9 ng/mL. Two other
studies evaluated the cut-off of serum ferritin in young chil-
CRP C-reactive protein
Hb Hemoglobin
RCS Restricted cubic spline
dren (0-5 years) using the fifth percentile of the distribution
of serum ferritin as the cut-off for defining iron deficiency. One
study identified the fifth percentile to correspond to a serum
ferritin of <10 µg/L, whereas the other found this value to be
12 µg/L.12,13 A study targeting a younger age group of children
(9-12 months) has suggested 2 SD cut-off values and has re-
ported values as low as 5 µg/L for serum ferritin in this age
group.14
The evidence behind these values is weak. They were not
developed based on clinical relevance and are not specific for
infants 12-36 months of age (the age of peak prevalence). In
2015, the National Institutes of Health established an iron ini-
tiative to prioritize research gaps, including current chal-
lenges in measuring and screening for iron status in young
children, highlighting the importance of re-evaluating serum
ferritin cut-off values.15
It also is important to understand the relationship between
serum ferritin cut-offs and important child health outcomes
such as neurodevelopment. Some evidence suggests that low
serum ferritin alone may have significant impact on chil-
dren’s neurodevelopment5,16,17; however, this evidence is not
conclusive and further research is needed.18
Considering the extended time course for development of
long-term health outcomes that may be related to iron defi-
ciency in children, identifying clinically relevant cut-off values
From the 1Child Health Evaluative Sciences, The Hospital for Sick Children Research
Institute; 2Institute of Health Policy, Management and Evaluation, University of
Toronto; 3Dalla Lana School of Public Health, University of Toronto; 4Pediatrics
Outcomes Research Team, Division of Pediatric Medicine, Department of Pediatrics,
The Hospital for Sick Children; 5Department of Pediatrics, Faculty of Medicine,
University of Toronto, Toronto, Canada; 6Applied Health Research Centre, Li Ka
Shing Knowledge Institute of St. Michael’s Hospital, Toronto, Canada; 7Department of
Pediatrics, St. Michael’s Hospital, University of Toronto; 8Division of Developmental
Pediatrics, Holland Bloorview Kids Rehabilitation Hospital, Bloorview Research
Institute, Toronto; 9Department of Nutritional Sciences, University of Toronto, Toronto,
Canada; and 10Division of Endocrinology, Department of Medicine, University of
Toronto, Toronto, Canada
Supported by The Hospital for Sick Children Foundation ([SP05-602] with a grant to
the Pediatric Outcomes Research Team), St Michael’s Hospital Foundation (2012-
0051-GF), and Canadian Institutes for Health Research (CIHR; FRN-115059). The
authors declare no conflicts of interest.
0022-3476/$ - see front matter. © 2017 Elsevier Inc. All rights reserved.
http://dx.doi.org10.1016/j.jpeds.2017.03.028

CRP 5.4.0 DTD ■ YMPD9084_proof ■ April 18, 2017

THE JOURNAL OF PEDIATRICS • www.jpeds.com
for serum ferritin will enhance the laboratory diagnosis of iron
deficiency in young children. The objective of this study was
to identify clinically relevant serum ferritin cut-off values for
the diagnosis of iron deficiency in children aged 12-36 months
by examining its relationship with Hb concentration.
Methods
This was a cross-sectional study of healthy, urban children living
in Toronto, Canada. Data were collected from May 2010 to July
2014. Study participants included children aged 12-36 months
recruited during a scheduled health supervision visit with a
physician participating in the TARGet Kids! primary care prac-
tice based research network (www.targetkids.ca).19 Excluded
children were those diagnosed with anemia other than iron
deficiency, having C-reactive protein (CRP) ≥10 mg/L, previ-
ously diagnosed with a hematologic disorder (thalassemia and
other disorders of Hb), developmental disorder, genetic, chro-
mosomal, or syndromic condition, and chronic medical con-
ditions (except asthma and allergy). Serum ferritin is an acute-
phase reactant, and concentrations of serum ferritin may be
elevated in the presence of chronic inflammation, infection,
malignancy, or liver disease.2 In this study we have combined
serum ferritin concentration with determination of another
acute-phase reactant (CRP) to exclude children whose serum
ferritin may have been elevated due to inflammation or
infection.
Data on children’s health, nutrition, and sociodemo-
graphic characteristics were collected prospectively with a stan-
dardized parent-completed survey instrument. A sample of
blood also was collected and analyzed to determine chil-
dren’s iron status by the use of iron specific indicators—Hb,
serum ferritin, and CRP. Serum ferritin was measured with a
Roche Modular platform (Roche Diagnostics Limited, Rotkreuz,
Switzerland), and Hb was measured using Sysmex platform
(Sysmex Canada, Mississuaga, ON, Canada).20,21 All diagnos-
tic assessments were performed at the Mount Sinai Services
Laboratory (www.mountsinaiservices.com). Approval for data
collection was received from the Hospital for Sick Children and
St Michael’s Hospital research ethics boards, and informed
consent was received from parents of participating children.
Statistical Analyses
Descriptive statistics were used to describe the distribution of
the variables: child age and sex, Hb, and serum ferritin. Non-
linearity assumptions were tested by visual inspection and by
performing a likelihood ratio test to compare the nested uni-
variate models with linear predictors vs restricted cubic spline
(RCS) regression models, which included nonlinear predictors.22
Some variables have a curvilinear relationship (nonlinear)
with each other. Increases in “X” variable initially produce in-
creases in “Y,” but after a while subsequent increases in “X”
produce declines in “Y.” RCS regression analysis can estimate
such nonlinear relationships. The graph of the relationship
between “X” and “Y” consists of a curve with one or more
bends, termed “knots,” at which the slope of the curve changes
2 Abdullah et al
CRP 5.4.0 DTD ■ YMPD9084_proof ■ April 18, 2017
Volume ■■
signs. If nonlinearity assumptions were satisfied, we planned
to apply RCS regression analysis to characterize the associa-
tions between serum ferritin (independent variable) and Hb
(dependent variable). We intended to add children’s age and
sex to the model as potential confounders13,23-25; however, the
spline function would be performed only for the continuous
variable serum ferritin.
By applying the RCS regression analysis (using 4 or 5 knots)
we planned to identify 2 specific serum ferritin values. First,
a serum ferritin value that predicted a maximum increase in
Hb as a function of serum ferritin was calculated by finding
the root of the derivative of the predicted mean Hb. This
maximum Hb level is known as the “Hb plateau point.” Second,
because a Hb cut-off of 110 g/L is used to define anemia, we
intended to find the corresponding serum ferritin value that
predicted a Hb of 110 g/L from the RCS regression models.
It has been recommended that for a large sample (n > 100),
4 or 5 knots is a good choice and provides enough flexibility
for a reasonable loss of precision caused by overfitting the
data.22,26 Therefore, we performed the analysis with both 4 and
5 knots.
Data analyses were performed with SAS, version 9.1 (SAS
institute Inc, Cary, North Carolina). R version 3.0.1 (http://
www.r-project.org/, Vienna, Austria) was used for RCS analy-
sis.27 Alpha (2-tailed) was set at values less than .05.
Results
Blood samples of 1257 children (12-36 months of age) were
analyzed for Hb and serum ferritin. The mean (±SD) age of
the children was 18.9 (5.9) months, and 53.5% were male. Mean
(±SD) serum ferritin and Hb were 27.7 (19.7) µg/L and 119.3
(8.8) g/L, respectively. Children’s health, nutrition, and so-
ciodemographic characteristics have been reported previ-
ously in the cohort profile.28 Compared with nested univariate
models of linear predictors, the model with RCS showed a like-
lihood ratio c2 value of 91.07 (P < .001), indicating that a non-
linear relationship (between serum ferritin and Hb) had a better
fit to the data.
Using RCS regression analysis, we found serum ferritin to
be significantly associated with Hb (P ≤ .001), age (P ≤ .001),
and sex (P = .004) in the 4-knots model and with Hb (P ≤ .001)
and age (P = .003) in the 5-knots model. From the model with
4 knots, we identified a serum ferritin value of 23.7 µg/L cor-
responding to the “Hb plateau point” of 121.2 g/L (see Methods
for definition). Figure 1 shows the plot of the RCS regres-
sion model with 4 knots, which depicts a 2-phase association
between serum ferritin and Hb with a strong increase up to
the “Hb plateau point,” followed by a much milder increase
afterwards. The serum ferritin value of 2.4 µg/L corresponded
to the Hb value of 110 g/L (mean age 18.9 months and
sex = male). Similarly, the model with 5 knots (Figure 2) iden-
tified a serum ferritin value of 17.9 µg/L corresponding to the
“Hb plateau point” of 121.0 g/L; and a serum ferritin value of
4.6 µg/L corresponded to the Hb value of 110 g/L (mean age
18.9 months and sex = male).
■■ 2017
Figure 1. Plot of the RCS regression model with 4 knots
(shaded areas inside the dashed lines represent 95% CIs).
Discussion
Using RCS modeling, we identified that the values of Hb in-
creased until reaching a maximum level and then plateaued
thereafter (Figures 1 and 2). In the RCS models (with 4 or 5
knots) the corresponding serum ferritin value for this maximum
level of Hb was 23.7 µg/L or 17.9 µg/L, respectively. In healthy
children, a value of serum ferritin at which Hb concentra-
tion is maximized may have important clinical implications.
Hb represents the functional form of iron in the body and
serum ferritin the storage form.29 A serum ferritin value that
corresponds to the maximum level of Hb may represent the
optimal amount of storage iron required to carry out the func-
tional processes associated with iron metabolism.29 Hence,
serum ferritin values less than these thresholds may nega-
tively affect functional processes.30 Therefore, a value of serum
ferritin <17.9 µg/L may be a clinically meaningful cut-off value,
because below this level the functional impact of iron defi-
ciency may begin to develop. The Hb values of the plateau point
Figure 2. Plot of the RCS regression model with 5 knots
(shaded areas inside the dashed lines represent 95% CIs).
Re-Evaluation of Serum Ferritin Cut-Off Values for the Diagnosis of Iron Deficiency in Children Aged 12-36 Months
CRP 5.4.0 DTD ■ YMPD9084_proof ■ April 18, 2017
CLINICAL AND LABORATORY OBSERVATIONS
Table. Serum ferritin cut-off levels identified by the RCS
regression analysis
Serum ferritin, µg/L
Model with Model with
Hemoglobin level 4 knots 5 knots
Corresponding to the Hb plateau point 23.7 17.9
Corresponding to Hb cut-off value of 110 g/L 2.4 4.6
in the 2 models (4 or 5 knots) were 121.2 g/L and 121.0 g/L,
respectively. These values being very close to each other
strengthens our physiologic explanation of the relevance of the
“Hb plateau point” in the association between serum ferritin
and Hb.
Another finding from our study was the identification of a
serum ferritin value that corresponded to a Hb concentra-
tion of 110 g/L. RCS models identified these values as 2.4 µg/L
(4 knots) or 4.6 µg/L (5 knots). In children with iron defi-
ciency, a Hb cut-off of 110 g/L has been used to define anemia
and has been found to be associated with adverse
neurodevelopmental outcomes.5,6 Hence, a child with iron de-
ficiency anemia based on a Hb level of <110 g/L may have a
serum ferritin as a low as 2.4 µg/L. These findings raise ques-
tions regarding the appropriateness of using Hb cut-off of
110 g/L to define clinically significant iron deficiency anemia.
This study reports the relationship between 2 iron status in-
dicators (Hb and serum ferritin) with the use of RCS regres-
sion analysis in a large sample of preschool children (n = 1257).
We have presented the association between serum ferritin and
Hb and their corresponding values using both the 4- and
5-knots models (Table).22 It is important to recognize that the
values of serum ferritin identified from the models do not cor-
respond to the serum ferritin cut-off values currently recom-
mended to diagnose iron deficiency in preschool children
(serum ferritin of <10 µg/L or 12 µg/L).2,3 Currently recom-
mended serum ferritin cut-off values were based on the dis-
tribution of serum ferritin in a sample of the population rather
than being clinically defined. Furthermore, from our study, the
corresponding serum ferritin value of a clinically important
Hb cut-off (110 g/L) showed how low the serum ferritin can
be when a child is diagnosed with anemia due to iron defi-
ciency. These low levels of serum ferritin may have a clinical
impact on the neurodevelopment of young children.
Based on the methodology we used, the serum ferritin cut-
off values identified from this study are statistical markers that
may be clinically relevant. Considering this as a limitation of
this analytic approach, these values may not represent optimal
thresholds for diagnosing iron deficiency. Hence, it is crucial
to evaluate the direct impact of the serum ferritin levels iden-
tified from this study on important clinical outcomes such as
children’s neurodevelopment and growth. Furthermore, the
diagnostic accuracy of these cut-off values need to be deter-
mined via epidemiologic methods such as receiver operating
characteristics analysis and likelihood ratios.31,32 Assessment of
the direct clinical impact and diagnostic properties will advance
our understanding of the optimal cut-off values of serum
3
THE JOURNAL OF PEDIATRICS • www.jpeds.com
ferritin for laboratory diagnosis of iron deficiency in this age
group. Further research using a longitudinal design will allow
us to understand the relationship between serum ferritin and
Hb over time. This study may be considered the initial step
towards finding serum ferritin cut-off values that signify clini-
cal relevance rather than being merely distribution-based. ■
We thank Yang Chen for his contribution in regard to the statistical analy-
sis and interpretation of the data. We also thank all of the participat-
ing children and families for their time and involvement in the TARGet
Kids! primary care practice based research network and all practice site
physicians, research staff, collaborating investigators, trainees, method-
ologists, biostatisticians, data management personnel, laboratory man-
agement personnel, and advisory committee members.
Submitted for publication Nov 8, 2016; last revision received Jan 27, 2017;
accepted Mar 9, 2017
Reprint requests: Patricia Parkin, MD, FRCPC, The Hospital for Sick Children
Research Institute, Peter Gilgan Centre for Research and Learning, 686 Bay
St, Toronto, ON M5G 0A4, Canada. E-mail: patricia.parkin@sickkids.ca
References
1. Centers for Disease Control and Prevention. Iron deficiency—United States,
1999-2000. Morbid Mortal Wkly Rep MMWR 2002;51:897-9.
2. Baker RD, Greer FR. Diagnosis and prevention of iron deficiency and iron-
deficiency anemia in infants and young children (0-3 years of age). Pe-
diatrics 2010;126:1040-50.
3. World Health Organization. Iron deficiency anaemia: assessment, pre-
vention and control. A guide for programme managers. Geneva: World
Health Organization; 2001 WHO/NHD/01.32001.
4. Centers for Disease Control and Prevention. Recommendations to prevent
and control iron deficiency in the United States. Morbid Mortal Wkly Rep
MMWR 1998;47(RR–3):1-36.
5. Idjradinata P, Pollitt E. Reversal of developmental delays in iron defi-
cient anaemic infants treated with iron. Lancet 1993;341:1-4.
6. Lozoff B, Smith JB, Kaciroti N, Clark KM, Guevara S, Jimenez E. Func-
tional significance of early-life iron deficiency: outcomes at 25 years. J
Pediatr 2013;163:1260-6.
7. Martins S, Logan S, Gilbert RE. Iron therapy for improving psychomo-
tor development and cognitive function in children under the age of
three with iron deficiency anaemia. Cochrane Database Syst Rev
2001;(2):CD001444.
8. McDonagh M, Blazina I, Dana T, Cantor A, Bougatsos C. Routine iron
supplementation and screening for iron deficiency anemia in children ages
6 to 24 months: a systematic review to update the U.S. Preventive Ser-
vices Task Force recommendation. Evidence Synthesis No 122 AHRQ Pub-
lication No 13-05187-EF-1. Rockville (MD): Agency for Healthcare
Research and Quality; 2015.
9. Lynch S. The rationale for selecting and standardizing iron status indi-
cators. World Health Organization Report: priorities in the assessment
of vitamin A and iron status in populations, Panama City, Panama, 15-
17 September 2010. Geneva: World Health Organization; 2012.
10. Dallman PR, Siimees MA, Stekel A. Iron deficiency in infancy and child-
hood. Am J Clin Nutr 1980;33:86-118.
11. Siimees MA, Addiego JE, Dallman PR. Ferritin in serum: diagnosis of iron
deficiency and iron overload in infants and children. Blood 1974;43:581-
90.
4 Abdullah et al
CRP 5.4.0 DTD ■ YMPD9084_proof ■ April 18, 2017
Volume ■■
12. Looker AC, Dallman PR, Carroll MD, Gunter EW, Johnson CL.
Prevalence of iron deficiency in the United States. JAMA 1997;277:973-
6.
13. Sherriff A, Emond A, Hawkins N, Golding J. Haemoglobin and ferritin
concentrations in children aged 12 and 18 months. Arch Dis Child
1999;80:153-7.
14. Domellof M, Dewey KG, Lonnerdal B, Cohen RJ. The diagnostic crite-
ria for iron deficiency in infants should be reevaluated. J Nutr
2002;132:3680-6.
15. National Institutes of Health. Iron screening and supplementation of iron-
replete pregnant women and young children. https://ods.od.nih.gov/
Research/Iron.aspx. Accessed September 11, 2016.
16. Oski FA, Honig AS, Helu B, Howanitz P. Effect on iron therapy on be-
havior performance in nonanemic, iron-deficient infants. Pediatrics
1983;71:877-80.
17. Akman M, Cebeci D, Okur V. The effects of iron deficiency on infants’
developmental test performance. Acta Paediatr 2004;93:1391-6.
18. Abdullah K, Kendzerska T, Shah P, Uleryk E, Parkin PC. Efficacy of oral
iron therapy in improving the developmental outcome of pre-school chil-
dren with non-anemic iron deficiency: a systemic review. Public Health
Nutr 2013;16:1497-506.
19. TARGet Kids! Collaboration. About TARGet Kids! Our Team, Our Re-
search, Participant Resource, Publication. [Internet]. http://www
.targetkids.ca. Accessed March 18, 2015. 2014.
20. Roche Diagnostics [internet]. http://www.roche.com/diagnostics. Ac-
cessed January 2014.
21. Sysmex [Internet]. https://www.sysmex.com/us/en/Brochures/Brochure
_XE-2100_MKT-10-1134.pdf. Accessed January 2014.
22. Harrell FE. Regression modeling strategies: with application to linear
models, logistic regression, and survival analysis. New York: Springer-
Verlag; 2010.
23. Freeman VE, Mulder J, van Hof MA, Hoey HMV, Gibney MJ. A longi-
tudinal study of iron status in children at 12, 24 and 36 months. Public
Health Nutr 1998;1:93-100.
24. Pasricha S, Black J, Muthayya S, Shet A, Bhat V, Nagaraj S, et al. Deter-
minants of anemia among young children in rural India. Pediatrics
2010;126:e140-9.
25. Agho KE, Dibley MJ, D’Este C, Gibberd R. Factors associated with hae-
moglobin concentration among Timor-Leste children aged 6-59 months.
J Health Popul Nutr 2008;26:200-9.
26. Durrleman S, Simon R. Flexible regression models with cubic splines. Stat
Med 1989;8:551-61.
27. R Core Team. R: A language and environment for statistical computing;
2017. R Foundation for Statistical Computing, Vienna, Austria. https://
www.R-project.org/. Accessed March 31, 2017.
28. Abdullah K, Thorpe KE, Maguire JL, Birken CS, Fehlings D, Hanley AJ,
et al. Risk factors, practice variation and hematological outcomes of chil-
dren identified with non-anemic iron deficiency (NAID), following screen-
ing in primary care setting. Paediatr Child Health 2015;20:302-6.
29. Munoz M, Garcia-Erce JA, Remacha AF. Disorders of iron metabolism.
Part I: molecular basis of iron homeostasis. J Clin Pathol 2011;64:281-
6.
30. Beaton GH, Corey PN, Steele C. Conceptual and methodological issues
regarding the epidemiology of iron deficiency and their implications for
studies of the functional consequences of iron deficiency. Am J Clin Nutr
1989;50:575-88.
31. Thomas C, Thomas L. Biochemical markers and hematological indices
in the diagnosis of functional iron deficiency. Clin Chem 2002;48:1066-
76.
32. Guyatt GH, Oxman AD, Ali M, Willan A, McIlroy W. Laboratory diag-
nosis of iron deficiency anemia. J Gen Intern Med 1992;7:145-53.