The American Academy of Pediatrics recommends universal screening for anemia using hemoglobin at 12 months. However, hemoglobin lacks diagnostic accuracy for iron deficiency, and the optimal age for screening has not been determined. Our objective was to assess a screening strategy for iron deficiency using serum ferritin.
We conducted a cross-sectional study of children 1 to 3 years old attending a health supervision visit. We examined the relationship between child age and serum ferritin, age and hemoglobin, hemoglobin and serum ferritin, and the prevalence of elevated C-reactive protein (CRP).
Restricted cubic spline analysis (n = 1735) revealed a nonlinear relationship between age and serum ferritin (P < .0001). A linear spline model revealed that from 12 to 15 months, for each 1-month increase in age, serum ferritin levels decreased by 9% (95% confidence interval [CI]: 5 to 13). From 15 to 24 months, the rate of change was nonsignificant. From 24 to 38 months, for each month increase in age, serum ferritin increased by 2% (95% CI: 1 to 2). For hemoglobin, from 12 to 24 months, the rate of change was nonsignificant. From 24 to 38 months, for each 1-month increase in age, hemoglobin increased by 20% (95% CI: 9 to 32). Compared with the serum ferritin cutoff of <12 μg/L, the hemoglobin cutoff of <110 g/L had a sensitivity of 25% (95% CI: 19 to 32) and a specificity of 89% (95% CI: 87 to 91). Elevated CRP ≥10 mg/L occurred in 3.3% (95% CI: 2.5 to 4.2).
Screening for iron deficiency using serum ferritin at 15 or 18 months may be a promising strategy. For children at low risk for acute inflammation, concurrent measurement of CRP may not be necessary.
The American Academy of Pediatrics recommends universal screening for anemia with measurement of hemoglobin at 12 months of age. However, hemoglobin lacks diagnostic accuracy for iron deficiency, and the optimal age for screening has not been determined.
In children 1 to 3 years screened in primary care, serum ferritin was lowest between 15 and 24 months, and hemoglobin did not change significantly. Screening for iron deficiency using serum ferritin at 15 or 18 months may be a promising strategy.
Iron deficiency is a common nutritional deficiency worldwide and is considered a significant public health concern in both developing and developed countries.1,–5 For young children in the United States, the prevalence of iron deficiency is ∼15%, and the prevalence of iron deficiency anemia is ∼2%.4,5
The prevalence of iron deficiency peaks in early childhood, a sensitive time for the rapidly developing brain.6,–8 Iron deficiency (anemic and nonanemic) has a negative impact on neurodevelopmental outcomes.8,9 Iron is vital for the processes of monoamine metabolism, myelin synthesis, and metabolic function of the brain.6,7 Animal studies show that early postnatal iron deficiency alters brain development and cognition.10,11 Human studies reveal that iron deficiency is associated with poor outcomes across multiple domains of child development.12,13 Impairments in cognitive, social, and emotional functioning may persist into adolescence and young adulthood among individuals who had iron deficiency in early childhood.14,–17 There is some evidence that iron supplementation in infants and young children with iron deficiency is associated with improvements in motor and cognitive functioning when treatment is provided for a therapeutically appropriate duration.18,–20
The American Academy of Pediatrics (AAP) recommends universal screening for anemia through measurement of hemoglobin at 12 months of age.5 There are several concerns related to this strategy. First, hemoglobin lacks sensitivity and specificity for iron deficiency, because levels overlap for individuals with iron sufficiency and iron deficiency (sensitivity), and there are several other causes of anemia (specificity).5,21 Second, the developing brain may be iron deficient by the time anemia is detected.8 Third, the optimal age for screening in primary care has not been determined.
In contrast to the AAP recommendation, in 2015, the US Preventive Services Task Force found insufficient evidence to recommend routine screening for iron deficiency anemia in young children.22 To address knowledge gaps identified by the US Preventive Services Task Force, in 2016, the National Institutes of Health convened an Iron Workshop and an expert panel.23 The workshop report recognized serum ferritin as the most commonly used indicator of iron deficiency.24,25 In addition, although there are limited data in young children, studies in adults suggest that the commonly recommended cutoff for serum ferritin is specific as compared with bone marrow aspirate.24,25 The workshop report also recognized that serum ferritin is an acute phase reactant; therefore, concomitant inflammation complicates the use of serum ferritin in the diagnosis of iron deficiency.24,25 To address this, the report provided approaches to adjusting serum ferritin for inflammation using C-reactive protein (CRP) in population surveys.26
In Canada, there has been no recommendation for screening for iron deficiency or anemia in young children in primary care. Therefore, we had an opportunity to assess a screening strategy for iron deficiency in young children, 1 to 3 years of age, using serum ferritin. Our primary objective was to investigate the optimal age for screening by examining the relationship between age and serum ferritin. The second objective was to evaluate the AAP screening recommendation by examining the relationship between age and hemoglobin level and by comparing the diagnostic accuracy of a hemoglobin cutoff for anemia against a serum ferritin cutoff for iron deficiency. Our third objective was to assess the prevalence of acute systemic inflammation as measured by CRP in young children attending scheduled health supervision visits.
Methods
Design and Data Source
A cross-sectional study design was used. Our sample was drawn from the ongoing open longitudinal cohort The Applied Research Group for Kids (TARGet Kids!) (www.targetkids.ca).27 This cohort consists of healthy children from birth to age 5 years recruited while attending scheduled health supervision visits at a TARGet Kids!–affiliated family physician or pediatrician primary care practice in Toronto, Canada. In Canada, there are 11 recommended visits between birth and 5 years.28 Cohort exclusion criteria are health conditions affecting growth (eg, failure to thrive), any chronic health condition (except asthma and mild autism spectrum disorder), severe developmental delay, unscheduled visit due to acute illness, and parents unable to communicate in English. The TARGet Kids! cohort was started in 2008. Details on the cohort profile have been previously published.27
TARGet Kids! research assistants are embedded in each practice site and collect questionnaire data, anthropometric data, and blood samples on each study participant. Parents complete a questionnaire (developed by TARGet Kids! investigators on the basis of the Canadian Community Health Survey29), which contains questions pertaining to child and family sociodemographic information as well as questions related to the child’s diet and feeding practices.
Blood samples for hemoglobin, serum ferritin, and CRP are collected during the health supervision visits, refrigerated at the practice sites and transported to the laboratory at Mount Sinai Services the same day (www.mountsinaiservices.com/). Hemoglobin is analyzed on the Sysmex XN-9000 Hematology Analyzer (Sysmex America, Inc, Kobe, Japan); serum ferritin and high-sensitivity CRP are analyzed on the Roche platform (The Roche Group, Basel, Switzerland). The lower limit of detection for CRP is 0.15 mg/L.
Consent was obtained from parents of all children participating in TARGet Kids!. Ethics approval was obtained from the Research Ethics Boards at the Hospital for Sick Children and St Michael’s Hospital, Toronto. The cohort study is registered at www.clinicaltrials.gov (identifier NCT01869530).
Participants
For this analysis, we included data from children attending 1 of the following scheduled health supervision visits: 12, 15, 18, 24, or 36 months. Data were selected in this way to simulate the AAP recommendations for universal screening for anemia with a measurement of hemoglobin at 12 months, followed by continued risk assessment and selective screening up to 3 years of age.5 Additional exclusion criteria for this study were gestational age <37 weeks, concurrent iron supplementation, no serum ferritin data, and CRP ≥10 mg/L (signifying acute systemic inflammation30,31). The first visit with complete data were used for analysis.
Variables
Age was included in the analyses both as a continuous and as a categorical variable. We focused on children attending health supervision visits at the ages of 12, 15, 18, 24, and 36 months; however, parents rarely bring their children at these exact ages. Therefore, a priori for the analyses, we created more realistic age categories for these visits as follows: 12 to 13, 14 to 16, 17 to 19, 23 to 25, and 34 to 38 months.
Hemoglobin and serum ferritin were used either as continuous or as binary variables on the basis of the planned analysis. Anemia was defined as hemoglobin <110 g/L, and iron deficiency was defined as serum ferritin <12 μg/L, as recommended by the AAP.5
Descriptive variables included child characteristics (age, sex, birth weight, BMI z score [zBMI], day care attendance), child feeding practices (breastfeeding duration, current bottle use, daily cow’s milk intake), maternal characteristics (age, ethnicity, education), and family characteristics (family income, number of siblings).
Statistical Analysis
Means (SD) and proportions were calculated to describe the composition of the study population.
To address the primary objective of investigating the optimal age for screening using serum ferritin, we first calculated mean serum ferritin as well as the proportion of children with serum ferritin <12 μg/L for the total sample and for each of the 5 age categories. We then tested for a difference in the proportion of children with serum ferritin <12 μg/L by performing 4 pairwise comparisons using χ2 analysis. The age category of 12 to 13 months was used as the reference. The Bonferroni correction technique was applied, and a P value <.0125 was considered significant. (This technique lowers the α value when performing several comparisons and accounts for the effect of multiple testing.)
Next, we undertook restricted cubic spline (RCS) analysis to test for a relationship between age and serum ferritin. RCS is a useful test for examining variables that have a nonlinear (curvilinear) relationship with one another. “Knots” are applied to each point at which the slope changes along the curve, allowing for separate regression lines.
Finally, a multivariable linear spline regression model was used to examine the relationship. After nonlinearity was confirmed by loess curves, we selected 3 knot points to correspond to the age at scheduled visits (15, 18, and 24 months). The model was adjusted for variables selected a priori regardless of statistical significance (including child, maternal, and family characteristics).32 In addition, the model was adjusted for CRP, according to currently recommended approaches.26,33 Given nonnormally distributed residuals, serum ferritin was log transformed to fit the model and then back transformed to describe the results.
To address the second objective, we examined the relationship between age and hemoglobin using the same approach described above. Because there was no violation of normality of residuals for the outcome variable hemoglobin, it was not log transformed. We then examined the diagnostic accuracy of hemoglobin compared with serum ferritin. We calculated the sensitivity and specificity (and 95% confidence intervals [CIs]) of the hemoglobin cutoff <110 g/L (definition of anemia) against the serum ferritin cutoff <12 μg/L (definition of iron deficiency) as the criterion measure.
To address the third objective of describing the prevalence of acute systemic inflammation, we calculated the proportion (95% CI) of our cohort with CRP ≥10 mg/L, as well as the mean and median of these elevated CRP values.
A maximum of 10% of subjects had missing data across any variable included in the analysis. Multiple imputation was used for regression analyses. The fully conditional specification method was used to impute missing variables by using a separate conditional distribution for each imputed variable34; the fully conditional specification method employed logistic regression for binary or categorical variables and linear regression for continuous variables. Statistical significance was defined as P < .05; all statistical tests were 2 sided. SAS (SAS Institute, Inc, Cary, NC) version 9.4 was used for statistical analysis.
Results
From the total TARGet Kids! cohort of 6679 children, there were 4116 children enrolled at 1 of the prespecified 5 scheduled health supervision visits (age range: 12–38 months). Of these, 2322 were excluded: 412 had a gestational age of <37 weeks, 36 were receiving iron supplementation, 1874 did not have a blood sample for serum ferritin (parents of these children declined because screening blood tests are not the current standard of care in Canadian primary care), and 59 had a CRP value ≥10 mg/L. For analyses examining the relationship between age (as a continuous variable) and serum ferritin, the sample size was 1735 (Table 1). Child characteristics, diet, and feeding practices, as well as maternal and family characteristics for the full sample are shown in Table 2. For analyses examining age (as a categorical variable) and serum ferritin, the sample size was 1430 (82% of the full sample). Of 1735 children, 280 did not have data for hemoglobin. For analyses examining the relationship between age (as a continuous variable) and hemoglobin, the sample size was 1455 (84% of the full sample); for age (as a categorical variable) and hemoglobin the sample size was 1192 (69% of the full sample).
Participant Recruitment and Selection of Patients for Inclusion
| Characteristic | No. |
|---|---|
| Participants recruited at scheduled health supervision visits: 12, 15, 18, 24, and 36 mo | 4116 |
| Exclusion criteria | |
| Gestational age <37 wk | 412 |
| Receiving iron supplementation | 36 |
| No serum ferritin data | 1874 |
| CRP ≥10 mg/L | 59 |
| Final sample for serum ferritin analyses | 1735 |
| No hemoglobin data | 280 |
| Final sample for hemoglobin analyses | 1455 |
| Characteristic | No. |
|---|---|
| Participants recruited at scheduled health supervision visits: 12, 15, 18, 24, and 36 mo | 4116 |
| Exclusion criteria | |
| Gestational age <37 wk | 412 |
| Receiving iron supplementation | 36 |
| No serum ferritin data | 1874 |
| CRP ≥10 mg/L | 59 |
| Final sample for serum ferritin analyses | 1735 |
| No hemoglobin data | 280 |
| Final sample for hemoglobin analyses | 1455 |
Child and Family Characteristics (N = 1735)
| Value | |
|---|---|
| Child characteristics | |
| Age, mo, mean ± SD | 21.54 ± 8.38 |
| Sex, n (%) | |
| Female | 845 (48.7) |
| Male | 890 (51.3) |
| Birth wt, kg, mean ± SD | 3.36 ± 0.63 |
| zBMI, mean ± SD | 0.125 ± 1.11 |
| Missing, n (%) | 80 (4.6) |
| Daycare attendance, n (%) | |
| Yes | 599 (34.5) |
| No | 1076 (62.9) |
| Missing | 60 (3.5) |
| Child feeding practices | |
| Breastfeeding duration, n (%) | |
| None | 73 (4.2) |
| 0–<6 mo | 343 (19.8) |
| 6–<12 mo | 395 (22.8) |
| ≥12 mo | 841 (48.5) |
| Missing | 83 (4.8) |
| Current bottle use, n (%) | |
| Yes | 793 (45.7) |
| No | 820 (47.3) |
| Missing | 122 (7.0) |
| >2 cups cow’s milk intake, n (%) | |
| Yes | 407 (23.5) |
| No | 1234 (71.1) |
| Missing | 94 (5.4) |
| Maternal and family characteristics | |
| Maternal age, y, mean ± SD | 33.74 ± 4.64 |
| Maternal ethnicity, n (%) | |
| European | 1060 (61.1) |
| Asian | 244 (14.1) |
| African | 73 (4.2) |
| Latin American | 53 (3.1) |
| Middle Eastern | 36 (2.1) |
| Other | 100 (5.7) |
| Missing | 169 (9.7) |
| Maternal education level, n (%) | |
| High school or less | 140 (8.3) |
| Postsecondary | 1547 (89.2) |
| Missing | 48 (2.8) |
| Family income (CAD), n (%) | |
| <14 999 | 23 (1.3) |
| 15 000–29 999 | 82 (4.7) |
| 30 000–79 999 | 1299 (74.9) |
| >80 000 | 210 (12.1) |
| Missing | 121 (7.0) |
| Siblings, n (%) | |
| 0 | 741 (42.7) |
| 1 | 741 (42.7) |
| 2+ | 239 (13.8) |
| Missing data | 14 (0.8) |
| Value | |
|---|---|
| Child characteristics | |
| Age, mo, mean ± SD | 21.54 ± 8.38 |
| Sex, n (%) | |
| Female | 845 (48.7) |
| Male | 890 (51.3) |
| Birth wt, kg, mean ± SD | 3.36 ± 0.63 |
| zBMI, mean ± SD | 0.125 ± 1.11 |
| Missing, n (%) | 80 (4.6) |
| Daycare attendance, n (%) | |
| Yes | 599 (34.5) |
| No | 1076 (62.9) |
| Missing | 60 (3.5) |
| Child feeding practices | |
| Breastfeeding duration, n (%) | |
| None | 73 (4.2) |
| 0–<6 mo | 343 (19.8) |
| 6–<12 mo | 395 (22.8) |
| ≥12 mo | 841 (48.5) |
| Missing | 83 (4.8) |
| Current bottle use, n (%) | |
| Yes | 793 (45.7) |
| No | 820 (47.3) |
| Missing | 122 (7.0) |
| >2 cups cow’s milk intake, n (%) | |
| Yes | 407 (23.5) |
| No | 1234 (71.1) |
| Missing | 94 (5.4) |
| Maternal and family characteristics | |
| Maternal age, y, mean ± SD | 33.74 ± 4.64 |
| Maternal ethnicity, n (%) | |
| European | 1060 (61.1) |
| Asian | 244 (14.1) |
| African | 73 (4.2) |
| Latin American | 53 (3.1) |
| Middle Eastern | 36 (2.1) |
| Other | 100 (5.7) |
| Missing | 169 (9.7) |
| Maternal education level, n (%) | |
| High school or less | 140 (8.3) |
| Postsecondary | 1547 (89.2) |
| Missing | 48 (2.8) |
| Family income (CAD), n (%) | |
| <14 999 | 23 (1.3) |
| 15 000–29 999 | 82 (4.7) |
| 30 000–79 999 | 1299 (74.9) |
| >80 000 | 210 (12.1) |
| Missing | 121 (7.0) |
| Siblings, n (%) | |
| 0 | 741 (42.7) |
| 1 | 741 (42.7) |
| 2+ | 239 (13.8) |
| Missing data | 14 (0.8) |
CAD, Canadian dollars.
For the primary objective, the mean (SD) serum ferritin as well as the proportion of children with serum ferritin <12 μg/L are shown in Table 3 for the total sample and for each of the 5 age categories. Pairwise comparison revealed that the proportion of children with serum ferritin <12 μg/L at the 15-, 18-, and 24-month visits were significantly greater than the proportion at the 12-month visit (all Bonferroni corrected P values were <.0125). To assess the association between age and serum ferritin, the RCS model revealed a nonlinear relationship (P < .0001; Fig 1). The multivariable linear spine regression model, with age as the predictor and log-transformed serum ferritin as the outcome, was adjusted for the following variables: child age, sex, CRP, zBMI, daily cow’s milk intake, breastfeeding duration, current bottle use, and family income (Table 4). The model revealed that from 12 to 15 months, for each 1-month increase in age, serum ferritin levels decreased by 9% (95% CI: 5% to 13%; P < .0001). The rate of change was not significant from 15 to 18 months (P = .5) or from 18 to 24 months (P = .3). From 24 to 38 months, for each 1-month increase in age, serum ferritin increased by 2% (95% CI: 1% to 2%; P < .0001).
Mean ± SD Serum Ferritin and Proportion of Children With Serum Ferritin <12 µg/L
| Age | N | Mean ± SD | % <12 µg/L | Pa |
|---|---|---|---|---|
| Total sample | 1735 | 28.1 ± 19.0 | 12.1 | — |
| Age category, mo | 1430 | — | — | — |
| 12–13 | 358 | 34.4 ± 22.2 | 6.4 | — |
| 14–16 | 220 | 25.3 ± 19.1 | 16.4 | .0004 |
| 17–19 | 252 | 26.4 ± 19.5 | 14.7 | .003 |
| 23–25 | 307 | 24.1 ± 13.9 | 15.6 | .0004 |
| 34–38 | 293 | 29.2 ± 16.7 | 6.1 | .99 |
| Age | N | Mean ± SD | % <12 µg/L | Pa |
|---|---|---|---|---|
| Total sample | 1735 | 28.1 ± 19.0 | 12.1 | — |
| Age category, mo | 1430 | — | — | — |
| 12–13 | 358 | 34.4 ± 22.2 | 6.4 | — |
| 14–16 | 220 | 25.3 ± 19.1 | 16.4 | .0004 |
| 17–19 | 252 | 26.4 ± 19.5 | 14.7 | .003 |
| 23–25 | 307 | 24.1 ± 13.9 | 15.6 | .0004 |
| 34–38 | 293 | 29.2 ± 16.7 | 6.1 | .99 |
—, not applicable.
Pairwise comparisons using χ2 analysis. The age category of 12–13 mo was used as the reference. The Bonferroni correction technique was applied to account for the effect of multiple testing. A P value <.0125 was considered significant.
RCS model of change in serum ferritin by age. The RCS model reveals a nonlinear relationship between age and serum ferritin (P < .0001).
RCS model of change in serum ferritin by age. The RCS model reveals a nonlinear relationship between age and serum ferritin (P < .0001).
Multivariable Linear Spline Regression Model of Change in Serum Ferritin by Age
| Age, mo | βa | SE | 95% CI | P |
|---|---|---|---|---|
| 12–15 | −.09 | 0.02 | −0.13 to −0.05 | <.0001 |
| 15–18 | .01 | 0.02 | −0.03 to 0.05 | .5 |
| 18–24 | −.008 | 0.008 | −0.02 to 0.008 | .3 |
| 24–38 | .02 | 0.004 | 0.01 to 0.02 | <.0001 |
| Age, mo | βa | SE | 95% CI | P |
|---|---|---|---|---|
| 12–15 | −.09 | 0.02 | −0.13 to −0.05 | <.0001 |
| 15–18 | .01 | 0.02 | −0.03 to 0.05 | .5 |
| 18–24 | −.008 | 0.008 | −0.02 to 0.008 | .3 |
| 24–38 | .02 | 0.004 | 0.01 to 0.02 | <.0001 |
Adjusted for child age, sex, CRP, zBMI, daily cow’s milk intake, breastfeeding duration, current bottle use, and family income.
Based on log-transformed serum ferritin.
For the second objective, the mean hemoglobin (SD) as well as the proportion of children with hemoglobin <110 g/L are shown in Table 5 for the total sample and for each of the 5 age categories. The age category 12 to 13 months had the highest proportion of children with hemoglobin values <110 g/L (19%), and only the 34 to 38 months age category was significantly lower (5%; Bonferroni corrected P value <.0004). To assess the association between age and hemoglobin, the RCS model revealed a linear relationship (P = .7). Although this relationship was linear, our goal was to show the effects of age on hemoglobin by using the same knot points (corresponding to age at scheduled visits) used in the analysis for age and serum ferritin. The multivariable linear spine regression model, with age as the predictor and hemoglobin as the outcome, was adjusted for the same variables. The rate of change was not significant from 12 to 15 months (P = .4), from 15 to 18 months (P = .7), or from 18 to 24 months (P = .8). From 24 to 38 months, for each 1-month increase in age, hemoglobin increased by 20% (95% CI: 9% to 32%; P = .0006; Table 6). Comparing the diagnostic accuracy of a hemoglobin cutoff of <110 g/L against the serum ferritin cutoff of <12 μg/L as the criterion measure resulted in a sensitivity of 25% (95% CI: 19% to 32%) and a specificity of 89% (95% CI: 87% to 91%).
Mean ± SD Hemoglobin and Proportion of Children With Hemoglobin <110 g/L
| Age | N | Mean ± SD | % <110 g/L | Pa |
|---|---|---|---|---|
| Total sample | 1455 | 118.9 ± 8.6 | 12.5 | — |
| Age category, mo | 1192 | — | — | — |
| 12–13 | 308 | 117.7 ± 8.8 | 19.2 | — |
| 14–16 | 173 | 119.3 ± 8.2 | 9.3 | .02 |
| 17–19 | 219 | 118.7 ± 8.8 | 14.2 | .5 |
| 23–25 | 252 | 118.8 ± 7.6 | 10.3 | .02 |
| 34–38 | 240 | 121.1 ± 7.5 | 5.0 | <.0004 |
| Age | N | Mean ± SD | % <110 g/L | Pa |
|---|---|---|---|---|
| Total sample | 1455 | 118.9 ± 8.6 | 12.5 | — |
| Age category, mo | 1192 | — | — | — |
| 12–13 | 308 | 117.7 ± 8.8 | 19.2 | — |
| 14–16 | 173 | 119.3 ± 8.2 | 9.3 | .02 |
| 17–19 | 219 | 118.7 ± 8.8 | 14.2 | .5 |
| 23–25 | 252 | 118.8 ± 7.6 | 10.3 | .02 |
| 34–38 | 240 | 121.1 ± 7.5 | 5.0 | <.0004 |
—, not applicable.
Pairwise comparisons using χ2 analysis. The age category of 12–13 mo was used as the reference. The Bonferroni correction technique was applied to account for the effect of multiple testing. A P value <.0125 was considered significant.
Multivariable Linear Spline Regression Model of Change in Hemoglobin by Age
| Age, mo | βa | SE | 95% CI | P |
|---|---|---|---|---|
| 12–15 | .28 | 0.31 | −0.33 to 0.90 | .4 |
| 15–18 | .14 | 0.31 | −0.47 to 0.74 | .7 |
| 18–24 | −.03 | 0.13 | −0.29 to 0.22 | .8 |
| 24–38 | .20 | 0.06 | 0.09 to 0.32 | .0006 |
| Age, mo | βa | SE | 95% CI | P |
|---|---|---|---|---|
| 12–15 | .28 | 0.31 | −0.33 to 0.90 | .4 |
| 15–18 | .14 | 0.31 | −0.47 to 0.74 | .7 |
| 18–24 | −.03 | 0.13 | −0.29 to 0.22 | .8 |
| 24–38 | .20 | 0.06 | 0.09 to 0.32 | .0006 |
Adjusted for child age, sex, CRP, zBMI, daily cow’s milk intake, breastfeeding duration, current bottle use, and family income.
Based on log-transformed serum ferritin.
For the third objective, 3.3% (59 of 1794; 95% CI: 2.5% to 4.2%) had a CRP ≥10 mg/L. For these 59 children, the mean CRP value was 22 mg/L (SD: 13), and the median was 19 mg/L (range 10–64).
Discussion
We evaluated a screening strategy for iron deficiency in young children, 1 to 3 years of age, attending scheduled primary care health supervision visits using serum ferritin. In this population, we found that serum ferritin is lowest between 15 and 24 months; the change in hemoglobin between 12 and 24 months is not statistically or clinically significant; the hemoglobin cutoff of <110 g/L is specific but not sensitive for iron deficiency, as compared with the serum ferritin cutoff of <12 μg/L; and the prevalence of acute systemic inflammation is low (as measured by CRP ≥10 mg/L).
The AAP recommends universal screening for anemia through measurement of hemoglobin.5 For children 1 to 3 years of age, anemia has long been defined as a hemoglobin concentration of <110 g/L.5,35 Using this cutoff, we found the highest prevalence of anemia at 12 months of age. However, the mean hemoglobin for the entire cohort (118 g/L) was stable between 12 and 24 months. Furthermore, using the hemoglobin cutoff of <110 g/L to indicate iron deficiency resulted in a high rate of false-positives (75%) as compared with a serum ferritin cutoff of <12 μg/L. These findings suggest that hemoglobin is not an ideal measure of iron deficiency, and the cutoff of 110 g/L for anemia may need to be reconsidered. Hemoglobin at 12 months of age may be more reflective of an infant’s developing capacity for iron homeostasis, rather than pathologic states, which may improve with increasing age.36 In addition, anemia as measured by hemoglobin, lacks specificity for iron deficiency, because there are other causes of low hemoglobin such as hemoglobinopathies and other nutritional deficiencies.5
It is known that acute inflammation may falsely elevate serum ferritin.25,26 CRP values ranging from 3 to 30 mg/L have been suggested for determining the validity of a serum ferritin level in the diagnosis of iron deficiency.30,31,33 In our assessment, we used a cutoff for CRP of ≥10 mg/L. At this cutoff, acute inflammation appears to be uncommon among healthy young children attending scheduled primary care health supervision visits. Therefore, serum ferritin appears to be valid for screening for iron deficiency in this population. However, for screening other populations of children, such as those at risk for acute or low-grade (chronic) inflammation, serum ferritin may not be valid. For example, a relationship between low-grade inflammation (as measured by CRP between 1 and 10 mg/L) and obesity beginning by 3 years of age has recently been identified.37 In addition, for population-level surveys of prevalence of iron deficiency, several approaches have been proposed to adjust serum ferritin for the influence of inflammation.26,33 In our analysis, we used both exclusion of high values of CRP and the addition of CRP in our adjusted regression models.
On the basis of these data, we propose a reexamination of the AAP policy and a screening strategy to detect iron deficiency at the 15- or 18-month health supervision visit by using serum ferritin. Given our finding that iron deficiency peaks between 15 and 24 months, this would allow most cases to be identified. Although the AAP recommends concurrent measurement of CRP when interpreting serum ferritin, for children at low risk for acute or low-grade inflammation, this may not be necessary.
The World Health Organization recommends using serum ferritin for measuring iron status in population surveys and program evaluation.21 Using serum ferritin to screen for iron deficiency in primary care settings is also supported by the classic screening principles published by the World Health Organization in 1968.38 Iron deficiency is an important public health problem, with high prevalence in early childhood and is associated with detrimental consequences to cognitive, emotional, social, and motor development.1,–17 Nonanemic iron deficiency is a known latent stage that can be identified before progression to iron-deficiency anemia. Serum ferritin is an inexpensive, readily available, noninvasive test with high specificity for iron deficiency.24,25 Oral iron supplements are widely available in a variety of administration forms and effective in correcting iron deficiency.39 Finally, policies recommending screening for iron deficiency have already been established by professional organizations such as the AAP.5
It should be noted that the serum ferritin level at which the diagnosis of iron deficiency can be confirmed remains unknown. We have previously described reference intervals for hemoglobin and serum ferritin following the Clinical and Laboratory Standards Institute guidelines and found that ∼10% of children 1 to 3 years of age were misclassified (underestimated) by using the lower limit of the reference intervals rather than the currently recommended AAP cutoff values for hemoglobin and serum ferritin.40 We have also examined the relationship between serum ferritin and hemoglobin in this age group and found that a serum ferritin cutoff of 18 to 24 μg/L corresponds to the point at which hemoglobin plateaus at ∼120 g/L.41
Strengths of our study include the large sample size of 1735 children attending health supervision visits, with data collected prospectively from several primary care practices where screening for iron deficiency is not current standard of practice. Our study was limited by the lack of additional laboratory tests to diagnose other causes of anemia in children who were not iron deficient. In addition, although serum ferritin is a good measure of iron status, it is not known if it is an accurate measure of brain tissue iron deficiency.8 Future research is warranted regarding several issues not examined in our study, including cost, convenience, and missed opportunities to identify other hematologic disorders when using serum ferritin rather than hemoglobin for screening.
Conclusions
Findings from our study suggest that serum ferritin, rather than hemoglobin, may be a more promising screening test for iron deficiency in early childhood. The optimal age for screening using serum ferritin appears to be the 15- or 18-month health supervision visit. For children at low risk for acute inflammation, concurrent measurement of CRP may not be necessary. Further evaluation of this screening strategy is warranted.
Appendix
Collaborators
The following members of the TARGet Kids! Collaboration are nonauthor contributors.
Science contributors: Mary Aglipay, Laura N. Anderson, David W.H. Dai, Karen Eny, Charles Keown-Stoneman, Christine Kowal, and Dalah Mason. Site investigators: Murtala Abdurrahman, Gordon Arbess, Tony Barozzino, Imaan Bayoumi, Sylvie Bergeron, Joey Bonifacio, Ashna Bowry, Caroline Calpin, Douglas Campbell, Brian Chisamore, Evelyn Constantin, Karoon Danayan, Paul Das, Anh Do, Kathleen Doukas, Sloane Freeman, Sharon Gazeley, Rajesh Girdhari, Charlie Guiang, Leah Harrington, Sheila Jacobson, Paul Kadar, Tara Kiran, Holly Knowles, Sheila Lakhoo, Margarita Lam-Antoniades, Eddy Lau, Denis Leduc, Renata Leong, Fok-Han Leung, Patricia Li, Elise Mok, Rosemary Moodie, Katherine Nash, Sharon Naymark, James Owen, Marty Perlmutar, Andrew Pinto, Cristina Pop, Michelle Porepa, Adam Pyle, Julia Rackal, Noor Ramji, Danyaal Raza, Jane Ridley, Alana Rosenthal, Caroline Ruderman, Janet Saunderson, Michael Sgro, Barbara Smiltnieks, Carolyn Taylor, Joshua Tepper, Stephen Treherne, Suzanne Turner, Meta van den Heuvel, Joanne Vaughan, Karim Vellani, Zoe Von Aesch, William Watson, Karen Weyman, Patricia Windrim, Peter Wong, John Yaremko, Ethel Ying, and Elizabeth Young.
Dr Oatley contributed to the conception and design, interpretation of data, and drafting the initial manuscript; Drs Borkhoff and Parkin contributed equally as cosenior authors to the conception and design, acquisition of data, analysis and interpretation of data, drafting the initial manuscript, and revising the manuscript critically for important intellectual content; Ms Chen contributed to the analysis and interpretation of data and revision of the manuscript critically for important intellectual content; Drs Macarthur and Persaud contributed to the conception and design, interpretation of data, and revision of the manuscript critically for important intellectual content; Drs Birken and Maguire contributed to the conception and design, acquisition of data, interpretation of data, and revision of the manuscript critically for important intellectual content; and all authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.
This trial has been registered at www.clinicaltrials.gov (identifier NCT01869530).
FUNDING: Supported by a grant from the Canadian Institutes of Health Research (FRN 115059). Funding to support The Applied Research Group for Kids was provided by multiple sources, including the Canadian Institutes for Health Research, The Hospital for Sick Children Foundation (which supports the Pediatric Outcomes Research Team), and the St Michael’s Hospital Foundation. These funders had no role in the design and conduct of this work; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the article; nor the decision to submit the article for publication.
Acknowledgments
TARGet Kids! Collaboration details may be found on our website (www.targetkids.ca). We thank all participating 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, methodologists, biostatisticians, data management personnel, laboratory management personnel, and advisory committee members.
Comments