Professional Documents
Culture Documents
net/publication/23571503
CITATIONS READS
143 465
3 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Phuong Hong Nguyen on 30 May 2016.
ABSTRACT feeding, ie, at least 5 times/d for young children, provision of high-
Background: Micronutrient interventions have received much atten- quality complementary foods that are rich in both macro- and
tion as a cost-effective and promising strategy to improve child health, micronutrients, and promotion of exclusive breastfeeding, require
but their roles in improving child growth remain unclear. behavior change that is often difficult to accomplish (2, 3). It is in
Objective: Meta-analyses of randomized controlled trials were con- this context that the potential of micronutrient interventions as
ducted to evaluate the effect of micronutrient interventions on the cost-effective measures to prevent child undernutrition has re-
growth of children aged ,5 y old. cently received considerable attention (3–5).
Am J Clin Nutr 2009;89:191–203. Printed in USA. Ó 2009 American Society for Nutrition 191
192 RAMAKRISHNAN ET AL
METHODS calculated from SE or 95% CI. If studies did not report the SD of the
anthropometric changes, we calculated the SD for change, as-
Identification of studies suming that the correlation between the pre- and post-test variances
The studies considered for possible inclusion in the current meta- was equal to the average correlation found in available studies.
analyseswereidentifiedbycombiningtheresultsof2separatesearch To make sure our assumptions did not bias the results, we per-
strategies that were completed by April 15, 2008, with the PubMed formed the sensitivity analysis noted by Sachdev et al (9) using the
(National Library of Medicine, Bethesda, MD) database (1966 to following different assumptions for the correlation between the pre-
present). The first search for each intervention included the word and post-test variances: 1) using the average correlation in data sets
vitamin A or iron or zinc or multi-micronutrients in the title and the from available studies, 2) assuming a correlation ¼ 0.5 (14), 3)
words growth, infant, or child or children in any field. The second assuming no correlation (14), 4) using postintervention values to
search contained all articles that had the word supplement or sup- calculate the effect size, and 5) calculated effect size only for the
plemental in the title and the words growth or weight or length or subsample of studies that reported the changes and SD of change.
height in any field. The results of these 2 searches were then merged Effect sizes were calculated for individual studies by dividing the
and examined for inclusion and exclusion as described below. Us- difference between the mean change in treatment and control groups
ing similar specifications, we repeated the search using EMBASE by the pooled SD. This value is known as Cohen’s effect size or
(Elsevier, Amsterdam. Netherlands) and the Cochrane review for Cohen’s d, and it is useful in meta-analyses because it eliminates the
randomized controlled trials (Wiley InterScience, Hoboken, NJ). problems of units of measurement and duration, which may vary
Additional studies that were identified through the bibliographies of across studies (15). The overall mean effect size and 95% CI across
review articles were also included. studies was then estimated by assuming a random effects model that
TABLE 2
Characteristics of 19 data sets from 17 intervention studies included in the meta-analyses of vitamin A and child growth of children aged ,5 y old1
Mean Mean Mean Mean Mean
Mean initial initial initial initial initial
Authors Year Country Subjects initial age Dose Doses Duration height weight WAZ HAZ WHZ
n mo mg n wk cm kg
Fahmida2 (20) 2007 Indonesia 378 5 30 1 24 — — 20.48 21.02 0.40
Alarcon2 (21) 2004 Peru 229 17 60 1 18 76.8 10.4 20.18 20.82 0.47
Pangaribuan (22) 2003 Indonesia 400 36.5 60 1 16 89.2 12.1 21.52 21.48 20.72
Villamor (23) 2002 Tanzania 687 18.7 60 4 52 76.1 9.4 21.62 22.1 20.67
Yang3 (24) 2002 China 63 48 0.2 260 52 95.15 13.78 21.31 21.55 20.29
Rahman (I) (25) 2002 Bangladesh 317 23.7 60 1 24 — — 22.38 22.31 21.24
Rahman (II)4 (25) 2002 Bangladesh 336 23.7 60 1 24 — — 22.46 22.51 21.22
Hadi (26) 2000 Indonesia 1407 27 60 6 16 — — — — —
Smith (27) 1999 Central 51 44 3 24 24 92 13.9 21.04 21.94 0.23
America
Donnen (28) 1998 Zaire 235 36 60 2 52 — — — — —
West (29) 1997 Nepal 3497 36 60 4 64 84.3 10.83 22.23 22.46 21.01
Fawzi (30) 1997 Sudan 21,251 39 60 3 72 88.8 11.79 21.92 21.99 20.91
Bahl (31) 1997 India 900 26.3 60 1 12 77.9 9.04 22.87 22.83 21.38
Kirkwood/Health5 (32) 1996 Ghana 1500 30 60 3 52 — — 21.78 22.04 20.59
Kirkwood/Survival5 (32) 1996 Ghana 15,000 30.6 60 3 104 — — 21.82 — —
Ramakrishnan (33) 1995 India 592 17.7 60 2 52 — — 21.79 21.34 21.22
Lie (34) 1993 China 172 21 60 2 52 — — — — —
Rahmathullah (35) 1991 India 15,419 33 2.5 52 52 — — — — —
West (36) 1988 Indonesia 2766 36 60 2 52 — — — — —
1
WAZ, weight-for-age z score; HAZ, height-for-age z score; WHZ, weight-for-height z score.
2
Intervention group received vitamin A 1 iron 1 zinc; control group received iron 1 zinc.
3
Intervention group received vitamin A 1 calcium 1 zinc; control group received calcium 1 zinc.
4
Intervention group received vitamin A 1 zinc; control group received zinc.
5
Two data sets (‘‘Health’’ and ‘‘Survival’’) from one publication.
194 RAMAKRISHNAN ET AL
placebo and iron 1 zinc to zinc only. Three studies used fortifica-
tion-based strategies, namely, comparisons of iron-fortified formula
with non-fortified formula (48, 52, 56); the one by Adish et al (49)
compared cooking in an iron pot with cooking in an aluminum pot.
x 6 SD: 24.3 6
The duration of intervention varied from 8 to 52 wk (
10.4 wk). The initial ages of the children ranged from 1 to 48 mo.
Effect sizes for change in height were calculated for 34 data sets
(Figure 2) and ranged from 25.00 to 0.99. The weighted mean
effect size for height was 20.01 (95% CI: 20.17, 0.15) and in-
creased to 0.01 (95% CI: 20.08, 0.10) after excluding the study by
Majumdar et al (41) that was considered an outlier because its
effect size (25.00; 95% CI: 25.80, 24.20) was more than 5 times
smaller than the next smallest effect size.
Effect sizes for weight gain ranged from 28.14 to 2.18 and,
similar to height, the study by Majumdar et al (41) was defined an
outlier because of its extreme value. The overall weighted mean
effect size with and without the outlier was 0.07 (95% CI: 20.17,
0.30) and 0.08 (95% CI: 20.11, 0.27), respectively. Fourteen
studies had sufficient data to calculate effect sizes for WHZ; the
Zinc
FIGURE 1. Effect sizes for height gain in vitamin A intervention trials
among children aged ,5 y old. Data are presented as means with 95% CIs. Of the 602 potential studies identified from the database search
All (I) is a calculation with inclusion of outliers [Pangaribuan et al (22) and and 3 studies from bibliographies of review articles, only 89 were
Smith et al (27)]; All (II) is a calculation without outliers. zinc intervention trials among young children; an additional 46
studies were excluded because of zinc not being a unique factor in
suggested weight loss (effect size ¼ 20.34; 95% CI: 20.66, treatment (n ¼ 7), underlying diseases or severe malnutrition (n ¼
20.01) among those who received vitamin A compared with 31), lack of growth data (n ¼ 6), or short duration of intervention
placebo in the subsample of studies conducted in underweight and follow up (n ¼ 2), resulting in a final sample of 43 studies (20,
children (baseline WAZ , 22; n ¼ 3) compared with no effect 21, 24, 27, 37–40, 46, 50, 59–91) with 56 data sets for inclusion in
(effect size 0.00; 95% CI: 20.17, 0.17) among the studies where the zinc meta-analysis (Table 1). Selected characteristics of the
baseline WAZ 22 (n ¼ 10). There was no overall effect on studies included in this meta-analysis are described in Table 4. All
change on WHZ (weighted mean effect size was 0.01; 95% CI: studies were published between 1976 and 2008, of which 15 were
20.06, 0.09), but data were available for only 5 studies. conducted in Asia, 13 in Latin America and the Caribbean, 8 in
North America and Europe, and 7 in Africa. Mean initial age
ranged from 0 to 48 mo with most of the studies being conducted in
Iron younger children (initial age , 24 mo). Eight studies were con-
Sufficient information was available from the 36 data sets from ducted in newborns that were either term or small-for-gestational
27 studies (10–13, 20, 37–58) for calculating the effect of iron age infants. The intervention was provided daily as a liquid sup-
supplementation on child growth (Table 3). Most of the studies plement (syrup mixed into a beverage); the dosage, however,
were conducted in developing countries since the 1990s with more varied from as low as 20 mg/wk to 20 mg/d. A few studies also
than half (n ¼ 17) in Asia, followed by Africa (n ¼ 3) and Latin provided zinc in fortified formula (24, 68, 91) or as a fortified
America and the Caribbean (n ¼ 3). Two studies were conducted cereal porridge (60).
in North America and 3 in Europe. Most studies delivered iron in Similar to iron, several studies had more than one intervention
the form of a tablet or syrup taken daily, and the most common resulting in 2 or 3 data sets each. For example, Wuehler et al (59)
dosage was 10 mg/d; higher doses (20–60 mg/d) were used in compared 3 different doses (3, 7, and 10 mg) of zinc with placebo,
some of the studies with children older than 15 mo. Aweekly dose and Brown et al (60) compared 2 modes of interventions allowing
was used in 2 studies (40, 51). Seven studies (20, 37–40, 46, 50) us to compare zinc fortification in iron-fortified cereal porridge
also used different combinations of iron and zinc and therefore plus MM versus MM alone, and zinc supplementation added in
yielded 2 data sets each, namely, comparisons of iron only to MM versus MM alone. Similarly, several studies used different
MICRONUTRIENT INTERVENTIONS AND CHILD GROWTH 195
TABLE 3
Characteristics of the 36 data sets from 27 intervention studies included in the meta-analysis of iron and child growth of children aged ,5 y old1
Mean Mean Mean Mean Mean Mean Mean
initial initial initial initial initial initial initial
Authors Year Country Subjects age Dose2 Duration height weight WAZ HAZ WHZ Hb
n mo mg wk cm kg g/L
Fahmida3 (20) 2007 Indonesia 378 5 10 24 — — 0.05 20.97 0.40 96.6
Wasantwisut (I) (37) 2006 Thailand 306 4.5 10 24 62.40 6.60 20.11 20.66 0.47 117.4
Wasantwisut (II)3 (37) 2006 Thailand 303 4.5 10 24 61.95 6.45 20.20 20.69 0.40 113.8
Berger (I) (38) 2006 Vietnam 399 5.9 10 24 63.85 6.90 20.58 21.02 0.27 109.7
Berger (II)3 (38) 2006 Vietnam 385 5.9 10 24 63.80 6.90 20.57 21.05 0.31 109.1
Uronto (11) 2005 Indonesia 134 9.2 10 23 — — 20.95 20.73 20.49 109.1
Smuts (10) 2005 South A 99 8.4 10 24 68.70 9.10 0.48 20.73 1.30 112.3
Lopez de Romana (13) 2005 Peru 146 9 10 24 — — 20.27 20.99 0.62 106.0
Hople (12) 2005 Vietnam 148 8 10 24 — — 20.91 20.93 20.27 99.85
Lind (I) (39) 2004 Indonesia 327 6 10 24 — — 20.41 20.35 20.07 —
Lind (II)3 (39) 2004 Indonesia 323 6 10 24 — — 20.37 20.35 20.01 —
Black (I) (40) 2004 Bangladesh 43 6.5 20 (/wk) 24 — — 21.05 21.20 0.10 105.5
Black (II)3 (40) 2004 Bangladesh 53 6.5 20 (/wk) 24 — — 21.00 21.20 0.10 105.0
Majumdar (41) 2003 India 100 15 2 (/kg/d) 16 — — — — — 139.0
Friel (42) 2003 Canada 77 1 7.5 20 53.50 4.75 1.05 20.23 1.36 125.0
combinations of zinc and iron (20, 37–40, 46, 50) that yielded 2 did not identify any predictors that could explain the variation in
data sets each, namely, comparisons of zinc only to placebo and effect sizes.
zinc 1 iron to iron only. The duration of intervention varied from 8 Effect sizes for zinc intervention on weight gain were calcu-
to 64 wk with a median of 24 wk. lated for 54 data sets (n ¼ 41 studies) and ranged from 20.78 to
Effect sizes for change in height were calculated for 53 data sets 3.89. Approximately 61% of the data sets (n ¼ 33) had positive
(n ¼ 40 studies) and ranged from 20.80 to 1.12 (Figure 3). Thirty effect sizes, and 10 were statistically significant. The study by
(56.6%) studies had a positive effect size and 11 were statistically Shrivastava et al (86) was considered an outlier because of the
significant; however, the overall weighted mean effect (0.07; 95% large effect size (3.89; 95% CI: 2.95, 4.83) that was at least 2 times
CI: 20.03, 0.17) was small and not statistically significant. Sen- larger than the next largest effect size. The overall weighted mean
sitivity analyses using different assumptions did not alter the ob- effect sizes were 0.09 (95% CI: 20.11, 0.25) and 0.06 (95% CI:
served effect sizes and conclusions (results not shown). There was 20.10, 0.23) with and without the outlier, respectively, indicating
significant heterogeneity (P , 0.001), but the stratified analysis no statistically significant effect of zinc on weight gain. Sensitivity
196 RAMAKRISHNAN ET AL
WHZ. The overall weighted mean effect sizes for the combination
of vitamin A and zinc were 0.10 (95% CI: 20.41, 0.61) for height,
0.11 (95% CI: 20.58, 0.80) for weight, and 0.05 (95% CI: 20.12,
0.22) for WHZ. The overall weighted mean effect sizes for the
combination of iron and folic acid were 0.16 (95% CI: 20.05,
0.38) for height and 0.79 (95% CI: 20.35, 1.94) for weight.
n mo mg/d wk cm kg lmol/L
Wuehler (I)2 (59) 2008 Ecuador 251 21 3 24 77.45 9.75 21.25 22.30 20.15 —
Wuehler (II) (59) 2008 Ecuador 253 21 7 24 77.40 9.70 21.25 22.30 20.15 —
Wuehler (III) (59) 2008 Ecuador 253 21 10 24 77.40 9.70 21.25 22.25 20.20 —
Brown (I)3 (60) 2007 Peru 302 7.5 3 24 65.2 7.6 20.65 21.19 0.60 11.96
Brown (II) (60) 2007 Peru 302 7.5 3 24 65.4 7.6 20.75 21.19 0.46 11.87
Fahmida (20) 2007 Indonesia 399 5 10 24 — 0.0 20.06 20.99 0.28 15.30
Berger (I) (38) 2006 Vietnam 393 5.9 10 24 63.9 6.9 20.57 21.01 0.26 14.50
Berger (II)4 (38) 2006 Vietnam 391 5.9 10 24 63.8 6.9 20.58 21.07 0.31 14.18
Heinig (61) 2006 USA 85 4 5 24 64.1 7.0 0.74 0.49 0.30 —
Olney (I)5 (62) 2006 Zanzibar 433 8.8 10 52 — 0.0 21.25 21.45 20.15 —
Olney (II) (62) 2006 Zanzibar 443 8.8 10 52 — 0.0 21.15 21.45 20.06 —
Silva (63) 2006 Brazil 58 23.5 10 16 — 0.0 0.00 21.95 21.20 8.61
Wasantwisut (I) (37) 2006 Thailand 304 4.5 10 24 62.3 6.6 20.18 20.70 0.42 11.30
Wasantwisut (II)4 (37) 2006 Thailand 305 4.5 10 24 62.1 6.5 20.13 20.66 0.45 10.85
Brooks (64) 2005 Bangladesh 1665 5.3 70 (/wk) 52 62.7 6.4 20.80 21.10 0.04 9.80
Gardner (65) 2005 Jamaica 114 19 10 24 77.1 8.6 20.87 21.43 21.65 —
Alarcon (21) 2004 Peru 213 17 3 18 76.8 10.5 20.26 21.04 0.52 —
weight change did not report WHZ, which may explain some of
FIGURE 3. Effect sizes for height gain in zinc intervention trials among the observed publication bias.
children aged ,5 y old. Data are presented as means with 95% CIs.
DISCUSSION
A, iron, and zinc showed a similar mean effect size of 20.002
The key findings of this review are summarized in Table 6.
(95% CI: 20.08, 0.07).
Interventions containing iron only, vitamin A only, and combi-
For all 3 outcomes, there was no evidence of heterogeneity
nations of iron and zinc, iron and vitamin A, and zinc and vitamin
(P ¼ 0.08 for height and weight and P ¼ 0.51 for WHZ).
A do not improve growth in height, weight, or WHZ in children
Sensitivity analyses using different assumptions confirmed the
aged ,5 y. Interventions containing zinc only have a small posi-
overall findings.
tive effect (0.06; 95% CI: 0.006, 0.11) on change in WHZ but do
not improve height or weight gain in young children. Finally, MM
interventions have a small effect only on growth in height (0.09;
Publication bias 95% CI: 0.008, 0.17).
Except for the studies of effects of zinc on WHZ, the funnel plot A major strength of the current review is the sample size for the
in each meta-analysis was relatively symmetrical, thus indicating various meta-analyses, except for the 2-way combinations. This is
the absence of publication bias (data not shown). The Egger’s especially true for MM interventions, because several studies were
weighted regression method and Begg’s rank correlation method completed since the previous review that had only 5 studies, of
further confirmed the symmetrical observation of the funnel plot which only 3 were conducted in young children (98, 99, 101). The
with P values . 0.05. For the effect of zinc on WHZ, the funnel plot present review included 20 studies that contributed 27 data sets,
was quite asymmetrical, indicating evidence of publication bias (P and many of them were conducted in infants. Although most
of Egger’s method ¼ 0.01 and P of Begg’s method ¼ 0.02). The of these studies (19 data sets) provided the intervention as ‘‘sup-
effects of zinc on WHZ were reported in only 22 studies (32 data plements,’’ some used food-based approaches and the findings did
sets); many studies that reported effects of zinc on height and not differ by mode of delivery (results not shown).
MICRONUTRIENT INTERVENTIONS AND CHILD GROWTH 199
TABLE 5
Selected characteristics of 27 data sets from 20 intervention studies included in the meta-analyses of multiple micronutrients on growth of children aged
,5 y old1
Mean Mean Mean Mean Mean Mean Mean
initial initial initial initial initial initial initial
Authors Year Country Subjects age Duration height weight WAZ HAZ WHZ Hb
n mo wk cm kg g/L
Begin (I)2 (93) 2008 Guatemala 69 6.8 32 64.0 6.89 21.07 21.52 0.18 115.5
Begin (II) (93) 2008 Guatemala 63 6.8 32 64.1 6.94 21.05 21.54 0.22 115.5
Fahmida (20) 2007 Indonesian 386 5 24 — — 20.59 21.04 0.28 95.6
Giovannini (92) 2006 Cambodia 127 6 52 — — 20.87 20.79 20.38 101.2
Olney (62) 2006 Zanzibar 114 8.8 52 — — 21.25 21.50 . 93.9
Faber (94) 2005 South Africa 288 9 24 68.8 9.05 0.33 20.86 1.23 111.0
Hop Le (I)3 (12) 2005 Vietnam 149 8 24 — — 20.94 20.92 20.31 99.4
Hop Le (II) (12) 2005 Vietnam 150 8 24 — — 20.93 20.89 20.31 98.7
Lopez de Romana (I)3 (13) 2005 Peru 135 9 24 — — 20.16 20.79 0.59 106.0
Lopez de Romana (II) (13) 2005 Peru 137 9 24 — — 20.20 20.91 0.65 105.0
Nesamvuni (95) 2005 South Africa 36 24 52 76.15 10.2 21.63 22.78 0.09 111.5
Smuts (I)3 (10) 2005 South Africa 99 8.4 24 68.1 8.7 0.21 20.82 1.08 111.2
Smuts (II) (10) 2005 South Africa 96 8.4 24 68.25 8.9 0.39 20.77 1.24 112.4
Uronto (I)3 (11) 2005 Indonesian 132 9.2 23 — — 20.88 20.69 20.44 108.7
Overall, our findings confirm the results of the earlier meta- The most surprising result of the current review is the one for
analyses conducted by Ramakrishnan et al (8) but contradict the zinc interventions. We did not find the significant positive
findings presented by Brown et al (6) regarding the benefits of zinc effects that were seen both in height and weight gain in the
supplementation. The findings for vitamin A are not surprising; earlier meta-analyses by Brown et al (6). Our findings are
we added 2 new studies (20, 21) and only one study (103) in the timely and important in light of the fact that the Lancet series
earlier review was not eligible because it was conducted in older on Maternal and Child Undernutrition recommended zinc
children (9–12 y). In the case of iron, although the results are supplementation as an effective intervention to reduce mor-
similar, the current meta-analyses included 13 new studies (10– bidity and prevent stunting (3). The Lancet series estimated
13, 20, 37, 38, 40, 42, 44, 52, 54, 56), and 7 studies (103–109) that universal coverage with zinc supplementation in the 36
included in the earlier review were ineligible for the current meta- countries with the highest burden of undernutrition would re-
analyses because they were conducted among older children. Our duce the prevalence of stunting by 9.1% at 12 mo, by 15.5% at
findings are also consistent with the conclusions by Sachdev et al 24 mo, and by 17% at 36 mo (3). The recommendation of zinc
(9), who conducted a meta-analysis of iron supplementation trials supplementation as an effective intervention to reduce mor-
on physical growth and found no positive effect. That study in- bidity rates was based on a new meta-analysis that showed very
cluded 25 studies, of which 15 were conducted among children large preventive effects, namely, reductions of 14% and 20% in
,5 y old, in contrast to the 27 studies that contributed 36 data sets episodes of diarrhea and lower respiratory infections, re-
in the current meta-analysis. Many of these were the recently spectively (3). In the case of zinc supplementation as an in-
completed IRIS trials that were not included in the earlier meta- tervention to prevent stunting, no new analyses were carried
analyses. It is important to note that there were no significant out; instead, the Lancet series accepted the results of the meta-
differences when we stratified by baseline hemoglobin level for analysis by Brown et al (6), which reported an effect of zinc
the various study populations (results not shown). supplementation on growth rates in height of 0.35 (95% CI:
200 RAMAKRISHNAN ET AL
TABLE 6
Summary estimates of mean effect sizes (95% CI) for all growth outcomes by type of micronutrient intervention1
Outcomes Vitamin A Iron Zinc Iron 1 zinc MM
Height gain (cm/y) 0.08 (20.18, 0.34) 0.008 (20.08, 0.10) 0.07 (20.03, 0.17) 0.0004 (20.21, 0.21) 0.09 (0.008, 0.17)
Weight gain (kg/y) 20.03 (20.23, 0.18) 0.08 (20.11, 0.27) 0.06 (20.10, 0.23) 0.05 (20.10, 0.21) 0.04 (20.05, 0.12)
Changes in WHZ (/y) 0.01 (20.06, 0.09) 20.02 (20.15, 0.12) 0.06 (0.006, 0.11) 0.07 (20.05, 0.18) 20.001 (20.07, 0.07)
1
MM, multiple micronutrients; WHZ, weight-for-height z score.
MICRONUTRIENT INTERVENTIONS AND CHILD GROWTH 201
breastfeeding for survival and reduced morbidity but at the same 14. Follmann D, Elliott P, Suh I, Cutler J. Variance imputation for over-
time noted that breastfeeding promotion interventions do not views of clinical trials with continuous response. J Clin Epidemiol
1992;45:769–73.
improve growth. Of the nutrition interventions that we have at our 15. Hunter JE, Schmidt FL. Methods of meta-analysis: correcting error and
disposal, only direct efforts to improve complementary feeding bias in research findings. Newbury Park, CA: Sage Publications, 1990.
have been shown to improve growth. On the basis of a limited 16. Hedges LV, Pigott TD. The power of statistical tests in meta-analysis.
number of studies, it is known that education and counseling of Psychol Methods 2001;6:203–17.
caretakers in food-secure populations can improve growth in 17. World Health Organization. Physical status: the use and interpretation
of anthropometry. Report of a WHO Expert Committee. World Health
height (0.25; 95% CI: 0.01, 0.49) and providing complementary Organ Tech Rep Ser 1995;854.
food, with or without education and counseling, can improve 18. Sterne JAC, Egger M, Smith GD. Investigating and dealing with
height in food-insecure populations (0.41; 95% CI: 0.05, 0.76) (3). publication and other biases in meta-analysis. In: Egger M, Davey
What we do not know is whether education and counseling alone Smith G, Altman DG, eds. Systematic reviews in health care. London,
can be effective in food-insecure populations. Thus, while we United Kingdom: BMJ Books, 2001:189–208.
19. Sterne JAC, Egger M. Regression methods to detect publication and
move forward to correct micronutrient deficiencies to improve other bias in meta-analysis. In: Rothstein HR, Sutton AJ, Borenstein M,
child survival and health, we must, at the same time, focus on eds. Publication bias in meta-analysis: prevention, assessment and ad-
interventions to improve complementary feeding. justment. New York, NY: John Wiley & Sons, Ltd, 2005:99–110.
20. Fahmida U, Rumawas JS, Utomo B, Patmonodewo S, Schultink W.
The authors’ responsibilities were as follows—UR: contributed to the study Zinc-iron, but not zinc-alone supplementation, increased linear growth
design, including the development of the analytical approach, reviewing the of stunted infants with low haemoglobin. Asia Pac J Clin Nutr 2007;16:
studies, interpretation of findings, and writing the article; PN: conducted the 301–9.
literature searches to identify the studies, extracted data from identified studies, 21. Alarcon K, Kolsteren PW, Prada AM, et al. Effects of separate delivery