Meta-Analysis Approach on Iron Fortified Foods and

its Effect on Hemoglobin Concentration in Pregnant

Women
Project report submitted for partial fulfillment of the requirement of

POST GRADUATE DIPLOMA IN

STATISTICAL METHODS AND ANALYTICS

Submitted

by

Anindita Mazumder
Roll No: DST-16/17-002
May, 2017

INDIAN STATISTICAL INSTITUTE

NORTH-EAST CENTRE, TEZPUR, ASSAM 7840128

I
 Date: 22-05-2017

CERTIFICATE

This is to certify Ms. Anindita Mazumdar has done the project under my

supervision and guidance (from Jan, 2017 to May, 2017). This is an original project report

based on work carried out by him / her in partial fulfillment and requirement of the Post

Graduate Diploma in Statistical Methods and Analytics programme of the Indian Statistical

Institute, North-East Centre, Tezpur, Assam.

(Dr Athe Ramesh)

II
 ACKNOWLEDGEMENT

I would like to express my special thanks of gratitude to my project supervisor

Dr. Ramesh Athe, Faculty, ISINE Centre Tezpur. Without his help and encouragement,

the present work would not have been possible.

I am also grateful to Prof .Nityananda Sarkar, Centre Head, ISINE Centre,

Tezpur, Dr. Ch Holendro Singh, Students In charge and other faculty members of the

Centre for their support and suggestions.

Finally, I thank my parents and my friends for their support and for being my

constant source of inspiration.

(Anindita Mazumder)

III
 CONTENTS

S.No. TITLES Page No.

1 INTRODUCTION 1-8

2 METHODOLOGY 9-20

2.1 Literature Search

2.2 Selection criteria

2.3 Objective of the study

2.2 Data Extraction and Quality Assessment

2.3 Statistical Analysis

3 RESULTS & CONCLUSIONS 21-27

3.1 Results

3.2 Algorithm For Meta-Analysis

4 REFERENCES 28-30

IV
 List Tables
S.No. Page No.

1 Prevalence of the three major micronutrient deficiencies in South Asia 5

2 Indicators for assessing iron status at the population level 8

3 Checklist of items to include when reporting a systematic review 17-19

4 Summary of trials assessing the effect of iron fortified foods on 23

measurements in pregnant women.

List Figures
S.No. Page No.

1 Obstacles faced in reaching a favourable outcome 2

2 Venn diagram depicting the relationship between systemic review and 4

meta-analysis

3 Flow of information through the different phases of a systematic 20

review

4 Flow diagram for inclusion in the present meta-analysis of randomized 22

controlled trials

5 Forest Plot: Effect of iron fortification on mean difference in Hb 24

concentration in comparison with no intervention on placebo control in
pregnant women.

6 Funnel plot of all individual studies in the Meta-analysis. 25

V
INTRODUTION:

Nutrition Science aim to create new knowledge, but scientist rarely sit back to reflect
on what nutrition research has achieved in recent decades. Meta-analysis, a statistical method
of combining data from multiple sources, has been used for many decades in the nutritional
sciences, social sciences, environmental sciences, and agriculture. In early 1990’s, meta-
analysis also started to gain traction in medicine as a valid and useful way to combine the
results of randomized controlled studies. Researchers trying to summarize the constantly
growing body of research in the social, nutritional, behavioral, and health sciences are
increasingly using this technique. Meta-Analysis provides an entire set of statistical methods
for aggregating and comparing the results from several relevant studies, allowing researchers
to determine whether a treatment is actually effective overall and whether the effectiveness of
a treatment depends on certain study and/or subjects characteristics (moderators). Potentially
relevant studies to determine the most promising directions for future researcher.

One requirement is to prepare a validated and reliable checklist that can assess the
quality of meta-analysis in terms of reporting, methodology, science and most especially, the
actual statistical analysis. Literature review reveals that existing checklists mainly focus on
other aspects of quality with little or no attention to the quality of statistical methodology.
Consequently, this thesis attempts to cover this gap.

Meta-analysis is an essential tool that facilitates clinicians, medical experts and

decision makers to cope with the information overload in the public and healthcare sectors.
The publication of meta-analyses is increasing rapidly every day with less or more of
methodological rigor. Clinicians and medical experts depend wholeheartedly on the results
and conclusions obtained from analyzing meta-analysis in order to assess the clinical
effectiveness of healthcare intervention on a daily basis. Meta-analysis provides a specific
estimate of a relationship which may also indicate if there is any need for further research.

1
 The pressures of moral–ethical obligations, legal liability and health economic
rationing have heralded the advent of evidence-based healthcare in the last few decades. To
ensure the best possible outcomes for patients, clinicians are increasingly required to
implement best practices and continuous quality improvement processes within the clinical
environment. This inextricably involves the application of the best available knowledge,
usually in the form of scientific research, to guide clinical decision making. Hence, the use of
clinical research is no longer an option but a necessity. However, problems remain for a
practicing clinician as to what constitutes ‗best available knowledge ‘and in particular which
type of research should be used (Fig.1)

Fig. 1. Obstacles faced in reaching a favourable outcome (Source 1)

2
 Meta-analysis first came into the scientific community‘s spotlight with Smith and
Glass‘s (2) review of the effects of psychotherapy. Smith and Glass reported the results of a
massive quantitative summary of 375 studies of psychotherapy effectiveness. The studies
were conducted over many years, over different geographic areas with diverse clients, using
different types of therapies and outcome measures. Over all this variability, clients receiving
psychotherapy had better outcomes than control clients who received no treatment. Therapy
outcomes were similar whether the therapies were behavioural or non-behavioural, over
varying levels of therapist experience, and over both brief and long-term therapy. With a few
exceptions, the effects of therapy generalized over great heterogeneity in many moderators
across these studies.
Meta-analysis practice has a long history. In the early 18th century, the English
mathematician Roger Cotes computed weighted averages of measurements made by different
astronomers. In 1904, Sir Karl Pearson used quantitative methods to average results from six
studies of the effects of a newly developed inoculation against typhoid (3-4). However,
widespread adoption of such methods did not occur until Glass (2) coined the term meta-
analysis to describe quantitative techniques for cumulating results over studies.
Glass (2) distinguished between primary, secondary, and meta-analysis, with these
three phrases intended to differentiate between three different types of statistical analyses.
According to Glass (1976), primary analysis is ―the original analysis of data in a research
study‖. Secondary analysis is ―the re-analysis of data for the purposes of answering the
original research question with better statistical techniques, or answering new questions with
old data‖. Meta-analysis is ―the statistical analysis of a large collection of analysis results
from individual studies for the purpose of integrating the findings‖.
In a systematic review, two types of synthesis can be performed: a qualitative
synthesis where primary studies are summarized like in a narrative review and a quantitative
synthesis where primary studies are statistically combined. This quantitative synthetic
component is termed a meta-analysis: the statistical quantitative integration of individual
study findings to get an overall summary result (2). A common misunderstanding is that a
meta-analysis is exactly identical to a systematic review and can be used interchangeably as
synonyms. In truth, a meta-analysis is actually a subset component of a systematic review as
illustrated in Fig. 2.

3
Fig. 2 Venn diagram depicting the relationship between systemic review and meta-analysis

Food fortification is usually regarded as the deliberate addition of one or more

micronutrients to a particular food, so as to increase the intake of these micro-nutrient(s) in
order to correct or prevent a demonstrated deficiency and provide a health benefit. The extent
to which a national or regional food supply is fortified varies considerably. The concentration
of just one micronutrient might be increased in a single foodstuff (e.g. the iodization of salt),
or, at the other end of the scale, there might be a whole range of food–micronutrient
combinations. The public health impact of food fortification depends on a number of
parameters, but predominantly the level of fortification, the bioavailability of the fortificants,
and the amount of fortified food consumed.

Food fortification and supplementation with nutrients has been carried out for
centuries, often before the scientific rationale became available. There are a number of
examples of food fortification from different parts of the globe. In Central Europe during the
middle ages, mothers were known to push iron nails into apples, leave them there for a while
and then feed the apples to listless or ailing daughters. There are reports as early as 1824 of
the indigenous Indian population of Columbia, South America, and treating goiter with a
specific source of salt, which was later found to have high iodine content.
4
Table 1. Prevalence of the three major micronutrient deficiencies in South Asia (Source 5)

5
 Deficiencies of micronutrients are a major public health problem in developing
countries. Fe-deficiency anaemia is a worldwide public health problem; global prevalence is
estimated at 24.8% and the highest prevalence occurs in sub-Saharan Africa and south central
Asia (6). In the case of breast-fed children aged 6 months, complementary food is expected to
meet the requirements for almost all micronutrients (7). Although modest information exists
on the nutritional status of schoolchildren, studies from several developing countries have
demonstrated a high prevalence of micronutrient deficiencies in this age group (8–10).
Under-nutrition in general and Fe deficiency in particular among schoolchildren can lead to
anaemia and negatively affect growth (11), motor and cognitive development (12) and
immune function (13). All of these can adversely affect academic performance (14).
However, on account of suboptimal feeding practices in terms of quantity and quality,
children often tend to receive Fe much below their daily requirement. Therefore an
alternative method of providing Fe for this vulnerable segment of the population is required.
Among the various strategies, Fe fortification of foods has been suggested as a cost-effective,
long-term, population-based strategy with better compliance to improve Fe status and to
prevent Fe deficiency worldwide (15, 16).

In India, limited studies have evaluated the effect of Fe-fortified iodized common salt
on Hb levels. The results suggested progressive increase in net Hb concentration in the range
of 10–30 g/l after 1 year in rural communities and 5g/l in urban schoolchildren (17). Thus,
there is evidence that children might benefit more from Fe fortification like their counterparts
in developed countries.
In recent years robust study design and meta-analysis have advanced significantly and are
readily available (18). In the present study, we aimed to systematically review the current
literature and to perform a meta-analysis to estimate the effect of Fe-fortified foods on Hb
concentration in children. Since we expected heterogeneity among studies, we also explored
whether factors such as age, duration of the study and levels of fortification could predict the
effects on Hb concentration in pregnancy.

6
A fortified food should:

centrally processed with minimal stratification of the fortificant;

-economic status; and

Selection of an appropriate vehicle is a critical step in successful fortification. In

many cases identification of suitable vehicles is made difficult by the absence of reliable

information on dietary habits of the target population.

The need for fortification

Mineral and vitamin deficiencies are a serious public health problem in many

developing countries and often occur in industrialized countries. Common mineral and trace

element deficiencies include iron, zinc, selenium, iodine and calcium. The most important

vitamin deficiencies today are probably vitamin A, vitamin D and folic acid, although niacin

deficiency in maize eating populations, thiamine deficiency in rice-eating populations, and

scurvy caused by lack of fresh fruits and vegetables, have claimed many lives in the past and

still appear occasionally in refugee camps. Various strategies have been developed to prevent

micronutrient deficiencies.

7
Risk factors for deficiency

The main risk factors for iron deficiency are:

vitamin C enhances the absorption of iron from the diet);

iets high in phytate (including legumes and cereals) or

phenolic compounds (present in coffee, tea, sorghum and millet);

struation, or parasite infections such as hookworm,

ascaris and schistosomiasis.

Table 2. Indicators for assessing iron status at the population level (Source 19, 20)

As mentioned above, acute or chronic infections, including malaria, can also lower
hemoglobin concentrations. The presence of other micronutrient deficiencies, especially of
vitamins A and B 12, folate and riboflavin, also increases the risk of anaemia (21).

8
MEHDOLOGY:

2.1. Literature Search:

This is by far the most important, and time-consuming part of the meta-analysis. We

identified existing reviews to support our practice by searching the resources, once relevant

study have been identified, the quality of their methods should be appraised, their evidence

should be examined and their findings should be assessed for application in practice. A

structured search strategy was used. The Cochrane Library (www.thecochranelibrary.com)

has several databases of published and ongoing systematic reviews (22). CDSR contains the

full-text of regularly updated systematic reviews of healthcare interventions carried out by the

Cochrane Collaboration, plus protocol for reviews currently in preparation (23). MEDLINE

(available freely via PubMed at www.ncbi.nlm.nih.gov/PubMed) Bibliographic records (with

and without abstracts) of biomedical literature from 1966 onwards. EMBASE

(www.embase.com) Records of biomedical literature from 1974 onwards. Science Citation

Index

Relevant studies found through electronic or manual searches can be used to identify

further relevant citations by electronically locating other citations on the same topic through

citation search on Science Citation Index. Initially, the selection criteria should be applied

liberally to the citation lists generated from searching relevant literature sources. Citations

often contain only limited information, so any titles (and abstracts) which seem potentially

relevant should provisionally be included for consideration on the basis of the full text

articles.

The steps in this process were conducted according to the PRISMA (Preferred

Reporting Items for Systematic reviews and Meta-Analysis) guidelines for meta-analysis

(24). We searched both PubMed and the Cochrane Library databases from 1990 up to

December 2016, and also reviews and the reference lists of the articles, using the keyword

‘food fortification’ paired with ‘iron’ or ‘haemoglobin’ or ‘dual fortification’ or ‘triple

fortification’ or ‘multiple micronutrient fortification’ and ‘fortification trial’.

9
2.2. Selection Criteria

The search was regardless of language and publication status. These studies included

multiple intervention groups with other micronutrients that were administered

simultaneously; the outcome measure was the effect of Fe fortification on Hb concentration

only. The studies were limited to publications where RCT evaluated Fe fortification among

pregnancy for its effect on Hb concentration.

2.3. Objectives of the Study:

Has the intake of iron fortified foods on hemoglobin (Hb) concentration in during

pregnancy of interest increased to expected levels as a result of the sub-clinical fortification

programme?

2.4. Data Extraction and Quality Assessment

It is one of the most important steps in conducting a meta-analysis, and the methods

of data abstraction that were used by the authors should be described in detail. The reader of a

meta-analysis should be provided with enough information to determine whether the studies

that were included appropriate for a combined analysis. Two or more authors of a meta-

analysis should abstract information from studies independently. The results from the data

abstraction are compared only after completing the review of the articles. The article should

state any discrepancies between authors and how the discrepancies were resolved.

Results should be collected only from separate sets of patients, and the authors should

be careful to avoid studies that published the same subjects or overlapping groups of subjects

that appeared in different studies under duplicate publications. Raw numbers, in addition to

risk ratios, should be recorded. Results from intention-to-treat analyses should be reported,

when possible. The gold standard of data abstraction in a meta-analysis is to include patient

level data from the studies combined in the meta-analysis, which usually requires contacting

the authors of the original studies (25). Obtaining patient level data may reveal differences

among the trials that other-wise would not have been detected.

10
 The title and abstracts of the studies identified in the web search were read and
irrelevant studies were excluded. Full texts of the remaining studies were retrieved. To avoid
publication bias, only peer-reviewed published studies were included. The extraction of data
consisted of obtaining sample size, age, duration of intervention, levels of fortification and
mean change and standard deviation of Hb concentration in the intervention and control
groups. The search, data extraction and quality assessment were completed independently by
two content experts according to the inclusion criteria and confirmed by using recommended
criteria for RCT (26). Concealment of allocation was classified as ‘adequate’, ‘unclear’,
‘inadequate’ or ‘not used’, based on randomization, blinding and reporting of withdrawals.
Blinding was classified as ‘double blinding’, ‘single blinding’, ‘no blinding’ or ‘unclear’. In
designs employing two or more different intervention groups (different levels of fortification
or administration regimens) and a single control group, the sample size of the control group
was equally allotted to the number of intervention groups while retaining the same mean
value for the change and its standard deviation. In reporting such designs, each intervention
subgroup was analysed separately. Thus, some studies contributed more than one intervention
component with a single control group for the statistical analysis and resulted in a greater
number of trials than the number of studies included.
2.5. Statistical Analysis

There are many different types of effect sizes but those discussed here provide
information about the magnitude and direction of the difference between two groups or the
relationship between two variables. An effect size can be a difference between means, a
percentage, or a correlation. Researchers usually want to show there is a difference between
the groups they are studying, or that some variables they are investigating are correlated.
Effect sizes provide this information by assessing how much difference there is between
groups or how strong the relationship is between variables. In other words, effect sizes assess
the magnitude or strength of the findings that occur in research studies.
While estimating the effect sizes, two commonly used statistical methods for
combining data fixed effects and random effects models (27). One of the goals of these
methods is to provide a summary statistic of an intervention‘s effect or exposure, as well as a
confidence interval. The fixed effects model examines whether the treatment produced a

11
benefit in the studies that were conducted. In contrast, the random effects model assumes that
the studies included in the meta-analysis are a random sample of a hypothetical population of
studies. The summary statistic is typically reported as a risk ratio, but it can be reported as a
rate difference, or percentage.

When the studies report means and standard deviations, the preferred effect size is
usually the raw mean difference or weighted mean difference (WMD) or Mean
difference, the standardized mean difference (SMD) or standard weighted mean
difference (SWMD). Once an effect size was estimated for each trial, the overall effects of
these results were assessed by Q statistic. Another strategy for quantifying the heterogeneity
in a meta-analysis consists of estimating variance (2) in between-studies. The I2 quantifies
the extent of heterogeneity from a collection of effect sizes, which is interpreted as
approximately the percentage of total variation in study estimates due to heterogeneity rather
than sampling error.

In a meta-analysis the results of every study is quantified by means of an effect size index
(e.g., weighted mean difference) that can be applied to all studies enabling us to give the
study results in the same metric. A first step in meta-analysis is determination of the effect to
be analyzed over studies. The simplest measure is Effect Size (Ei). Once an effect size is
estimated for each study, the next step is to summarize these results from the individual
studies to gain an understanding of the overall effect.

A common measure is the standardized mean difference. There are many variations on this
effect measure, but most often used with the best statistical properties is known as effect size
2
(Ei). Given means (𝑋̅𝑒, 𝑋̅𝑐), standard deviations (𝑆 𝑒 2 , 𝑆 𝑐 ), and sample sizes (𝑁 𝑒 , 𝑁 𝑐 ) for an
experiment and control group, respectively, the effect size calculated as

(𝑋̅ 𝑒 -𝑋̅𝑐)
Ei= --------------------------------------- (1)
𝑆

√(Nc -1)Sc 2 +(Ne -1)Se 2

S= --------------------------- (2)
√Nc +Ne -2

12
Once an effect size is estimated for each study, the next is summarizing these results from the
individual studies to gain an understanding of the overall effect. In the following, Ei indicates
the effect size for the ith study, regardless of which effect measure is used, and vi indicates its
variance. In Meta-analysis, studies are weighted differentially depending on their accuracy.
The simplest and least controversial method of estimating their accuracy is variance; studies
with low variance are given high weight and those with high variance are given low weight.
Specially, the weight for the ith study is the inverse of its variance,

𝑊𝑖 = 1⁄𝑉𝑖 ---------------------------------- (3)

The grand measure of effect is simply a weighted average of the individual effect sizes,

∑ 𝑤𝑖𝐸𝑖
𝐸̅ = ∑ 𝑤𝑖 ----------------------------------- (4)

This is average effect for all studies. Confidence intervals can be created to test whether the
mean differs from the hypothesis of the no effect using standard distribution or bootstrapping.
To test whether the sample effect sizes are themselves homogeneous (from a single
population), one use Q statistics, and a form of weighted sum-of-squares. The total
heterogeneity in a Meta-analysis study is

𝑄𝑇 = ∑ 𝑤𝑖 (𝐸𝑖 − 𝐸̅ )2 --------------------------- (5)

QT can be tested against a χ2 distribution with n-1 degrees of freedom. This tests whether the
effect sizes for all of the studies are equal; a significant result indicates that the variance is
greater than expected due to sampling error. Simple calculation of 𝐸̅ assumes that all of the
studies come from a single homogeneous sample. Often variation in the effect size may be
explained by additional variables. The simplest way to evaluate this allows the framework of
meta-regression analysis.
There are usually some differences between studies in the key characteristics of their
populations, interventions and outcomes (clinical heterogeneity), and their study designs and
quality (methodological heterogeneity). These are discovered during tabulation of
information from the studies. These variations in study characteristics and quality are likely
to have some influence on the observed effects. Investigation of heterogeneity is about this
variation of effects between studies and its reasons.

13
 Meta-regression is an extension of subgroup analysis. Using meta-regression, a better
understanding of the causes for heterogeneity between study groups can be found. However,
meta-regression has a number of significant limitations. Firstly, the initial decision to perform
a meta-regression on a certain variable is entirely observer dependent and hence is also prone
to selection bias. Meta-regression requires many studies (>10) and there is a risk of obtaining
a spurious correlation for a variable especially when many characteristics (confounders) are
studied (23).
A fixed effects meta-analysis was applied to obtain pooled effects size with 95% CI or
else a random effects meta-analysis was performed (heterogeneous, 2 > 0). If the
heterogeneity existed (I2>50%), a meta-regression approach was used to test the study
heterogeneity by relating study characteristics (28, 29). The confounders were identified and
a covariate meta-analysis was performed to estimate the net pooled effect size, after removing
the effect of covariates (confounders) (30). A commonly used test for publications bias is
Egger‘s regression test. Publication bias is one issue that researchers face when
conducting a literature review, designing a new study, or conducting a meta-analysis
(31).
Graphical display
Forest plot
This is a commonly used, easy to understand, graphical display of individual effects
observed in studies included in a systematic review along with the summary effect of meta-
analysis. The results of individual trials are ordered and presented one below the other with
their respective 95% confidence intervals. The pooled estimate is presented at the bottom
with the 95% confidence interval. A vertical line is displayed to indicate no-effect and to
differentiate between the studies that favour the intervention group or the control group and
describes the Q-test statistic, Ƭ2 (variation among the included studies), d.f , I2 (variation in
percentage among the included studies), Z (test statistics), and p-value. For each study, a box
representing the point estimate of effect lies in the middle of a horizontal line which
represents the confidence interval of the effect.

14
 To have a more appropriate tool for the assessment of heterogeneity by using Q
statistic value, I2 = (Q - (k-1))/Q has been introduced. I2 is interpreted as the proportion of the
existing variability due to heterogeneity between studies. I2 ranges between 0% and 100%;
0% indicates no observed heterogeneity, and larger values show increasing heterogeneity.
Values of 25%, 50% and 75% may be taken to represent low, moderate and high levels of
heterogeneity. With I2 dependence on the number of included studies is avoided. However, I2
still hinges on the precision of the studies, or in other words, on the size of the studies. This
limits the use of I2 when comparing values across meta-analyses with different study sizes
(29).

Funnel plot
Funnel Plot, or, funnel graph, is the frequently used method for detecting the
publication bias. The basic idea is that if the point estimates from individual studies are
plotted against the inverse of the variances, or another surrogate for sample size, the points
visualized together should produce a funnel shape, so they are scattered around the true value
of the point estimate with the scattering narrowing as the standard errors decrease (32). That
is, in such a plot, the effect size of studies is plotted against study sample size. If there is no
publication bias, the plot would resemble an inverted funnel with a wide dispersion of results
among studies of small size and a narrower range of study results for large studies. If the plot
shows an asymmetrical and skewed shape, publication bias may present. This usually takes
the form of a gap in the wide part of the funnel, which indicates the absence of small studies
showing no benefit or harm.
In fact, the funnel plot is a graphical test for any type of bias that is associated with sample
size. The publication bias and sampling bias are more likely to affect smaller studies than
large trials and may thus lead to funnel plot asymmetry. Another source of asymmetry arises
from differences in the methodological quality. Smaller studies are, on average, conducted
and analysed with less methodological rigor than larger studies, and trials of lower quality
tend to show larger effects. Other factor, such as heterogeneity in treatment effect between
low and high risk groups can also lead to asymmetry in the funnel plot.
The major advantage of funnel plot is that it is easy to be performed which only
requires published data. But the method is practically limited to meta-analysis with large
enough numbers of studies to allow one to visualize a funnel shape to the data. The symmetry
of funnel plot is defined informally. So, if the number of studies included in a meta-analysis
is small, it is difficult to detect the symmetry of funnel plot through visual examination.

15
Egger proposed a linear regression model to measure funnel plot asymmetry. It is a formal
test for asymmetry in funnel plot (31).

Publication bias

Meta-analyses are subject to publication bias because studies with negative results are

less likely to be published and, therefore, results from meta-analyses may overstate a

treatment effect. One strategy to minimize publication bias is to contact well-known

investigators in the field of interest to discover whether they have conducted a negative study

that remains unpublished. With the development of the National Library of Medicine‘s

clinical trial registry (www.clinicaltrials.gov), researchers conducting meta-analyses have

better opportunities to identify trials that are unpublished. Even though there is no consistent

relationship between the publication of a study with study design, methodological quality,

study size or number of study centres, the publication of a trial is more likely when it shows

either a statistically significant large effect in the outcome for a new treatment (a positive

trial) or when compared to existing treatments (a non-inferiority trial). The publication of a

trial is less likely when there are non-significant findings, results with small effect sizes or

negative findings (a negative trial). Usually Egger‘s regression test and funnel plots are

applied to detect such a publication bias (31).

Reporting the Results

Systematic reviews and meta-analyses are essential to summarize evidence relating to

efficacy and safety of health care interventions accurately and reliably. A systematic review

is a review of a clearly formulated question that uses systematic and explicit methods to

identify, select, and critically appraise relevant research, and to collect and analyze data from

the studies that are included in the review. Statistical methods (meta-analysis) may or may

not be used to analyze and summarize the results of the included studies. Meta-analysis refers

to the use of statistical techniques in a systematic review to integrate the results of included

studies. The steps in this process we conducted according to the PRISMA guidelines for

meta-analysis.
16
Table 3: Checklist of items to include when reporting a systematic review

Section/Topic # Checklist item

TITLE
Identify the report as a systematic review, meta-
Title 01 analysis, or both
ABSTRACT
Provide a structured summary including, as
applicable: background; objectives; data sources;
study eligibility criteria, participants, and
interventions; study appraisal and synthesis methods;
results; limitations; conclusions and implications of
Structured summary 02 key findings; systematic review registration number.
INTRODUCTION
Describe the rationale for the review in the context of
Rationale 03 what is already known.
Provide an explicit statement of questions being
addressed with reference to participants,
interventions, comparisons, outcomes, and study
objectives 04 design (PICOS).
METHODS
Indicate if a review protocol exists, if and where it
can be accessed (e.g., Web address), and, if available,
provide registration information including
Protocol and registration 05 registration number.
Specify study characteristics (e.g., PICOS, length of
follow-up) and report characteristics (e.g., years
considered, language, publication status) used as
Eligibility Criteria 06 criteria for eligibility, giving rationale.
Describe all information sources (e.g., databases with
dates of coverage, contact with study authors to
identify additional studies) in the search and date last
information sources 07 searched.
search 08 Present full electronic search strategy for at least one
17

study selection
Data collection process
data items
Risk of bias in individual
studies
Summary measures
Synthesis of results
Risk of bias across studies
Additional analyses
RESULTS
study selection
database, including any limits used, such that it could
be repeated.
State the process for selecting studies (i.e., screening,
eligibility, included in systematic review, and, if
09 applicable, included in the meta-analysis).
Describe method of data extraction from reports (e.g.,
piloted forms, independently, in duplicate) and any
processes for obtaining and confirming data from
10 investigators.
List and define all variables for which data were
sought (e.g., PICOS, funding sources) and any
11 assumptions and simplifications made.
Describe methods used for assessing risk of bias of
individual studies (including specification of whether
this was done at the study or outcome level), and how
12 this information is to be used in any data synthesis.
State the principal summary measures (e.g., risk
13 ratio, difference in means).
Describe the methods of handling data and
combining results of studies, if done, including
measures of consistency (e.g., I2) for each meta-
14 analysis.
Specify any assessment of risk of bias that may affect
the cumulative evidence (e.g., publication bias,
15 selective reporting within studies).
Describe methods of additional analyses (e.g.,
sensitivity or subgroup analyses, meta-regression), if
16 done, indicating which were pre-specified.
Give numbers of studies screened, assessed for
eligibility, and included in the review, with reasons
for exclusions at each stage, ideally with a flow
17 diagram.
18

Study characteristics
Risk of bias within studies
Results of individual studies
Synthesis of results
Risk of bias across studies
Additional analysis
DISCUSSION
Summary of evidence
limitations
conclusions
FUNDING
Funding
For each study, present characteristics for which data
were extracted (e.g., study size, PICOS, follow-up
18 period) and provide the citations.
Present data on risk of bias of each study and, if
available, any outcome-level assessment (see Item
19 12).
For all outcomes considered (benefits or harms),
present, for each study: (a) simple summary data for
each intervention group and (b) effect estimates and
20 confidence intervals, ideally with a forest plot.
Present results of each meta-analysis done, including
21 confidence intervals and measures of consistency.
Present results of any assessment of risk of bias
22 across studies (see Item 15).
Give results of additional analyses, if done (e.g.,
sensitivity or subgroup analyses, meta-regression [see
23 Item 16]).
Summarize the main findings including the strength
of evidence for each main outcome; consider their
relevance to key groups (e.g., health care providers,
24 users, and policy makers).
Discuss limitations at study and outcome level (e.g.,
risk of bias), and at review level (e.g., incomplete
25 retrieval of identified research, reporting bias).
Provide a general interpretation of the results in the
context of other evidence, and implications for future
26 research.
Describe sources of funding for the systematic
review and other support (e.g., supply of data); role
27 of funders for the systematic review.
19
Figure 3. Flow of information through the different phases of a systematic review.

20
RESULTS & CONCLUSIONS

3.1 Results:
Search results

A total of 963 articles were identified, of which 885 were excluded because they were
not RCT or their interventions were not relevant to the purpose of the current analysis. 78
potentially relevant articles were selected for full text evaluation, out of which 8 relevant
articles were submitted to meta-analysis after employing the inclusion and exclusion criteria
(Fig. 4).

Study characteristics and data quality

Characteristics of the eight studies included in the analysis (33-40) are shown in Table
4. All of these were RCT. Of the eight RCT, all had similar Hb concentrations in intervention
and control groups at baseline. A total of 5845 pregnant women with levels of Fe fortification
such that daily Fe intake through fortified food ranged between 10 and 60 mg per pregnant
women, with intervention duration ranging between two and six months, were studied. These
studies have been carried out over the past 16 years. All 22 studies evaluated the effect of
various levels of fortification on Hb concentration. Out of eight, six studies of multiple
interventions with multiple micronutrients were included and only those groups that received
Fe were considered (35, 38-40). Of these 6 included studies, each one had more than one
trial. The trials were either based on different levels of intervention of food fortification
conducted on two occasions (before and after) or they were compared with placebo. In each
trial the number of participants, mean and standard deviation of Hb concentration were
estimated for conducting meta-analysis.

21
 Titles and abstracts retrieved from electronic and bibliographies: (n=963)

Excluded (n=885);
Titles and abstracts were very unlikely to be
relevant

Titles and abstracts that appeared potentially relevant, ordered as full

text papers: (n=78)

Excluded (n=17):
Commentaries (n=1); Reports (n=1);
Reviews (n=15)

Full papers assessed for inclusion (n=61)

Papers excluded with reasons (n=53):

Fortification on children & adults (n=10),
Non RCT (n=18),
Iron Supplementation (n=19), and
Fortification without iron (n=6)

Full papers included in Review (n=8)

Included in meta-analysis (n=8)

Fig. 4 Flow diagram for inclusionNot

in randomized
the present controlled
meta-analysis
trialsof
or randomized
irrelevant controlled trials
assessing the effect of iron-fortified foods on mean Hb concentration in pregnant women.

Not randomized controlled trials or irrelevant

22
Effects of Fe fortification on Hb concentration
The meta-analysis results indicated that the mean change in Hb concentration was
significantly in the Fe-fortified group than in the control group (n 5845; WMD1.59 g/dl, 95%
CI 2.60, 5.78 g/dl; P; 0.46), as depicted on the forest plot (Fig. 5). There was significant
heterogeneity for the mean Hb concentration reported among the included trials. All
statistical tests of heterogeneity – such as the Q statistic (χ2 1185.03, df; 21), which was more
than df; τ2 greater than zero (τ2 =97.75); and I2 greater than 50% (I2 =98 %) – were higher
than the expected value, indicating heterogeneity among the studies. Increased duration of
receiving fortified foods might have resulted in higher levels of Hb.

Table 4. Summary of trials assessing the effect of iron fortified foods on measurements

in pregnant women

Study Averag Dose of

Food Study
Study Year duration e age fortification
Vehicle Design
(months) (years) (mg/L)

Awadallah SM et al.33 2003 6 31 50 Milk RCT

Makola D et al.34 2003 2 25 10 Beverage RCT

Latham MC et al.35 2003 2 32 54 sachet RCT

Hoa PT et al.36 2005 4 26 60 Milk RCT

Zhou SJ et al.37 2007 2 27 60 Meals RCT

Choudhury N et al.38 2012 2 27 60 M. powder RCT

Silva CLD et al.39 2012 6 27 37 Wheat RCT

Araujo CRMA et al.40 2013 3 32 11 Flour RCT

RCT, Randomized controlled trial; M.powder : Micronutrient powder

23
Fig. 5: Effect of iron fortification on mean difference in Hb concentration in comparison with
no intervention on placebo control in pregnant women. Random effects Meta analysis of
WMD on Hb concentration with iron fortified food intervention compared with control
group. The sizes of data markers indicate the weight of each study in the analysis. The
horizontal lines represents 95% CI. Blob indicates best estimate and diamond indicates the
summary estimator of the WMD.

Publication bias
The funnel plot (Fig. 6) was asymmetrical, indicating the probable presence of
publication bias.

24
Fig. 6: Funnel plot of all individual studies in the Meta analysis. Studies that evaluated the
effect of iron fortification on Hb concentration in pregnant women were plotted with their
WMD on the X-axis and the corresponding standard error of the WMD along the Y-axis.

25
Conclusion

The Fe fortification intervention varied across trials; hence, results should be

interpreted accordingly. The present meta-analysis of eight studies consisting of 22 trials

found that Fe fortification was significantly associated with increased Hb concentration in

intervention pregnant women compared with the controls. However, there was heterogeneity

in the results across the trials.

The present study adopted sequential statistical methods to verify that implementation of

fortified foods with Fe improves Hb concentration in the pregnant women beneficiaries.

There is emerging evidence to suggest that reports of RCT from certain countries mostly have

statistically positive results on micronutrients (41). This may be due to the fact that in those

countries, anaemia is largely due to Fe deficiency (single nutrient) rather than the

multifactorial aetiology reported from developing countries.

However, the present study suggests the possibility of a positive effect of Fe fortification on

Hb in pregnant women. This phenomenon needs to be investigated further. There is a need to

conduct more trials in developing countries so as to investigate whether Fe fortification

improves Hb concentration and also to assess other confounders, such as intakes of protein,

energy and folic acid, which could contribute in improving Hb levels.

This present study suggests that consumption of Fe-fortified foods significantly

increases the Hb concentration in pregnant women. Further research efforts should

concentrate on higher quality and more rigorous randomized trials with longer follow-up to

resolve the uncertainty regarding the safety and clinical effectiveness.

26
3.2 ALGORITHM FOR META-ANALYSIS:

This algorithm helps to understand the statistical methods and principles of

systematic review and meta-analyses. In the dissertation that follows there is a step-by-step
explanation of the review process. This algorithm will be a useful companion in framework
development as well as throughout the five steps in the review process given below:

1. First, the problems to be addressed have to be specified in the form of well-structured

questions (step-1). This is a key step, as all other aspects of the review follow directly
from the questions.

2. Second, through literature searches have to be conducted to identify potentially

relevant studies which can shed light on the questions (step-2). This is an essential
feature that makes a review systematic.

3. Third, the quality of the selected studies is assessed (step-3).

4. Fourth, the evidence concerning study characteristic and results is summarized (step-
4). When feasible and appropriate, statistical methods (i.e. meta-analysis) help in
collating results, and

5. Finally, inferences and recommendations for practice are generated by interpreting

and exploring the clinical relevance of the findings (step-5).

Acknowledgement

I am gratefully acknowledge the support extended by the Indian Statistical

Institute, North East Centre (ISINE), Tezpur, India, for carrying out the study.
Ramesh Athe contributed to develop the study protocol, supervised the study and
guided dissertation preparation. It gives me immense pleasure to acknowledge the
support and suggestions from our head Prof. Nityananda Sarkar and Director of
Indian Statistical Institute Prof. Sanghamitra Bandyopadhyay. I am also thankful to
all faculty members for their insightful inputs.

27
REFERENCES

1. Athanasiou, Thanos, Darzi, Ara (2011). Evidence synthesis in Healthcare. A practical

handbook for clinicians. Chap 2:69, Springer.
2. Glass G. (1976). Primary, secondary, and meta-analysis of research. Educational
Researcher. 5I: 3-8.
3. Cooper H, Hedges LV. (Eds.). (1994).The handbook of research synthesis. New York
Russell Sage Foundation.
4. Dickerson K, Higgins K, Meinert CL (1990). Identification of meta-analyses: The
need for standard terminology. Control Clin Trials. 11: 52-66.
5. Gragnolati M, Shekar M, Gupta MD, Bredenkamp C, et al (2005). India‘s
undernourished children: a call for reform and action. The International Bank for
Reconstruction and Development / The World Bank, Washington, DC.
6. World Health Organization (2008) Worldwide Prevalence of Anaemia 1993–2005.
Geneva: WHO.
7. Brown K, Dewey K & Allen L (1998) Complementary Feeding of Young Children in
Developing Countries. A Review of Current Scientific Knowledge. Geneva: WHO.
8. Le HT, Brouwer ID, Verhoef H et al. (2007) Anemia and intestinal parasite infection
in school children in rural Vietnam. Asia Pac J Clin Nutr 16, 716–723.
9. Neumann CG, Bwibo NO, Murphy SP et al. (2003) Animal source foods improve
dietary quality, micronutrient status, growth and cognitive function in Kenyan school
children: background, study design and baseline findings. J Nutr 133, 11 Suppl. 2,
3941S–3949S.
10. Sivakumar B, Nair KM, Sreeramulu D et al. (2006) Effect of micronutrient
supplement on health and nutritional status of school children: biochemical status.
Nutrition 22, 1 Suppl., S15–S25.
11. Lawless JW, Latham MC, Stephenson LS et al. (1994) Iron supplementation improves
appetite and growth in anemic Kenyan primary school children. J Nutr 124, 645–654.
12. Black MM (2003) Micronutrient deficiencies and cognitive functioning. J Nutr 133,
11 Suppl. 2, S3927–S3931.
13. Thurnham DI (1997) Micronutrient and immune function: some recent developments.
J Clin Pathol 50, 887–891.
14. Popkin BM & Lim-Ybanez M (1982) Nutrition and school achievement. Soc Sci Med
16, 53–61.

28
15. Allen L, de Benoist B, Dary O et al. (2006) Guidelines on Food Fortification with
Micronutrients, 2nd ed. Geneva: WHO and FAO.
16. Horton S (2006) The economics of food fortification. J Nutr 136, 1068–1071.
17. Nair KM (2001) Alternate strategies for improving iron nutrition: lessons from recent
research. Br J Nutr 85, Suppl. 2, S187–S191.
18. Hedeker D & Gibbons RD (2006) Longitudinal Data Analysis. Hoboken, NJ: Wiley.
19. Benoist de B, Andersson M and Egli I, et al. (2004), eds. Iodine status worldwide.
Global Database on Iodine Deficiency. Geneva, World Health Organization.
20. Ottaway PB (2008). Food fortification and supplementation. Technological, safety
and regulatory aspects. CRC Press, New York.
21. Allen LH, Casterline-Sabel JE (2000). Prevalence and causes of nutritional anemias.
In: Ramakrishnan U, ed. Nutritional Anemias. Boca Raton, FL, CRC Press. pp: 17–
21.
22. Chalmers TC, Blum A, Buyse M, et al (1988). Data Analysis for Clinical Medicine:
The Quantitative Approach to Patient Care. Rome, Italy: International University
Press. pp: 75-84.
23. Higgins JP, Green S (2009). Cochrane Handbook for systematic reviews of
interventions. Version 5.0.0, John Wiley & Sons.
24. Liberati A, Altman DG, Tetzlaff J et al. (2009) The PRISMA statement for reporting
systematic reviews and meta-analyses of studies that evaluate health care
interventions: explanation and elaboration. BMJ 339, b2700.
25. Schulz KF, Chalmers I, Hayes RJ, Altman D (1995). Empirical evidence of bias.
Dimensions of methodological quality associated with estimates of treatment effects
in controlled trials. JAMA. 273:408–12.
26. Juni P, Altman DG & Egger M (2001) Assessing the quality of randomized controlled
trials. In Systematic Reviews in Health Care: Meta-analysis in Context, 2nd ed., pp.
87–108 [M Egger, G Davey Smith and DG Altman, editors]. London: BMJ Books.
27. Petitti DB (1994). Statistical Methods in Meta-analysis. Meta-analysis, Decision
Analysis, and Cost-effectiveness Analysis. New York, NY: Oxford University Press.
pp: 90-114.
28. DerSimonian R & Laird N (1986) Meta-analysis in clinical trials. Control Clin Trials
7, 177–188.
29. Borenstein M, Hedges LV, Julian PT et al. (2009) Introduction to Meta-Analysis.
Hoboken, NJ: John Wiley & Sons.

29
30. Athe R, Rao MV, Nair KM. Impact of iron-fortified foods on Hb concentration in
children (<10 years): a systematic review and meta-analysis of randomized controlled
trials. Public Health Nutr. 2014; 17:579-86.
31. Egger M, Davey Smith G, Schneider M et al. (1997) Bias in meta-analysis detected
by a simple, graphical test. BMJ 315, 629–634.
32. Light RJ, Pillemer DB (1984). Quantitative procedures. Summing Up: The Science of
Reviewing Research. Harvard University Press, Cambridge, MA.
33. Awadallah SM, Abu-Elteen KH, Elkarmi AZ et al. (2003). Maternal and Cord Blood
Serum Levels of Zinc, Copper, and Iron in Healthy Pregnant Jordanian Women.
34. Makola D, Deborah MA, Tatala SR et al. (2004). A Micronutrient-Fortified Beverage
Prevents Iron Deficiency, Reduces Anemia and Improves the Haemoglobin
Concentration of Pregnant Tanzanian Women. The Journal of Trace Elements in
Experimental Medicine 17:1–8.
35. Latham MC, Deborah MA, Makola D et al., (2003). Efficacy trials of a micronutrient
dietary supplement in schoolchildren and pregnant women in Tanzania. American
Society for Nutritional Sciences.
36. Hoa PT, Khan NC, Beusekom CV et al., (2005). Milk fortified with iron or iron
supplementation to improve nutritional status of pregnant women: An intervention
trial from rural Vietnam. Food and Nutrition Bulletin, vol. 26, no. 1. The United
Nations University.
37. Zhou SJ, Gibson RA, Crowther CA et al.,(2007). Should we lower the dose of iron
when treating anaemia in pregnancy? A randomized dose–response trial. European
Journal of Clinical Nutrition (2009) 63, 183–190.
38. Choudhury N, Aimone A, Hyder SMZ et al. (2012) Relative efficacy of micronutrient
powders versus iron–folic acid tablets in controlling anemia in women in the second
trimester of pregnancy. Food and Nutrition Bulletin, vol. 33, no. 2. The United
Nations University.
39. Silva CLD, Saunders C, Szarfarc SC et al., (2012). Anaemia in pregnant women
before and after the mandatory fortification of wheat and corn flours with iron. Public
Health Nutrition: 15(10), 1802–1809.
40. Araujo CRMA, Uchimura TT, Fujimori E et al., (2013). Haemoglobin levels and
prevalence of anemia in pregnant women assisted in primary health care services,
before and after fortification of flour. Rev Bras Epidemiol 2013; 16(2): 535-45.
41. May GS, Demets DL, Friedman LM, Furberg C, et al., (1981). The randomized
clinical trial: bias in analysis. Circulation. 64:669–73.

30