SPECIAL THEME

How to Conduct a Systematic Review and Meta-analysis

Krysten Marie R. Yayen, MD, DFM*; John Michael D. Deblois, MD, DFM** and
Ma. Rosario Bernardo-Lazaro, MD, MBAH, FPAFP***

Systematic reviews and meta-analysis combine results and analysis of data from different primary studies (e.g. cross-sectional
studies, case-control studies, cohort studies) conducted on similar or related research topics. They are secondary studies that
guide clinical decision-making, delivery of care and policy development. This article aims to discuss how to conduct a systematic
review and meta-analysis. The steps in conducting a systematic review and meta-analysis include: 1) Identify the purpose
including formulating the research question and validating the purpose of the literature scan, 2) Formulate the objectives,
3) Literature search including selection of studies based on population, intervention, comparison and outcome, 4) Retrieval of
full text articles, 5) Critical appraisal of articles, 6) Data extraction, 7) Data analysis and 8) Writing the final report. The Preferred
Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) is a useful guide in conducting and write systematic review
and meta-analysis. While ethics approval is not usually required for systematic review and meta-analysis, authors of such study
should still observe good practices including avoiding plagiarism, maintaining transparency and ensuring data accuracy.

Key words: Systematic review, meta-analysis

Introduction to produce a single estimate like the effect of treatment or risk factor.
Meta‑analysis reduces the quantity of data by summarizing data
from multiple resources, helps to make efficient use of existing data,
Systematic reviews and meta-analysis are research methods that
ensuring generalizability, helping to check consistency of relationships,
combine and analyze data from different studies conducted on similar
explaining data inconsistency, and quantifies the data and helps to
or related research topics. These are powerful tools that have been
improve the precision in estimating the risk by using explicit methods.4
actively performed in various fields as they overcome difficulties in
Although systematic reviews are published in academic forums, there
performing large-scale studies.1 Systematic reviews involve a detailed
are also organizations and databases specifically developed to promote
and comprehensive plan and search strategy aimed at reducing bias
and disseminate them. For example, the Cochrane Collaboration (www.
through the identification, appraisal, and synthesis of all relevant
cochrane.org) is a widely recognized and respected international and
studies on a particular topic. They differ from narrative reviews which
not-for-profit organization that promotes, supports, and disseminates
are mainly descriptive and often only focus on a subset of studies in an
systematic reviews and meta-analyses on the efficacy of interventions
area chosen based on availability or author selection.2
in the health care field.
Systematic reviews may or may not include a statistical synthesis
called meta-analysis, which involve techniques to synthesize the data
Why Conduct a Systematic Review and Meta-analysis
from several into a single quantitative estimate or summary effect
size. The type of effect size calculated depends on the type of outcome
Systematic reviews and meta-analyses can be used to produce
and intervention being examined as well as the data available from
statements to guide clinical decision-making, the delivery of care, as
published trials.3 Thus, a meta-analysis is the combination of data from
well as policy development. Systematic reviews and meta-analyses may
several independent primary studies that address the same question
be conducted to: 1) Investigate conflicting results of studies; 2) Address
_______________
variations to clinical practice; 3) Inform areas for future research; and
* Healthway Manila 4) Produce statements to guide decision-making and policy-making. A
**Philippine General Hospital, University of the Philippines Manila systematic review and meta-analysis may be conducted to determine
**Ateneo School of Medicine & Public Health whether or not current practice is based on relevant evidence and to

VOL. 62 NO. 1 JUNE, 2024 51

address any variation in practice that may be occurring. Conducting will be defined by the research question. The FINER criteria (Feasible,
a systematic review and meta-analysis may also identify gaps, Interesting, Novel, Ethical, and Relevant) (Refer to Table 1 & Figure 1)
deficiencies, and trends in the current evidence and can help underpin are used to capture the engaging issues a systematic review should
and inform future research in the area.5 cover.10
An example of published systematic review and meta-analysis
in family and community practice is “The effectiveness of community- Table 1. FINER criteria in formulating a research question.
based interventions in the control of hypertension: A systematic review
and meta-analysis, published in the Filipino Family Physician and may Feasible The study must be doable in terms of the breadth of the review
be accessed in this link. https://thepafp.org/journal/abstract/the- being planned.
effectiveness-of-community-based-interventions-in-the-control-of-
Interesting There is commitment from the authors to see a study to its
hypertension-a-systematic-review-and-meta-analysis/6
conclusion.

Authorship in Meta-analysis: Team Size and Dynamics Novel Made to address a genuine gap in knowledge.

While Cochrane recommends a team of two review authors Ethical The opportunity costs must be considered, especially research
for systematic reviews and meta-analyses, the Philippine Academy that can have political implications that can widen health
of Family Physicians (PAFP) advocates for a team of three review inequalities
authors, consisting of two residents and one consultant who serves
Relevant Involve stakeholders in defining the focus and questions it will
as the group’s adviser.7,8 In instances where the initial two reviewers address and to translate the writing to inform clinical decisions
encounter differing viewpoints or disagreements during the review
process, it is important to resolve these through thorough discussion
or consultation with the third reviewer, who serves as an impartial
arbiter.1 This collaborative approach promotes consensus and ensures
that the final extracted data represent a collective agreement which
reduces subjectivity and strengthens the overall validity of the review.

Steps in Conducting a Systematic Review and Meta-analysis

1. Identify the Purpose

There are two main goals in mind when making a systematic

review and meta-analysis. First is the assessment of the evidence for
the effectiveness of specific interventions for a certain condition or
causal associations attributed to this specific condition. Second is to
paint the big picture using broad generalizations across larger numbers
of study outcomes. These two are often contrasting in their overarching
purposes and must be clarified early in the study.9
When the goal is determining effectiveness of specific
interventions or determining causal associations, PICO (population,
intervention, comparator, outcome) is usually the framework being used
in meta-analysis research in primary care practice in the Philippines.
The approach is different when attempting to do broad generalizations.
Here, the focus is mainly in identifying sources of heterogeneity in large
swathes of knowledge, with deliberate incorporation of heterogeneous Figure 1. Overview of the steps in conducting a systematic review and meta-analysis.
populations so that factors that modify their outcomes can be examined.
This is usually done in social sciences and EEC (ecology, evolutionary
biology, and conservation). For this text, the focus will be on the first Validating the Purpose with Literature Scan
type which is more relevant in family medicine practice in primary care
settings. Doing a literature scan gives the investigator a rough picture of
the current available knowledge for their review. This is a continuous
Formulating the Research Question and overlapping process involving the desire of the reviewer to do
broad or narrow topics. Broader topics allow better, and more accurate
Formulating the topic is an important step and is anchored by generalizations based on the available data, but this may involve a wider
the formulation of the research question. The parameters of the study breadth of literature that needs to be processed. Narrower topics are

52 THE FILIPINO FAMILY PHYSICIAN

generally more feasible, but due to its more specific nature, literature outcomes, such as outcomes of interest, may result in a study being
availability and generalizability are an issue.10 Thus, a literature scan included, while those that prevent certain outcomes may be excluded.
allows investigators to gain methodological ground by identifying Outcomes that are likely to be meaningful to all stakeholders involved
important issues to answer in their chosen topic and balance feasibility should be included. Trivial outcomes should not be included.
of a research question without sacrificing a meaningful output.11
Population
2. Formulate the Objectives
Disease or conditions of interest should be defined using explicit
The primary objective is ideally stated in one sentence and in criteria that establishes their presence or absence. Broad population
the form of “To assess the effects of [intervention or comparison] for and setting of interest should be defined. Factors such as age, sex, race,
[health problem] in [types of people, disease, or problem and setting educational status, or the presence of a particular condition such as pain
if specified].” This may be followed by secondary objectives. The PICO or dyspnea. Settings may include primary care outpatient settings in
(Population, Intervention, Comparator, Outcome) pneumonic is usually low-middle income countries.
the format followed. Different PICO constructs are used at different
The following are useful criteria from Cochrane14
stages of the review:
• How is the disease/condition defined?
• Review PICO – basis of study eligibility to be included.
• What are the most important characteristics that describe
• Synthesis PICO – determining how each synthesis will be
these people (participants)?
structured.
• Are there any relevant demographic factors (e.g. age, sex
• Studies PICO – what was actually investigated in the studies
ethnicity)?
included.
• What is the setting (e.g. outpatient, community, etc)?
• Who should make the diagnosis?
3. Literature Search
• Are there other types of people who should be excluded
from the review (because they are likely to react to the
Search strategies for meta-analysis should be systematic,
intervention in a different way)?
transparent, and reproducible. This must include as many relevant
• How will studies involving only a subset of relevant
studies as possible, including, possibly, unpublished studies. The most
participants be handled?
common search strategy being deployed in research practice is the
use of keywords in electronic databases. The use of Boolean operators
Consider characteristics that might be expected to modify the
is advisable to narrow down search results and make the load more
size of the intervention effects. This means identifying subpopulations
manageable. Databases used for clinical Systematic Review and Meta-
important for the implementation of the intervention which may
analysis research include the following as principal search systems: ACM
modify its applicability or its expected effects. Once subpopulations are
Digital Library, BASE, ClinicalTrials.gov, Cochrane Library, EbscoHost
identified, the researcher may either maintain the breadth of the review
(tested for ERIC, Medline, EconLit, CINHAL Plus, SportsDiscus), OVID
or exclude these subpopulations.
(tested for Embase, Embase Classic, PsychINFO), ProQuest (tested for
Nursing & Allied Health Database, Public Health Database), PubMed,
Intervention
ScienceDirect, Scopus, TRID, Virtual Health Library, Web of Science
(tested for Web of Science Core Collection, Medline), and Wiley Online
Interventions should be predefined and unambiguous in the
Library.
studies to be included. Experimental interventions should be defined by
Other search systems, like Google scholar, are not recommended
specific comparators such as placebo, no treatment, standard of care, or
as a principal search system due to various performance limitations.12
if there is an active control such as a different type of intervention.
Selection of Studies General considerations should include:
• What is delivered?
Studies should include all available literature, including gray • Who delivered it?
literature, in order to prevent selection biases or “Matthew effect,” or • How is it delivered?
the bias towards highly cited articles.13 Methodological peculiarities • Where is it delivered?
and appropriateness of the study in answering the research question • Can the intervention be adapted or tailored?
can serve as reasons for study exclusion. This is conducted in a stepwise
manner: 1) Removal of duplicate citations from different databases; Factors to consider when developing criteria for types of interventions:14
2) Abstract screening and 3) Full text screening • What are the experimental and control (comparator)
For in-depth criteria eligibility, the population, intervention, and interventions of interest?
comparison serve as the primary basis in the eligibility criteria for study • Does the intervention have variations (e.g. dosage/intensity,
inclusion. Outcomes are rarely used as an eligibility criterion. Hence mode of delivery, personnel who deliver it, frequency,
studies should be included irrespective of their outcomes. Meaningful duration or timing of delivery)?

VOL. 62 NO. 1 JUNE, 2024 53

 • Are all variations to be included (for example, is there a the judgment of the review authors about the presence or absence of six
dose below which the intervention may not be clinically domains: 1) Random sequence generation; 2) Allocation concealment;
appropriate, will all providers be included)? 3) Blinding participants or investigators; 4) Incomplete outcome data;
• Will studies including only part of the intervention be 5) Selective reporting; 6) Other biases.
included?
• Will studies including the intervention of interest combined PRISMA Flow Diagram
with another?
• Have the different meanings of phrases such as ‘control’, The PRISMA (Preferred Reporting Items for Systematic Reviews
‘placebo’, ‘no intervention’ or ‘usual care’ been considered? and Meta-Analyses) flow diagram is a crucial element in documenting
and reporting the retrieval, inclusion, and exclusion process of studies
Minor differences can occur across different types of studies and in a systematic review. It provides a concise visual summary of the
should not hinder overall synthesis. However, what constitutes an screening process, ensuring transparency by documenting decisions
important difference requires judgment from the reviewer. Generally, made at each stage.16 Initially, the diagram reports the number of
differences that alter decisions how an intervention is implemented or articles identified during the search. As the review progresses, it
whether an intervention is used are important. tracks the flow of articles, highlighting the number of included and
excluded studies at each stage. It is essential to document the reasons
Comparison for exclusion when excluding articles during the full-text assessment.
By presenting this information in a clear and structured manner, the
Comparisons such as intervention versus placebo, intervention flow diagram enhances the reproducibility of the systematic review.
versus control, or intervention A versus intervention B must be Figure 3 shows the recommended format of the PRISMA 2020 flow
considered. In some studies, multiple comparisons may result from diagram, depicting the selection process, the corresponding numbers
multiple interventions. In these cases, a limited subset can be selected. of included and excluded reports, and the reasons for exclusions. This
The comparisons from these subsets can then be used for the review. diagram should be included in the Results section of the final write-up
of a Systematic Review and/or Meta-analysis
Retrieval of Full Text Articles
Data Extraction
Retrieval of full text articles can be done via OpenAccess articles,
Institutional subscriptions, or personal contact to the author. The extraction of data involves systematically gathering pertinent
information from a selected set of articles to establish a comprehensive
Critical Appraisal of Studies understanding of the research topic. This process ensures that the
review is founded on meticulous and accurate data, thereby minimizing
After retrieving the full text of selected articles, the next step in potential biases. Currently, there are a number of online software that
conducting a systematic review and meta-analysis involves critically manage articles for systematic review including functions for data
appraising the individual studies. The caliber of the individual studies extraction. Some of these are Covidence, Rayyan, and Cachii.
determines how reliable the systematic review and meta-analysis are. When performing data extraction manually, a standardized
Meta-analyses can produce false conclusions if the findings of the initial Data Extraction Form for each study should be used by all review
studies are skewed. authors as seen in Figure 2. The reviewers should plan and discuss
Critical appraisal encompasses systematically assessing the risk what data are to be extracted from the studies and independently
of bias using predefined criteria, evaluating the relevance of the study extract data from each study. 7,8 In situations where initial reviewers
populations, interventions, and outcome measures, and assessing encounter differing viewpoints or disagreements during the review
the fidelity of intervention implementation. This process ensures the and extraction process (e.g., whether a study should be included
synthesis of credible and reliable evidence, enhancing the validity and or excluded), a consultation with the third reviewer can be done to
impact of the review’s findings. settle the decision. Table 3 presents the essential information that
must be obtained at a minimum.
Assessment of Bias
Data Analysis
Bias is defined as a systematic error or deviation from the truth
in the results obtained.15 The presence of bias might cause the actual As previously discussed, a meta-analysis synthesizes results
intervention effect to be underestimated or overestimated. To evaluate from different studies in order to derive a conclusion that has greater
the reliability of the studies, review authors may include a section on accuracy and precision than may be possible when looking at individual
Assessment of Bias in the data extraction form, as shown in Figure 2. studies. To do this, it is important to assess the direction of the effect,
The study limitations are evaluated by review authors using the size of the effect, homogeneity of effects among studies, and strength
“Risk of Bias” method proposed by Cochrane 2022 (Table 2). Risk of bias of evidence.1 This evaluation sets the foundation for a comprehensive
is classified in randomized studies as “low”, “unclear”, or “high” based on review of the data, both qualitatively and quantitatively.

54 THE FILIPINO FAMILY PHYSICIAN

 Figure 2. Sample Data Extraction Form, adapted from the study “The Effectiveness of Community-based Interventions in the Control of Hypertension - A
Systematic Review and Meta-analysis” by Arteza, Yayen, and Mendoza (2022) published in The Filipino Physician Journal (Arteza, et al, 2022)

Table 2. Cochrane Risk of Bias Assessment Tool adapted from Cochrane 2022.

VOL. 62 NO. 1 JUNE, 2024 55

Table 3. Essential information to be obtained from each study (Mendoza 2022;
Espallardo 2023).

Information Purpose

1. The title of the article / Provides the title of the study, aiding
Name of the Journal in the identification and referencing
of the specific article

2. The name of the first author Proper identification and citation of

the study

3. The year the article was published Provides temporal context and
establishes the chronological timeline
of research on the topic

4. The population on whom

the study was conducted Specifies the characteristics of the
individuals or groups involved in the
study

5. The type of research Identifies the study design (e.g.,

randomized controlled trials, cohort
studies, case-control studies)

6. Brief description of the intervention Provides information on the

intervention’s dosage, form,
Figure 3. PRISMA 2020 flow diagram for new systematic reviews which included
administration method, and other
searches of databases and registers only (Figure recreated from PRISMA 2020 Flow
relevant details
Diagram template.)

56 THE FILIPINO FAMILY PHYSICIAN

 Figure 4. PRISMA flow diagram from The Effectiveness of Community-based Interventions in the Control of Hypertension: A Systematic Review and Meta-
analysis by Arteza, Yayen, Mendoza (2022); The Filipino Family Physician Journal.

A qualitative review is done if the different research outcomes uploads/protected_file/RevMan5.4_user_guide.pdf. Additionally,

cannot be combined due to inherent differences.1 This involves online learning resources on using RevMan 5.4 can be found at https://
presenting the results and characteristics of the individual studies training.cochrane.org/online-learning/core-software/revman.
in a tabulated or a descriptive format allowing for a comprehensive
understanding of each study’s contribution to the topic at hand. Establish Types of Data and Effect Measures: Continuous and Dichotomous
Conversely, if the results are comparable and can be combined, then the Variables
review authors can proceed to create a quantitative review or a meta-
analysis. When evaluating data, there are two types of outcome variables:
A meta-analysis typically involves two stages. In the first stage, continuous variables and dichotomous variables. Continuous variables
a summary statistic or an effect measure is computed to effectively are measurements that can take any value within a specific range,
describe the observed intervention effect in the same manner across such as blood pressure or cholesterol levels. (Table 3) In contrast,
all studies. This summary statistic may take the form of a risk ratio for dichotomous variables are outcomes with only two possible categories,
dichotomous data or a difference between means for continuous data, such as the presence or absence of disease. In studies, the outcomes
ensuring uniformity in reporting the effect size. In the second stage, of participants are summarized by an estimate of effect or association,
the clinical effectiveness of the treatment is assessed by calculating a these are referred to as Effect Measures or Measures of Association. The
weighted pooled estimate based on individual studies.15 This pooled types of Effect Measures for each outcome data are enumerated in Table
estimate serves as the primary outcome of the meta-analysis and 4.
is often visually represented using a forest plot, providing a clear When combining data from continuous variables, the effect
overview of the collective evidence. The calculations and generation of measures used are the mean difference (MD) and standardized mean
a forest plot are done using statistical software (e.g., Review Manager) difference (SMD). The mean difference represents the absolute
that researchers can either download for free or access through a difference in mean values between groups and is used when all studies
paid subscription. An example of a forest plot generated through the utilize the same scale of measurement (e.g., blood pressure in mmHg).1
software Review Manager v 5.4 is illustrated in Figure 5. Alternatively, the standardized mean difference is employed when
For detailed instructions on how to use Review Manager, version studies assess the same outcome using different scales or instruments
5.4, you can refer to the official user guide provided by Cochrane, (e.g., WHO Quality of Life Measurement and the McGill Quality of Life
which is available at https://training.cochrane.org/system/files/ Questionnaire).

VOL. 62 NO. 1 JUNE, 2024 57

 Figure 5. Example of a Forest plot illustrating the pooled estimate of community-based interventions in the control of hypertension,adapted from the study
“The Effectiveness of Community-based Interventions in the Control of Hypertension - A Systematic Review and Meta-analysis” by Arteza, Yayen, and Mendoza
(2022) published in The Filipino Family Physician (Arteza et al, 2022)

Table 4. Outcome Variables with corresponding definitions, examples, and effect measures.

●
●
● ●
●
●
●

●
●
● ●

The effect measures used in dichotomous variables are enumerated a systematic review and meta-analysis, ultimately providing valuable
in Table 4. For randomized controlled trials (RCTs), quasi-experimental insights into the efficacy of interventions or the impact of risk factors.
studies, or cohort studies The Risk Ratio and Risk Difference can be used.
On the other hand, in case-control studies or cross-sectional studies, the Analyze the Effect Size: Fixed-effects and Random-effects Models
Odds Ratio is typically used.1 It should be noted that the interpretation
of the Odds Ratio can be challenging; hence, using the Risk Ratio or Risk In a meta-analysis, each study is summarized by an estimate, and the
Difference is recommended when possible. overall measure of effect is calculated as a weighted average of these
When choosing an effect measure for a meta-analysis, it is individual study results. 15 When analyzing the effect size, review
important to consider whether it conveys the necessary clinically useful authors can choose between two models: the Fixed Effects Model and
information, is appropriate for the study design, and is statistically the Random Effects Model. Before choosing the type of model to be used
suitable and convenient for analysis. By understanding the nature of in the meta-analysis, researchers must consider assumptions intrinsic
continuous and dichotomous variables and selecting appropriate effect in each model as well as factors within each study and differences
measures, review authors can effectively combine and analyze data in between studies. Refer to Table 5.

58 THE FILIPINO FAMILY PHYSICIAN

 Table 5. Comparison of fixed effects model and random effects model.

The Fixed Effects Model assumes that all studies in the meta- Evaluate Heterogeneity
analysis share a common true effect size.7 This means that the
treatment’s effect is consistent across all studies, and any apparent Heterogeneity arises from variations in experimental effects
variation across studies is solely attributed to random errors inherent analyzed across studies included in the meta-analysis, which cannot be
in each study. In essence, if it were not for random (sampling) error, solely attributed to random error. It results from differences in clinical
all study results would be identical. Consequently, the Fixed Effects and methodological aspects across each study.15 Clinical variations
Model assigns greater weight to larger studies, as they provide more can lead to different intervention effects measured in different types
precise estimates.1 This model is appropriate when the studies are of studies, while methodological differences reflect varying degrees
homogeneous, meaning they are drawn from the same population, of bias. However, methodological diversity does not necessarily imply
share similar characteristics, adopt the same design and methodology, variations in the true outcome across studies. Thus, a valid meta-
or when there is minimal variability in results within each study and any analysis should include studies that are sufficiently homogeneous in
remaining variance is thought to be caused by random error. terms of participants, interventions, and outcomes. Evaluating the
On the other hand, the Random Effects Model assumes that there heterogeneity of the included studies is therefore crucial for each
is heterogeneity or variation in the true effect size across studies, in assessed outcome.
addition to sampling error. This model is used when the studies are Qualitatively, heterogeneity can be assessed by comparing the
considered different from each other.1 It considers both within-study confidence intervals of individual studies in the forest plot. Poor overlap
sampling error and between-study variation when estimating the among confidence intervals suggests statistical heterogeneity. In Table
overall effect size. Unlike the Fixed Effects Model, the Random Effects 6, studies with smaller confidence intervals show limited overlap,
Model does not significantly decrease the weight assigned to studies indicating possible statistical heterogeneity.
with a small number of patients, as it acknowledges the potential Quantitatively, the chi-square (χ2 or Chi2) test is used to evaluate
heterogeneity among studies. the presence of statistically significant heterogeneity across the included
In summary, the choice between these models is determined studies. A p-value below the threshold of statistical significance
by the underlying assumptions regarding the similarity of studies (commonly p < 0.05 or p < 0.10) indicates significant heterogeneity
and the presence of heterogeneity. If the studies are assumed to be among the outcome estimates. For instance, in Figure 5, the p-value is
homogeneous, the Fixed-Effects Model is appropriate. However, if p = 0.004, indicating a statistically significant degree of heterogeneity.
substantial heterogeneity is deemed present between studies, the Conversely, the Higgins I2 test is employed to assess the degree of
Random Effects Model is preferred as it provides a more comprehensive statistical heterogeneity. Interpretation of I2 values varies across
estimate by incorporating both within-study and between-study references, but the Cochrane Handbook suggests the following
variation. thresholds listed in Table 6. In Figure 5, the I2 value is I2 = 74%,

VOL. 62 NO. 1 JUNE, 2024 59

indicative of moderate to substantial heterogeneity across the included A funnel plot is a scatter plot that compares the intervention effect
studies which the researchers attributed to differences in the study point estimates of each study against a measure of its size or precision.
population, study location, and intervention.6 The plot is based on the understanding that the precision of the
calculated outcome is directly proportional to the study’s size. A funnel
Table 6. I2 values with their corresponding degrees of heterogeneity according to
plot can be constructed when there are more than five studies included
Cochrane Handbook 2022. in the meta-analysis.7,8
Typically, this plot is generated using statistical software such as
I2 Value Degree of Heterogeneity Review Manager, providing a graphical representation of potential bias
by displaying the distribution of study results relative to their sample
0%-40% Might not be important size or precision. In the funnel plot, the intervention effect estimates are
plotted on the vertical axis, while the sample size or precision is plotted
30%-60% May represent moderate heterogeneity
on the horizontal axis for each study. Consequently, smaller studies may
50%-90% May represent substantial heterogeneity be spread out at the bottom of the graph, with the spread narrowing
as studies increase in size or precision. Interpreting the funnel plot
75%-100% Considerable heterogeneity is qualitative: the presence of a symmetric inverted funnel suggests
the absence of publication bias (Figure 6-A), while an asymmetrical
appearance indicates potential publication bias (Figure 6-B). The extent
Assess for Publication Bias of asymmetry corresponds to the magnitude of bias.

Publication Bias is the most common bias in meta-analyses Sensitivity Analysis

(Ahn, 2018), it refers to the selective publication or non-publication
of research results based on their nature and direction (Cochrane, A Sensitivity Analysis examines the robustness of the conclusions
2022). It has been widely acknowledged that published trials tend to with respect to the overall quality of the included studies.15 It is a
present results favoring the intervention and are more likely to show procedure used to determine how the values of an independent variable
statistical significance compared to unpublished trials. Studies with will influence the significance of a given dependent variable when one
positive findings are also more likely to receive funding and subsequent or more studies are removed from the meta-analysis.17 Following the
publication, while those with unfavorable results may be less likely to individual article appraisals, certain studies may be identified as having
be published or may be published later. Additionally, smaller studies a high risk of bias, which could potentially impact or invalidate the
with inconclusive or negative findings face higher chances of non- overall result, particularly if it is statistically significant. In such cases,
publication or delayed publication compared to larger studies with a sensitivity analysis can be conducted by reassessing the relevant
favorable experimental results.7,8 Therefore, a funnel plot is generated outcome after excluding the studies with a high risk of bias from the
to assess publication bias among the included studies in the meta- analysis. If the sensitivity analysis demonstrates that the overall result
analysis. and conclusions remain unchanged despite the removal of studies with

Figure 6. Theoretical Funnel Plots showing the effect size on the x-axis and sample size on the y-axis as a scatter plot. (A) Funnel plot without publication bias. The individual
plots are broader at the bottom and narrower at the top. (B) Funnel plot with publication bias. The individual plots are located asymmetrically.

60 THE FILIPINO FAMILY PHYSICIAN

a higher risk of bias, the results are considered more robust, and their interventions and (b) the types of interventions employed. Identification
interpretation becomes more reliable. and analysis of subgroups based on characteristics offer insights into
factors contributing to observed heterogeneity, refining the precision
Subgroup Analysis and applicability of findings.
The plan for subgroup analysis should be explicitly stated
Subgroup analysis involves classifying participant data into in the research protocol. A priori decisions enable transparency,
subsets for comparison based on criteria such as gender, geographical methodological rigor, and the mitigation of data-driven issues. Criteria,
area, or other relevant parameters.15 It addresses issues about specific rationale, and anticipated differences in treatment effects based on
patient groups, interventions, or study designs, particularly when faced chosen factors should be clearly outlined in the research protocol. This
with persisting heterogeneity despite a random-effects model. In such pre-specification reduces bias and increases the credibility of study
cases, subgroup analysis becomes critical, assisting in the exploration results. If additional relevant subgroups emerge during the research,
of potential causes of variability by determining whether the treatment this should be documented as exploratory post hoc analyses.
impact varies across study subgroups.
In Figure 7, the forest plots illustrate the subgroup analysis Ethical Issues
conducted by Arteza et al in their study entitled ‘The Effectiveness of
Community-based Interventions in the Control of Hypertension: A A systematic review and meta-analysis study does not require
Systematic Review and Meta-analysis.’ The subgroup analyses were ethics board approval as it involves analyzing data that has been
based on two key factors: (a) who is responsible for implementing the previously published, assuming it was approved by respective review

Figure 7. Forest plots illustrating the Reduction of DBP with subgroups based on (a) who is responsible for the interventions, (b) type of interventions, adapted
from the study “The Effectiveness of Community-based Interventions in the Control of Hypertension - A Systematic Review and Meta-analysis” by Arteza, Yayen,
and Mendoza (2022) published in The Filipino Family Physician (Arteza et al, 2022)

VOL. 62 NO. 1 JUNE, 2024 61

boards. Nevertheless, ethical considerations play a crucial role in Lastly, accurate data synthesis and analysis techniques are
conducting and writing the final report of a systematic review and meta- essential for maintaining the credibility of the review. Researchers
analysis to maintain the credibility of the research process. Maintaining should employ robust statistical methods and clearly report their
academic honesty and intellectual integrity is paramount to ensure findings. Misrepresentation or manipulation of data can significantly
the accurate representation of ideas and findings. Good practices undermine the trustworthiness of the review and compromise its
that research authors should observe include: 1) Avoiding plagiarism; scientific value. By adhering to rigorous data synthesis and analysis
2) Maintaining transparency; and 3) Ensuring data accuracy.18 practices, researchers demonstrate their commitment to producing
Plagiarism is defined as the act of using another person’s reliable and unbiased results.
original words, ideas, data, or other creations without proper credit
or permission. Review authors must correctly cite and acknowledge Write the Final Report using the PRISMA Guidelines
prior work, giving credit to the original authors, in order to avoid
plagiarism. It is acceptable to describe other works using the review The final report of a systematic review and meta-analysis is vital
author’s own words, as long as proper citations are provided to attribute in disseminating the findings to the scientific community and guiding
the information to its original source.18 By adhering to rigorous evidence-based decision-making. This final report is done based on the
citation practices, researchers demonstrate respect for the intellectual Preferred Reporting Items for Systematic Reviews and Meta-Analyses
contributions of others and uphold the principles of academic integrity. (PRISMA) 2020 guideline.16 The PRISMA 2020 guideline is a widely
Another ethical consideration that should be maintained throughout recognized and recommended framework for reporting systematic
the research process is transparency. Transparent reporting ensures reviews and meta-analyses. By adhering to this guideline, researchers
the accuracy and reliability of the review. Review authors should can ensure transparency and reproducibility in their final reports,
provide detailed descriptions of the methods used, including the facilitating the effective communication and interpretation of research
criteria for study selection, data extraction, and analysis techniques. findings. The checklists presented in Tables 7 and 8 summarize the
By openly disclosing the procedures and decision-making processes key components of the 2020 Preferred Reporting Items for Systematic
involved, readers can assess the reliability and validity of the review. Reviews and Meta-Analyses (PRISMA) guideline. Each section of the
Additionally, any potential conflicts of interest should be disclosed to guideline corresponds to specific items that are crucial for ensuring the
ensure transparency and minimize bias. transparency, thoroughness, and reproducibility of the review process.

Table 7. PRISMA 2020 item checklist.

62 THE FILIPINO FAMILY PHYSICIAN

VOL. 62 NO. 1 JUNE, 2024 63
64 THE FILIPINO FAMILY PHYSICIAN
 Table 8. PRISMA 2020 for abstracts checklist.

References 7. Mendoza N. How to conduct meta-analysis and systematic reviews

[Lecture/Presentation]. RECAP Online Webinar The Philippine Academy of
1. Ahn E, Kang H. Introduction to systematic review and meta-analysis. Family Physicians; 2022.
Korean J Anesthesiol 2018 Apr;71(2):103-12. doi: 10.4097/kjae.2018. 8. Espallardo N. How to Conduct Meta-analysis and Systematic Review
71.2.103. [Lecture/Presentation]. RECAP Online Webinar The Philippine Academy of
2. Uman and Lindsay. Systematic reviews and meta-analyses. J Can Acad Family Physicians; 2023.
Child Adolesc Psychiatr 2011; 20:1. 9. Gurevitch J, Koricheva, J, Nakagawa S & Stewart G. Meta-analysis and the
3. Petticrew M & Roberts H. Systematic Reviews in the Social Sciences: A science of research synthesis. Nature 2018; 555(7695): 175–82. https://
Practical Guide. Malden, MA: Blackwell Publishing 2008. doi.org/10.1038/nature25753
4. Gopalakrishnan S, Ganeshkumar P. Systematic reviews and meta-analysis: 10. Thomas J, Kneale D, McKenzie JE, Brennan SE, Bhaumik S. Chapter 2:
Understanding the best evidence in primary healthcare. J Fam Med Prim Determining the scope of the review and the questions it will address.
Care 2013. In: Higgins JPT, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, Welch
5. Munn, et al. Systematic review or scoping review? Guidance for authors VA (editors). Cochrane Handbook for Systematic Reviews of Interventions
when choosing between a systematic or scoping review approach. BMC version 6.3 (updated February 2022). Cochrane, 2022. Available from
Med Res Method 2018; 18:143. www.training.cochrane.org/handbook.
6. Arteza MAJ, Yayen KMR, Mendoza NRNP. The effectiveness of community- 11. Steel P, Beugelsdijk S & Aguinis H. The anatomy of an award-winning
based interventions in the control of hypertension: A systematic review meta-analysis: Recommendations for authors, reviewers, and readers
and meta-analysis. The Filipino Family Physician 2022; 60(1):1-10. of meta-analytic reviews. J Int Bus Stud 2021; 52: 23–44. https://doi.
org/10.1057/s41267-020-00385-z

VOL. 62 NO. 1 JUNE, 2024 65

12. Gusenbauer M & Haddaway NR. Which academic search systems are 15. Higgins JPT, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, Welch VA.
suitable for systematic reviews or meta-analyses? Evaluating retrieval Cochrane Handbook for Systematic Reviews of Interventions version 6.3
qualities of Google Scholar, PubMed, and 26 other resources. Research (updated February 2022). Cochrane, 2022. Available from www.training.
Synthesis Methods 2020; 11(2): 181–217. https://doi.org/10.1002/ cochrane.org/handbook.
jrsm.1378 16. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD,
13. Hansen C, Steinmetz H & Block J. How to conduct a meta-analysis in et al. The PRISMA 2020 statement: an updated guideline for reporting
eight steps: a practical guide. Manag Rev Q 2022; 72: 1–19. https://doi. systematic reviews. BMJ 2021;372:n71. doi: 10.1136/bmj.n71.
org/10.1007/s11301-021-00247-4 17. Tawfik GM, Dila KAS, Mohamed MYF, Tam DNH, Kien ND, Ahmed AM, Huy
14. McKenzie JE, Brennan SE, Ryan RE, Thomson HJ, Johnston RV, Thomas J. NT. A step by step guide for conducting a systematic review and meta-
Chapter 3: Defining the criteria for including studies and how they will be analysis with simulation data. Trop Med Health 2019 Aug 1;47:46. doi:
grouped for the synthesis. In: Higgins JPT, Thomas J, Chandler J, Cumpston 10.1186/s41182-019-0165-6.
M, Li T, Page MJ, Welch VA (editors). Cochrane Handbook for Systematic 18. Wager E, Wiffen PJ. Ethical issues in preparing and publishing systematic
Reviews of Interventions version 6.3 (updated February 2022). Cochrane, reviews. J Evid Based Med 2011 May;4(2):130-4. doi: 10.1111/j.1756-
2022. Available from www.training.cochrane.org/handbook. 5391.2011.01122.x.

66 THE FILIPINO FAMILY PHYSICIAN