Narrative reviews

Narrative review is one of the most traditional and common types of review in academic writing, with a history spanning thousands of years. The primary aim of this type of review is to provide a comprehensive, qualitative synthesis of the existing literature on a particular topic, without employing a systematic search strategy or predetermined criteria for study selection (Mahdavinoor Reference Mahdavinoor, Ladan-Moghadam and Mahdavinoor2025).

Narrative reviews are highly susceptible to bias. For example, suppose we intend to examine the association between overweight and depression through a narrative review. If, worldwide, 100 articles have been published on this topic, of which only 15 report a significant association and 85 do not, in a narrative review, one could selectively cite from either set of studies and present a conclusion aligned with one’s viewpoint (Mahdavinoor Reference Mahdavinoor, Ladan-Moghadam and Mahdavinoor2025). Another challenge with this type of review is its lack of reproducibility. Scientific research should be replicable by other researchers, but narrative reviews lack this feature because of the absence of a transparent methodology. Moreover, the possibility of conflict of interest is present because of the selective inclusion of studies (Mahdavinoor Reference Mahdavinoor, Ladan-Moghadam and Mahdavinoor2025).

For these reasons, some journals may be less inclined to publish narrative reviews. Even when they do accept them, conditions are often imposed; for instance, they may only consider submissions from recognised experts, or specify that such reviews can only be submitted by invitation. Nevertheless, narrative reviews can be useful for readers unfamiliar with a topic who seek a broad overview of a particular field. They may also provide researchers with a platform to articulate new ideas within a given domain.

For practical guidance on writing narrative reviews, see Chaney (Reference Chaney2021), Sarkar & Bhatia (Reference Sarkar and Bhatia2021) and Ferrari (Reference Ferrari2015).

Systematic reviews

A systematic review is a structured and rigorous methodology for synthesising research evidence using explicit, transparent and reproducible methods. Unlike narrative reviews, systematic reviews follow a predefined protocol in which the research question, inclusion and exclusion criteria, data extraction procedures, methods for quality assessment and strategies for data synthesis are specified in advance.

In a systematic review, all records retrieved from specified databases for a defined time frame are screened, and studies that meet the predefined inclusion criteria are included, regardless of the direction of their results. This minimises bias and makes the findings reproducible and reliable. For example, if we aim to conduct a systematic review on the effectiveness of cognitive–behavioural therapy for adolescents with depression, we first develop a protocol in which the methods are described in detail (including the research question, eligibility criteria, search strategy, etc.). We then conduct the search in accordance with this protocol. In this specific example, we might search Scopus and PubMed and screen records indexed between 2000 and 2025. If the search yields, for instance, 3000 records, we are required to screen all retrieved records against the predefined inclusion criteria to determine which studies should be included. This structured process helps minimise researcher selection bias and improves transparency and reproducibility.

Meta-analyses

Meta-analysis refers to the statistical combination of results from two or more studies (Higgins Reference Higgins, Thomas and Chandler2024). In practice, when multiple primary studies exist in a given field (e.g. the effectiveness of drug X on depression) and these studies have reported different results, we can combine the findings and assign weights to the individual studies in order to obtain a pooled estimate along with a confidence interval (Dekkers Reference Dekkers2018). As another example, suppose there are 31 provinces in Iran, and in each province a cross-sectional study has been conducted to estimate the prevalence of suicide among university students. By performing a meta-analysis of these prevalence estimates across provinces, we can calculate an overall percentage that represents the approximate prevalence of suicide among university students in Iran.

Meta-analysis serves several purposes, including increasing statistical power, exploring heterogeneity across studies and providing a more precise estimate of effect (Box 1) (Borenstein Reference Borenstein, Hedges and Higgins2009; Higgins Reference Higgins, Thomas and Chandler2024). A systematic review may lead to a meta-analysis if the quantitative data extracted are sufficiently similar and not highly heterogeneous. Sometimes, researchers proceed to a meta-analysis without adequately considering the degree of methodological heterogeneity among the included studies, which can call the credibility of the results into question. When methodological heterogeneity is high, it may be preferable to present a systematic review without meta-analysis (Borenstein Reference Borenstein, Hedges and Higgins2009; Higgins Reference Higgins, Thomas and Chandler2024).

For practical guidance on conducting a meta-analysis, see Higgins et al (Reference Higgins, Thomas and Chandler2024) and Tawfik et al (Reference Tawfik, Dila and Mohamed2019).

Umbrella reviews

The findings of primary studies are often inconsistent, which can create confusion in clinical decision-making and health policy (Mahdavinoor Reference Mahdavinoor, Ladan-Moghadam and Mahdavinoor2025; Ioannidis Reference Ioannidis2005). To address this challenge, systematic reviews and meta-analyses – considered second-level studies – have been designed to synthesise and analyse the results of primary studies, thereby providing a more comprehensive and reliable picture of the available evidence. For example, suppose that 20 primary studies worldwide have investigated the effect of a particular drug on reducing blood pressure. A systematic review and meta-analysis would attempt to summarise these 20 studies and statistically combine their results to generate an overall and trustworthy estimate of the drug’s effectiveness. However, in some cases, multiple independent systematic reviews and meta-analyses are published on the same topic. This is where the umbrella review comes into play: a third-level study that synthesises, evaluates and compares the findings of existing systematic reviews and meta-analyses to provide a higher-level summary of the evidence (Fernandez Reference Fernandez, Sharifnia and Khalil2025).

An umbrella review is a type of review that focuses not on primary studies, but rather on systematic reviews and meta-analyses (Fernandez Reference Fernandez, Sharifnia and Khalil2025). In other words, an umbrella review examines systematic reviews and meta-analyses rather than primary studies. Compared with meta-analyses, the research questions in umbrella reviews are usually broader (Asgarizadeh Reference Asgarizadeh and Zardkhaneh2024). For instance, a single meta-analysis may assess whether one particular factor is a risk factor for depression. Globally, multiple meta-analyses have been conducted to identify various risk factors for depression, and even for a single factor, several meta-analyses may exist. An umbrella review, by contrast, would examine all risk factors for depression and synthesise the results of all relevant meta-analyses to provide a comprehensive and precise picture of the risk factors for depression.

Currently, umbrella reviews sit at the top of the hierarchy of evidence (Paolo Reference Paolo and Joaquim2018; Asgarizadeh Reference Asgarizadeh and Zardkhaneh2024) because they synthesise findings from existing systematic reviews and meta-analyses; however, this position refers to the level of synthesis, not to inherent methodological superiority or certainty of evidence. It is important to emphasise that no review type is intrinsically better than another; the most appropriate approach depends on the research question, the level and state of the available evidence, and practical constraints such as time and resources.

For practical guidance on conducting an umbrella review, see Fernandez et al (Reference Fernandez, Sharifnia and Khalil2025), Paolo and Joaquim (Reference Paolo and Joaquim2018), Biondi-Zoccai (Reference Biondi-Zoccai2016) and Aromataris et al (Reference Aromataris, Fernandez and Godfrey2015).

Network meta-analyses

Suppose there are several different treatments for depression, each shown to be effective in individual studies. How can we determine which of these treatments is most effective? Network meta-analysis enables the simultaneous comparison of multiple interventions within a single analysis by combining both direct and indirect evidence (Lumley Reference Lumley2002).

The advantage of network meta-analysis is that it maximises the use of available evidence and allows for comparisons that have not been directly tested, thereby producing a coherent analysis of an entire set of interventions. For example, if one trial compares treatment A with treatment B, and another trial compares treatment B with treatment C, a network meta-analysis allows us to indirectly compare treatment A with treatment C. In addition, by ranking interventions, network meta-analysis provides clinicians with a clearer sense of the relative priority of treatment options (Box 2) (Salanti Reference Salanti2012; Higgins Reference Higgins, Thomas and Chandler2024).

One mistake that researchers may unintentionally make when choosing this review type is to conduct a network meta-analysis simply because multiple interventions exist, even when the available evidence is too limited, uncertain or unsuitable for reliable indirect comparisons and treatment rankings (Mbuagbaw Reference Mbuagbaw, Rochwerg and Jaeschke2017; Higgins Reference Higgins, Thomas and Chandler2024).

For practical guidance on conducting a network meta-analysis, see Higgins et al (Reference Higgins, Thomas and Chandler2024).

Meta-epidemiological studies

Why is it recommended that blinding (masking) be performed in trials whenever possible? What advantage does blinding provide? How about random allocation? Do lower-quality studies tend to produce larger effect estimates compared with higher-quality studies? Meta-epidemiological studies are conducted to answer such questions. Put simply, meta-epidemiology investigates how study characteristics influence study results across multiple studies (Naylor Reference Naylor1997; Puljak Reference Puljak, Makaric and Buljan2020). In this approach, the focus shifts from a clinical question such as ‘Is this intervention effective?’ to a methodological one: ‘Does the way studies are conducted influence the results we obtain?’

For practical guidance on conducting a meta-epidemiological study, see Zhang (Reference Zhang2016).

Scoping reviews

The JBI Scoping Review Methodology Group defines scoping reviews as a form of evidence synthesis conducted with the aim of systematically identifying and mapping the breadth of studies available on a particular topic, field or issue, without being restricted to a specific type of source (such as primary studies, reviews or non-empirical evidence). These reviews can help clarify key concepts and definitions in the literature and identify important features or factors related to the topic under investigation, including methodological aspects (Munn Reference Munn, Pollock and Khalil2022).

Put more simply, a scoping review is an exploratory method of evidence synthesis that seeks to determine ‘what exists’ in a research area, rather than ‘what works’ (Munn Reference Munn, Pollock and Khalil2022). In contrast to systematic reviews, which typically address a precise question (e.g. ‘Is treatment X effective for condition Y?’), a scoping review aims to chart the entire field (e.g. ‘What types of study have been conducted on condition Y?’ or ‘What concepts have been discussed in the literature regarding treatment X?’).

The objectives of scoping reviews include (Munn Reference Munn, Peters and Stern2018b):

1. (a) identifying the types of evidence available in a particular field;

2. (b) refining and clarifying essential terms and definitions used in the literature;

3. (c) clarifying key concepts and definitions in the literature;

4. (d) identifying important features or factors related to a concept;

5. (e) serving as a preliminary step towards a systematic review;

6. (f) identifying and analysing existing knowledge gaps in the field.

A fundamental distinction between scoping reviews and systematic reviews lies in the research question and the scope of evidence eligible for inclusion. In a systematic review, the research question is usually narrow and specific, focusing on evaluating the effectiveness of an intervention, a risk factor or a particular association between variables. Accordingly, the included evidence is usually limited to the study designs or publication types most appropriate for answering that specific question; for example, a systematic review may include only randomised controlled trials (RCTs), only observational studies, or another clearly defined type of evidence, depending on its aim. In scoping reviews, however, the question is usually broader and more exploratory, and a wider range of evidence sources and publication types may be included within the same review, from letters to the editor to meta-analyses and RCTs (Munn Reference Munn, Pollock and Khalil2022; Pollock Reference Pollock, Peters, Tricco, Aromataris, Lockwood and Porritt2024). For instance, suppose we wish to investigate interventions for reducing anxiety among university students. If our aim is solely to assess the effectiveness of each intervention, a systematic review and meta-analysis would be appropriate. However, if the goal is to identify the full range of interventions implemented, the types of population studied, the methods of delivery, the assessment tools used and the contexts of research, then a scoping review would be the method of choice.

For practical guidance on conducting a scoping review, see Munn et al (Reference Munn, Pollock and Khalil2022), Pollock et al (Reference Pollock, Peters, Tricco, Aromataris, Lockwood and Porritt2024) and Mak & Thomas (Reference Mak and Thomas2022).

Mapping reviews

To my knowledge, this type of review was first introduced by the Evidence for Policy & Practice Information Centre (EPPI Centre) at University College London (Gough Reference Gough, Oliver and Thomas2017; EPPI Centre 2005). Its primary aim is to answer questions such as ‘What do we know about a topic?’ or ‘What research exists, and where, within a specific field?’ (Campbell Reference Campbell, Tricco and Munn2023). In other words, the main purpose of a mapping review is to identify research gaps in a given area. For example, suppose we plan to conduct a mapping review on studies of suicide among university students in Iran. We would need to address questions such as: What types of study design have been conducted most frequently? What measurement tools have been used? In which regions have most studies been carried out? What has been the trend in the growth and publication of relevant articles over time? By reviewing studies related to suicide among Iranian students and answering these and similar questions, we can identify existing research gaps.

Mapping reviews may sometimes have a broader focus than scoping reviews but extract more limited data from the included studies (Campbell Reference Campbell, Tricco and Munn2023).

For practical guidance on conducting a mapping review, see James et al (Reference James, Randall and Haddaway2016) and Khalil et al (Reference Khalil, Welch and Grainger2025).

Rapid review

A rapid review is a form of knowledge synthesis that accelerates the process of conducting a traditional systematic review by simplifying or omitting certain methodological steps to produce evidence in a resource-efficient and timely manner (Hamel Reference Hamel, Michaud and Thuku2021).

Traditional systematic reviews, by adhering to rigorous guidelines, can provide relatively reliable evidence, but their completion can take a long time – often more than a year. At times, however, policymakers need to make urgent decisions and therefore require trustworthy evidence quickly (Box 3). In such situations, conducting a traditional systematic review may not be feasible, and instead, a type of review is needed that can be carried out within a shorter time frame. Rapid reviews fulfil this need by expediting the systematic review process through the simplification of certain steps (e.g. data extraction carried out by a single reviewer instead of two) or omission of others (e.g. skipping the quality appraisal of included studies), making them feasible within a limited time (Hamel Reference Hamel, Michaud and Thuku2021). For instance, suppose a novel infectious disease is spreading across a country, and the government needs to decide whether drug X is effective enough to be included in treatment guidelines. In such circumstances, where quick decision-making is essential, conducting a rapid review would be the most suitable option.

For practical guidance on conducting a rapid review, see Hamel et al (Reference Hamel, Michaud and Thuku2021), Tricco et al (Reference Tricco, Langlois and Straus2017) and Garritty et al (Reference Garritty, Hamel and Trivella2024).

State-of-the-art reviews

This type of review, which is time-oriented, examines the current state of knowledge about a phenomenon and suggests directions for future research (Grant Reference Grant and Booth2009). Questions such as ‘Where are we now in a particular field of science? Where were we before, and how did we get here? And where are we heading?’ are typically addressed in state-of-the-art reviews (Barry Reference Barry, Merkebu and Varpio2022a). For example, suppose we intend to conduct a state-of-the-art review of treatments for depression. We would first examine when treatments for depression were initially introduced and trace their historical trajectory to the present point (for instance, beginning with psychoanalysis, then the emergence of cognitive–behavioural therapy, followed by pharmacotherapy and, most recently, artificial intelligence-based interventions). Next, we would analyse why the treatment pathways developed as they did (e.g. why pharmacological treatments became dominant). Finally, we would assess where the field currently stands and what potential directions lie ahead for the future. A common mistake is to provide only a descriptive history or a fragmented summary of the literature. A state-of-the-art review should instead interpretively show where the field currently stands, how it has reached this point and where it may move in the future.

For practical guidance on conducting a state-of-the-art review, see Barry et al (Reference Barry, Merkebu and Varpio2022a, Reference Barry, Merkebu and Varpio2022b, Reference Barry, Merkebu and Varpio2022c).