Writing a Successful Systematic Review Manuscript for a Radiology Journal

Systematic reviews (SRs) are valuable tools in summarizing evidence in the literature on a selected topic. SRs are considered one of the strongest forms of scientific evidence because they provide high statistical power and precise results, and they help resolve conflicting results across individual studies. A well-conducted SR can improve patient care by providing strong evidence. The SR publication rate in imaging journals has increased >23-fold in the last two decades. ¹ The most frequent subtype is diagnostic accuracy SR, which accounts for almost half of all published SRs in radiology journals.^1,2 In this editorial, we highlight important components of SRs, with an emphasis on the diagnostic accuracy subtype.

General Considerations

The inception of any SR starts with the research question. Performing a careful literature review ensures no similar studies have been conducted. Furthermore, it is essential to prepare a protocol for the study a priori, which can be published online using resources, such as Prospero (https://www.crd.york.ac.uk/prospero/) or the Open Science Framework (https://osf.io).

Title

The title of the SR should explicitly state ‘systematic review’ and ‘meta-analysis’ (if applicable). Of note, meta-analysis and systematic review are often but incorrectly used interchangeably. The subtype, such as diagnostic test accuracy, can also be included. The title should identify the population(s), intervention(s) and outcome(s) of interest.

Abstract

The abstract is a summation of your hard work. To a reviewer and the general readership, it is also the first part of the manuscript they read. ³ Consequently, it is paramount to ensure your abstract is polished, as it could influence the acceptance of your manuscript and readership interest. The abstract should have a clearly stated Purpose/Objective. The Methods should succinctly describe the inclusion/exclusion criteria, literature search, screening and data extraction process, risk of bias assessment and statistical methods. The protocol may be referenced in this section. The Results should summarize the number of studies and observations/individuals included, basic demographic data, risk of bias assessment and statistical analysis. The Conclusion should provide a broad statement of your findings. The sections mentioned in the Abstract are discussed in more detail below.

Introduction

The Introduction of any manuscript can be challenging to write. In its essence, it should justify the reason for your study to the reader. A succinct but informative background on your topic should be presented to highlight what is established, as well as the gaps, in the literature. This background should support your Purpose/Objective, including how your topic can address gaps in the literature.

Methods

The Methods section should paint a picture of how the study was conducted in a manner that would allow another researcher to replicate your study. The appropriate methodologic guidance protocols and reporting practices should be followed. For example, the Cochrane Handbook and the Preferred Reporting Items for a SR and Meta-analysis (PRISMA) of Diagnostic Test Accuracy Studies may be utilized.^4,5 It is essential to register a study protocol a priori for your SR, using resources like Prospero or the Open Science Framework. In general, ethics approval for SRs is not required, as all the data are available in the public domain.

The population, intervention, comparison/control and outcomes (PICO) relevant to the research question should be outlined, as well as the inclusion and exclusion criteria. Details of the literature search should be provided, including the time frame, databases searched (including grey literature, with at least two electronic sources, such as Medline, Central, Embase, and Cochrane databases) and the search strategies used (Keywords and/or MeSH index terms). Ideally, the literature search should be conducted with the assistance of a librarian. The title and abstract (phase 1), as well as full-text (phase 2) screening process should be presented, including the use of a pilot phase, duplicate screening and resolution process for indeterminate cases. For the data extraction of included studies, the data metrics being collected must be clearly listed, and the use of two independent reviewers for study selection is important to avoid potential selection or interpretation bias of a single reviewer. Of note, it is recommended to indicate the experience of the two reviewers in the topic and how discordances were resolved. Each included study in a SR should undergo a risk of bias assessment, for example, using the QUality Assessment of Diagnostic Accuracy Studies (QUADAS) tool. ⁶ The ‘high risk’ and ‘low risk’ criteria for risk of bias tailored to the research question can be outlined for each category. For the data synthesis and statistical analysis, the assistance of an experienced statistician must be obtained. Forest plots and summary receiver operating characteristic curves, where applicable, can be synthesized. For diagnostic accuracy SRs, bivariate random effects meta-analysis is generally performed. ⁷ Meta-regression and/or subgroup analysis can be conducted to compare diagnostic tests and assess for sources of heterogeneity. Statistical programs, such as R (‘mada’ package) and STATA (‘midas’ and ‘metandi’ packages), can be used for analysis.

Results

The Results section should clearly report what has been outlined in the Methods. A summary of the literature search and the screening phases should be provided, including a study flow diagram as your ‘Figure 1’. Common reasons for exclusion of studies at screening can be listed in the text and/or study flow diagram. A breakdown of the final number of included studies and cumulative number of observations/patients should be described, as well as the demographic features of the included studies in your ‘Table 1’. The findings of the risk of bias assessment must then be presented, including the most common reasons for a high risk of bias in each category. The risk of bias for each study by category can be presented in a Table or Figure. The results of the data synthesis, including meta-analysis, meta-regression and/or subgroup analysis, can then be reviewed in an organized manner, following a similar approach to the Methods.

Discussion

The Discussion section of any manuscript can also be challenging. It is important to provide insightful points while being concise. The approach described is just one of many potentially successful options. In the first paragraph, a summary of the study’s intent and important findings is provided. This is followed by 1–2 paragraphs which compare the study findings to the literature, highlighting the novel findings of this study and relevant future directions. Subsequently, a single paragraph is dedicated to the ‘real world’ or clinical practice implications of the study. The limitations section follows as the penultimate paragraph, as every study has room for improvement. Finally, a conclusion paragraph may be a part of the Discussion or a separate section, depending on journal preference.

Closing Remarks

Designing and conducting a systematic review must be performed according to appropriate methodological standards. We hope this basic framework for writing a successful SR for a Radiology journal will serve as a helpful guide for new researchers.