The two terms are often used as if they were interchangeable. They are related but not the same, and the distinction affects how a manuscript is designed and reported.
Systematic review: a method for finding and appraising evidence
A systematic review follows a documented protocol: a defined question, explicit eligibility criteria, a reproducible search across named databases, screening of records, structured data extraction and a synthesis. Reporting guidance such as PRISMA 2020 describes what should be disclosed so that readers can judge and repeat the work.
Meta-analysis: a statistical technique for combining results
A meta-analysis pools quantitative results from studies that measured comparable outcomes, producing a combined estimate with a confidence interval and an assessment of heterogeneity. It is appropriate only when pooling is statistically and physically meaningful.
| Aspect | Systematic review | Meta-analysis |
|---|---|---|
| Nature | Methodology for identifying and synthesising studies | Statistical pooling of results |
| Requires numeric pooling? | No | Yes |
| Typical output | Structured evidence summary | Pooled effect size with heterogeneity assessment |
| Relationship | A meta-analysis often forms part of a systematic review, but a systematic review need not include one. | |
Why engineering studies are tricky to pool
Engineering experiments differ in specimen size, test rate, environment and measurement protocol, so results may not be comparable. When they are not, a narrative or tabular synthesis is the honest choice.
Practical advice
Decide early whether pooling is feasible, write the protocol before searching, and record every screening decision. These habits make the manuscript easier to write and defend.
Further detail
Two different things that are often confused
A systematic review is a method for finding, selecting, appraising and summarising all relevant studies on a defined question, using a transparent, reproducible procedure. A meta-analysis is a statistical technique for combining quantitative results from multiple studies into a pooled estimate. A systematic review may include a meta-analysis, but it does not have to, and a meta-analysis done without a systematic review of the evidence risks being biased by selective inclusion.
What makes a review systematic
Key features are a focused question, a pre-specified protocol, a comprehensive and documented search, explicit eligibility criteria, screening of records by more than one person where feasible, structured data extraction, assessment of quality or risk of bias and a synthesis that considers study quality. The process is reported in a way that others can follow and repeat.
A narrative review by an expert may be valuable, but it is not systematic unless it follows these steps.
When a meta-analysis is appropriate
Meta-analysis is appropriate when studies address the same question, measure comparable outcomes and are similar enough in design and population (or materials and conditions) that pooling makes sense. It increases precision and can explore why results differ. It is inappropriate when studies are too different, when data are missing or when quality is poor, in which case a narrative or tabular synthesis is better.
In engineering, differences in materials, conditions and measurement methods are common, so many reviews do not pool.
Effect measures and models
Common effect measures are standardised mean differences, ratios of means, correlation coefficients, proportions and hazard ratios, chosen according to the outcome. Fixed-effect models assume one true effect; random-effects models assume a distribution of effects across studies and are usually more appropriate with heterogeneity. Weights depend on study variance.
Report the model chosen and why, and show individual study results in a forest plot.
Heterogeneity, bias and sensitivity
Heterogeneity is variation between studies beyond chance. Statistics such as I squared and tau squared quantify it, and subgroup analyses or meta-regression explore sources, with caution when there are few studies. Publication bias, where studies with positive results are more likely to be published, is assessed with funnel plots and related tests when enough studies exist. Sensitivity analyses test how results change when studies are removed or assumptions altered.
Interpretation should reflect uncertainty, not only the pooled estimate.
Scoping, mapping and other review types
Not every question needs a systematic review. A scoping review maps the extent and nature of evidence and identifies gaps. A mapping review categorises studies. An integrative review combines diverse sources. An umbrella review synthesises existing reviews. A realist review examines mechanisms and context. Choose the type that fits the question and resources.
Naming the type correctly avoids reviewer criticism.
Reporting guidelines and registration
Follow recognised reporting guidelines for the type of review, including a flow diagram of records. Registering the protocol where a suitable registry exists, or publishing it, supports transparency. Share extraction forms and analysis scripts.
Report deviations from the protocol and the reasons.
Common mistakes
Typical mistakes are a vague question, non-reproducible search, no quality appraisal, pooling incomparable studies, counting studies as votes, ignoring heterogeneity, selective reporting and conclusions beyond the evidence. Another is calling a review systematic without a protocol.
Honest labelling and transparent methods are better than overstated rigour.
A short decision guide
Do you have a focused question and enough comparable studies? If yes and you have resources, consider a systematic review, with meta-analysis if appropriate. If the field is broad or new, consider a scoping review. If you want to give an expert overview, write a narrative review, with transparent methods. If you want to understand the structure of a field quantitatively, consider bibliometric analysis.
Our manuscript services cover each type, from protocols to writing.
How we can help
We support protocol design, search strategy, screening workflows, extraction forms, synthesis and manuscript writing. The conduct and conclusions remain the authors' responsibility, and we cannot guarantee publication.
An illustrative example in an engineering context (hypothetical)
Suppose researchers ask: does adding a given type of fibre improve the compressive strength of concrete, and by how much? A systematic review would define the population (concrete mixes), the intervention (fibre type), the comparison (mixes without the fibre) and the outcome (compressive strength at 28 days). They would search databases, screen records, extract data and appraise quality. If enough studies report mean strength, standard deviation and sample size under comparable conditions, a meta-analysis could pool the ratio of means, with subgroup analysis by fibre content. If studies differ too much, they would present a structured narrative synthesis.
This example is hypothetical and does not report real findings; it shows how the questions, steps and choices fit together.
Quality appraisal for engineering studies
Medical tools for risk of bias do not map directly to engineering studies. Reviewers adapt criteria: whether specimens were randomised or matched, whether testing followed standards, whether sample sizes and variability were reported, whether control groups existed, whether measurement methods were validated and whether conflicts of interest were declared. Document the adapted tool and its rationale, and test it on a few studies first.
Appraisal should inform the synthesis, for example through sensitivity analyses that exclude low-quality studies.
Frequently asked questions
Do I need two reviewers? It is recommended, particularly for screening and extraction. How many studies are needed for a meta-analysis? Two can be combined mathematically, but few studies give imprecise estimates of heterogeneity; interpret cautiously. Can I include grey literature? Yes, if justified and documented. Can I combine different materials? Only with caution and clear justification.
Is a systematic review publishable in engineering journals? Many journals accept them, especially if the question is relevant and the method is rigorous.
Planning resources and a timeline
A rough plan for a systematic review: protocol development, search and deduplication, title and abstract screening, full-text screening, extraction and appraisal, synthesis and writing. The screening stages dominate when records number in the thousands. Two screeners working part-time can take months. Software for screening helps but does not replace judgement.
A rapid or scoping review may be a better fit when time is limited, as long as the method is transparent and the claims are modest.
Reporting the synthesis honestly
Present the characteristics of included studies in a table, the quality appraisal results, the main findings by outcome and the overall strength of evidence. State where evidence is absent or weak. Avoid claiming a pooled effect is the truth; present it as the best estimate under the review's assumptions, with uncertainty.
Offer implications for practice and research that follow from the evidence, and note that new studies may change conclusions.
Next steps
Write your question in a structured form, decide whether studies are likely to be comparable enough to pool, run a pilot search to estimate the number of records, and draft a protocol. If you are not sure which review type fits, describe your topic through the contact page and we will discuss options without commitment.
Common questions about authorship and ethics in reviews
Reviews are secondary research, but they carry ethical responsibilities: accurate representation of the included studies, transparent methods, declaration of conflicts and respect for copyright when reproducing figures. Contributions such as screening and extraction merit authorship if they meet journal criteria. If a review draws on unpublished data from authors, obtain permission.
Be careful with text and figure reuse; paraphrase and cite, and request permission for reproduced material. Generative tools used in screening or drafting should be disclosed according to policy and verified by a person.
Ask for an independent check of your protocol
Before starting, have a colleague or information specialist read the protocol and search strategy. Small changes to inclusion criteria or search terms can change the evidence base substantially. An independent check is inexpensive compared with repeating screening.
Summary
A systematic review is a transparent method for finding and appraising evidence; a meta-analysis is a statistical method for pooling comparable results. Use the first when the question is focused and the second only when studies are comparable. Report honestly, including heterogeneity and limits.
Choose the review type that fits the question and the resources available.