What is a systematic review?
The aim of a systematic review is to identify, evaluate and synthesise all the available empirical evidence that meets pre-defined eligibility criteria relating to a specific research question to provide informative and evidence-based answers. Many systematic reviews include meta-analyses. A meta-analysis is a statistical method, summarising the findings of independent studies.
Why do a systematic review?
The purpose of a systematic review in the context of healthcare is to provide the best available and up-to-date evidence on interventions for use by consumers, health professionals and policy makers to inform healthcare decisions.
In this section, you should start by describing the condition to be addressed with reference to its biology, diagnosis, prognosis and prevalence or incidence in the context of public health importance.
Next, all the current groups of interventions that are available to prevent, treat or manage the chosen condition should be critically discussed. You should also highlight the interventions that can be used in combination and how they would work in combination if appropriate. The relative importance of the different interventions in current healthcare practice should be discussed as well as identifying the intervention(s) to be reviewed systematically.
Finally, you should explain theoretically how intervention(s) being reviewed might produce better clinical outcome(s) with reference to up-to-date literature. The explanation should focus on the underlying biology. You should also make clear why the review is worthwhile doing or why it is needed with reference to any perceived improved clinical outcome(s) or to providing information, ensuring that a particular audience make evidence-based healthcare decisions.
This should be a precise statement of the primary objective of the systematic review in one sentence. This might be followed by a series of secondary objectives relating in particular to a participant group(s). There is no need to state a specific hypothesis but the objective should be phrased as a research question if appropriate.
In this section it can be helpful to think about what patient outcome data you are going to extract from your studies. Is it your objective to examine response to treatment, patient survival, quality of life or all of these factors?
Quantity of research available
For full-text excluded from your review, you should produce a table, which includes author and year, study design, other relevant information and reasons for exclusion. This table could be placed in the appendix of your review. For publications to be included in your review, you need to assess the quality of their research.
How many papers should I have?
Some research questions that could be addressed by systematic review could have hundreds of papers on the topic, others may have one or two. Therefore in the early stages of planning your review topic you need to get a sense from searching the literature on the topic how many relevant studies there are that you would need to include in your systematic review.
For an undergraduate dissertation we suggest students aim to design their research question and search strategy such that they have 5-15 studies in their review.
More studies can of course be included but you need adequate time to review them all, as well as space in your word count to discuss the diversity of your findings. Fewer studies will mean it is less likely to make meaningful conclusions and perform any kind of statistical analysis.
Note: You might end up with fewer studies than you thought once you have obtained the full text of all the papers. Some studies that look like they meet the eligibility criteria from the abstract but are not suitable once read in detail. How to narrow a search when you have too many studies:
Only look at a very specific intervention: Only one drug or way of administering that drug Look at a more specific population of patients: Only look at one type of cancer or heart disease, look at a group of patients that have failed a previous kind of treatment.
How to broaden a search when you have too few studies: Look at a broader range of interventions: Look at a class of drugs or a group of similar drugs
Look at several groups of patients: Look at a second kind of cancer or a range of heart diseases.
Don’t limit your search to any specific country or region
You may need to carry out a meta-analysis to estimate the overall effect of an intervention. You can use this statistical method to find out the overall treatment/size effect for a minimum of two studies. However, if you reviewed only two studies, you would need to discuss the limitations of having only two studies. If you have only identified one suitable study for your review, you must simply describe the results or redefine your research question.
You should only use this statistical method if the following are same in all studies:
- the types of participants;
- intervention and comparator;
- primary outcomes over the same timeframe;
- size effects/treatment effects are generally in the same direction and the confidence intervals overlap.
This plot can be generated using a statistical software package. On the next page is an example of a forest plot (See Figure 3).
If the above criteria are not met, you should not do a meta-analysis but present the data narratively. You could do this as follows:
- Summarise the conclusion of the paper
- Do a vote-call of all available results
- Plot all available data with publication against effect size with precision
- Weighted average of effect size if possible
You should try to avoid doing a narrative analysis because they highly subjective and are more likely to be opened to selection and interpretation bias. If you opt for this approach, you should try to plot effect size or determine the weighted average at least, if possible. For further guidance on how to present and write up narrative review, please consult your supervisor.