How to find a novel, valid and impactful meta-analysis topic

Most unhelpful reviews are not badly conducted. They answer a question that did not need answering: someone already did it, or there are too few studies to say anything, or the answer would not change what anyone does. The months spent on the review are the real cost.

A topic worth the effort passes three tests at once. It is novel (nobody has answered this question, or what exists is out of date or unreliable). It is valid (enough comparable studies of adequate quality exist to answer it). And it is impactful (it adds to the literature and answers a question that matters).

Why this is harder than it looks

The evidence base grows faster than reviews can keep up with, which is why gaps exist. A well-known 2010 paper by Bastian and colleagues put the problem in its title: seventy-five trials and eleven systematic reviews a day. At the same time, many topics attract several overlapping reviews, and Ioannidis (2016) argued that a large share of systematic reviews and meta-analyses are redundant or misleading.

So both problems are real: important questions without a review, and popular questions with too many. Browsing for ideas will not tell you which you are looking at. You need a method that checks each idea.

The five steps

Start from a question, not a drug. Write a specific question as population, intervention, comparator and outcome. If you cannot name all four, the question is not ready.
Apply the novelty test. Search for existing reviews and registered protocols that answer the same question. Decide whether to skip, update, extend or proceed. The guide to checking whether a review already exists gives the search steps.
Apply the validity test. Count the primary studies, check that they measure the same outcome in a comparable way, and look at their quality. Without this, a novel topic may be novel only because the studies do not exist.
Apply the impact test. Ask who would use the answer and what they would do differently. A review is impactful when it adds to the literature and answers an important question.
Decide, register and write a protocol before you start. Registering on PROSPERO or a similar register also lets others see that the topic is taken.

Let SynthGap run the first pass

Enter an area. SynthGap proposes candidate topics, checks each against existing systematic reviews, and lists the primary studies behind it. The free plan includes 3 discovery runs a month.

Start free

A one-page checklist

Can I state the question as population, intervention, comparator and outcome?
Have I searched PubMed, the Cochrane Library, PROSPERO and Epistemonikos for the same question, and recorded the date?
If a review exists, is it out of date, narrow, or at high risk of bias, and do I have a clear reason mine adds something?
Are there at least five comparable primary studies, and do they report the outcome in a form I can pool?
Are the studies good enough that a pooled result would mean something?
Would a clinician, guideline group or patient group use the answer?
Do I have the time and skills for the full review, including screening and data extraction by two people?

How the three tests relate

Topics usually fail one test while passing another. A question can be novel because nobody studies it, which makes it invalid. It can be valid and impactful but already reviewed three times, which makes it not novel. It can be novel and valid but of little interest to anyone, which makes it low impact.

That is why the tests need to be applied in order and every time. A topic that looks promising after one test is only a candidate.

Where SynthGap fits

SynthGap automates the first pass. It searches the published literature, checks candidate topics against existing systematic reviews, and proposes topics with the primary studies behind them. For the topics you choose, the deep analysis captures the related studies, selects the final included studies, runs a small meta-analysis and reports whether the topic looks worth a full review.

Its suggestions are AI-generated, so treat them as candidates to verify, not as conclusions.

Questions

How many primary studies do I need for a meta-analysis?

Two studies can be pooled technically, but with fewer than about five the estimate of heterogeneity is unreliable, and a funnel plot to look for publication bias is generally not recommended with fewer than ten. Plan for more than the minimum.

Can I do a review on a topic that already has one?

Yes, if you can say why yours adds something: substantial new evidence since the last search, a different population or outcome, or a more reliable method. Repeating the same question with the same studies adds nothing.

Is novelty enough?

No. A topic can be novel only because no studies exist, or because nobody needs the answer. Novelty, validity and impact all have to hold.

Keep reading

General methodological guidance, not medical advice. SynthGap's suggestions are AI-generated from the published literature, so verify them before registering a protocol.