Hi. It is indeed a difficult topic. Overlap is usually a phenomenon seen in overviews of systematic reviews, but as you say, it can also happen in systematic reviews of primary studies (specially of large observational primary studies).

I am not so confident in providing a specific recommendation, but here are two references that may help you:

• El Ansari W, El-Ansari K, Lock M. Mind the Overlap! Meta-Analyses That Synthesize the Findings of Primary Studies Based on Large Data Registries: the Case of Metabolic and Bariatric Surgery. Obes Surg. 2023 Nov;33(11):3689-3691. doi: 10.1007/s11695-023-06819-5. Epub 2023 Oct 5. PMID: 37796374.
• Lock M, El Ansari W. New world of big data-new challenges for evidence synthesis: impact of data duplication on estimates generated by meta-analyses and the development of a framework for its identification and management. J Clin Epidemiol. 2024 Dec 17;179:111641. doi: 10.1016/j.jclinepi.2024.111641. Epub ahead of print. PMID: 39701398.

Dear Alireza,

The inclusion of the same sample in one meta-analysis results in duplicate publication bias and can question the validity of your findings. My personal experience in order to avoid such cases is to separate studies that use the same database and compare their inclusion criteria. The data used from these databases in the multiple publications on the subject could differ in the following ways:

1) date of inclusion (i.e. one study used registry data from 2010-2012, while another used 2014-2018).

2) demographics (i.e. one study only included older adults from the database, another focused on adolescents, etc.).

3) disease characteristics (i.e. one study focused on early stage cancer patient while another focused in late stage patients only).

4) setting (i.e. one study used hospital X and Y data, while the other only focused on hospital Z, yet they are all included in the database).

Based on these differences the samples could be entirely different, even though they have been sampled from the same database, and can be pooled together in one meta-analysis. In case the same sample has been used by multiple studies, you should choose the larger study for the meta-analysis and exclude the other study. In case you wish to keep the duplicate study in your systematic review and only exclude it from the meta-analysis, you should note this in your methods and your results to avoid confusion.

In your particular situation, you should first check whether the hospitals in PearlDiver and TrinetX are the same as the hospitals in the other studies and whether they collected data in the same year and on the same patients. If so, you can only include one of the duplicate studies and it should be the one with the larger sample size.

Also as a side note, studies using non-registry based samples may also be using a duplicate sample. Authors will typically report this in their methods section and/or discussion, and having the same author/authors on two papers should prompt a closer look into their included populations.

The first task would be to develop a clear definition of your topic and identify the types of data each article must include for you to use it in your study. For example, if your meta-analysis was looking at cause/effect, should all the studies you include be experimental? It is likely that including descriptive studies would bias your result with non-causal results.