EU MDR literature reviews – best practice for database selection and search strategies

The EU’s Medical Device Regulation (MDR) 2017/745 raises the bar for device manufacturers looking to stay compliant. With its introduction into law now postponed to May 2021, device manufacturers have more time to get their literature reviews for Clinical Evaluation Reports (CERs) up to the standard required. Best practice for EU MDR literature reviews means using multiple databases and ensuring search strategies are as refined and comprehensive as possible. We discuss this change in more detail in our post EU Medical Device Regulation: What does it mean for manufacturers?

In this post, we look at best practice for EU MDR literature reviews (including the MEDDEV 2.7/1 Revision 4 guidelines) and discuss how broad database choice and detailed search strategies can be easily combined to create a best-practice Clinical Literature Search for a CER.


The value of a broad set of sources

MDR 2017/745 requires that at least two databases are searched. As the nature of devices means that references to them may extend beyond medical and pharmaceutical literature, there is value in extending any search to the broadest possible range of sources, for example engineering or other databases.

The Dialog strategy below (which we explain in more detail shortly) searches for adverse effects relating to excimer lasers used for eye surgery.

S1 (EMB.EXACT.EXPLODE(“excimer laser — adverse device effect”)) 35*
S2 (MESH.EXACT(“Lasers, Excimer — adverse effects”)) 166*
S3 (“excimer laser” or “ArF laser” or “argon fluoride excimer laser” or “argon fluoride laser” or “argon fluoride lasers” or “EC-5000 Quest” or “excimer laser device” or “excimer lasers” or “KrCl laser” or “KrF laser” or “krypton chloride excimer laser” or “krypton chloride laser” or “krypton fluoride excimer laser” or “krypton fluoride laser” or “XeCl laser” or “XeF excimer laser” or “XeF laser” or “xenon chloride excimer laser” or “xenon chloride laser” or “xenon fluoride laser”) n/5 (adverse or complicati* or untoward or unexpected or injur* or death or fatal or fatality or degrade or degrading or degradation or corrosiv* or corrodes) 1038*
S4 S1 or S2 or S3 1185*
S5 S4 and py(>2018) 82*

* Duplicates are removed from the search and from the result count.

The first aspect of this search to note is the number of databases, 11 in total. Beyond databases typically associated with a CER search, for example Embase and MEDLINE, this strategy additionally takes in broader sources, including Current Contents and Health Research Full Text Professional. It also adds sources with engineering content, such as EI Compendex and Inspec. The outcome of this search was 82 relevant articles, 16 more than when the same search was carried out in just Embase, MEDLINE and Biosis – illustrating the value of a broader universe of sources.

The chart shows that nearly 23% of total documents retrieved come from sources beyond just Embase, MEDLINE and Biosis.


How accurate synonyms expand the results available

This search also shows how an accurate universe of medical synonyms is crucial to retrieving all possible references relevant to a CER. Here’s a step-by-step explanation of how the search strategy above works:

S1 and S2 are searches in Embase and MEDLINE that use the thesauri in those databases to retrieve relevant references. However, the broader universe of databases that this search also covers do not offer the same thesauri. So to capture relevant articles, S3 applies a set of synonyms.

The medical device synonyms are added automatically by the Dialog Medical Synonyms feature, a curated list selected from a variety of specialist sources and taxonomies. The search then looks for adverse events and relevant synonyms within five words’ proximity to the device terms. Free-text terms related to devices such as degradation and corrosion are also added.

Bringing S1, S2 and S3 together in S4 finds 1,185 records, which a date restriction of the last two years narrows down to a manageable and relevant 82 articles.

Finally, it’s worth noting that in this example, duplicate references are removed from both the search and result count, taking away a potentially onerous deduplication task that can otherwise be necessary when searching across multiple databases.

And with synonyms clearly displayed as in this strategy example, the search strategy can be easily in line with the latest MDR requirements.

So, with a wide universe of databases and precision search that also employs an accurate and up-to-date set of synonyms for medical devices, manufacturers are better equipped to compile MDR-compliant literature searches with minimum effort.

In our next post covering the requirements for MDR 2017/745, we’ll look at best practice in the literature review process and how this can help device manufacturers comply with the upcoming regulation.