Your literature search will fail audit. Here’s why.

Written by HATEM RABEH, MD, MSc Ing

Your Clinical Evaluation Expert And Partner

in
S

Most deficiencies I see in literature searches don’t come from laziness or lack of effort. They come from a misunderstanding of what the Notified Body is actually assessing.

Reviewers don’t just check if you found literature. They check if your process could reliably identify all relevant literature. That’s a different standard. And it’s the one that determines whether your clinical evaluation survives scrutiny.

The Real Standard: Demonstrable Reproducibility

When a Notified Body reviews your literature search, they’re not judging whether you found enough studies. They’re judging whether someone else, following your documented process, would arrive at the same results.

This is what MDR Annex XIV Part A, Section 1 means when it requires a “comprehensive review of relevant scientific literature.” Comprehensive doesn’t mean exhaustive. It means systematic and defensible.

If your search strategy can’t be reproduced, it can’t be trusted. And if it can’t be trusted, your entire clinical evaluation rests on uncertain evidence.

Where Most Search Strategies Break Down

I see three recurring failure points in literature searches that reach my desk for review.

1. The Search String Looks Impressive But Lacks Logic

Manufacturers often present complex Boolean strings with multiple operators, wildcards, and nested terms. They look sophisticated. But when you ask why specific terms were included or excluded, the answer is usually: “We wanted to be thorough.”

That’s not a strategy. That’s guessing.

Every term in your search string must be justified. Why this synonym? Why this truncation? Why this combination? If you can’t explain the logic, the reviewer won’t accept the results.

Here’s what goes wrong: Teams build search strings by brainstorming keywords. They add more and more terms to “catch everything.” But they never test whether those terms actually improve sensitivity without destroying precision.

The result is noise. Thousands of irrelevant hits. And worse—no clear rationale for why the search was constructed that way.

2. Database Selection Isn’t Justified

PubMed is almost always included. Embase often appears. Sometimes Cochrane. But why those databases? And more importantly—why not others?

The choice of databases must match the clinical question and the type of evidence you need. If you’re evaluating a surgical device, excluding engineering databases might miss critical safety data. If you’re assessing clinical performance, ignoring gray literature might miss post-market surveillance reports.

Notified Bodies look for this gap. They ask: “Did you consider clinical trial registries? Did you check regulatory databases for adverse events? Did you search the manufacturers’ websites for instructions for use or safety communications?”

If the answer is no, they ask why not. And “we didn’t think of it” isn’t acceptable.

3. The Selection Process Isn’t Documented

This is the most common breakdown. The search returns 800 articles. The final report includes 23. What happened to the other 777?

If you can’t show the path from search results to included studies, the reviewer has no way to assess whether your selection was appropriate. They have to assume it was arbitrary.

I’ve seen clinical evaluation reports with detailed search strings and no record of how studies were screened. No flowchart. No inclusion/exclusion log. Just a jump from search to final list.

That’s an automatic deficiency.

What Notified Bodies Actually Look For

When I train teams on literature search documentation, I tell them to think like an auditor. What would make you trust someone else’s search?

Here’s what reviewers check:

Explicit Scope Definition

Before the search even starts, you need to define what you’re looking for. Not just the device. The clinical questions.

What safety outcomes matter for this device? What performance endpoints are relevant? What patient populations are in scope? What comparators are appropriate?

These questions define your search boundaries. If they’re not documented upfront, your search strategy has no anchor.

MDCG 2020-6 on sufficient clinical evidence reinforces this. The literature review must address specific clinical questions tied to the device’s intended purpose and risk profile. Generic searches don’t meet that standard.

Transparent Search Execution

The reviewer needs to see exactly what you did. That means:

• Full search strings for each database
• Date ranges and filters applied
• Number of results from each source
• Date the search was performed
• Name of the person who performed it

This isn’t bureaucracy. It’s proof that the search happened as described. Without this documentation, there’s no evidence the search was conducted at all.

Justified Selection Criteria

Your inclusion and exclusion criteria must connect to the clinical questions you defined. If you exclude animal studies, explain why human data is necessary for your evaluation. If you exclude studies older than 10 years, justify that cutoff based on the evolution of the technology or clinical practice.

Every filter must have a reason. And that reason must be documented before you apply the filter, not after.

Reviewers can tell when criteria were reverse-engineered to justify what you already decided to include.

A Clear Selection Flowchart

The PRISMA flow diagram was developed for systematic reviews, but it’s become the de facto standard for demonstrating literature selection in clinical evaluations.

It shows: total records identified, duplicates removed, records screened, articles excluded with reasons, full texts assessed, and final studies included.

If your clinical evaluation report doesn’t have this diagram, or something equivalent, expect a deficiency notice.

Why? Because without it, the Notified Body has no visibility into how you filtered the evidence. They can’t assess whether relevant studies were inappropriately excluded. And they can’t verify that your process was systematic.

The Question That Exposes Weak Searches

Here’s the question I ask when I review a literature search strategy:

“If a landmark study was published one day after your search date, how would you know?”

If the answer is: “We wouldn’t, but we did the search on this date and that’s what we found,” the strategy is incomplete.

A defensible literature search doesn’t end with a one-time query. It includes a mechanism for updates. That’s what MDCG 2020-6 and MDCG 2020-13 on PMCF emphasize. Clinical evaluation is a continuous process. The literature landscape changes.

Your search strategy must account for that. Notified Bodies expect to see how you’ll monitor for new evidence. If it’s not in the plan, they’ll flag it.

The Difference Between a Search and a Strategy

A search is an event. You run a query. You get results.

A strategy is a system. It’s reproducible, documented, justified, and updatable.

Most manufacturers submit searches. Notified Bodies require strategies.

The gap between those two things is where deficiencies appear.

What a Defensible Strategy Looks Like in Practice

Here’s what I see in clinical evaluation reports that pass Notified Body review without literature search deficiencies:

A clear protocol written before the search. It defines the clinical questions, the scope, the databases, the search terms, and the selection criteria. It’s signed and dated.

Search strings with explanatory notes. Each term is justified. Synonyms are listed with rationale. Truncations are explained. Combinations are tied to specific clinical questions.

A multi-database approach with reasoning. PubMed for clinical evidence. Embase for European studies. Engineering databases for device-specific technical performance. Regulatory databases for safety signals. Each choice is defended.

A documented two-stage screening process. Title and abstract screening by two independent reviewers. Disagreements resolved by discussion. Full-text review with reasons for exclusion recorded for every rejected article.

A PRISMA-style flowchart showing the entire selection path. Numbers at each stage. Reasons for exclusion categorized and counted.

A plan for ongoing literature monitoring. Defined intervals. Trigger events for ad-hoc searches. Responsibility assigned.

That’s not theoretical. That’s what survives audit.

Why This Matters More Than Most Teams Realize

The literature search isn’t just one section of the clinical evaluation report. It’s the foundation.

If the Notified Body doesn’t trust your literature search, they can’t trust your appraisal of the evidence. If they can’t trust your appraisal, they can’t accept your conclusions about safety and performance.

A weak literature search doesn’t just generate a deficiency. It undermines the entire clinical evaluation.

And here’s what I see happen next: The manufacturer responds by adding more databases or more search terms. But they don’t fix the underlying problem—the lack of a documented, reproducible, justified process.

So the deficiency persists. Another round of questions. Another delay.

The frustration I hear from regulatory teams is real. But the solution isn’t more effort. It’s more structure.

A Final Thought on Reproducibility

When Notified Bodies assess your literature search, they’re not trying to catch you out. They’re assessing risk.

If your search process is unclear, they can’t be confident that you identified all relevant safety and performance data. That’s a risk to patients. And it’s a risk they won’t accept.

The standard isn’t perfection. It’s demonstrable rigor.

If you can show that your search was planned, executed, and documented in a way that another expert could follow and validate, you’ll pass scrutiny.

If you can’t, you won’t. No matter how many studies you found.

The manufacturers who get this right don’t treat the literature search as a task to complete. They treat it as a process to defend.

That’s the difference between a file that stalls in review and one that moves forward.

Peace,
Hatem
Clinical Evaluation Expert for Medical Devices
Follow me for more insights and practical advice.