# Clinical Evaluation Report: Literature Search - SaMD Author: Sebastian Skorka Published: 2025-11-06 Updated: 2025-11-06 Source: https://openregulatory.com/articles/doing-the-literature-review-for-clinical-evaluation-reports-for-software-as-a-medical-device --- Clinical Evaluation Report: Literature Search for Software as a Medical Device You managed to identify risks, the software requirements are defined and documented and the first usability tests have been conducted (successfully). The technical documentation is nearly complete, now off to the next exercise: the Clinical Evaluation. This document is basically a miniature bachelor thesis about your medical device. It shall demonstrate the evidence of the medical device's safety, performance, and clinical benefits. In this article we will take a closer look into the process of writing the CER (and CEP) with a specific focus on SaMDs. Regulatory References  Here I summarize the regulatory requirements and references. I recommend reading them first or in parallel. Article 61 - Clinical Evaluation The Article introduces the principle of the clinical evaluation and some of the basic requirements. Annex XIV - Clinical Evaluation (Plan and Report) The most relevant article in the MDR about clinical evaluation. It introduces the CE Plan and highlights a rough chapter structure. MDCG 2020-1 Guidance on clinical evaluation (MDR) / Performance evaluation (IVDR) of medical device software This guidance provides helpful information on the concepts that apply to SaMDs only. In addition, the guidance explains some concepts with example devices.  MDCG 2020-13 Clinical evaluation assessment report template Do you want to see your CER through the eyes of an auditor? Take a look at this MDCG document. Basically, it is a checklist the auditors follow. Use it as a  rough structure for the CEP and the CER. In addition, you will understand the most important points you need to cover in your CER.  Literature Review: Step by Step Process I will use a simple example to demonstrate the steps and make it easier to follow. A simple mobile phone app to prevent back pain will do the job. You enter your symptoms and the app suggests some exercises. Patients follow the videos and can track symptoms or pain. Nothing special. Define the Scope! At first you define the scope of the literature review based on the intended use of your SaMD.  The MDR wants you to identify relevant literature about the state of the art, the performance of your device, risks, clinical benefits and similar/equivalent devices. Let´s go through the point individually and connect them with our example. The state of the art - The background idea is that you need to show that your software and the underlying technology is well established and not outdated. You will need to create a separate chapter in the CER and you can consider it to be something like an introduction to your SaMD.  Example: This means that you can take some basic text book chapters about the benefits of physiotherapy and the respective exercises. There is clearly a scientific link between an exercise program and the reduction of back pain. You do not necessarily need to set up a search term for this. Performance & clinical benefit - The SaMD should have a clinical benefit and clinical performance. You can use literature data to support the clinical performance of the SW. However, this is mostly useful when your SW is already on the market and mentioned in publications. Read the MDCG 2020-1 for more information regarding the clinical performance of your SaMD. In most cases, the usability testing is sufficient! Risks - Of course, “with great power comes great…” The idea is to break down your technology and search for unknown risks that might be associated with your MD. Regulators want you to search publications for stuff (risks) that other people screw up and you can learn from (input to your RM). Example: Search the publication databases for risks, side effects, incidents etc. that might be associated with self-guided video based therapy. To be honest, in my experience this rarely brings new discoveries to the surface… Similar/equivalent devices - The MDR allows you to use (clinical performance) data from other manufacturers. However, only if you have access to the technical documentation (see MDCG 2020-5). No access? No equivalence. “But there are like 20 others app out there, is it impossible to use this information?”  Gladly, you can (and have to) identify similar devices. As a rule of thumb: Take 3 to 5 similar devices, run their names through the databases and screen the results for new risks. You tell me to search databases. Which databases? The MDR and the MDCG 2020-13 require you to identify relevant scientific databases and recall databases. Let's look at the scientific databases first. There are PubMed, Embase, Google Scholar and Cochrane (in the order of importance hehe). I personally think that one database is enough - the MDCG 2020-13 has a different opinion on this and tells you to search two databases… I recommend PubMed and Embase. You can spice it up by including cochrane. I never identified any useful publication in Cochrane but the search is done quickly and you can tell auditors that you put some effort into it. I do not like Google Scholar because the search filters are not very helpful and the overall screening of the abstracts is a bit of a pain. Personal taste I guess.  The second kind of database covers recalls, incidents etc. Of course there is no module in EUDAMED yet, so you need to rely on the German BfArm, the British MHRA or the FDA recall database. Take the similar devices and run them through these databases. The idea is again to identify new risks that you have overlooked in your risk assessment. Ok, cool. How do I identify the publications in PubMed/Embase? Based on the search categories (see above) you will create one or more search terms. These are structured key words to find the information in the millions of publications.  Let us look at an example. We want to find new risks associated with the mobile phone app. A potential search term could be:  [“Risk” OR “side effect” or “incident”] AND “physiotherapy” AND “back pain”  By combining the search words with boolean operators you can shape (to your advantage) the scope of the search and the number of results. The same principle applies to the Embase search. Identify the relevant publications and apply some filters. You are free to use filters in your literature search and the most common filters are the publication date (the last 5 years are sufficient) or the study design (review articles for the state of the art). It highly depends on the device and the search category. Potentially relevant vs. relevant articles You are nearly done! The only thing to do is to read all abstracts and exclude all non relevant articles from the list. Subsequently, you read the full texts and decide again if this publication is relevant for your CER or not. Provide always a justification why you exclude an article. Now you end up with the final list of relevant articles that find their way into the clinical evaluation. You can use them in the respective chapters state of the art, risks, benefits and similar devices.