Literature monitoring is an essential part of medical research and development of new and improved treatment methods, including medicinal products and medical devices. Review and analysis of literature can be used to replace or supplement non-clinical or clinical testing and literature references are part of legally required structure of study reports.

The legislation also suggests that monitoring of scientific literature is an important part of pharmacovigilance activities. This is further clarified in the good pharmacovigilance practice, where it is stated that the expectation is that weekly literature monitoring continues throughout the life-cycle of the product.

The two-sided nature of literature screening

Medical literature is one source of potential new safety related information. An article can contain a new benefit-risk assessment or the article can contain a case report where a medicinal product or medical device has caused an adverse event to the patient.

Indispensable research

In general, scientific literature is one of the most important ways to communicate new hypotheses,and so it is easy to see why it is such a good source to look for new safety information.

And because literature review is such an important part of research, there are very good tools available for performing the activity. There are huge databases aggregating articles directly from scientific journals or other databases and libraries. These often offer different kinds of tools to search the data and get alerted when new articles related to specific topics are being published.

Redundant efforts

As we have written in several of our blog posts, when it comes to pharmacovigilance requirements, there is also “the other kind” of literature monitoring happening outside of these great databases and easy-to-use tools: local literature screening.

Monitoring of obscure national publications is a European legal requirement that has been enforced by the European Medicines Agency, ballooning into an industry of itself within the world of pharmacovigilance. It is an expensive industry with local language experts spread out in different countries geographically and lots of manual process steps.

We have also pointed out in our posts that this legal requirement doesn’t seem to be going away, even if it’s clear that it has little to no added value. What makes the situation worse is that there are several developing countries that see the European way of doing PV as the “right way” and have adopted the requirements in their national legislation.

This means that while it seems we are stepping into a new age of large AI-powered language models with all information easily accessible through a conversation with an all-knowing chatbot, the literature screening requirements for the global pharmacovigilance system are diving deeper into the stone age, where a lot of the industry seems to be in.

Tepsivo is here to fix that

We have developed the first truly automated solution for local literature screening, Tepsivo Literature.

Why do we say “truly” ?

There are some who claim to have automated solutions for local literature screening, while their solution is actually mainly human based with people scanning journals and uploading text to a database. While in certain world regions this might be a cheaper solution and faster to set up than building real automation, there are reasons why this is an inferior approach. Continue reading to learn more.

In the pharmacovigilance provider industry, literature monitoring is usually split into two separate services: global literature monitoring and local literature screening.

Global literature monitoring refers to the screening of widely used reference databases for biomedical literature, for example PubMed (MEDLINE).

Local literature screening refers to sources that are not indexed in the widely used reference database used for global literature monitoring and this group of sources includes a mixed bunch of local journals, websites and newsletters, traditionally requiring a lot of people involved for language knowledge and time effort spent on these additional sources.

With Tepsivo Literature’s automated literature monitoring, this division is history. Tepsivo Literature offers better coverage than any other system on the market and offers the same easy and unified experience for both the global and local literature screening.

How do we do it?

The concept is fairly simple, actually. The global literature monitoring, though not review, is already automated by the large databases such as PubMed. You just need to subscribe to the keyword strings that you wish to receive or access the databases directly programmatically.

The local literature is traditionally the trickier part because of the variety of the sources, poor availability of the texts, and the large number of different languages involved. We’ve analyzed tens of thousands of sources globally and defined what the best way to access the article information is.

Most often, the journals provide the article information to a local library or database, and the articles can be accessed through data aggregators. In other cases the article information is accessed directly through the source. We have solved the language issue using a combination of the latest translation technology, ATC Code based terminology and advanced search technology.

Why is automated the way to go?

Cost

Automated screening and searching is cheaper than manual; much cheaper. With Tepsivo Literature, since the process of accessing the articles and searching them with the keywords is fully automated, the only cost is related to running the software.

The traditional local literature screening involves local language experts who define a list of local journals, read and search the articles weekly, provide weekly results to a central team, the central team reviews the reports, creates a summary report which is attached to a monthly report. And none of these highly trained professionals work for free. You can do the math of what it can cost you running this in 30 European countries as legally required for many MAHs.

Tepsivo Literature cost takes that pain away.

Accuracy and repeatability

In systematic reviews automated search has the same or better accuracy as humans when it comes to identifying relevant articles. This means identifying articles that contain one or more keywords specified for the search from a large group of articles. Tepsivo Literature is an accurate machine that will pick up the safety relevant articles every day following the same logic and rules.

When a human performs the search, on the best days they might have the same accuracy as a machine when they use the same search methods as the automation, for example database search with an agreed keyword string. In some specific situations, they can in fact even discover more relevant results than the machine-based search thanks to the real understanding of the text, for example when manually browsing through articles of an individual issue of a journal and noticing an article that discusses the topic using terms or language that would not be understandable to a machine or containing pictures or graphs that would require a built-for-purpose algorithm to be picked up by automation, but on their worst days the human might miss something completely.

When humans are involved in any process, there are always mistakes given enough time. And even if there is an exact process describing how to perform literature screening, what safety information is and how the search results should be documented, the human brain is essentially a black box. If you ask a person, or their equally qualified and trained deputy, to repeat the same search a year later, they might come up with completely different results. When a machine performs a process, it is always done in the exact same way, no matter if it’s done ten times or million times.

Of course, there can be issues with automated processes, as well. The automation is set up by a person originally, so it is possible that this person made a mistake. However, even if a mistake was made in the beginning, the following errors are accurate and repeatable. The mistakes made by humans can vary day-to-day in severity and type and they can be very difficult to notice and costly to fix.

This is why Tepsivo Literature was built by PV experts who have had their fair share of manual searches and why its functioning is human controlled and subject to continuous revalidation.

Timeliness

The saying goes that computers never sleep. This is also true when it comes to automated local literature screening. The search results are always up to date.

When people are performing local literature searches, it usually happens weekly or monthly. This is actually a sufficient frequency in practice, but there are several advantages for having the searches run continuously, and why not do it when you can?

Almost always when there are people involved, someone will be late with the delivery of their report or maybe they deliver the wrong report or some relevant information is missing. Rest assured, Tepsivo Literature is never late.

Thoroughness

Automated search really goes through everything. By everything, we mean many more sources, all articles, and all the text in all of those articles. When a local literature screening expert does the analysis, they usually select a handful of sources that they are aware of, and then based on the format of the journal, they either search it with a web search engine, CTRL + F or by skimming through the text (if they read through the articles, you will see it in the hours invoiced or work time wasted). With advanced search engines, it is possible to search large quantities of text with several different variations of the keywords in a blink of an eye, and that’s exactly what Tepsivo Literature achieves.

Centralization

With automated local literature screening, everything is centralized to one place. The articles and their metadata is in one database, the search results are delivered to one place for systematic safety review. With manual local literature screening there are trackers and reports for each country. In some cases you would only see the end result: no new safety relevant articles found. There can already be dozens of people involved in these two steps (review and report delivery). Then someone still needs to put together a summary report out of these separate files that have been delivered by email or by uploading them to a folder somewhere. All that time can be easily saved by Tepsivo Literature.

Language capabilities

With fully automated local literature monitoring, it is possible to utilize the latest technologies when it comes to language. It is possible to automatically detect the language of the article and then determine whether translation to English is required or not. If translation is performed, the translated text is stored with the original text.

It is possible to search both the original and translated text with thousands of keywords in different languages to ensure that all relevant information is identified and tagged in the text. If manual literature screening is performed, the process again requires dozens of local language experts. Qualifying, training and managing these experts requires resources and time. Even then, the local language experts can also make mistakes and often they also use machine translation tools at least to support their translation work.

Think of it like this: would you rather have one expert translator who speaks 20 languages on C level who analyzes the text and can make the decision on their own, or twenty translators who speak two languages on C or B level, who all analyze the text independently and then all the 20 translators need to have a conversation together to make a decision?

Traceability

When the process is fully automated from beginning to the end, all the steps from accessing the text to final delivery with full article data and tagged keywords are being tracked to an audit log. You can view the details of every step.

With traditional local literature monitoring you just have to trust that the local person for literature screening is performing the search. Or you have to request them to take screenshots which takes as much time as the actual search (and, actually, still doesn’t prove they’ve actually read the text).

Usually, the article search result is 0 safety relevant articles, so how can you really know they performed the screening? The local expert sending the report by email or uploading it to SharePoint is the first truly auditable action. Then happens the summarization of the local reports. While the source data and the end result are available and can be reviewed in an audit, the process of compiling the report usually happens in a way that does not leave a trace. With Tepsivo Literature and Tepsivo Platform full audit trail of all actions, missing activities, lacking evidence of documentation, or even retrospectively “faking of reports”, is literally impossible.

Quick summary

Comparison table – Manual Literature Screening and Tepsivo Literature