Thanks to the experience with our customers and many other pharmaceutical companies we talk to, it seems to be a trend that pharma companies are questioning the necessity of the regulatory requirement of screening local literature (LLS) for safety information, even more so with the requirement applying to approved products, not necessarily placed on the market.
Their product may be used by a small patient population, it can be a well-established product, or it may not be on the market at all. Especially in these cases, it seems that there is very little or no new information gained from the few non-indexed local sources. And no surprise, the activity is expensive which reasonably seems disproportionate to the output.
We are a company that manages local literature screening in all EEA countries for our customers, so the topic is quite close to us, and we keep thinking of what we can do to improve it, and where the value of the activity is in general.
With that in mind, Martti, our COO and literature screening geek, has performed an analysis of some of the publicly available EV data to discover more detail. The analysis is limited, but it does show some interesting data and point to what the reality and value of LLS is.
In conclusion, while the local literature screening requirement still needs clarification and the activity should be more streamlined and less expensive, it should not be omitted altogether. Admittedly, to our own surprise, our analysis shows that non-indexed journals have been a source of a meaningful number of total safety relevant data, even if it is not a substantial source.
What is Local Literature Screening?
In Pharmacovigilance, the literature monitoring can be split into two activities: global literature monitoring and local literature screening. Global literature monitoring refers to the screening of one or two global databases for biomedical literature, for example PubMed (MEDLINE) and Embase.
Local literature screening refers to sources that are not indexed in Embase or MEDLINE. This group of sources includes mixed bunch of local journals, websites and newsletters.
The requirement for literature screening is based on European Commission directive 2001/83 Art 107 3. that states “For medicinal products containing the active substances referred to in the list of publications monitored by the Agency pursuant to Article 27 of Regulation (EC) No 726/2004, marketing authorisation holders shall not be required to report to the Eudravigilance database the suspected adverse reactions recorded in the listed medical literature, but they shall monitor all other medical literature and report any suspected adverse reactions.”[i]
In the GvP Module VI, the guidance and requirement justification for local non-indexed sources is provided as: “…marketing authorisation holders should have procedures in place to monitor scientific and medical publications in local journals in countries where medicinal products have a marketing authorisation, and to bring them to the attention of the company safety department as appropriate.”[ii]
Taking into account the substantial costs that pharmaceutical companies typically face in order to meet the above requirement and the likelihood of an inspection finding if the requirement isn’t met, it is not unfair to expect a more thorough justification and analysis.
Why is it a Problem?
The first problem with local literature monitoring is that there is no definitive list of sources that should be screened. In fact, there has been hardly any support from the EMA or the NCAs related to recognizing which sources should be screened by the MAHs.
Second, the variety of the sources and in some cases the low number of relevant sources. Large amounts of resources and time effort are currently spent by all EU MAHs on local literature monitoring because they cannot be searched in one database and the search is very difficult to automate with the current technology. On the other hand, in many of the EU countries, large amounts of resources are wasted to screen 1 or 2 sources just to fulfil a regulatory requirement and to avoid inspection findings.
The third issue would be the low amount of relevant safety information found in the local literature screening. For professionals involved in the activity, it seems that relevant new safety information is almost never found in the non-indexed sources. This is not necessarily a problem as a low amount of adverse reactions is a good thing but it does make us question the need for the requirement to be as strict as is today.
Communication with the EMA
To tackle the first and the second problems, we have started a project to modernize the local literature screening. As a first step, we have contacted EMA a few weeks ago with the information presented here to see if they have plans related to the local literature screening.
We will follow up on our blog once we get the response from EMA. This should be within 3 months of our contact. Once we have more information from EMA regarding their possible plans (or the lack thereof) related to LLS we will have a better understanding where we should move with this initiative.
Analysis of EudraVigilance Data
We wanted to have more information on the third problem: is any relevant safety information really found from the local literature screening? We performed an analysis of the EudraVigilance data for 2020 to see how many adverse events were reported based on local literature screening.
Our definition for a local literature screening case is a literature case that is from a source that is not found on the Embase indexed journal list which is publicly available here and includes journals that are indexed on Embase and/or MEDLINE.
EMA 2020 Annual Report on EudraVigilance (EV)
EMA published their latest annual EudraVigilance report[iii] on 11th March. The report includes information about the operation and development of the system and key figures from the database with data analysis.
EudraVigilance currently contains over 18.6 million ICSRs. In 2020, 1.8 million post-marketing ICSRs were collected and managed. Out of those new ICSRs, 812,760 ICSRs originated from the EEA.
Public Data from EudraVigilance at www.adrreports.eu
We downloaded ICSRs from www.adrreports.eu.[iv] The website includes line listing information directly from EudraVigilance. It does not include reports from studies (e.g. clinical trial, non-interventional study) or other types of reports. In other words, the web site includes only spontaneous reports.
We downloaded 128,522 individual spontaneous cases originating from EEA for centrally authorized products (the data is split between centrally and non-centrally authorized products). 3,867 of these cases included a literature reference. This means that approximately 3 % of spontaneous cases were literature cases.
Analysis of Literature Cases
Visual examination of these 3,867 cases quickly reveals that one journal article can lead to several individual reports. An extreme example of this is article called “Amato M, Goretti B […], Effects of 2‐year treatment with dimethyl fumarate […], Neurological Sciences” that is the source of at least 391 ICSRs. This means that last year 10 % of the ICSRs originating from literature were submitted based on this one article. There are dozens of other articles that resulted to 10-20 individual cases.
To reduce the noise created by these articles, we removed duplicate references with Excel. 2,269 cases remained after the duplicate removal but still a visibly large number of duplicate articles remained.
The remaining duplicate articles were caused by non-compliance to the reference guide. This is one of the challenges with the literature based ICSRs in EudraVigilance: there are errors in the reference or the reference is missing partially.
The non-compliance to the referencing guide creates a challenge when the sources are analysed with a script. We tried to compare the references to the Embase indexed journals list with a simple script, but the test was not successful due to the heterogenicity of the EudraVigilance reference data. The intended full analysis of all cases was not possible.
Because the automatic comparison was not possible, we decided to analyse smaller amount of cases with a manual comparison to the Embase content list[v].
Out of the 2,269 remaining cases we took two separate samples:
- 100 random cases
- 200 first cases alphabetically ordered by the reference[vi]
In the sample of 100, 5 cases did not have a recognizable source. Out of the remaining 95 cases, 23 (24%) were non-indexed. Based on this sample, 0.7 % of all cases would be originating from local literature screening. 26 % (6) of the cases from non-indexed literature were from congresses or similar events. For indexed articles the same percentage of congresses as sources 7 % (5).
In the sample of 200, 12 cases did not have a recognizable source. Out of the remaining 188 cases, 41 (22%) were non-indexed. Based on this sample, 0.8 % of all cases would be originating from local literature screening. 12 % (5) of the cases from non-indexed literature were from congresses or similar events. For indexed articles the same percentage of congresses as sources was 2 % (3).
Notes from the qualitative analysis of the samples:
- There are source articles that are old (more than 15 years old)
- Some cases include several articles from different sources as a reference
- Some countries (with highest amount of local sources) stand out: Czech Republic, France, Germany, Poland and Portugal.
- Some of the non-indexed sources that were not in Embase were non-medical journals like fashion magazines.
The answer to the main question of this blog, is ‘Yes’, local literature screening is necessary. ICSRs are found from local non-indexed sources and those can lead to relevant signals.
However, it can be questioned, how likely this is if 0.75 % of the EV cases are from local literature. As described in our previous blog, according to the 2020 EV Annual report 1.8 million post-marketing ADR reports lead to 81 validated signals.
It should be noted that data presented at congresses leads to 12-26 % of the local literature ICSRs. How often is the data presented at these events already published in an article or it is going to be published in an article or in the best case, already reported by the healthcare professional or patient who originally observed the adverse event? Would it be a good practice that as a part of the poster / presentation submission to an event, the submitter would declare that the data presented has been appropriately reported beforehand?
Another thing to consider is the cost and efficiency of current local literature screening process.
As set in the Directive 2001/83/EC and Regulation 726/2004, EMA supports the global literature monitoring with MLM but there has been no support for local literature monitoring. Currently, based on our estimates, the local literature screening for one API can cost even 10 times more than global literature screening.
In the light of the data presented in this text, our suggestion is that the EMA would start performing monitoring of the local literature screening. Other option would be to define which of the non-indexed sources shall be monitored in each of the countries. This activity could be coordinated by the EMA with the NCAs. Third option would be to exclude certain products (for example orphan drugs) of this requirement or allow a risk-based approach to the monitoring based on the marketing status in the country and/or the indication of the product.
And, without doubt, we should strengthen the efforts to automate the process behind local literature monitoring. So far, the industry has been severely lagging in this area. If you already know anything about Tepsivo – stay posted on this.
[i] Directive 2001/83/EC, Article 107 3., as amended, available at: https://ec.europa.eu/health/sites/health/files/files/eudralex/vol-1/dir_2001_83_consol_2012/dir_2001_83_cons_2012_en.pdf
[ii] Guideline on good pharmacovigilance practices (GVP) Module VI – Collection, management and submission of reports of suspected adverse reactions to medicinal products (Rev 2), available at: https://www.ema.europa.eu/en/documents/regulatory-procedural-guideline/guideline-good-pharmacovigilance-practices-gvp-module-vi-collection-management-submission-reports_en.pdf
[vi] The initial analysis of the references was performed by Renuka Natarajan