The New EMA Protect ADR Database
On February 18, 2013 the EMA released a new Adverse Drug Reaction (ADR) database. See: http://www.imi-protect.eu/methodsRep.shtml
It is called the Protect Database and was developed by the EMA and some partners. It consists of all ADRs listed in section 4.8 (Undesirable Effects) of the Summary of Product Characteristics (SPC) for all centrally approved products. The database was updated as of 30 June 2012 based on the SPCs found at the European Commission’s website. It will be routinely updated at least once a year as new or amended SPCs are produced. Additional updates will be done to correct errors.
It is based on MedDRA PTs though LLTs may also be used if further precision is required. As not all the terms in the SPCs are in MedDRA so some mapping was done using “fuzzy text matching” and expert review of the mapping.
The EMA summarized its objectives in preparing this database. They note that the determination of whether an ADR is included in the SPC is time consuming and that the existence of a structured database for this would be useful. This allows the filtering and flagging of signals for unlisted (unexpected) reactions only and makes the signal detection process more efficient.
They further note in their letter describing the database (see the URL above for the document entitled “Introduction”): “A data set of established ADRs also allows a comparison to coincidental or unidentified drug-adverse event combinations only, an adjustment of statistical signals for known ADRs, and an evaluation of the effect of background restriction on the performance of statistical signal detection.”
Interestingly, they also say that the objective is NOT to provide a continuously updated list of ADRs.
A disclaimer notes that this database is not required by law is done only as an additional signaling tool. They note that the EMA and their partners cannot be held responsible for failure to some ADRs, inaccuracies in MedDRA mapping and “inappropriate use of this database.”
The database is an Excel spreadsheet which you can download immediately from the website (URL above). The structure includes:
- The brand name
- The active substance
- The date of the SPC
Complexity: if the product is generic, the date may also be the date when the CHMP issued a positive opinion for a variation type IB. If there is no decision, the SPC will be found in the EMA website.
- The ADR as it is written in the SPC.
- The MedDRA PT text corresponding to the ADR.
- Sometimes the MedDRA LLT and rarely the MedDRA SOC, HGLT and HLT if the term in the SPC is one of these.
- The MedDRA code number for the PT and SOC are given also for all SPC listed terms.
- The age group if indicated in the SPC ADR (children, adults, elderly only) or if not noted with the ADR: not specified.
- Gender if mentioned otherwise not specified/unclear/both genders.
- Causality if noted. The choices are: established ADR (which is the default for all ADRs in this database) unless it is specified that the relationship is either possible, doubtful or unassessable/unclear.
- Frequency using categories of unknown/not mentioned, very rare, rare, uncommon, common, very common. If there are conflicts in frequency (CF=conflicting frequencies) this is noted. This can occur if there are several clinical trials with different results (sometimes for different indications) or if ADRs are listed for each individual component of a product. Interestingly, if a percentage is described the database calls this “unknown” as “figures have not been interpreted”. That is, if a percentage is reported in the SPC it will not be in the database.
- Class warning: yes/no.
- Whether the ADR is listed in a table presenting results of clinical trials or in the context of trials: yes/no.
- Whether the ADR is listed in a table of post-marketing surveillance: yes/no. If seen in trials and post-marketing surveillance this will be noted as “O” in both.
The spreadsheet can be downloaded to your computer. The spreadsheet has 20 columns and 45,299 rows. The left-most column is the product name and it is repeated each time there is an ADR. So, for example, the first drug in the table, Abilify, covers the first 99 rows because it has 99 ADRs.
In a nice touch, the spreadsheet has as its first row a dropdown search box allowing one to easily sort the list by each of the columns. So one could look for all, say, headaches in the SPC term, PT, LLT etc. One can also color code selected cells.
This is a lovely database and clearly a lot of work was put into it. But several questions now arise.
This database has been prepared by the EMA and, in spite of the disclaimer, some may interpret this database as now being the “official” one for expected (listed) and unexpected determinations. Companies may use this database not just for signaling but also for routine case processing.
If so, this database may differ from the company’s own internal database of what is expected or not, particularly if the SPC terms are not MedDRA but rather verbatim terms. It may be a large effort for a big company to check their own list vs the EMA’s.
If this is now a “semi-official” database and it differs from how a company has done expected/unexpected this may produce problems. Are some cases which were considered by the company to be expected now considered unexpected based on this database? Or vice versa. The likelihood of this is small but with tens of thousands of terms that are drug specific, this will undoubtedly occur. The biggest issue will occur when there was not a direct mapping of the SPC ADR to MedDRA.
So what should a company do? Recode cases? Submit what are now late expedited reports to the EMA? Might a company now have more cases that are expedited in the EMA and not in the US or vice versa?
Should a company now match up this database with their own? Will the EMA consider that if a company uses the PROTECT database for expectedness instead of its own as “inappropriate use”?
Are there liability or legal ramifications? How will this be handled during regulatory PV inspections?
What if there is a disagreement or an error? Presumably, the company should point this out to the EMA at the email listed for correspondence (Protect_support@ema.europa.eu).
Are there any implications for the Investigator Brochure if the approved drug is being used in clinical trials also?
Can companies “ignore” or not use this database if they so choose?
Should it be used in the preparation of RMPs and PSURs/PBRERs? If so, in which sections?
Will the EMA and member states’ competent authorities use this for signaling and expectedness determinations? If so, what happens if the company does not? What if the MS CA finds a signal that the company doesn’t due to a difference in this database vs. the company’s?
These and other questions will undoubtedly be discussed over the next few months and hopefully, EMA will issue some clarifications, perhaps as a Q&A or FAQ document.
Yet another lesson that every action, particularly governmental actions, may have unintended consequences.