hero image

Earlier this year FDA announced that it is planning to make access to much of its structured data, including the adverse event data in its safety database FAERS, available more easily to the public.

Dr. Taha Kass-Hout, the Chief Health Informatics Officer of FDA wrote on the healthdata.gov website that FDA is developing an initiative called “openFDA”.  This program will make some of its data available to the public for upload.  The initial data will include:

The access will be available as raw files to be uploaded as well as “APIs” (“An Application Programming Interface, or API, is a set of software instructions and standards that allows machine to machine communication”) see here.

The first datasets are expected to be available in the summer of 2014 and a larger release in the fall of 2014.

FDA anticipates that it will also highlight projects using these data both by the government and private groups.  FDA also states that “openFDA will provide a platform for public challenges issued by the FDA and a place for the community to interact with each other and FDA domain experts with the goal of spurring innovation around FDA data”.   Not entirely clear what this means, particularly the first part about “challenges”. Click to see the new website (though there’s not much on it yet).

Currently FDA does make safety data available from FAERS.  Unfortunately these are raw zipped ASCII data files that are issued quarterly and not cumulative.  One must have an expert available with knowledge of the creation of relational databases in order to use these files.  These files contain some but not all information submitted to FDA in MedWatch forms or as E2B files.  The narrative, in particular, is not included.  This is problematic as the narrative is often the critical part of the case containing a sequential and detailed medical summary.  See this site for the files.

The only other way to get information from FAERS is to submit a Freedom of Information (FOI) request for either individual cases or a summary report on an individual product.  This too is complex.  Click here for an FOI request.

For individual cases, there is yet further complexity since the narrative in the MedWatch (or E2B) report must be reviewed by FDA to remove personal information and identifiers that are confidential.  The mechanism to obtain this information is available on the FDA website here and here.

Briefly, getting detailed AE data requires that the requestor identify each individual report desired.  Thus, in practice, one must get either a product summary report from FDA (as above) or use the quarterly files loaded into the requestor’s (or a third party’s) database to obtain the FAERS case report number.  In addition, the FDA would also like in each request the suspect drugs.  They prefer that a case list in Excel format be used for multiple cases.  The FDA does not want other case numbers (e.g. company case numbers, tracking numbers from the Spontaneous Reporting System data extracts sold by the National Technical Information Service).

The requestor must pay the fees for this which are not known in advance as the cost is a function of how much time it takes FDA to prepare the information from FAERS.  The requestor must submit “a statement concerning willingness to pay fees, including any limitations.”  FDA estimates each case takes about 15 minutes to prepare.

This is usually then a two step procedure (first getting a summary report to obtain the FDA FAERS number) followed by a second request to obtain the individual cases; this can take weeks or longer.  It is also possible that follow-ups to the initial case will not be picked up in the information sent by FDA if the follow-ups are not well linked to the initial case.

Well, to say that this is complex, not-user friendly and difficult to do is an understatement.  This has been clearly recognized by FDA and has led to the openFDA initiative; this is to be applauded.

What is not stated is whether the new files will contain the narratives.  One suspects not, as the redaction process must still presumably be done by hand.  As there are millions of cases, it is not clear how this could be done for the narratives in order to make them public.  We may still have to do an FOI request to get the narrative data.

It is also not clear what, if any, time lag will occur in making the data available.  For comparison, the latest dataset available at this time on FDA’s FAERS webpage is for the January – March 2013 quarter.  So there appears to be a lag of up to 9 months or more.

Since the FDA’s database is one of the largest in the world, making this information available will be a very valuable service to the public (and companies too).

So all of this is to be applauded but one can’t help but wonder whether there might not be other useful ways to make the data available.  One can look at other safety databases that make their data easily (or at least more easily) available.  Each has its limitations and none make full cases easily available:

Health Canada

The Canadian government makes extracts of its MedEffect Database available here.  This webpage allows one to search the database of adverse reactions from 1965 through September 30, 2013 (as of this writing).  It is easy to do and there are clear instructions on how to do this on this webpage.  After going through a couple of caveat and instruction screens one is led to a search page where one fills in information on the date range, patient information (gender, outcome, age range), the health product(s) in question either by brand name or active ingredient and AE search terms.  Not all information is required and one can do a search asking for everything on a particular drug or ingredient without any other search filters.

The data is delivered either in a standard format or in a customized format and can be exported.  A table is delivered (based on MedDRA) with the Canadian AR report number, the initial received date, age, gender, suspect product(s) and AR terms.  The report number is a hyperlink which, when clicked, brings the individual case information up.  This information contains more information on the patient, seriousness, products taken and AR terms.  There is no narrative.

All in all this works very well and is reasonably up to date (a 6 month lag more or less) and is available immediately.

EMA Portal

The EU has made data from EudraVigilance available in a limited aggregate form. See this page.  After a disclaimer page, one is led to an alphabetical list by name or substance where one can search for a product.  Not all marketed products are available.  One then clicks on the name and a pdf file is downloaded.  This file can be opened in Adobe. These are interactive pdf files! It has much aggregate information such as number of cases by age group, sex, geographical area, by SOC and then by more narrow AR terms (PT/LLT).  Again, the listings are not complete for all drugs and there are no individual case records available.  Nonetheless the data is useful and interesting.  The drug I chose at random to look at was up to date through February 2014 (one month ago).


The MHRA also has immediately available summaries of suspected ARs called Drug Analysis Prints (DAPs).  They contain complete listings of all suspected adverse drug reported to the MHRA via the Yellow Card Scheme for a particular drug substance. This includes all reports received from healthcare professionals, members of the public, and pharmaceutical companies.  The products are listed alphabetically by active ingredient.

Each report is a pdf of summary tabulations of ARs by SOC/PT (sometimes HLT).  There is little individual patient data.  See here.

The Uppsala Monitoring Centre

The UMC also has a drug safety database with limited information.  It contains data from national health agencies.  It is not on-line and must be requested from the UMC and each search is customized.  There is a fee for the information. See this page.

Bottom line: There is limited information available on-line directly from health agencies on drug adverse events/reactions.  The FDA is promising to make its FAERS data more easily available.  What we will be able to obtain remains to be seen but hopefully we will be able to get much more data more easily.

Related Articles