Eudract and Safety XML

Simon Bond



The European Clinical Trials Database Eudract is run by the European Medicines Agency. All studies that are officially registered clinical trials have to enter the results of the final analysis into the website to be made public.

An equivalent database hosted by the U.S. government is

There is a large amount of documentation online that will not be repeated here.

The detailed entering of safety information is an onerous task if it is to be done by hand. However we do have the facility to upload an XML file to automate this step. This R package seeks to enable the production of an XML file from a standard structure of Safety data that is recorded on a patient-level.

The key functions are:

We provide a dummy safety data set that is one line per patient-event: ?safety. The format of the data, specifically the variable names is described in the help file printed below. This needs to be turned into several frequency tables: one given information at a group level; one given at the event level, broken down into serious and non-serious events. The term “group” here corresponds to treatment arms in a randomised control trial. You need to define a group within EudraCT even if it is a one-armed study, in which case it can be a “dummy” label.

Some information is defined in a hard-coded fashion below, but it is understood that this will be generated by code if applied in real life. Each entry in the vectors below correspond to counts in each of the two groups.

subjectsExposed <- c("Control"=99,"Experimental"=101)
#count of deaths not in the Safety data. Could be c(0,0)
deathsExternal <- c("Control"=3,"Experimental"=5)

Coded adverse events are required to be helpful and avoid the task of reconciling minor spelling or text inconsistencies. This package and vignette assumes this is the case, and will not work in the absence of coding. We cannot provide the full MedDRA dictionary, due to copy right reasons. But normally this is available to sponsors. However, for upload into EudraCT, as a minimal requirement, only the System Organ Class (SOC) needs to be fully coded into the EudraCT internal version coding system. We have provided an internal data set, derived from the eutct site in the package to use this; see ?soc_code.

Safety data set

safety R Documentation

Example of safety data


A dataset containing some example data of safety event in raw source format




a data frame with 8 columns and 16 rows


meddra preferred term code


a unique subject identifier


a logical indicating if the event is related to the treatment


the meddra code for the System Organ Class


a numerical 0/1 to indicate if the event was fatal


a numerical 0/1 to indicate if the event was serious


the treatment group for the subject


a text description of the event. Needs to be matching 1-1 with the pt code


The data contains one row per patient-event. So the numbers exposed in each arm cannot be inferred from these data, as patients with no events will not be included in these data.

The variable names and formats are those required by safety_summary. The variable pt is not strictly required. An alternative to soc would be the equivalent character string from soc_code

Calculate Summary Statistics

We provide a function that derives the patient and event counts as required in a format internal to R.

safety_statistics <- safety_summary(safety, 
                                    excess_deaths = deathsExternal, 
                                    freq_threshold = 1
## Group-Level Statistics
##          title subjectsAffectedBySeriousAdverseEvents
## 1      Control                                     15
## 2 Experimental                                     33
##   subjectsAffectedByNonSeriousAdverseEvents deathsResultingFromAdverseEvents
## 1                                        15                                9
## 2                                        24                               22
##   subjectsExposed deathsAllCauses
## 1              99              12
## 2             101              27
## Non-serious event-level statistics (intial rows)
##   groupTitle subjectsAffected occurrences                              term
## 1    Control                1           1           Acute coronary syndrome
## 2    Control                1           1            Intestinal perforation
## 3    Control                1           1                Laryngeal stenosis
## 4    Control                1           1 Lower respiratory tract infection
## 5    Control                1           1               Lung adenocarcinoma
## 6    Control                2           2                         Pneumonia
##        eutctId
## 1 100000004849
## 2 100000004856
## 3 100000004855
## 4 100000004855
## 5 100000004855
## 6 100000004862
## Serious event-level statistics (intial rows)
##   groupTitle subjectsAffected occurrences                       term
## 1    Control                1           1             Abdominal pain
## 2    Control                1           1   Aortic valve replacement
## 3    Control                1           1            B-cell lymphoma
## 4    Control                0           0          Bladder papilloma
## 5    Control                1           1             Cardiac arrest
## 6    Control                0           0 Cardiac failure congestive
##        eutctId occurrencesCausallyRelatedToTreatment deaths
## 1 100000004856                                     0      0
## 2 100000004865                                     0      0
## 3 100000004851                                     0      0
## 4 100000004864                                     0      0
## 5 100000004849                                     0      1
## 6 100000004849                                     0      0
##   deathsCausallyRelatedToTreatment
## 1                                0
## 2                                0
## 3                                0
## 4                                0
## 5                                0
## 6                                0

Convert to XML

If you have produced these statistics through separate coding, then you can use the eudract:::create.summary_statistics() function to put them into the correct internal format and start the conversion to XML directly.

First we export the safety_statistics to a XML document that is human readable “simple.xml”. Then we convert to the EudraCT and formats.

simple <- tempfile(fileext = ".xml")
eudract_upload_file <- tempfile(fileext = ".xml")
ct_upload_file <- tempfile(fileext = ".xml")
simple_safety_xml(safety_statistics, simple)
## '/tmp/Rtmpw8QfUO/file15bd56da62b.xml' is created or modified
## '/tmp/Rtmpw8QfUO/file15bd3bb5fa98.xml' is created or modified
## Please email to tell us if you have successfully uploaded a study to EudraCT.
## This is to allow us to measure the impact of this tool.
                       original=system.file("extdata", "1234.xml", package ="eudract"),
## '/tmp/Rtmpw8QfUO/file15bd36869b1c.xml' is created or modified
## Please email to tell us if you have successfully uploaded a study to .
## This is to allow us to measure the impact of this tool.

Note that for the, there must first be a study set-up within website, and then a download of the XML taken. This is the original argument. Then the original file has the safety events data over-written, and can be manually uploaded back into

Alternatively, if you have a user account within, then the initial study needs to be set up within there, but we can use the API to directly upload without needing to manually interact with the site.

# Not actually run. It needs real user account details: the ones below are fictitious.


The key outputs are

We can validate the output against the XML schemas provided by EudraCT and, although the calls to eudract_convert() and clintrials_gov_convert() also do this behind the scenes, returning the value TRUE if there are no errors against the schema validation.

note these are semi-readable files of code/data rather than a standard web page.

myschema <- xml2::read_xml(system.file("extdata","adverseEvents.xsd", package="eudract"))
aes <- xml2::read_xml(eudract_upload_file)
check <- xml2::xml_validate(aes,myschema)
if(check){print("Validation against eudraCT schema has passed!")}
## [1] "Validation against eudraCT schema has passed!"
myschema <- xml2::read_xml(system.file("extdata","ProtocolRecordSchema.xsd", package="eudract"))
aes <- xml2::read_xml(ct_upload_file)
check <- xml2::xml_validate(aes,myschema)
if(check){print("Validation against schema has passed!")}
## [1] "Validation against schema has passed!"

Manual Upload

To use the resulting eudraCT xml file navigate and log in online to the study specific area of the EudraCT site. On the top banner is a link “Upload XML” which you follow. Choose the option “Adverse Events” rather than “Full data set”, and select the file xml you have produced. The resulting information can be viewed in the browser interactively or with a static pdf file (note this is a fictitious study and fictitious data). This is not the only step in completing the EudraCT report, as the description of the study, baseline characteristics and efficacy analysis will all need to be added. That is not the remit of this package though.

For the, once logged in, there is a button titled “Records”, near the top. From there select “Upload Record (XML)”. On the new page, use the “Choose File” button to select the newly created XML file, and click “Upload”.

To extract the original study record for over-writing, you need to go into the specific study record from the home page. In there beneath the initial section titled “Record Status”, there is a link “Download XML”, which will enable you to save locally the required file.