OHDSI Study: Levetiracetam and Risk of Angioedema in patients with Seizure Disorder

jon_duke · May 18, 2016, 10:04pm

Good afternoon OHDSI researchers!

We are pleased to announce the official start of the Keppra and Angioedema study! See full details on the wiki including study rationale, protocol, and code.

So far we have participation from UCLA, Columbia University, Regenstrief, and Janssen. We would be delighted for you to join!

If you have any questions, please respond via this thread.

Thanks,

Jon, Martijn, Marc, Patrick, George

Christian_Reich · May 25, 2016, 10:03pm

Gents:

If we (IMS) were to join, what data assets would be of interest? Don’t say “all”.

Patrick_Ryan · May 25, 2016, 10:13pm

Ok, not all…just all with exposures to levetieacetam and events of
angioedema:)

The more the merrier, I would think any and all ims assets would provide an
important perspective on the problem not reflected by current data
participants who have already shared results (truven ccae, mdcr, mdcd,
optum, cprd, regenstrief).

schuemie · May 26, 2016, 12:25pm

Hi @Christian_Reich,

Given how rare the outcome is, bigger would be better!

msschwalm · May 31, 2016, 4:44pm

Hi Jon, I’m working with Christian Reich at IMS, I read the protocol of the project and it sounds very interesting. I’m not familiar with this type of collaboration but very familiar with European EMR data. Can you guide me on next step and timelines? should we perform first “assessment” or count in order to double check feasibility with our Databases? What is the best approach? Thanks. Regards Marie-Sophie

jon_duke · May 31, 2016, 6:36pm

Hi Marie-Sophie,

Thanks for your interest in joining this study! A good way to run a
feasibility check is through the OHDSI web platform (specifically the
‘Calypso’ tool). While this is not strictly necessary prior to running the
R package, it is nice to get a sneak preview. Do you have the OHDSI web
platform (e.g., WebAPI, Atlas, etc) set up at IMS?

Thanks!

Jon

msschwalm · June 2, 2016, 9:15pm

Hi Jon,
Ok thanks for your advise. We don’t have the OHDSI web Platform set up for now at IMS.
I will go and have a look at the Calypso tool first and see how to handle this in the most efficient way.

MS

Patrick_Ryan · June 3, 2016, 12:34am

If you don’t have calypso running, then I recommend you just move directly
into running the r package that contains the study. Thay way, you’ll
generate the results and can evaluate them before sharing. That’ll be much
more efficient than trying that install other tool to evaluate feasibility.

jon_duke · June 3, 2016, 12:10pm

Yes I agree with Patrick. If you do not have Calypso already I would go
straight to running the R package. Let me know if you run into any
problems.

YuriK · July 12, 2016, 6:46am

Hi Jon, I’m working with Christian Reich and Marie-Sophie on this research and I have some issues with R package, while trying to run test example “TestCodeImeds.R”. It returns following output

source(‘TestCodeImeds.R’)
Loading required package: DatabaseConnector
Loading required package: RJDBC
Loading required package: DBI
Loading required package: rJava
Creating exposure and outcome cohorts
Connecting using Redshift driver
Error in .jcall(drv@jdrv, “Ljava/sql/Connection;”, “connect”, as.character(url)[1], :
java.sql.SQLException: Amazon Error setting/closing connection: UnknownHostException.

What can be the reason for this error?

schuemie · July 12, 2016, 7:55am

Hi @YuriK,

That test script is specific to the IMEDS lab. Are you running your code in the IMEDS lab? If not, you’ll need to provide the appropriate connection details for your environment.

YuriK · July 13, 2016, 1:30pm

Martijn, I was trying to execute it on my local machine (Windows) outside the IMEDS lab. Which connection details should be used in this case?

Thanks!

Yuriy

schuemie · July 14, 2016, 6:02am

Hi @YuriK,

You will need to connect to a database that conforms to the Common Data Model. Someone in your organization should be able to tell you the information you need to connect to that database. What you’ll (probably) need is:

Type of database system (SQL Server, Oracle, PostgreSQL, or Redshift)
Server address
Database / schema name
User name
Password

(Depending on the database system some more or less information is needed.)

yonghui · July 14, 2016, 8:22pm

Hi ALL,

We got an error when running the script:

Using prior: None
Using 4 thread(s)
Fitting outcome model took 0.254 secs
Error in open.ff(X[[i]], …) : unable to open
Calls: execute … open -> open.ffdf -> sapply -> lapply -> FUN -> open.ff
Execution halted

Anybody have the experience or solution?

Thanks,
Yonghui

schuemie · July 15, 2016, 8:22am

Hi @yonghui,

It seems you’re almost done running the analysis (450 of the 808 outcome models have completed), but for some reason it crashed for the 451st model because it could no longer open the data files on the local file system it needed to fit the model.

My guess is this is because we may have some issues on Linux / MacOs with file handles not being released. These handles are automatically released when you close your R session. Could you:

Restart R
Rerun the execute function, but this time set createCohorts = FALSE, for example:

execute(connectionDetails,
         cdmDatabaseSchema = "OMOPV5_DE",
         workDatabaseSchema = "OHDSI",
         studyCohortTable = cohortTable,
         oracleTempSchema = "OHDSI",
         outputFolder = outputDir,
         maxCores = 4,
         createCohorts = FALSE)

The program should just begin where it stopped.

Please let me know if this helps

yonghui · July 15, 2016, 4:40pm

Thanks, Martijn. It works for the model training. All the models were finished after restart R.

But I got the following new error regarding to X11. Our server is a command line Linux without X11. Is there an option to disable the X11?

Thanks,
Yonghui

Full running log:

Loading required package: DatabaseConnector
Loading required package: RJDBC
Loading required package: methods
Loading required package: DBI
Loading required package: rJava
Running analyses
*** Creating cohortMethodData objects ***
*** Creating study populations ***
*** Fitting shared propensity score models ***
*** Adding propensity scores to study population objects ***
*** Trimming/Matching/Stratifying ***
*** Computing covariate balance ***
*** Fitting outcome models ***

Packaging results in export folder for sharing
Connecting using PostgreSQL driver
Error in .External2(C_X11, paste(“png::”, filename, sep = “”), g$width, :
unable to start device PNG
Calls: execute … packageResults -> -> -> dev ->
In addition: Warning message:
In grDevices::png(…, res = dpi, units = “in”) :
unable to open connection to X11 display ‘’
Execution halted

yonghui · July 15, 2016, 5:42pm

It seems that xvfb can solve this x11 issue:

Install xvfb
install x11 fonts:
yum install xorg-x11-fonts*
Run R in xvfb:
xvfb-run R --no-save

Then the script finished with no error but some warnings.

Yonghui

yonghui · July 19, 2016, 7:41pm

Hi ALL,

I’m trying to run this protocol using an Oracle database. However, the username contains a special symbol ‘#’.
How can I fill the information for oracle?

connectionDetails <- createConnectionDetails(
dbms = “oracle",
user = “C##NER", # Assumes environmental variables set before running R,
password = “pass", # otherwise fill-in with correct user/password pair.
server = “server_ip/SID",
port = “1521")

execute(connectionDetails,
cdmDatabaseSchema = “public”,
workDatabaseSchema = “results_schema”,
studyCohortTable = “ohdsi_keppra_angioedema”,
oracleTempSchema = NULL,
outputFolder = “~/project/angioedema/run/study_results",
maxCores = 4)

Thanks,
Yonghui

schuemie · July 20, 2016, 5:59am

Hi @yonghui,

I’m surprised it doesn’t just work like you’re using it right now. Could you tell me what the error message is?

yonghui · July 20, 2016, 3:33pm

Hi Martijn,

The user name used in the script is:
user = “C##NERCDMW”,

However, the error message reported was:
01918: user ‘CF6TX5KHA#NERCDMW’ does not exist

So, I guess the ‘#’ symbol caused the problem.

Here is the full running log:

Loading required package: DatabaseConnector
Loading required package: RJDBC
Loading required package: methods
Loading required package: DBI
Loading required package: rJava
Creating exposure and outcome cohorts
Connecting using Oracle driver

using THIN to connect
Error executing SQL: Error in .local(conn, statement, …): execute JDBC update query failed in dbSendUpdate (ORA-01918: user ‘CF6TX5KHA#NERCDMW’ does not exist
)

An error report has been created at /data2/user_data/ywu4/project/angioedema/run/errorReport.txt
Error in value[3L] : no loop for break/next, jumping to top level
Calls: execute … tryCatch -> tryCatchList -> tryCatchOne ->
Execution halted

Thanks,
Yonghui