13 replies to this topic

[Michael Kong]

Hi all,

I have a dataset with embedded BLQs (sometimes 2 or 3 consecutive embedded BLQs!!). How do I set the BQL rule set in Phoenix to set these as missing?

Thanks for your help!

Mike

[Simon Davis]

Hi Mike,
Can you perhaps share the data file then I could present back a PHXPROJ with the BQL object configured as you desire; perhaps as the screenshot below.

I would comment that if the BQL record is text e.g. "<10" it is already, for numerical purposes 'missing', so by default would not be plotted or used in calculations

Simon

Attached Thumbnails

• 

[Michael Kong]

Thanks Simon. Please see attached. [file name=raw_data.txt size=797]http://www.pharsight.com/extranet/media/kunena/attachments/legacy/files/raw_data.txt[/file]

Attached Files

•  raw_data.txt   797bytes   0 downloads

[Michael Kong]

Hi Simon,

I attached the data file in the last post. Did you get a chance to look at it?

Cheers,

Mike

[Simon Davis]

yes Mike - I got it and I set something up but there's something I want to confirm before I post it back. Been out all week but will try to get to it either tomorrow or Monday, sorry.

 

 Simon.

[Simon Davis]

Apologies Mike, I was trying to recreate something weird I saw in the attached project before I realised the problem; in this specific profile (Subject 3), the highest conc was the last timepoint, so therefore the rule below is executing as expected.

Before Tmax set BQL to zero,
Set first BQL after Tmax to BQL
Set all data points after two or more BQLs (post Tmax) to Zero.

I made a copy of this profile as Subject 3.1 to compare and contrast where the last point in the profile is less than a previous Tmax.


I'm not sure this example warrants changing anything with the BQL wizard since having the highest conc at the end makes it a pretty weird profile to be censoring the BQLs, what do you (and others) think?

Simon

[file name=mikeBQL.phxproj size=214830]http://www.pharsight.com/extranet/media/kunena/attachments/legacy/files/mikeBQL.phxproj[/file]

Attached Thumbnails

• 

Attached Files

•  mikeBQL.phxproj   209.79KB   0 downloads

[Michael Kong]

Thanks for this Simon.

My main problem is setting BLQs to missing for the NCA. However, it looks like you are setting embedded BLQs to zero. How would you set the BLQ rule set to do this, particularly when there are 2 or 3 consecutive embedded BLQs (eg: Subj 2, tp 5 & 8; Subj 3, tp 5, 8 ,12).

I suppose this also relates to your question on Subj 3 - here, I would rather not set BLQ to zero but set them as missing for the NCA. For plotting purposes however, all BLQs would be set to zero.

I feel this is the most appropriate way of handling these but let me know if you disagree.

Yes, these profiles are indeed odd but I suppose this is partly due to the increased analytical variability when you are close to the LLQ.

 

Mike

[Simon Davis]

Mike,
If your "main problem is setting BLQs to missing for the NCA" then you don't need to do anything since it is already text then it is already numerically missing and you could use the conc column as it is directly in NCA, you don't need to prep it first with the BQL wizard

However, if you do want BQLs to be replaced according to rules dependent on their position then you can use the BQL wizard. Here is a simpler example that would be suitable for extravascular admin as it helps report Tlag if present (if IV i wouldn't set PreDose values to zero since I want to back extrapolate to a predicted C0)

Before Tmax set BQL to zero,
Set all other BQL after Tmax to a missing value e.g. < 2.00

the created column PKconc is the one you would map to you NCA

For plotting purposes if you want all BLQs to be set to zero then you could feed the output into a second BQL object (or DataWizard since there is no need to identify position within the profile) and create GraphConc with your required substitution.

The final set has RawConc e.g. for a table or listing, PKconc for your NCA and GraphConc for any graphs, remember that you may want to consider some nominal number other than zero if you want to be able to plot them on a log scale. Remember yo can use Reference lines in Phoenix graphs to display LoQ

Simon [file name=BQL_Result.xls size=1256]http://www.pharsight.com/extranet/media/kunena/attachments/legacy/files/BQL_Result.xls[/file]

Attached Thumbnails

• 
• 

Attached Files

•  BQL_Result.xls   1.23KB   0 downloads

[Michael Kong]

Simon ,

In your example, the 12 hr tp for subject 2 is set to <2.00. Can you set up the rule set to display this as the original value (ie:4.83) in the pkconc column?

Mike

[Simon Davis]

yes Mike, simply leave the optional setting I have circled blank.

this option is proposed where the user wants to censor quantifiable values observed after two or more consecutive BQL values and was added, at user requests, in 6.3

Simon

Attached Thumbnails

• 

[Michael Kong]

Thank you Simon - I should have been more clear that this is what I wanted to achieve.

Returning to your question about subject 3, I agree that censoring BLQs here would be odd given the last timepoint Cmax and that setting to missing would be appropriate. However, since our rule set defines pre-Tmax BLQs to be set to zero, this exceptional case would require a separate rule set. I suppose I could change the '<2.00' for this subject to some other non-numeric code to define a rule set specifically for this subject. Is there a better way? Appreciate all the support..

[Simon Davis]

no problem Mike I was actually just giving an illustrative example; I'd encourage you to play with it to check it process the data as you are expecting. Regarding subject 3, with the rules that you described it will process fine so you don't need another rule.

If in some other case you did need that then I would probably write an explicit "custom transformation" in a data wizard to pick out the values that you needed to edi so that people could easily see and track these 'one offs'

alternatively you could implement some data checks to compare e.g. Tmax and Tlast and flag that if it's something that you want to highlight in your report. I've also used the selective capabilities of inner join to pick out timepoints corresponding to Cmax (see the project on this post for fun!)

http://www.pharsight...34&id=1387#1394

Really Phoenix has a lot of flexibility and once you've got some of its capabilities you can design the workflow that works best for your needs.
Simon

Attached Thumbnails

• 

[Michael Kong]

Since Tmax is at 24 hr, for this subject I wanted all embedded BLQ before that to be set to missing and the one at pre-dose to be set to zero. This is not the case if we use the current rule set as you've shown under column pkconc. Am I missing something?

I haven't really dealt with embedded BLQ before Tmax, but I presume this is the best way to handle them - if you have insight on this, that would be great.

I'll definitely look at your post, thanks.

[Simon Davis]

the advantage of setting BQLs before Tmax to Zero is to get a reported TLag but if you simply want the predose time point to be set to zero when BQL i would just use the data wizard with the custom transformation;

if(time=0,0,conc)

Simon

PS did you see the announcement about Phoenix Software user group meeting Tuesday, 29 October 2013 8AM - 5PM The Pyramid Club Philadelphia, Pennsylvania 19103

http://www.pharsight...0&catid=46#2040

Attached Thumbnails

•