Bioequivalence and Bioavailability Forum • Health Canada: Data format

Helmut
★★★

Vienna, Austria,
2018-01-15 13:32
(2147 d 02:31 ago)

Posting: # 18172
Views: 3,227

Health Canada: Data format [Regulatives / Guidelines]

Dear all,

does anyone have experience with the 2004 „Draft Guidance for Industry. Preparation of Comparative Bioavailability Information for Drug Submissions in the CTD Format”?

Appendix B: Computer Format for the Submission of Data for Comparative Bioavailability Studies
2. Detailed Specifications

states

• Concentrations below the limit of quantitation (LOQ) should be entered as 0.0.

What? Also strange:

The following provides an example of a hard copy of the data set and should be included with the submission:

01 AB 1 A 000.00 000.00² 052.012³ 095.03 122.20 .⁴ 065.15 046.24 019.20 014.99 000.00 000.00
─────────
² Concentrations below the limit of quantitation (LOQ) should be entered as 0.0.
³ The entered drug concentrations must be the measured (uncorrected) concentration and should not be below the lowest or above the highest nominal concentrations of the standard curve.
⁴ Missing data must be entered as a period (.).

What I don’t get:

The format of the second file does not allow for giving the actual sampling times (only the scheduled ones in the first file).
BQLs as zero? Whilst it makes sense to set BQLs before the first measured concentration to zero I’m not so sure about later ones.

Which software is HC using? I expect that the results of NCA will not match the standard output of commercial software (e.g., Phoenix WinNonlin) and R-package bear.

—
Dif-tor heh smusma 🖖🏼 Довге життя Україна!
Helmut Schütz

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes

ElMaestro
★★★

Denmark,
2018-01-15 17:53
(2146 d 22:11 ago)

@ Helmut
Posting: # 18173
Views: 2,601

Health Canada: Data format

Post reply

Hi Hötzi,

"because of reasons" :-D

You are of course right, but for purposes of importing such data it may be more convenient to rely on only cells with something that can be converted to doubles or a single category of "something else". I know this is not optimal, but there are such technical reasons and restraints.
As a relevant perspective, you try and write a function in R which reads a textfile, figures out numbers of columns and rows, imports the table and uploads every cell value in memory without using the usual read.table or read.csv etc functions. Of course you need at all times to be able to do numerical gymnastics like max(Sneaky.Table[,7], na.rm=T) etc. You'll get sick and tired within 2 minutes and you will be loving the for nice little simplifying rules about the nature of input. Now imagine doing the same in fortran or whatever where you have 1/1000 of the usual R functions. Big trouble...

Yes, life would be easier if the guy using the data you are submitting would be as IT-literate as you are. Well, life just isn't always optimal :-)

—
Pass or fail!
ElMaestro

Helmut
★★★

Vienna, Austria,
2018-01-16 01:16
(2146 d 14:48 ago)

@ ElMaestro
Posting: # 18174
Views: 2,649

Data integrity?

Post reply

Hi ElMaestro,

❝ "because of reasons" :-D

❝

❝ etc. etc. etc.

❝

❝ Yes, life would be easier if the guy using the data you are submitting would be as IT-literate as you are.

I have no technical problems preparing the data in this format. What I find hard to swallow is that I

have to manipulate (!) the data (my files contain the string BQL for all values <LLOQ after t_max) and
have no means to give actual sampling times.

Concerning #1 hopefully HC’s software is “clever” enough to ignore late BQLs (replaced by zeros) and will not add a triangle after t_last (i.e., apply Pharsight’s unspeakable invention AUC_all instead of AUC_last). #2 will always lead to discrepancies to my results since I abandoned using scheduled sampling times in NCA decades ago.

❝ Well, life just isn't always optimal :-)

All too true.

BTW, another goody in Section 11.4.2

The regression lines, based upon at least four points during the terminal log-linear phase of the curve, used to estimate the terminal disposition rate constant (λ) should also be displayed.

Oops!

—
Dif-tor heh smusma 🖖🏼 Довге життя Україна!
Helmut Schütz

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes

Helmut
★★★

Vienna, Austria,
2018-01-16 14:07
(2146 d 01:57 ago)

@ ElMaestro
Posting: # 18181
Views: 2,616

Health Canada: 2012 guidance

Post reply

Hi ElMaestro and all,

can we ignore at least part of the 2004 guidance? In Appendix 1 (Tables A1-B&C) to the 2012 BE-guidance BQLs are given as such with an exploratory footnote:

Lower limit of quantitation is 5 ng/mL. Any concentration below this limit is reported as Below Limit of Quantitation (BLQ) except at time 0. Zero is used in the calculation of area under the curve (AUC) for times preceding the first observed concentration and in the calculation of summary statistics.

What do we have?

2004 guidance
0.00 0.00 52.01 95.03 122.20 77.88 65.15 46.24 19.20 14.99 0.00 0.00
2012 guidance
0.00 BLQ 52.01 95.03 122.20 77.88 65.15 46.24 19.20 14.99 BLQ BLQ
Output of Phoenix/WinNonlin
0.00 0.00 52.01 95.03 122.20 77.88 65.15 46.24 19.20 14.99 BQL BQL
That’s PHX/WNL’s most commonly applied BQL Rule Set.
With unconditional substitution (BQL → 0) one could comply with the 2004 guidance.
To comply with the 2012 guidance one could keep the data as reported (BQLs) and set up a replacement in a Data Wizard (if t=0: BQL → 0).
Output of bear
0.00 NA 52.01 95.03 122.20 77.88 65.15 46.24 19.20 14.99 NA NA
Complies with the 2012 guidance.
To comply with the 2004 guidance the data have to be modified outside of bear (in R, a spreadsheet, …) before import.

—
Dif-tor heh smusma 🖖🏼 Довге життя Україна!
Helmut Schütz

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes