Bioequivalence and Bioavailability Forum

Main page Policy/Terms of Use Abbreviations Latest Posts

 Log-in |  Register |  Search

Back to the forum  Query: 2017-10-17 05:59 CEST (UTC+2h)
 
msmnainar
Regular

India,
2017-01-20 07:49

Posting: # 16985
Views: 1,711
 

 LSM differences in SAS vs Phoenix [Software]

Dear all

Could you please clarify the reason for getting results of LSM differences and confidence limit values in opposite to one another (eg., negative vs positive or positive vs negative values) when you evaluate the same data using SAS and Phoenix WinNonlin.


Thanks

Sundar
Shuanghe
Regular

Spain,
2017-01-20 12:31

@ msmnainar
Posting: # 16990
Views: 1,442
 

 LSM differences in SAS vs Phoenix

Hi Sundar,

From what you said, I guess you should check your SAS code first. Maybe you coded treatment as A and B for test and reference but you copy/paste the sas code from guideline or somewhere else? If so, change the order of -1 1 to 1 -1 in estimate statement since many of those code assume that the treatment is coded as T and R, so the order is reversed. e.g.ESTIMATE 'Test vs. Reference' treat -1 1 to ESTIMATE 'Test vs. Reference' treat 1 -1.

If your sas code is correct then I couldn't think of any reason how one can screw up Phoenix setting since you would visually assigned which formulation is reference, whether it is coded as A/B, T/R or 1/2. You should know which is which.

All the best,
Shuanghe
Helmut
Hero
Homepage
Vienna, Austria,
2017-01-20 13:34

@ msmnainar
Posting: # 16991
Views: 1,461
 

 Lexical order

Hi Sundar,

» Could you please clarify the reason for getting results of LSM differences and confidence limit values in opposite to one another (eg., negative vs positive or positive vs negative values) when you evaluate the same data using SAS and Phoenix WinNonlin.

Both software packages by default give the LSM difference in lexical order. If you coded test with T and reference with R, the difference will be given as R – T (since R < T).
It might be that your SAS-code changed the order by a statement similar to this one:

lsmeans formulation / pdiff=control("R") CL alpha=0.10;

In Phoenix there is no native method the change the order which you will get in
Bioequivalence > Output Data > LSM Differences.

[image]


My method to recode the LSM differences (which works for any specification of T and R):
LSM Differences > Send To > Data > Data Wizard
Rename to Reordered LSM Differences and add these Custom Transformations:
  1. R: left(Level_Level, search(Level_Level, '-', 1)-2)
  2. T: right(Level_Level, search(Level_Level, '-', 1)-2)
  3. Relevel: concatenate(T, ' - ', R)
  4. Difference: -Estimate
  5. Lower: -Upper_CI
  6. Upper: -Lower_CI
Cosmetics:
  • Filter: Exclude
    ⦿ Level_Level
    ⦿ Estimate
    ⦿ Lower_CI
    ⦿ Upper_CI
    ⦿ R
    ⦿ T
  • Properties
    Relevel => Level_level
    Difference => Estimate
    Lower => Lower_CI
    Upper => Upper_CI

[image]



PS: You selected ☑ E-mail notification if there has been a reply to this message
However, both replies bounced from the E-mail server (xxxxxxxx wasn’t found at yyyyyyyy.com. 550 5.1.10 RESOLVER.ADR.RecipientNotFound; Recipient not found by SMTP address lookup).
Since you registered in 2011 with your company’s e-mail address: Did you change your job? If yes, consider updating your profile (once logged in, click on msmnainar in the upper right hand corner [image] Edit User data E-mail click on [ Change e-mail address ]) or refrain from further E-mail notifications. THX.

[image]All the best,
Helmut Schütz 
[image]

The quality of responses received is directly proportional to the quality of the question asked. ☼
Science Quotes
ElMaestro
Hero

Denmark,
2017-01-20 20:12

@ Helmut
Posting: # 16992
Views: 1,395
 

 Lexical order

Hi Hötzi,

» Both software packages by default give the LSM difference in lexical order. If you coded test with T and reference with R, the difference will be given as R – T (since R < T).

Hey I did not know this :-)

Perhaps this explains why many companies are still using A and B for treatments:cool::ponder:

I could be wrong, but…


Best regards,
ElMaestro

- since June 2017 having an affair with the bootstrap.
Back to the forum Activity
 Thread view
Bioequivalence and Bioavailability Forum | Admin contact
17,394 Posts in 3,725 Threads, 1,071 registered users;
27 users online (0 registered, 27 guests).

The rise of biometry in this 20th century,
like that of geometry in the 3rd century before Christ,
seems to mark out one of the great ages or critical periods
in the advance of the human understanding.    R.A. Fisher

The BIOEQUIVALENCE / BIOAVAILABILITY FORUM is hosted by
BEBAC Ing. Helmut Schütz
XHTML/CSS RSS Feed