Dear all,

I am trying to calculate the sample size necessary to proof non-
inferiority for two survival curves. I have downloaded the 7-days
trial version of the PASS software but it does not work. I get the
following message: "A value of 0 is out of range", I have tried the
example that I find in the PASS documentation and even with this data
it does not work.

I have thought also about calculating by myself this sample size but I
found only a little number of papers regarding this problem, all of
these too difficult to understand for me.

Could some body guide me about how to obtain this sample size?? Maybe
statisticians are used to use some approximation as it is usual in
these calculations?

Thanks in advance for your help!!

Sincerely,
Juan Vicente Torres.

Dear all,

I am trying to calculate the sample size necessary to proof non-
inferiority for two survival curves. I have downloaded the 7-days
trial version of the PASS software but it does not work. I get the
following message: "A value of 0 is out of range", I have tried the
example that I find in the PASS documentation and even with this data
it does not work.

I have thought also about calculating by myself this sample size but I
found only a little number of papers regarding this problem, all of
these too difficult to understand for me.

Could some body guide me about how to obtain this sample size?? Maybe
statisticians are used to use some approximation as it is usual in
these calculations?

Thanks in advance for your help!!

Sincerely,
Juan Vicente Torres.

On May 6, 10:30 am, "juanvte... at (no spam) googlemail.com"



juanvte... at (no spam) googlemail.com> wrote:
Dear all,

I am trying to calculate the sample size necessary to proof non-
inferiority for two survival curves. I have downloaded the 7-days
trial version of the PASS software but it does not work. I get the
following message: "A value of 0 is out of range", I have tried the
example that I find in the PASS documentation and even with this data
it does not work.

I have thought also about calculating by myself this sample size but I
found only a little number of papers regarding this problem, all of
these too difficult to understand for me.

Could some body guide me about how to obtain this sample size?? Maybe
statisticians are used to use some approximation as it is usual in
these calculations?

Thanks in advance for your help!!

Sincerely,
Juan Vicente Torres.

I found this paper to be helpful.

"Sample Size Computation for Two-sample Noninferiority Log-rank Test"
by Jung, Kang,
McCall and Blumenstein in J. of Biopramaceutical Statistics, 15:
969-979, 2005

On May 6, 6:16 pm, "Kevin E. Thorpe" <kevin.tho... at (no spam) utoronto.ca> wrote:



On May 6, 10:30 am, "juanvte... at (no spam) googlemail.com"

juanvte... at (no spam) googlemail.com> wrote:
Dear all,

I am trying to calculate the sample size necessary to proof non-
inferiority for two survival curves. I have downloaded the 7-days
trial version of the PASS software but it does not work. I get the
following message: "A value of 0 is out of range", I have tried the
example that I find in the PASS documentation and even with this data
it does not work.

I have thought also about calculating by myself this sample size but I
found only a little number of papers regarding this problem, all of
these too difficult to understand for me.

Could some body guide me about how to obtain this sample size?? Maybe
statisticians are used to use some approximation as it is usual in
these calculations?

Thanks in advance for your help!!

Sincerely,
Juan Vicente Torres.

I found this paper to be helpful.

"Sample Size Computation for Two-sample Noninferiority Log-rank Test"
by Jung, Kang,
McCall and Blumenstein in J. of Biopramaceutical Statistics, 15:
969-979, 2005

Thank you very much Bruce and Kevin:

I used both recommendations and I got the solution by means these two
different paths.

Jerry from PASS support replied my e-mail and the solution was as easy
as setting my windows xp to run with english settings instead the
spanish ones.

Moreover, I found another paper: Design and analysis of equivalence
clinical trials via the SAS system, a SUGI paper and I found the
following formula:

m = 4(Z(1-alfa)+Z(1-beta))^2 / ln(delta)^2

Also surprising for me, it is only necessary to set delta (the non-
inferiority range for the hazard ratio) to obtain the sample size.
Z(x) is the inverse normal standard distribution.

I made my own excel (more understandable for me) and I obtained the
same results as using PASS.

Best regards,
Juan Vicente Torres

On May 6, 9:30 pm, "juanvte... at (no spam) googlemail.com"



juanvte... at (no spam) googlemail.com> wrote:
On May 6, 6:16 pm, "Kevin E. Thorpe" <kevin.tho... at (no spam) utoronto.ca> wrote:

On May 6, 10:30 am, "juanvte... at (no spam) googlemail.com"

juanvte... at (no spam) googlemail.com> wrote:
Dear all,

I am trying to calculate the sample size necessary to proof non-
inferiority for two survival curves. I have downloaded the 7-days
trial version of the PASS software but it does not work. I get the
following message: "A value of 0 is out of range", I have tried the
example that I find in the PASS documentation and even with this data
it does not work.

I have thought also about calculating by myself this sample size but I
found only a little number of papers regarding this problem, all of
these too difficult to understand for me.

Could some body guide me about how to obtain this sample size?? Maybe
statisticians are used to use some approximation as it is usual in
these calculations?

Thanks in advance for your help!!

Sincerely,
Juan Vicente Torres.

I found this paper to be helpful.

"Sample Size Computation for Two-sample Noninferiority Log-rank Test"
by Jung, Kang,
McCall and Blumenstein in J. of Biopramaceutical Statistics, 15:
969-979, 2005

Thank you very much Bruce and Kevin:

I used both recommendations and I got the solution by means these two
different paths.

Jerry from PASS support replied my e-mail and the solution was as easy
as setting my windows xp to run with english settings instead the
spanish ones.

Moreover, I found another paper: Design and analysis of equivalence
clinical trials via the SAS system, a SUGI paper and I found the
following formula:

m = 4(Z(1-alfa)+Z(1-beta))^2 / ln(delta)^2

Also surprising for me, it is only necessary to set delta (the non-
inferiority range for the hazard ratio) to obtain the sample size.
Z(x) is the inverse normal standard distribution.

I made my own excel (more understandable for me) and I obtained the
same results as using PASS.

Best regards,
Juan Vicente Torres

After further study I am amazed with the above equation and its
implications. This equation is telling me that the sample size
calculation for a non-inferiority test with survival data depends only
of the non-inferiority range (delta). It means that the sample size is
independent to the observed Hazard Rate or possible deviations,
parameters which are very important in other non-inferiority tests as
for example non-inferiority for two means.

Does somebody know if these results are correct? I think that it is a
hard affirmation and I would to be confident before do it.

Thanks again for your answer Kevin!

I tried to get the paper you suggested but in my university this
journal is not available. The results that I obtain with this formula
are almost identically to the results obtained by PASS software.
Therefore, I can guess that this calculations are not incorrect.
Another surprising thing was that PASS asks for a reference hazard
rate but it is not used. I tried with different hazard rates but I
obtained the same results.

Maybe Bruce can add something!

On May 9, 10:00 am, "juanvte... at (no spam) googlemail.com"

juanvte... at (no spam) googlemail.com> wrote:
Thanks again for your answer Kevin!

I tried to get the paper you suggested but in my university this
journal is not available. The results that I obtain with this formula
are almost identically to the results obtained by PASS software.
Therefore, I can guess that this calculations are not incorrect.
Another surprising thing was that PASS asks for a reference hazard
rate but it is not used. I tried with different hazard rates but I
obtained the same results.

Maybe Bruce can add something!

I have nothing to add. Sorry.

You might want to ask Jerry Hintze (from PASS) to explain why you're
getting the same results when you enter different reference hazard
rates. His answer to that could be interesting.

--
Bruce Weaver
bwea... at (no spam) lakeheadu.cawww.angelfire.com/wv/bwhomedir
"When all else fails, RTFM."