Bioequivalence and Bioavailability Forum • Freeze-thaw studies

martin
★★

Austria,
2013-09-04 13:29
(4676 d 06:26 ago)

Posting: # 11430
Views: 6,262

Freeze-thaw studies [Bioanalytics]

Dear all,

I have a off-topic general statistical question regarding analysis of freeze-thaw studies.

I have a freeze-thaw experiment where two lots are freezed and thawed for five times. For each freeze-thaw step three replicates are available. No ‘acceptance’ criteria from a biological point of view are available.

Does somebody know how this kind of studies is usually analyzed? Are there
any guidelines and/or papers available? Any suggestions are highly appreciated!

best regards

martin

Edit: Category changed. [Helmut]

Helmut
★★★

Vienna, Austria,
2013-09-04 15:27
(4676 d 04:27 ago)

@ martin
Posting: # 11432
Views: 5,296

Freeze-thaw studies

Post reply

Hi Martin!

❝ I have a freeze-thaw experiment where two lots are freezed and thawed for five times. For each freeze-thaw step three replicates are available. No ‘acceptance’ criteria from a biological point of view are available.

Do you know the initial concentration?

❝ Does somebody know how this kind of studies is usually analyzed? Are there

❝ any guidelines and/or papers available?

Nope. As en entry point see EMA’s bioanalytical GL. There the limits are the ones common for (in)accuracy / (im)precision: ±15%/15% (or ±20%/20% at the LLOQ).

—
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

martin
★★

Austria,
2013-09-04 16:03
(4676 d 03:51 ago)

@ Helmut
Posting: # 11433
Views: 5,404

Freeze-thaw studies

Post reply

Dear Helmut,

Nope, no "true" initial concentration available (just 3 replicates prior first freezing step).

I thought on linear regression and test the slope to be different for zero. However, I do not like this approach for obvious reasons ;-)

. Another approach may be to calculate the 95% CI for the mean at step 0 (kind of plausibility interval) and check if CI for the model-predictions is within that interval.

What do you think?

Best regards

PS.: Here is an artificial data set for illustration


# generate example data # 

set.seed(54757) 

step <- c(rep(0,3),rep(1,3), rep(2,3), rep(3,3), rep(4,3), rep(5,3))

x <- rnorm(n=length(step), mean=0, sd=1)

data <- data.frame(lot='Lot 1', step=step, x=x)



# linear regression # 

mod <- lm(x ~step, data=data)

ci95 <- t.test(subset(data, step==0)$x)$conf.int 

plot(x ~ step, data=data, ylim=c(-4, 4))

for(i in 1:3){points(predict(mod, interval='confidence')[,i]~data$step, type='l')} 

abline(h=ci95, lty=3)

Helmut
★★★

Vienna, Austria,
2013-09-04 16:43
(4676 d 03:11 ago)

@ martin
Posting: # 11435
Views: 5,186

Freeze-thaw studies

Post reply

Hi Martin,

❝ Nope, no "true" initial concentration available (just 3 replicates prior first freezing step).

OK, expected that.

❝ I thought on linear regression and test the slope to be different for zero. However, I do not like this approach for obvious reasons ;-) .

Yes, doesn’t make sense.

❝ Another approach may be to calculate the 95% CI for the mean at step 0 (kind of plausibility interval) and check if CI for the model-predictions is within that interval.

❝ What do you think?

Much better. See also the Hsin-ya’s and Yung-jin’s stab for R.

—
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