Bioequivalence and Bioavailability Forum

shri ☆ 2008-09-05 09:42 (6151 d 07:45 ago) Posting: # 2316 Views: 12,443	Treat significance [General Statistics] Post reply
	Dear all, how can we interprete treatment significance from CI? when CI does not include 100? plese explain in brief. thanks in advance — shri

ElMaestro
★★★

Denmark,
2008-09-05 10:50
(6151 d 06:36 ago)

@ shri
Posting: # 2317
Views: 11,095

Treat significance

Post reply

Hi Shri,

in general, when 100 (1.00) is not a part of your CI it is not a problem.
You have shown that the two products are not identical in terms of AUC or Cmax, but still they may be considered bioequivalent, provided of course that the CI is contained within the usual boundaries.

Often this is just a symptom of overpowering (the variability was lower than expected, more completers than expected or similar). This just makes the data material better, certainly not worse. Otherwise we end up in situations where the company 'should have included fewer subjects' and got a wider CI.

Denmark at some point signaled that they wouldnot accept BE if 100 (1.00) was not part of the CI. May the powers that be forgive their sins and have mercy upon them.

EM.

--
Edit: Full quote removed. Please see this post! [Jaime]

Ohlbe ★★★ France, 2008-09-05 12:27 (6151 d 05:00 ago) @ ElMaestro Posting: # 2318 Views: 11,067	Denmark Post reply
	Dear Shri and EM, For more details on this Danish requirement: a link to the page on their web site is given on the guidance page. Regards Ohlbe

ElMaestro ★★★ Denmark, 2008-09-05 14:02 (6151 d 03:24 ago) @ Ohlbe Posting: # 2319 Views: 11,082	Denmark Post reply
	Thanks Ohlbe, I must admit I thought the Danish Medicines Agency had seen the light and removed these requirements. But the headline of the Danish web page makes me wonder: Do they apply these requirements solely in relation to substitution or do the requirements apply to approval of generics generally? EM.

Ohlbe
★★★

France,
2008-09-05 14:31
(6151 d 02:56 ago)

@ ElMaestro
Posting: # 2320
Views: 11,099

Denmark

Post reply

Dear ElMaestro,

❝ Do they apply these requirements solely in relation to substitution or do the requirements apply to approval of generics generally?

This is quite unclear to me ! The title of the web page only refers to substitution. But in the first paragraph it says:
The requirements apply to both authorisation of generic medicinal products and the subsequent labelling with regard to generic substitution on the Danish market.

Then you have a table with acceptance limits for 90 % CI for marketing authorisation and for substitution, where the limits for marketing authorisation are kept to 80-125 for all drugs except immunosuppressives, and are set to 90-111 for a number of drugs for substitution. But this criterion of the 90 % CI including 100 is not mentioned in the table.

If somebody has an experience with Denmark, or has time to call the contact person mentioned at the bottom of the web page, please let us know the outcome !

Regards
Ohlbe

ElMaestro ★★★ Denmark, 2008-09-05 15:03 (6151 d 02:24 ago) @ Ohlbe Posting: # 2321 Views: 10,953	Denmark Post reply
	❝ If somebody has an experience with Denmark (...) From Shakespeare's Hamlet (Act 1/scene 4): MARCELLUS: Something is rotten in the state of Denmark. HORATIO: Heaven will direct it. Why don't we just light some candles? EM.

shri ☆ 2008-09-06 12:25 (6150 d 05:02 ago) @ Ohlbe Posting: # 2327 Views: 10,968	Denmark Post reply
	dear ohlbe, please give me information about the treat significance when CI lies below 100. means i what scence we make conclusion from CI about treat significance. thannks in advance. — shri

Ohlbe
★★★

France,
2008-09-06 18:43
(6149 d 22:43 ago)

@ shri
Posting: # 2329
Views: 11,014

Statistically different or clinically different

Post reply

Dear Shri,

If you have a significant treatment effect, it just means that the test product and the reference product are statistically different. If the CI lies below 100, this means that the bioavailability of the test product is statistically lower than the reference product.

Statistically different doesn't mean that the difference has any clinical relevance, I mean that the difference may not have any impact on the patients taking the drug. That's why 90 % CI are calculated, and we are not just testing for treatment effect in the ANOVA. As long as your 90 % CI lies within the acceptance limits (usually 80-125), the products are still considered bioequivalent.

Regards
Ohlbe

shri ☆ 2008-09-08 08:29 (6148 d 08:58 ago) @ Ohlbe Posting: # 2333 Views: 11,022	Statistically different or clinically different Post reply
	Thanks ohlbe for your response, Its really helpful for me. If you have any information about it please provide. ok. thanks again. — shri

kumarnaidu
★

Mumbai, India,
2014-03-14 07:54
(4135 d 08:32 ago)

@ shri
Posting: # 12618
Views: 9,346

Statistically different or clinically different

Post reply

Hi all,
I have one study result with me of diuretic class drug. The study was refrence replicated study with N=50. The result obtained is given below
Cmax - 91.41(83.04 - 100.62), CV=34.25, Power=96.82
AUC(0 to t)- 90.14(85.78 - 94.72) , CV=17.31, Power=100.00

As we understand, in Nordic countries, it is a requirement that the interval cover 100% in order to obtain generic substitution. In view of this requirement and our study results what should be the possible justification we can present to the agency if query come.

—
Kumar Naidu

ElMaestro
★★★

Denmark,
2014-03-14 08:10
(4135 d 08:16 ago)

@ kumarnaidu
Posting: # 12619
Views: 9,304

Statistically different or clinically different

Post reply

Hi Kumarnaidu,

❝ As we understand, in Nordic countries, it is a requirement that the interval cover 100% in order to obtain generic substitution. In view of this requirement and our study results what should be the possible justification we can present to the agency if query come.

You have shown BE by ordinary standards. Just submit the dossier. DK can only theoretically kill it if they are the RMS in a DCP/MRP. But I doubt they will try.

—
Pass or fail!
ElMaestro

kumarnaidu ★ Mumbai, India, 2014-03-14 13:36 (4135 d 02:50 ago) @ ElMaestro Posting: # 12622 Views: 9,312	Statistically different or clinically different Post reply
	Hi ElMaestro, Thanks a lot for your comments. — Kumar Naidu

Helmut
★★★

Vienna, Austria,
2014-03-14 14:31
(4135 d 01:56 ago)

@ kumarnaidu
Posting: # 12623
Views: 9,314

Post hoc power #666

Post reply

Hi Kumar,

❝ As we understand, in Nordic countries, […]

Only in Denmark.

As stated by ElMaestro – submit the study. I would suggest to remove “power” from the report for two reasons:

Educational. Post hoc power is meaningless. Don’t submit nonsense.
Your numbers are likely wrong. How did you derive them?
Assuming a fully replicated design I got:
- C_max (reference-scaled) 95.33%, unscaled 87.95%
- AUC_0–t unscaled 99.94%
If the design was a partial replicate, “power” would be lower.

—
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

d_labes
★★★

Berlin, Germany,
2014-03-14 14:53
(4135 d 01:34 ago)

@ kumarnaidu
Posting: # 12624
Views: 9,234

Post hoc power - power of which test?

Post reply

Dear Kumar Naidu,

❝ I have one study result with me of diuretic class drug. The study was refrence replicated study with N=50. The result obtained is given below

❝ Cmax - 91.41(83.04 - 100.62), CV=34.25, Power=96.82

❝ AUC(0 to t)- 90.14(85.78 - 94.72) , CV=17.31, Power=100.00

I'm not a friend of calculating post-hoc power :no:

. Search the forum to see why.

But if you nevertheless insist to do so, I think it should be the right power, namely the power of the BE test (90% confidence intervals contained in the acceptance range 80-125 or, alternatively, two one-sided t-tests TOST). And sorry, I couldn't reproduce your numbers in calculating that power!

Using PowerTOST (not considering the unbalancedness within sequence groups) I obtain:

#Cmax

power.TOST(CV=0.3425, theta0=0.9141, n=50, design="2x3x3")

#[1] 0.7421506

# AUC

power.TOST(CV=0.1731, theta0=0.9014, n=50, design="2x3x3")

# [1] 0.9902765

If your numbers are from Phoenix/WinNonlin see this thread to find out that the power values from that software are statistically nonsense.

—
Regards,

Detlew

kumarnaidu ★ Mumbai, India, 2014-03-15 11:56 (4134 d 04:30 ago) @ d_labes Posting: # 12633 Views: 9,277	Post hoc power - power of which test? Post reply
	❝ If your numbers are from Phoenix/WinNonlin Hi D_labes and Helmut thanks for your valuable comments. All above calculations were provided by CRO and I think they use SAS or WinNonlin. — Kumar Naidu

Treat significance [General Sta­tis­tics]

Treat significance

Denmark

Denmark

Denmark

Denmark

Denmark

Statistically different or clinically different

Statistically different or clinically different

Statistically different or clinically different

Statistically different or clinically different

Statistically different or clinically different

Post hoc power #666

Post hoc power - power of which test?

Post hoc power - power of which test?

Treat significance [General Statistics]