Non-Inferiority Assessment of Patch Adhesion Non-Inferiority - - PDF document

5/21/2015 1

1

Non-Inferiority Assessment of Patch Adhesion and Dermatology Irritation Evaluation

Non-Inferiority Assessment of Patch Adhesion and Dermatology Irritation of Post-Market Change and Generic Drug Evaluation

• Presented by Mark Liu, Senior Director of Biostatistics

» Mylan Pharmaceuticals

• Coauthored by Russ Rackley, VP of Global PK

» Mylan Pharmaceuticals

2

Non-Inferiority Assessment of Patch Adhesion and Dermatology Irritation Evaluation

Disclaimer: This presentation reflects the views

• f the authors and not necessarily the Mylan’s
• fficial standing on the issues.

5/21/2015 2

3

Outlines

• Overview of current Dermal Irritation Evaluation

– Study Design and Irritation Scales – Current Statistical Method

• Simulated Examples will show the problem of the current method
• Our proposed Statistical method to overcome the problem.
• Adhesion issues and an easy solution
• Conclusions

Assessment of Irritation

5/21/2015 3

5

FDA Irritation Scales

Dermal Response

Scale Irritation No evidence of irritation 1 Minimal erythema, barely perceptible 2 Definite erythema, readily visible; or minimal edema; or minimal papular response 3 Erythema and papules 4 Definite edema 5 Erythema, edema, and papules 6 Vesicular eruption 7 Strong reaction spreading beyond test (i.e., application) site

Other Effects

Scale Appearance A (0) Slighty glazed appearance B (1) Marked glazed appearance C (2) Glazing with peeling and cracking F (3) Glazing with fissures G (3) Film of dried serous exudates covering all or part of the patch site H (3) Small petechial erosions and/or scabs

• Typically evaluated for multiple wear periods over 21 days.
• Low scores associated with good performance.
• Sum of the Dermal Response and Other Effects as final analysis

value.

6

Current OGD Statistical Method

• The analyses for cumulative irritation are intended to demonstrate

that ’the upper bound of the one-sided 95% CI of the mean Test score minus 1.25 times the mean RLD score must be less than or equal to 0’. 95% 1.25 ∗ 0 rearranged… 95%

• 0.25
• The rearrangement demonstrates that the result of this metric relative

to acceptance criterion can become excessively stringent as the mean RLD score approaches zero.

5/21/2015 4

7

Hypersensitivity of the Current Assessment Criteria

• In situations of low or minimal irritation response, the margins allowed

are far lower than would be permitted relative to products with worse performance (i.e. scores >1).

0.5 1 1.5 2 2.5 0.5 1 1.5 2 2.5 Test - Mean Irritation Reference - Mean Irritation Line of Identity FDA Limit

8

Hypersensitivity of the Current Assessment Criteria

0.1 0.2 0.3 0.4 0.5 Non‐Inferiority Margin

Absolute Margins with Different Levels of Mean Scores

2 1.5 1 0.5 0.2 0.1

5/21/2015 5

9

Issues with current Method

• For assessment of irritation of Transdermal product, issues exist with

current recommended scoring system and statistical methodology such that a RLD with good clinical performance will predominately have low scores which creates a condition that is prohibitive to generic entry.

• This situation flourishes at levels of irritation that are clinically negligible.

And while infinitesimally small differences may exist between a Test and Reference product, they occur at levels for which clinical safety is self- evident.

• The overly stringent nature of OGD’s method will be demonstrated via

simulated examples.

10

Assessment of Irritation Simulated Examples – general rules

• Five separate studies were used to generate five

simulated examples.

• Four of these five studies have more than 200

subjects completed the study, the other one has more than 100 subjects completed the study.

• One Bootstrap simulation sample was selected from

each study with desired mean values.

5/21/2015 6

Assessment of Irritation Simulated Example One

12

Assessment of Irritation Simulated Example One

• 21-day continuous cumulative irritation study
• 3.5 days per patch application on same skin site, for

total 6 patches per treatment

• 200 subjects total with simultaneous wear of both Test

and RLD patches.

5/21/2015 7

13

Assessment of Irritation Example One – simulated dataset

Irritation Values

Patch number 1 2 3 4 5 6 194 194 188 181 177 152

1

6 3 10 16 16 39

2

1 1 5 7

3

2 2 2 2 2

Total

200 200 200 200 200 200

Irritation Values

Patch number 1 2 3 4 5 6 192 192 192 178 171 159

1

8 8 6 20 15 32

2

2 2 14 4

3

5

Total

200 200 200 200 200 200

Test formulation RLD formulation

14

Assessment of Irritation Example One – Summary Statistics

Summary Statistics on cumulative Means

Treatment N Mean Std Dev CV Range Test 200 0.123 0.325 264% 0 - 2.67 RLD 200 0.123 0.227 225% 0 - 1.67

5/21/2015 8

15

Assessment of Irritation Example One - results

• Statistical results using OGD’s method

– Upper one-sided 95% confidence interval of cumulative adhesion for the mean Test adhesion score minus 1.25 times the mean Reference adhesion score should not exceed zero.

Test Reference Parameter Upper 95% CI Criteria Pass/Fail 0.123 0.123 Test – 1.25*Ref 0.0016 ≤ 0 Fail

16

Assessment of Irritation Example One – Bootstrap simulation

Number of Subjects Passing Rate 36 13% 60 18% 120 30% 200 44% 400 72%

All based on 2000 simulation samples

RLD mean irritation ~ 0.125

5/21/2015 9

Assessment of Irritation Simulated Example Two

18

Assessment of Irritation Simulated Example Two

• 21-day continuous cumulative irritation study
• 3.5 days per patch application on the same skin site,

total 6 patches per treatment

• 200 subjects total with simultaneous ware of both

Test and RLD

5/21/2015 10

19

Assessment of Irritation Example Two – simulated dataset

Test formulation RLD formulation

Irritation Values Patch number 1 2 3 4 5 6 167 176 171 165 162 158 1 27 14 20 24 24 29 2 6 10 9 9 9 5 3 3 6 4 5 6 1 1 1 7 1 1 1 200 200 200 200 200 200 Irritation Values Patch number 1 2 3 4 5 6 163 177 174 159 163 146 1 28 16 18 32 26 41 2 9 7 7 8 6 6 3 4 6 4 1 5 1 6 1 7 1 200 200 200 200 200 200 20

Assessment of Irritation Example Two - summary

Summary Statistics on cumulative Means

Treatment N Mean Std Dev CV Range Test 200 0.25 0.552 220% 0 - 4.167 RLD 200 0.25 0.469 188% 0 - 4.000

5/21/2015 11

21

Assessment of Irritation Example Two – Bootstrap Simulation

Number of Subjects Passing Rate 36 32% 60 35% 120 41% 200 50% 400 74%

All based on 2000 simulation samples

RLD mean irritation ~ 0.25

Assessment of Irritation Simulated Example Three

5/21/2015 12

23

Assessment of Irritation Simulated Example Three

• 21-day continuous cumulative irritation study
• 2-days, 2-days, and 3-days per week patch

application on same site, total 9 patches per treatment

• 200 subjects total with simultaneous wear of both

Test and RLD

24

Assessment of Irritation Example Three – simulated dataset

Test formulation RLD formulation

Irritation Values Patch number 1 2 3 4 5 6 7 8 9 110 14 4 1 86 183 188 170 135 3 18 9 2 4 3 8 16 30 131 106 63 56 3 17 39 55 67 4 12 30 14 6 5 4 5 30 43 53 51 6 1 7 7 2 5 1 1 8 5 5 5 5 200 200 200 200 200 200 200 200 200 Irritation Values Patch number 1 2 3 4 5 6 7 8 9 103 5 9 2 2 2 1 85 166 162 129 92 63 22 14 26 2 12 29 29 54 61 68 85 70 58 3 9 8 12 2 9 4 6 37 51 79 81 81 5 14 25 6 1 3 3 3 7 2 5 7 7 7 8 200 200 200 200 200 200 200 200 200

5/21/2015 13

25

Assessment of Irritation Example Three – Statistical summary

Summary Statistics on cumulative Means

Treatment N Mean Std Dev CV Range Test 200 2.00 0.841 42% 0 – 5.444 RLD 200 2.00 0.773 39% 0 – 4.667

26

Assessment of Irritation Example Three – Statistical results

• Statistical results using OGD’s method

– Upper one-sided 95% confidence interval of cumulative adhesion for the mean Test adhesion score minus 1.25 times the mean Reference adhesion score should not exceed zero.

Test Reference Parameter Upper 95% CI Criteria Pass/Fail 2.000 2.001 Test – 1.25*Ref

• 0.4402

≤ 0 Pass

5/21/2015 14

27

Assessment of Irritation Example Three – Bootstrap simulative

Number of Subjects Passing Rate 36 100% 60 100%

All based on 2000 simulation samples

RLD mean irritation ~ 2.0

28

Assessment of Irritation Summary results of all Examples

All estimates based on 2000 simulation samples each

Number of Subjects RLD Means and Test mean 0.123 0.25 0.50 1.00 2.00 36 13% 32% 60% 77% 100% 60 18% 35% 79% 93% 120 30% 41% 96% 100% 200 44% 50% >99% 400 72% 74% 100% Boot-strap simulation summary of power, depending on mean RLD score

5/21/2015 15

29

Assessment of Irritation Summary results of all Examples

All based on 2000 simulation samples

13 18 30 44 72 32 35 41 50 74 60 79 96 100 100 77 93 100 100 100 100 100 100 100 100 10 20 30 40 50 60 70 80 90 100 36 60 120 200 400

Statistical Power Graph

0.123 0.25 0.5 1 2

Assessment of Irritation Proposed alternative method

5/21/2015 16

31

Analysis of Irritation Proposed alternative method

• OGD’s current statistical method for analyzing cumulative irritation was

found to be overly sensitive (or stringent) when the RLD data trended toward zero.

• Without altering the current irritation scale, an alternative method is

proposed.

• We have analyzed the data, according to this method.
• Compared the OGD analysis to the proposed alternative method.

32

Analysis of Irritation Criteria Adjusted by Threshold

• If Reference mean score is

below a clinically relevant threshold of 1, then the upper statistical bound would be based

• n an absolute value of 0.25 (i.e.

fixed limit based on 25%*1.0).

• If the Reference mean were to

be greater than one, then OGD recommended limit, based on 25% of the Reference mean, would apply.

• 95%
• 0.25

5/21/2015 17

33

Example one OGD and alternative method Results comparison

Number of Subjects Passing Rate per OGD method Passing Rate Per Alternative method 36 13% 100% 60 18% 100% 120 30% 100% 200 44% 100% 400 72% 100% All based on 2000 simulation samples

RLD mean irritation ~ 0.125

34

Example one OGD and Alternative method Results comparison

• When Test are 10% more irritating

Number of Subjects Passing Rate per OGD method Passing Rate per Alternative method 36 10% >99% 60 13% 100% 120 19% 100% 200 26% 100% 400 46% 100% All based on 2000 simulation samples

For RLD mean irritation ~ 0.125

5/21/2015 18

35

Example Two FDA and Mylan method Results comparison

Number of Subjects Passing Rate per FDA method Passing Rate per Mylan method 36 32% 84% 60 35% 96% 120 41% >99% 200 50% 100% 400 74% 100% All based on 2000 simulation samples

RLD mean irritation ~ 0.25

36

Example Two FDA and Mylan method Results comparison

• When Test are 10% more irritating

Number of Subjects Passing Rate per FDA method Passing Rate per Mylan method 36 19% 93% 60 21% >99% 120 28% 100% 200 36% 100% 400 58% 100% All based on 2000 simulation samples

For RLD mean irritation ~ 0.25

5/21/2015 19

Assessment of Adhesion Evaluation of Current Guidance

38

Analysis of Adhesion

• OGD’s current statistical method for analyzing cumulative adhesion was

found to be overly sensitive (or stringent) when the RLD data trended toward zero (this was found to be problematic for adhesion analysis when OGD’s adhesion scale is used).

• Since an alternative scoring scale can be implemented, so is the

statistical method. We found EMA’s current guidance is reasonable and practical, in which the scale is directly related to area of adhesion.

• If we have to keep the current OGD’s scales, then alternative method

can be implemented, similar to Irritation analysis.

5/21/2015 20

39

Analysis of Adhesion FDA Adhesion Scale

Adhesion Scale Specified in the FDA guidance Score Adhesion

≥ 90% adhered (essentially no lift off the skin)

1

≥ 75% to < 90% adhered (some edges only lifting off the skin)

2

≥ 50% to < 75% adhered (less than half of the patch lifting off the skin)

3

> 0% to < 50% adhered but not detached (more than half of the patch lifting off the skin without falling off)

4

0% adhered - patch detached (patch completely off the skin)

• Typically evaluated at multiple time-points over one wear period.
• Low scores associated with good performance.

40

Adhesion Evaluation Hypothetical Example

• Test product has essentially best possible performance, yet the data

fails to demonstrate non-inferiority, when OGD’s method and criteria are applied.

Summary of adhesion results (N=252)

Scores 100% scales Scores FDA Scales 36 Subjects with 7 Observation each Treatment Test RLD 100% 252 251 85% 1 1

5/21/2015 21

41

Adhesion Evaluation Hypothetical Example

• OGD’s scale (per current guidance)

– OGD method: Upper one-sided 95% confidence interval of cumulative adhesion for the mean Test adhesion score minus 1.25 times the mean Reference adhesion score should not exceed zero. Test Reference Parameter Upper 95% CI Criteria Pass/Fail 0.0000 0.0040 Test – 1.25*Ref 0.0026 ≤ 0 Fail Test Reference Parameter Upper 95% CI Criteria Pass/Fail 0.0000 0.0040 Test – Ref 0.0027 ≤ 0.25 Pass – Alternative Method: Upper one-sided 95% confidence interval of cumulative adhesion for the mean Test adhesion score minus the mean Reference adhesion score should not exceed 0.25.

42

Adhesion Evaluation Hypothetical Example

• When a method is based on 100% scales, we find the European

method has a similar interpretation, to what we’ve proposed:

– Lower boundary of two-sided 90% confidence interval of cumulative adhesion for the mean Test adhesion score minus the mean Reference adhesion score should be more than -10.00.

Test Reference

Parameter

Lower 95% CI Criteria Pass/Fail 100.00 99.96

Test – Ref

• 0.0274

≥ -10 Pass

The European method evaluated the last observation of the adhesion period, whereas our US proposed method evaluates the cumulative mean adhesion over the adhesion period.

5/21/2015 22

Conclusions

44

Conclusion

• OGD’s current method for assessing non-inferiority of Irritation and

Adhesion is excessively stringent when both the Test product and the RLD have low irritation or good adhesion properties.

• Examples and Simulations provided demonstrate that this has led to

the inability to demonstrate non-inferiority even they are essentially same.

• This current statistical methodology creates a unnecessary conflict

between statistical and clinical relevance.

• We propose alternatives, and believes the FDA Statisticians can

approve a practical statistical method that continues to satisfy the requirements of the applicable statutes and regulations for demonstration of bioequivalence.