Rudolf Koehne-Volland Head of Biostatistics & CEO ...worldbil.com/CTEF/wp-content/uploads/2016/10/Rudolf-Koehne.pdf · Head of Biostatistics & CEO Metronomia Clinical Research GmbH

Best practice in dose finding studies

Rudolf Koehne-Volland

Head of Biostatistics & CEO

Metronomia Clinical Research GmbH

13-Oct-2016 2

Best practice

s in dose

finding studies || CTEF Munich

2016

Best practice in dose finding studies

Agenda

What is the problem?

Dose finding requirements

Dose finding approaches

Challenges

Conclusion

13-Oct-2016 3 Best practices in dose finding studies || CTEF Munich 2016

Success rates in drug development Trends over time

13-Oct-2016 Best practices in dose finding studies || CTEF Munich 2016 4

Source: CMR International 2015 Global R&D Performance Metrics Programme

Attrition reasons 2013 & 2014


In many cases homework was not done in Phase II

Not learning as much as necessary about a new treatment

Not generating a good data basis or know-how for planning Phase III

Reason Phase II Phase III & Reg review

Efficacy 49% 53%

Safety 31% 25%

Strategy 19% 5%

Source: Cortellis Competitive Intelligence & Thomson Reuters Integrity

2000 − 2012 around 16% of FDA NME filings failed to be approved first time due to “uncertainties related to dose selection” (Sacks et al., 2014)

21% of drugs filed in 1980 − 1999 had post-approval dosage changes or were removed from sale (Cross et al., 2002)

Poor dose finding and failure


Source: Cortellis Competitive Intelligence & Thomson Reuters Integrity

Characterizing dose response for safety and efficacy is critical

dose too high safety and tolerability problems are likely to result

dose too low may lead to inadequate efficacy

Therapeutic window: doses that are both efficacious and safe

Optimal dose: a trade-off between efficacy and safety


Dose finding challenge


clinically relevant effect

efficacy

dose d1

Response

safety threshold

safety

dose d2

Pr(

dose t

ole

rate

d)

1

Agenda




Challenges

Conclusion


Paracelsus (1493 – 1541)


“All things are poison and

nothing is without poison,

only the dose permits something not to be poison.”

ICH E4 Dose Response Information to Support Drug Registration


Selecting the right dose

Paracelsus: ~ 500 years ago

ICH E4: >20 years ago

2000 − 2012: ~ 16% of FDA NME filings failed due to “uncertainties related to dose selection”

Are we doing ok?

We are probably not doing ok!



Source: http://www.ema.europa.eu/docs/en_GB/document_library/Report/2015/04/WC500185864.pdf

EMA Report from Dose Finding Workshop 2014









Times are changing

Better dose finding is required by …

Regulatory guidance

Medical & scientific needs

Drug development business

Biotech business model

potential pharma investors look more carefully into dose finding

… and even request certain study designs and methods


Agenda




Challenges

Conclusion


Dose finding approaches in Phase I

Most common objective

Identification of the maximum tolerated dose (MTD)

Most frequently used study design

3+3 design (and its modifications) are used since decades

Alternatives

Accelerated titration design

Adaptive dose finding designs incl. Bayesian model based methods (e.g. continual reassessment method)



Source: http://meetinglibrary.asco.org/sites/meetinglibrary.asco.org/files/edbook/156/pdf/EdBookAM2015353.pdf

Agenda




In Phase I

In Phase II

Challenges

Conclusion


Multiple, parallel dose groups

Placebo controlled, sometimes also with active controll

Randomized

Phase II dose finding studies

Standard study design


Establish evidence of a dose-response relationship (proof of concept)

Is there a dose related effect at all?

What is the maximum effect size?

What is the nature of the dose response shape?

Go / no go decision to Phase III

If “go”: Selection of optimal dose for Phase III


Objectives


Pairwise comparisons

Pairwise comparisons of active doses against placebo (e.g. ANOVA-based hypothesis testing or hierarchical testing)

Drawback: Available information across all active doses is not considered efficiently

Questions

What happens in-between the doses?

What is the dose-response curve?

Do you want to rely on point estimates?


Most frequently used statistical method


Dose

*

*

* Statistically significant better than placebo, but no difference between the two





Statistical testing is normally done in Phase III and requires large sample sizes Testing in Phase II with much lower sample sizes is often not

adequately powered

Pairwise comparisons in dose finding trials are open to by chance findings – particularly if not adequately powered

Often only a small number of doses is studied


Issues with pairwise comparisons


Model based analysis!

Modeling provides more information

Smoothens dose estimates

Interpolation between doses

Confidence intervals for quantities of interest


How to improve?

Dose

Sta

nd

ard

ize

d d

iffe

ren

ce

to

pla

ce

bo


True shape of dose response model is typically unknown

Selecting a working model may have a substantial impact on the final dose estimate

Model selection using observed data needs to account for the inherent uncertainty

Useful to have a unified approach combining the advantages of dose response signal testing and modeling

MCP-Mod: Combination of

Multiple Comparison Procedures and Modelling


How to improve?


Dose

Sta

nd

ard

ize

d d

iffe

ren

ce

to

pla

ce

bo


MCP-Mod approach


Source: http://www.ema.europa.eu/docs/en_GB/document_library/Regulatory_and_procedural_guideline/2014/02/WC500161027.pdf

MCP step

Testing dose response signal based on a set of candidate models

Out of the significant models, the model with the best fit is selected

Mod step

The dose response curve is estimated using the selected model (or by model averaging of significant models)

Relevant parameters like ED50 (dose that achieves 50% of the maximum achievable effect) or the Minimal effective dose (MED) can be derived


MCP-Mod approach



MCP-Mod approach


Source: http://www.ema.europa.eu/docs/en_GB/document_library/Regulatory_and_procedural_guideline/2014/02/WC500161027.pdf

MCP-Mod

Example


Dose

MCP-Mod

Example


MCP-Mod

Example


Dose

MCP-Mod

Example


Moderate sample sizes allow well powered MCP-Mod studies for proof of concept, i.e. for establishing a dose response relationship

A meaningfull precision for dose estimates (e.g. the MED) can only be achieved with higher sample sizes

There is nothing like a free lunch

MCP-Mod is clearly preferable to pairwise comparisons

However, the full potential is associated with higher sample sizes

MCP-Mod

Points to consider


Planning is important and requires a joint effort of clinicians, pharmacologists and statisticians

What are candidate dose response curves?

What is the dose range, e.g. the maximum dose to be included?

Which doses and number of doses shall be studied?

What is the minimal clinically important effect?

What precision is expected, e.g. for the MED?

Involves simulation studies

Takes more time

MCP-Mod

Points to consider


Agenda




Challenges

Conclusion


Implementing proper dose finding studies

Challenges

Time and financial pressure

Lack of experience with methods

Implementation and statistical analysis



Challenges

Create awareness of importance of scientifically sound and meaningful dose finding studies

Overcome adherence to what has been done in the past

Regulatory guidances

Examples



Challenges

Convince your team that its worth the additional effort

Much better decision making

Better regulatory acceptance

Better treatments

„If you buy cheaply (in Phase II), you pay dearly (in Phase III)!”


Agenda




Challenges

Conclusion


Conclusions

Phase II dose finding studies should be designed and analysed in order to get a good understanding of the dose response relationship

More dose levels and cover the entire dose range

Use modelling techniques (e.g. MCP-Mod) rather than pairwise comparions

Proper dose finding studies allow much better decision making

We could do better in drug research, we just have to start doing it

Better methods are available and strongly supported by regulatory bodies

Better research in phase II will pay off


Thank you very much!

Rudolf Koehne-Volland

[email protected]

+49 89 829265 222


Rudolf Koehne-Volland Head of Biostatistics & CEO ...worldbil.com/CTEF/wp-content/uploads/2016/10/Rudolf-Koehne.pdf · Head of Biostatistics & CEO Metronomia Clinical Research GmbH

Documents