pierre gressens modèles animaux : intérêts et limites
TRANSCRIPT
Pierre Gressens
Modèles animaux :
Intérêts et limites
• Neuroprotective strategies as an example
• False positive studies : what should we learn from
them ?
• True negative studies : why are they important ?
• False negative studies : what do they tell us ?
Focus & plan
• Adult stroke field : huge failure in clinical trials with drugs
protective in animal models (except for tPA)
False positive studies
• Adult stroke field : huge failure in clinical trials with drugs
protective in animal models (except for tPA)
• Pessimistic interpretation : animal models not predictive of
humans
False positive studies
• Adult stroke field : huge failure in clinical trials with drugs
protective in animal models (except for tPA)
• Pessimistic interpretation : animal models not predictive of
humans
• Scientific approach : why ?
False positive studies
• Animal studies
- “wrong” compound : PK, PD, BD, BBB, …, wrong target, …
False positive studies
• Animal studies
- “wrong” compound : PK, PD, BD, BBB, …, wrong target, …
- “wrong” design : blinded, randomized, stats, controls (KOs,
behavior)
False positive studies
• Animal studies
- “wrong” compound : PK, PD, BD, BBB, …, wrong target, …
- “wrong” design : blinded, randomized, stats, controls (KOs,
behavior)
- confounding variables
False positive studies
• Animal studies
- “wrong” compound : PK, PD, BD, BBB, …, wrong target, …
- “wrong” design : blinded, randomized, stats, controls (KOs,
behavior)
- confounding variables
- T°
- time of the day, season, …
- sex
- maternal stress, maternal care, maternal feeding, …
- person performing model, tests, analyses, …
False positive studies
Temperature
Thoresen et al., unpublished data
0
1
2
3
4
Mean
Glo
bal P
ath
olo
gy S
core
Control 32°C 37°C 38°C 39°C
Post-HI Recovery Temperature
3
Time of the day
Bednarek & Gressens, unpublished data
Maternal stress
Rangon et al., J Neurosci 2007
The ALS lesson
Scott et al., ALS 2008
SOD1 mutant = ALS modelRiluzole protection(increased lifespan)
The ALS lesson
Scott et al., ALS 2008
SOD1 mutant = ALS modelRiluzole protection(increased lifespan)
5429 miceRiluzole efficacy
computer analysis
The ALS lesson
Scott et al., ALS 2008
SOD1 mutant = ALS modelRiluzole protection(increased lifespan)
5429 miceRiluzole efficacy
computer analysis
confounding biological factors
optimal study design
The ALS lesson
Scott et al., ALS 2008
SOD1 mutant = ALS modelRiluzole protection(increased lifespan)
5429 miceRiluzole efficacy
computer analysis
confounding biological factors
optimal study design
optimal study design
8 « protective » drugswell-powered study
The ALS lesson
Scott et al., ALS 2008
SOD1 mutant = ALS modelRiluzole protection(increased lifespan)
5429 miceRiluzole efficacy
computer analysis
confounding biological factors
optimal study design
optimal study design
8 « protective » drugswell-powered study
no effect on lifespan !!!
The ALS lesson
Scott et al., ALS 2008
SOD1 mutant = ALS modelRiluzole protection(increased lifespan)
5429 miceRiluzole efficacy
computer analysis
confounding biological factors
optimal study design
optimal study design
8 « protective » drugswell-powered study
no effect on lifespan !!!
? previous studies = biased
• Animal studies
- “wrong” compound : PK, PD, BD, BBB, …, wrong target, …
- “wrong” design : blinded, randomized, stats, confounding
variables
- healthy vs sick animals
False positive studies
Impact of systemic inflammation on neuroprotection
Gressens et al., Eur J Pharm 2008Gressens et al., unpublished
Impact of systemic inflammation on neuroprotection
Gressens et al., Eur J Pharm 2008Gressens et al., unpublished data
Impact of systemic inflammation on neuroprotection
Gressens et al., Eur J Pharm 2008Gressens et al., unpublished data
Impact of systemic inflammation on neuroprotection
Gressens et al., Eur J Pharm 2008Gressens et al., unpublished data
• Animal studies
- “wrong” compound : PK, PD, BD, BBB, …, wrong target, …
- “wrong” design : blinded, randomized, confounding variables
- healthy vs sick animals
• Human clinical trials
- too “stringent” outcome
- death vs survival of impaired patients
False positive studies
The catch 22
0
Damage
Death
Insult
Neuroprotection
Protective effect on mortality?
• allow to rule out potential pathways and targets
True negative studies
• allow to rule out potential pathways and targets
… if studies correctly performed !
• good rationale (hypothesis to test)
• good design :
- sufficient power !!!
- multiple models
- multiple species
True negative studies
NADPH oxidase
• oxidative stress is deleterious for the brain
• inhibition of NADPH oxidase = neuroprotective in
adults
• ? good target in neonates
NADPH oxidase: not a good target in neonates
Doverhag et al., NBD 2008
NADPH oxidase: not a good target in neonates
Doverhag et al., NBD 2008
• what do they tell us ?
False negative studies
• what do they tell us ?
• different case scenarios …
False negative studies
• power calculation taking into account
- variability of procedure
- variability of outcome variable
Methodological biases
Power
(n=8/group)
Power
p = 0.0764(n=8/group)
Power
p = 0.0764(n=8/group) (n=16/group)
Power
p = 0.0764(n=8/group)
p = 0.0088(n=16/group)
• power calculation taking into account
- variability of procedure
- variability of outcome variable
• appropriate outcome & readout, combined R/
Methodological biases
Cx Hipp Cer Bs.g Thal0
4
3
2
1
Bra
in p
atho
logy
sco
reHypothermia + drug
Haland et al., Pediat Res 1997
Cx Hipp Cer Bs.g Thal0
4
3
2
1
Bra
in p
atho
logy
sco
reHypothermia + drug
Haland et al., Pediat Res 1997
- optimized HT- drug effect ? (complex paradigms & analyses or -)
Cx Hipp Cer Bs.g Thal0
4
3
2
1
Bra
in p
atho
logy
sco
reHypothermia + drug
Haland et al., Pediat Res 1997
- optimized HT- drug effect ? (complex paradigms & analyses or -)
- « human efficacy » HT- effect of drug on a cooled brain
• power calculation taking into account
- variability of procedure
- variability of outcome variable
• appropriate outcome & readout, combined R/
• dose-response curve
Methodological biases
Dose-response : U-shape curve
Sokolowska et al., submitted
• power calculation taking into account
- variability of procedure
- variability of outcome variable
• appropriate outcome & readout, combined R/
• dose-response curve
• BD (BBB penetration, degradation, …), PK, species
specificities
Methodological biases
Administration schedule
Gressens et al., unpublished data
Administration schedule
Gressens et al., unpublished data
• pre-clinical drug testing ≠ search for targets
Mixed effects
• pre-clinical drug testing ≠ search for targets
• cell type : neurons vs microglia / astroglia
=> cell type-specific conditional KOs
Mixed effects
• pre-clinical drug testing ≠ search for targets
• cell type : neurons vs microglia / astroglia
=> cell type-specific conditional KOs
• timing issue : early M1 microglia vs late M2 microglia
=> time-course of lesions
Mixed effects
M1 & M2 microglia
Kigerl et al., J Neurosci 2009
M1 & M2 microglia
Kigerl et al., J Neurosci 2009
• pre-clinical drug testing ≠ search for targets
• cell type : neurons vs microglia / astroglia
=> cell type-specific conditional KOs
• timing issue : early M1 microglia vs late M2 microglia
=> time-course of lesions
• responders vs non-responders
Mixed effects
• ! p>0.05 ≠ groups are similar
= groups are not statistically different
Responders & non-responders
• ! p>0.05 ≠ groups are similar
= groups are not statistically different
Responders & non-responders
p = 0.7182
• ! p>0.05 ≠ groups are similar
= groups are not statistically different
Responders & non-responders
p = 0.7182
• ! p>0.05 ≠ groups are similar
= groups are not statistically different
Responders & non-responders
p = 0.7182
• experimental bias
• maternal care bias
• other bias
Responders & non-responders
• experimental bias
• maternal care bias
• other bias
• ? mimics some human situation
Responders & non-responders
• experimental bias
• maternal care bias
• other bias
• ? mimics some human situation
• ? mechanism : epigenetics
Responders & non-responders
• experimental bias
• maternal care bias
• other bias
• ? mimics some human situation
• ? mechanism : epigenetics
• ad hoc statistical tools to confirm R vs non-R
• mechanistic approaches
Responders & non-responders
Acknowledgements
Vincent DegosAngela KaindlCatherine VerneyVincent El GhouzziStéphane PeineauStéohanie SigautAnne-Marie BodiouValérie BiranPascal DourneauSophie LebonLeslie SchwendimannTiffen Le Charpentier
Olivier BaudRomain FontaineJérémie Dalous
Cobi HeijnenAnnemieke KavelaarsCora Nijboer
Elie SalibaGéraldine Favrais
Petra HuppiStéphane Sizonenko
Yvan van den Loojj
Bernard Thébaud
Ulrika AdenMax Winerdal
Jon Lampa
Ursula Felderhoff-MueserMatthias Keller
Olaf DammannChristiane Dammann
Wolfgang Bueter
Henrik HagbergDavid EdwardsDenis AzzopardiMary Rutherford
Catie RoussetEtienne Jacotot
Michael SpeddingPhilippe Delagrange
Esther Shenker
Shyamala ManiParthiv Haldipur
Carina Mallard