Tools for risk analysis in the context of melt extrusion

kiszélesítése
és
hosszú
távú
szakmai
fenntarthatóságának
megalapozása

 a
kiváló
tudományos
utánpótlás
biztosításával”


Gyógyszertudományok Doktori Iskola Ph.D. kurzus (GYTKDIE16)

„Introduction to melt extrusion and application of quality by design principles”

2012. 03. 27. – 03. 30.

„Solid dispersions: Types, production, characterization”

Prof. Dr. Peter Kleinebudde


TÁMOP‐4.2.2/

B‐10/1‐2010‐0012
projekt


InsJtute
of
PharmaceuJcs
and
BiopharmaceuJcs
 Heinrich‐Heine‐University


Düsseldorf,
Germany
 Peter
Kleinebudde


Introduc.on
to
melt
extrusion
and
applica.on
 of
quality
by
design
principles


Szeged,
March
26‐30,
2012



PharmaceuJcal
Solid
State



PSSRC
Research
Cluster


Content


•  Solid
dispersions:
Types,
producJon,
characterizaJon


•  IntroducJon
to
melt
extrusion:
Equipment,
process,
 materials,
properJes
of
extrudates,
downstream
 processing,
applicaJons


•  PAT
applicaJons
for
melt
extrusion

and
mulJvariate
 analysis
of
spectral
data


•  Tools
for
risk
analysis
in
the
context
of
melt
extrusion


•  Approaches
to
develop
a
design
and
a
control
space


3


PharmaceuJcal
Solid
State



PSSRC
Research
Cluster


Risk
analysis


•  ICH
Q8,
Q9,
Q10


•  DefiniJons


•  Risk
analysis


•  Design
of
experiments
remarks


•  Examples
for
Design
space


4


© ICH, November 2010

Nov
2005
&
Nov
2008


November
2005


June
2008


ICH
Q8,
Q9
and
Q10


•  High
level
guidances

 (not
prescripJve)


•  Science
and
risk‐based


•  Encourages
systemaJc
 approaches


•  Applicable
over
enJre
product
 lifecycle


•  Intended
to
work
together
to
 enhance
pharmaceuJcal
 product
quality


PharmaceuJcal
Development
‐
Q8(R2)


•  Describes
science
and
risk‐based
approaches
for
 pharmaceuJcal
product
and
manufacturing
process
 development


•  Introduced
concepts
of
design
space
and
flexible
 regulatory
approaches


•  Introduced
concepts
of
Quality
by
Design
(QbD)
and
 provided
examples
of
QbD
development
approaches
 and
design
space


© ICH, November 2010

Quality
Risk
Management
Process
‐
Q9


Process Development

Control Strategy Development

Continual Improvement of the product

© ICH, November 2010

PharmaceuJcal
Quality
System
‐
Q10


•  Describes
key
systems
that
facilitate
establishment
and


maintenance
of
a
state
of
control
for
process
performance
and
 product
quality


•  Facilitates
conJnual
improvement


•  Applies
to
drug
substance
and
drug
product
throughout
product
 lifecycle


•  Sound
pharmaceuJcal
development
(Q8R(2))
in
combinaJon
with
a
 robust
PQS
(Q10)
provide
opportuniJes
for
flexible
regulatory
 approaches.
Relevant
PQS
elements
include
systems
for:




–  Track
and
trend
product
quality
 –  Maintain
and
update
models
as
needed


–  Internally
verify
that
process
changes
are
successful


© ICH, November 2010

PharmaceuJcal
Quality
System
‐
Q10


ICH
Q8,
Q9
and
Q10
Working
Together


Formula.on
Ac.vi.es:


• 
QTPP
DefiniJon


• 
Pre‐FormulaJon
Studies


• 
FormulaJon
Screening


• 
OpJmizaJon
&
SelecJon
 Process
Development
Ac.vi.es:


• 
Process
Screening


• 
Lab
Scale
Development


• 
Scale‐Up
Studies
 Manufacturing
Ac.vi.es:


• 
Commercial
Scale
 

Manufacturing


• 
Batch
Release


Q8
PharmaceuJcal
Development
 Q9
Quality
Risk

Management
 Q10
PharmaceuJcal
Quality
Systems


International Conference on Harmonisation of Technical

Requirements for Registration of Pharmaceuticals for Human Use

Implementation of ICH Q8, Q9, Q10

Breakout
A
 Design
Space

© ICH, November 2010

QbD
Story
per
Unit
OperaJon


Process
 Variables


Design
of
 Experiments


Quality
 Risk
Management


IllustraJve
Examples
of
Unit
OperaJons:


QTPP



&
CQAs


Design

 Space


Control



Strategy
 Batch



Release


Compression
 Real
Time

 Release
tes.ng


(Assay,
CU,
DissoluJon)


Blending
 API


Crystalliza.on


© ICH, November 2010

DS
development
‐
Prior
knowledge


•  Key
messages


– Prior
knowledge
may
include
:



•  internal
knowledge
from
development
and
manufacturing



•  External
knowledge:
scienJfic
and
technical
publicaJons
(including
 literature
and
peer‐reviewed
publicaJons)


– CitaJon
in
filing:
regulatory
filings,
internal
company
report
or
 notebook,
literature
reference


– No
citaJon
necessary
if
well
known
and
accepted
by
scienJfic
 community


DS
development
‐
Prior
knowledge


•  What
might
be
applicable
sources
of
Prior
Knowledge
?


•  IdenJfy
other
type
of
prior
knowledge
that
can
be
used
 in
DS
development


Example
from
Case
Study
:
Crystalliza.on
of
the
drug
substance


-  ParJcle
size
control
needed
during
crystallizaJon


-  Prior
knowledge/1^st
principles
shows
that
other
unit
operaJons
(Coupling
 reacJon,
aqueous
workup,
filtraJon
and
drying)
have
low
risk
of
affecJng
 purity
or
PSD.




> Knowledge
from
prior
filings


> Knowledge
from
lab
/
piloJng
data,
including
data
from
other


compounds
using
similar
“plasorm”
technologies


> 

© ICH, November 2010

DS
development
‐
QRM


•  Risk
assessment
is
based
on
prior
knowledge
and


relevant
experience
for
the
product
and
manufacturing
 process


– Gaps
in
knowledge
could
be
addressed
by
further
 experimentaJon


– Assignments
of
risk
level
must
be
appropriately
jusJfied


•  Risk
assessments/control
will
iterate
as
relevant
new
 informaJon
becomes
available


– Final
itera.on
shows
control
of
risks
to
an
acceptable
level


PharmaceuJcal
Solid
State



PSSRC
Research
Cluster


Ishikawa
diagram


16
 Melt
extrusion
and
QbD
principles
I


© ICH, November 2010

IllustraJon
from
the
Case
Study
‐
Risk
Assessment
for
PSD
Control


To
be
invesJgated
 in
DOE


Detailed
working
documents
like
this
would
likely
not
be
included
in
the
submission


DS
development
–
DOE
&
Modeling


•  Target
the
desired
quality
avribute
range
from
QTPP


•  DeterminaJon
of
edge
of
failure
is
not
required


•  Modeling
is
not
required
to
develop
a
Design
Space



•  Models
need
to
be
verified,
updated
and
maintained


© ICH, November 2010

DS
development
–
DOE
&
Modeling


– Does

the
DOE
results,
as
presented
in
the
case
study,
provide
 sufficient
informaJon
to
define
a
design
space?


– Describe
which
parameters
are
addressed
by
univariate
vs.


mulJvariate
DOEs
and
how
these
are
factored
into
the
design
 space


– Model
implementaJon:
Describe
how
variability
due
to
the
 process
operaJons
and/or
analyJcal
method
is
considered
in
 use
of
the
model



– Describe
the
process
for
maintenance
&
updaJng
of
the
model


© ICH, November 2010

DS
development
–
Process
parameter
&
quality
 avributes


– Design
space
presenta.on
in
the
submission
could
include
 cri.cal
and
non‐cri.cal
parameters


•  CriJcal
parameter
ranges/model
are
considered
a
regulatory


commitment
and
non‐criJcal
parameter
ranges
support
the
review
of
 the
filing


•  CriJcal
parameter
changes
within
design
space
are
handled
by
the
 Quality
System
and
changes
outside
the
design
space
need
appropriate
 regulatory
noJficaJon



– Non‐cri.cal
parameters
would
be
managed
by
Quality
System


© ICH, November 2010

IllustraJon
from
case
study
:
QTPP
and
CQAs


Dosage form and strength

Immediate release tablet containing 30 mg of active ingredient.

Specifications to assure safety and efficacy during shelf-life

Assay,

Uniformity of Dosage Unit (content uniformity) and dissolution.

Description and hardness Robust tablet able to withstand transport and handling.

Appearance Film-coated tablet with a suitable size to aid patient acceptability and compliance.

Total tablet weight containing 30 mg of active ingredient is 100 mg with a diameter of 6 mm.

Drug
Product
CQAs


• Assay


• Content
Uniformity


• DissoluJon


• Tablet
Mechanical
Strength


CQAs
derived
using
Prior
Knowledge
(e.g.


previous
experience
of
developing
tablets)
 CQAs
may
be
ranked
using
quality
risk
assessment.


QTPP


API
CrystallizaJon:



Design
Space
&
Control
Strategy


© ICH, November 2010

Large
square
 
shows
the
ranges
tested
in
the
DOE
 Red
area
shows
points
of
failure


Green
area
shows
points
of
success.


• In
the
idealized
example
at
lex,
the
 oval
represents
the
full
design
space.

It
 would
need
to
be
represented
by
an
 equaJon.




• AlternaJvely,
the
design
space
can
be
 represented
as
the
green
rectangle
by
 using
ranges


- a
porJon
of
the
design
space
is
not
 uJlized,
but
the
benefit
is
in
the
 simplicity
of
the
representaJon


Seed
wt%


IllustraJon
from
the
case
study
:
 OpJons
for
DepicJng
a
Design
Space


© ICH, November 2010

Key
Messages


•  Quality
Risk


Management
is
the
 full
process


•  Quality
Risk
 Assessment,


Control,
Review
etc.


represent
only
 individual
steps


ICH
Q9


© ICH, November 2010

Key
Messages


•  QRM
is
an
iteraJve
process
and
not
a
one
off
acJvity


•  UJlisaJon
of
QRM
acJviJes
should
lead
to
a
greater
 assurance
of
quality
through
risk
control



–  Facilitate
the
awareness
of
risks
 –  Risk
does
not
go
away


–  Risk
can
be
predicted,
prevented
and
controlled


•  QRM
processes
should



–  Focus
on
what
is
important
to
establish
the
manufacturing
process
and
 controls
and
maintain
them
over
the
life
cycle


–  Be
integrated
in
PharmaceuJcal
Quality
System
elements


Key
Messages


•  QRM
used
by
company
can
provide
regulators
with
 greater
assurance
of
a
company’s
product
and
 process
understanding
and
the
ability
to
assure
 quality
of
manufactured
products


•  QRM
should
be
used
by
regulators
(both
assessors
 and
inspectors)
to
guide
regulatory
acJviJes


independent
of
the
industry
uJlisaJon
of
QRM


© ICH, November 2010

Key
Messages


•  Regulators
should
use
QRM
methods
appropriately
to
 reach
raJonal
and
jusJfied
regulatory
decisions
^e.g.

– Risk
based
regulatory
decisions
 (suspected
quality
defects
etc.)
 – Assessment
of
regulatory
filing



– Planning
and
conducJng
of
inspecJons

 – PrioriJsaJon
of
inspecJon
findings


© ICH, November 2010

Opportuni.es
to
apply
Quality
Risk
Managements


QRM
in
the
Product
Life
Cycle


Pa.ent
 needs
 Business



needs


Quality Target Product Profile (QTPP)

Critical Quality Attribute

(CQA)

Critical Process Parameters

(CPP) Product


design


Manu‐


facturing
 Process


design


Control

 Strategy


Technical regulatory Filing & Review

Performance Review &

Change Control

Commercial

Manufacturing


Research
and

 clinical
studies


Process
understanding
 PAI



Inspec.ons
 GMP



Inspec.ons
 Knowledge
management


Technical
 Transfer


approx.



life
cycle
.me


1/4
 3/4


© ICH, November 2010

Key
Messages


Two
primary
principles
of
QRM
are


•  The
evaluaJon
of
the
risk
to
quality
should
be
based
 on
scienJfic
knowledge
and
ulJmately
link
to
the
 protecJon
of
the
paJent



•  The
level
of
effort,
formality
and
documentaJon
of
the
 quality
risk
management
process
should
be


commensurate
with
the
level
of
risk
 ICH
Q9


Key
Messages


•  Reduce
subjec.vity
of
implemen.ng
QRM
by
making
sure
the
right
people
 are
at
the
table



(e.g.
mulJ‐discipline,
include
respecJve
stakeholders,
as
applicable)


•  Use
QRM
methods
appropriately
and
present
the
conclusions
and
 jus.fica.ons
clearly



–  Be
clear
and
consistent
in
wording
/
terms
used
based
on
internaJonally
 agreed
definiJons


–  Transparency
on
the
logic
of
the
methodology
and
the
decision
making
 –  QRM
can
not
be
used
to
jusJfy
failure



•  Use
QRM
proac.vely
for
increasing
the
knowledge
of
your
product
and
 processes


Quality
by
Design


Karlsruhe,
Oktober,
2011



Dr.
Iris
M.
ZieglerNasr
Moheb,
ImplementaJon
of
Q8
:
FDA
PerspecJve;

ISPE
and
PDA
Washington,
D.C.,
December
2006

 



DESIGN
SPACE 

ICH
Q8(R2):
Design
space


•  The
mulJdimensional
combinaJon
and
interacJon
of
input
 variables
(e.g.,
material
avributes)
and
process
parameters
that
 have
been
demonstrated
to
provide
assurance
of
quality.


•  The
relaJonship
between
the
process
inputs
(material
 avributes
and
process
parameters)
and
the
criJcal
quality
 avributes
can
be
described
in
the
design
space.



•  Working
within
the
design
space
is
not
considered
as
a
change.


•  Movement
out
of
the
design
space
is
considered
to
be
a
change
 and
would
normally
iniJate
a
regulatory
postapproval
change
 process.



Presenta.on
of
Design
space


Presenta.on
of
Design
space


Desired:
DissoluJon
>
80%


Parameter
1:
44‐53
 Parameter
2:
0‐1.1


Presenta.on
of
Design
space


Presenta.on
of
Design
space


Presenta.on
of
Design
space


International Conference on Harmonisation of Technical

Requirements for Registration of Pharmaceuticals for Human Use

Implementation of ICH Q8, Q9, Q10

Breakout
A
 Design
Space

presentaJons.html


Tools


•  StaJsJcal
thinking


•  Design
of
Experiments
(DoE)


•  MulJvariate
data
analysis
(MVDA:
PCA,
PLS
etc.)


•  Neural
networks
including
self
organizing
maps
(SOM)


•  Process
simulaJon
(numerical
simulaJon
of
physical
 processes)


–  ConJnuum
approach,
e.g.
CFD
 –  Discrete
approach,
e.g.
DEM
 –  MulJscale
modeling


•  Semi‐empirical
models


•  First
principle
models


Design
of
Experiments
(DoE)


•  Full
factorial
design


•  FracJonal
factorial
design


•  Response
surface
design


•  two
factor
levels:
linear
and
interacJon
models


•  three
and
more
factor
levels:
quadraJc
models


Informa.on
about
DoEs


•  For
DoEs
involving
single‐
or
mulJple‐unit
operaJons
that
 are
used
to
establish
CPPs
and/or
to
define
a
Design
Space
 (DS),
the
inclusion
of
the
following
informaJon
in
the
 submission
will
greatly
facilitate
assessment
by
the
 regulators:



–  raJonale
for
selecJon
of
DoE
variables
(including
ranges)
that
 would
be
chosen
by
risk
assessment
(e.g.,
consideraJon
of
the
 potenJal
interacJons
with
other
variables).




–  Any
evidence
of
variability
in
raw
materials
(e.g.,
drug
substance
 and/or
excipients)
that
would
have
an
impact
on
predicJons
made
 from
DoE
studies.



–  LisJng
of
the
parameters
that
would
be
kept
constant
during
the


Informa.on
about
DoEs


–  Type
of
experimental
design
used
and
a
jusJficaJon
of
its
 appropriateness,
including
the
power
of
the
design.




–  Factors
under
study
and
their
ranges
can
be
presented
in
a
 tabular
format.
Submivers
should
indicate
if
the
factors
are
 expected
to
be
scale‐dependent.



–  Reference
to
the
type
of
analyJcal
methods
(e.g.,
HPLC,
NIR)
 used
for
the
evaluaJon
of
the
data
and
their
suitability
for
 their
intended
use
(e.g.,
specificity,
detecJon
limit).



–  Results
and
staJsJcal
analysis
of
DoE
data
showing
the
 staJsJcal
significance
of
the
factors
and
their
interacJons,
 including
predicJons
made
from
DoE
studies
relevant
to
 scale
and
equipment
differences.


ICH
Quality
IWG:
Points
to
Consider
for
ICH
Q8/Q9/Q10
ImplementaJon


DESIGN
OF
EXPERIMENTS


REMARKS 

collinearity


Factors
x₁
and
x₂
are
perfectly
correlated:
R=1


?


collinearity


Model
equa.on


constant
 main
effects,
 linear
terms
 error


linear
model


Model
equa.on


constant
 main
effects,
 linear
terms
 two
factor
 interacJons


error


interac.on
model


Model
equa.on


constant
 main
effects,
 linear
terms
 two
factor
 interacJons


quadraJc
terms
 error


quadra.c
model


Coding
of
variables


Coding
of
variables


Factor
space


High
level
pressure:


1
bar
or
5
bar


StaJsJca


Evalua.on
with
coded
factors


ALL
MODELS
ARE
WRONG,
 BUT
SOME
ARE
USEFUL.



 
 
 
 G.E.P.
BOX


Factor
levels


Y


threshold:
quadra.c
model


‐10
 0
 10
 20
 30
 40
 50
 60


0
 20
 40
 60
 80
 100
 120


linear
model


quadraJc
model


EXAMPLES 

Example


Dynamic
co‐precipita.on
process


Thermodynamics
and
crystal
growth
 model:


‐ 
semi‐empirical
nucleaJon
rate
equaJon


Naproxen
and
Eudragit
L100
 in
alcohol:
precipitaJon
with
water



Dynamic
co‐precipita.on
process


•  3³
full
factorial
design


– Slurry
temperature
 – Slurry
sJrring
rate


– Non‐solvent
addiJon
rate


•  InteracJon
model


‐
CL
ranges
calculated


‐ 
ANOVA


‐ 
General
Linear
Modeling


Dynamic
co‐precipita.on
process


GLM


Neural
network
modeling


Dynamic
co‐precipita.on
process


As
a
proof
of
concept,
this
simplified
case
study
was
mainly
focused
on
 the
effects
of
process
variables
while
keeping
other
variables
fixed
 based
on
risk
analysis
and
iniJal
formulaJon
development
results.


Maximum
Rate7
was
obtained
around
the
following
process
window:


‐1
<
x1
<
0.1,
‐0.25
<
x3
<
0.95


Example


uniformity


•  Discrete
element
modeling
(DEM)‐simulaJon


•  2
potenJal
criJcal
input
parameters:
weight
raJo,
 blending
Jme


•  2
component
system:
ASA
and
lactose


•  Blending
end
point


•  Target:
RSD
<
5%


Blend
homogeneity
and
content
 uniformity


Ishikawa
diagram


FMEA


uniformity


SimulaJon
parameters:


‐ 
spherical
parJcles
of
4mm
diameter


‐ 
loading
40%
volume,
i,e,
30,000
parJcles


‐ 
density:
1.53
g/cm³
for
ASA
 





1.40
g/cm³
for
lactose


‐ 
maximal
parJcle
overlap:
5%


‐ 
coefficient
of
resJtuJon:
0.7
ASA;
0.6
lactose


‐ 
sliding
fricJon
coefficient:
0.4
ASA;
0.3
 lactose


‐ 
15
rpm


‐ 
10%
‐
50%
ASA


‐ 
0
–
60
s
blending
Jme


‐ 
mulJple
samples
with
constant
sampling
size


‐ 
nine
samples
axer
each
revouluJon

 

of
120
parJcles
(=
6g
sample
weight)


Blend
homogeneity
and
content
 uniformity


RSDs=σ_s

W with σs= p(1−p)

( ) completely



random




simulaJon


uniformity


10%
ASA
axer
15
revoluJons


ASA
and
lactose
parJcles 
 
 
only
ASA
parJcles 



Blend
homogeneity
and
content
 uniformity


Lacey‘s
Index:


raJo
of
 
‚mixing
achieved‘
to





 
‚mixing
possible‘


Example


Bootstrapping



•  Bootstrapping
is
the
pracJce
of
esJmaJng
properJes
of
an
 esJmator
(such
as
its
variance)
by
measuring
those
properJes
 when
sampling
from
an
approximaJng
distribuJon.



•  One
standard
choice
for
an
approximaJng
distribuJon
is
the
 empirical
distribuJon
of
the
observed
data.



•  In
the
case
where
a
set
of
observaJons
can
be
assumed
to
be
 from
an
independent
and
idenJcally
distributed
populaJon,
 this
can
be
implemented
by
construcJng
a
number
of
 resamples
of
the
observed
dataset
(and
of
equal
size
to
the
 observed
dataset),
each
of
which
is
obtained
by
random
 sampling
with
replacement
from
the
original
dataset.



Bootstrapping



•  Bootstrapping
allows
one
to
gather
many
alternaJve
versions
of
the
single
 staJsJc
that
would
ordinarily
be
calculated
from
one
sample.



•  For
example,
assume
we
are
interested
in
the
height
of
people
worldwide.


As
we
cannot
measure
all
the
populaJon,
we
sample
only
a
small
part
of
it.


From
that
sample
only
one
value
of
a
staJsJc
can
be
obtained,
i.e.
one
 mean,
or
one
standard
deviaJon
etc.,
and
hence
we
don't
see
how
much
 that
staJsJc
varies.



•  When
using
bootstrapping,
we
randomly
extract
a
new
sample
of
n
heights
 out
of
the
N
sampled
data,
where
each
person
can
be
selected
at
most
t
 Jmes.



•  By
doing
this
several
Jmes,
we
create
a
large
number
of
datasets
that
we
 might
have
seen
and
compute
the
staJsJc
for
each
of
these
datasets.



•  Thus
we
get
an
esJmate
of
the
distribuJon
of
the
staJsJc.
The
key
to
the
 strategy
is
to
create
alternaJve
versions
of
data
that
"we
might
have
seen".


Example


Example


Robust
design
space
by
bootstrap


techniques


Example


Data‐driven
modeling


•  High‐Dimensional
Model
RepresentaJons


•  Response
Surface
Methodology


•  Kriging
Methodology


–  Capability
of
modeling
complex
funcJons
and
providing
 error
esJmates


–  FuncJon
values
for
a
sampling
point
located
close
to
the
 test
point
is
weighted
more
heavily
in
contrast
to
a
sampling
 point
located
farther
away.


•  Modeling
with
Discrete
Design
Variables


•  Case
study:


–  ConJnuous
mixer



•  Provide
a
general
framework
for
mapping
the
design
 space
of
any
process
operaJon
for
which
first‐

principle
models
are
not
yet
available.


Example


Example


Example


Example


Mul.scale
modeling


•  Based
on
the
analysis
of
the
current
status
of
the
process
 modeling
tools
and
systems
in
the
area
of
solids


processes,
the
following
conclusions
can
be
drawn:



•  the
ulJmate
goal
of
the
modeling
is
simulaJon
of
a
whole
 process
plant
on
the
macroscale;



•  to
adequately
consider
material
properJes
and
apparatus
 geome‐tries,
the
mulJscale
methodology
should
be
used.



•  As
a
main
simulaJon
approach
in
SolidSim,
the
sequenJal‐

modular
method
is
used
(Hillestad
&
Hertzberg,
1986),
for
 which
the
calculaJon
of
units
is
iteraJvely
repeated
unJl
 convergence
is
reached.



Mul.scale
modeling


Mul.scale
modeling


Mul.scale
modeling


Mul.scale
modeling


Solid
Sim
 flowsheet


Mul.scale
modeling


Mul.scale
modeling


PharmaceuJcal
Solid
State



PSSRC
Research
Cluster


Düsseldorf


NASA,
2004


Contact


Prof.
Dr.
Peter
Kleinebudde


InsJtute
of
PharmaceuJcs
and
BiopharmaceuJcs
 Universitaetsstrasse
1


40225
Duesseldorf

 Germany


tel.:
+49‐211‐8114220
 fax:
+49‐211‐8114251
 e‐mail:
kleinebudde@hhu.de


Acknowledgements


Dr.
Iris
Ziegler
 Dr.
Markus
Thommes
 Markus
Wirges


Thank
you
for
your
kind
avenJon!