Model-based treatment planning in reproductive medicine
Transcript of Model-based treatment planning in reproductive medicine
Model-based treatmentplanning
in reproductive medicine
Dr. Susanna Roblitz Zuse Institute Berlin
Computational Systems Biology Group
ZIB,FU
FU Berlin
September 14, 2015
Motivation
SciCade 2015 2 Susanna Roblitz
The human menstrual cycle
Exactly timed interplay ofphysiological processes
I follicle development
I ovulation and fertilization
I formation of corpus luteum
I embryonic attachment andgrowth in the uterus
⇒ coordination between neuraland endocrine systems
(http://www.websters-online-dictionary.org/definitions/Menstrual Cycle)
SciCade 2015 3 Susanna Roblitz
Endocrine disorders
Unwanted childlessness among couples in Europe: 12-15%
Female health problems: 50%, thereof 40% endocrinological diseases
Infertility due to female problems
EU infertile couplesin reproductive age
Infertility due to endocrinological
diseases
I PCOS (Polyzystic Ovarian Syndrom):main cause for hyperandrogenism, leading to cycle disorders and infertility(4-12% of women in reproductive age)
I Endometriosis (uterine lining outside uterus):about 40% of women at reproductive age, thereof 30-50% infertility
I Hyperprolactinemia (increased blood levels of prolactin):in about 20% of women with reproductive disorders
I External factors: smoking, BMI, age
SciCade 2015 4 Susanna Roblitz
Reproductive medicine
Increased chance for successful pregnancy by modern techniques:
I In-vitro fertilization (IVF)
I Intracytoplasmic sperm injection (ICSI)
Success rates: 8 - 35%
Depending on the clinic due to differenttreatment strategies!
Aim: supply of model-based clinical decision support system forreproductive endocrinologists
I better understanding of complex processes
I simulation and optimization of treatment strategies in silico(cost-saving and efficient)
SciCade 2015 5 Susanna Roblitz
Virtual hospital
I input: patient data,treatment protocol
I output: actions to beperformed on thepatient
based on
I a population ofvirtual patients
I a virtual doctor
Patient-specific model
Treatment protocol
Physician Patient
Actions
Measurements
Actio
ns
Op
tio
ns
Individualised Treatment Protocol
Pass orFail + Counterexample
Mo
de
lling
+C
linic
al
exa
ms
Observable Outputs
Patient-specificmodel (plant)
Controllable Inputs(Controllable External Factors)
Treatment protocolmodel (controller)
Uncontrollable Inputs (Uncontrollable External Factors)
Model-Based Verification of Treatment Protocols
Observable Outputs
Patient-specificmodel (plant)
Controllable Inputs(Controllable External Factors)
Treatment protocolmodel (controller)
Uncontrollable Inputs (Uncontrollable External Factors)
Model-Based Design of Individualised Treatment Protocols
?
Modelling
Virtual Hospital
Parameter identification
SciCade 2015 6 Susanna Roblitz
Experimental data in vivo
SciCade 2015 7 Susanna Roblitz
Untreated cycles: healthy women
−10 0 10 20 30 400
20
40
60
80
100
IU/L
LH
−10 0 10 20 30 400
5
10
15
20
IU/L
FSH
−10 0 10 20 30 400
100
200
300
400
500
pg/m
l
E2
−10 0 10 20 30 400
5
10
15
20
25
30
ng/m
l
P4
SciCade 2015 8 Susanna Roblitz
Untreated cycles: women with PCOS
SciCade 2015 9 Susanna Roblitz
Treatment protocol data
0" 1" 2" 3" 4" 5" 6" 7" 8" 9" 10" 11" 12" 13" 14"
Fer0lity'treatment'
Downregula0on'with'GNrH''Analoga'(Long'protocol)''or'Primolut'N'(short'protocol)'
E2"P"PRL"T"A'
E2'P'T'
DHEAS'A'
UltraKsound'
' ' ' Daily' ' ' ' injec0on' ' ' ' of' ' ' ' FSH' ' ' or' ' ' HmG' ((9K' 14' days)' ' .''''
E2'T'
DHEAS'A'
UltraKsound'
Verfica0on'of'successful'
downregula0on'
Ovula0on'induc0on'between'day'9'and'day'14'(depending'on'follicles'size,'number'and'E2'level)'
E2'T'
DHEAS'A'
UltraKsound'
Individual'0ming'of'controls'every'1K2'(3)'days'
Egg're0reval'35h'aZer'ovula0on'induc0on'
Embryo'transfer'2K3'days'aZer'egg'
re0eval''
Ovarrechts Ovar links
GnRHa hMG/ Tag Datum BT E2 P4 < 10 12 14 16 18 ≥ < 10 12 14 16 18 ≥FSH pmol/L nmol/L 10 11 13 15 17 19 20 10 11 13 15 17 19 20
.
.
.1 225 Fr 07.06.13 8 2841 4 1 1 5 1 11 225 Sa 08.06.13 91 225 So 09.06.13 101 225 Mo 10.06.13 11 6062 2 1 1 1 2 3 5 3 3
SciCade 2015 10 Susanna Roblitz
Treatment cycle: ultrasound measurements
SciCade 2015 11 Susanna Roblitz
Drug data
Single dose Nafarelin (GnRH agonist)
4 5 6 7 8 90
0.5
1
1.5
2
2.5
4 5 6 7 8 90
50
100
150
4 5 6 7 8 95
10
15
20
25
30
35
40
4 5 6 7 8 90
50
100
150
200
250
300
Single and multiple dose Cetrorelix (GnRH antagonist)
0 20 40 600
5
10
15
20
25
0 20 40 60 800
2
4
6
8
0 20 40 602.5
3
3.5
4
4.5
5
5.5
6
6.5
0 20 40 60 800
50
100
150
measurements: drug, LH, FSH, E2
SciCade 2015 12 Susanna Roblitz
Model development for thehuman menstrual cycle
SciCade 2015 13 Susanna Roblitz
Conceptual model
Compartments: blood, ovaries,uterus, pituitary, hypothalamusComponents:
I Estradiol
I Progesterone
I Inhibin A and B
I LH + receptor binding
I FSH + receptor binding
I GnRH + receptor binding
I 6 follicular stages
I 6 luteal stages (corpusluteum)
HYPOTHALAMUS
PITUITARY
CORPUS LUTEUM
OVARIES
inhibin
activin
follistatin
FSH
LH
GnRH
estradiol
progesterone
estradiol
progesterone
TEUM
ovulation
SciCade 2015 14 Susanna Roblitz
Model GynCycle
����
����
Lut1 Lut2Sc2OvFPrF
GnRH antagonist
CENTRAL COMPARTMENT
GnRH antagonist
DOSING COMPARTMENT
PERIPHERAL COMPARTMENT
GnRH antagonist
GnRH agonist
DOSING COMPARTMENT
GnRH Ant−RecComplex
inactive GnRH−Rec
complex
complex
active GnRH−Rec
active Ago−Rec
complex
GnRH agonist
CENTRAL COMPARTMENT
AF1 AF2 AF3 AF4 Sc1 Lut3 Lut4
inactive
GnRH Receptors
GnRH Receptors
active
inactive Ago−Rec
complex
GnRH (G)
Progesterone (P4)
Estradiol (E2)
Inhibin B (IhB)
Inhibin A (IhA)
effective IhA (IhA )e
free LH receptors
LH(R )
LH receptor complex
(LH−R)
desensitized rec.
LH,des
pit
pituitary LH
(LH )blood
serum LH
pit(FSH )pituitary FSH
blood(FSH )
serum FSH free FSH receptors
(R )FSH
FSH receptor complex
(FSH−R)
(R )
FSH,des(R )desensitized rec.
(LH )
(freq)
( s )
foll. LH sensitivity
(mass)GnRH mass
GnRH frequency
GynCycle: 33(+8) ODEs, 114 parameters [Roblitz et al. (2013)]
I a model for the idealized cycle of a healthy womanI computation of hormone profiles and follicle development over
time
SciCade 2015 15 Susanna Roblitz
Submodel for follicular development
xi : radius of follicle i
dxi
dt= A(G − D), i = 1, . . . , n
A := µH+i (FSH)
G :=
νκH−(P4) + (ν + β)xi
n∑j=1
xj
xi
D :=
(νβ + x2i )
n∑j=1
xj + κxi
xi
H−(P4) := cη5
P4(t)5 + η5, H+
i (FSH) :=FSH(t)5
δi5 + FSH(t)5
initial FSH sensitivity: ∼ N (µ, σ)
# follicles created in a fixed time interval: ∼ Poisson
0 2 4 6 8 10 12 140
0.02
0.04
0.06
0.08
0.1
FSH Sensitivity Mean100 150 200 250 300 350
Ovu
l per
per
iod
0.5
1
1.5
2
2.5
3
3.5
4
Number of ovulations per period
SciCade 2015 16 Susanna Roblitz
Parameter estimation
Model: y(t, θ) = (y1(t, θ), . . . , yn(t, θ)) ∈ Rn
Parameters: θ = (θ1, ..., θq) ∈ Rq
Data: zkl ≈ yk(tl , θ), k = 1, . . . , n, l = 1, . . . ,mk
(i) direct minimisation of least squares error
‖F (θ)‖22 =
n∑k=1
mk∑l=1
(zkl − yk(tl , θ))2
2σ2kl
θ−→ min
⇒ ill-posed problem(ii) computation of joint probability distributions according toBayes’ theorem P(θ|z) ∝ P(z |θ)P(θ) with likelihood
P(z |θ) ∝ exp(−‖F (θ)‖2
2
)SciCade 2015 17 Susanna Roblitz
Results of in silico experiments
SciCade 2015 18 Susanna Roblitz
Generic model for normal cycles
Single parametrization from real patient normal cycle data.
I normal cycle simulation
0 10 20 300
50
100
150
mIU
/mL
LH
0 10 20 300
5
10
15
20
mIU
/mL
FSH
0 10 20 300
5
10
15
20
25
30
ng/m
L
P4
0 10 20 300
100
200
300
400
500
pg/m
L
E2
I simulating the effect ofbirth control pills
0 50 100 150 200 250 3000
20
40
60
80
100
120
day
LH P4 estrogens
SciCade 2015 19 Susanna Roblitz
Validation of generic model
Validate generic model with real patient treatment data.
I single dose agonist (nafarelin)
−20 0 20 40 60 80 1000
50
100
150
200
day−20 0 20 40 60 80 1000
50
100
150
200
day−20 0 20 40 60 80 1000
50
100
150
200
day
I multiple dose agonist (nafarelin)
−20 0 20 40 60 80 100 120 1400
5
10
15
20
days
ng/m
L
datasimulated P4
[Roblitz et al. (2013)]
I single dose antagonist(cetrorelix)
−30 −20 −10 0 10 200
100
200
300
days
pg/m
L
dataE2
SciCade 2015 20 Susanna Roblitz
Uncertainty quantification
0 20000 40000 6000021000
21500
22000
22500 par16
0 20000 40000 60000
9.2
9.4 par71
0 20000 40000 600004.2
4.3
4.4
4.5 par87
21000 21500 22000 225000.000
0.002
0.004
0.006 par16
9.2 9.40
5
10
15 par71
4.2 4.3 4.4 4.50
5
10
15 par87
I 82 random variables, uniform prior distribution
I posterior sampling: Metropolis-Hastings algorithm with lognormalproposal distribution
I alignment of LH peaks in each sample; only acceptance of periodicsolutions with cycle length 20-50 days
SciCade 2015 21 Susanna Roblitz
Virtual patients
Generate model instances (parametrizations) compatible with realpatient data for the normal cycle [Mancini et al. (2014)].
finite set of biologically admissable parameter sets
Real Patient
Virtual patient
(a) Medical level
(b) Computation level
offline −→ online
SciCade 2015 22 Susanna Roblitz
Virtual patients
Validate virtual patient models with real patient data fromtreatment cycles.
Long protocol: downregulation cycle days 23 to 50 with Triptoreline, then 14 days
stimulation, finally Ovitrelle (drug database!)
−→ model refinement −→
SciCade 2015 23 Susanna Roblitz
Model-based treatment verification & design
I treatment verificationthe treatment model (closedloop system) reaches a statein which some desiredproperty is satisfied(treatment goals)
I treatment designfinding values for treatmentparameters (type, dose andtime of drug) that optimizesome key performanceindicators (KPIs): E2 levels,number and size of follicles,total amount of drug
SciCade 2015 24 Susanna Roblitz
Model-based treatment verification
Verify that a given treatment protocol reachesits goal for the largest possible number of(virtual) patients → evaluate success rate
����
�����
��
��
SciCade 2015 25 Susanna Roblitz
Model-based treatment design
Synthesised generic down-regulation treatments require 40% of theinjections and <25% of the overall Decapeptyl amount required byreference treatment. Individualised treatments even lighter, stillachieving clinical goals!
SciCade 2015 26 Susanna Roblitz
Model-based treatment design
incremental change oftreatment parameters:
I age class
I AMH level
I AFC class
I dose ofstimulation drug
→ set ofPareto-optimaltreatments, in which atleast one performanceindicator is better
SciCade 2015 27 Susanna Roblitz
Conclusion
Benefit
I the virtual hospital as a trainingtool for physicians
I suggestions for new clinical studies
Future work
I improve the model
I improve the mathematicalalgorithms
I perform model-based comparisonof treatment protocols
I extend approach toendocrinological diseases
SciCade 2015 28 Susanna Roblitz
Acknowledgement
Computational Systems Biology Grouphttp://www.zib.de/numeric/csb
in particular:
Thomas Dierkes, Rainald Ehrig, Stefan Schafer,
Claudia Stotzel
Contact: [email protected]
Partners in the EU-project PAEON-Model-Driven Computation ofTreatments for Infertility RelatedEndocrinological Diseases
I Enrico Tronci (La SapienzaRome)
I Brigitte Leeners (UniversityHospital Zurich)
I Tillmann Kruger (HannoverMedical School)
I Marcel Egli (University Lucerne)
SciCade 2015 29 Susanna Roblitz