utility-based optimization of combination therapy using ...utility-based optimization of combination...
TRANSCRIPT
![Page 1: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/1.jpg)
Utility-Based Optimization of CombinationTherapy Using Ordinal Toxicity and Efficacy in
Phase I/II Clinical Trials
Peter F. ThallDepartment of Biostatistics
University of Texas, M.D. Anderson Cancer Center
Memorial Sloan Kettering Cancer Center
October 2, 2009
![Page 2: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/2.jpg)
UTILITY-BASED DOSE-FINDING 1
Collaborators
Hoang Nguyen – Biostatistics Dept, M.D. Anderson, TexasComputer Programming, Prior Calibration, Model Refinements,Graphics
![Page 3: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/3.jpg)
UTILITY-BASED DOSE-FINDING 1
Collaborators
Hoang Nguyen – Biostatistics Dept, M.D. Anderson, TexasComputer Programming, Prior Calibration, Model Refinements,Graphics
Nadine Houede – Institut Bergonie, Bordeaux, FranceMedical Application, Utility Elicitation and Restaurant Selection
![Page 4: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/4.jpg)
UTILITY-BASED DOSE-FINDING 1
Collaborators
Hoang Nguyen – Biostatistics Dept, M.D. Anderson, TexasComputer Programming, Prior Calibration, Model Refinements,Graphics
Nadine Houede – Institut Bergonie, Bordeaux, FranceMedical Application, Utility Elicitation and Restaurant Selection
Xavier Paoletti – Institut Curie, Paris, FranceGaussian Copula Formulas and Wine Selection
![Page 5: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/5.jpg)
UTILITY-BASED DOSE-FINDING 1
Collaborators
Hoang Nguyen – Biostatistics Dept, M.D. Anderson, TexasComputer Programming, Prior Calibration, Model Refinements,Graphics
Nadine Houede – Institut Bergonie, Bordeaux, FranceMedical Application, Utility Elicitation and Restaurant Selection
Xavier Paoletti – Institut Curie, Paris, FranceGaussian Copula Formulas and Wine Selection
Andrew Kramar – Parc Euromedecine, Montpellier, FranceDetailed Editing and La Joie de Vivre
![Page 6: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/6.jpg)
UTILITY-BASED DOSE-FINDING 2
OUTLINE
![Page 7: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/7.jpg)
UTILITY-BASED DOSE-FINDING 2
OUTLINE
1. Combination Bio-Chemotherapy for Bladder Cancer
![Page 8: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/8.jpg)
UTILITY-BASED DOSE-FINDING 2
OUTLINE
1. Combination Bio-Chemotherapy for Bladder Cancer
2. Data Structure and Probability Models
![Page 9: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/9.jpg)
UTILITY-BASED DOSE-FINDING 2
OUTLINE
1. Combination Bio-Chemotherapy for Bladder Cancer
2. Data Structure and Probability Models
• Bivariate Almost Ordinal (Toxicity, Efficacy)
![Page 10: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/10.jpg)
UTILITY-BASED DOSE-FINDING 2
OUTLINE
1. Combination Bio-Chemotherapy for Bladder Cancer
2. Data Structure and Probability Models
• Bivariate Almost Ordinal (Toxicity, Efficacy)• A Flexible Parametric Model
![Page 11: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/11.jpg)
UTILITY-BASED DOSE-FINDING 2
OUTLINE
1. Combination Bio-Chemotherapy for Bladder Cancer
2. Data Structure and Probability Models
• Bivariate Almost Ordinal (Toxicity, Efficacy)• A Flexible Parametric Model• Establishing Priors
![Page 12: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/12.jpg)
UTILITY-BASED DOSE-FINDING 2
OUTLINE
1. Combination Bio-Chemotherapy for Bladder Cancer
2. Data Structure and Probability Models
• Bivariate Almost Ordinal (Toxicity, Efficacy)• A Flexible Parametric Model• Establishing Priors
3. Trial Design and Conduct
![Page 13: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/13.jpg)
UTILITY-BASED DOSE-FINDING 2
OUTLINE
1. Combination Bio-Chemotherapy for Bladder Cancer
2. Data Structure and Probability Models
• Bivariate Almost Ordinal (Toxicity, Efficacy)• A Flexible Parametric Model• Establishing Priors
3. Trial Design and Conduct
• Utilities
![Page 14: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/14.jpg)
UTILITY-BASED DOSE-FINDING 2
OUTLINE
1. Combination Bio-Chemotherapy for Bladder Cancer
2. Data Structure and Probability Models
• Bivariate Almost Ordinal (Toxicity, Efficacy)• A Flexible Parametric Model• Establishing Priors
3. Trial Design and Conduct
• Utilities• Safety Rules
![Page 15: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/15.jpg)
UTILITY-BASED DOSE-FINDING 2
OUTLINE
1. Combination Bio-Chemotherapy for Bladder Cancer
2. Data Structure and Probability Models
• Bivariate Almost Ordinal (Toxicity, Efficacy)• A Flexible Parametric Model• Establishing Priors
3. Trial Design and Conduct
• Utilities• Safety Rules
4. Computer Simulations
![Page 16: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/16.jpg)
UTILITY-BASED DOSE-FINDING 2
OUTLINE
1. Combination Bio-Chemotherapy for Bladder Cancer
2. Data Structure and Probability Models
• Bivariate Almost Ordinal (Toxicity, Efficacy)• A Flexible Parametric Model• Establishing Priors
3. Trial Design and Conduct
• Utilities• Safety Rules
4. Computer Simulations
![Page 17: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/17.jpg)
UTILITY-BASED DOSE-FINDING 3
Dose-Finding for Combination Therapy of Bladder Cancer
![Page 18: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/18.jpg)
UTILITY-BASED DOSE-FINDING 3
Dose-Finding for Combination Therapy of Bladder Cancer
Goals: Find an optimal dose pair of combination therapy =chemotherapeutic (chemo) agents gemcitabine + cisplatin +biological agent for untreated advanced bladder cancer, basedon Toxicity and Efficacy (a “phase I/II” trial)
![Page 19: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/19.jpg)
UTILITY-BASED DOSE-FINDING 3
Dose-Finding for Combination Therapy of Bladder Cancer
Goals: Find an optimal dose pair of combination therapy =chemotherapeutic (chemo) agents gemcitabine + cisplatin +biological agent for untreated advanced bladder cancer, basedon Toxicity and Efficacy (a “phase I/II” trial)
Treatment Regime: In each 28-day cycle, the patient receives
1. biological agent orally each day at dose levels d1 = 1, 2, 3 or 42. gemcitabine on days (1, 8, 15) at dose levels d2 = 1, 2 or 3(750, 1000 or 1250 mg/m2/day)3. a fixed dose of 70 mg/m2 cisplatinum on day 2
=⇒ ddd = (d1, d2) ∈ {1, 2, 3, 4} × {1, 2, 3}
![Page 20: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/20.jpg)
UTILITY-BASED DOSE-FINDING 4
The Dose Pair Domain D
![Page 21: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/21.jpg)
UTILITY-BASED DOSE-FINDING 4
The Dose Pair Domain D
(1, 3) (2, 3) (3, 3) (4, 3)
↑ (1, 2) (2, 2) (3, 2) (4, 2)
d2 (1, 1) (2, 1) (3, 1) (4, 1)
d1 −→
d1 = dose of biological agent
d2 = dose of chemo agent gemcitabine
![Page 22: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/22.jpg)
UTILITY-BASED DOSE-FINDING 5
Clinical Outcomes (evaluated over two 28-day cycles )
![Page 23: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/23.jpg)
UTILITY-BASED DOSE-FINDING 5
Clinical Outcomes (evaluated over two 28-day cycles )
Toxicity includes AEs (fatigue, diarrhea, mucositis) related to thebiological agent and chemo-related AEs (renal tox, neurotoxicity)
Y1 = 0 if no grade 3,4 (severe) TOXY1 = 1 if grade 3,4 TOX occurs, but resolved within 2 weeksY1 = 2 if grade 3,4 TOX occurs, & not resolved within 2 weeks
![Page 24: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/24.jpg)
UTILITY-BASED DOSE-FINDING 5
Clinical Outcomes (evaluated over two 28-day cycles )
Toxicity includes AEs (fatigue, diarrhea, mucositis) related to thebiological agent and chemo-related AEs (renal tox, neurotoxicity)
Y1 = 0 if no grade 3,4 (severe) TOXY1 = 1 if grade 3,4 TOX occurs, but resolved within 2 weeksY1 = 2 if grade 3,4 TOX occurs, & not resolved within 2 weeks
Efficacy is evaluated by the end of two cycles (day 56)
Y2 = 0 if progressive disease (PD) at any time in the first 2 cyclesY2 = 1 if stable disease (SD) at day 56Y2 = 2 if complete or partial remission (CR/PR) at day 56
![Page 25: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/25.jpg)
UTILITY-BASED DOSE-FINDING 6
Interim Within-Patient Treatment Modifications
![Page 26: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/26.jpg)
UTILITY-BASED DOSE-FINDING 6
Interim Within-Patient Treatment Modifications
1. If grade 1 or 2 non-haematologic TOX in cycle 1, the dose ofthe biological agent is reduced
2. If the patient does not recover from a grade 3, 4non-haematologoc TOX in two weeks, the bio agent is stopped,but the patient may continue to receive the chemo agents(physician decision)
3. If Y1 = 2 (unresolved gr. 3,4 TOX) or Y2 = 2 (PD) thentreatment is stopped
![Page 27: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/27.jpg)
UTILITY-BASED DOSE-FINDING 6
Interim Within-Patient Treatment Modifications
1. If grade 1 or 2 non-haematologic TOX in cycle 1, the dose ofthe biological agent is reduced
2. If the patient does not recover from a grade 3, 4non-haematologoc TOX in two weeks, the bio agent is stopped,but the patient may continue to receive the chemo agents(physician decision)
3. If Y1 = 2 (unresolved gr. 3,4 TOX) or Y2 = 2 (PD) thentreatment is stopped
4. Y1 = 2 and no PD before day 56 =⇒ Y2 is inevaluable
![Page 28: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/28.jpg)
UTILITY-BASED DOSE-FINDING 7
The Outcome Domain
Y2
0 = PD 1 = SD 2 = CR/PR
0 (0, 0) (0, 1) (0, 2) –
Y1 1 (1, 0) (1, 1) (1, 2) –
2 (2, 0) (2, 1) (2, 2) (2, Ineval)
10 elementary outcomes, including {Y1 = 2 and Y2 inevaluable}
![Page 29: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/29.jpg)
UTILITY-BASED DOSE-FINDING 8
A General Probability Model
![Page 30: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/30.jpg)
UTILITY-BASED DOSE-FINDING 8
A General Probability Model
Outcomes Y = (Y1, Y2) where ordinal Yj ∈ {1, ...,mj}
where mj = # levels of outcome j = 1, 2
![Page 31: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/31.jpg)
UTILITY-BASED DOSE-FINDING 8
A General Probability Model
Outcomes Y = (Y1, Y2) where ordinal Yj ∈ {1, ...,mj}
where mj = # levels of outcome j = 1, 2
Doses ddd = (d1, d2)
![Page 32: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/32.jpg)
UTILITY-BASED DOSE-FINDING 8
A General Probability Model
Outcomes Y = (Y1, Y2) where ordinal Yj ∈ {1, ...,mj}
where mj = # levels of outcome j = 1, 2
Doses ddd = (d1, d2)
Marginals πk,y(ddd, θθθ) = Pr(Yk = y | ddd, θθθ)
![Page 33: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/33.jpg)
UTILITY-BASED DOSE-FINDING 8
A General Probability Model
Outcomes Y = (Y1, Y2) where ordinal Yj ∈ {1, ...,mj}
where mj = # levels of outcome j = 1, 2
Doses ddd = (d1, d2)
Marginals πk,y(ddd, θθθ) = Pr(Yk = y | ddd, θθθ)
Joint pmf πππ(yyy| ddd, θθθ) = Pr(Y = yyy | Z = 1, ddd, θθθ)
where yyy = (y1, y2) is observed if Z = 1
![Page 34: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/34.jpg)
UTILITY-BASED DOSE-FINDING 9
Likelihood
Z = I(Y2 is evaluable), ζ = Pr(Z = 1)
δ(yyy) = I(Y = yyy, Z = 1)
δ1(y1) = I(Y1 = y1)
L(Y, Z | ddd, θθθ) =
1∏Z=0
[ζ
∏yyy
{πππ(yyy| ddd, θθθ)
}δ(yyy)]Z [
(1− ζ)2∏
y1=0
{π1,y1(ddd, θθθ)
}δ1(y1)]1−Z
![Page 35: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/35.jpg)
UTILITY-BASED DOSE-FINDING 10
![Page 36: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/36.jpg)
UTILITY-BASED DOSE-FINDING 11
The Aranda-Ordaz (AO) Model (1983)
![Page 37: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/37.jpg)
UTILITY-BASED DOSE-FINDING 11
The Aranda-Ordaz (AO) Model (1983)
Given linear term η(d,ααα) the AO model is
Pr(Y = 1|d,ααα) = ξ{η(d,ααα), λ} = 1− (1 + λeη(d,ααα))−1/λ, λ > 0
1. For η(d,ααα) = α0 + α1log(d), d = 0 =⇒ ξ = 0 =⇒ The outcome isimpossible if no treatment is given
![Page 38: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/38.jpg)
UTILITY-BASED DOSE-FINDING 11
The Aranda-Ordaz (AO) Model (1983)
Given linear term η(d,ααα) the AO model is
Pr(Y = 1|d,ααα) = ξ{η(d,ααα), λ} = 1− (1 + λeη(d,ααα))−1/λ, λ > 0
1. For η(d,ααα) = α0 + α1log(d), d = 0 =⇒ ξ = 0 =⇒ The outcome isimpossible if no treatment is given
2. For η(d,ααα) = α0 + α1d, d = 0 =⇒ ξ = 1− (1 + λeα0)−1/λ =baseline Pr(Y=1) without treatment
![Page 39: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/39.jpg)
UTILITY-BASED DOSE-FINDING 11
The Aranda-Ordaz (AO) Model (1983)
Given linear term η(d,ααα) the AO model is
Pr(Y = 1|d,ααα) = ξ{η(d,ααα), λ} = 1− (1 + λeη(d,ααα))−1/λ, λ > 0
1. For η(d,ααα) = α0 + α1log(d), d = 0 =⇒ ξ = 0 =⇒ The outcome isimpossible if no treatment is given
2. For η(d,ααα) = α0 + α1d, d = 0 =⇒ ξ = 1− (1 + λeα0)−1/λ =baseline Pr(Y=1) without treatment
3. λ=1 =⇒ ξ{η(d,ααα), 1} = eη(d,α)
1+eη(d,α) (logistic)
![Page 40: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/40.jpg)
UTILITY-BASED DOSE-FINDING 11
The Aranda-Ordaz (AO) Model (1983)
Given linear term η(d,ααα) the AO model is
Pr(Y = 1|d,ααα) = ξ{η(d,ααα), λ} = 1− (1 + λeη(d,ααα))−1/λ, λ > 0
1. For η(d,ααα) = α0 + α1log(d), d = 0 =⇒ ξ = 0 =⇒ The outcome isimpossible if no treatment is given
2. For η(d,ααα) = α0 + α1d, d = 0 =⇒ ξ = 1− (1 + λeα0)−1/λ =baseline Pr(Y=1) without treatment
3. λ=1 =⇒ ξ{η(d,ααα), 1} = eη(d,α)
1+eη(d,α) (logistic)
4. limλ→0 ξ(η(d,ααα), λ) = 1− exp{−eη(d,α)} (compl. log-log)
![Page 41: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/41.jpg)
UTILITY-BASED DOSE-FINDING 12
![Page 42: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/42.jpg)
UTILITY-BASED DOSE-FINDING 13
Dose
Prob
(Res
pons
e)
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
Dose
Prob
(Res
pons
e)
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
Dose
Prob
(Res
pons
e)
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
DosePr
ob(R
espo
nse)
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
Dose
Prob
(Res
pons
e)
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
Dose
Prob
(Res
pons
e)
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
![Page 43: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/43.jpg)
UTILITY-BASED DOSE-FINDING 14
A Generalized Aranda-Ordaz (GAO) Model
![Page 44: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/44.jpg)
UTILITY-BASED DOSE-FINDING 14
A Generalized Aranda-Ordaz (GAO) Model
To Accommodate Two Linear Terms:
η1 ≡ η1(d1, ααα1) for d1 and η2 ≡ η1(d2, ααα2) for d2
![Page 45: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/45.jpg)
UTILITY-BASED DOSE-FINDING 14
A Generalized Aranda-Ordaz (GAO) Model
To Accommodate Two Linear Terms:
η1 ≡ η1(d1, ααα1) for d1 and η2 ≡ η1(d2, ααα2) for d2
ξ∗{η1, η2, λ, γ} = 1− {1 + λ(eη1 + eη2 + γeη1+η2)}−1/λ
γ accounts for interaction between the two agents
γ = 0 =⇒ Additive effects eη1 and eη2 in the GAO model
![Page 46: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/46.jpg)
UTILITY-BASED DOSE-FINDING 15
Linear terms determining the marginal of Yk
![Page 47: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/47.jpg)
UTILITY-BASED DOSE-FINDING 15
Linear terms determining the marginal of Yk
For j = dose, k = outcome, y = value of Yk
η(j)k,y(dj, ααα
(j)k ) = α
(j)k,y,0 + α
(j)k,y,1 (dj − d̄j)
![Page 48: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/48.jpg)
UTILITY-BASED DOSE-FINDING 15
Linear terms determining the marginal of Yk
For j = dose, k = outcome, y = value of Yk
η(j)k,y(dj, ααα
(j)k ) = α
(j)k,y,0 + α
(j)k,y,1 (dj − d̄j)
1. α(1)k,y,0 and α
(2)k,y,0 are intercepts
![Page 49: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/49.jpg)
UTILITY-BASED DOSE-FINDING 15
Linear terms determining the marginal of Yk
For j = dose, k = outcome, y = value of Yk
η(j)k,y(dj, ααα
(j)k ) = α
(j)k,y,0 + α
(j)k,y,1 (dj − d̄j)
1. α(1)k,y,0 and α
(2)k,y,0 are intercepts
2. α(1)k,y,1 and α
(2)k,y,1 are dose effects
![Page 50: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/50.jpg)
UTILITY-BASED DOSE-FINDING 15
Linear terms determining the marginal of Yk
For j = dose, k = outcome, y = value of Yk
η(j)k,y(dj, ααα
(j)k ) = α
(j)k,y,0 + α
(j)k,y,1 (dj − d̄j)
1. α(1)k,y,0 and α
(2)k,y,0 are intercepts
2. α(1)k,y,1 and α
(2)k,y,1 are dose effects
3. ααα(j)k = (α(j)
k,1,0, α(j)k,1,1, α
(j)k,2,0, α
(j)k,2,1)
![Page 51: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/51.jpg)
UTILITY-BASED DOSE-FINDING 15
Linear terms determining the marginal of Yk
For j = dose, k = outcome, y = value of Yk
η(j)k,y(dj, ααα
(j)k ) = α
(j)k,y,0 + α
(j)k,y,1 (dj − d̄j)
1. α(1)k,y,0 and α
(2)k,y,0 are intercepts
2. α(1)k,y,1 and α
(2)k,y,1 are dose effects
3. ααα(j)k = (α(j)
k,1,0, α(j)k,1,1, α
(j)k,2,0, α
(j)k,2,1)
4. θθθk = (ααα(1)k , ααα
(2)k , λk, γk)
![Page 52: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/52.jpg)
UTILITY-BASED DOSE-FINDING 16
Marginal Distributions
![Page 53: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/53.jpg)
UTILITY-BASED DOSE-FINDING 16
Marginal Distributions
For each outcome k = 1, 2 and levels y = 1, ...,mk,
Pr(Yk ≥ y | Yk ≥ y−1, ddd, θθθk) = ξ∗{η(1)k,y(d1, ααα
(1)k ), η(2)
k,y(d2, ααα(2)k ), λk, γk}
≡ ξ∗k,y(ddd, θθθk) =⇒
![Page 54: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/54.jpg)
UTILITY-BASED DOSE-FINDING 16
Marginal Distributions
For each outcome k = 1, 2 and levels y = 1, ...,mk,
Pr(Yk ≥ y | Yk ≥ y−1, ddd, θθθk) = ξ∗{η(1)k,y(d1, ααα
(1)k ), η(2)
k,y(d2, ααα(2)k ), λk, γk}
≡ ξ∗k,y(ddd, θθθk) =⇒
πk,0(ddd, θθθk) = 1− ξ∗k,1(ddd, θθθk)
πk,y(ddd, θθθk) = {1− ξ∗k,y+1(ddd, θθθk)}∏y
j=1 ξ∗k,j(ddd, θθθk), 1 ≤ y ≤ mk − 1
πk,mk(ddd, θθθk) =
∏mkj=1 ξ∗k,j(ddd, θθθk)
![Page 55: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/55.jpg)
UTILITY-BASED DOSE-FINDING 17
In the three-level ordinal outcome case, the unconditionalmarginal probabilities are
πk,0(ddd, θθθk) = 1− ξ∗k,1(ddd, θθθk)
πk,1(ddd, θθθk) = ξ∗k,1(ddd, θθθk)− ξ∗k,1(ddd, θθθk)ξ∗k,2(ddd, θθθk)
πk,2(ddd, θθθk) = ξ∗k,1(ddd, θθθk) ξ∗k,2(ddd, θθθk)
![Page 56: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/56.jpg)
UTILITY-BASED DOSE-FINDING 18
Bivariate Distribution of (Y1, Y2)
![Page 57: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/57.jpg)
UTILITY-BASED DOSE-FINDING 18
Bivariate Distribution of (Y1, Y2)
Use a Gaussian copula, given by
Cρ(u, v) = Φρ{Φ−1(u),Φ−1(v)} for 0 ≤ u, v ≤ 1.
Φρ = cdf of a bivariate std normal with correlation ρ
Φ = usual N(0,1) cdf
![Page 58: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/58.jpg)
UTILITY-BASED DOSE-FINDING 19
The cdf of each Yk is
Fk(y | ddd, θθθk) =
0 ify < 01− πk,1(ddd, θθθk)− πk,2(ddd, θθθk) if0 ≤ y < 11− πk,2(ddd, θθθk) if1 ≤ y < 21 ify ≥ 2
![Page 59: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/59.jpg)
UTILITY-BASED DOSE-FINDING 19
The cdf of each Yk is
Fk(y | ddd, θθθk) =
0 ify < 01− πk,1(ddd, θθθk)− πk,2(ddd, θθθk) if0 ≤ y < 11− πk,2(ddd, θθθk) if1 ≤ y < 21 ify ≥ 2
The joint pmf of Y is
πππ(yyy| ddd, θθθ) =2∑
a=1
2∑b=1
(−1)a+bCρ(ua, vb)
u1 = F1(y1 | ddd, θθθ), v1 = F2(y2 | ddd, θθθ)
u2 = F1(y1 − 1 | ddd, θθθ), v2 = F2(y2 − 1 | ddd, θθθ)
![Page 60: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/60.jpg)
UTILITY-BASED DOSE-FINDING 20
πππ(yyy) = Φρ{Φ−1 ◦ F1(y1), Φ−1 ◦ F2(y2)}
− Φρ{Φ−1 ◦ F1(y1 − 1), Φ−1 ◦ F2(y2)}
− Φρ{Φ−1 ◦ F1(y1), Φ−1 ◦ F2(y2 − 1)}
+ Φρ{Φ−1 ◦ F1(y1 − 1), Φ−1 ◦ F2(y2 − 1)}
θθθ = (θθθ1, θθθ2, ρ) and dim(θθθ) = 21
![Page 61: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/61.jpg)
UTILITY-BASED DOSE-FINDING 21
Establishing Priors
![Page 62: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/62.jpg)
UTILITY-BASED DOSE-FINDING 21
Establishing Priors
1. Elicit prior means of πk,1(ddd) and πk,2(ddd) for each ddd andoutcome k = 1, 2
![Page 63: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/63.jpg)
UTILITY-BASED DOSE-FINDING 21
Establishing Priors
1. Elicit prior means of πk,1(ddd) and πk,2(ddd) for each ddd andoutcome k = 1, 2
2. Use an extension of the least squares method of Thall andCook (2004) to solve for prior means of the 21 modelparameters
![Page 64: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/64.jpg)
UTILITY-BASED DOSE-FINDING 21
Establishing Priors
1. Elicit prior means of πk,1(ddd) and πk,2(ddd) for each ddd andoutcome k = 1, 2
2. Use an extension of the least squares method of Thall andCook (2004) to solve for prior means of the 21 modelparameters
3. Calibrate prior variances to obtain reasonably small prioreffective sample sizes of the priors on πk,1(ddd) and πk,2(ddd) =⇒ESS = .20 to .70
![Page 65: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/65.jpg)
UTILITY-BASED DOSE-FINDING 21
Establishing Priors
1. Elicit prior means of πk,1(ddd) and πk,2(ddd) for each ddd andoutcome k = 1, 2
2. Use an extension of the least squares method of Thall andCook (2004) to solve for prior means of the 21 modelparameters
3. Calibrate prior variances to obtain reasonably small prioreffective sample sizes of the priors on πk,1(ddd) and πk,2(ddd) =⇒ESS = .20 to .70
![Page 66: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/66.jpg)
UTILITY-BASED DOSE-FINDING 22
Establishing Utilities
![Page 67: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/67.jpg)
UTILITY-BASED DOSE-FINDING 22
Establishing Utilities
The Delphi method: A tool for establishing consensus amongexperts (Dalkey, 1969; Brook, Chassin, Fink, et al., 1986)
![Page 68: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/68.jpg)
UTILITY-BASED DOSE-FINDING 22
Establishing Utilities
The Delphi method: A tool for establishing consensus amongexperts (Dalkey, 1969; Brook, Chassin, Fink, et al., 1986)
1. Elicit utilites from each of several individuals
![Page 69: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/69.jpg)
UTILITY-BASED DOSE-FINDING 22
Establishing Utilities
The Delphi method: A tool for establishing consensus amongexperts (Dalkey, 1969; Brook, Chassin, Fink, et al., 1986)
1. Elicit utilites from each of several individuals
2. Show all elicited values (anonymously) to all individuals, andallow them to adjust their utilties
![Page 70: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/70.jpg)
UTILITY-BASED DOSE-FINDING 22
Establishing Utilities
The Delphi method: A tool for establishing consensus amongexperts (Dalkey, 1969; Brook, Chassin, Fink, et al., 1986)
1. Elicit utilites from each of several individuals
2. Show all elicited values (anonymously) to all individuals, andallow them to adjust their utilties
3. Repeat 2 or 3 times, and compute the mean across experts ofthe final values
![Page 71: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/71.jpg)
UTILITY-BASED DOSE-FINDING 22
Establishing Utilities
The Delphi method: A tool for establishing consensus amongexperts (Dalkey, 1969; Brook, Chassin, Fink, et al., 1986)
1. Elicit utilites from each of several individuals
2. Show all elicited values (anonymously) to all individuals, andallow them to adjust their utilties
3. Repeat 2 or 3 times, and compute the mean across experts ofthe final values
Nadine Houede did this via questionnaire with 8 French medicaloncologists who treat bladder cancer patients, using a utilityscale of 0 to 100, and finished after 2 rounds
![Page 72: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/72.jpg)
UTILITY-BASED DOSE-FINDING 24
French Utilities Elicited Using the Delphi Method
![Page 73: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/73.jpg)
UTILITY-BASED DOSE-FINDING 24
French Utilities Elicited Using the Delphi Method
Y2 = 0 Y2 = 1 Y2 = 2 Y2
PD SD CR/PR Inevaluable
Y1 = 0 25 76 100 –
Y1 = 1 10 60 82 –
Y1 = 2 2 40 52 0
![Page 74: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/74.jpg)
UTILITY-BASED DOSE-FINDING 25
Using Utilities to Select Dose Pairs
![Page 75: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/75.jpg)
UTILITY-BASED DOSE-FINDING 25
Using Utilities to Select Dose Pairs
For U(y) = the numerical utility of outcome y, the mean utilityfor a patient treated with ddd is
u(ddd, θθθ) = EY{U(Y) | ddd, θθθ} =∑yyy
U(yyy)πππ(yyy | ddd, θθθ)
![Page 76: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/76.jpg)
UTILITY-BASED DOSE-FINDING 25
Using Utilities to Select Dose Pairs
For U(y) = the numerical utility of outcome y, the mean utilityfor a patient treated with ddd is
u(ddd, θθθ) = EY{U(Y) | ddd, θθθ} =∑yyy
U(yyy)πππ(yyy | ddd, θθθ)
Given current datan = {(Y1, ddd1), · · · , (Yn, dddn)}, select
dddopt(datan) = argmaxddd∈D
∫θθθ
u(ddd, θθθ)p(θθθ | datan)dθθθ
= argmaxddd∈D
∑yyy
U(yyy)∫θθθ
πππ(yyy | ddd, θθθ)p(θθθ | datan)dθθθ
![Page 77: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/77.jpg)
UTILITY-BASED DOSE-FINDING 26
Additional Safety Rules
![Page 78: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/78.jpg)
UTILITY-BASED DOSE-FINDING 26
Additional Safety Rules
1. Do Not Skip Untried Dose Pairs:
If (d1, d2) is the current dose pair, then escalation is allowed to asyet untried pairs (d1 + 1, d2), (d1, d2 + 1), or (d1 + 1, d2 + 1).
![Page 79: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/79.jpg)
UTILITY-BASED DOSE-FINDING 26
Additional Safety Rules
1. Do Not Skip Untried Dose Pairs:
If (d1, d2) is the current dose pair, then escalation is allowed to asyet untried pairs (d1 + 1, d2), (d1, d2 + 1), or (d1 + 1, d2 + 1).
2. Stop The Trial if All Dose Pairs are Too Toxic:
For πmax1,2 = fixed upper limit on Pr(Y2 = 2), and pU = a fixed upper
probability cut-off (e.g. .80 to .90), terminate accrual if
minddd Pr{π1,2(ddd, θθθ) > πmax1,2 | data} > pU
![Page 80: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/80.jpg)
UTILITY-BASED DOSE-FINDING 27
Simulations
![Page 81: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/81.jpg)
UTILITY-BASED DOSE-FINDING 27
Simulations
Nmax = 48, choose ddd for cohorts of 3 patients, start at ddd = (2,2),safety stopping rule applied with πmax
1,2 = .33 and pU = .80.
![Page 82: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/82.jpg)
UTILITY-BASED DOSE-FINDING 27
Simulations
Nmax = 48, choose ddd for cohorts of 3 patients, start at ddd = (2,2),safety stopping rule applied with πmax
1,2 = .33 and pU = .80.
A simulation scenario is a vector πππtrue(ddd) of fixed probabilityvalues for the 12 ddd pairs, with ρtrue = 0.10.
utrue(ddd) =∑
yyy U(yyy) πππtrue(yyy | ddd)
![Page 83: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/83.jpg)
UTILITY-BASED DOSE-FINDING 27
Simulations
Nmax = 48, choose ddd for cohorts of 3 patients, start at ddd = (2,2),safety stopping rule applied with πmax
1,2 = .33 and pU = .80.
A simulation scenario is a vector πππtrue(ddd) of fixed probabilityvalues for the 12 ddd pairs, with ρtrue = 0.10.
utrue(ddd) =∑
yyy U(yyy) πππtrue(yyy | ddd)
1000 runs per scenario
![Page 84: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/84.jpg)
UTILITY-BASED DOSE-FINDING 28
![Page 85: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/85.jpg)
UTILITY-BASED DOSE-FINDING 29
57.3
2 0.7Che
mo
Age
nt D
ose
3
True UtilityProb SelectNum Patients
56.9
2 0.8
Scenario 1
60.4
8 3.3
64.2
17
6.0
60.5
2 1.2Che
mo
Age
nt D
ose
2
60.1
2
6.2
63.6
12
7.9
67.4 2811.3
54.6 0 0.3
Biol Agent Dose 1
Che
mo
Age
nt D
ose
1
54.2 0 0.3
Biol Agent Dose 2
57.8
2 1.2
Biol Agent Dose 3
61.7
26
8.8
Biol Agent Dose 4
![Page 86: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/86.jpg)
UTILITY-BASED DOSE-FINDING 30
![Page 87: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/87.jpg)
UTILITY-BASED DOSE-FINDING 31
47.2
3 1.0Che
mo
Age
nt D
ose
3
True UtilityProb SelectNum Patients
46.2
1 1.0
Scenario 2
44.2 3 1.2
42.2 1 0.6
54.6
12 3.5
Che
mo
Age
nt D
ose
2
65.8 42 20.7
61.8
30 13.4
43.4 1 1.6
44.6 4 1.2
Biol Agent Dose 1
Che
mo
Age
nt D
ose
1
44.9
1 1.0
Biol Agent Dose 2
45.1
2 1.7
Biol Agent Dose 3
39.1 1 0.9
Biol Agent Dose 4
![Page 88: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/88.jpg)
UTILITY-BASED DOSE-FINDING 32
![Page 89: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/89.jpg)
UTILITY-BASED DOSE-FINDING 33
48.6
1 0.7Che
mo
Age
nt D
ose
3
True UtilityProb SelectNum Patients
53.1
2 0.9
Scenario 3
59.6
11 4.1
64.9 2711.2
43.9 0 0.3C
hem
o A
gent
Dos
e 2
50.8
1
4.0
55.2
3
4.2
63.5 28
11.4
41.4 0 0.1
Biol Agent Dose 1
Che
mo
Age
nt D
ose
1
48.9
0 0.1
Biol Agent Dose 2
53.4
1 0.8
Biol Agent Dose 3
58.8
27
10.2
Biol Agent Dose 4
![Page 90: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/90.jpg)
UTILITY-BASED DOSE-FINDING 34
![Page 91: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/91.jpg)
UTILITY-BASED DOSE-FINDING 35
61.2
12 3.5
Che
mo
Age
nt D
ose
3
51.0
2 1.0
Scenario 4
38.0
1 2.026.0 2
2.9
True UtilityProb SelectNum Patients
65.4
14 4.5
Che
mo
Age
nt D
ose
2
53.6
2
5.7
45.3
1
5.3 37.3
3
5.0
73.9 53
12.1
Biol Agent Dose 1
Che
mo
Age
nt D
ose
1
61.9
1 0.4
Biol Agent Dose 2
55.2
1 1.1
Biol Agent Dose 3
44.9
8 4.5
Biol Agent Dose 4
![Page 92: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/92.jpg)
UTILITY-BASED DOSE-FINDING 36
Additional Simulation Results
![Page 93: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/93.jpg)
UTILITY-BASED DOSE-FINDING 36
Additional Simulation Results
1. Safe : πmax1,2 = .33 and pU = .80 =⇒ Large PSTOP in toxic
scenarios
![Page 94: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/94.jpg)
UTILITY-BASED DOSE-FINDING 36
Additional Simulation Results
1. Safe : πmax1,2 = .33 and pU = .80 =⇒ Large PSTOP in toxic
scenarios
2. Insensitive to Cohort Size : c = 1 versus c = 3 give thevirtually same results under Scenarios 1 – 4, but slightly largerPSTOP for c = 1 (93% versus 90%) under the toxic Scenario 5
![Page 95: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/95.jpg)
UTILITY-BASED DOSE-FINDING 36
Additional Simulation Results
1. Safe : πmax1,2 = .33 and pU = .80 =⇒ Large PSTOP in toxic
scenarios
2. Insensitive to Cohort Size : c = 1 versus c = 3 give thevirtually same results under Scenarios 1 – 4, but slightly largerPSTOP for c = 1 (93% versus 90%) under the toxic Scenario 5
3. Consistent : Pr{Select ddd having largest utrue(ddd)} ↑ with Nmax
![Page 96: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/96.jpg)
UTILITY-BASED DOSE-FINDING 36
Additional Simulation Results
1. Safe : πmax1,2 = .33 and pU = .80 =⇒ Large PSTOP in toxic
scenarios
2. Insensitive to Cohort Size : c = 1 versus c = 3 give thevirtually same results under Scenarios 1 – 4, but slightly largerPSTOP for c = 1 (93% versus 90%) under the toxic Scenario 5
3. Consistent : Pr{Select ddd having largest utrue(ddd)} ↑ with Nmax
4. Stupid Priors Give Stupid Designs : A naive “uninformative”prior with E(θθθ) = 0 and all prior standard deviations = 1000 givesterrible results
![Page 97: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/97.jpg)
UTILITY-BASED DOSE-FINDING 37
Conclusions
![Page 98: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/98.jpg)
UTILITY-BASED DOSE-FINDING 37
Conclusions
1. Utilities including Future Patient Benefit, Cost, ExpectedProfit, etc. may lead to better designs and more publications.
![Page 99: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/99.jpg)
UTILITY-BASED DOSE-FINDING 37
Conclusions
1. Utilities including Future Patient Benefit, Cost, ExpectedProfit, etc. may lead to better designs and more publications.
2. French utilities are not necessarily the same as Americanutilities =⇒ Think globally but elicit utilities locally.
![Page 100: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/100.jpg)
UTILITY-BASED DOSE-FINDING 37
Conclusions
1. Utilities including Future Patient Benefit, Cost, ExpectedProfit, etc. may lead to better designs and more publications.
2. French utilities are not necessarily the same as Americanutilities =⇒ Think globally but elicit utilities locally.
3. Many extensions are possible, and this also may lead to betterdesigns and more publications.
![Page 101: Utility-Based Optimization of Combination Therapy Using ...Utility-Based Optimization of Combination Therapy Using Ordinal Toxicity and Efficacy in Phase I/II Clinical Trials Peter](https://reader031.vdocuments.us/reader031/viewer/2022041801/5e51a3c682320242626745ad/html5/thumbnails/101.jpg)
UTILITY-BASED DOSE-FINDING 37
Conclusions
1. Utilities including Future Patient Benefit, Cost, ExpectedProfit, etc. may lead to better designs and more publications.
2. French utilities are not necessarily the same as Americanutilities =⇒ Think globally but elicit utilities locally.
3. Many extensions are possible, and this also may lead to betterdesigns and more publications.