Clinical Research Methods

PHASE I CLINICAL TRIALS

Connie Lee Batlevi

David R. Spriggs

Phase 0 Trials

Proof of concept trials designed to be first in human (Nat Rev Cancer 2007;7:131) administered while toxicology studies are being performed for IND submission
Single doses of new drug given at low concentrations not expected to cause tox: Correlated w/biomarker assay, imaging technique, of drug activity w/drug administration in vivo

Phase I Trials

One of the initial phases of clinical experimentation
- First-in-human trials: Promising preclinical data, clinical efficacy unproven novel Rx combinations of existing and/or new agents (eg, carboplatin/paclitaxel + PI3K inhibitor)
Primary objectives: Identify drug tox, describe drug pharmacology (PK, PD), identify RP2D
Population: Heterogeneous (eg, nonhistology specific), newer trials of molecular agents may select pts based on genotype, often includes pts who have failed multiple lines of prior tx
Starting dose: Dose chosen to treat first cohort of pts based on animal studies, commonly 1/10-1/3 of mouse LD10 (dose resulting in death of 10% of mice), other animal species may be used

Definition of Terms
Dose-limiting toxicity (DLT)	Unacceptable tox that limit further dose escalation, protocol-specific, defined based on CTCAE criteria (note: Not all grade 3 tox are dose-limiting), occur w/in a specific time-frame (eg, w/in first cycle)
Maximum tolerated dose (MTD)	Highest dose level at which a defined proportion of pts (eg, 33%) experience a DLT
Target tox level	Typically 20-33%, maximum probability of DLT considered acceptable
Recommended phase 2 dose (RP2D)	Dose recommended for subsequent phase II trials: Either the MTD or one dose level below MTD
Optimal biologic dose (OBD)	Dose a/w most desirable prespecified effect on a biomarker
Dose-efficacy curve	Reflects relationship between dose & probability of efficacy for a specific agent
Dose-tox curve	Reflects relationship between dose & probability of tox for a specific drug
PK	Drug effects on absorption, distribution, metabolism, & excretion, several PK levels taken along study course
PD	Drug effects on host processes (eg, molecular correlates or biomarkers, imaging endpoints, includes hematologic & nonhematologic tox)
Therapeutic index	Dose required to produce critical tox divided by dose yielding defined antitumor effect
(Adapted from J Natl Cancer Inst 2009;101:708)

Types of Phase I Designs

Rule-based designs
- Egs: Standard 3+3, AT, PGDE
- Traditional phase I designs w/no prior assumptions of dose-tox curve, allow dose escalation & reduction, easily implemented w/o need for special mathematical software → relatively inefficient at establishing MTD, RP2D based on current dose level & does not incorporate all tox data
Model-based designs
- Egs: CRM, escalation w/O/D control, TITE-CRM, EffTox, TriCRM
- Statistical model: Starts w/initial dose-tox curve → model modified by tox experienced by pts on trial; allows the incorporation of all available tox data, RP2D estimated w/confidence interval

Traditional 3+3

Classic phase I clinical trial design where pts are treated initially in cohorts of 3s at escalating dose levels, dose level escalation or de-escalation follows modified Fibonacci sequence: 100% → 67% → 50% → 40% → 33%, intra-pt dose escalation is typically not allowed

Relatively simple to implement, provides more info on PK inter-pt variability per dose level, defined sequence of escalation & de-escalation

3+3 Design Based on Number of Patients Experiencing DLT Per Cohort
Cohort	Dose Escalate (↑ to next higher dose level)	Expand Cohort (↑ cohort by 3 pts)	Dose De-escalate (↓ to next lower dose level)
3-Pts	0 DLTs	1 DLT	≥2 DLTs
6-Pts (expanded)	1 DLT	N/A	≥2/6 DLTs

MTD definition = Highest dose where 0-1/16 pts w/DLT (estimates MTD at 33%) or next higher dose level
Pitfalls: Grossly underestimate MTD, dose escalation is slow, many pts treated at subtherapeutic (below RP2D) dose levels

Accelerated Titration (J Natl Cancer Inst 1997;89:1138)

Affords rapid initial dose escalation, wide variety of designs, intra-pt dose escalation permitted in some designs
Design eg: Initial dose-escalation phase w/single-pt cohorts, dose is escalated rapidly at 96-100% of previous dose → defined stopping rule is hit (eg, 1 DLT or 2 mod. tox observed) → reverts to 3+3 design w/40% dose increments between cohorts
MTD is estimated using all tox data collected during trial, some designs may use 3+3

rules to determine MTD
Advantages: Reduces number of pts treated at subtherapeutic doses, designs w/intra-pt dose escalation affords tx at higher & potentially more effective doses, disadvantages: Intra-pt dose escalation may mask cumulative/delayed tox, aggressive & may expose more pts to higher doses, requires acute tox (chronic/cumulative tox not captured)

Pharmacologically-Guided Dose Escalation (J Natl Cancer Inst 1990;82:1321)

Assumes DLT predictable by plasma drug concentrations: Plasma exposure defined by AUC extrapolated from preclinical data
Design: Single-pt dose escalation (based on PK data, typically 100%) → stopping rule hit (eg, target AUC reached or DLTs occur) → switch to 3+3
Provides rapid dose escalation but not widely used due to difficulty obtaining real-time PKs, dose escalation may be hampered by inter-pt variability

Continual Reassessment Method (Biometrics 1990;46:33)

Bayesian approach: Starts w/defined acceptable probability of DLT (20-33%), 1-3

pts selected per cohort depending on accrual & desired DLT rate, starts at dose thought to be close to MTD
Mathematical model for dose-tox curve continuously modified as each enrolled pt experiences a DLT → dose is then escalated & de-escalated accordingly (next pt is treated at new estimated dose close to MTD) → trial stopped if prespecified condition met or precision in probability of DLT at estimated MTD level is achieved
Requires intensive statistical assistance, thought to provide more precise MTD estimate, may be safer w/fewer pts treated at toxic doses

Other Model-Based Designs

Modified CRM design: Escalation w/O/D control (Stat Med 1998;17:1103)
TITE-CRM (Biometrics 2000;56:1177): Incorporates time-to-event of tox for each pt, assumes hazard of tox remains constant
Efficacy & tox models: Eff Tox defines acceptable dose combination based on trade-offs between probabilities of efficacy & tox, TriCRM w/3 potential trial outcomes (no efficacy no tox, efficacy only, tox only)

PHASE II and III CLINICAL TRIALS

Connie Lee Batlevi

Robert J. Motzer

Definition of Terms
Historical controls	Pt prev treated (or not) used as standard for comparison to pt being treated in current experimental design
Type I error (α)	FP, ↓ type I error by increasing sample size
Type II error (β)	FN, ↓ type II error by increasing sample size
Effect size	Difference between null & alternative hypothesis values
Binomial proportion test	Assuming equal distribution, determines statistical significance of deviations from expected distribution
Hazard ratio	Ratio of standard/experimental tx failure rates
Logrank test	Nonparametric test (assumes data is skewed) establish efficacy of new tx vs. control when measurement is time-to-event