Design and Analysis of Phase I Clinical Trials in Cancer Therapy

1
Design and Analysis of Phase I Clinical Trials in
Cancer Therapy

• Merrill J. Egorin, M.D.
• University of Pittsburgh Cancer Institute

• Modified from previous talks given by Alex Adjei
  others
• I get to do it this year because Alex is not
  participating in the course no one else was
  silly enough to agree to do it.

2
Exposures Disclosures

• I have the following financial relationships to
  disclose
• Consultant for Novartis, Bristol-Myers Squibb,
  Saladax, Amplimed, Adherex, Johnson and Johnson,
  Daiichi Sankyo
• Grant/Research support from Novartis,
  Bristol-Myers Squibb, Merck, Infinity, The NCI
• Honoraria from Novartis, Bristol-Myers Squibb

3
Questions I Will Try to Cover

• What are Phase I studies?
• This is sort of easy to address.
• Why do we do Phase I studies?
• This seems sort of easy to address.
• What do we need to do Phase I studies?
• I think this should be rather straightforward.
• How do we do Phase I studies?
• This is a bag of worms!

4
Phase I Studies in Oncology

• First evaluation of a new cancer therapy in
  humans
• First-in-human single agent study
• Last year someone said they were doing the
  second first-in-human study of some agent (I am
  still thinking about that!
• Combination of novel agents
• Combination novel agent and approved agent
• Combination of approved standard agents?
  Combination of novel agent and radiation therapy
• Eligible patients usually have refractory solid
  tumors of any type
• Not necessarily the same definition used in other
  branches of medicine.

5
Why Do a Phase I Study?(Sort of Like Why Did the
Chicken Cross the Road?)

• Primary objective (there can only be one of
  these.)
• The critical information needed to do a phase II
  study.
• MTD/toxicities
• Secondary objectives (These are sexy, but
  secondary!)
• PK (my favorite)
• PD and biomarkers
• Response (even the Pirates win sometimes

6
What Is Needed To Do A Phase I Study?

• An Investigational New Drug Application (IND)
• Provides a means of advancing from pre-clinical
  to clincal testing
• Required for unmarketed/unaproved products
• May be required for already marketed products
• A formal application to study and intervention in
  patients
• Commercial-sponsor usually a pharmaceutical
  company
• Non-commercial
• Investigator
• Emergency use (e.g. single patient)
• Treatment

7
IND Application

• Should include
• Chemistry, manufacturing, and control information
• Animal pharmacology and toxicology
• Genotoxicity
• Toxicology
• Histopathology
• Non-clinical justification for proposed
  dose/schedule
• Justification for duration of treatment
• Prior use in humans if applicable
• Clinical protocol and investigator information

8
IND Application

• Informed consent (More on this later in the
  course)
• Continued treatment beyond protocol?
• Charging for intervention?
• 30-day review clock
• Clinical hold
• Safety concerns
• Design will not allow protocol objectives to be
  met
• Teleconference between sponsor and division
  director about what is required to lift the hold
• IND exemptions for studies of lawfully marketed
  cancer drug
• Based on interpretation of risk
• Dose, schedule, patient population

9
Phase I Patient Population

• Conventional eligibility criteria
• Advanced solid tumors unresponsive to standard
  therapies or for which there is no known
  effective treatment
• Performance status (e.g. ECOG 0 or 1)
• Adequate organ functions (e.g. ANC, platelets,
  creatinine, AST/ALT, bilirubin)
• Specification about prior therapy allowed
• Specification about time interval between prior
  therapy and initiation of study treatment
• No serious uncontrolled medical disorder or
  active infection

10
Phase I Patient Population

• Agent-specific eligibility criteria - examples
• Restriction to certain patient populations must
  have strong scientific rationale
• Specific organ functions
• Example cardiac function restrictions (QTc lt
  450-470 ms, LVEF gt 45, etc) if preclinical data
  or prior clinical data of similar agents suggest
  cardiac risks
• Example no recent (6-12 months) history of
  acute MI/unstable angina, cerebrovascular events,
  venous thromboembolism no uncontrolled
  hypertension no significant proteinuria for
  antiangiogenic agents
• Prohibited medications if significant risk of
  interaction with study drug

11
A New Agent Merits Clinical Study If

• It is biologically plausible that the agent may
  have activity in cancer (target seems valid and
  agent affects it)
• There is reason to expect benefit for patients
  (preclinical or other evidence of efficacy)
• There is reasonable expectation of safety
  (toxicology)
• There are sufficient data on which to base a
  starting dose
• Hirschfeld S, 2004

12
The 3 Basic Tenets Of Phase I Studies

• Define a recommended dose
• SAFELY (minimum of serious toxicities)
• EFFICIENTLY (smallest possible of pts)
• RELIABLY (high statistical confidence)
• SAFETY TRUMPS EVERYTHING ELSE

13
Phase I Study Basic Design Principles

• Start with a safe dose
• Minimize of pts treated at sub-toxic/therapeutic
  doses
• Escalate dose rapidly in the absence of toxicity
• Escalate dose slowly in the presence of toxicity
• Expand patient cohort at recommended phase II
  dose
• Do you buy this argument?
• If so, why?

14
Phase I Trials Fundamental Questions

• If you believe the stuff on the previous slide,
  you must consider
• At what dose do you start?
• What are the endpoints?
• How many patients per cohort?
• How quickly do you escalate and by what method?

15
At what dose do you start ?
16
Preclinical Toxicology

• Typically a rodent (mouse or rat) and non-rodent
  (dog or non-human primate) species
• Animal organ-specific toxicities
• Are not great predictors of human toxicity
• Myelosuppression and gastrointestinal toxicity
  more predictable
• Hepatic and renal toxicities large false
  positive rate
• Toxicologic parameters
• LD10 lethal dose in 10 of animals
• TDL (toxic dose low) lowest dose that causes
  any toxicity in animals

17
Phase I Trials Starting Dose

• 1/10th of the LD10 in rodents,
• or
• 1/3rd of the minimal toxic dose in large animals
• Expressed as mg/m2
• Freireich, et.al., Quantitative comparisons of
  toxicity of anticancer agents in mouse, rat, dog,
  monkey and man. Cancer Chemother. Repts.
  50219-244, 1966
• These have historically been safe doses

18
Freireich EJ, et al, Cancer Chemother Rep
50219-244, 1966
19
What Are The Endpoints/Objectives?
20
Phase I Study Endpoints

• Classical goals (we have not been smart enough to
  understand primary and secondary.
• Primary
• Identify the maximally tolerated dose (MTD) and
  recommended Phase II dose (RP2D)
• Identify dose-limiting toxicities (DLTs)
• Secondary
• Pharmacokinetics
• We have not been smart enough to recognize drug
  metabolism a
• Pharmacodynamics (Clinical and now molecular)
• Target modulation
• Efficacy

21
Defining Toxicities NCI Common Toxicity
Criteria

• Grade 1 mild
• Grade 2 moderate
• Grade 3 severe
• Grade 4 life-threatening
• Grade 5 fatal
• Version 4 of the NCI Criteria are here.

22
Dose-Limiting Toxicities (DLT)

• Toxicities that, due to their severity or
  duration, are considered unacceptable, and limit
  further dose escalation
• Defined in advance of beginning trial
• Classically based on cycle 1 toxicity
• Examples
• ANC lt 500/ml for ? 5 or 7 days
• ANC lt 500/ml of any duration with fever
• PLT lt 10,000 or 25,000/ml
• Grade 3 or greater non-hematological toxicity
• Inability to re-treat patient within 2 wks of
  scheduled treatment

23
Definition of DLT is Dynamic

• Examples DLTs in 2008
• Diarrhea grade 3 in spite of adequate
  antidiarrheal therapy (loperamide)
• Nausea and vomiting grade 3 in spite of
  adequate anti-emetic prophylaxis and therapy
  (steroids, 5HT3 antagonists)
• Hypertension grade 3 in spite of adequate
  anti-hypertensive therapy
• Inability to take at least 90 of drug doses in a
  cycle (continuous oral meds)
• Grade 2 chronic unremitting toxicity

24
Maximally Tolerated Dose (MTD)

• Inconsistently defined as either
• Dose at which ? 33 of pts experience
  unacceptable toxicity (DLT in ? 2 of 3 or ? 2 of
  6)
• OR
• 1 dose level below that
• MTD level _at_ DLT (in Europe or Japan)
• MTD level below DLT (in US)
• 6-10 pts treated at the recommended Phase II dose
  (MTD or 1 dose level below)

25
Recap Trans-AtlanticDifferences in Terminology

• Important to note that
• Maximum tolerated dose (MTD)
• Usually means recommended dose in US
• Usually means dose level above recommended dose
  in Europe and some other jurisdictions

26
How many patients per cohort?
27
Patients per Cohort Guiding Principles

• Minimum needed to provide adequate toxicity
  information
• Classically 3 patients per cohort
• In some designs 1 patient per cohort until
  toxicity seen
• If correlative studies are a major aim, may
  increase up to 6 patients per cohort

28
Phase I Standard 3 3 Design
Eisenhauer et al.
29
How quickly do you escalate?
30
Phase I Trial Design Dose Escalation

• Escalation in decreasing steps (Hansen HH et
  al. Cancer Res. 1975)
• Attributed to a 13th century Italian
• Leonardo Fusano, aka
• Leonardo di Pisa
• Fibonacci
• How many pairs of rabbits can be produced from
  a single pair under specified conditions? (1, 1,
  2, 3, 5, 8, 13, 21, 34, 55, 89, 144..) (Liber
  abaci)

31
Phase I Trials Dose Escalation
The Modified Fibonacci Schedule
32
Cohort Dose Escalation
33
Problems and Pitfalls
34
Phase I Study Assumptions

• The higher the dose, the greater the likelihood
  of efficacy
• Dose-related acute toxicity is regarded as a
  surrogate for efficacy
• The highest safe dose is the dose most likely to
  be efficacious

35
Dose-response Efficacy and Toxicity
36
Modified Fibonacci Dose Escalation

• Problems
• Requires many patients
• Takes a long time (really?)
• May expose a substantial proportion of patients
  to low, ineffective doses

37
Classic Phase I Trials Design Limitations

• Wide confidence intervals
• Patients treated at ineffective doses in first
  cohorts
• High risk of severe toxicities at late cohorts

38
Classic Phase I Trials Design Limitations

• Chronic toxicities usually cannot be assessed
• Cumulative toxicities usually cannot be
  identified
• Uncommon toxicities will be missed

39
Phase I Studies and Infrequent Toxicities
Probability of overlooking a toxicity
POT(p) (1-p)n n sample size, p
true toxicity rate
40
Alternate Designs

• Starting dose
• Number of patients per dose level
• Method/rapidity of dose escalation

41
Selection of Starting Dose for Phase I Trials
Retrospective analysis of 21 trials using
modified Fibonacci dose escalation
Unsafe defined as reaching MTD in ? 3 dose
levels Eisenhauer et al JCO (18), 2000
42
Intra-patient dose escalation

• Treat patients at dose level 1
• Dose level 2 is well tolerated and patients at
  dose level 1 have no toxicities
• Patients at level 1 are escalated to level 2
• WHY NOT DO THIS ALWAYS ?
• Makes evaluation of chronic toxicities difficult
• The proverbial 1 responder at dose level 1

43
Phase I Trial DesignAccelerated Titrated Design
(Rule-based)

• First proposed by Simon et al (J Natl Cancer Inst
  1997)
• Several variations exist
• usual is doubling dose in single-patient cohorts
  till Grade 2 toxicity
• then revert to standard 33 design using a 40
  dose escalation
• intrapatient dose escalation allowed in some
  variations
• More rapid initial escalation

44
Accelerated Titrated Design
45
Accelerated Titrated Design Phase I Study of
Lonafarnib (SCH66336)
Adjei AA et al, Cancer Research, 2000
46
Phase I Trial DesignModified Continual
Assessment Method (MCRM Model-based)

• Bayesian method
• Pre-study probabilities based on preclinical or
  clinical data of similar agents
• At each dose level, add clinical data to better
  estimate the probability of MTD being reached
• Fixed dose levels, so that increments of
  escalation are still conservative

47
Modified Continual Assessment Method (MCRM
Model-based)

• Example Pre-set dose levels of 10, 20, 40, 80,
  160, 250, 400
• If after each dose level, the statistical model
  predicts a MTD higher than the next pre-set dose
  level, then dose escalation is allowed to the
  next pre-set dose level
• Advantages
• Allows more dose levels to be evaluated with a
  smaller number of patients
• More patients treated at or closer to
  therapeutic dose
• Disadvantages
• Does not save time, not easily implemented if
  without access to biostatistician support

48
Phase I Trial DesignDose Escalation with
Overdose Control (EWOC Model-based)

• Bayesian method
• After each cohort of patients, the posterior
  distribution is updated with DLT data to obtain
  ?d (probability of DLT at dose d). The
  recommended dose is the one with the highest
  posterior probability of DLT in the ideal
  dosing category
• The overdose control mandates that any dose that
  has gt 25 chance of being in the over-dosing or
  excessive over-dosing categories, or gt 5
  chance of being in the excess-overdosing
  category, is not considered for dosing

49
Estimated MTD Based on Bayesian Logistic Method
(Parameter Estimation With Overdose Control)
EXAMPLE of Probability of DLTs (Bayesian design)
50
Phase I Trial Design-Bayesian EWOC Model (Example
With 1 DLT in 4 Patients)

• Posterior summaries for probabilities of DLT (in
  )
• __________________________________________________
  ______________
• Dose 0-0.16 0.16-0.33 0.33-0.6 0.6-1
  Median Overdose
• mg Under- Targeted Over- Excessive DLT
  rate Control
• /day dosing Toxicity dosing toxicity
• __________________________________________________
  ______________
• 2.5 95.7 3.3 0.5 0.4
  0.0 ok
• 5 91.1 5.5 3.1 0.4
  0.2 ok
• 10 82.9 9.1 6.7 1.3
  1.9 no3
• 17.5 62.5 17.7 12.5 7.2
  8.8 no2,3
• 25 40.0 21.2 19.3 19.5
  22.3 no1-3
• 35 15.6 19.7 20.4 44.3
  53.7 no1-3
• 45 9.9 12.5 19.9 57.7
  73.0 no1-3
• 60 5.6 10.9 12.0 71.5
  86.5 no1-3
• 75 3.2 7.7 12.7 76.4
  92.8 no1-3
• 100 2.4 3.3 13.3 80.9
  96.6 no1-3
• __________________________________________________
  ________________
• 1 Escalation to this dose not possible, overdose
  control criterion
• (P(overdosing or excessive tox)gt25) violated

• 0/3

• No first cycle DLTs
• Bayesian model supports dose escalation in cohort
  2 (100 dose increment)

51
Phase I Trial Design-Bayesian EWOC Model Example
With ¼ DLTs

• Posterior summaries for probabilities of DLT (in
  )
• __________________________________________________
  ______________
• Dose 0-0.16 0.16-0.33 0.33-0.6 0.6-1
  Median Overdose
• mg Under- Targeted Over- Excessive DLT
  rate Control
• /day dosing Toxicity dosing toxicity
• __________________________________________________
  ______________
• 2.5 48.1 28.4 16.4 7.1
  16.5 no2
• 5 30.5 31.5 27.5 10.5
  26.4 no1,2
• 10 16.5 23.7 37.3 22.4
  38.3 no1-3
• 17.5 8.8 14.7 41.6 34.9
  50.3 no1-3
• 25 6.1 9.9 37.6 46.4
  58.3 no1-3
• 35 5.6 7.7 29.6 57.1
  65.6 no1-3
• 45 3.7 7.6 24.5 64.1
  69.6 no1-3
• 60 3.3 6.5 18.9 71.2
  75.1 no1-3
• 75 2.4 6.1 15.3 76.1
  78.1 no1-3
• 100 2.4 4.9 13.6 79.1
  82.3 no1-3
• __________________________________________________
  ________________
• 1 Escalation to this dose not possible, overdose
  control criterion

• 1/4

• First cycle DLT in first cohort
• Bayesian model does not support dose escalation
  in cohort 2

52
Challenges to Developing Novel Non-Classical
Antineoplastic Agents

• General requirement for long-term administration
  pharmacology and formulation critical
• Difficulty in determining the optimal dose in
  phase I MTD versus OBD
• Absent or low-level tumor regression as single
  agents problematic for making go no-go decisions
• Need for large randomized trials to definitively
  assess clinical benefit need to maximize chance
  of success in phase III

53
Phase I Trial Design Non-Cytotoxic Agents

• MTD may not be the goal of Phase I because
  specificity of effect may be lost at MTD
• Pharmacologic effect may not equal biologic
  effect
• Goal identify optimal biologically effective
  dose (OBED)
• Paradox requires early development and
  integration of (usually unvalidated) measures of
  biologic effect into Phase I

54
Do We Need Correlative Studies ?

• Conventional cytotoxic drugs have led to
  predictable effects on proliferating tissues
  (neutropenia, mucositis, diarrhea), thus enabling
  dose selection and confirming mechanisms of
  action
• Targeted biological agents may or may not have
  predictable effects on normal tissues and often
  enter the clinic needing evidence/proof of
  mechanisms in patients

55
Do we need correlative studies ?

• Therefore, biological correlative studies may be
  used to derive the best dose and schedule of an
  agent, and
• To determine whether the drug is inducing the
  intended biological effect in the patient, and
• To predict clinical benefit, although this
  generally requires testing in large randomized
  studies.

56
Alternative Endpoints

• Minimum blood levels/AUC or other PK measure
• Inhibition of target
• In normal tissue
• In tumor tissue
• Need enough preclinical evidence to suggest that
  the above are reasonable endpoints with
  sufficient clinical promise
• Must also pay attention to toxicity

57
Phase I Trial Design Non-Cytotoxic Agents -
Examples

• Pre-clinically define target drug exposure
  Matrix Metalloproteinase Inhibitors
• Define pharmacodynamic endpoint Bortezomib (70
  of 26S proteasome inhibition in PBMCs)
• Use functional imaging as endpoint Vatalanib
  (DCE-MRI)
• Use cumulative toxicities CI-1040 MEK
  Inhibitor, 800mg TID intolerable after 3 cycles
  of therapy
• Problem If drug works, youre a genius. If it
  doesnt, youre a goat

58
Vatalanib (PTK/ZK) VEGF Receptor Tyrosine
Kinase Inhibitor
Extracellular
Intracellular
Adapted from Dvorak H. J Clin Oncol.
200220(21)4368-4380.
59
PTK/ZK Induced Significant Reduction in Tumor
Blood Flow in Metastatic Colorectal Cancer by
DCE-MRI
Reduction in tumor blood flow through liver
metastases secondary to colorectal cancer at day
2 is significantly correlated with improved early
clinical outcome
Thomas AL, et al. Semin Oncol. 20033032-38. Morg
an B, et al. J Clin Oncol. 2003213955-3964.
60
Using DCE-MRI to Establish the Optimal
Therapeutic Dose for Vatalanib
Morgan B, et al. J Clin Oncol. 2003213955-3964.
61
The Conundrum of Optimal Biologic Dose Phase
III PTK787/FOLFOX in CRC
PTK787 1250 mg QD
62
Phase I Trials of Agent Combinations

• Initial dose-finding component often needed if
  you are planning a new combination for a phase II
  trial
• Patients for dose-finding phase
• Advanced solid tumors (all comers)
• Advantage fast accrual
• Disadvantage may not be representative of your
  patient population of interest
• Specific patient population (e.g. same as phase
  II cohort)
• Advantage population of interest, and early
  glimpse at antitumor activity in disease of
  interest
• Disadvantage slow down accrual especially if
  rare/uncommon tumors

63
Phase I Trials of Agent Combinations

• Dose escalation
• New drug A Standard combination BC
• Ideally keep standard combo doses and escalate
  the new drug (e.g. 1/3, 2/3, full dose)
• Need to provide rationale why add A to BC?
• Need to think about overlapping toxicity in your
  definition of DLT
• Do you need PK assessment to determine if A, B
  and C interact with each other?

64
How are Phase I Studies Designed Now ?

• 31 targeted agents representative of most common
  targets
• Reports (papers or abstracts) of completed single
  agent phase I trials in non-hematologic
  malignancies
• 57 phase I reports identified
• Parulekar and Eisenhauer JNCI July 2004

65
Agents/Targets
Target Agent Trials
EGFR/HER2 ZD1839OSI-774C225MAB225EMD7200EKB569RG83852trastuzumab 51111111
MMP BAY 12-9566MarimastatCOL-3BMS-275291 3211
Farnesyltrans-ferase BMS-214662R115777L778,123SCH66336 2323
Target Agent Trials
Angiogenesis EndostatinZD6474PTK787SU6668SU5416BevacizumabSU11248 3113311
BCL-2 G3139 2
PKC alpha ISIS 3521 2
raf kinase BAY 43-9006 ISIS 5132 33
DNA MTase MG98 2
MEK CI-1040 1
mTOR CCI-779 1

• Parulekar and Eisenhauer JNCI July 2004

66
Results Reason for Halting Dose Escalation
Reason No. Trials No. Agents
Toxicity 36 20
PK 7 5
Other Design (max. planned dose) Drug Supply Other phase I results Active dose 3421 3221
Not stated 4 4
TOTAL 57
Parulekar and Eisenhauer JNCI July 2004
67
Basis for Recommending Phase II Dose
Recommended phase II dose? Recommended phase II dose? No. Trials No. Agents
Recommended Recommended 50 27
Basis ToxicityPK (blood levels)Other trial toxicityClinical activityPBMC findingsTumor measuresConvenience 35921111 19721111
Not stated 5 5
Not recommended Not recommended 2 1
Parulekar and Eisenhauer JNCI July 2004
68
Secondary Information to Support Dose
Recommendation
No. Trials Primary Basis No. Trials Secondary Support From No. Trials Secondary Support From No. Trials Secondary Support From No. Trials Secondary Support From No. Trials Secondary Support From
No. Trials Primary Basis Toxicity PK Tumor measures Surrogate measures Other
Toxicity 35 - 12 2 7 8
PK 9 - - 1 1 3
Tumor measures 1 - - - - -
Surrogate measures 1 - - - - -
Other 4 2 1 1 -
TOTAL 50 1 12 4 9 11
Other includes clinical effects, results other
trials, convenience, etc.
69
Laboratory and Imaging Studies
Total No. Trials Primary Basis of Dose Supportive of dose No Help
Tumor Tissue 5 1 4 -
Surrogate Tissue PMBC Skin Buccal 1122 1-- 522 5--
Imaging 6 - 1 5
15
Parulekar and Eisenhauer JNCI July 2004
70
Summary ReviewPhase I Trials Targeted Agents

• Toxicity
• Most common reason to halt escalation and primary
  basis for dose recommendation (35/50 trials)
• PK (Blood levels)
• Second most common basis for dose recommendation
  (9/50 trials)
• Laboratory studies
• May provide information to support dose
  recommendation

Parulekar and Eisenhauer JNCI July 2004
71

• Phase I Study - Ethics

72
Phase I Study - Ethics

• Patient benefit or antitumor activity is not a
  primary goal of the study, but therapeutic
  intent is an important feature
• Desperate patients cannot make a truly informed
  decision
• Historically low probability of response in Phase
  I trials
• lt 5 response rate
• Majority of responses occur within 80-120 of
  the recommended phase II dose

73
Response Rates and Toxic Deaths in Phase I
Oncology Trials (Horstman et al, NEJM 352 2005)
Trial No. of Trials of Pts Assessed for Response Overall Response Rate No. of Patients Assessed for Toxic Events Deaths from Toxic Events no.
TotalFirst use of an agent in humans 117 3164 4.8 3498 9 (0.26)
Cytotoxic chemotherapyFirst use of an agent in humans 43 1298 5.0 1422 7 (0.49)
ImmunomodulatorFirst use of an agent in humans 16 404 7.4 431 1 (0.23)
Receptor or signal transductionFirst use of an agent in humans 27 742 3.8 853 1 (0.12)
AntiangiogenesisFirst use of an agent in humans 8 200 7.0 228 0
Gene transferFirst use of an agent in humans 0 0 0 0 0
VaccineFirst use of an agent in humans 23 520 3.1 564 0
74
Phase I Study - Ethics

• Investigators have an inherent conflict of
  interest
• Funding
• Academic promotion
• Publicity

75
Phase I Study Ethics Partial Solutions to the
Dilemma

• Youre the patients physician 1st and a scientist
  2nd
• Scientific goals should never take precedence
  over the patients best interest
• Only pts for whom no life-prolonging or curative
  therapy exists are eligible for Phase I trials
• Informed consent is obtained from every patient
• No new agent can hit the pharmacy shelves without
  going through Phase I clinical evaluation

76
Phase I Clinical Trials - Summary

• Most drugs tend to follow the MTD/DLT paradigm
• Alternative designs continue to be explored. Most
  times they are more complex.
• Correlative studies are increasingly important in
  the comprehensive evaluation of new agents
• Patient benefit/wellbeing trumps all the science
• Being a good phase I trialist is not as simple as
  you may think