Transcript Forrest-ISAP-post-IC..

Patient Studies Modeling Surrogates and
Their Linkages: MIC, Clinical Scoring and
PK/PD Indices of Effect
Alan Forrest, Pharm.D.
State University of New York at Buffalo
School of Pharmacy & Pharmaceutical Sciences,
School of Medicine & Biomedical Science
Institute for Clinical Pharmacodynamics
Ordway Research Institute; Albany, NY
CPL Associates, LLC; Buffalo, NY
Overview
• Difficulties in studying PK/PD of anti-infectives in humans
• PD endpoints, in ID, which might be modelled
– Advantages/disadvantages
– Current status in drug development
– Examples
• Proposal to co-model pathogen replication and death, the timecourse of severity of disease manifestation & the effects of drug
therapy, on these processes (PD)
• Enabling study designs
• Summary/Conclusions
– Implications, applications
– What is needed?
Informative PK/PD ID Studies, in Humans, are Difficult
• PD models (drug exposure & pathogen susceptibility versus
outcome) usually need large numbers of evaluable patients (“100s”)
– Adequate PK data usually needed in all (dose-response analyses are
too insensitive, would require even larger numbers)
– Positive cultures with susceptibilities needed (must consider drug
concentrations in relation to the MIC, ED50, etc; concentration-response
analyses are also usually too insensitive)
– Relevant covariates & outcomes determined & documented
• A large range of drug exposures is needed (dose ranging plus PK
variability is better than relying on PK variability, alone)
• “Placebo effect” (spontaneous resolution) for certain infections such
as acute bacterial exacerbation of chronic bronchitis (ABECB)
• To model exposure & susceptibility vs probabilities of success &
failure, “adequate” numbers of both outcomes are required (&
obviously, we usually cannot design for more clinical failures)
Candidate PD Endpoints
• Probabilities of good & bad events (categorical/ordinal responses)
– Examples: mortality, microbiologic or clinical success or failure, yes/no
adverse event (AE), emergence of resistance
– Most available type of endpoint, in drug development trials, but least
sensitive/informative (thus requires the largest sample size, especially
with mortality as an endpoint!; “All-cause mortality is the gold standard”)
– Accepted definitions generally available, usually largely based on
observations made AFTER the course of treatment (eg, TOC visit)
– Probability of clinical cure is a “noisier” endpoint than microbiologic cure,
but usually has required similar activity breakpoints (eg, AUC/MIC)
– Clinical resolution, after two weeks, is “counted” as being equivalent to
needing only two days to achieve substantial resolution
Candidate PD Endpoints (continued)
• Time to good or bad events (interval, AKA ‘survival’ analyses)
– Examples: eradication, clinical resolution, resistance, AE
– Under-utilized, more difficult to obtain; (much) more informative;
probably less sensitive to placebo effect
– Time to eradication & resolution is important to patients, their family &
clinicians; likely associated with probability of emergent resistance &
other AE; may be associated with needed duration of treatment;
probably a factor in total costs of the infectious episode
– Not yet widely accepted; needs standard definitions & validation
• Change in a continuous numeric or ordinal variable (regression)
– Examples: pathogen titer (quantitative or semi-), such as viral load;
composite disease scores; a lab value sensitive to an AE (eg, a change
in creatinine clearance or platelet count)
– Least commonly available, but most sensitive/informative type of data
– Smallest required sample sizes
Ciprofloxacin vs Lower Respiratory Tract Infections
Microbiologic Cure
100
100
% probability of micro cure
% probability of clinical cure
Clinical Cure
80
60
40
20
0
1
2
3
4
80
60
40
20
0
1
Log10 AUC/MIC
Forrest A, et al, Antimicrob Agents Chemother.1993. 37:1073–1081
2
3
Log10 AUC/MIC
4
% of patients remaining culture-positive
Ciprofloxacin: Time to Negative Cultures vs AUC/MIC
100
75
AUC/MIC < 125, n=21
50
125 < AUC/MIC < 250, n=15
25
AUC/MIC > 250,
n=28
0
0
2
4
6
8
10
Days of treatment
Forrest A, et al, Antimicrob Agents Chemother.1993. 37:1073–1081.
12
14
Time to Emergence of Resistance
Percent Susceptible
100
AUIC >100 (n=97)
80
60
GNR treated with
-lactam (n=14)
40
20
0
0
AUIC <100 (n=17)
5
10
15
20
Days From Initiation of Therapy
Thomas, et al. Antimicrob Agents Chemother.1998.42(3):521-527
Population Pharmacokinetics of Linezolid in Seriously Ill
Adult Patients from a Compassionate Use Protocol
Alison K. Meagher, Alan Forrest, Craig R. Rayner,
Mary C. Birmingham, Jerome J. Schentag
AAC 47(2):548-53 2003.
Population Pharmacodynamics of Linezolid in Seriously Ill
Adult Patients from a Compassionate Use Protocol
Craig R. Rayner, Alan Forrest, Alison K. Meagher,
Mary C. Birmingham, Jerome J. Schentag
Clin PK 42(15):1411-23 2003.
Pharmacostatistical Modelling of Hematologic Effects
of Linezolid in Seriously Ill Patients
Alan Forrest, Craig R. Rayner, Alison K. Meagher,
Mary C. Birmingham, Jerome J. Schentag
40th ICAAC, Toronto, Ontario, Canada, Sept 2000.
Probability of Eradication
1.0
1.0
0.8
0.8
0.6
0.6
0.4
0.4
Bacteremia
N=68
0.2
0.2
0.0
Probability of Eradication
10
100
1000
0.0
10
1.0
1.0
0.8
0.8
0.6
0.6
0.4
Lower respiratory
tract infection
N=11
0.2
0.0
Bone infection
N=15
10
100
AUIC (SIT-1  hr)
1000
0.4
100
1000
Skin and skin
structure infection
N=36
0.2
0.0
10
100
AUIC (SIT-1  hr)
1000
Linezolid AUC/MIC and Time to Negative
Cultures, in Patients with Bacteremia
1.0
>105
N=39
51-105
N=23
0.8
<51
N=6
0.6
0.4
AUC/MIC
T50%Erad
T75%Erad
0.2
<51
4.0
-
51-105
3.0-4.0
8.0-9.0
>105
0.5-1.0
4.0-5.0
P=0.0210 *
0.0
0
10
20
Days of Therapy
30
* Log-rank test
% Reduction in Platelets versus AUC and Duration
90
80
70
60
% in
Platelets
50
40
30
20
10
120
100
1000
80
Duration of
Therapy
(Days)
800
60
600
40
400
20
200
0
AUC
(mg/Lx24hr)
Decision Analysis
Decision Analysis
% Probability or % Decrease
Predicting Population Outcomes:
Bacteremia, 600mg IV q.12hr x 3wk
120
% Probability of Failure
80
MIC (mg/L): 1.0
2.0
4.0
% Decrease in:
Platelets
40
Hemoglobin
0
10
100
24hr Average AUC (mg/ Lxhr )
1000
4000
2000
0
0
7
14
21
0
7
14
21
0
7
14
21
0
7
14
21
1500
1000
500
0
80
40
0
40
20
0
Time (Days)
System of Models for Pathogen Growth & Death, Disease
State Progression & the Effects of Drug Therapy
• 1st biomath model: in the infected host, pathogen titers initially
increase rapidly (output is culture +/-, semi- or quantitative titers)
• 2nd model: pathogen produces toxins, etc, stimulates inflammatory
processes, etc, which result in appearance & worsening of signs &
symptoms of infection (output is mostly ordinal, probably a
composite disease severity score)
• 3rd model (PK/PD): effective treatment reduces titer (inhibits
replication &/or enhances rate of death), drives titer to BLQ
• Disease signs & symptoms (usually) begin to lessen later than titers
start to drop & resolve more slowly (time to resolution correlated with
but NOT EQUAL to time to eradication)
METHODS: Protocol Highlights
• Non-comparative, open label, single-center, pilot study, of
gatifloxacin (for 5 days) to treat acute maxilary sinusitis
• Symptoms for greater than 7 days (purulent nasal discharge from
maxillary sinus orifice, nose, or back of the throat)
• Radiological documentation with opacification, an air fluid level, or
mucosal thickening of  5 mm
• Indwelling sinus catheter allowed sampling of sinus aspirate (SA),
for drug concentration & semi-quantitative culture results
• C&S of SA daily x5; on day 4 serial samples of plasma & SA were
obtained (6 samples over 6 hours) for gatifloxacin assay
Ambrose PG, et al. Use of Pharmacodynamic Endpoints in the Evaluation of Gatifloxacin
for the Treatment of Acute Maxillary Sinusitis. Clin Infect Dis. 2004. 38:1513-20
Gatifloxacin Exposure: Plasma & Sinus Aspirate
Time to Bacterial Eradication
Time to Resolution Correlated with Time to Eradication
• 8 other signs & symptoms were also followed
• 91% (29/32) of the total number of signs & symptoms
resolved by the end of therapy
• Time to sinus sterilization correlated with median time to
resolution of signs and symptoms (rs = 0.85)
Change of pneumonia score
10
5
Addition
of Synercid
VAN AUC/MIC ~180
0
-8
-6
-4
-2
0
2
4
6
8
10
12
14
16
18
20
22
24
26
-5
Culture (-)
VAN AUC/MIC ~450
-10
Culture (-)
-15
Treatment day
VAN AUC/MIC ~80
+ SYN AUC/MIC ??
28
Proposed Enabling Study Designs
• Data which should ‘usually’ be obtained
– Plasma PK and baseline culture and sensitivities (C&S)
– Yes/No clinical and microbiologic success/failure
– Serial disease severity scores
• Data to be obtained when possible
– Serial C&S from the ‘site of infection’ (eg, ELF, CSF, blood)
• Quantitative or semi-quantitative titers, if possible
– PK data from the ‘site of infection’
Summary and Conclusions
• We believe that time to eradication and time to (substantive) clinical
resolution, for example, are strongly associated and, if they could be
acceptably defined and validated, would be superior and relevant
endpoints in ID drug development research
– more sensitive, specific and informative
• Composite ID disease severity scores
– Components would differ by site/type of infection
– Perspective: patients’ (how do you feel?); clinicians (signs & symptoms
plus results of tests & procedures used to monitor patients); other?
• What is needed?
– Develop new definitions/standards, evaluate existing ones
– Develop, with regulatory scientists, an approach to evaluate & validate
these “new” endpoints (eg, against what/which Gold Standard(s)?)
• Retrospective application to existing data? Prospective only?