Using an External Control to Evaluate the Effectiveness of Posaconazole for Refractory Invasive Fungal Infections

1
Using an External Control to Evaluate the
Effectiveness of Posaconazole for Refractory
Invasive Fungal Infections

• Kenneth J. Koury
• Jagadish P. Gogate
• Schering-Plough Research Institute

2
Refractory Invasive Fungal Infections(rIFI)

• Life-threatening infections
• Limited therapeutic options, especially in
  salvage setting
• Randomized study
• straightforward interpretation of results
• not practical to complete in a timely fashion
• serious ethical issue potential enrollment of
  subjects into a failing regimen
• No randomized trials completed for salvage
  therapy for any antifungal drug

3
Description of Study P00041

• Open-label, non-randomized, multinational,
  salvage therapy study
• Posaconazole oral suspension 400 mg bid
• Subjects with invasive fungal infections
  refractory to standard antifungal therapies or
  subjects who were intolerant of standard
  antifungal agents
• Started In January 1999 (compassionate use)
• Enrollment expanded based on investigator
  interest, completed by April 2001

4
Advice from Regulatory Authorities

• CPMP (Committee for Proprietary Medicinal
  Products)
• lack of randomized trial always makes it
  difficult to evaluate a new agent
• well-conducted external-control study may provide
  a useful comparative reference
• sponsor must demonstrate that any potential
  selection bias was minimized
• show that two groups are clinically similar

5
Advice from Regulatory Authorities

• FDA
• treatment for rIFI may fulfill an unmet medical
  need
• response rates from a control group required
• numerous discussions regarding design and
  analysis of external-control study (P02387)
• comparability of patient populations
• minimization of potential bias

6
External-Control Study (P02387)

• Provide reference (benchmark) for comparison to
  P00041
• Procedures and methods prospectively implemented
  to minimize potential sources of bias
• Exhaustive review of all cases of IFI at study
  centers over target time period
• 2073 screened for possible enrollment
• 279 fulfilled criteria

7
Control of Potential Bias

• External Data Review Committee (DRC)
• 15 independent clinical experts and 2
  radiologists
• Concurrent, blinded review of data from 330
  posaconazole-treated subjects (P00041) and 279
  external-control subjects (P02387)
• Determined patient eligibility
• met MSG/EORTC criteria for proven or probable IFI
• met criteria for refractory disease or
  intolerance of standard therapy or both
• Determined patient outcome global response
  status at end of treatment

8
Selection Bias

• Exhaustive review of all cases of IFI at study
  centers during comparable time period
• Similar inclusion and exclusion criteria
• Autopsy data not used for identifying control
  subjects
• Exclude control subjects who died within 72 hours
  of after the initiation of therapy

9
Information Bias

• Inherent differences in the amount of data
  available in a retrospective study vs.
  prospective study
• Develop uniform format for data presentation to
  maintain blinding of DRC
• Outcome data from a given subject are not
  available for review by DRC until eligibility of
  that subject has been determined

10
Temporal and Geographic Bias

• Use same time frame (almost concurrent rather
  than historical controls)
• Use mostly same centers
• Expect controls to have similar disease
  characteristics to posaconazole-treated subjects

11
Other Important Factors

• Primary pathogen (type of fungus)
• Infection site
• Underlying disease and associated treatments
• BMT, Solid Organ Transplant
• Hematologic malignancies
• HIV/Aids
• Use of immune response modifiers, growth factors

12
Baseline and Disease Characteristics(With
Potential to Influence Outcome)

• Demographic Variables Age, Gender, Race and
  Weight.
• Refractory Status
• Intolerance Status
• Period of Enrollment
• Primary Pathogen Aspergillus, Candida and Other
  Yeast, Fusarium, Cryptococcus, Zygomycetes, Other
  Filamentous Fungi, Other Endemic Fungi and
  Multiple Pathogens
• Infection Site pulmonary, extra pulmonary,
  disseminated with pulmonary involvement and
  disseminated without pulmonary involvement

13
Baseline and Disease Characteristics (With
Potential to Influence Outcome)

• Underlying Disease 9 variables each with two
  categories (present and absent Hematologic
  Malignancies, Non-Hematologic Malignancies,
  Nonmalignant Hematologic Disorders, Bone Marrow
  Transplant, Solid Organ Transplant, Renal
  Disease, Hepatic Disease, Immunocompromised-
  Acquired, Immunocompromised-Congenital
• Duration of prior effective antifungal therapy.
• Baseline Neutropenia
• Prior systemic corticosteroid use and Concurrent
  systemic corticosteroid use

14
Primary Efficacy Endpoint

• DRC-adjudicated Global Response Status at the
  end of treatment
• subject considered a responder if global
  response at EOT was classified as a complete or
  partial response
• subject considered a non-responder if
  classified as stable disease, failure, or unable
  to determine

15
Global Response Status

• Complete Response
• Resolution of all attributable clinical signs
  and symptoms, radiological and mycological
  abnormalities, if present at baseline.
• Partial Response
• Clinically meaningful improvement in attributable
  clinical signs and symptoms, radiological and
  mycological abnormalities, if present at
  baseline.
• Stable Disease
• No improvement in attributable clinical signs
  symptoms, radiological and mycological
  abnormalities, if present at baseline
• Failure
• Deterioration in attributable clinical signs and
  symptoms, radiological and/or mycological
  abnormalities necessitating alternative
  antifungal therapy or resulting in death
• Unable to Determine
• If for any reason the global response cannot be
  assessed (e.g., insufficient medical records to
  determine outcome)

16
Primary Analysis

• Modified ITT Population
• subjects with proven or probable aspergillosis
• refractory to standard therapy or intolerant
• Analysis of Global Response Rate
• Logistic regression model includes
• treatment
• key prognostic factors
• other potential prognostic factors (with
  treatment imbalances at baseline)
• Assessment of treatment differences based on
  odds ratio for treatment effect and its 95
  confidence interval

17
Prognostic Variables

• Key Variables
• Refractory status
• Intolerance status
• Site of infection
• Presence of baseline neutropenia
• Duration of prior antifungal therapy
• Region

• Other Variables
• All other baseline and disease characteristics
  considered as potential prognostic variables
• Included as covariates in primary statistical
  analysis if treatment imbalance at baseline is
  significant (plt0.10)

18
Supportive Analyses

• Two step procedure
• Preliminary analysis to identify prognostic
  indicators with potential to influence global
  response
• Based on a logistic regression procedure
• All variables (except treatment) included
• Variable selection option STEPWISE in PROC
  LOGISTIC of SAS (significance level 0.15 for
  entry and 0.10 for stay)
• Statistical Modeling
• Logistic regression on global response with
  treatment and other covariates selected from the
  above procedure.

19
Secondary Analyses

• Survival Analyses
• Time to Death (logrank test)
• Cox Proportional Hazards Model
• Treatment Group
• Site of Infection
• Refractory Status, Intolerance Status
• Baseline Neutropenia

20
Additional Analyses

• Forming cohorts based on a Prognosis Score (low,
  intermediate, high risk)
• Use of Propensity scores

21
Baseline and Disease Characteristics
22
Baseline and Disease Characteristics
23
Baseline Comparisons

• Both studies are comparable with respect to the
  distribution of demographic and almost all of the
  baseline disease characteristics
• Majority of the subjects are refractory to prior
  antifungal therapies
• Study P00041 has generally sicker patients (e.g.,
  hepatic disease and renal disease)

24
Global Response Rates (CompletePartial)
Region P00041 (N107) P02387 (N86)
US 38/94 (40.4) 16/68 (23.5)
Ex-US 7/13 (53.8) 6/18 (33.3)
All 45/107 (42.1) 22/86 (25.6)

• Unadjusted OR2.11
• Confidence Interval for OR(1.15, 3.92)
• p-value 0.018

25
Primary Efficacy Analysis

• Logistic Regression on Global Response
• Key Prognostic Factors included
• Refractory Status, Intolerance Status, Site of
  Infection, Neutropenia, Duration of Prior
  Antifungal Therapy, Age and Region of
  Enrollment.
• Potential Prognostic Factors with Baseline
  Imbalance
• Enrollment Time, Race, Non-malignant Hematologic
  Disorders, Renal Disease and Hepatic Disease

26
Results of Primary Efficacy Analysis

• Only two significant variables in the model
• Treatment and Baseline Neutropenia
• No interaction between Treatment and Neutropenia
  based on a logistic regression model with
  Treatment, Neutropenia and interaction between
  them
• Odds Ratio for Treatment 4.06
• P-value for the Treatment Effect 0.006
• 95 Confidence Interval (1.50, 11.04)

27
Impact on Odds Ratio for Treatment Effect
Sequential Addition of Prognostic Variables
Variable Added in the Model Sequentially Odds Ratio P-value
No covariate 2.11 0.0178
Neutropenia 1.94 0.040
Region 2.10 0.025
Hepatic Disease 2.55 0.010
Refractory Status 2.64 0.007
Intolerance Status 2.80 0.005
Infection Site 2.70 0.008
Non-malignant Hematologic Disorders 2.89 0.006
Age 2.88 0.007
Renal Disease 3.05 0.008
Prior AF duration 3.34 0.006
Enrollment Time 4.05 0.006
Race Group 4.06 0.006
28
Supportive Analysis

• Variables associated with the Global Response
  selected using the Stepwise Logistic Regression
• Baseline Neutropenia, Prior Corticosteroid Use,
  Solid Organ Transplant
• Logistic regression on Global Response with
  Treatment and the above variables in the Model
• Odds Ratio for Treatment 2.05
• P-value for the Treatment Effect 0.034
  Confidence Interval (1.06, 3.97)

29
Sensitivity Analyses
No. Variations of the Prognostic Factors OR CI/P-value
1 Excluded Non-malignant Hematologic Disorders, Renal and Hepatic Disease 2.68 (1.15, 6.27)/ 0.023
2 Binary infection site, Continuous Age and Prior AF duration 3.67 (1.41, 9.54)/0.008
3 Binary Enrollment and Infection site, Continuous Age and Ordinal Prior AF duration 3.73 (1.51, 9.21)/0.004
4 Binary Enrollment and Infection Site, Continuous Age and Prior AF duration 3.45 (1.42, 8.41)/0.006
5 Ordinal Enrollment, Binary Infection site, Continuous Age and Ordinal Prior AF duration 3.48 (1.39, 8.72)/0.008
6 Ordinal Enrollment, Binary Infection site, Binary Prior AF duration (lt30, gt30) and Continuous Age 3.69 (1.46, 9.35)/0.006
30
Global Response Rates by Baseline Neutropenia
Study protocol Study protocol Study protocol Study protocol
P00041 P00041 P02387 P02387
N N
Neutropenia Status 1/3 33.3
Missing 1/3 33.3
Neutropenia 5/21 23.8 2/26 7.7
No Neutropenia 35/78 44.9 19/58 32.8
Unable to Determine 4/5 80.0 1/2 50.0
Total 45/107 42.1 22/86 25.6
31
Global Response Rates by Corticosteroid Use
Study protocol Study protocol Study protocol Study protocol
P00041 P00041 P02387 P02387
N N
Prior Corticosteroid 15/29 51.7 12/26 46.2
No prior use 15/29 51.7 12/26 46.2
Prior use 30/78 38.5 10/60 16.7
Concomitant Corticosteroid 13/23 56.5 7/27 25.9
No Concomitant use 13/23 56.5 7/27 25.9
Concomitant use 32/84 38.1 15/59 25.4
Total 45/107 42.1 22/86 25.6
32
Global Response Rates by Infection Site
Study protocol Study protocol Study protocol Study protocol
P00041 P00041 P02387 P02387
N N
Infection Site
Pulmonary 31/79 39.2 17/67 25.4
Extra-Pulmonary 10/19 52.6 5/11 45.5
Disseminated with Pulmonary Involvement 1/3 33.3 0/3 0
Disseminated w/o Pulmonary Involvement 3/6 50 0/5 0
Total 45/107 42.1 22/86 25.6
33
Global Response Rates by Underlying Disease
Conditions
Study protocol Study protocol Study protocol Study protocol
P00041 (n107) P00041 (n107) P02387 (n86) P02387 (n86)
N N
Bone Marrow Transplant 21/55 38.2 7/38 18.4
Solid Organ Transplant 7/12 58.3 5/7 71.4
Hematologic Malignancies 29/79 36.7 16/70 22.9
Non-malignant Hematologic Disorders 36/91 39.6 11/52 21.2
Non-hematologic malignancies 6/13 46.2 1/5 20.0
Immunosuppressive-acquired 13/28 46.4 6/19 31.6
Immunosuppressive-congenital 1/2 50.0 0/3 0.0
Renal Disease 27/68 39.7 5/21 23.8
Hepatic Disease 16/43 37.2 0/6 0.0
No Identifiable Risk Factors 1/1 100.0
34
(No Transcript)
35
Variables used for Prognosis Score

• BMT, Solid Organ Transplant, Acquired
  Immunodeficiency, Congenital Immunodeficiency,
  Hematologic Malignancy, Non-hematologic
  Malignancy, Non-malignant Hematologic Disorder,
  Renal disease risk factor OR Creatinine gt Grade
  1, Liver disease risk factor OR Total Bilirubin
  gt Grade 1 OR SGOT gt Grade 1 OR SGPT gt Grade 1,
  Prior Systemic Steroid use, Prior
  Immunosuppressive therapy, Prior Myelosuppressive
  therapy, Baseline Neutropenia, Baseline
  Mechanical Ventilation, and Baseline GVHD
• Prognosis Score one point added to subjects
  score for each condition that is present

36
Global Response Rate by Prognosis Score
Prognosis Score Group P00041 Posaconazole P02387 Control
Low Risk (0-1) 2/2 (100) -
Intermediate Risk (2-4) 10/21 (58) 8/25 (32)
High Risk (gt5) 33/48 (39) 14/61 (23)
All 45/107 (42) 22/86 (26)
P-value 0.0202 based on CMH stratified analysis
for general association.
37
Propensity Scores

• Propensity Score is defined as the conditional
  probability of being treated given the
  covariates.
• Estimated propensity score can be used to reduce
  the bias through matching, stratification,
  regression adjustment or some combination of all
  three.
• Advantage of using propensity scores in
  estimating treatment effect is its simplicity in
  interpreting the results.
• Quantiles of the distribution of the propensity
  scores can be used as covariates in logistic
  regression analysis or as strata in assessing the
  treatment effect.

38
Global Response Rate by Propensity Score Group
Propensity Score Group P0041 P02387
lt0.177 0/1 (0) 18/57(32)
0.177- lt0.608 10/16 (63) 3/18 (17)
0.608- lt0.936 12/33 (36) 1/9 (11)
gt0.936 23/57 (40) 0/2 (0)
All 45/107 (42) 22/86 (26)
P-value using the logistic regression is
0.0223 P value based on the CMH analysis is
0.0207
39
Summary and Conclusions

• Posaconazole is substantially more effective than
  control for treatment of rIFI due to aspergillus
• The adjusted odds ratio obtained from the primary
  analysis is almost twice as large as the
  unadjusted odds ratio
• The treatment effect based on primary analysis is
  robust in the sense that it is not sensitive to
  the classification of baseline and disease
  characteristics used to specify the way in which
  the prognostic factors were incorporated in the
  analyses.

40
Summary and Conclusions

• The treatment effect is statistically significant
  even when the influential factors non-malignant
  hematologic disorders, hepatic disease and renal
  disease are excluded from the primary model (odds
  ratio2.70, p-value0.0227).
• Alternate methods of controlling for prognostic
  factors confirm effectiveness of posaconazole.
• Survival (based on the Kaplan-Meier estimates) is
  substantially longer in subjects treated with
  Posaconazole than control subjects. Kaplan-Meier
  estimates of the survival rates at the end of one
  year are 38 and 22 in P0041 and P02387
  respectively.