multiple interventions for multiple targets
Skip this Video
Download Presentation
Multiple Interventions for Multiple Targets

Loading in 2 Seconds...

play fullscreen
1 / 37

Multiple Interventions for Multiple Targets - PowerPoint PPT Presentation

  • Uploaded on

Multiple Interventions for Multiple Targets. Emilia Bagiella, PhD Columbia University. Traumatic Brain Injury (TBI). An injury to the head arising from blunt or penetrating trauma or from acceleration-deceleration forces Damage can be focal or diffuse Closed head injury

I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
Download Presentation

PowerPoint Slideshow about ' Multiple Interventions for Multiple Targets ' - cascata-firbis

An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.

- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
multiple interventions for multiple targets
Multiple Interventions for Multiple Targets

Emilia Bagiella, PhD

Columbia University

traumatic brain injury tbi
Traumatic Brain Injury (TBI)
  • An injury to the head arising from blunt or penetrating trauma or from acceleration-deceleration forces
  • Damage can be focal or diffuse
  • Closed head injury
  • Penetrating head injury
traumatic brain injury tbi1
Traumatic Brain Injury (TBI)
  • An estimated 1.5 million head injuries occur every year in the United States
  • TBI is the leading cause of death and disability in children and adults ages 1 to 44
  • Approximately 52,000 deaths every year
traumatic brain injury tbi2
Traumatic Brain Injury (TBI)
  • More than 5.3 million Americans, 2% of the U.S. population, currently live with disabilities resulting from TBI
  • Direct medical costs and indirect costs (e.g. lost productivity) due to TBI is estimated at $56.3 billion annually
how does tbi happen
How does TBI happen?
  • Falls (28%)
  • Motor vehicle-traffic crashes (20%)
  • Struck by/against events (19%); and
  • Assaults (11%)
short term effect of tbi
Short term effect of TBI

Initial acute insult (diffuse axonal injury) followed by a cascade of events involving multiple secondary injuries

  • Significant tissue damage
  • Ischemia/hypoxia
  • Brain swelling
  • Brain hemorrhage
long term consequences
Long term consequences
  • Wide range of functional changes affecting thinking, sensation, language, and/or emotions.
  • Epilepsy and increase the risk for Alzheimer’s and Parkinson’s disease, and other brain disorders
  • Virtually no injury is without consequence
  • Unpredictable
  • No known risk factors
    • Low socio-economic class
    • Drug and alcohol abuse
    • War
  • Affects mostly otherwise healthy individuals
clinical trials in tbi
Clinical Trials in TBI

So far…

  • Difficult to design and perform
  • Hampered by many problems
  • Most trials have failed to show improvement in (any) outcomes
clinical trials in tbi1
Clinical Trials in TBI
  • Have learned more from clinical practice than from clinical trials
  • Guidelines based on common sense
  • Back to square 1
possible interventions
Possible interventions


Physiological - Surgical

  • Neuroprotective agents
  • Steroids
  • Free radicals scavangers
  • Insuline like growth factor (IGF)
  • Progesterone
  • Biomarkers(?)
  • Hypothermia
  • Decompressive craniectomy
  • O2 monitoring
  • ICP/CBF management
  • Pre-clinical and clinical data are disconnected
  • Need adequate pre-clinical TBI models
  • Virtually no early phase trials in TBI
  • Dosing
  • Duration of treatment
  • Time of treatment initiation
design problems
Design Problems
  • Weaknesses in study design
  • Insufficient power/sample size
  • Inadequate outcome measures or lack of sensitivity of the outcomes measure
  • Too small effect sizes
  • Too variable population
  • No single measure can capture the multidimensional nature of TBI outcome
  • Combination of drugs are needed for the treatment of TBI
glasgow outcome scale gos
Glasgow Outcome Scale (GOS)
  • 5-category scale
  • Gold standard for trial in TBI
  • The only measure of functional recovery accepted by the FDA
  • Very broad and not specific
  • Non differential misclassification
  • Considerable loss of statistical power and the attenuation of the true treatment effect
  • To identify measures that together would reflect the “global” status of TBI patients
  • Functional, physical, emotional, cognitive, and social spheres
global statistical analysis approach
Global Statistical Analysis Approach
  • Offers a method to utilize several outcome measures without need to pre-specify one as primary
  • Avoids loss of power due to multiple comparisons
  • Easily interpreted and of direct clinical interest
binary outcome
Binary Outcome

Let Yi = 1(0) denote the success(failure) for the

i th measure (i =1, … ,k)

Let T = 1(0) for experimental(standard) Tx.

binary outcome1
Binary Outcome
  • Interest centers on treatments where (even approximately)

b1 = b2 = … = bk = b

  • The results of the analysis are estimate of the common odds ratio, exp(b ), together with a standard error, z-score, p-value and confidence intervals

Statistical power is greatest when the odds ratios are all equal, but that is nor required to test H0: all bi = 0 (no treatment effect with any measure)

  • Analogous to the Mantel-Haenszel procedure with correlated tables

Power is reduced if some bi = 0 or are of opposite signs.

But that is an ambiguous situation where careful judgment is required, in which case reduced power may not be inappropriate


The methods can be easily extended to clinical trials with stratified randomization.

  • Estimation of b is absolutely conventional using weighted least squares.
  • Further adjustment for additional covariates can be handled by GEE methods.
  • Targets multiple areas of recovery
  • Reduces the trial size
  • Has greater statistical power than any single outcome measure
  • Need to know the correlation among outcomes
  • (May) still need large sample size
  • FDA (?)
example cobrit trial
Example: COBRIT trial
  • Fix OR=1.4 (8% improvement on the GOS)
  • Choose 9 measures of functional and cognitive status
  • Fix the type I error at 0.05 and the power at 85%
  • Needed 1124 patients to detect the effect size of interest.
  • Would need 1836 participants to run a trial with the same power, type I error and effect size with GOS alone.

Single treatments do not work

  • “Cocktails” or combination treatment may work
  • What to combine and at what dose?
sequential selection procedure
Sequential Selection Procedure
  • Dichotomize the outcome as success or failure
  • Use a coin tossing model for the outcome data
  • Determine the number of treatment combinations to be tested (k)
  • Determine the number of treatment combinations to be chosen (b)
levin robbins leu lrl procedure
Levin-Robbins-Leu (LRL) Procedure
  • Choose a positive integer r to guarantee high probability of correct selection
  • Compare all k treatment combinations at the same time
  • Eliminate “inferior” treatments as soon as they follow r successes behind the treatment with the currently b largest tally
levin robbins leu lrl procedure1
Levin-Robbins-Leu (LRL) Procedure
  • Recruit “superior” treatments as soon as they pull r successes ahead of the treatments with the currently held (b+1) largest tally
  • The procedure stops when b treatments are recruited and k-b treatments are withdrawn
  • Very efficient in pre-clinical studies
  • It allows early stopping
  • Reduces total number of experiments
  • Maximize probability of correct selection of the best treatment
  • Difficult to apply to clinical setting
  • Must start from a small number of treatments
  • No hypothesis testing
  • 13 agents
  • 3 doses
  • Start from 558 two-way treatment/dose combinations
  • Assume 10 combinations are truly beneficial with 63% probability of success
  • LRL procedure requires 2502 experiments compared to 11160 for fixed sample size
open problem
Open Problem