Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life...
Transcript of Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life...
![Page 1: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/1.jpg)
Sample Copy. Not For Distribution.
![Page 2: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/2.jpg)
i
Recent Statistical Techniques
in Clinical Research
Sample Copy. Not For Distribution.
![Page 3: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/3.jpg)
ii
Publishing-in-support-of,
EDUCREATION PUBLISHING
RZ 94, Sector - 6, Dwarka, New Delhi - 110075
Shubham Vihar, Mangla, Bilaspur, Chhattisgarh - 495001
Website: www.educreation.in
________________________________________________________________
© Copyright, Authors
All rights reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any form by any means, electronic, mechanical, magnetic, optical, chemical, manual, photocopying, recording or otherwise, without the prior written consent of its writer.
ISBN: 978-1-61813-740-1
Price: ` 522.00
The opinions/ contents expressed in this book are solely of the authors and do not represent the opinions/ standings/ thoughts of Educreation or the Editors . The book is released by using the services of self-publishing house.
Printed in India
Sample Copy. Not For Distribution.
![Page 4: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/4.jpg)
iii
Recent Statistical
Techniques In Clinical
Research
Dr Basavarajaiah D M
EDUCREATION PUBLISHING (Since 2011)
www.educreation.in
Sample Copy. Not For Distribution.
![Page 5: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/5.jpg)
iv
Sample Copy. Not For Distribution.
![Page 6: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/6.jpg)
v
TO
PARENTS AND WIFE
WITH LOVE
Sample Copy. Not For Distribution.
![Page 7: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/7.jpg)
vi
PREFACE
Statistical methods play an essential role in all stages of a
quantitative health care and clinical research from design analysis
and interpretation of real life data sets. The clinical trial is “ the
most definitive tool for evaluation of the applicability of clinical
research” It represents “a key research activity with the potential
to improve the quality of health care and management through
careful comparison of alternative treatments” It has been called on
many occasions “the gold standard” against which all other
clinical research is measured. Although many clinical trial are of
high quality, a careful reader of the medical literature will notice
that a large number have deficiencies in design, conduct, analysis,
presentation, and /or interpretation of results. Improvements have
occurred over the past few decades, but too many trials are still
conducted without adequate attention to its fundamental principles.
Certainly, numerous studies could have been upgraded if the
authors had a better understanding of the fundamentals. This book
covers the essential principles and methods required for clinical
research. The underlying concepts of statistical analysis including
basic and some more advanced analysis techniques are also
covered. This book is an attempt to present the recent statistical
techniques and tools with suitable examples from real life data
sets, which the clinical researchers and academicians need. A
considerable part of the book is devoted therefore to the design of
experiments in phase I, II and III clinical trials. The reader will be
able to follow the sequence of ideas in the latter part of the book. It
is probably desirable to list these topics explicitly because of the
tendency of uncritical readers to regard either the whole of a book
as new, or none of it. The specific topics, most of which have been
dealt with formulation of the model and its applications in the
context of clinical trials or medical research. The book have been
framed with seven chapters, chapter -I describes the clinical
research design and statistical methods, it covers brief concept of
clinical study, research design, different phases of clinical trials,
drug development process, statistical issues and methods etc.,
Sample Copy. Not For Distribution.
![Page 8: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/8.jpg)
vii
Chapter –II describes the different experimental design approach
to clinical research; basic elements of experimental design, CRD,
RCBD and LSD with mathematical models. Chapter-III covers
the recent simulation models viz., sensor –noise estimation by snap
shot techniques, which covers the pharmacokinetic model,
formulation of epidemiological probability model, Sensor fusion
with normally distributed simulation model by real life animal
science data sets and also we discuss the practical utility of the
sensor fusion model in medical science. Chapter –IV covers the
statistical methods for clinical research. Chapter -V briefly covers
the survival analysis, hazard rate, censored data, parametric –
weibull distribution, KPM, Gompertzmakeham distribution
models. Chapter –VI deals with the different statistical models
approach to life threatened diseases -an Indian perspective viz.,
neural network modeling, application of neural network modeling
in medical research etc. Chapter-VII covers the image processing
modeling on radiographic features; analog, digital image
processing, different stages of digital image processing, signal,
artificial intelligence (AIs) etc. Chapter-VIII describes an
introduction to the database management in clinical research,
model structures, different types of model, Data base management
system flow,security,sequences,statistical modeling on DBMS,data
base environment etc.Finally, the Chapter IX deals with the
ethical issues and perspective of clinical research.
There are many fascinating problems that remain and we
would have liked to have solved these. However, the solution
might suggest themselves in the forthcoming days. Some problems
are discussed briefly for all chapters. My basis debt with regard to
whole book is to all scientific and academic community. I am
indebted to our University officers, Karnataka Veterinary Animal
and Fisheries Sciences University, Bidar, India for moral support
and permission to entitled of this publication.
*****
Sample Copy. Not For Distribution.
![Page 9: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/9.jpg)
viii
ACKNOWLEDGEMENTS
„Recent statistical techniques in clinical research’ text book is the
product of shared network. We would like to give special thanks to
our university KVAFSU (B) officers who helped to build it. We
would like to express our gratitude to Dr.B.Parabhakar former
Professor and head, Department of Medicine, Bengaluru Medical
College and Research Institute, Fort Road, Bengaluru. Thanks to
our parents, wife and kids for their love and support.
Finally, we would like to thank our readers. We hope you
enjoy reading this book and find it useful. We wish all our readers
all the very best in their future endeavour .We would love to hear
your comments and suggestions .You can send your mails to
*****
Sample Copy. Not For Distribution.
![Page 10: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/10.jpg)
ix
CONTENTS AT A GLANCE
S. No. Content Page
1
Clinical Research Design And Statistical
Methods
1
1.1 Introduction
1.2 Statistical historical perspectives of clinical
trial
1.3 Brief concept of study design
1.3.1 Cross sectional study
1.3.2 Longitudinal study
1.3.3 Prospective study
1.3.4 Retrospective study: Case control
1.4 Different phases of clinical trial
1.5 Drug development process
1.6 Study population
1.7 Statistical issues and methods
1.8 Application of multinomial distribution in
clinical trial
1.8.1 Gehans‟s two stage design
1.8.2 Two stage Simon‟s experimental design
1.9 Random variable
1.9.1 Conditional expectation
1.9.2 Conditional variance
1.10 Randomized clinical trial (RCT)
1.10.0 Simple randomization
1.10.1 Block randomization
1.10.2 Minimization method stratification
1.10.3 Stratified randomization method
2 Different Experimental Design Approach
To Clinical Research
27
2.10 Experimental design
Sample Copy. Not For Distribution.
![Page 11: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/11.jpg)
x
2.11 Replication
2.12 Randomization
2.13 Control of error
2.13.1 Blocking
2.13.2 Proper experimental techniques
2.13.3 Data analysis
2.14 Design on single factor clinical experiments
2.15 Complete randomized design
2.15.1 Randomization
2.16 Analytical method
2.17 Analysis of variance (ANOVA)
2.18 Randomized complete block design (RCBD)
2.19 Randomization and lay out
2.20 Analytical method
2.21 Analysis of variance-ANOVA RCBD
2.22 Latin square designs-LSD
2.23 Randomization and lay out
2.24 Analytical method-ANOVA LSD
2.25 Cross over design –Switch back design
2.26 Bio-equilancy trial
2.27 Design approach to human genetics
2.28 Multinomial distribution in clinical trial
2.29 Moment generating function of crosses between
males and females genotypes (MGF)
2.30 Simulation-Monte Carlo method (SMCM)
2.31 Expected values
2.32 Discrete series: Expected
2.33 Continuous series: Expected
2.34 Effect of genetic inheritance on new drug trail
2.35 Genetic correlation
2.36 Formulation of the model-genetics
2.36.1 Assumption of the model
2.37 Analytical procedure or methods
2.38 Genetic correlation
Sample Copy. Not For Distribution.
![Page 12: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/12.jpg)
xi
2.38.1 ANOVA and co variances of cd4 count
and RNA viral load
2.39 Estimation of heritability
2.40
2.41
Method of analysis of clinical research -non
orthogonal life data sets
Model formulation of non-orthogonal clinical
data
3 Sensor Fusion -Noise Estimation By
Snapshots Techniques
60
3.10 Introduction
3.11 Model formulation: diseases/infection
susceptible model
3.12 Pharmacokinetic model for FMD
3.12.1 Model formulation
3.13 Formulation of epidemiological probability
model
3.14 Sensor fusion with normally distributed
3.15 Practical utility of the sensor fusion model in
animal science
4 Statistical Methods For Clinical Research 73
4.10 Bio equilancy trial
4.11 Equivalence testing- a new gold standard
4.12 Comparing two response rates
4.13 Characteristics of normal distribution
4.14 Log normal distribution
5 Survival Analysis 86
5.10 Introduction
5.11 Model description
5.12 Survival analysis
5.13 Survival function
5.14 Concept of the model
5.14.1 Survival function
5.14.2 Estimation of s(t):
5.14.3 Probability density function:
Sample Copy. Not For Distribution.
![Page 13: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/13.jpg)
xii
5.15 Hazard Function(Hz)
5.16 Hazard rate
5.17 Exponential distribution
5.18 Gompertz –Makeham distribution model
5.19 Censoring and life table methods
5.20 Kaplan Meir analysis
5.21 Estimation with censored data
5.22 Non informative censoring
5.23 Mantel-haenszel test
5.24 K-Sample Mantel-Haenszel test
6 Statistical Models Approach To Life
Threatened Diseases -An Indian Perspective
111
6.10 Introduction
6.11 Demographic features of HIV infected women
6.12 Model description
Table 6.13.1 Probabilities of transmitting HIV,
at of before birth by ASSA model.
Table 6.13.2 Probabilities of transmitting HIV
of African infants infected at 4-6 weeks, after
birth to mothers on HAART
6.13 Intrauterine and intrapartum transmission
6.14 Transmission probability at or before birth, in
the absence of ARV-prophylaxis.
6.15 Neural network modeling in HIV/AIDS
6.16 Application of neural network in medical
science
6.17 Equation of survival analysis –model constructs
6.18 Neural network model –corollary
6.19 Hierarchical neural nets for survival analysis
6.20 Nonhierarchical neural nets for survival analysis
6.21 Salient findings of neural net work model fitting
6.22 Application of the neural network in HIV or
medical research
6.23 Modeling on assessment of quality of life of
Sample Copy. Not For Distribution.
![Page 14: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/14.jpg)
xiii
patients
6.24 Quality life domain model
6.25 Model formulation
6.26 Model structure
6.27 Principle component analysis
6.28 Model discussion
7 Image Processing Modeling On
Radiographic Features
154
7.10 Introduction
7.11 Analog image processing
7.12 Digital image processing
7.13 Application of digital image processing
7.14 Different stages of digital image processing
7.15 Signal
7.16 Relationship
7.17 How a digital image is formed
7.18 Overlapping fields-Machine computer vision
7.19 Computer graphics
7.20 Artificial intelligence
7.21 Signal processing
7.22 Signals
7.23 Analog signal
7.24 Human voice
7.25 Digital signal
7.26 System
7.27 Sampling
7.28 Quantization
7.29 Application of digital image processing
7.30 Image sharpening and restoration
7.31 UV imaging
7.32 Transmission and encoding
7.33 Machine /Robot vision
7.34 Hurdle detection
7.35 Line follower robot
Sample Copy. Not For Distribution.
![Page 15: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/15.jpg)
xiv
7.36 Color processing
7.37 Pattern recognition
7.38 Video processing
7.39 Modeling on image processing
7.40 Camera pixels
7.41 Oversampling.
7.42 Zooming
7.43 Pixel
7.44 Resolution
7.45 Megapixels: We can calculate mega pixels of a
camera using pixel resolution
7.46 Advantage
7.47 Kernel regression model for image processing
7.48 Advanced model of image processing
8 Data Base Management In Clinical Research 181
8.10 Introduction
8.11 Data collection and documentation
8.12 Database management system (DBMS) and its
Applications
8.13 Database management system models (DBMS)
8.14 Hierarchical model
8.15 Network model
8.16 Object oriented model
8.17 DBMS-Architecture
8.18 Data types and diversity of clinical research
8.19 Structured and unstructured Sequences
8.20 Diagrammatic expression (graphs)
8.21 High-Dimensional Data sets
8.22 Data dimension
8.23 Data- temporal
8.24 Scalar and Vector Fields
8.25 Statistical and mathematical Models
8.26 Constraints
8.27 Data Provenance
Sample Copy. Not For Distribution.
![Page 16: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/16.jpg)
xv
8.28 Workflow Management
8.29 Data Integration
8.30 Data ware housing
8.31 Time variant
8.32 Major components
8.33 Minor components-Schematic diagram of ware
housing
8.34 Data base security
8.35 Data base environment
8.36 Data security risks
8.37 Clinical data tempering
8.38 Falsifying user identies
8.39 Password related threats
8.40 Unauthorized access to data matrix
8.41 Different dimension of clinical data base
security
8.42 Requirement of data security
8.43 Masking or blinding data sets
9 Ethical Issues And Perspective On Clinical
Research
205
9.10 Introduction
Experimental unit
Treatment
9.11 Evaluation
9.12 Principles of good clinical practices (GCD)
9.13 Basic consideration of clinical research-
objective of trial:
9.14 Target population and patient selection
9.15 Eligibility criteria for anti-infective agents
Inclusion criteria
Exclusion criteria
9.16 Design –basic consideration of clinical research
Selection of control
Statistical consideration
Sample Copy. Not For Distribution.
![Page 17: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/17.jpg)
xvi
Other consideration
9.17 Different design of trails: parallel group design:
9.18 Cross overdesign
9.19 Different design of trails: parallel group design:
9.20 Factorial design
9.21 Randomization
9.22 Merits of randomization
9.23 Blinding or masking
9.24 Ethical issues and perspectives on clinical
research Milestones
9.25 Ethical issues
9.26 Flow diagram of informed consent
9.27 Indian scenario
9.28 Ethical issues
Sample Copy. Not For Distribution.
![Page 18: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/18.jpg)
xvii
TABLE AND FIGURES CONTENTS
Index Contents
TABLES
2.39.1 Genetic correlation matrix of CD4 count
follow up of infected mother who are
transfer HIV infection to her child
2.39.2 Genetic correlation matrix of RNA Plasma
viral load of infected mother who are
transfer HIV infection to her child
5.16.18 Survivability of different CD4 category
with HAART duration
5.22.2 Means and Medians for Survival Time
5.22.4 Area under the Curve (AUC) on different
mortality associated parameters.
6.13.1 Probabilities of transmitting HIV, at of
before birth by ASSA model.
6.13.2 Probabilities of transmitting HIV of
African infants infected at 4-6 weeks, after
birth to mothers on HAART
6.18 .1 Table 6.18(f) Confidence interval of Mean
survivability with gender wise and
different age class
6.28.2 Total scores of different domains of QOL
(Transformed scale)
6.28.3 Total scores of different domains of QOL
(Transformed scale)
6.28.4 Correlation matrix of categorical variables
of PLHIV Table 6.28(e) Strong correlation
of categorical variables of HIV patients.
6.28.5 Associated parameters in PLHIV.
6.28.6 Principle component analysis of
associated parameters in PLHIV.
6.28.7 Component transformation matrix of
Sample Copy. Not For Distribution.
![Page 19: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/19.jpg)
xviii
associated parameters in PLHIVs
5.16.2 Kaplan-Meier survival curve and its 95%
confidence interval for patients initiating
therapy from different CD4 category
5.22.1 Mortality in relation to the low birth
weight (kgs)
5.22.3 ROC curve depicted for weight of the baby
and defined parameters with respect to
mortality
6.15.1 Probability plot for base line CD4 count at
the time of HAART Initiation of pregnant
women
1.15.2 Probability plot for CD4 count at the time
of pregnancy
6.18.1 Descriptive statistics of PLHA five year
Cohort
FIGURES
6.18.2 Survivability at Five years
6.18.3 Survivability function at Five years
6.18.4 Log survival function and Hazard function
6.18.5 Cohort Survivability and survival function
among PLHA with different age class
6.18.6 Log survival function and Hazard function
with different age class
6.18.7 Multilayer Neural net work Output and
hidden layer
6.18.9 Radial basis function neural net work
Output and hidden layer
6.18.10 Gender wise relation with Survivability
6.18.11 Mean scores of QOL
Sample Copy. Not For Distribution.
![Page 20: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/20.jpg)
Recent Statistical Techniques in Clinical Research
19
1
Clinical Research Design
And Statistical Methods ______________________________________________________
1.1 Introduction
Statistical methods are backbone of clinical research and will infer
the sound knowledge on formulation and randomization of clinical
trials. The statistical tools are very important for clinicians and
policymakers for the implementation of new RCT-guidelines and
hands-on training programme for clinicians, specialists and young
researchers. Many clinicians and researchers will fail to use the
advanced statistical tools in their research work, due to non
availability of statistical literatures and reference books. In this
section we discussed the different statistical tools, study designs
and advanced statistical methods for formulating the clinical trials.
The clinical trial is a study in human subjects, in which
treatment (intervention) is initiated specially for therapy
evaluation. Practically, we define a clinical trial as a type of study
for comparing the effect and value of intervention against a control
in human beings (Lind et al., 1996). It is an experiment involving
testing medical treatment in human subjects (Zhang et al., 2005).
Further the clinical trial may be elucidated in many ways viz., the
comparison of Stavudine fixed dose combination versus placebo
consideration on the length of survival in patients with HIV/ AIDS,
evaluating the effectiveness of a new antifungal medication on
Athlete's foot, and hormonal therapy on the reduction of breast
cancer etc. The clinical trial is more important perspective for
Sample Copy. Not For Distribution.
![Page 21: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/21.jpg)
Dr Basavarajaiah D M
20
clinicians and researchers. Many clinicians obtained evidence
based results and drew the conclusion based on the practical
approach. However, the trial is very easy for anecdotal information
about the benefit of therapy accepted for standard treatment care.
Many literature reported that high concentration of oxygen was
useful for therapy in premature infants until a clinical trial
demonstrated. Tsiatis et al. (2004) reported that, the prolonged use
of hormone replacement therapy for women followed menopause
problem.
1.2 Statistical Historical Perspectives of
Clinical Trial
Scurvy experiment was the first clinical trial conducted during
1747 by James F. Lind, a physician onboard of the “Salisbury“.
From1920onwards,R. A. Fisher introduced randomization as a core
principle in the statistical theory of the design of experiments.
During 1947-1948, Streptomycin was invented by randomized
controlled clinical trial (RCT) for curing tuberculosis and it was
published in the British Medical Journal, 1948.Statisticians, Sir
Austin Bradford Hill et al.(1951) reported that, a total of 1.80
million children participated in the largest trial to assess the
effectiveness of Salk vaccine in preventing paralysis or death from
poliomyelitis. Such large number was needed, because the
incidence rate of polio was about 1/2000 and evidence of treatment
effect was needed as soon as possible so that, the vaccine could be
routinely administered. As per the results, they observed that there
were two components (randomized and non-randomized) of the
trial. For the non-randomized component, one million children in
the first to third grades participated. The second grade was offered
vaccine, where as first and third grades formed the control group.
There was also randomized component where 0.8 million children
were screened in double blind placebo trial. The incidence of polio
in the randomized vaccinated group was less than half that in the
control group and even larger differences were seen in the decline
of paralytic polio.
Sample Copy. Not For Distribution.
![Page 22: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/22.jpg)
Recent Statistical Techniques in Clinical Research
21
1.3 Brief Concept of Study Design
1.3.1 CROSS SECTIONAL STUDY: In cross sectional study, the
data were obtained from a random sample of the population at one
point of time. This gave a snapshot of the population. Based on
single survey of a specific population or a random sample thereof,
we determined the proportion of individuals with heart disease at
one point of time. This is referred to as the prevalence of disease.
Collection of demographic and other information would allow to
break down the prevalence by age, race, sex, socioeconomic status
and geographical area etc. Important special case, where the
exposure and diseases are dichotomous, the data from a cross
sectional study can be represented by using 2X2 contingency table:
Characteristics Disease No disease Total
Exposed A B (A+B)
Unexposed C D (C+D)
Total (A+C) (B+D) GT
Prevalence of disease (%) = , Probability of exposure = ,
Prevalence of disease among exposed population = ,
Prevalence of diseases among unexposed population = . We
can also assess the association between the exposure and disease
using data from a cross sectional study. One such measure is
relative risk was estimated by:
It is easy to conclude that the relative risk has following
properties:
1 positive association; i.e., the exposed population has
higher disease probability than the unexposed population.
1 no association; i.e., the exposed population has the same
disease probability as the unexposed population.
1 negative association; i.e., the exposed population has
lower disease probability than the unexposed population.
Sample Copy. Not For Distribution.
![Page 23: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/23.jpg)
Dr Basavarajaiah D M
22
1.3.2 LONGITUDINAL STUDY: In a longitudinal study,
subjects are followed over time and single or multiple
measurements of the variables of interest are obtained.
Longitudinal epidemiological studies generally fall into two
categories, prospective (moving forward in time) and retrospective
(going backward in time).
1.3.3 PROSPECTIVE STUDY: In a prospective study, a cohort
of individuals are obtained, who are free of a particular disease
under study and data are collected on certain risk factors i.e.
children were exposed to contaminants with respect to age, sex,
race etc. The individuals are then followed over for some specified
period of time, to determine whether they get the disease or not.
The relationship between the probability of getting disease during
certain time period is called as incidence of the disease.
1.3.4 RETROSPECTIVE STUDY: CASE CONTROL:A very
popular design in many epidemiological and RCT studies is the
retrospective cross sectional, case control design. In such a study,
individuals with disease (cases) and the individuals without disease
(controls) were identified. Using records or questionnaires the
investigators go back in time and ascertain the exposure status and
risk factors from their past. Such data are used to estimate relative
risk factors.
Drug
response Cases Controls Total
Responded A B (A+B)
Not
responded C D (C+D)
Total (A+C) (B+D) GT
Sensitivity (%) = , Specificity (%) = , Drug responded
among cases = ,Drug not responded among cases = ,
relative risk was estimated in accordance with the above
mentioned formula.
Sample Copy. Not For Distribution.
![Page 24: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/24.jpg)
Recent Statistical Techniques in Clinical Research
23
1.4 Different Phases of Clinical Trial
The process of drug development can be broadly classified as
preclinical and clinical. Preclinical refers to experimentation that
occurs before exposure to human subjects, whereas clinical refers
to experimentation after exposure to the human subjects. It will be
assumed that, the drug has already been developed by the chemist
or biologist, tested in the laboratory for biological activity (invivo)
and the new drug or therapy is found to be classified and
sufficiently promising to be introduced into the humans. Within the
paradigm, we can classify the clinical trial into four phases;
Phase I: To explore possible toxic effects of drugs and determine a
tolerated dose for further experimentation. During this phase the
pharmacology of the drug may also be explored.
Phase II:Screening and feasibility study by initial assessment for
therapeutic effects and further assessment of toxicities.
Phase III: Comparison of new intervention (drug or therapy) to
the current standard of treatment; both with respect to efficacy and
toxicity.
Phase IV: Observational study of morbidity/ adverse effects on
human intervention.
1.5 Drug Development Process
Sample Copy. Not For Distribution.
![Page 25: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/25.jpg)
Dr Basavarajaiah D M
24
1.6 Study Population:
Population is an universe. The target population of the study must
be defined in advance with unambiguous inclusion and exclusion
criteria. The impact of these criteria on the study design will be
visualized on the basis of location. Individual population group or
sample should be generalized before the recruitment procedure in
accordance with human ethical guidelines.
1.7 Statistical Issues And Methods
Our goal in clinical trial is to estimate an endpoint related to the
treatment efficacy and sufficient precision to aid the investigations
for determining, whether the proposed treatment to be studied for
further period. The following points highlighted for the recruitment
of the patients during study interventions;
1) The proportion of patients responding to the treatment arm
[response has to be unambiguously defined].
2) The proportion of patients with known side effects.
3) Average decreased level of serological markers over a period
of time.
Many literatures quoted that “the reliable statistical tools
inform sound decisions and better outcomes” incorrect /unethical
use of statistics can produce misleading results, poor advice and
worse choice.
For example, suppose we consider the patients with esophageal
cancer, who were treated with chemotherapy prior to surgical
resection. A complete response is defined as an absence of
macroscopic tumor at that time of surgery; we suspect that this
may occur with 35 per cent probability using a drug under
investigation in phase-II. The 35 per cent is just a guess, possibly
based on similar acting drugs used in the past and the goal is to
estimate the actual response rate with sufficient precision, in this
case we want 95 per cent confidence interval to be within 15 per
cent of the true or positives value.
As researchers, we start by positioning a statistical model; i.e.,
let 'p' denote the complete population response rate. Suppose we
Sample Copy. Not For Distribution.
![Page 26: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/26.jpg)
Recent Statistical Techniques in Clinical Research
25
conduct an experiment having η patients with known esophageal
cancer treated with the chemotherapy prior to surgical resection.
The data was collected based on the incumbent changes of
serological values and associated casual indicators. The result of
the experiment yields a random variable 'X', the number of patients
in a sample of size 'η' that have a complete response rate. A
popular discrete probability model for this scenario is to assume
that:
, X Binomial (η ,p)
Where, n = number of cancer patients, p = probability of
treatment success, (1-p) = probability of treatment failure and 'x' is
random variable with different time period.
We assumed that, the probability of treatment success is equal
to the probability of failure. The treatment effect should be
dichotomous and the probability of success and failure is always
equal to one, the mean of the probability was 'np' and standard
deviation was „npq‟. If the number of patients increased in
treatment arm, the binomial distribution tends to be poison
distribution. When „n‟ is sufficiently large, the distribution of the
sample proportion is well approximated by a normal
distribution with mean „np‟ and variance p(1-p)/n,
This approximation is useful for inference
regarding the population parameter „np‟. Because of the
approximate normality, the estimator „p‟ will be within 1.96
standard deviations of „np‟, approximately 95 per cent of the time
(Approximation gets better with increasing sample size).
Therefore, the population parameter „np‟ will be within the interval
, with approximately 95 per cent
probability. Since, the value „p‟ is unknown to us, we approximate
using „p‟ to obtain the approximate 95 per cent confidence interval.
Switch back to above mentioned example, where best guess for the
response rate is about 35 per cent, if we want the precision of our
estimator to be such that the 95 per cent confidence interval with in
the 15 per cent of true „p‟, then we need to calculate sample size.
Sample Copy. Not For Distribution.
![Page 27: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/27.jpg)
Dr Basavarajaiah D M
26
= 39 patients (needed for our
experimentation).
If the number of patients increased in phase II and III trial, the
binomial distribution tends to be poison approximation the
probability of success is very small or nearer to zero. Binomial
distribution showed to be poison approximation, the density
function is given by:
, so that the binomial distribution tends to the
Poisson distribution with mean .
It was showed that the expected number of patients of ‟ is
equal to and that variance of x is also . This distribution
tends to the normal distribution as „ gets large in essentially the
same sense as the binomial distribution tends to normality. The
Poisson law is useful in getting quick approximation to binomial
probabilities.
1.8 Application Of Multinomial Distribution In
Clinical Trial
Let us, consider the simple situation of clinical trial conducted in
varied geographical locations and obtaining the data as per the
norms. We assumed „n‟ clinical trials and there were „t‟ possible
results viz., a1, a2, a3,……at with probability that we shall get n1a1‟s
n2a2‟s ..anat. Again the matter was solved by permutation and
combination for any possible arrangements of n1 a1‟s, n2a2
‟s …and
so on, the probability of trial is equal to:
Then, the cumulative probability for pooled trial is calculated
by the following density function
For example, the new-clinical trials were conducted in the
following geographical location, the resulted observations are
mentioned below:
Sample Copy. Not For Distribution.
![Page 28: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/28.jpg)
Recent Statistical Techniques in Clinical Research
27
1.8.1 GEHAN’S TWO STAGE DESIGN: Unlikely treatment
will achieve some low level of response because due to recursive
changes of clinical and biological parameters, the researcher
favorably should stop the trial as early as possible. For example,
suppose 20 per cent response rate is the lowest i.e., we consider the
acceptable new treatment experimentation. Suppose, no responses
in „n patients and if „n‟ is sufficiently large, then we confident that
the treatment is ineffective. Statistically how „n‟ is drawn from
parent population, so that if there are zero responses among „n‟
patients, we relatively confident that the response rate will not be
achieved 20 per cent or better
if and p≥0.20,
Choose „n‟ so that =0.05 or . This
leads to (rounding up), thus with the patient is unlikely
(≤0.50) that no one could response, if the treatment response rate
was greater than 20 per cent. Thus, zero patients responding among
Sl.
No. Location
No. of
patients
Sample
number
Probability
values
1. Northern
region 14 n1 0.23
2. Southern
region 15 n2 0.22
3. Middle
region 19 n3 0.36
4. Eastern
region 28 n4 0.44
5. Western
region 32 n5 0.52
Total 108
Sample Copy. Not For Distribution.
![Page 29: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/29.jpg)
Dr Basavarajaiah D M
28
15 might be used as evidence to stop the clinical trial and declare
the treatment failure. This was the logic behind Gehan‟s two stage
design and he also suggested the following strategy, if minimal
acceptable response rate is then we choose the first stage with
patients such that:
=0.5;
If there is zero response among the first patients, then stop
and treatment can be declared as failure. Otherwise, continue with
recruiting additional patients that will ensure a certain degrees of
freedom of predetermined accuracy in 95 per cent confidence
interval.
Figure 2: Normal distribution simulated curve from Gehan’s
two stage design with 99 per cent confidence interval
From figure 2, Gehan‟s design was treated 15 patients
initially. If none responded, the treatment would be declared as
failure and the study stopped.
Sample Copy. Not For Distribution.
![Page 30: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/30.jpg)
Recent Statistical Techniques in Clinical Research
29
Get Complete Book At Educreation Store
www.educreation.in
Sample Copy. Not For Distribution.
![Page 31: Sample Copy. Not For Distribution. · techniques and tools with suitable examples from real life data ... RCBD and LSD with mathematical models. ... would have liked to have solved](https://reader031.fdocuments.net/reader031/viewer/2022011803/5b90bfab09d3f2b6628cb211/html5/thumbnails/31.jpg)
Sample Copy. Not For Distribution.