Stat

Statistical Methods in Clinical
Research
Dr Ranjith P
DNB Resident ACME Pariyaram ,
Kerala
www.dnbpediatrics.com

Overview
 Data types
 Summarizing data using descriptive statistics
 Standard error
 Confidence Intervals

Overview
 P values
 Alpha and Beta errors
 Statistics for comparing 2 or more groups with
continuous data
 Non-parametric tests

Overview
 Regression and Correlation
 Risk Ratios and Odds Ratios
 Survival Analysis
 Cox Regression

 Forest plot
 PICOT
overview

Types of Data
 Discrete Data-limited number of choices
 Binary: two choices (yes/no)
 Dead or alive
 Disease-free or not
 Categorical: more than two choices, not ordered
 Race
 Age group
 Ordinal: more than two choices, ordered
 Stages of a cancer
 Likert scale for response
 E.G. strongly agree, agree, neither agree or disagree, etc.

Types of data
 Continuous data
 Theoretically infinite possible values (within
physiologic limits) , including fractional values
 Height, age, weight
 Can be interval
 Interval between measures has meaning.
 Ratio of two interval data points has no meaning
 Temperature in celsius, day of the year).
 Can be ratio
 Ratio of the measures has meaning
 Weight, height

Types of Data
 Why important?
 The type of data defines:
 The summary measures used
 Mean, Standard deviation for continuous data
 Proportions for discrete data
 Statistics used for analysis:
 Examples:
 T-test for normally distributed continuous
 Wilcoxon Rank Sum for non-normally distributed
continuous

Descriptive Statistics
 Characterize data set
 Graphical presentation
 Histograms
 Frequency distribution
 Box and whiskers plot
 Numeric description
 Mean, median, SD, interquartile range

Histogram
Continuous Data
No segmentation of data into groups

Frequency Distribution
Segmentation of data into groups
Discrete or continuous data

Box and Whiskers Plots

Box and Whisker Plots
Popular in Epidemiologic Studies
Useful for presenting comparative data graphically

Numeric Descriptive Statistics
 Measures of central tendency of data
 Mean
 Median
 Mode
 Measures of variability of
data(dispersion)
 Standard Deviation, mean deviation
 Interquartile range, variance

Mean
 Most commonly used measure of central tendency
 Best applied in normally distributed continuous data.
 Not applicable in categorical data
 Definition:
 Sum of all the values in a sample, divided by the number of
values.

 Eg mean Height of 6 adolescent
children 146 ,142,150,148,156,140
 Ans ?
 882/6 =147

Median
 Used to indicate the “average” in a
skewed population
 Often reported with the mean
 If the mean and the median are the same,
sample is normally distributed.

 It is the middle value from an ordered
listing of the values
 If an odd number of values, it is the middle
value 1.2.3.4.5 ie 3
 If even number of values, it is the average
of the two middle values.1,2,3,4,5,6 ie
3+4/2 = 3.5
 Mid-value in interquartile range

Mode
 Infrequently reported as a value in studies.
 Is the most common value eg. 1,3,8,9,5,8,5,6
 mode = 5
.

Interquartile range
 Is the range of data from the 25th percentile
to the 75th percentile
 Common component of a box and whiskers
plot
 It is the box, and the line across the box is the
median or middle value
 Rarely, mean will also be displayed.

Mean deviation(standard
deviation )
 Mean deviation(SD) = £I X- Ẍ I / n
 n is the no of observations Ẍ is the mean ,
X each observation
 Square mean deviation= variance=
£I X- Ẍ I² / n
Root mean square deviation =√£I X- Ẍ I² / n

Variance
 Square of SD(standard deviation )
Coefficient of variance = SD/ mean x 100
Eg. If sd is 3 mean is 150
Variance is 9, coefficient of variance is
300/150 = 2

Standard Error
 A fundamental goal of statistical analysis is to
estimate a parameter of a population based
on a sample
 The values of a specific variable from a
sample are an estimate of the entire
population of individuals who might have
been eligible for the study.
 A measure of the precision of a sample

Standard Error
 Standard error of the mean
 Standard deviation / square root of (sample
size)
 (if sample greater than 60)
 Sd/ √n
 Important: dependent on sample size
 Larger the sample, the smaller the standard
error.

Clarification
 Standard Deviation measures the
variability or spread of the data in an
individual sample.
 Standard error measures the precision
of the estimate of a population
parameter provided by the sample
mean or proportion.

Standard Error
 Significance:
 Is the basis of confidence intervals
 A 95% confidence interval is defined by
 Sample mean (or proportion) ± 1.96 X standard error
 Since standard error is inversely related to the
sample size:
 The larger the study (sample size), the smaller the
confidence intervals and the greater the precision of the
estimate.

 Mean +/- 1 sd = 68.27% value
Mean +/- 2 sd = 95.49% value
 Mean +/- 3 sd = 99.7% value
 Mean +/- 4 sd = 99.9% value

Confidence Intervals
 May be used to assess a single point
estimate such as mean or proportion.
 Most commonly used in assessing the
estimate of the difference between two
groups.

Confidence Intervals
Commonly reported in studies to provide an estimate of the precision
of the mean.

P Values
 The probability that any observation is
due to chance alone assuming that the
null hypothesis is true
 Typically, an estimate that has a p
value of 0.05 or less is considered to
be “statistically significant” or unlikely
to occur due to chance alone. Null
hypothesis rejected

 The P value used is an arbitrary value
 P value of 0.05 equals 1 in 20
chance
chance
chance.

Errors
 Type I error
 Claiming a difference between two
samples when in fact there is none.
 Remember there is variability among samples-
they might seem to come from different
populations but they may not.
 Also called the  error.
 Typically 0.05 is used

Errors
 Type II error
 Claiming there is no difference between
two samples when in fact there is.
 Also called a  error.
 The probability of not making a Type II
error is 1 - , which is called the power of
the test.
 Hidden error because can’t be detected
without a proper power analysis

Errors
Null
Hypothesis
H0
Alternative
Hypothesis
H1
Null
Hypothesis
H0
No Error Type I

Alternative
Hypothesis
H1
Type II

No Error
Test result
Truth

Sample Size Calculation
 Also called “power analysis”.
 When designing a study, one needs to
determine how large a study is needed.
 Power is the ability of a study to avoid a
Type II error.
 Sample size calculation yields the
number of study subjects needed, given
a certain desired power to detect a
difference and a certain level of P value
that will be considered significant.www.dnbpediatrics.com

Sample Size Calculation
 Depends on:
 Level of Type I error: 0.05 typical
 Level of Type II error: 0.20 typical
 One sided vs two sided: nearly always two
 Inherent variability of population
 Usually estimated from preliminary data
 The difference that would be meaningful
between the two assessment arms.

One-sided vs. Two-sided
 Most tests should be framed as a two-
sided test.
 When comparing two samples, we usually
cannot be sure which is going to be be
better.
 You never know which directions study results
will go.
 For routine medical research, use only two-
sided tests.

Statistical Tests
 Parametric tests
 Continuous data normally distributed
 Non-parametric tests
 Continuous data not normally distributed
 Categorical or Ordinal data

Comparison of 2 Sample Means
 Student’s T test
 Assumes normally distributed continuous
data.
T value = difference between means
standard error of difference
 T value then looked up in Table to
determine significance

Paired T Tests
 Uses the change before
and after intervention in a
single individual
 Reduces the degree of
variability between the
groups
 Given the same number of
patients, has greater
power to detect a
difference between
groups

Analysis of Variance(ANOVA)
 Used to determine if two or more
samples are from the same
population-
 If two samples, is the same as
the T test.
 Usually used for 3 or more
samples.

Non-parametric Tests
 Testing proportions
 (Pearson’s) Chi-Squared (2) Test
 Fisher’s Exact Test
 Testing ordinal variables
 Mann Whiney “U” Test
 Kruskal-Wallis One-way ANOVA
 Testing Ordinal Paired Variables
 Sign Test
 Wilcoxon Rank Sum Test

Use of non-parametric tests
 Use for categorical, ordinal or non-normally
distributed continuous data
 May check both parametric and non-
parametric tests to check for congruity
 Most non-parametric tests are based on
ranks or other non- value related methods
 Interpretation:
 Is the P value significant?

(Pearson’s) Chi-Squared (2) Test
 Used to compare observed proportions of an
event compared to expected.
 Used with nominal data (better/ worse;
dead/alive)
 If there is a substantial difference between
observed and expected, then it is likely that
the null hypothesis is rejected.
 Often presented graphically as a 2 X 2 Table

Non parametric test
For comparing 2 related samples
-Wilcoxon Signed Rank Test
For comparing 2 unrelated samples
-Mann- Whitney U Test
For comparing >2groups
-Kruskal Walli Test

Mann–Whitney U test
 Mann–Whitney–Wilcoxon (MWW), Wilcoxon
rank-sum test, or Wilcoxon–Mann–Whitney
test) is a non-parametric test especially that a
particular population tends to have larger
values than the other.
 It has greater efficiency than the t-test on non-
normal distributions, such as
a mixture of normal distributions, and it is
nearly as efficient as the t-test on normal
distributions.

STUDENT T TEST
 A t-test is any statistical hypothesis
test in which the test statistic follows
a normal
distri bution if the null hypothesis is
supported.
 It can be used to determine if two sets of
data are significantly different from each
other, and is most commonly applied
when the test statistic would follow
a normal distribution

 The Kaplan–Meier estimator,also known
as the product limit estimator, is
an estimator for estimating the survival
function from lifetime data.
 In medical research, it is often used to
measure the fraction of patients living for a
certain amount of time after treatment.
 The estimator is named after Edward L.
Kaplan and Paul Meier.

 A plot of the Kaplan–Meier
estimate of the survival function is
a series of horizontal steps of
declining magnitude which, when
a large enough sample is taken,
approaches the true survival
function for that population.

 ODDS RATIO
In case control study –
measure of the strength of the
association between risk factor
and out come

Odds ratio
Lung
cancer(case
s)
No lung
cancer
(controls)
smokers 33 (a) 55 (b)
Non smokers 2 (c) 27 (d)
TOTAL 35(a+c) 82(b+d)

 Odds ratio =ad/bc
=33*27/55*2
=8.1
ie smokers have 8.1 times have the
risk to develop lung cancer than non
smokers

RELATIVE RISK
 Measure of risk in a cohort
study
 RR=lncidence of disease
among exposed /
incidence among non exposed

Cigarette
smoking
Developo
d lung
cancer
Not
Developo
d lung
cancer
total
Yes 70 (a) 6930 (b) 7000
(a+b)
No 3 (c) 2997 (d) 3000
(c+d)

 Incidence among
smokers=70/7000=10/1000
 Incidence among non
smokers=3/3000=1/1000
 Total incidence= 73/10000=7.3/1000

RR=lncidence of disease among exposed/
incidence among non exposed
Relative risk of lung cancer=10/1=10
Incidence of lung cancer is 10 times higher in
exposed group (smokers) , ie having a
Positive relationship with smoking
Larger RR ,more the strength of association

Attributable risk
 It is the difference in incidence
rates of disease between exposed
group(EG) and non exposed
group(NEG)
 Often expressed in percent

 (Incidence of disease rate in EG-
Incidence of disease in NEG/incidence
rate in EG ) * 100
AR= 10-1/10=90%
I.e. 90% lung cancers in smokers was
due to their smoking

Population attributable Risk
 It is the incidence of the disease in total
population - the incidence of disease
among those who were not exposed to
the suspected causal factor/incidence of
disease in total population
 PAR=7.3-1/7.3=86.3%, ie 86.3 %
disease can be avoided if risk factors like
cigarettes were avoided

Mortality rates & Ratios
 Crude Death rate
 No of deaths (from all cases )per 1000
estimated mid year population(MYP) in
one year in a given place
 CDR=(No. deaths during the year
/MYP)*1000

 CDR in Panchayath A is 15.2/1000
 Panchayath B is 8.2/1000 population
Health status of Panchayath B is better
than A

 Specific Death rate=(No of diseases due to
specific diseases during a calendar year/
MYP)*1,000
Can calculate death rate in separate diseases
eg . TB, HIV 2/1000, 1/1000 resp
Age groups 5-20yrs, <5yrs - 1/1000, 3/3000
resp.
Sex eg. More in males,
Specific months,etc

Case fatality rate(ratio)
 (Total no of deaths due to a particular
disease/Total no of cases due to same
disease)*100
 Usually described in A/c infectious
diseases
 Dengue, cholera, food poisoning etc
 Represent killing power of the disease

Proportional mortality rate(ratio)
 Due to a specific disease=(No of
deaths from the specific disease in a
year/ Total deaths in an year )*100
 Under 5 Mortality rate=(No of deaths
under 5 years of age in a given
year/Total no of deaths during the
same period)*100

Survival rate
 (Total no of patients alive after 5yrs/Total
no of patients diagnosed or treated)*100
 Method of prognosis of certain disease
conditions mainly in cancers
 Can be used as a yardstick for
assessment of standards of therapy

INCIDENCE
 No of new cases occurring in a defined
population during a specified period of time
 (No of new cases of specific disease during a
given time period / Population at risk)*1000
 Eg 500 new cases of TB in a population of
30000, Incidence is (500/3000)*1000
ie 16.7/1000/yr expressed as incidence rate

Incidence-uses
 Can be expressed as Special incidence
rate , Attack rate , Hospital admission rate
, case rate etc
 Measures the rate at which new cases are
occurring in a population
 Not influenced by duration
 Generally use is restricted to acute
conditions

PREVALENCE
 Refers specifically to all current
cases (old & new) existing at a
given point of time, or a period of
time in a given population
 Referred to as a rate , it is really a
a ratio

 Point prevalence=(No of all currant cases
(old& new) of a specified disease existing
at a given point of time / Estimated
population at the same point of time)*100
 Period prevalence=(No of existing cases
(old& new) of a specified disease during
a given period of time / Estimated mid
interval population at risk)*100

jan1
dec 31 case1
3 4

5
case 6
case 7
case 8

 Incidence - 3,4,5,8
 Point prevalence at jan 1- 1,2& 7
 Point prevalence at Dec 31- 1,3,5&8
 Period prevalence(jan-Dec)-
1,2,3,4,5,7&8

Relationship b/n Incidence &
prevalence
 Prevalence=Incidence*Mean duration
 P=I*D I=P/D D=P/I
 Eg: Incidence=10 cases/1000
population/yr
 Mean duration 5 yrs
 Prevalence=10*5 =50/1000 population

PREVALENCE-USES
 Helps to estimate magnitude of
health/disease problems in the community,
& identify potential high risk populations
 Prevalence rates are especially useful for
administrative and planning purposes
eg: hospital beds, man power
needs,rehabilation facilities etc.

Statistical significance
 P value (hypothesis)
 95% CI (Interval)

P value & its interpretation
“it is the probability of type 1 error”
 The chance that, a difference or
association is concluded , when actually
there is none.

 Study of prevalence of obesity in male &
female child in a classroom.
50 students
of 25 boys- 10 obese
of 25 girls - 16 obese
p value : 0.02

Null hypothesis: “no difference in obesity
among boys & girls in the classroom”

study ,Bubble vs conventional CPAP for
prevention of extubation Failure( EF) in
preterm very low birth weight infants.
EF bCPAP =4(16)
cCPAP =9(16)
p value-0.14

Null hypothesis: “ no difference in EF
among preterm babies treated with
bCPAP &cCPAP.”

95% CI
95%CI= Mean ‡1.96SD(2SD)
= Mean ‡ 2SE
1) 100 children attending pediatric OP.
mean wt=15kg SD=2
95%CI =?

Interpretation of 95%CI
 If a test is repeated 100times , 95 times
the mean value comes between this
value.
 If CI of 2 variables overlap, the chance
of significant difference is very less.

Measures Of Risk
 case control study- Odds ratio
 Cohort study -RR,AR

Chi-Squared (2) Test
 Chi-Squared (2) Formula
 Not applicable in small samples
 If fewer than 5 observations per cell, use
Fisher’s exact test

BREAK

Correlation
 Assesses the linear relationship between two variables
 Example: height and weight
 Strength of the association is described by a correlation
coefficient- r
 r = 0 - .2 low, probably meaningless
 r = .2 - .4 low, possible importance
 r = .4 - .6 moderate correlation
 r = .6 - .8 high correlation
 r = .8 - 1 very high correlation
 Can be positive or negative
 Pearson’s, Spearman correlation coefficient
 Tells nothing about causation

Correlation
Source: Harris and Taylor. Medical Statistics Made Easy

Correlation
Perfect Correlation
Source: Altman. Practical Statistics for Medical Research

Regression
 Based on fitting a line to data
 Provides a regression coefficient, which is the slope of the
line
 Y = ax + b
 Use to predict a dependent variable’s value based on the
value of an independent variable.
 Very helpful- In analysis of height and weight, for a known
height, one can predict weight.
 Much more useful than correlation
 Allows prediction of values of Y rather than just whether
there is a relationship between two variable.

Regression
 Types of regression
 Linear- uses continuous data to predict continuous
data outcome
 Logistic- uses continuous data to predict
probability of a dichotomous outcome
 Poisson regression- time between rare events.
 Cox proportional hazards regression- survival
analysis.

Multiple Regression Models
 Determining the association between two
variables while controlling for the values of
others.
 Example: Uterine Fibroids
 Both age and race impact the incidence of
fibroids.
 Multiple regression allows one to test the impact of
age on the incidence while controlling for race
(and all other factors)

Multiple Regression Models
 In published papers, the multivariable models are
more powerful than univariable models and take
precedence.
 Therefore we discount the univariable model as it does not
control for confounding variables.
 Eg: Coronary disease is potentially affected by age, HTN,
smoking status, gender and many other factors.
 If assessing whether height is a factor:
 If it is significant on univariable analysis, but not on
multivariable analysis, these other factors confounded the
analysis.

Survivial Analysis
 Evaluation of time to an event (death,
recurrence, recover).
 Provides means of handling censored data
 Patients who do not reach the event by the end of
the study or who are lost to follow-up
 Most common type is Kaplan-Meier analysis
 Curves presented as stepwise change from
baseline
 There are no fixed intervals of follow-up- survival
proportion recalculated after each event.

Survival Analysis
Source: Altman. Practical Statistics for Medical Researchwww.dnbpediatrics.com

Kaplan-Meier Curve
Source: Wikipedia

Kaplan-Meier Analysis
 Provides a graphical means of comparing the
outcomes of two groups that vary by intervention or
other factor.
 Survival rates can be measured directly from curve.
 Difference between curves can be tested for
statistical significance.

Cox Regression Model
 Proportional Hazards Survival Model.
 Used to investigate relationship between an event
(death, recurrence) occurring over time and possible
explanatory factors.
 Reported result: Hazard ratio (HR).
 Ratio of the hazard in one group divided the hazard in
another.
 Interpreted same as risk ratios and odds ratios
 HR 1 = no effect
 HR > 1 increased risk
 HR < 1 decreased risk

Cox Regression Model
 Common use in long-term studies
where various factors might predispose
to an event.
 Example: after uterine embolization, which
factors (age, race, uterine size, etc) might
make recurrence more likely.

True disease state vs. Test result
not rejected rejected
No disease
(D = 0)

specificity
X
Type I error
(False +) 
Disease
(D = 1)
X
Type II error
(False -) 

Power 1 - ;
sensitivity
Disease
Test

Specific Example
Test Result
Pts with
disease
Pts without
the disease

Test Result
Call these patients “negative” Call these patients “positive”
Threshold

Test Result
without the disease
with the disease
True Positives
Some definitions ...

Test Result
without the disease
with the disease
False
Positives

Test Result
without the disease
with the disease
True
negatives

Test Result
without the disease
with the disease
False
negatives

Test Result
without the disease
with the disease
‘‘-’’ ‘‘+’’
Moving the Threshold: right

Test Result
without the disease
with the disease
‘‘-’’ ‘‘+’’
Moving the Threshold: left

TruePositiveRate
(sensitivity)
0%
100%
False Positive Rate
(1-specificity)
0% 100%
ROC curve

TruePositiveRate
0
%
100%
False Positive Rate
0
%
100%
TruePositiveRate
0
%
100%
False Positive Rate
0
%
100%
A good test: A poor test:
ROC curve comparison

Best Test: Worst test:
TruePositiveRate
0
%
100%
False Positive
Rate
0
%
100
%
TruePositive
Rate
0
%
100%
False Positive
Rate
0
%
100
%
The distributions
don’t overlap at all
The distributions
overlap completely
ROC curve extremes

FOREST PLOT
114www.dnbpediatrics.com

 An example forest plot of five odds ratios
(squares) with the summary measure
(centre line of diamond) and associated
confidence intervals (lateral tips of
diamond), and solid vertical line of no
effect. Names of (fictional) studies are
shown on the left, odds ratios and
confidence intervals on the right.

 A forest plot (or blobbogram[1]) is a
graphical display designed to illustrate
the relative strength of treatment effects
in multiple quantitative scientific studies
addressing the same question. It was
developed for use in medical research
as a means of graphically representing a
meta-analysis of the results of
randomized controlled trials.

 i. Probably a small study, with a wide CI,
crossing the line of no effect (OR = 1).
Unable to say if the intervention works
 ii. Probably a small study, wide CI , but
does not cross OR = 1; suggests
intervention works but weak evidence
 iii. Larger study, narrow CI: but crosses
OR = 1; no evidence that intervention
works

 iv. Large study, narrow confidence
intervals: entirely to left of OR = 1;
suggests intervention works
 v. Small study, wide confidence
intervals, suggests intervention is
detrimental
 vi. Meta-analysis of all identified studies:
suggests intervention works.

PICOT
 Used to test evidence based research
 Population
 Intervension or issue
 Comparison with another intervention
 Outcome
 Time frame

Stat

Recommandé

Recommandé

Contenu connexe

Tendances

Tendances (20)

En vedette

En vedette (12)

Similaire à Stat

Similaire à Stat (20)

Plus de Ajay Agade

Plus de Ajay Agade (20)

Stat