Meta-analysis. Systematic review.

The concept of clinical

hypotheses: types and formulation algorithm.
Clinical guidelines and recommendations.

CREATED BY OLES VAKAROV

Evidence-based medicine is
an appropriate sequential
chain (algorithm) of actions
from the formulation of a
clinical problem to its
solution
Formulation of a clear clinical problem, based on the
specifics of the patient.

Search for literature on the problem.

Evaluation of research for their validity and usefulness.

Application of useful findings in clinical practice.

Meta-analysis

Traditional descriptive analysis of

medical information often leads
to distortion of information. An
alternative approach, which is
becoming increasingly popular,
is a systematic analysis with
statistical generalization of data
(meta-analysis)
 cumulative meta-analysis

Types of
meta- prospective meta-analysis
analysis:

meta-analysis of individual
data
 Cumulative
meta-analysis
ALLOWS YOU TO DEMONSTRATE
THE CURVE OF ACCUMULATION
OF ESTIMATES IN THE PROCESS
OF NEW EVIDENCE
Prospective
useful not only for summarizing previous clinical
trials, but also for those trials that are planned to
be performed in the future. This approach can be
meta- implemented where a network of information
exchange and joint programs already exists. In
analysis practice, instead of prospective, hybrid
prospective-retrospective meta-analysis is often
used, which combines the obtained results with
previously published ones. This allows the use of
general evidence for individuals in each of the
group trials, although they may differ in better
planning
Meta-analysis of individual data

u Are based on the study of the

results of treatment of individual
patients, in contrast to the
literature.
u This approach is only available to
participants in a clinical trial
network who have access to a
medical history
Stages of statistical research

drawing up a plan and developing a

research program

data registration

data development and compilation

statistical analysis
Drawing up a plan and developing a
research program

1 2 3 4 5 6 7
Select the Select Select object Select Select Select Select
purpose objectives of of observation organizationa research research
the study observation unit l elements methods program
(types)
 Formulate a
hypothesis

u null hypothesis (there

is not difference
between the group)

u alternative hypothesis
(there is a difference
between the groups)
Data registration

choose a method and

design
Data development
and compilation

collect the data

Statistical analysis
(analyse the data and draw conclusion)
interpretation and comparison of data

formulation of conclusions

literary design of results

proposals for implementation in practice

Statistical tests work by
calculating test statistics
A number that describes how the relationship between the variables in your
test differs from the null hypothesis of no relationship

It then calculates the p-value (probability value). The value of p estimates how
likely it is that you will see the difference described by the test statistics if the
null hypothesis of no connection were true

If the value of the test statistics is more extreme than the statistics calculated
by the null hypothesis, then we can conclude that there is a statistically
significant relationship between the variables of the forecast and the results

If the value of the test statistic is less extreme than the value calculated from
the null hypothesis, then you cannot conclude that there is a statistically
significant relationship between the forecast variables and the results
The data checked in
statistical tests allow to
make assumptions
Variants of assumptions based on
data

u Independence of observations (multiple measurements of one

subject are not independent, while measurements of several
different subjects are independent)
u Homogeneity of variance (if one group has much more variation
than others, it will limit the effectiveness of the test)
u Data normality (data follow a normal distribution – only to
quantitative data)
Types of variables

Quantitative Categorical
variables - variables -
represent the represent the
number of things grouping of things
 Continuous (also known as
variable ratios) - represent
measures and can usually be
divided into units of less than one
Quantitative
variables Discrete (also integer variables) -
represent the number and usually
cannot be divided into units less
than one
 Sequential: submit data in
order ( ratings).

Categorical Nominal: represent group

variables names

Binary: represents data with

a result of yes / no or 1/0
Level of measurement

Parametric tests - to draw NON-Parametric tests – to

stronger conclusions from draw are not as strong, do
the data, they can be not make as many
performed only with data assumptions about the data
that meet the general and are useful when one or
assumptions of statistical more general statistical
tests assumptions are violated
Parametric tests

Interval – Results placed along a regular scale, with an arbitrary zero-point

Multiplication/ Divishion is meaningless but the ratio of differences is valid as
differences are consistent

Ratio – Results placed along a regular scale with an absolute zero-point

Allows for statements such as 6 units is twice as much as 3 units
The most common types of tests

Regression Comparative Correlation

tests tests tests
Regression tests

They look for the

effect of one or
Used to check
more continuous
causation
variables on
another variable
 Simple linear regression uses
one independent variable to
explain or predict the outcome
of the dependent variable
Types of
regression Multiple linear regression
uses two or more independent
variables to predict the
outcome
Comparative tests

They can be used to

test the effect of a
Look for differences categorical variable
among groups on the average of
some other
characteristic
 T-tests are used when comparing the
means of two groups (for example, the
average height of men and women).
Comparative
tests The ANOVA and MANOVA tests are
used when comparing the
performance of more than two groups
(for example, the average height of
children, adolescents and adults)
Correlation tests

They can be used to

Check whether two check the automatic
variables are related, correlation of two
without suggesting variables that you want
causal relationships to use in (for example) a
multiple regression test
Correlation tests

Pearson’s r test

The correlation coefficient (r) tells you the strength of a

linear relationship between two quantitative variables
 Nominal - labelled result or characteristics
- can be used for equivalence and set
membership
NON- - can show commonality but leaves it
undefined
Parametric
tests Ordinal – results in some sort of rank order
- results do not show differences between
place!
- No arithmetic is valid, you cannot say 3rd
place is the same as 1st and 2nd combined.
 Mann-Whitney U Test (is a nonparametric version of the
independent samples t-test. The test primarily deals with
two independent samples that contain ordinal data)

Wilcoxon Signed Rank Test (counterpart of the paired

samples t-test. The test compares two dependent
NON- samples with ordinal data)

Parametric The Kruskal-Wallis Test (is a nonparametric alternative to

tests the one-way ANOVA, is used to compare more than two
independent groups with ordinal data)

Chi square test of independence (checks whether two

variables are likely to be related or not)
Methods of showing results:

Tables and graphs

Pie charts and component band charts
Spot maps and rate maps
Bar charts
Line graphs
Frequency distributions and histograms
Normal distributions
 Advantages of meta-analysis

u allows a scientifically sound and reproducible way to summarize

information obtained from different sources
u it can combine studies whose data are statistically inaccurate, which will
provide a reliable total result
u at generalization the heterogeneity of results which studying of the
reasons allows to reveal other clinical problems can be shown
u during the meta-analysis, the authors constantly communicate with their
research colleagues to clarify certain aspects of their published
messages or to search for other research
u information networks are formed, which in the future will facilitate
individual and long-term meta-analysis
Systematic
reviews
These are scientific and
analytical researches, the
material for which are the
results of previously
conducted and already
published original trials.
The main provisions of the
systematic review
Defining the main goal

Defining methods for evaluating results

Systematic information search

Generalization of quantitative information

Generalization of evidence using appropriate statistical methods

Interpretation of results
Algorithm for creating systematic
reviews
Formulation of the Summary of the main
purpose of the characteristics of all
Search for studies Exclusion of studies
review and identified studies and
development of that meet the assessment of the that do not meet the
inclusion criteria selection criteria
criteria for inclusion quality of their
in it methodology

Collection of a
complete set of Analysis of the results
Sensitivity analysis in
necessary data, if of selected studies Preparation of a
general and in
possible, obtaining (use of statistical structured report of
subgroups - if
them from the synthesis) (meta- the review
possible
authors of the analysis) - if possible
original publications

Preparation of
Description of all
conclusions,
possible limitations
recommendations Preparation of a
and discrepancies in
for medical practice structured abstract
the existing
and subsequent
database
research
Systematic review

u Answers a clearly formulated clinical question

u It is based on the search results of all sources of
information in different languages
u Analyzes the reliability of research results by assessing
the reliability of methods for collecting and
processing clinical information
u Summarizes only benign data
u It is regularly updated as the results of new clinical
trials appear
The advantages of systematic
reviews are:
• the presence of a total reliable result;
• the ability to identify problems of individual sensitivity;
• promoting the formation of integrated information networks
and expanding professional communication;
• evaluation (critical approach) of research on the validity and
benefits;

• application of useful findings in clinical practice.

Clinical recommendations/
clinical practice guidelines

These are provisions for medical practice that are systematically developed and make
it easier for physicians and patients to make optimal decisions in any particular situation

Are one of the ways to summarize the available data and evaluate objective
information to create modern technologies for diagnosis, prevention and treatment, as
well as to improve the health system, increase the professionalism of health
professionals
Clinical recommendations are
based on the results of clinical trials
of drugs or methods of treatment of
relevant diseases, especially the
results of randomized controlled
clinical trials
Clinical recommendations are
created according to the appropriate
methodology, which guarantees their
modernity, reliability, complete
generalization of the best world
experience and knowledge, the
possibility of application in clinical
practice, ease of use.
To comply with the requirements of good
faith clinical guidelines, they must provide:

u Maximum openness and transparency of the development process

u Reconciling the interests and capabilities of consumers and buyers of medical
services, as well as those who provide them
u Correspondence of the recommended medical-diagnostic and preventive
methods and technologies to a modern condition of medical science and
practice
u Objectivity and reliability of information on the effectiveness of clinical
technologies used
u Choosing the most cost-effective medical technology
Clinical recommendations

A - high reliability - based on hierarchy I

evidence

B - moderate reliability - based on

hierarchy II evidence

C - limited reliability - based on hierarchy III

evidence

D - a minor reliability - based on hierarchy

IV