METHODOLOGY OF CLINICAL RESEARCH

1
METHODOLOGY OF CLINICAL RESEARCH

• Dr.Ramdas Pisharody

2
WHAT IS RESEARCH A planned activity leading
to generation of information that will help in
answering a specific question
3
RESEARCH DOMAINS The Iterative Loop
Efficacy
4
RESEARCH QUESTION Articulating or transcribing
the research idea to communicate the nature of
the study, the subjects involved, the exposure
and outcome of interest and describe what is
being assessed or compared.
5
ISSUES IN SELECTING A PROBLEM Rationale
- Scientific Research Feasible - Do-able
research Has it been done before ? If so, Is
it useful applicable to your patients
Generalizable Take a realistic piece of the
problem
6

• STEPS IN FORMULATING A QUESTION
• Identifying a relevant problem
• Formulating a hypothesis
• Specify the sample
• Specify the maneuver or exposure
• Specify the outcomes of interest
• Specify nature of comparison or estimates to be
  made
• Should suggest the appropriate design

7

• DESIGN ARCHITECTURE
• The framework in which investigation is
• planned and carried out.
• It is necessarily based on the type of research
• question
• Guided by the subjects under study, methods
• of investigations and the statistical
• principles
• An appropriate research design is a
• prerequisite for a valid study

8

• RESEARCH DESIGN
• Descriptive Studies
• Case studies and Case series
• Surveys
• Cross Sectional
• Longitudinal
• B. Analytical Studies
• Observational
• Experimental

9
Case Series

• Australian gynaecologist, Dr. McBride
  of Sydney, suspected that thalidomide was the
  cause of limb and bowel malformations in three
  children he had seen at Crown Street Women's
  Hospital. 1956

10

• Pneumocystis carinii pneumonia and mucosal
  candidiasis in previously healthy homosexual men
  evidence of a new acquired cellular
  immunodeficiency1981 Dec 10305(24)1425-31.
• MS Gottlieb, R Schroff, HM Schanker, JD Weisman,
  PT Fan, RA Wolf, and A Saxon

11

• SURVEYS
• Cross sectional survey
• a one time measurement conducted on a sample
  derived from a population
• Usually done to estimate prevalence and to
  describe
• disease spectrum.
• The prevalence thus computed is called point
  prevalence.
• If repeated, can estimate cumulative prevalence
• Prevalence No. of persons with disease
• Population at risk

12

• SURVEYS
• Longitudinal Survey Follow up study
• A group of subjects under surveillance over a
  period of time.
• Intended to measure new cases occur over a
  specified duration
• Denominator is the product of population at
  risk and the duration of
• observation.
• Longitudinal surveys are single group cohorts of
  people at risk of
• developing a disease.
•  
• Incidence rate Number of new cases occurring
  over the period
• Population at risk X time of
  observation

13

• REQUISITES OF A SURVEY
•  
• Study population to be clearly defined
• Sample if any should be appropriate and large
  enough to
• measure the disease and will depend on
  disease frequency.
• Pre-tested questionnaire to obtain all
  relevant information
• should be designed
• Criteria for diagnosis of disease should be
  laid down.

14

• OBSERVATIONAL STUDY DESIGN
•  
• Comparative studies in populations used to
  study prognosis,
• risk and causation
• Investigators observe events in a specified
  population and
• come to conclusions.
• Types
• - Case control study
• - Cohort study
• - Ecological study

15
Cassava is an aetiological agent in Tropical
pancreatitis (TP)

• Group of patients with TP
• ask for exposure to cassava
• Group of persons without TP
• ask for exposure to cassava
• Compare the two groups.

(cases)
(controls)
16
Case Control Studies
17

• ADVANTAGES
•    
•   Easier, cheaper and more feasible.
•    Useful to study rare diseases.
•    Most useful for diseases with
• prolonged exposure time.
•    Can look at multiple risk factors.

18

• LIMITATIONS
• Choice of controls sometimes difficult Eg.
  Cancer, CAD
• Polygenic diseases-Cluster effect difficult to
  choose
• appropriate controls.
• Cannot compute risk of disease.
• Temporality questionable.

19
(No Transcript)
20
Cohort Studies
21
Dialysis Outcomes and Practice Patterns
(DOPPS)

• Prospective cohort study
• 12 countries- 900 facilities
• 38,000 patients recruited since 1996
• Patient outcomes and factors influencing
• gt 75 publications

22

• ADVANTAGES
• Temporality as in real life.
• Useful for diseases with high prevalence
• When one exposure gives risk to multiple diseases
• When exposure is rare
• Measures absolute and relative risk of disease.

23

• LIMITATIONS
• Slow, time consuming, costly.
• Not useful in diseases with long latency.
• Difficult to assess independent effect of
  multiple risk
• factors. Eg. Framingham study
• .

24
Association or causation?
Ear lobe crease means increased heart disease
Will surgical excision help?
25
Bradford Hills Criteria of Causation

• Consistency
• Biologic plausibility
• Strength of association
• Temporal relationship
• Specificity
• Coherence
• Biologic gradient
• Experimental evidence

26
James Lind, 1753

• Twelve sailors with scurvy
• Pairs treated with
• cider
• elixir vitriol
• vinegar
• sea water
• 2 oranges lemon
• nutmeg
• Pair on citrus well after 1 week

27
MRC trial of Streptomycin in TB 1948

• Patients randomly allocated to streptomycin or
  placebo
• Results favoured streptomycin (p0.0001)
• bed rest only 4/52 cured
• bed rest streptomycin 28/55 cured

Bradford Hill
28
Randomized Controlled Trials
29
Blinding

• Participants
• Administering
• Following up
• Outcome measurement
• Analysis
• Writing up

• Open.
• Single blind
• Double blind
• Triple blind
• Quadruple blind

30
(No Transcript)
31
FACTORIAL DESIGN
32
Meta-Analysis
33
Evidence Pyramid
34
STUDY QUESTIONS AND APPROPRIATE DESIGNS
Type of Question Appropriate Study
Design Burden of illness - Prevalence
Cross Sectional Survey -
Incidence Longitudinal survey Causation,
Risk Prognosis Case Control Study,
Cohort study Occupational risk, Environmental
Risk Ecological studies Treatment
Efficacy Randomized Controlled
study Diagnostic Test Evaluation Randomized
Controlled study Cost Effectiveness
Randomized Controlled study
35
Misconception!

• For every research project there is one single
  best design to answer the question.

36
RESEARCH IDEA Role of high flux dialysis in
maintenance hematodialysis Patients RESEARCH
HYPOTHESIS Would patients using high flux
dialysers fare better than those using
conventional dialysers
37
RESEARCH QUESTION Would the QOL as measured by
KDQ be better by 25 in patients on thrice weekly
maintenance hemodialysis using high
flux dialysers than those using conventional
dialysers ?
Design a. Concurrent Randomized Controlled
study b. Cross-Over RCT
38
Registry

• Advantages
• Denominator will be sufficient Number
• We can make more subgroup analysis hypothesis
  testing
• More precise estimates (standard error will be
  less)

39
Registry

• Disadvantages
• More misclassification
• More systematic error
• Redundancy of data
• More missing values
• Accuracy of collected data may be poor
• Internal validity can be less

40
Methods
41
STATISTICAL INFERENCE
Observed state of affairs in sample
True state of affairs in population
Random variability Systematic Bias
42
CAVEAT Observed Measurement Truth Bias
Noise Statistical methods are useful to explain
Noise. Methodological flaws result in Bias
cannot be over come by Statistics Appropriate
methodology reduces or eliminates
Bias Appropriate Statistics reveals the Truth
embedded in the Noise
43
Sample and population

• Impossible to actually compute the mean value of
  any variable
• in the entire population.
• Hence means and variability obtained from sample
  is applied to
• the population
• If repeated studies are done on different
  samples drawn from
• the same population the distribution of the
  sample means also
• follow a normal distribution.
• The mean of all the sample means is assumed as
  population mean
• The dispersibility of the sample means is
  expressed as Standard Error

44
Variability and Distribution
Sample Population Measure of
Mean Mean Central tendency X
? Measure of Standard Deviation Standard
Error Dispersion s (? s / ?n
) Limits centiles Confidence quartiles
interval deciles
45
STATISTICAL INFERENCE
 
46
STATISTICAL INFERENCE
47
Type I Error and level of alpha
48
Steps in Statistical Inference

• Generating NULL and ALTERNATIVE hypothesis
• Determining the level of alfa
• Testing the hypothesis using appropriate
  statistical tests
• Obtaining p value
• Concluding from the p value.

49

Generating NULL and ALTERNATIVE hypothesis

• Null hypothesis states
• There is no real difference between the two
  populations in terms of the outcome measures
• Alternative hypothesis states
• There is a real difference between the two
  populations in terms of the outcome measures

50
STATISTICAL INFERENCE
51
Test statistic
Design Nature of Statistical
Statistic variable Test
derived Qualitative Chi square ?2
Two independent (nominal) groups
Quantitative Student t test t
(continuous) Qualitative Mc Nemars
?2M (nominal) Chi square Two related
groups Quantitative Paired t test
t (continuous)
52
Getting the p value   Each test statistic has
a sampling distribution from which p values for
the corresponding value of the statistic can be
read off from available tables.
53
Inference If the obtained p value is small,
i.e. (smaller than the level of ?) it is
statistically significant i.e. probability of
null hypothesis being true is small and hence it
is rejected.
54
Conclusion If results are statistically
significant, see if the observed differences
are clinically important and see if the
differences observed would be acceptable to
others
55
Conclusion If not statistically significant,
see if the sample size was adequate enough not
to have missed a clinically important
difference In other words see if the probability
of the Type II error (?) is within acceptable
limits. The complement of ? i.e. 1 - ? is
called the power of the study which tells you the
strength with which you concluded that there is
no difference between the two groups.
56
Confidence interval
1.96 CI tells you that .. If the study is
repeated several times the values of the sample
mean would fall within this range 95 out of 100
times.
57
SUMMARY

• Research is a pre-planned, logical sequence of
  actions leading to conclusions that are not far
  from the truth.
• Minimisation of systematic error involves
  several strategies in selection of study
  subjects, proper measurement of outcomes etc.

58
Summary (contd..)

• Validity and precision are the key objectives of
  a good study.
• Generalizability or External Validity relates
  to acceptability of study results to similar
  patient groups in different settings.
• Statistical tests measure the random error and
  quantify the precision of a study

59

• If you want to know the taste of a pear, you
  must change the pear by eating it yourself. . . .
  
• If you want to know the theory and methods of
  revolution, you must take part in revolution.
• All genuine knowledge originates in direct
  experience
• Mao Tse-Tung