An Introduction to Clinical Epidemiology and Research

1
An Introduction to Clinical Epidemiology and
Research Methodology

• F. Farrokhyar, MPhil, PhD, PDoc
• Department of Surgery
• Department of Clinical Epidemiology and
  Biostatistics

2
Evidence-based Medicine

• Explicit and judicious use of the best current
  evidence in making decisions about the care of
  the individual patient.
• EBM and patient-derived outcomes assessment
  movements has led to an increased focus on the
  effectiveness of clinical care.
• Clinical epidemiology provides the methodology
  with which we assess this effectiveness.

3
Pyramid of Evidence
4
Clinical Research

• Experimental
• RCTs
• Non-RCTs
• Observational
• Analytical
• Cohort studies
• Case-control studies
• Descriptive studies no control group
• Case-series
• Longitudinal studies

5
Clinical Research . . .

• If treatment effect is very large, this is likely
  to be identified reliably in an observational
  study, if they are properly designed to control
  biases.
• However, most interventions and procedures have
  moderate or small treatment effect and these
  biases would matter considerably.

6
What is a case series?
Exposure

• A case-series is a common way of describing the
  clinical picture of a disease.
• Case-series provide the least quality clinical
  evidence

7
Research question

• How many patients with prostate cancer will have
  urinary tract infection after prostate resection
  surgery?

8
What is a case series?
21
140
21/140 0.15 15
Major Limitations - no control group -
retrospective
9
Research question

• Does the use of pre-op antibiotic decrease the
  urinary tract infection after prostate resection
  surgery?

10
Study of Cause-and-Effect Relationship
11
What is a case-control study?
95 CI
12
Case-control studies

• Advantages
• Feasible for rare conditions
• Smaller sample size
• Can assess outcome with many exposures
• Overcome temporal delays
• Cheap

• Disadvantages
• Retrospective
• Potential threats to internal validity
• Selection bias
• Measurement bias
• Data extraction bias
• Missing data
• Recall bias

13
To control biases in case-control studies?

• At the stage of study design to control for
  selection bias
• Restriction
• Matching
• At the stage of data analysis
• Multivariable analysis
• Stratified analysis

14
Study of Cause-and-Effect Relationship
15
What is a cohort?
16
What is a cohort study?
Sample
Does the use of pre-op antibiotic decrease the
urinary tract infection after prostate resection
surgery?
17
What is a cohort study?
Sample
215 185
400
Odds ratios? (0.40)
18
Cohort studies

• Advantages
• Prospective
• Estimate incidence
• Feasible when randomization not possible
• Can assess the relationship between exposure/
  intervention and many outcomes

• Disadvantages
• Not feasible for rare diseases
• Expensive
• Results not available for a long time.
• Potential threats to internal validity
• Selection bias
• Performance bias
• Detection/ascertainment bias
• Measurement bias
• Attrition bias

19
What are the potential threats to internal
validity?

• Selection bias systematic differences in
  comparison groups.
• Measurement bias systematic differences in
  measuring variables
• Performance bias systematic differences in care
  provided apart form the intervention being
  evaluated.
• Detection - systematic differences in outcome
  assessment
• Attrition bias systematic differences in
  withdrawals from the study.

20
To control biases in cohort studies?

• At the stage of study design
• Restriction
• Matching
• At the stage of data analysis
• Propensity score analysis
• Multivariable analysis
• Stratified analysis

21
(No Transcript)
22
Study of Cause-and-Effect Relationship
23
What is Randomization?
24
Why Randomization is Important?

• The process of randomization provides comparable
  groups for most factors so that the differences
  in the outcome at the end of the trial can be
  attributed to the intervention alone.
• Consequently, preventing biased assignment
  permits a more definitive interpretation of the
  trials results.

25
What is an RCT?
Clinical uncertainty? Clinical equipoise?
Calculation RR, RRR, OR, RD, NNT/ NNH
26
What are the potential sources of bias in RCTs?

• Concealment
• Blinding care providers
• Blinding outcome assessors
• Minimize loss to follow-ups
• . blind patients, investigators, data entry
  personnel, data analyst

• Selection bias
• Performance bias
• Detection/ascertainment bias
• Attrition bias

27
In a well-designed RCT

• Inclusion and exclusion criteria are predescribed
  and identical between the treatment groups.
• All patients have the equal chance of receiving
  either treatment.
• Known and unknown confounders are similar between
  the treatment groups. Any differences would be by
  chance.
• Treatments are concurrent, avoiding temporal
  trends.
• Data collection is prospective, concurrent, high
  quality and eliminates the differences in
  definitions and types of variables.
• Assumptions underlying the statistical comparison
  tests are met.

28
Types of RCT in Surgery?

• Three types of RCTs are commonly described
  surgical
• Type I comparison of medical treatments in
  surgical patients traditional clinical trials
• Type II comparison of a surgical technique with
  a medical treatment or no treatment
• Type III comparison of two surgical techniques

29
Surgical trials vs. clinical
trials
30
What are the Challenges of Surgical Trials?

• Clinical trials do not risk any differential
  skills in administering the intervention.
• A major difference between clinical trials and
  surgical trials is that surgical trials require
  skills and training to administer the
  intervention.
• Surgery is a skilled and multi-step process and
  this makes the design of surgical trials
  challenging for the following reasons.

31
Ethical issues

• Helsinki Declaration all study subjects have
  the right to withdraw from a study at any time,
  without prejudice.
• This is hardly possible in a surgical trial.
• Sham surgery? is it ethical?
• In drug trials, most likely the drug company is
  responsible in case of an adverse event or a bad
  outcome.
• In surgical trials, surgeons, personally and
  legally, are responsible for their patients.

32
Learning curve
33
Why learning curve?

• It takes training and experience to develop
  expertise in a surgical technique.
• Surgeons tend to predominantly use a single
  surgical approach to treat a specific problem.
• Even for a fully trained surgeon, there is a
  learning process to become an expert with a new
  technique.
• If the participating surgeon is performing both
  new and standard techniques and has limited
  experiences with the new one, the results will
  be biased towards the standard technique.

differential bias
34
Expertise-based RCTs

• Patients are randomized to surgeon who is an
  expert in that intervention.
• Advantages
• Surgeons will perform only the procedure in which
  they have expertise
• Procedural crossovers are less likely to occur
  because surgeons are doing the procedure they are
  most comfortable with.
• It makes it easier to obtain informed consent and
  recruit patients into the trial

35
Surgeons Skill Variation
36
Surgeons Skill Variation

• There is an inherent variation in performing a
  specific procedure by different experts.
• There is also the effects of trainees, fellows
  and others in the operating room
• It requires more sophisticated statistical
  analysis to adjust for the effect of surgeons
  skill variation in multicenter surgical trials.

37
(No Transcript)
38
Blinding

• Clinical trials are traditionally double blinded
  patients and investigators are blinded to the
  type of treatment
• In surgical trials
• Surgeons cannot be blinded.
• It is hard and unethical not to tell patients.
• sham surgery has been used in the evaluation
  of fetal tissue transplants in Parkinsons
  disease
• Blinded trials are not possible when comparing
  surgical and non-surgical treatments.
• Outcome assessors and data analyst could be
  blinded

39
Outcome assessment

• Patients knowledge of the procedure often
  influences their behaviour Hawthorne effect
• Investigators knowledge of the procedure
  influences their assessment of the outcome.
• Co-intervention patients often require extra
  treatment such as physiotherapy or rehabilitation
  in orthopaedic surgery.

40
Follow up time

• It is hard to determine the appropriate duration
  of follow up in surgical trials.
• The loss to follow up is higher when the duration
  of follow up is very long.
• The loss to follow up is higher when no treatment
  is required after surgery.

41
Different types of RCTs

• Equality trials
• Superiority trials
• Non-Inferiority trials
• Equivalence trials
• Cluster randomized trials
• Cross-over trials

42
Study of Cause-and-Effect Relationship
43
In summary,

• Randomized controlled trials should have a very
  important role in the evaluation of surgical
  interventions. If properly designed, conducted,
  and interpreted the results are likely to make a
  substantial impact on the health of patients.

44
However,

• In those situations when it is deemed infeasible
  to conduct a randomized controlled trial, we need
  to choose alternatives such as inception cohort
  or case-control studies that may be less rigorous
  than the randomized controlled trial but more
  rigorous than the uncontrolled case-series.

45
Message to take home

• In designing a research project consider
• Frequency of the disease
• Treatment effect
• Follow-up time
• and
• Ethics

Pilot study
46
(No Transcript)