Designing Clinical Research and Interpreting Clinical Evidence

1
Designing Clinical Research and Interpreting
Clinical Evidence

• Advances in Maternal and Neonatal Health

2
Objective of Using Clinical Evidence

• To promote care based on best available evidence
• To encourage practitioners to develop their
  skills in obtaining, evaluating and incorporating
  evidence into daily clinical practice
• In order to achieve that, practitioners should
  understand and intellectually evaluate new
  clinical data as it becomes available

3
In an Ideal World

• The most effective care for every condition would
  be known
• Every clinician would know the most effective
  care for every patient
• Every clinician would practice the most effective
  care that she/he knows

4
In the Real World

• Much of what should be known is not known
• Much that is known, is not known by most
  clinicians
• Clinicians often fail to practice what they know
  to be the most effective form of care

5

• Evidence-based medicine is the systematic,
  scientific and explicit use of current best
  evidence in making decisions about the care of
  individual patients.

6
Levels of Evidence and Grades of Recommendations

7
Conducting Research Steps

• Formulate research question
• Design study
• Choose subjects
• Choose variables to be measured
• Collect data
• Analyze data
• Draw conclusions

8
Definitions

• Quality of test
• Sensitivity Likelihood that diagnostic test will
  indicate presence of disease when disease is
  actually present (true positive rate)
• Specificity Likelihood that diagnostic test will
  indicate absence of disease when disease is
  actually absent
• Usefulness of result
• Positive predictive value Likelihood that
  positive test result actually means that disease
  is present
• Negative predictive value Likelihood that
  negative test result actually mean that disease
  is absent

9
Prototype 2 X 2 Table

• Example Doing a culture for Group B
  streptococcus (GBS)
• Sensitivity Chance that test will indicate GBS
  when woman has it
• Specificity Chance that test will indicate no
  GBS when woman does not have it
• Positive predictive value Chance that positive
  culture GBS colonization
• Negative predictive value Chance that negative
  culture no GBS

PPV
a b c d
- NPV
Spec.
Sens.
10
2 X 2 Table (continued)
Sensitivity 67 Specificity 52 PPV 20 NPV
90
11
Meta Analysis

• Is one tool that may allow useful information to
  be obtained from multiple studies
• Is systematic evaluation of collection of several
  studies which are similar in design, study
  populations and outcomes examined
• Combines data appropriately to find answer to
  important question in cumulative information in
  literature
• Is systematic review of medical literature

12
Measures of Statistical Significance Relative
Risk

• Rate of risk of outcome in exposed individuals to
  risk in unexposed individuals
• From cohort study only
• Reflects risk to total population
• Relative risk
• 1 No difference in outcome between 2 groups
• lt 1 Less risk of outcome
• gt 1 Higher risk of outcome

13
Measures of Statistical Significance Odds Ratio

• Compares likelihood of outcome being studied
  occurring in group receiving intervention
  (experimental) with group not receiving (control)
• From case control
• Does not reflect total population, but can
  closely estimate relative risk
• Graphically represented on logarithmic scale
• Vertical line at 1 No difference in outcome
  between two groups
• Ratios less than 1 (left of vertical line)
  Improvement of outcome
• Ratios more than 1 Worse outcome

14
Measures of Statistical Significance 95
Confidence Interval

• Confidence interval Range in which true effect
  size can be found
• 95 chance that true effect size lies within 95
  confidence interval
• If confidence interval overlaps 1.0, then there
  is a gt 5 possibility that observed outcome
  difference is due to chance
• Very wide results less believable
• Very narrow more believable

15
Reduced Risk
Increased Risk
Confidence Interval
Results consistent with chance
0.01
0.1
1
10
100
Odds Ratio
No Difference
16
Measures of Statistical Significance p Value

• The probability, under a null hypothesis, of data
  as extreme or more extreme than was observed in
  one study
• Conventionally set as 0.05
• Equivalent to 5
• Difference is significant if p value is less than
  0.05 (lt 0.05)
• Means that there is less than a 5 chance that
  the result obtained is due to chance, or
• 95 certain that result obtained by the
  intervention is true

17
Statistical Significance vs. Clinical Significance

• It is up to the practitioner to decide whether a
  statistically significant or non-significant
  result is clinically significant

18
Summary

• Evidence-based medicine should be used to set a
  standard of care
• New data should be evaluated critically to
  determine whether to change standards

19
References

• Hulley SB and SR Cummings. 1988. Designing
  Clinical Research An epidemiological approach.
  Williams and Wilkins Baltimore, Maryland.