Statistics

Question 1

Define Confidence Interval

Answer 1

We are 95% confident that the true odds ratio/relative risk lies in our given range. If we were to repeat this test indefinitely, 95% of odds ratios/relative risk will lie between the CI

Question 2

Define Odds Ratio

Answer 2

The ratio of the odds of the intervention compared with the odds of the control

OR = AD/BC

Question 3

Define Relative Risk Ratio

Answer 3

The ratio of the risk of the intervention compared with the risk of the control

RRR = A(C+D)/(C(A+B))

Question 4

Define p-value

Answer 4

This measures how likely it is that any observed difference between groups is due to chance.

In other words, the P value is the probability of seeing the observed results just by chance if there is genuinely no difference between the groups (null hypothesis is true).

A P-value of 0 indicates that the difference is not due to chance (the groups are different) and a value of 1 indicates that groups are the same (but there may be random variation)

Question 5

Likelihood Ratios

Answer 5

LR+ = (Prob + test in diseased) / (Prob + test in nondiseased)
LR+ = SEN / (1 - SPE)

LR- = (Prob - test in diseased) / (Prob - test in nondiseased)
LR- = (1 - SEN) / SPE

Tests where the likelihood ratio lie close to 1 have little practical significance as the post-test probability (odds) is little different from the pre-test probability.

• A likelihood ratio of < 1 indicates the result is associated with absence of the disease
• A likelihood ratio of > 1 indicates the result is associated with the disease

Question 6

Types of Evidence Based Studies

Answer 6

• Ecologic Study
• Case Study (individual)
• Case Report (several individuals)
• Cross-Sectional
• Case-Control Study
• Cohort Control
• Controlled Clinical Trial
• • Intervention
• Randomised Controlled Trial
• • Intervention
• Systemic Reviews

Question 7

Define Ecologic Study

Answer 7

Concerning populations, not individuals

Question 8

Define Case Study (individuals)

Answer 8

Individuals, profile of subjects for rare conditions

Question 9

Define Case Report (several individuals)

Answer 9

Profile of several subjects

Question 10

Define Cross-Sectional

Answer 10

Measure exposure AND disease at one point.

Opens up the possibility of the ecological fallacy, where group characteristics are inferences onto individuals. For example, most Australian’s are bogans. Therefore John, an Australian, is a bogan.

Question 11

Define Case-Control Study

Answer 11

Outcome first then exposure.

Unfortunately the temporal relationship is unclear

Question 12

Define Cohort Study

Answer 12

Exposure first then outcome.

The temporal relationship is clear

Question 13

Systemic Review Types and Differences

Answer 13

Narrative vs Systemic

Narrative

• Lacks rigor
• Broad in scope
• Limited critical appraisal
• Broad summary of state of affairs

Systemic

• Rigorous-minimise bias
• Answers a specific PICO question
• Protocol
• Search strategy documented
• Quality of studies appraised according to set criteria

Question 14

Systemic Review Bias Source

Answer 14

• Inclusion criteria
  
  • Inadequate literature search
  • Publication bias
  • Inadequate assessment of study quality

Question 15

Systemic Review Outcome Measurements

Answer 15

• Dichotomous outcomes
  * RR, AR, OR and NNT
  • Continuous outcomes
    
    • Weighted mean difference (outcomes on the same scale) or standardised mean difference (outcomes are conceptually the same, but on a different scale)

Question 16

Define Statistical Heterogeneity and Homogeneity

Answer 16

Identifies whether individual studies within a meta-analysis are similar enough to validate a meaningful interpretation from the meta-analysis.

• Statistical heterogeneity
  
  • Exists when there is greater variation between results in the trial than is compatible with chance
• Statistical homogeneity (good sign)
  
  • Where the studies are similar enough for us to be able to combine them

Question 17

Measures of Statistical Heterogeneity

Answer 17

• Chi-squared test
  * Used to assess statistical heterogeneity
  * If chi-squared > degrees of freedom -> evidence for heterogeneity (bad sign)
  • I-squared test
    
    • 70-100% indicates heterogeneity
    • < 70% indicates homogeneity

Question 18

Further Systemic Review Analyses

Answer 18

• Sensitivity analyses
  * Examine how the results vary under different assumptions
  * E.g. Re-analysing data using low, medium and high studies
  • Subgroup analyses
    
    • Meta-analyses on subgroups of the studies
    • E.g. Sex, age and drug doses

Question 19

Difference between Testing and Screening

Answer 19

• Testing
  * Testing in the presence of symptoms or to confirm the presence/absence of a disease
  • Screening
    
    • Testing in the absence of symptoms
      
      • Opportunistic (i.e. PSA test at a general check-up)
      • Selective (i.e. Mammography in women 50-69)
      • Mass screening (neonatal screening)

Question 20

Primary vs Secondary vs Tertiary Prevention

Answer 20

Primary Prevention
To prevent new diseases developing

Secondary Prevention
To find new diseases at an early stage

Tertiary Prevention
To minimise the effects of established disease(s)

Question 21

Benefits of Screening

Answer 21

• Early detection of the disease
  
  • Early treatment of the disease
  • Psychological well being

Question 22

Limitations of Screening

Answer 22

• Over-diagnosis (over treat)
  * The identification of slow growing cancers that may never have become apparent, ‘false positive’
  • Cannot detect interval disease
    
    • Interval cancers occur between screening as they are found in the time interval between screens. These can be overcome by
      
      • Increasing screening frequency (but this further increases over diagnosis)
  • False negative/positives
  • Side effects

Question 23

Screening Biases

Answer 23

• Screening bias
  
  • Length-time bias
  • Lead-time bias

Question 24

Define Screening Bias

Answer 24

Participants in screening programmes tend to be healthier than those who don’t volunteer or comply

Question 25

Define Length-Time Bias

Answer 25

There may be different patterns of growth in the same disease. Screening is more likely to find slow growing tumours than rapidly growing ones.

Question 26

Define Lead-Time Bias

Answer 26

Lead time is the length of time between the detection of a disease and its usual clinical presentation. Lead-time bias e of the disease - it may appear to prolong survival but instead it just diagnoses earlier.