STATS lec 19- clinical stats

Question 1

Risk

Answer 1

• In clincal research risk is the probability of something
• P =n/N
• Where P is the risk (probability) n is how often it occurs
• N is the total population under study
• The risk of rolling a 6 is ??

Question 2

Odds

Answer 2

• Another way of expressing the likelihood of occurrence
• Odds = n/N-n
• The odds of rolling a 6 are

Question 3

Example- Excelin (high risk)

Answer 3

• Disease occurs (event rate) 40% in the control group e.g. 5 yrs death rate
• But only occurs at 30% with drug
• Absolute risk in control group (CER=0.4) = 40% without drug
• Intervention = drug
• Absolute risk in intervention group (IER=0.3) = 30% with drug

Question 4

The cure study

Answer 4

• Effects of clopidogrel in addition to aspirin in patients with acute coronary syndromes without ST-segment elevation
• Patients with ACS have major vascular events.. does giving them clopidogrel reduce the risk of these events
• from 11.4% to 9.3% over 12 months
• So CER = 0.114, IER=0.093

Question 5

Example 1- 2x2 contingency table

Answer 5

• Outcome is always on the top

Question 6

Example 2- Excelin (low risk)

Answer 6

• The Excelin trial is repeated in a more healthy population
• Rat eofdeath is 10% in the control group and 7.5% in the intervention group
• CER= 0.1 IER= 0.075
• CER = 10/100
• IER = 7.5/100

Question 7

Absolute risk- AR

Answer 7

• AR is how likely something will occur
• The probability that an individual will experience the specified outcome during a specified period
• Range 0 to 1 or %
• In contrast to common usage, the word risk may refer to adverse events or desirable events (MI or CURE)

Question 8

Absolute risk reduction (or increase)

Answer 8

• Control_risk - Intervention_risk
• # 1 (high risk) = 40% - 30% = 1-%
• # 2 (low risk)= 10%-7.5% = 2.5%
• ARR = CER - IER
• Can be misleading since it depends on the population characteristics, tells you more about disease in cohort as oppose to drug

Question 9

Relative risk (risk ratio)- RR

Answer 9

• The number of times more likely (RR >1) or less likely (RR<1) an event is to happen in one group compared with another
• It is the ratio of the absolute risk (AR) for each group
• Think of it as the proportional risk
• RR= IER/CER (for both groups RR=0.75)
• RR>1 means increased risk
• RR= 1 means no difference in risk
• RR <1 Means risk reduced

Question 10

Relative risk reduction

Answer 10

Control_risk - Intervention_risk /Control_risk

• RRR= 40-30/40 = 25% for population 1
• RRR = 10-7.5%/10%= 25% for population 2
• RRR Can be misleading since it doesn’t tell you how many patients will benefit

Question 11

absolute v relative risk

Answer 11

• Disadvantage of RRR
  
  • Doesn’t take into account baseline risk of population groups- end up an insignificant result appearing significant
• NB- Large difference between RRR & ARR only occurs when the outcome is rare
• Undue emphase on either RRR or ARR can be misleading- check both before deciding on real benefit of a drug

Question 12

Example 3- superal

Answer 12

• Drug superal has RR reduction for stroke of 33% but increased risk of severe gastric bleeding 3-fold
• The baseline risk of gastric bleeding in the general populaiton is 1%/year
• Who should we treat with superal
• What are the RR and AR for stroke in P1 and P2
• What are the RR and AR for adverse effects in P1 and P2
• Calc: CER, IER, ARR, RRR NNT, NNH

Question 13

Superal

Answer 13

• Superal has RRR for stroke of 33% but increased risk of severe gastric bleeding- 3 fold
• P1 Primary prevention
  
  • 3%/3yrs, down to 2%… Net effect down by 1%
  • Bleed risk is 3%/3yrs increased to 9%/3yrs, net effect up by 6%
  • Stroke down by 1%, bleed up by 6%
• P2 Secondary prevention
  
  • 30% risk/3yrs down to 20% net effect down by 10%
  • Stroke down by 10 %, bleed up by 6%

Question 14

Numbers needed to treat (NNT)

Answer 14

• A more useful way of thinking about effect
• NNT = 1/ARR
• What is the NNT for superal?
• What is the number needed to harm for superal (NNH= 1/ARR)
• NNT is the number of fo subjects who must be treated with the intervention, compared with the control for 1 addiotnal subject to experience the beneficial outcome

Question 15

Superal Numbers needed to

Answer 15

• P1-NNT = 100
• P1-NNH = 16
• P2-NNT= 10
• P2-NNH= 16
• Describe in whole numbers
• If necessary round NNT up; Round NNH down

Question 16

Odds ration- OR

Answer 16

• Is the ratio of
• Odds of outcome in intervention group … against …
• Odds of outcome in the control group
• Used in cross-sectional studies and case-control studies
• In a (non-real) study of 2,500 patients taking aspirin it was found that 42 had GI bleeding
• The number of people with GI bleeds in the general population is 1 in 100
• What are the odds ration for aspirin causing GI bleeds

Question 17

Odds ratios

Answer 17

• RRR, NNT preferred as more intuitive
• Still needed for
  
  • Meta-analyses (As event rates differ between populations)
  • Case-control studies- especially where population event rate unknown
  • Multiple regression

Question 18

Confidence Intervals- CI

Answer 18

• Accuracy of measurement
• CI decreases as sample size increases
• All measures (ARR, RRR, OR) should have confidence intervals
• If the confidence interval for an OR crosses 1, then a significant efect has not been found

Question 19

Ways to cheat on statistical tests

Answer 19

• Throw all data in the computer and report all results p<0.05 (data dredging)
• If the two groups are different and this benefits the intervention group, forget to adjust
• Ignore all drop-outs, only analyse subjects that complete treatment
• If outliers (unusal results) are messing up the results then get rid of them. On the other hand, if they make the results better then keep them
• If you gain significance early; stop the trial. If you almost gain significance, the extend the trial
• If the whole group isn’t significant, look for sub-groups that are