Step3 7 Statistic

Question 1

Number needed to treat

Answer 1

Need to assess Absolute Risk Reduction first

ARR = Group not treated - Group treated (in decimals)

NNT = 1/ARR

Question 2

How do you interpret the Relative Risk

Answer 2

The risk compare from the numerator to the denominator A/B (usually exposed to unexposed)

< 1 means a decreased risk in group A
> 1 means increased risk in group A

Confidence interval must not include 1

A RR of 0.4 means that the risk of A is 40% of B

Question 3

Benefits of Intention to treat analysis

Answer 3

Keeps randomization and reduce bias (due to cross over)

Patients are analyzed exclusively within the group they started the trial

Question 4

Canaglifozin

Answer 4

SGLT2 inhibitor: inhibit glucose reabsorption

Increase risk of UTI, candidiasis, hypoglicemia

Question 5

Likelihood ratio of a test

Answer 5

+LR: sensitivity / 1 - specificity
-LR: 1-sensitivity / specificity

Interpretation:
>10: Strong evidence to rule in disease
5-10: Moderate...
2-5: Weak...
0.5-2: No evidence to rule in or out
0.2-0.5: Weak 
0.1-0.2: Moderate
<0.1: Strong evidence to rule OUT disease

Question 6

Size of sample needed in each case

Magnitude of effect
P-value
Power
Standard deviation

Answer 6

The magnitude of effect:
The Bigger the magnitude of the effect, the smaller the sample you will need

P-value:
If you set a P value of 0.5 vs 0.1, you will need a bigger sample for the latter

Power:
The bigger the power you want your study, the bigger the sample you will need

Standard deviation:
If SD is smaller than expected, it will be easier to detect difference, smaller sample will work

Question 7

Validity vs. Reliability

Answer 7

Validity (accuracy) is the ability of a test to provide correct results

Reliability: the ability of a test to always reproduce the same result. Doesn’t matter if it is right or wrong

Question 8

What is the kappa statistic

Answer 8

Kappa statistics measure inter-rater reliability, AKA inter-rater concordance

Value -1 to +1
0= results due to chance alone
<0= disagreement
>0= agreement

0-0.2: negligible

0. 21-0.4: minimal
1. 41-0.6: fair
2. 61-0.8: good
3. 81-1: excellent

Question 9

How do you adjust for confounders in a clinical study?

Answer 9

Use a multiple regression analysis

Question 10

How do people that do not respond to questionnaire in a study, can affect the study

Answer 10

They can differ in a big way from respondants

Question 11

Test property that it is not modified by incidence

Answer 11

Likelihood ratio

Sensitivity and specificity

Question 12

Observer bias

Definition
Fix

Answer 12

Occurs when the observer collecting data is influenced by knowledge of patient hx or something else

FIX:
Double-blind

Question 13

Selection bias

Answer 13

Occurs in the manner participants are selected or lost during the trial
Susceptibility bias is a type of selection bias

FIX:
Randomization