LC Exam 2: Breast Cancer Screening and Stats Flashcards
Requirement of successful screening programs
Advances time to diagnosis for cancers with consequences
Early treatment is beneficial compared to after sx onset
Lead time bias
Length time bias
Overdx bias
Lead time: find earlier, live longer with dz (appear to live longer)
Length time: slower growing -> better prognosis
Overdx: benign natural hx -> better prognosis (wasn’t going to cause trouble in first place)
Sensitivity and specificity
Disease No disease
+ Test True + False +
- Test False - True -
Sen: TP/(TP+FN) – Disease column
Spe: TN/(TN+FP) – No disease column
True always on top
PPV and NPV
Disease No disease
+ Test True + False +
- Test False - True -
PPV: TP/(TP+FP) = Chance I have dz when test is +
NPV: TN/(TN+FN) = Chance I don’t have dx when test is -
SPIN and SNOUT
SPecific tests rule IN (less false positives)
SeNsitive tests rule OUT (higher false positives)
More sensitive test has higher incidence of:
False positives or false negatives?
More false positives
As prevalence increases what happens to predictive value?
PPV increases as prevalence increases
NPV decreases
(Therefore false negatives more common)
Likelyhood ratio
Moves from pretest to posttest probability
Prob of test result in person with disease/
Prob of same test result in person without disease
LR = 1 no impact on likelihood of disease
LR >1 more likely
LR<1 less likely
LR <0.1 = great at ruling out
LR >10 = great at ruling in
LR calculation
LR+ = sen/1-spe (positive test result)
LR- = inverse (negative test result)
What is relative risk reduction?
Reduction of risk attributable to an intervention compared to a control
RRR=1-RR
RR = [a/(a+b)]/[c/(c+d)]
When does OR = RR?
When prevalence is LOW OR=RR
What is the absolute risk reduction?
Difference in risk (not proportion) between intervention and control
aka: difference in risk between groups:
ARR= Group1%-Group2% or
c/(c+d) - a/(a+b)
NNT
Number of patients that need to be treated or screened for one patient to benefit
NNT = 1/ARR
NNS (screen) variation with prevalence?
As prevalence increases, NNS goes down
Easier to find
Guideline for screening breast MRI?
Guideline for prophylactic tamoxifan?
lifetime risk of developing breast cancer is >20%
5 yr risk is >3%