Biostatistics Flashcards
What are 3 principles of clinical trial design?
Physician-Patient relationship must exist, Physician must act in patient’s best interest, Written consent must be obtained from patient after full disclosure of clinical trial
When should median be used in place of mean?
When data is highly-skewed
Relationship between mean, median, mode in setting of negative skew?
Mean < Median < Mode
Relationship between mean, median, mode in setting of positive skew?
Mode < Median < Mean
Benefit of pooling data from several different studies in a meta-analysis?
Increases power of study … by increasing N
Describe the difference between effect modification and confounding?
Effect modification = SIGNIFICANT difference between strata; Confounding = NO significant difference between strata … (in confounding, the significance disappears)
Equation for Positive Predictive Value (PPV)?
A/(A+B) … TP/(TP+FP)
Definition of PPV?
Probability that patient actually has a disease, given a (+) test result
Observer bias typically occurs in which setting?
Absence of blinding
Observer bias is particularly likely when the studied outcome is …
Qualitative
Evaluation of multiple endpoints during a study increases the risk of …
Type 1 error
Describe a Type 1 error?
Reject the null, when the null was correct … (say there’s a difference, when there really wasn’t)
___ refers to increased likelihood of making a Type 1error when evaluating multiple end points?
Multiplicity
Define net clinical benefit?
Measure of a medication’s possible benefit, minus its possible harm
What is the most important aspect of a study, according to the Intention To Treat (ITT) principle?
Randomization
Intention To Treat (ITT) principle attempts to avoid the effects of …
Crossover, Dropout
What is the goal of Phase 4 clinical trials?
Post-marketing surveillance to monitor safety of medications that have already been released on market
Why might some clinical trials fail to detect serious side effects of new medications before Phase 4?
Inadequate power
Calculations for Positive Predictive Value (PPV) requires knowledge of …
Prevalence
___ is an expression of sensitivity and specificity that can be used to assess the value of a diagnostic test, independent of prevalence
Likelihood ratio
Equation for (+) likelihood ratio?
Patient with disease testing (+) / Patient without disease testing (+) … A / B
Equation for (–) likelihood ratio?
Patient with disease testing (–) / Patient without disease testing (–) … C / D
Smaller LR corresponds to ___ disease prevalence
Less likely
Most valid measure of central location for normally distributed data?
Mean
Most valid measure of central location for date with skewed distribution?
Median
___ refers to bias in which working populations are generally healthier than general population, exhibit lower mortality rates
Healthy worker effect
Healthy worker effect is a type of ___ bias
Selection
Describe sensitivity?
Probability of detecting a disease if it is present
Area under curve (AUC) if a Reciever-Operating Characteristic (ROC) is a reflection of …
Diagnostic accuracy
Describe statistical power?
Ability to detect an association, if that association exists
Value of meta-analysis?
Increase power of study
Higher SN means fewer …
False (–)
Higher SP means fewer …
False (+)
Relationship between SP and PPV?
Higher SP will increase PPV
Description of length-time bias?
Survival benefits of a diagnostic test are overstated … due to detection of a disease that is benign, slowly-progressive
Description of lead-time bias?
Test diagnoses a disease earlier … time from diagnosis to death appears prolonged … appears like early detection improves survival … BUT actually, disease was just diagnosed EARLIER in course
Randomized clinical trials are designed to test ___, but not ___
Efficacy; AEs
___ occurs in a meta-analysis when studies with significant (+) results are included, but studies with (-) non-significant results are not included
Publication bias
Greatest source of bias in case-control studies?
Recall bias
Quadriplegic patient remarks that that 5 years in his current state are equivalent to 1 year in full health – what is he referring to?
Quality-adjusted life years (QALYs)
Quadriplegic patient remarks that that 5 years in his current state are equivalent to 1 year in full health – what is his Time Trade-Off?
1/5 = 0.2
A patient had full health until 30 yo, then developed disability with TTO = 0.2; What is QALY at 35 yo?
30(1) + 5(0.2) = 31
Question for calculating Disability-Adjusted Life Years (DALYs)?
Years of Life Lost (YLL) + Years Lived with Disability (YLD) = YLL + YLD
Description of a factorial study design?
Study design that utilizes 2+ interventions with ALL possible combinations of those interventions … in a study with 2 interventions, a factorial design has 4 possible groups … Treatment A, Treatment B, Treatments A+B, Placebo (neither)
Alternate name for factorial study design?
Fully-crossed study
Equation for NNT?
1/ARR (absolute risk reduction)
What is the NNT for control group rate = (8/1000) and treatment group rate (6/1000)?
1/ ((8/1000) – (6/1000)) = 1/(0.002) = 500
In which situation would you need a larger sample size to show a significant difference between 2 groups … decreasing power, increasing a from 0.01 to 0.05, showing difference between groups = 5 instead of 8, using 1-tailed test instead of 2-tailed test?
Showing difference between groups = 5 instead of 8
Does using a 1-tailed or 2-tailed test require MORE power (aka larger sample size)?
2-tailed test
___ refers to repeating primary analysis calculations in a study by modifying certain criteria or variable ranges to determine whether those modifications significantly affect the results initially obtained
Sensitivity analysis
___ refers to bias that occurs when test A diagnoses a disease earlier than test B; As a result, the time from initial diagnosis to death appears prolonged with test A, although there is no actual improvement in survival
Lead-Time Bias
Recall bias is also known as a misclassification of ___
Exposure
Equation for attributable risk percent (ARP)?
[(Risk in exposed) – (Risk in unexposed)] / (Risk in total population)
Prevalence of smoking in a population is 50%; 5-year risk of ischemic CVA is 1:1000 in smokers, 0.5:1000 in non-smokers – what is the attributable risk percent of smoking on ischemic CVA?
(0.001-0.0005)/[(0.001.50) + (0.0005.50)] = 33%
Equation for Absolute Risk Reduction?
ARR = control rate – treatment rate
Equation for Relative Risk Reduction?
RRR = ARR/control rate
Equation for Relative Risk?
RR = treatment rate/control rate
Equation for NNT?
NNT = 1/ARR
Definition of positive likelihood ratio?
(probability of patient with disease testing +) / (probability of patient without disease testing +)
Definition of negative likelihood ratio?
(probability of patient with disease testing –) / (probability of patient without disease testing –)
Relationship between likelihood ratios and prevalence of disease?
LRs are independent of prevalence
Equation for Positive LR?
Positive LR = SN/(1-SP)
Equation for Negative LR?
Negative LR = (1-SN)/SP
LR < 1 indicates …
Decreased likelihood that disease is present
LR = 1 indicates …
No change in likelihood of disease
LR > 1 indicates …
Increased likelihood that disease is present
Main measure of association for case-control studies?
OR
Main measure of association for cohort studies?
RR
Main measure of association for cross-sectional studies?
OR
How can you tell if a difference between 2 groups is significant, looking at standard error bars?
Overlap of standard error bars = non-significant difference
Definition of Hazard Ratio?
Likelihood of an event occurring in a treatment group relative to a control group
What is the null value for Hazard Ratio?
1
Interpretation of Hazard Ratio < 1.0?
Event is less likely to occur in a treatment group vs. control group
Interpretation of Hazard Ratio > 1.0?
Event is more likely to occur in a treatment group vs. control group
Value of multiple linear regress?
Evaluate association between 1 quantitative dependent variable and 2+ independent variables (either quantitative or qualitative)
How can you make a confidence interval more narrow?
Increase the power (sample size)
Asymmetry in a funnel plot suggests …
Publication bias
Utility of the ANOVA test?
Comparing mean value of continuous variable across several (3+) groups
Which statistic does ANOV test generate?
F statistic … which is then used to calculate a p-value
___% of data in a normal distribution lies within 1 standard deviation
68
___% of data in a normal distribution lies within 2 standard deviations
95
___% of data in a normal distribution lies within 3 standard deviations
99.7
Relationship between mean, median, mode in a R-skewed distribution?
Mode < Median < Mean
___ refers to probability of an individual testing positive given the presence of disease
Sensitivity
___ refers to probability of an individual testing negative given the absence of disease
Specificity
___ refers to probability that an individual who tests positive for disease actually has the disease
Positive Predictive Value
___ refers to probability that an individual who tests negative for disease actually does not have the disease
Negative Predictive Value
___ refers to the probability of an individual with the disease testing positive divided by the probability of individual without the disease testing positive
Positive likelihood ratio
___ refers to the probability of an individual with the disease testing negative divided by the probability of individual without the disease testing negative
Negative likelihood ratio
Equation that relates Positive likelihood ratio to SN and SP?
LR+ = SN/(1-SP)
Equation that relates Negative likelihood ratio to SN and SP?
LR- = (1-SN)/SP
2 values that must be known to calculated the PPV and NPV?
Prevalence, Pre-test probability
Relationship between sample size (n) and Type 1 error?
Type 1 error increases as sample size increases
Equation for attributable risk percentage?
[(Risk exposed) - (Risk nonexposed)] / (Risk exposed)
Equation for hazard ratio?
(Hazard rate in intervention group) / (Hazard rate in control group)
RCT investigates anti-inflammatory intervention as treatment for neuropathic pain; Prelim results shows significant reduction in pain score among treatment group (p=0.06); Significant reduction in pain among non-obese subjects (0.01), but non-significant increase in pain scores among obese subjects (p=0.06) - what is best explanation for difference in finding between obese and non-obese subjects?
Effect modification
Relationship between PPV and R (correction coefficient)?
Higher R, Higher PPV
What is non-ABCD equation for RR?
(rate of outcome in group 1) / (rate of outcome in group 2)
Measure of association for case-control studies?
OR
Primary endpoint of study is 30% reduction in PTH after treatment with Drug X or Drug Y; Primary efficacy for Drug X = 74%, for Drug Y = 85.4% with p<0.001; What is an appropriate 95% confidence interval for this study?
11.4 must be included in CI and at center of CI; CI cannot include 0
DHS investigates an outbreak of Gi illness; Total of 25 individuals with GI bug and 75 individuals without GI bug are studied; 96% of patients with GI bug ate a local restaurant 2-4 days prior, 4% of patients without GI bug ate at same restaurant - what is the study design here?
Case-Control
Highly sensitive tests are useful for___; highly specific tests are useful for ___
Screening; Confirmation
Measure of association for case-control studies?
OR
Measure of association for cohort and RCT studies?
RR
Description of a nested case-control study?
Case-control study that draws its cases and controls from people participating in a follow-up (cohort or RCT) study
A screening test with high ___ is the best for early detection of breast CA in a population with stable incidence of disease
SN
Group of researchers wants to investigate an acute outbreak of acute diarrhea; Approximately 50 people developed severe hemorrhagic diarrhea, thought to be related to local seafood; Which study design is most appropriate to test this hypothesis?
Case-control
Difference between indications for case-control vs. cross-sectional studies?
Case-control = investigate outbreaks … Cross-sectional = compare prevalence
Utility of case-control studies?
Investigate outbreaks
Cross-sectional studies begin with …
Risk factor/exposure … investigate outbreaks
Utility of cross-sectional studies?
Compare prevalence
Which study design is helpful for studying rare exposures?
Cohort
Equation for Standardized Mortality Ratio?
(Observed # of Deaths) / (Expected # of Deaths)
Purpose of pre-clinical trials?
Explore if & how new treatment may work
Study population of pre-clinical trials?
Laboratory and animal models
Purpose of phase 1 clinical trials?
Evaluate safety
Study population of phase 1 clinical trials?
Healthy subjects, small number
Purpose of phase 2 clinical trials?
Explore efficacy (does it work?)
Study population of phase 2 clinical trials?
Small number of affected subjects
Purpose of phase 3 clinical trials?
Compare new treatment to current Standard of care
Study population of phase 3 clinical trials?
Large number of affected subjects randomly assign to treatment or control
Purpose of phase 4 clinical trials?
Identify rare and long-term adverse effects
A study has power of 80%; what is the probabilty of finding a significant effect if none actually exists?
Type 1 error = alpha; alpha = P value
Description of (+) LR?
(probability of diseased patient testing positive) / (probability of healthy patient testing positive)
Description of (-) LR?
(probability of diseased patient testing negative) / (probability of healthy patient testing negative)
Equation for (+) LR?
SN / (1-SP)
Equation for (-) LR?
(1-SN)/SP
What is the meaning of the kappa statistic?
Inter-rater reliability (concordance)
11 mo female presents shortly after diagnosed with rare disease; Parents ask about factors that mayhave contributed to developing of disease - which study design would be best?
Case Control
___ studies are best at identifying potential risk factors for rare diseases
Case Control
Description of “per protocol” analysis?
Compares treatment groups in an RCT by including ONLY patients who strictly adhered to protocol
___ refers to an analysis that compares intervention groups by including all subjects as initially allocated after randomization, regardless of what happens during the study period
Intention to Treat (ITT)
Intention to Treat (ITT) analysis leads to a ___ estimate of intervention effect
Conservative
The results of a Intention to Treat (ITT) analysis reflect the __ effect of intervention as intended in the population
Real
___ refers to an analysis that compares intervention groups by including data only from subjects who complete the intervention originally allocated at time of randomization
Per protocol
___ refers to an analysis in which subjects are evaluated based on intervention they received, rather than the intervention to which they were randomized
As-treated analysis
Which type of analysis is used to help preserve the benefits of randomization in superiority trials?
Intention to Treat (ITT)
When comparing performance measures of 2 different tests, which statistic can show that there is no significant difference between the tests?
Overlapping CIs
Which type of screening test will identify individuals without disease?
Screening test with high SP and high NPV
2 tests for heterogeneity?
Q statistic, I2 index
Which result of Q statistic suggests no heterogeneity?
P > 0.05
Which result of I2 index suggests mild heterogeneity?
25%
Which result of I2 index suggests moderate heterogeneity?
50%
Which result of I2 index suggests high heterogeneity?
75%
What is an epidemiologic parameter that does not vary with disease prevalence and provides useful information for individual patients?
Likelihood ratio (LR)
Equation for positive LR?
Positive LR = SN/(1-SP)
Equation for negative LR?
Negative LR = (1-SN)/SP
___ refers to type of measurement bias that occurs when a study uses gold standard testing selectively in order to confirm a positive (or negative) result of preliminary test
Verification bias
Effect of confounding variable on statistical significance?
Difference disappears when patients are stratified (based on the confounding variable)
Confounding threatens the ___ of a study
Internal validity
An epidemiologic study looks at adverse events in 10,000 patients who underwent PCI; According to results, 2% of patients who received metal stents developed thrombosis; 40% of patients who developed stent thrombosis passed away – the risk of death in a patient who develops stent thrombosis is best described as …
Case fatality rate
How would you describe mortality rate?
Likelihood in general population of dying from disease
Equation for FPR?
FPR = 1 – SP
Equation for FNR?
FNR = 1 – SN
Definition of standardized incidence ratio?
Observed # of cases / Expected # of cases
What is the value of standardized incidence ratio?
Determine if occurrence of disease in a small population is high or low compared to expected value derived from a larger comparison population
___ refers to the probability of detecting a true difference between normal and pathologic groups
Power
Equation for power?
1 - b
Relationship between variability and power?
Lower the variability, higher the power
Relationship between precision and power?
Higher the precision, higher the power
Diagnostic test shows that 92% of all skin tissue samples with (+) result were cancerous, while 99% of all skin tissue samples with (-) result were normal – what is best interpretation of these findings?
92% is PPV, 99% is NPV
If RR of an outcome in group A as compared to group B is x, then RR in group B compared to group A is …
1/x
RR of SCD in low-dose HCTZ group as compared to high-dose HCTZ is 0.4; how much does high-dose HCTZ increase risk of SCD compared to low-dose group?
Increased by factor of 2.5
SN of study is 50% (2/4), SP is 98% (394/400) – what is PPV?
25%
Ecologic (correlational) studies are based on ___ level
Population
___ refers to the mistake of applying population-level information to individual level
Ecologic fallacy … results of ecologic study cannot provide individual-level conclusions
___ refers to misclassification of events that results from knowing exposure status of patient in RCT
Observer bias
Best way to reduced observer bias?
Blinding
Description of Hawthorne Effect?
Patients change behavior when they know they’re being studied
Hawthorne Effect poses a threat to the ___ of a study
Internal validity
Association between risk factor and disease is more likely to be causal if …
Strength increases as exposure level increases (dose response relationship)
__Refers to the incidence rate of an event and intervention group divided by the incidence rate of the same event in the control group
Rate Ratio
