UAMS P3 - EBM E2 Practice

True/False: Correlation analysis can establish causality between two variables.

False. Correlation describes how two variables change together; it does NOT measure causality.

Multiple Choice: What is the allowable range of the Pearson correlation coefficient r?

A) 0 to 1

B) -1 to 1

C) -0.5 to 0.5

D) 1 to 10

B) -1 to 1.

Fill in the blank: r = 0 implies __________.

No linear relationship (there may still be a nonlinear relationship).

Multiple Choice: Pearson correlation is most appropriate when the two variables are:

A) Ordinal

B) Binary

C) Continuous and normally distributed

D) Count data following Poisson

C) Continuous and normally distributed.

Multiple Choice: Spearman's correlation is preferred when:

A) Both variables are continuous and normal

B) At least one variable is ordinal or the data are skewed/non-normal

C) The relationship must be strictly linear

D) The goal is to infer causality

B) At least one variable is ordinal or the data are skewed/non-normal.

True/False: Spearman's correlation requires the relationship to be linear.

False. Spearman measures monotonic relationships and does not require linearity.

Select all that apply: Which statements correctly describe the direction of correlation?

A) Positive correlation: as one variable increases, the other increases

B) Positive correlation: as one variable increases, the other decreases

C) Negative correlation: as one variable increases, the other decreases

D) Negative correlation: as one variable increases, the other increases

A and C.

Applied: You measured pain on a 5‑point Likert scale and age in years. The age distribution is heavily right‑skewed. Which correlation test is most appropriate and why?

Spearman's correlation, because at least one variable is ordinal and the distribution is non‑normal.

Applied interpretation: r = -0.62 for (hours studied, number of errors on an exam). Interpret the direction.

Negative direction: as hours studied increase, the number of errors tends to decrease.

Select all that apply about Spearman's correlation:

A) Uses ranks rather than raw values

B) Measures monotonic relationships

C) Requires both variables to be continuous and normal

D) Does not have to be linear

A, B, and D.

Fill in the blank: The symbol r measures the ________ of association between two variables.

Strength (and direction) of the correlation.

Multiple Choice: Which scenario could yield r ≈ 0 despite a strong association?

A) A strictly linear upward trend

B) A curved (U‑shaped) relationship

C) A strictly linear downward trend

D) Perfect positive alignment of all points

B) A curved (U‑shaped) relationship (nonlinear).

Applied significance: A study reports r = 0.25 with a 95% CI of (0.05, 0.44). Is this correlation statistically significant at α = 0.05?

Yes. The 95% CI does not include 0.

Fill in the blank: Spearman's correlation evaluates the ________ relationship between two ranked variables.

Monotonic.

True/False: Pearson focuses on raw values, whereas Spearman focuses on ranks.

True.

Multiple Choice: Which data conditions best support Pearson over Spearman?

A) Two continuous variables with approximate normality

B) One ordinal variable and one skewed continuous variable

C) A monotonic but curved relationship with outliers

D) Binary outcomes

A) Two continuous variables with approximate normality.

Applied: You have systolic BP (mmHg) and age (years), both roughly normal. Which correlation would you choose and why?

Pearson correlation, because both variables are continuous and approximately normally distributed.

Select all that apply about correlation vs. causation:

A) Correlation measures how variables change together

B) Correlation proves that one variable causes the other

C) Correlation does not measure causality

D) Causality implies a direct cause‑effect mechanism

A, C, and D.

Fill in the blank: As X increases, Y decreases describes a ________ correlation.

Negative.

Applied: A scatterplot shows a strictly increasing but curved trend with non‑normal data. Which correlation is preferable and why?

Spearman's correlation, because it assesses monotonic (not necessarily linear) relationships and does not require normality.

True/False: Regression analysis examines the one-way relationship of one variable with multiple other variables.

True. Regression explores how a dependent variable (Y) changes with one or more independent variables (X's).

Multiple Choice: In regression terminology, the dependent variable (Y) is also referred to as:

A) Covariate

B) Predictor

C) Outcome variable

D) Independent variable

C) Outcome variable.

Fill in the blank: A variable that influences both the likelihood of receiving treatment and the occurrence of the outcome, but is not in the causal pathway, is called a ________.

Confounding factor.

Multiple Choice: Linear regression should be used when the dependent variable is:

A) Nominal

B) Binary

C) Continuous (interval/ratio)

D) Ordinal

C) Continuous (interval/ratio).

True/False: Logistic regression is used when the dependent variable is binary and reports results as Odds Ratios (OR).

True.

Multiple Choice: Which regression model is most appropriate for time-to-event (survival) data?

A) Linear regression

B) Logistic regression

C) Cox proportional hazards model

D) Poisson regression

C) Cox proportional hazards model.

Select all that apply: Which are primary purposes of regression analysis?

A) Reduce bias

B) Estimate independent effects of risk factors

C) Adjust for imbalances in baseline characteristics

D) Ignore confounders

A, B, and C.

Fill in the blank: In linear regression, a positive coefficient (> 0) means X has a ________ association with Y.

Positive association - as X increases, Y increases.

Applied interpretation: If a logistic regression reports OR = 2.5 (95% CI 1.2-3.4), this means:

The odds of the outcome are 2.5 times higher in the treatment group; statistically significant since CI does not include 1.

True/False: In Cox regression, HR > 1 indicates a decrease in hazard of the outcome with X.

False. HR > 1 indicates an *increase* in the hazard (risk) of the outcome.

Multiple Choice: Which of the following is NOT a confounding factor?

A) Variable influencing both treatment and outcome

B) Variable within the causal pathway

C) Variable not related to outcome

D) Variable unrelated to exposure

B) A variable within the causal pathway is not a confounder.

Fill in the blank: In regression, the independent variable(s) are also called ________.

Predictors or explanatory variables.

Select all that apply: Which statements about statistical inference in regression are correct?

A) p < 0.05 indicates statistical significance

B) For ratios (OR, HR), the CI must not include 1

C) For coefficients, the CI must not include 0

D) Significance proves clinical importance

A, B, and C are correct; D is incorrect.

Applied: A Cox model reports HR = 0.70. Interpret this finding.

There is a 30% reduction in hazard of the outcome with the treatment compared to control.

15) Multiple Choice: The dependent variable in logistic regression is:

A) Continuous

B) Binary (yes/no)

C) Categorical with > 2 levels

D) Ordinal only

B) Binary (yes/no).

True/False: Regression can help adjust for confounders and improve precision by reducing standard errors.

True.

Applied: A linear regression coefficient of -5.2 means:

For every 1‑unit increase in X, the dependent variable (Y) decreases by 5.2 units.

Multiple Choice: What does it mean if the 95% CI for a regression coefficient includes 0?

A) Statistically significant effect

B) No statistically significant effect

C) Confounding bias present

D) Model mis-specified

B) No statistically significant effect.

Fill in the blank: Logistic and Cox regressions report measures of association as ________ and ________, respectively.

Odds Ratios (OR) and Hazard Ratios (HR).

Select all that apply: Which regression models correspond to the outcome variable type?

A) Linear → Continuous outcome

B) Logistic → Binary outcome

C) Cox → Time‑to‑event outcome

D) Logistic → Time‑to‑event outcome

A, B, and C.

True/False: In survival analysis, 'survival' refers to patients who have not yet experienced the event of interest.

True. 'Survival' represents individuals who have not had the outcome event occur.

Fill in the blank: When the time to an outcome event is not fully observed, this is known as ________.

Censoring.

Multiple Choice: Which of the following is a *nonparametric* method used to construct the survival function?

A) Linear regression

B) Kaplan-Meier estimator

C) Cox proportional hazards model

D) Log‑rank regression

B) Kaplan-Meier estimator.

Multiple Choice: Which test is used to evaluate overall differences between survival curves?

A) Student's t‑test

B) Log‑rank test

C) Chi‑square test

D) Wilcoxon signed‑rank test

B) Log‑rank test.

True/False: In a Kaplan-Meier curve, the survival function starts at 0% at time 0 and increases as events occur.

False. It starts at 100% at t = 0 and never increases—it only decreases as events occur.

Select all that apply: Which statements about the Kaplan-Meier curve are true?

A) It estimates survival probabilities over time.

B) It accounts for censored observations.

C) It can increase when new subjects are added.

D) It never increases over time.

A, B, and D are true.

Fill in the blank: The regression model used for survival data that reports hazard ratios assumes that the hazard remains constant over time—this is known as the ________ assumption.

Proportional hazards assumption.

Applied interpretation: A Cox regression model yields HR = 1.5 (p = 0.02). Interpret this result.

The hazard of the outcome is 50% higher in the treatment group; statistically significant since p < 0.05.

Multiple Choice: Which of the following statements about censored data is TRUE?

A) Censored patients are excluded from analysis.

B) Censoring occurs only when a subject dies.

C) Censored patients contribute survival information up to their last known follow‑up.

D) Censoring invalidates survival models.

C) Censored patients contribute data up to their last follow‑up time.

True/False: In survival analysis, hazard ratios greater than 1 indicate increased hazard (risk) of the outcome in the treatment group compared with control.

True.

True/False: The PICO framework is used to formulate a clear and focused clinical question for evidence-based research.

True.

Fill in the blank: The 'P' in PICO stands for ________.

Population or Patient problem (e.g., age, condition, ethnicity).

Multiple Choice: Which of the following best represents the 'I' in PICO?

A) Independent variable

B) Intervention (test, drug, therapy)

C) Inclusion criteria

D) Interpretation

B) Intervention (test, drug, or therapy).

Multiple Choice: In PICO, the 'C' represents:

A) Cohort

B) Comparator or Control

C) Covariate

D) Correlation

B) Comparator or Control - the alternative to the intervention.

Fill in the blank: The 'O' in PICO stands for ________, describing what you aim to accomplish, measure, or improve.

Outcome.

True/False: Time (T) is always a required component in a PICO question.

False. Time is optional and included when duration is clinically relevant.

Select all that apply: Which elements make up a complete PICO question?

A) Population/Problem

B) Intervention

C) Comparison

D) Outcome

E) Time (if applicable)

A, B, C, D, and optionally E.

Applied: Formulate the PICO elements for this scenario - 'Does metformin compared to placebo reduce A1c levels in adults with type 2 diabetes over 6 months?'

Identify P, I, C, O, T.

P = Adults with type 2 diabetes; I = Metformin; C = Placebo; O = Reduction in A1c; T = 6 months.

Multiple Choice: Which of the following statements about the purpose of PICO is TRUE?

A) It helps design pharmacokinetic studies only.

B) It organizes literature search strategies and clarifies the clinical problem.

C) It replaces the need for statistical analysis.

D) It is used only in qualitative research.

B) It organizes literature searches and clarifies clinical questions.

Fill in the blank: The overall purpose of using the PICO model is to ________.

Formulate a concise, answerable clinical question that guides evidence-based practice.

True/False: 'Risk' is defined as the probability of an outcome of interest occurring within a specified period.

True.

Fill in the blank: Absolute Risk Reduction (ARR) = Risk in ________ - Risk in ________.

ARR = Risk in control group - Risk in treatment group.

Multiple Choice: Relative Risk (RR) is calculated as:

A) (Risk in control) / (Risk in treatment)

B) (Risk in treatment) / (Risk in control)

C) (Odds in treatment) / (Odds in control)

D) 1 - (Risk in treatment)

B) (Risk in treatment) / (Risk in control).

Applied Math: If 20 % of patients in the control group and 10 % in the treatment group experience an event, what is the ARR?

ARR = 0.20 - 0.10 = 0.10 (10 %).

Fill in the blank: Number Needed to Treat (NNT) = 1 / ______.

NNT = 1 / Absolute Risk Reduction (ARR).

Applied Math: Using the data in Question 4 (ARR = 0.10), calculate the NNT.

NNT = 1 / 0.10 = 10 → 10 patients must be treated to prevent 1 event.

Multiple Choice: A Relative Risk (RR) > 1 indicates:

A) Lower risk in treatment vs control

B) No difference

C) Higher risk in treatment vs control

D) Cannot determine

C) Higher risk in treatment vs control.

True/False: Relative Risk Reduction (RRR) = 1 - RR.

True.

Applied Math: If the RR = 0.75, what is the RRR?

RRR = 1 - 0.75 = 0.25 → 25 % relative risk reduction.

Select all that apply: Which are appropriate uses of RR?

A) Cohort studies

B) Case-control studies

C) Prospective trials

D) Cross‑sectional studies

A and C - RR is used in cohort and prospective studies.

Multiple Choice: Odds Ratio (OR) compares:

A) Risk of event in treatment vs control

B) Probability of event to probability of no event

C) Mean difference in outcomes

D) Hazard ratio of treatment vs control

B) The odds of an event to the odds of no event between groups.

Applied Math: If 30 % of treated patients have an event and 60 % of controls have an event, calculate the OR.

(Show all steps.)

Treatment odds = 0.30 / 0.70 = 0.43; Control odds = 0.60 / 0.40 = 1.5; OR = 0.43 / 1.5 = 0.29.

Fill in the blank: Number Needed to Harm (NNH) = 1 / ______.

NNH = 1 / Absolute Risk Increase (ARI).

Applied Math: If 10 % of treated patients and 5 % of controls experience a side effect, calculate the NNH.

ARI = 0.10 - 0.05 = 0.05; NNH = 1 / 0.05 = 20 → 1 additional harm per 20 patients treated.

True/False: ORs and RRs both equal 1 when there is no difference in risk or odds between groups.

True.

A clinical trial evaluating Drug X for hypertension designates reduction in mean systolic blood pressure at 12 weeks as its main outcome. All other outcomes (e.g., change in LDL, body weight) are exploratory.

What type of endpoint is the reduction in mean systolic blood pressure?

Primary endpoint.

A study of a new diabetes medication measures change in HbA1c as its main endpoint, and also tracks body‑weight change and fasting glucose for additional analysis.

Which of the following are secondary endpoints?

Body‑weight change and fasting glucose levels.

A heart‑failure trial measures a combined outcome of cardiovascular death, hospitalization for heart failure, or urgent visit for IV diuretics.

What type of endpoint is this?

Composite endpoint.

True/False: Composite endpoints are designed to accumulate a higher number of events and thereby increase study power.

True.

Select all that apply: Which of the following are appropriate characteristics of a composite endpoint?

A) Each component is clinically important

B) Each component's biological plausibility is similar

C) A soft endpoint drives the results despite minimal clinical relevance

D) All components move in the same clinical direction

A, B, and D are correct.

A trial evaluates Drug A vs Drug B for stroke prevention. The main outcome is time to first ischemic stroke. Secondary outcomes include functional recovery and bleeding events.

Identify each endpoint type.

Primary = Time to first ischemic stroke; Secondary = Functional recovery and bleeding events.

In a lipid‑lowering trial, the composite endpoint consists of myocardial infarction, stroke, and cardiovascular death. The trial finds a reduction mainly in non‑fatal MI but not in stroke or death.

Which limitation of composite endpoints does this illustrate?

That one component (a 'soft' endpoint) can disproportionately drive the overall result, potentially misleading interpretation.

Fill in the blank: The main purpose of designating a *primary endpoint* is to determine study ________ and calculate statistical ________.

Purpose and power.

Select all that apply: Which of the following endpoints would typically be *secondary endpoints* in a cancer trial?

A) Overall survival

B) Progression‑free survival

C) Quality‑of‑life score

D) Biomarker response rate

C and D — Quality‑of‑life and biomarker response are commonly secondary endpoints.

A clinical trial on an asthma inhaler lists the following outcomes:

• Change in FEV₁ at 6 months

• Rate of asthma exacerbations

• Hospitalizations for respiratory causes

Which outcomes could reasonably form a *composite endpoint*, and why?

FEV1 change alone is a primary endpoint;

the rate of exacerbations plus hospitalizations could form a clinically meaningful composite endpoint reflecting disease control and severity.

A study comparing Drug A vs placebo finds a p‑value = 0.03 and a mean BP reduction of 1 mmHg. Is this statistically or clinically significant?

Statistically significant (p < 0.05) but not clinically significant (minimal effect size).

The mean A1c change with Drug B vs placebo was -0.8 % (p = 0.10). Interpret the result.

Neither statistically nor clinically significant (p > 0.05).

A trial reports a mean pain score reduction of 2.5 points (95 % CI -3.8 to -1.2; p < 0.01). The MCID for pain is 2 points. Interpret the result.

Both statistically and clinically significant (CI excludes 0 and exceeds MCID).

A trial finds RR = 0.95 (95 % CI 0.85 - 1.06; p = 0.32). Interpretation?

Not statistically significant (CI includes 1); not clinically significant.

True/False: A statistically significant result always means the finding is clinically important.

False.

Select all that apply: Which criteria must be met for statistical significance?

A) p < 0.05

B) 95 % CI includes 0 (for mean diff)

C) 95 % CI excludes 0 or 1

D) Effect size is large

A and C are correct criteria for statistical significance.

A study shows Drug C reduces cholesterol by 18 mg/dL (p = 0.04). MCID = 10 mg/dL. Interpretation?

Both statistically and clinically significant (effect > MCID and p < 0.05).

A study reports a hazard ratio of 0.88 (95 % CI 0.75 - 1.12; p = 0.27). Interpretation?

Neither statistically nor clinically significant (CI includes 1 and p > 0.05).

A trial finds a mean difference of -4 points on a fatigue scale (95 % CI -8 to -0.5; MCID = 5). Interpretation?

Statistically significant (CI excludes 0) but not clinically significant (effect < MCID).

Fill in the blank: Clinical significance focuses on whether the observed effect is _______ to the patient.

Meaningful or beneficial to the patient.

Multiple Choice: Which of the following best explains why large sample sizes can yield statistical but not clinical significance?

A) They inflate variance

B) They reduce effect size

C) They make tiny differences detectable

D) They increase Type II error

C) They make tiny differences detectable as statistically significant.

A trial comparing two antidepressants finds a difference in response rates of 3 % (p = 0.02). Interpretation?

Statistically significant (p < 0.05) but not clinically significant (effect too small to impact care).

Select all that apply: Which statements are true about clinical significance?

A) Considers magnitude and relevance to patients

B) Determined by p‑value alone

C) May exist even if p > 0.05

D) Based on MCID or expert judgment

A, C, and D are true.

Applied: A study of Drug D for migraine shows p = 0.049 and a mean pain reduction of 1 point on a 10‑point scale (MCID = 2). Interpretation?

Statistically significant but not clinically significant.

True/False: Trials can be clinically significant even if p > 0.05 when the effect size is large and patient benefit is meaningful.

True.