Trial Design + Biostats Flashcards
- A quality improvement (QI) initiative is implemented
to decrease falls in patients recently discharged from
the hospital. As the pharmacist champion at a pharmacy benefit manager (PBM), you have worked with the formulary team to ensure that medications with a high risk of falls have criteria for use. Which factor will be most important in showing the effectiveness of this QI initiative?
A. Obtaining informed consent for participation in
the QI initiative.
B. Identifying patients at risk for falls.
C. Creating a community advertising campaign to
bring awareness to the initiative.
D. Determining the social value of the QI initiative.
1.
A QI study would likely show effectiveness using a pre/
post-intervention cohort design. In this type of study,
informed consent is generally not required because the treatment is provided to all patients as a standard of care (Answer A is incorrect). A community advertising campaign might improve the delivery of care and improve patient adherence, but it is unnecessary to measure the effectiveness of the QI initiative (Answer C is incorrect). The QI initiative is believed necessary and has therefore been deemed to have intrinsic social value (Answer D is incorrect). To show the effectiveness of the intervention, populations at risk for falls must be recognized before initiating an intervention (Answer B is correct).
Correct Answer: B
- A study seeks to determine the impact of adverse drug events on patient outcomes in FDA-approved drugs. Which would be the best approach to conducting this study?
A. A randomized controlled clinical trial (RCT) with
a test for continuous variables to determine the
difference in outcomes.
B. A retrospective case-control study with a test
for proportions to determine the difference in
outcomes.
C. A prospective observational study with survival
analysis to determine the difference between
cohorts.
D. A retrospective cohort study with a test for proportions to determine the difference in outcomes.
- To effectively determine the incidence and clinical impact
of adverse drug events on clinical outcomes, it would be
unethical to randomize patients to experience the event
(Answer A is incorrect). A retrospective design would
not be ideal because of the limitations in data extraction, assignment of events, and interpretation of causality
(Answers B and D are incorrect). A prospective observational design would allow the investigator team to identify the incidence of events and sequential events and determine causality (Answer C is correct).
Correct Answer: C
A case-control study is conducted to determine
whether proton pump inhibitor (PPI) use is associated
with an increased risk of developing C. difficile infection (CDI). The final analysis shows the odds ratio
(OR) for CDI with PPI exposure to be 1.3 (95% confidence interval [CI], 0.8–1.5). Which best describes
the results?
A. PPI exposure increases the risk of CDI by 130%.
B. PPI exposure reduces the risk of CDI by 20%.
C. PPI exposure increases the risk of CDI by 30%.
D. PPI exposure is not associated with an increased
risk of CDI.
- A correct interpretation of the results is recognizing that
even though the OR suggests an associated increase of 30% in the risk of being exposed to CDI, the 95% CI crosses 1, meaning that the odds of exposure to a PPI are as likely to increase the risk as to decrease the risk of developing CDI (Answer D is correct; Answers A–C are incorrect).
Correct Answer: D
Questions 4 and 5 pertain to the following case:
A retrospective case-control study was conducted to assess peripheral neuropathy in patients taking fluoroquinolone antibiotics. Risk factors for neuropathy were not assessed (however, both the cases and the controls of this study had identical diagnostic evaluations and were stratified according to the duration of quinolone use before the onset of
neuropathy). Which type of bias is this study design most susceptible to?
A. Confounding by indication
B. Recall bias
C. Diagnostic bias
D. Misclassification
- Recall bias is always a potential concern for case-control
studies because of the time that passes between the study and the drug “ingestion.” Because risk factors were not included in the study design, this is of concern (Answer B is correct). Although a study may be susceptible to many types of bias, the other choices would not pose as much risk (if any) as recall bias (Answers A, C, and D are incorrect).
Correct Answer: B
Questions 4 and 5 pertain to the following case:
A retrospective case-control study was conducted to assess peripheral neuropathy in patients taking fluoroquinolone antibiotics. Risk factors for neuropathy were not assessed (however, both the cases and the controls of this study had identical diagnostic evaluations and were stratified according to the duration of quinolone use before the onset of
neuropathy).
Which factor will be most affected by the type of bias
likely to occur in this study?
A. External validity
B. Internal validity
C. Assessment of exposure
D. Number of patients needed for the study
- Internal validity is greatly jeopardized because the study
is not designed to protect against this possible bias. Case-control studies are an inappropriate design to access the frequency of an event. In a sense, this design flaw eventually jeopardizes external validity (how well does a study apply to other patients with this condition/disease?), but a lack of internal validity is most essentially affected (Answer B is correct). The other answers can adequately be controlled for in the design and conduct of the study (Answers A, C, and D are incorrect).
Correct Answer: B
- When describing the results of an RCT, the investigators report using an intention-to-treat analysis to
analyze their data. The results of their investigation
comparing two antivirals for influenza-like illness
show no difference in the number of hospitalizations
for influenza between the treatment groups. Given
their method of data analysis, which statement is most
appropriate?
A. May be susceptible to issues regarding recall bias.
B. Provides a good measure of effectiveness under
typical clinical conditions.
C. Cannot estimate the method’s effectiveness.
D. May overestimate the actual treatment effect.
- Intention-to-treat analysis is a comparison of the treatment groups that includes all patients as originally allocated after randomization. This is the recommended method to estimate effectiveness under typical clinical conditions and avoid any bias. Per-protocol analysis is a comparison of treatment groups that includes only patients who completed the treatment originally allocated. If performed alone, per-protocol analysis leads to bias (Answer B is correct; Answer C is incorrect). Intention-to-treat analysis
is the most common approach to data analysis for RCTs and may underestimate the treatment effect (Answer D is incorrect). Recall bias is not a concern with RCTs (Answer A is incorrect).
Correct Answer: B
Questions 7 and 8 pertain to the following case:
A prospective randomized study compared furosemide
with hydrochlorothiazide when treating patients with newly diagnosed hypertension. One of the study’s end points was myocardial infarction one year after treatment initiation. The following table summarizes myocardial infarction rates in all patients.
(All patients, n (%)):
Furosemide: 17/332 (5.1)
Hydrochlorothiazide: 21/274 (7.7)
The 95% CI for the difference in myocardial infarction
rates between the two groups was −3.2% to 10.3%.
Which conclusion is most appropriate?
A. Furosemide is superior to hydrochlorothiazide.
B. Superiority of furosemide could not be established over hydrochlorothiazide.
C. Hydrochlorothiazide is not inferior to furosemide.
D. No conclusion can be drawn because p values are
unavailable.
- The CI of the difference in myocardial infarction rate
between the two groups includes zero; thus, there is no
statistically significant difference between the two groups
(Answer B is correct). Answer A is incorrect because the
95% CI contains zero and is therefore not statistically
significant. Answer C is incorrect because insufficient
information is provided. Answer D is incorrect because all
the previously stated information can be determined without the benefit of reported p values.
Correct Answer: B
Questions 7 and 8 pertain to the following case:
A prospective randomized study compared furosemide
with hydrochlorothiazide when treating patients with newly diagnosed hypertension. One of the study’s end points was myocardial infarction one year after treatment initiation. The following table summarizes myocardial infarction rates in all patients.
(All patients, n (%)):
Furosemide: 17/332 (5.1)
Hydrochlorothiazide: 21/274 (7.7)
According to the data and the result obtained, which
best represents the number of patients who would need to be treated with furosemide to prevent one case of myocardial infarction?
A. Number needed to treat (NNT) would be 3.
B. NNT would be 39.
C. NNT would be 250.
D. NNT should not be calculated because the result
was nonsignificant.
- Answers A and C are incorrect calculations. Calculating
the NNT to prevent one recurrence using furosemide is
as follows: 0.051 − 0.077 = 0.026 and 1/0.026= 38.46, or
around 39. However, the NNT should not be calculated
when the end point of interest is nonsignificant (Answer B is incorrect; Answer D is correct).
Correct Answer: D
- An RCT assesses the effects of different inhaled
medication regimens on pulmonary function in three
groups of subjects with chronic obstructive pulmonary
disease. Groups will be randomly assigned to 1) long-acting muscarinic antagonist alone, 2) long-acting
muscarinic antagonist combined with a long-acting
beta-2 agonist, or 3) long-acting beta-2 agonist combined with an inhaled corticosteroid. Study outcomes will be assessed 3 months after randomization. Investigators wanted to assess pulmonary function with the Global Initiative for Chronic Obstructive Lung Disease (GOLD) system, which is an ordered scale from 1 to 4. Which statistical test is most appropriate to assess differences in functional classification between the groups?
A. Kruskal-Wallis test
B. Wilcoxon signed-rank test
C. Analysis of variance (ANOVA)
D. Analysis of covariance (ANCOVA)
- The GOLD system is an ordinal scale from 1 (mild) to 4
(very severe). Neither ANOVA nor ANCOVA is appropriate for ordinal or noncontinuous data (Answers C and D are incorrect). The Wilcoxon signed-rank test is an appropriate nonparametric test to use for paired ordinal data, such as the change in GOLD score over time on the same person (Answer B is incorrect). The Kruskal-Wallis test is the nonparametric analog of a one-way ANOVA and is appropriate for this analysis (Answer A is correct).
Correct Answer: A
- You are evaluating a randomized, double-blind, parallel group-controlled trial that compares four different
antihypertensive regimens for their effect on systolic
blood pressure in patients with refractory hypertension:
• Regimen 1: hydrochlorothiazide/diltiazem/
metoprolol
• Regimen 2: furosemide/diltiazem/metoprolol
• Regimen 3: diltiazem/metoprolol/enalapril
• Regimen 4: hydrochlorothiazide/metoprolol/
enalapril
The authors conclude that regimen 1 is better than
regimen 2 (p<0.05) and that regimen 4 is better than
regimen 1 (p<0.01), but no difference is observed
between any other regimens. The investigators used
an unpaired (independent sample) t-test to test the
hypothesis that each antihypertensive regimen was
equivalent. Which statement is most appropriate?
A. Investigators used the appropriate statistical test
to analyze their data.
B. Regimen 4 is the most effective of the tested antihypertensive regimens.
C. ANOVA would have been a more appropriate test.
D. A paired t-test is a more appropriate test.
- You cannot determine which finding is more important (in this case, the best antihypertensive regimen) based on the p value (i.e., a lower p value does not mean it’s more important; Answer B is incorrect). All statistically significant results are interpreted as significant without respect to the size of the p value. This trial had four independent samples, and the unpaired (independent samples) t-test is not appropriate because it requires several unnecessary tests (multiple comparisons) and increases the chances of making a type I error (Answer A is incorrect). In this setting, ANOVA is the correct test (Answer C is correct), followed by a multiple comparisons procedure to determine where the actual differences between the groups lie. A paired t-test is inappropriate because this is a parallel group trial (Answer D is incorrect). Using ANOVA in this case assumes a normal distribution and equal variance in each of the four groups.
Correct Answer: C
- In the results of a randomized, double-blind, controlled clinical trial, the difference in 10-year mortality between the intervention group and the control group is 6% (p=0.01), and it is concluded that there is a statistically significant difference between the groups. Which statement is most consistent with this finding and conclusion?
A. The chance of making a type I error is 5 in 100.
B. The trial does not have enough power.
C. There is a high likelihood of having made a type
II error.
D. The chance of making a type I error is 1 in 100.
- The typical a priori error (type I) rate is 5% (i.e., when the study was designed, the error rate was designed to be 5% or less). The actual type I error rate is reported in the question as 0.01 (1%) (Answer A is incorrect). Information is insufficient to select Answer B (Answer B is incorrect). A type II error was not made because this error has to do with not finding a difference when one truly exists (Answer C is incorrect). In this question, the type I error rate is 1%, the value of the p value (Answer D is correct).
Correct Answer: D
- Researchers planned a study to evaluate the percentage of subjects who were cured from hepatitis C virus (HCV) infection with direct-acting antivirals (DAAs)
compared with interferon. In the study of 1000 subjects, the DAA group (n=500) and the interferon group
(n=500) were compared. The investigators used HCV
cure as their primary end point. Which is the most
appropriate statistical test to answer such a question?
A. Independent-samples t-test
B. Chi-square or Fisher exact test
C. Wilcoxon signed-rank test
D. One-sample t-test
- The primary end point in this study, the percentage of
subjects cured of HCV, is nominal data. This type of data requires either a chi-square test or a Fisher exact test, depending on the sample size or, more accurately, the number of counts in the individual contingency table cells. (Answer B is correct). An independent-samples t-test is not appropriate because continuous data (e.g., viral load) is not being compared for the primary end point (Answer A is incorrect). If we were comparing viral load between the two groups, the test might be appropriate if parametric assumptions were met. The Wilcoxon signed-rank test is the appropriate nonparametric test for comparing paired samples (usually in a crossover trial; Answer C is incorrect). Finally, a one-sample t-test is used to compare the
mean of a single group with the mean of a reference group. This is also incorrect in this situation because two groups are being compared (Answer D is incorrect).
Correct Answer: B
- An investigational drug is being compared with an
existing drug for the treatment of diabetes in subjects
> 65 years of age. The study is designed to detect a
minimum 20% difference in response rates between
the groups, if one exists, with an a priori alpha (α)
of 0.05. The investigators believe that a smaller difference (10% difference vs 20% difference) would be
more clinically meaningful. In revising their study,
they decide they want to be able to detect a minimum
10% difference in response. Which change to the
study parameters is most appropriate?
A. Increase the sample size.
B. Change the study design to compare the investigational drug with placebo.
C. Select an alpha (α) of 0.01 as a cutoff for statistical significance.
D. Decrease the sample size.
- Detecting the smaller difference between the treatments requires more power. Power can be increased in several different ways. Answer A is correct because the most common approach is to increase the sample size. Answer D is incorrect because smaller sample sizes diminish a study’s ability to detect differences between groups. Power can also be increased by increasing the α value, but doing so increases the chances of a type I error. Answer C decreases α, thus making it more difficult to detect differences between groups. Although the subjects are older patients with diabetes, Answer B is likely unethical because treatment options for diabetes are available and thus should not be withheld from study subjects.
Correct Answer: A
- You are designing a new computer alert across your
clinic system for patients on anticoagulants at high
risk for bleeding. You want to develop a model to predict which patients are most likely to have a bleed.
You plan to retrospectively assess the presence or
absence of several different variables based on data
from a cohort across your clinic system. The comparison will consist of all patients on anticoagulants who
had bleeds compared with all other patients on anticoagulants over the last 3 years. Which technique will be most useful in completing such an analysis?
A. Correlation
B. Kaplan-Meier curve
C. Regression
D. ANOVA
- Regression analysis is the most effective way to develop models to predict outcomes or variables (Answer C is correct). There are many different types of regression, but all share the ability to evaluate the impact of multiple variables simultaneously on an outcome variable. Correlation analysis is used to assess the association between two (or more) variables, not to make predictions (Answer A is incorrect). Kaplan-Meier curves are used to graphically depict survival curves or time to an event (Answer B is incorrect). Although ANOVA can adjust for covariates, it is limited on the number of variables. In addition, ANOVA evaluates continuous outcomes, and the outcome in the question is categorical (Answer D is incorrect).
Correct Answer: C
Null hypothesis is
Stating no difference between groups
All statistical conclusion are made in reference to the Null hypothesis
