Research Design, Biostatistics, and Literature Evaluation

Answer: D

This study requires consideration of options A, B, and

C. Informed consent may be waived because both drugs

may be considered standard of care, but some IRBs at

various institutions may require the use of community

consent measures; obtaining informed consent directly

from the patient or a family member may be impossible

due to the acute nature of the intervention. Differences

in local perceptions on why patients may be agitated

on presentation to the emergency department could

lead to heterogeneity in use of the study agents, affect-

ing results downstream. In addition, a multicenter trial

creates a risk of differences in patient characteristics

during enrollment that may need to be controlled for

in the study. Option D is not a valid practical consider-

ation based on the study design. The cluster crossover

trial may compare two known standards of care, and the

crossover design allows for hospitals/units to serve as

their own controls. Risk of bias is never zero, but blind-

ing would not introduce unnecessary bias into this trial’s

results (Answer D is correct; Answers A, B, and C are

incorrect).

The study goal is to determine whether exposure to

4F-PCC resulted in mortality, or the outcome of interest.

This is performed in a cohort of patients who all have

end-stage liver disease (Answer B is correct). A case-

control study involves structuring study groups by those

who did and did not have an outcome of interest in order

to determine relevant risk factors leading to that outcome

(Answer A is incorrect). A case study would be better

suited to a single group without a comparator and is typi-

cally used to describe a novel intervention or outcome

(Answer C is incorrect). Pharmacokinetic trials typically

involve sample collection and evaluation of drug concen-

trations to describe the kinetics of the drug as opposed to

clinical outcomes (Answer D is incorrect).

Answer: A

If patients who are bleeding more often receive 4F-PCC

according to institutional standards or physician prefer-

ence, there is likely a difference in the severity of illness

between the two study groups (those who did and did not

receive 4F-PCC), which could lead to selection bias in

the result of mortality (Answer A is correct). Although

measurement bias is possible because of multiple peo-

ple collecting data, it is less likely to have as large an

impact on the outcomes as long as a data collection plan

is implemented and followed (Answer B is incorrect).

Misclassification bias is also possible but is less likely to

affect outcomes at the same magnitude as selection bias

as long as a consistent data collection and chart review

plan is followed (Answer C is incorrect). Observation

bias would likely be more important in a prospective trial

than in a retrospective trial to blind those involved to

ensure all other standards of care are followed (Answer

D is incorrect).

Answer: C

Student t-test is appropriate to compare two independent

means as a measure of parametric or normally distrib-

uted data (Answer C is correct). A chi-square analysis

is used for categorical data (Answer A is incorrect).

Wilcoxon signed rank is used to compare two related

medians as a measure of nonparametric data (Answer B

is incorrect). Mann-Whitney U is used to compare two

independent medians as a measure of nonparametric

data (Answer D is incorrect).

Answer: B

A regression analysis can be used to identify independent

associations in the presence of other risk factors for the

outcome. Logistic regression is used when the outcome

of interest (i.e., patients with positive or negative procal-

citonin) is categorical in nature (Answer B is correct).

Linear regressions are used similarly to logistic regres-

sions for a continuous dependent variable (Answer A is

incorrect). A Cox proportional hazards model is used

as a survival analysis or time-to-event analysis, which

would not be applicable to the outcome or question in

this case (Answer C is incorrect). Although a correlation

could determine the strength of association between pos-

itive procalcitonin and infection, it would not consider

other risk factors that could be confounding the results

(Answer D is incorrect).

Answer: D

The sensitivity is around 94%, and the specificity is 50%

(Answer D is correct). The sensitivity is calculated by

dividing 189/201, and the specificity is calculated by

dividing 100/200. The PPV is around 65%, and the NPV