HYMR Biostatistics

Question 1

```

~~~

Which of the following best describes the proper use of an odds ratio (OR)?
A. Estimates the relative risk in retrospective studies

B. Estimates the absolute risk reduction in prospective studies

C. Use of the odds ratio in prospective studies will underestimate the risk

D. An odds ratio is used in a prospective study to show the strength of the relationship between 2 variables

Answer 1

Answer: A - Estimates the relative risk in retrospective studies

• OR is an estimate of the relative risk (RR) since the number of patients at risk for the condition is not always known, thus preventing the calculation of incidence
• OR are used in retrospective studies (ex. case-controlled studies), RR are used in prospective studies

Question 2

A group of investigators have designed a study to determine if ezetimibe (Zetia) was more effective than cholestyramine (Questran) for the treatment of hyperlipidemia. They designed the study so that 500 patients would be randomly assigned to one of 2 groups: ezetimibe 10 mg once daily or cholestyramine 4 g by mouth twice a day with meals. At baseline, patients had fasting lipid profile done. These patients were then prospectively followed for 6 months. At the end of the study (6-months) the patients had a follow up visit where fasting lipid profiles completed. The primary endpoint was to determine the change in LDL-c (mg/dL) from baseline. The secondary endpoints included: 1-The proportion of patients developing cholestasis or needing a cholecystectomy. Which of the following statistical tests would you recommend for answering the primary objective (assume sample or population studied is normally distributed)?

A. Student’s t-test
B. Paired t-test
C. Chi Square
D. ANOVA

Answer 2

A. Student’s T-test

Question 3

Which of the following best describe alpha?

A. The level or chance of making a Type II error

B. The ability to predict the effect of one variable on another

C. The level or chance of making a Type I error

D. The degree of external validity

Answer 3

C. The level/chance of making a Type I error

Type 1 error occurs when the study/investigators say there is a difference between two studied interventions, but in reality there is no difference at all

The alpha value is determined a-priori and is most commonly set to 0.05 (meaning there is a 5% chance the study will make a Type 1 error)

Answer choice A refers to beta. As beta decreases the power of the study increases and the chance of making a Type II error decreases.

Question 4

Which of the following provides the best interpretation of a p=0.001 (assuming alpha= 0.05)?

A. I am 99.9% confident that the results of the is study are clinically significant

B. I am 0.1% confident that the results of this study are clinically significant

C. There is a 99.9% chance that the results of this study are due to chance

D. There is 0.1% chance that the results of this study are due to random error

Answer 4

D. There is 0.1% chance that the results of this study are due to random error

*P-value NEVER indicates “clinical” significance.
*P-value ONLY helps you know whether the results that were found in the study were due to random error or due to chance alone
*P-value tells you how much of the sample data supports the null hypothesis to be true (ie are results due random error or chance alone)

Question 5

True/False: A “p-value” helps the investigator determine the degree of clinical significance found in the study?

A. True

B. False

Answer 5

B. False

Question 6

Which one of the following best defines Power= 1- Beta?

A. The probability that you will make a Type 1 error

B. The probability that you NOT make a Type II error

C. To estimate the percentage of baseline risk that was removed because of the treatment used

D. The strength of the relationship between 2 variables

Answer 6

B. The probability that you NOT make a Type II error

• The first answer choice is what an alpha value determines. Note: most
  clinicians/researchers will accept a 5% chance (or alpha = 0.05) that when we do a study that the results we found are not true or accurate.

*The second answer choice is correct because of the word “not” in the answer. Since Beta (B) by definition is the probability of a Type II error, then 1-Beta is the probability of NOT making a Type II error. Note: the larger the Beta the lower the power of study and greater chance of making a Type II error. Type II error occurs when a study states there is no difference between the groups assessed but in reality there is a difference. Any time a study fails to find a statistical difference between two groups the greater the chance that a Type II error has occurred.

• The third answer choice reflects the relative risk evaluation.
• The last answer choice reflects the correlation coefficient (r).

Question 7

A simple linear regression can be utilized to predict the outcome of the following?

A. There is a single independent and single dependent variable

B. There is more than one independent and only one dependent variable

C. There is a single dichotomous independent variable

D. There are two unrelated independent variable

Answer 7

A. There is a single independent and single dependent variable

*Regression analysis is a mathematic description or equation that provides “predictability” of one variable on another variable
*For simple linear regression analysis, this usually analyzes the relationship of two variables (1 being an independent variable and other being dependent variable)

*Multi-linear regression: relationship between >1 independent variable and 1 dependent variable

Question 8

A prospective, clinical trial comparing two treatment regimens and their ability to prevent the development of a DVT was done in post-operative patients who underwent a total hip replacement (THR). Patients were randomized to receive either warfarin at 5 mg by mouth daily (n=500) starting the day of surgery or enoxaparin 40 mg SC once daily (n=500) starting the day of surgery. Both treatment options were used for a total of 5 consecutive days. Fifty patients receiving warfarin developed a DVT, whereas only 25 patients receiving enoxaparin developed a DVT at the two week follow up evaluation. What is the absolute risk reduction for patients receiving enoxparin in this study?

A. 0.05

B. 0.10

C. 0.15

D. 0.25

Answer 8

A. 0.05

• First you must calculate the relative risk for each group. The warfarin group: 50/500 = 0.1. The enoxparin group: 25/500 = 0.05 (note this is half of the warfarin group, no calculation even needed).
• Now, subtract the two to find the difference giving an absolute risk reduction (ARR) = 0.1-0.05 -0.05.

*ARR is the difference in risk of an outcome between patients who received treatment and those of another treatment (or control group).

*Note: ARR is used to calculate number needed to treat (NNT), NNT = 1/ARR or 1/0.05= 20. You would need to treat 20 post-THR patients w/ enoxaparin for 5 days to prevent 1 patient from getting a DVT.

Question 9

Which of following best describes a 95% confidence interval?

A. The width of the confidence intervals is not dependent on the standard error of the mean (SEM)

B. It provides you the predictive properties of a test

C. If you were to repeat the study under ideal conditions, you would be 95% confident that your result would fall within that confidence interval.

D. It helps you to determine if your result was due to chance or random error

Answer 9

C. If you were to repeat the study under ideal conditions, you would be 95% confident that your result would fall within that confidence interval.

*The wider the confidence interval, the more likely that the “true” population value will be anywhere within that interval. HOWEVER, a wide confidence interval may indicate a problem w/ the study or data obtained. One of the main reasons for a wide CI is insufficient sample size.

*A: dependent and influenced by the standard error of the mean (SEM)
B: reflects of positive or negative predictive value for a test result
D: describes what a p-value is helping you to determine

Question 10

See Q10 picture.
A clinical trial was completed and a review of the data from the study is being analyzed to determine the right statistical test that should be used. Based on the graphic representation of the data, which of the following statements is most accurate?

A. The data is considered to be homogenous and can be analyzed by using parametric statistical analysis

B. The mean is greater than the median

C. The data is negatively skewed because the median is larger than the mean

D. The data is positively skewed because the median is smaller than the mean

Answer 10

C. The data is negatively skewed because the median is larger than the mean

Question 11

Which of the following statistical tests would be considered non-parametric?

A. 1-way ANOVA

B. Student’s T-test

C. Fisher’s Exact

D. Paired T-test

Answer 11

C. Fisher’s Exact

Question 12

See Q12 table

Question 13

A group of investigators have designed a study to determine if rosuvastatin (Crestor) was more effective than atorvastatin (Lipitor) for the treatment hyperlipidemia. They designed the study so that a 1000 patients would be randomly assigned to one of 2 groups: rosuvastatin 40 mg once daily or atorvastatin 40 mg once daily. At baseline, patients had fasting lipid profile done and were assessed for myopathy (muscle aches/pain). These patients were then prospectively followed for 3 months. At 3 months the patients had a follow up visit where fasting lipid profiles and patient assessments on myopathy (none, mild, moderate, or severe pain) were completed. The primary endpoint was to determine the change in LDL-c (mg/dL) from baseline. The secondary endpoints included: The proportion of patients achieving their LDL-c goal per guidelines and the patient’s rating of myopthy. What type of data is the primary endpoint?

A. Nominal

B. Ordinal

C. Continuous

D. Categorical

Answer 13

C. Continuous

• This question is specifically asking about the primary endpoint and thus you should focus on that specific wording in the case.
  In this case, the investigators specifically want to know the change in LDL-C or the amount of lowering in the LDL-c in mg/dL. This is a concentration and each mg/dL is the same in magnitude and thus considered to be continuous (note: the change can be to infinity, there is no ranking or scale here).
  
  • If the case had said the proportion of patients who achieved an LDL-c goal of < 100 mg/dL, then the investigators would be looking at this endpoint as a “yes” or “no” question (ie., the patient either achieved an LDL-c goal of less than 140 or they did not), thereby making it nominal or categorical.
  • If the investigators had defined the LDL-c into a sense of order or ranking such as <100 (mild), 100-130 (moderate), or > 130 (severe) then the endpoint would be treated as ordinal data.

High-Yield Core Concept:
* It is always important to read the endpoint in question carefully to see the subtle difference in wording as it changes everything as it relates to the type of data it is assigned. Board exams like to do this so make sure you understand this concept.

High-Yield Fast Fact:
* Nominal data is considered non-numerical data but is also sometimes referred to as “binary” or “dichotomas” data since it is either one thing or another (i.e., two options). For example, the use of mortality as an endpoint of the study. At the end of the study the subject is either dead or alive. There is no in-between.

Question 14

A prospective randomized controlled trial has set out to determine if warfarin caused more GI bleeds than placebo This study contained 2 groups independent of each other. Group A was assigned to take warfarin 2.5 mg by mouth once a day Group a was assigned to take placebo by mouth once a day. After 12 months of therapy, the incidence of Gl bleeds in Group A was 0.41 and Group B it was 0.28. Which of the following best describes the results of this data?

A. The relative risk is 0.68

B. About 32% of the risk for GI bleed was from the use of warfarin

C. The relative risk is 0.53

D. There was a 46% excess risk in GI bleeds with the use of warfarin

Answer 14

D. There was a 46% excess risk in GI bleeds with the use of warfarin

*This question requires that you calculate the relative risk (RR) which is the incidence of
the exposed group divided by the incidence non-exposed group. Therefore the RR
would be 0.41/0.28 = 146.
* Since the RR is greater than 1, the difference from or above “1.0’ is the “excess risk”,
which is 0.46 or 46%.

High-Yield Core Concept:
The relative risk (RR) is determined by calculating the incidence of the exposed group
divided by the incidence non-exposed group

Question 15

A prospective trial has set out to determine if warfarin (Coumadin) caused more minor bleeding than placebo his study contained 2 groups independent of each other Group A took warfarin 2.5 mg by mouth daily. Group a took placebo by mouth once daily After 12 months of therapy, the incidence of bleeding in Group A was 0.41 and Group a it was 0.28. Which of the following best describes the results of this data?

A. The relative risk is 0.68

B. 32% of the risk for bleeding was from the use of warfarin

C. The relative risk is 0.53

D. There was a 46% excess risk in bleeding with the use of warfarin

Answer 15

D. There was a 46% excess risk in bleeding with the use of warfarin

*They gave you the incidence for the events in each group studied. Now all you have to do is calculate the relative risk (RR), which is the incidence of Group A divided by the incidence of Group B (RR = 0.41/028 = 1.46).

*A RR of 1 means there is no difference in risk at all. If the RR is greater 1 then there is an excess risk with that intervention (i.e., use of warfarin in this case).

• If the RR is less than 1, the intervention of interest will take away risk of the outcome.
  Therefore, a RR of 1.46 means that patients taking warfarin have a 46% excess risk
  compared to those taking placebo. Do not make the mistake of calculating the absolute risk reduction (AAR) on accident which is determined by 0.41 - 028 = 13.

Question 16

Which of the following can have a significant impact on the power of a study?

A. Sample Size

B. Ethnicity of the subjects enrolled

C. Location of the study

D. Type of blinding method used

Answer 16

A. Sample Size

*Most common factors that impact the power of a study include: sample size and the a-priori setting of the alpha and beta values

Question 17

A prospective study was done over 3 years to determine if a new antihypertensive (Drug A) would offer any benefit over placebo for the treatment of hypertension. 500 patients were randomized to one of 2 groups: Group A (250 patients total) Drug A 50 mg by mouth once daily or Group B (250 patients total) placebo by mouth once daily. The primary endpoint was the development of a stroke At the end of the trial, 88 patients in Group A had a stroke and 95 patients in Group B had a stroke What would be an appropriate statistical test for analyzing the primary endpoint of this study?

A. Fisher’s Exact

B. Chi-Square

C. Students t-test

D. 2-way ANOVA

Answer 17

B. Chi-Square

Question 18

A study was completed and revealed a p-value between the two treatment groups to be p = 0.01. Assuming an alpha = 0.05, which of the following offers the best interpretation of this p value?

A. There is a 5% chance the study has made a type Il error

B. There is a 1% chance that the study results are due to random error

C. There is a 5% chance that the study results are due to chance alone

D. There is a 0.01% chance that the results are due to chance alone

Answer 18

B. There is a 1% chance that the study results are due to random error

A p-value only indicates the chance the results that were found are due to chance alone or random error. Thus the smaller the p-value, the more likely the results are real the p-value has nothing to do with *clinical significance’ or *clinical relevance”. That is determined by the researcher and/or reader of a study.

Question 19

When the data reveals a mean less than the median less than the mode it can be described as being?

A. Normally distributed

B. Skewed to the left

C. Skewed to the right

D. Positively skewed

Answer 19

B. Skewed to the left

• Being “skewed to the left” is the same as saying the data is ‘negatively skewed”.
  
  • To clarify further, when they say skewed to the left, this means the mean is less than the median and/or the tail of the curve is on the left side of the graph as identified in the image below.

*Thus, the opposite is true, if the data is skewed to the right the data is called positively skewed.
* Here the mean is now greater than the median (acting is an outlier) and thus skews the data to the right (e.g., where the “tail” of the curve would then be on the right side of the graph and look opposite to the image below)

*You should be able to quickly look at a graph like this and be able to interpret it any of the above explanations (or listed answer choices).

Question 20

The HOPE trial was a prospective, randomized controlled trial where patients were
randomized to either receive ramipril or placebo The results revealed that the relative risk for death from non-cardiovascular causes was 1.03 (95% C’: 0.85 - 1.26). Which of the following conclusions can be made from this result?

A. The p value will be less than 0.05

B. The p value will be greater than 0.05

C. Ramipril provides statistically significant reductions in death from noncardiovascular events

D. Ramipril provides a greater degree of clinical significance than placebo

Answer 20

B. The p value will be greater than 0.05

When evaluating a 95% confidence interval for a relative risk or risk ratio or hazards ratio, if the confidence crosses through and includes 1.00 in the range it cannot be statistically significant and thus the p-value would be greater than 0.05. You will not know what the exact p-value by looking at the 95% Cl, but you can know it is greater than 0.05.

Question 21

Dr Smart is interested in studying a new treatment (RABI-MAb) for rabies in dogs. A side effect
is that some of the dogs developed edema (swelling or fluid accumulation). He decides to
conduct a randomized double-blind placebo controlled study with 100 dogs in the control
group and 100 dogs receiving RABI-MAD 25 dogs in the placebo group develop edema and
50 dogs in the active group developed edema What is the excess relative risk (%) in causing
edema?

A. 25%

B. 50%

C. 100%

D. 200%

Answer 21

C. 100%

*This is easily done by setting up your 2 x 2 table and filling in the blanks so that you can
make the calculation accurately and most efficiently.
* RR = 0.5/0.25 = 2.0. Take the 2.0 and since 1-2 = 1 x 100 = 100%

Question 22

What is the primary difference between using a Kaplan-Meier and the Cox Proportional Hazards method of survival analysis?

A. None, the two are equivalent

B. The Kaplan-Meier method uses actual time of event occurrence where the Cox Proportional Hazards Method uses set time intervals

C. Cox Proportional Hazards Model allows for the control of continuous extraneous variables while Kaplan-Meier does not

D. All of the above are false

Answer 22

C. Cox Proportional Hazards Model allows for the control of continuous extraneous variables while Kaplan-Meier does not

*A Kaplan-Meier curve is used to assess the “time to an event.” Many times this “event” is death or mortality, but in reality it can be any event.

*The use of a Cox Proportional Hazards Model allows the investigators to control for confounding variables or factors that may be influencing the endpoint of their study in addition to the intervention (or independent variable) being studied

Question 23

A hospital medicine team needed to determine whether adding clopidogrel to aspirin would improve survival in a 70-year-old patient with acute coronary syndrome (ACS). Clinicians assessed the clinical relevance of results from a trial that compared the safety and efficacy of clopidogrel plus aspirin compared to clopidogrel alone in patients with ACS for 9 months. How should these results be interpreted?

See Question 23 Table

A. Treating 48 patients with ACS for 9 months with aspirin will prevent 1 death due to stroke, MI, or CV causes

B. Treating 100 patients with ACS for 9 months with clopidogrel plus aspirin will cause one major bleed

C. Treating 100 patients with ACS for 9 months with clopidogrel plus aspirin will prevent 1 death due to stroke, MI, or CV causes

D. NNT and NNH are not useful parameters to determine clinical value of study results (7%)

Answer 23

C. Treating 100 patients with ACS for 9 months with clopidogrel plus aspirin will prevent 1 death due to stroke, MI, or CV causes

• Number needed to treat (NNT) and number needed to harm (NNH) are clinically useful measures that help clinicians determine how much risk and benefit are present when a similar group of patients is treated with a medication or intervention.
• NNT is the inverse of the absolute risk reduction, while NNH is the inverse of the absolute risk increase.
• In this example, the absolute risk reduction of 2.1% between treatment groups means that treating 48 acute coronary syndrome (ACS) patients for 9 months with clopidogrel plus aspirin will prevent 1 death due to stroke, myocardial infarction (MI), or CV causes. Additionally, the absolute risk increase of 1% between treatment groups means that treating 100 patients for 9 months clopidogrel plus aspirin will cause one major bleed.

High-Yield Core Concept:
* The NNT is the inverse of the absolute risk reduction (ie., 1 divided by the absolute difference between the incidence of both groups), while NNH is the inverse of the absolute risk increase.

Question 24

A scatter plot (used for determining a correlation coefficient) depicts the relationship between a dependent and independent variable and is used to determine:

A. Clinical significance

B. Statistical significance

C. Association

D. Predictability

Answer 24

C. Association

• A correlation analysis generates a correlation coefficient (r), which can range from being -1 to +1.
• It is done to describe the strength of the “relationship” between 2 variables whereas regression analysis is done to provide the “predictability” of one variable on another variable.
• The closer to +1, the stronger the “correlation” or “relationship between 2 variables. It is also important to know that correlation does NOT imply anything about causation, thus making this question false.
• Lastly, the correlation (r) does not have the ability to determine which of the two variables came first in existence to influence the other.

Question 25

A group of investigators have designed a study to determine if simvastatin (Zocor) was more effective than pravastatin (Pravachol) for the treatment high cholesterol. They designed the study, so that 1000 patients would be randomly assigned to one of 2 groups: Zocor 40 mg once daily or Pravachol 40 mg once daily. At baseline, patients had fasting lipid profile done and were assessed for myopathy (muscle aches/pain). These patients were then prospectively followed for 3 months. At 3 months the patients had a follow up visit where fasting lipid profiles and patient assessments on myopathy (none, mild, moderate, or severe pain) were completed. The primary endpoint was to determine the change in LDL-c (mg/dL) from baseline. The secondary endpoints included: 1-The proportion of patients achieving their LDL-c goals per guidelines; 2-The patients rating of myopathy. Which of the following statistical tests should you use when analyzing secondary endpoint number 1 (the proportion of patients achieving their LDL-c goal)?

A. McNemar Test

B. Chi Square

C. Wilcoxin Rank Sum

D. Sign Test

Answer 25

B. Chi Square

Question 26

A prospective study was done over 3 years to determine if a new antibiotic (Drug A) would offer any benefit over placebo for the treatment of meningitis. 500 patients were randomly assigned to one of two groups: Group A (250 patients total) 2 g IV once daily x 7 days or Group B (250 patients total) placebo IV once daily x 7 days. The primary endpoint was overall mortality. At the end of the trial, 110 patients in Group A died and 120 patients in Group B died. What is the number to treat (NNT) for this study?

A. 25 patients

B. 11 patients

C. 1 patient

D. 2 patients

Answer 26

A. 25 patients

Always set up a 2 x 2 table so that you can calculate your relative risks (RR). You will need the absolute relative risk in order to calculate the number needed to treat (NNT).

Step 1. RR for Group A: 110/250 = 0.44

Step 2. RR for Group B: 120/250 = 0.48

Step 3. Calculate the ARR = RR-RR = 0.44 -0.48 = 0.04

Step 4. To calculate the NNT, take the reciprocal of ARR: so NNT = 1/0.04 (move decimal over two and get 100 on top, so 100/4 = 25

Therefore, you would have to treat 25 patients each for a period of 7 days with antibiotic Drug A in order for 1 person to be saved from dying due to sepsis. If each day 7 course of Drug A cost the system $150 per day, then the system would have to spend (7 days x $150/day = $1,050/patient: 25 patients x $1,050/patient = $26,250). Therefore, the system would have to spend $26,250 in order to prevent one person from dying.

Question 27

Which of the following values of a correlation coefficient is the best indicator of the predictive power of the correlation?

A. r = 0.3
B. r = 0.6
C. r = 0.9
D. None of the above

Answer 27

D. None of the above

• A correlation analysis generates a correlation coefficient (r), which can range from being -1 to +1.
• It is done to describe the strength of the “relationship” between 2 variables whereas regression analysis is done to provide the “predictability” of one variable on another variable.
  Since the question specifically used the words “correlation coefficient” and “predictive” power, none of the above is the only option because predictability is done with a regression analysis.
• All of the other answer options are correlation coefficients.

Question 28

A study was completed and revealed a p-value between the two treatment groups to be p = 0.01. Assuming an alpha = 0.05, which of the following offers the best interpretation of this p- value?

A. There is a 5% chance the study has made a type II error

B. There is a 1% chance that the study results are due to random error

C. There is a 5% chance that the study results are due to chance alone

D. There is a 0.01% chance that the results are due to chance alone

Answer 28

B. There is a 1% chance that the study results are due to random error

• A p-value only helps you to know whether the results that were found in the study were due to random error or due to chance alone.
• In order to interpret this p-value fully you have to convert it into a percentage. To do this all you have to do is multiple it by 100 or move the decimal point over to the right by two spaces. Therefore, the p-value of 0.01 is the same as 1%. Therefore based on the definition of a p-value above, there is a 1% chance that the study results are due to random error or chance.
  *The alpha value is the level or chance an investigator could be making a type I error. As a reminder, a type I error occurs when the study or investigators say there is a difference between two studied interventions, but in reality there is actually no difference at all. Its use in this case is simply basic information and not meant to be something to focus on.

Question 29

Which of the following is a common parametric statistical test used when studying the effect of an intervention in two independent groups?

A. Chi square test

B. Student’s t-test

C. Fisher’s exact test

D. ANOVA

Answer 29

B. Student’s t-test

• The chi square and Fisher’s exact test that are used when studying 2 independent groups, but are considered to be nonparametric statistical tests for nominal data.
• The ANOVA is a parametric statistical analysis when evaluating 3 or more groups (this question asked you about “two” independent groups.
• That leaves the Student’s t-test, which fits all criteria in the question.

*The student t-test is the most common statistical test used when comparing the effect of a continuous variable between two independent groups

Question 30

The HOPE trial was a prospective, randomized trial where patients were randomized to either receive ramipril or placebo. The results (comparing the incidence in the ramipril group to placebo group) revealed that the relative risk for death from noncardiovascular causes was 1.03 (95% CI: 0.85 - 1.26). Which of the following conclusions can be made from this result?

A. The p value will be less than 0.05

B. The p value will be greater than 0.05

C. Ramipril provides statistically significant reductions in death from noncardiovascular events

D. Ramipril provides a greater degree of clinical significance than placebo

Answer 30

B. The p value will be greater than 0.05

• This is a high-yield concept that you must know. All this question really needs you to do is simply look at the 95% confidence interval (CI) and interpret it.
• This 95% Cl is in reference to relative risk (as stated in the case) and as such, a relative risk equal to 1 means there is NO difference between the two treatments. Therefore, if the 95% CI contains “1.0” in the range provided, it cannot be statistically significant and thus would have a p-value > 0.05. This is the case here. The 95% CI of 0.85 to 1.26 includes the value “1.0” within that range and thus the p-value would be > 0.05 (you may not know what the exact p-value is, but it does not matter - it is just > 0.05.
  If the 95% CI had been 0.85 - 0.98, then it would not have contained “1.0” and thus would be considered “statistically significant or having a p-value of < 0.05.
• The case does NOT have to give you the p-value for you to answer this question. Simply looking at the 95% CI tells you whether the p-value is > 0.05 or < 0.05.
• As it relates to the final answer choice. An interpretation of a 95% CI and the level of p- value generated NEVER EVER indicates clinical significance. If you ever see the words *clinically significant” or “clinical significance” anywhere in the answer choice for a question pertaining to the interpretation of a p-value it is absolutely wrong. This is the most common trick question as too many people think that p-values reflect clinical significance because they fail to understand this concept. A p-value ONLY helps you to know whether the results that were found in the study were due to random error or due to chance alone.

High-Yield Core Concept:
A 95% CI for a relative risk (RR) cannot be statistically significant it crosses through and includes “1.00” within the interval range.

Question 31

Investigators collected data assessing efficacy of migraine abortive therapy in 6- to 12-year- old children (n = 132) compared to adolescents (n = 107) with an acute migraine. Treatment outcomes are summarized in the table below. How should these results be interpreted?

See Q31 Table: Efficacy of Migraine Abortive Therapy in Pediatric Patients (6-12 years of age)
Medication

A. Fewer children responded to a triptan as abortive therapy for migraine

B. Both codeine and ibuprofen are significantly better in managing acute migraine attacks in children compared to adolescents

C. Adolescents responded significantly better than children to ibuprofen for acute migraine management

D. Codeine response was comparable in children and adults as abortive therapy for migraine

Answer 31

A. Fewer children responded to a triptan as abortive therapy for migraine

• There were fewer children in this study who responded to sumatriptan compared to the number of adolescents.
  *When interpreting outcomes data, p-values less than 0.05 are statistically significant.
• Additionally, when addressing relative risk (RR) or odds ratios (OR), a 95% confidence interval that crosses “1.00” is NOT statistically significant.
• Consequently, results from this trial show that codeine is significantly better in aborting migraine in children compared to adolescents as the p-value is < 0.05 and the 95% CI does not cross 1.00.

High-Yield Core Concept:
* Proper interpretation of relative risk, risk ratios, hazard ratios, and odds ratios including evaluating the interval of the 95% or 99% confidence interval. If that 95% CI cross over or through 1.00 in the range, it cannot be statistically significant (ie., the p-value cannot be < 0.05)

Question 32

A group of investigators have designed a study to determine if ezetimibe (Zetia) was more effective than cholestyramine (Questran) for the treatment of hyperlipidemia. They designed the study so that 500 patients would be randomly assigned to one of 2 groups: ezetimibe 10 mg once daily or cholestyramine 4g by mouth twice a day with meals. At baseline, patients had fasting lipid profile done. These patients were then prospectively followed for 6 months. At the end of the study (6-months) the patients had a follow up visit where fasting lipid profiles completed. The primary endpoint was to determine the change in LDL-c (mg/dL) from baseline. The secondary endpoints included: 1-The proportion of patients developing cholestasis or needing a cholecystectomy. Which of the following best describes the study groups in this trial?

A. One-sample case

B. Two independent samples

C. Related samples

D. Paired samples

Answer 32

B. Two independent samples

• The study design described above clearly states that the patients enrolled in this study were randomly assigned to one of “two” groups.
• Also they go one to further that one group was to get ezetimibe and another groups was given cholestyramine.
• As such, there are 2 independent groups being evaluated here.
• Furthermore, they never said that the patients switch arms of the study as seen in a cross-over study where the groups would now be considered related samples since the patients would have served as their own controls from being in both arms of the study.

High-Yield Core Concept:
* The study design outlines what type of groups are present in the study. This is very important for determining which statistical test should be used to analyze the data obtained from that study.

Question 33

When calculating power, the reduction in Beta will likely result in what type of change to the sample size needed?

A. Increase

B. Decrease

C. No change

D. Beta is not used the calculation of Power

Answer 33

A. Increase

• Since Beta (B) by definition is the probability of a Type II error, then Power = 1-Beta is the probability of NOT making a Type II error.
• Note: the larger the Beta the lower the power of study and greater chance of making a Type II error.
• Therefore, the smaller the Beta the larger the power and less chance of making a type II error, results in the study that needs more patients or larger sample size in order to find a difference if it in fact does exist.

High-Yield Core Concept:
*Type II error occurs when a study states there is no difference between the groups assessed but in reality there is a difference.
*Any time a study fails to find a statistical difference between two groups the greater the chance that a Type II error has occurred.

Question 34

The HOPE trial was a prospective, randomized trial where pts were randomized to either received ramipril or placebo. The results (comparing the incidence in the ramipril group to placebo group) revealed that the relative risk from myocardial infarction, stroke, and death from cardiovascular causes was 0.78 (95% CI: 0.70 - 0.86). Which of the following conclusions provides the best interpretation of this result?

A. 22% of the risk for having a myocardial infarction, stroke, or death from cardiovascular causes was increased by the use of ramipril

B. 22% of the risk for having a myocardial infarction, stroke, or death from cardiovascular causes was removed by the use of ramipril

C. 78% of the risk for having a myocardial infarction, stroke, or death from cardiovascular causes was increased by the use of ramipril

D. 78% of the risk for having a myocardial infarction, stroke, or death from cardiovascular causes was removed by the use of ramipril

Answer 34

B. 22% of the risk for having a myocardial infarction, stroke, or death from cardiovascular causes was removed by the use of ramipril

• In this case, they tell you that the relative risk (RR) is 0.78, which is less than 1.
• Any RR less than 1 means that the intervention studied “removed” or “prevented” the risk of the study endpoint from happening.
  *The amount of risk removed from the patient is the difference of the RR from 1 (also known as relative risk “reduction”.
• Therefore, 1-0.78 = 0.22 or 22% (if you multiply 0.22 x 100 to get the percent).
• Therefore, the second answer option is the only one that explains that correctly.

High-Yield Core Concept:
*The risk of the event after the experimental treatment as a percentage of the original risk
* Relative Risk = Risk Ratio = RR

High-Yield Fast Fact:
* Relative risk Calculated for prospective studies including retrospective cohort studies (whereas an Odds Ratio is done for case controlled studies to estimate risk).

Question 35

True/False: A type 1 error occurs when a study finds that there is a statistically significant difference in two interventions, but in reality they are not different.

A. True

B. False

Answer 35

A. True

Question 36

Which of the following key assumptions must be met for a t-test to be used?

A. Nominal data

B. Categorical data

C. Ordinal data

D. Continuous data

Answer 36

D. Continuous data

• Several general rules must be met before a t-test can be appropriately used.
  
  • The first is having an adequate sample size (usually > 30-40 patients - depends on what you read).
  • The second includes having data that is both continuous and following a normal distribution.
  • If the data is skewed or has significant outliers then a Mann-Whitney U test should be done instead.

High-Yield Core Concept:
* A t-test (or historically known as “student’s t-test) is a parametric statistical analysis that can be used when making comparisons of ratio or interval (ie., continuous) data endpoints between two independent groups that meet all assumptions.

Question 37

A study of a randomized parallel design allocating patients (n = 100) to either enalapril or quinapril, and evaluating their effects on plasminogen activating inhibitor (PAI)-1 production (measured as ng/ml). Which statistical test should be utilized?

A. Sign Test

B. Paired t-test

C. Student’s t-test

D. Chi square

Answer 37

C. Student’s t-test

• The question describes a typical randomized, clinical trial where patients are randomly assigned to one of two groups.
• Since they never say the patients cross over in the study to other intervention, you must assume that they do not and thus are 2 independent groups (ie., the patients in each group or intervention are not the same people - they are independent of each other).
  *The dependent variable in question here is the effect on “PAI-1” production in a concentration of ng/ml. As such they are treating this data variable as “continuous (or ratio) data where the magnitude of difference between each ng/ml is exactly the same and can on for infinity (for example, the difference between 20 and 21 ng/ml is 1 ng/ml and is the same as the difference between 45 ng/ml and 46 ng/ml; the magnitude of difference is always 1 and is not changing and it is ‘measurable” unlike with ordinal or nominal data).
• Therefore, with 2 independent groups analyzing continuous data, a student’s t-test should be used.

Question 38

Which of the following best describe beta (B)?

A. The chance you have made a Type I error

B. The chance you have made a Type II error

C. The degree of clinical significance between two treatment groups

D. The degree of statistical significance between two treatment groups

Answer 38

B. The chance you have made a Type II error

• Beta by definition is the probability of a Type II error, then 1-beta is the probability of NOT making a Type II error. Note: the larger the beta the lower the power of study and greater chance of making a Type II error.
• A type II error occurs when a study states there is no difference between the groups assessed but in reality there is a difference. Any time a study fails to find a statistical difference between two groups the greater the chance that a Type II error has occurred. The alpha value is the level or chance an investigator could be making a type I error. As a reminder, a type I error occurs when the study or investigators say there is a difference between two studied interventions, but in reality there is actually no difference at all.
• The third and fourth answer options are distracters and have nothing to do with the definition of beta.

High-Yield Core Concept:
* Power is 1 - beta and determines the chance of not making a type II error.

Question 39

A p value of 0.001 is more ____ significant than a p-value of 0.01?

A. Statistically

B. Clinically

C. Neither a or b

Answer 39

C. Neither a or b

A p-value NEVER indicates “clinical” significance

Question 40

What type of data is usually analyzed by parametric statistical analysis?

A. Nominal data

B. Binomial data

C. Ordinal data

D. Continuous data

Answer 40

D. Continuous data

• Parametric statistical analysis assumes that the data generated from a sample of the general population follows a normal distribution (or does not have data points that are outliers where they can skew the curve).
• It also assumes that enough patients are available in the study (usually at least 30-40 patients for a sample size).
• Continuous data is the only data listed where an actual mean, median, and mode can be determined.
• Ordinal data should be analyzed by nonparametric statistical tests and can only be best described with a median and mode, not a mean (due to presence of outliers that shew the data).
  *Nominal data is also nonparametric and can only be described with mode, not median or a mean.

High-Yield Core Concepts:
* Continuous data (also commonly known as interval or ratio data) is known as numerical data endpoints that are measurable where the magnitude of difference between
numbers is the same.
* Ratio data (a type of continuous data endpoint) is different from interval data in that it has an absolute zero.
* Continuous (interval or ratio) data follows a normal distribution and is analyzed using
parametric statistical analysis.

High-Yield Fast Fact:
* Examples of ratio data include temperature and pulse because they have an absolute
0.0.

Question 41

Which of the following best represents the definition of survival analysis?

A. Approximates the effect of an intervention on the incidence of a categorical dependent variable in comparison to another group

B. Tests the independence of a nominal dependent variable and nominal independent variable

C. Approximates the probability of an event (dependent variable) occurring as a function of time

D. Compares the variability of two observed means to determine the risk for chance

Answer 41

C. Approximates the probability of an event (dependent variable) occurring as a function of time

• A Kaplan-Meier curve is used to assess the “time to an event.
• Many times this “event” is death or mortality, but in reality it can be any event.
  *We call it a curve, because the graphical representation of the even in relation to time generates two or more curves depending on how many independent variables are being studied.

High-Yield Core Concept:
* Kaplan-Meier curves are most commonly known as survival analysis where the lengths of the horizontal lines on the X-axis of serial times represent the survival duration for that interval and when the interval is terminated by the occurrence of the event of interest.

Question 42

True/False: The underlying assumption for nonparametric data is that the population being studied follows a normal distribution and is typically used with continuous data.

A. True

B. False

Answer 42

B. False

• This is false because only parametric data is normally distributed and where continuous data is normally assessed using parametric statistically tests such as a student’s t-test. * Nonparametric data is not homogenous and does not follow a normal distribution. Many of times there are data points that are considered to be outliers thus skewing the mean or average when compared to the median.
• Data that is skewed from an outlier is not homogenous and thus should be treated as not normally distributed.
• Sometimes continuous data will not be normally distributed due to sample size or outliers and is thus evaluated using nonparametric statistically analysis instead of the preferred parametric statistical analysis.

Question 43

A group of investigators have designed a study to determine if simvastatin (Zocor) was more effective than pravastatin (Pravachol) for the treatment high cholesterol. They designed the study, so that 1000 patients would be randomly assigned to one of 2 groups: Zocor 40 mg once daily or Pravachol 40 mg once daily. At baseline, patients had fasting lipid profile done and were assessed for myopathy (muscle aches/pain). These patients were then prospectively followed for 3 months. At 3 months the patients had a follow up visit where fasting lipid profiles and patient assessments on myopathy (none, mild, moderate, or severe pain) were completed. The primary endpoint was to determine the change in LDL (mg/dL) from baseline. The secondary endpoints included: 1-The proportion of patients achieving their LDL goals per guidelines: 2-The patients rating of myopathy (aka muscle pain; categorized as mild, moderate, severe). Which of the following statistical test should you use when analyzing secondary endpoint number 2 (the patient’s rating of myopathy)?

A. Mann-Whitney U

B. Chi-square

C. Paired t-test

D. Student’s t-test

Answer 43

A. Mann-Whitney U

• Since the endpoint as dependent on the patient’s “rating” of the level of myopathy they experienced (mild, moderate or severe) there is a sense of “ranking” or an “order” to the endpoint (i.e., mild is not as bad as moderate and moderate is not as bad as severe, however, the difference between mild and moderate may not be the same magnitude of difference between going from moderate to severe myopathy.
• Said another way, the magnitude of difference between each point is not the same and cannot be measured, thus making it ordinal data.
• Since the study design indicated having 2 independent groups and it is looking at an ordinal data variable, the correct statistical analysis would be a non-parametric statistical test such as Mann-Whitney U.

Question 44

When examining a set of data that used a continuous (ratio or interval) variable you find that the mean, median, and mode are approximately equal. In that case you expect that the date is more likely to be:

A. Parametric

B. Nonparametric

C. Neither

Answer 44

A. Parametric

• An important key word is used in the question - “continuous (or ratio or interval) data variable
• A continuous variable that generates data where the mean, median, and mode are very similar or equal, means that it follows a normal distribution or is homogenous. In this situation, there are no significant outliers that are skewing the data.
• As such, it would be considered to follow the rules of being parametric where parametric statistical analysis would be used.
• It is important to note that “ordinal” data does not have a mean, but instead a median and mode and that “nominal” or “categorical” data generates only a mode.
  Knowing this can help you also narrow down on the question’s answer.

Question 45

A randomized, placebo controlled clinical trial was conducted to evaluate the ability of a new lipid lowering medication to be able to achieve a desired goal of at least a 30% reduction in the LDL cholesterol when compared to baseline levels. What type of data is this primary endpoint based on the comparison being made in the study?

A. Nominal data

B. Ordinal data

C. Interval data

D. Continuous data

Answer 45

A. Nominal data

See table

Question 46

True/False: A correlation between two variables implies that the dependent variable is the cause for the observed change?

A. True

B. False

Answer 46

B. False

• A correlation analysis generates a correlation coefficient (r), which can range from being -1 to +1.
• It is done to describe the strength of the “relationship” between 2 variables whereas regression analysis is done to provide the “predictability” of one variable on another variable.
• The closer to +1, the stronger the “correlation” or “relationship between 2 variables.
• It is also important to know that correlation does NOT imply anything about causation, thus making this question false.
• Lastly, the correlation (r) does not have the ability to determine which of the two variables came first in existence to influence the other.

High-Yield Core Concept:
* Correlation analysis is used to determine the strength of the “relationship” between 2 variables, whereas regression analysis provides the “predictability of one variable on another variable.

Question 47

Which one of the following best defines or describes the definition of power in a study?

A. The probability the study will make a Type I error

B. The probability the study will not make a Type II error

C. To estimate the percentage of baseline risk that was removed because of the treatment used

D. The strength of the relationship between 2 variables

Answer 47

B. The probability the study will not make a Type II error

• By definition, power is 1 - beta.
• Beta is known as the chance of making a type 2 error.

High-Yield Core Concept:
* Power is 1 - the chance of making a type 2 error or said another way, power is the chance of NOT making a type 2 error.

High-Yield Fast Fact:
.A type 2 occurs when a study shows no difference between groups study, but in reality. a difference is truly present.