1. (D) Data from variables can be categorical (qualitative) or numerical (quantitative).

Categorical data include (1) nominal data that describe data that can be in categories but have no particular order or magnitude differences (such as blood groups) and (2) ordinal data that are data that can be allocated to an ordered set of categories (such as AHA classes I to IV or severity of AV valve regurgitation from mild to severe).

Numerical data include (1) discrete data that can only be certain whole numbers (such as number of surgeries or catheterizations) and (2) continuous data that can be any numerical value (such as cardiac indices or pulmonary vascular resistances). Pulmonary vascular resistance, therefore, would be numerical data of a continuous nature.

1. Permit the study of rare diseases.

2. Permit the study of diseases with long latency between exposure and manifestation.

3. Can be launched and conducted over relatively short time periods.

4. Relatively inexpensive as compared to cohort studies.

5. Can study multiple potential causes of disease.

1. Information on exposure and past history is primarily based on interview and may be subject to recall bias.

2. Validation of information on exposure is difficult, incomplete, or even impossible.

3. By definition, concerned with one disease only.

4. Cannot usually provide information on incidence rates of disease.

5. Generally incomplete control of extraneous variables.

6. Choice of appropriate control group may be difficult.

7. Methodology may be hard to comprehend for nonepidemiologists and correct interpretation of results may be difficult.

1. Allow complete information on the subject’s exposure, including quality control of data, and experience thereafter.

2. Provide a clear temporal sequence of exposure and disease.

3. Give an opportunity to study multiple outcomes related to a specific exposure.

4. Permit calculation of incidence rates (absolute risk) as well as relative risk.

5. Methodology and results are easily understood by none-pidemiologists.

6. Enable the study of relatively rare exposures.

1. Not suited for the study of rare diseases because a large number of subjects are required.

2. Not suited when the time between exposure and disease manifestation is very long, although this can be over-come in historical cohort studies.

3. Exposure patterns, for example, the composition of oral contraceptives, may change during the course of the study and make the results irrelevant.

4. Maintaining high rates of follow-up can be difficult.

5. Expensive to carry out because a large number of subjects are usually required.

6. Baseline data may be sparse because the large number of subjects does not allow for long interviews.

3. (B) A clinical trial is research involving administration of a test regimen to humans to evaluate both efficacy and safety. The several phases of a clinical trial are (1) phase I—safety and pharmacologic profiles; (2) phase II—pilot efficacy studies; (3) phase III—extensive clinical trial; and (4) phase IV—studies after FDA approval for distribution.

Phase 0—administration of a single subtherapeutic dose of the drug to a small group (10 to 15 subjects) to gather preliminary data on pharmacokinetics and pharmacodynamics; phase I—a small group (20 to 80 subjects) of volunteers to assess the safety and pharmacokinetic profile of the medication; phase II—a large group (20 to 300 of subjects) to assess safety in a larger group of patients as well as effectiveness of the drug; phase III—randomized controlled multicenter trial on a relatively large group (300 to 3,000 or more subjects) depending on the medical condition and to assess the effectiveness of the drug in comparison with an accepted therapy; and phase IV—involves safety surveillance and ongoing technical support of a drug after permission for it to be distributed.

4. (B) The Fisher exact test is used when the numbers in the contingency table of categorical variables are relatively small while the McNemar test is used for two groups with paired data. The Mantel-Haenszel test is an extension of the χ² test used when comparing several two-way tables (such as for meta-analysis studies). χ² test is a measure of the difference between actual and expected frequencies with categorical variables.

5. (E) Parametric tests are used to compare samples of normally (or Gaussian) distributed data. These tests include (1) the Student’s t-test (used to compare two samples to test the probability that the samples come from a population with the same mean value) and (2) the ANOVA (used to compare the means of two or more samples to see whether they are derived from the same population). The analysis of covariance (ANCOVA) is an extension of ANOVA to accommodate continuous variables. Note: The Kolmogorov-Smirnov test is used to test the hypothesis that the collected data are from a normal distribution so that the parametric statistics can be used.

Nonparametric tests are used when the data are not normally distributed so that the above tests are not appropriate. These tests include (1) the Wilcoxon signed-rank test (for comparing the difference between paired groups, as in t-test for parametric data); (2) the Mann-Whitney U-test or the Wilcoxon rank sum test (for comparing two sets of data that are derived from two different sets of subjects); and (3) the Kruskal-Wallis test (for comparing two or more independent groups, as in ANOVA for parametric data).

6. (E) Prevalence of a disease describes what proportion of the population has the disease at a specific point in time (disease burden of a specific population). The prevalence of a disease depends on both the incidence and duration of the disease (prevalence = incidence × duration of disease). Of note, incidence describes the occurrence frequency of new cases during a time period. Prevalence can be considered to be similar to pretest probability; the more disease in the population, the higher likelihood of getting a positive test. Thus, as prevalence increases, PPV increases and NPV decreases. The opposite is true for decreasing prevalence; PPV decreases and NPV increases.

7. (A) Correlation is often confused with “regression,” which quantifies the association between two variables. Regression analysis is used to delineate how one set of data relates to another through a best fit line, in which the regression coefficient is the slope of the line. While this describes a simple linear regression, other types of regression include (1) logistic regression (variation of linear regression when there are only two possible outcomes); (2) Poisson regression (variation of regression calculations to allow for frequency of rare events); and (3) Cox proportional hazards regression model (used in survival analysis to investigate the relationship between an event and several variables).

The most common survival curve method is the Kaplan-Meier curve, which graphically displays the survival of a cohort with calculation of survival estimates upon each death or event (as seen in Figure 17.1). This figure depicts Kaplan-Meier event-free survival curves for arrhythmic events according to the combination of left ventricular ejection fraction (LVEF) with nonsustained ventricular tachycardia and baroreflex sensitivity (BRS). The total population has been divided into four groups after dichotomization of LVEF according to <35% and >35% and BRS and SD of normal intervals according to the ATRAMI cutoff values of <3 ms/mm Hg and >3 ms/mm Hg. The probability value refers to differences in event rate between subgroups. A nonparametric test to compare the survival between two potential Kaplan-Meier curves is the log rank test.

8. (A) The sensitivity and specificity of a diagnostic test depends on more than just the “quality” of the test; they also depend on the definition of what constitutes an abnormal test. Look at the idealized graph in Figure 17.9 showing the number of patients with and without a disease arranged according to the value of the diagnostic test. These distributions overlap; the test (like most) does not distinguish normal from disease with 100% accuracy. The area of overlap between these two distributions represents where the test cannot distinguish normal from disease.

In practice, we choose a cutoff (indicated by the vertical black line) above which we consider the test to be abnormal and below which we consider the test to be normal. The position of the cut point will determine the number of true positive, true negatives, false positives, and false negatives. We may wish to use different cutoffs for different clinical situations if we wish to minimize one of the erroneous types of test results. The arrow points to the false negatives.

9. (A) A normal or Gaussian distribution is a well-recognized curve (Figure 17.10). This reflects a continuous probability distribution that is bell shaped (unimodal) and symmetrical about the mean with two parameters, the mean (µ) and the variance (σ²). The SD is the measure of dispersion or variability in a sample. The SD is used for data that are normally distributed (±1 SD = 68.2%, ±2 SD = 95.4%, and ±3 SD = 99.7% of data). The mean and the median of a normal distribution are equal.

Note: A quick check to see whether a distribution is normally distributed is to see whether two SD away from the mean are still within the possible range for the variable.

The t-distribution is similar to the normal distribution but more spread out with longer tails.

Examples of continuous probability distributions that are not normal include the χ² distribution (a right skewed distribution characterized by degrees of freedom); the F-distribution (also skewed to the right and used for comparing two variances); and the lognormal distribution (highly skewed to the right as it is the probability distribution of a random variable whose log follows the normal distribution). The binomial and Poisson distributions are types of discrete probability distributions.

10. (C) CI is the range that is likely to contain the true population mean value that would be present (if the data for the whole population is obtained). A 95% CI means that there is 95% chance that the population value lies within the stated limits. The SD indicates the variability in a sample. In a normal distribution, 95% of the distribution of the sample means it is within 1.96 SD of the population mean. The SD is the standard error of the mean (SEM) and the 95% CI for the mean is calculated by: sample mean – 1.96 × SEM to sample mean + 1.96 × SEM.

The size of the CI would be related to the sample size of the study (the larger the study population, the narrower the CI). Importantly, when assessing the averages or means of two groups and the CI includes zero, then no difference between the populations is detected (null hypothesis cannot be rejected). If using relative risk or odds ratio rather than means of two groups and the CI includes one, then no difference between the populations is detected (null hypothesis cannot be rejected). Therefore, in this case, the CI includes zero and no true difference can be stated; thus, the P-value would likely be >0.05 (5%).

11. (C) There are two types of applied statistics. Descriptive statistics (means, medians, modes, SD, quartiles, and histograms) describe the data in a sample. Inferential statistics are statistical methods that estimate whether the results suggest a real difference between populations (such as the Student’s t-test, ANOVA, and the χ² test).

12. (E) A type I error (α error) occurs when a null hypothesis that is correct is rejected (declaring that there is a difference when there is not). A type II error (β error) occurs when a hypothesis that is incorrect is accepted (declaring that a difference does not exist when in fact it does). The chance of making a type I error is the same as the P-value. Note: A type III error is a study design that produces the right answer to the wrong question.

The P (probability) value is the probability that defines how likely it is that a hypothesis (usually the null hypothesis) is true (that there is no difference between two treatments). The P-value is therefore the probability of an observed difference occurring solely by chance. The usual P-value at the significance level is 0.01 to 0.05. Note: A method used to adjust the P-value for multiple testing is the Bonferroni adjustment.

13. (B) The power of a study is the probability that it would detect a statistically significant difference. As the β value is the probability of accepting a hypothesis that is false, the power of the study (1 – β) is therefore the probability of rejecting the null hypothesis when it is false. Or in other words, power is the probability of avoiding a type II error. The power of a study should be at least 80% and is increased by several factors including larger significance level, larger effects, decreased variability or variance of the observations, and larger sample size.

14. (A) A CEA is an economic assessment method in which the costs and consequences of alternative interventions are expressed in costs per unit of health outcome. This commonly used methodology is applicable to health programs as well as health services to help determine the preferred action that requires the least cost to produce a given level of effectiveness.

Another economic tool is the CUA, which uses quality-of-life measurements expressed as utilities (such as QALY) in the value equation. A disability-adjusted life year (DALY) is also a measure used but is for the overall “burden of disease.” It quantifies the impact of not only premature death as in QALY but also disability on a population by combining them into a single, comparable metric.

A third economic assessment methodology is the CBA, which seeks to translate all relevant healthcare considerations into monetary terms by analyzing economic and social costs of medical care and benefits of reduced loss of net earnings due to prevention of premature death or disability.

Other less common methods of economic evaluation include cost-consequence analysis (CCA), CMA, and even CVA.

15. (C) A meta-analysis is a technique in which results from a number of studies that are similar in nature are gathered to give one overall estimate of the effect. The formal steps include the following: (1) decide on effect of interest, (2) check for statistical homogeneity, (3) estimate the average effect of interest with CIs, and (4) interpret the results and present the findings (forest plot). The advantages include refinement and reduction, efficiency, generalizability and consistency, reliability, and power and precision. The disadvantages include publication bias, clinical heterogeneity, quality differences, and lack of independence of study subjects.

A systemic review (such as the international network called the Cochrane collaboration with its Cochrane database of systematic reviews) often uses meta-analysis techniques to render well-informed clinical decisions; it is an essential part of evidence-based medicine. Major disease categories will often have a sufficient number of randomized clinical trials for a meta-analysis to be carried out to determine the value of such an intervention.

20. (A) The ROC curve is a two-way plot of the sensitivity (true-positive rate) against 1 – the specificity (false-positive rate) for different cutoff values for a continuous variable in a diagnostic test². The upper left corner or coordinate (0,1) is called the perfect classification (100% sensitivity or no false negatives, and 100% specificity or no false positives) (Figure 17.11).

21. (E) The variance is the square of the SD while coefficient of variation is the ratio of the SD to the mean. While the SD is a measure of spread away from the mean and is equal to the square root of the variance, the SEM is a measure of precision of the sample mean or how close the sample mean is likely to be to the population mean.

22. (B) Accuracy is the degree of closeness of measurements to that quantity’s true value while precision is the reproducibility of a study result with the study to be repeated under the same circumstances (measured by standard error of measurement) (Figure 17.12).

23. (A) Chi-squared (χ²) test is a measure of the difference between actual and expected frequencies with categorical variables; a contingency table is set up to calculate the χ² value. If there is no difference between actual and expected frequencies, then χ² would be 0. The larger the difference, the bigger the χ² value (but it is easier to note the P-value that accompanies the χ² value). The df is the number of independent comparisons that can be made between members of the sample and is used with χ² value to calculate the P-value. In this case, the df (of 1) is needed to calculate the P-value. The χ² test is sometimes used with Yates continuity correction to improve the accuracy of the P-value.

The Fisher exact test is used when the numbers in the contingency table of categorical variables are relatively small (<5) while the McNemar test is used for two groups with paired data. The Mantel-Haenszel test is an extension of the χ² test used when comparing several two-way tables (such as for meta-analysis studies).

24. (A) Correlation coefficient is the strength of the linear relationship between two variables and it measures the strength of an association between two variables without assuming cause-and-effect. This relationship is denoted by the letter r that ranges from -1 to +1 (R² is sometimes given to correct for negatively correlated relationships); 1 = strong positive relationship, -1 = strong negative relationship, and 0 = no relationship. The coefficient r cannot be calculated when there is neither a nonlinear relationship nor when there are multiple outliers.

When the degree of linear relationship is extended to several variables, it is known as multiple correlation coefficient. The Pearson correlation coefficient “r” is used if the values are sampled from a normally distributed population (if not, the Spearman correlation coefficient “rs” is used).

29. (B) Bias occurs when there is a systematic difference between the result of a study and the true result. It occurs when a systematic error is introduced into the study which results in selecting one outcome over the other. Types of bias include: observer bias (observer inaccurately assesses variable), confounding bias (spurious association), selection bias (selected study subjects not representative of true population), information bias (measurements incorrectly recorded), publication bias (only positive results are published), and others like recall bias (inaccurate remembering of facts or omitting details by subjects), and/or allocation bias (systematic difference in how subjects are assigned to comparison groups).

30. (E) An association is any relationship between two measured quantities that relates them to be statistically dependent, whereas correlation defines a linear relationship between the two quantities. Causation in addition to association includes the following criteria: temporality, strength of causality, dose-response, repetition in a different population, consistency with other studies, and biologic plausibility.

31. (A) The Belmont report elucidates three principles of research ethics: (1) respect for persons: protecting the autonomy of all people and treating them with courtesy and respect and allowing for informed consent (researchers must be truthful and conduct no deception); (2) beneficence: the philosophy of “do no harm” while maximizing benefits for the research project and minimizing risks for the research subjects; and (3) justice: ensuring reasonable, nonexploitative, and well-considered procedures are administered fairly and equally.

32. (C) The IRB, also known as the ethical review board, is a committee that is designated to approve and review research involving human subjects to protect the rights and welfare of human research subjects. The DSMB is an independent group of experts that continuously monitors the data from various aspects of a clinical trial to ensure patient safety as well as validity and scientific merit of the trial. The difference between the IRB and the DSMB is that the IRB is primarily responsible for the review of clinical protocols and related documents while the DSMB‘s main responsibility is to review the trial safety and efficacy data.

33. (C) Reliability is the consistency of a set of measurements/measurement tool, or the repeatability/reproducibility of such a methodology (inversely related to random error). Validity is the extent to which the study measures what it is intended to measure so that validity is a measurement of systematic error or bias (examples are confounding and selection bias). Accuracy is the degree of a measurement closeness to that quantity’s true or accepted value; while precision is the closeness of two or more measurements to each other when repeated under the same circumstances (measured by standard error of measurement). Power is the probability of correctly rejecting the null hypothesis when the null hypothesis is false.

34. (A) Calculation of a minimal sample size involves the following parameters: power (usually 0.80); significance level (usually 0.01 or 0.05); variability of the observations (or the standard deviation); and the smallest effect of interest (the standardized difference).

1. Allow complete information on the subject’s exposure, including quality control of data, and experience thereafter.

2. Provide a clear temporal sequence of exposure and disease.

3. Give an opportunity to study multiple outcomes related to a specific exposure.

4. Permit calculation of incidence rates (absolute risk) as well as relative risk.

5. Methodology and results are easily understood by nonepidemiologists.

6. Enable the study of relatively rare exposures.

1. Not suited for the study of rare diseases because a large number of subjects are required.

2. Not suited when the time between exposure and disease manifestation is very long, although this can be overcome in historical cohort studies.

3. Exposure patterns, for example, the composition of oral contraceptives, may change during the course of the study and make the results irrelevant.

4. Maintaining high rates of follow-up can be difficult.

5. Expensive to carry out because a large number of subjects are usually required.

6. Baseline data may be sparse because the large number of subjects does not allow for long interviews.

1. Permit the study of rare diseases.

2. Permit the study of diseases with long latency between exposure and manifestation.

3. Can be launched and conducted over relatively short time periods.

4. Relatively inexpensive as compared to cohort studies.

5. Can study multiple potential causes of disease.

1. Information on exposure and past history is primarily based on interview and may be subject to recall bias.

2. Validation of information on exposure is difficult, or incomplete, or even impossible.

3. By definition, concerned with one disease only.

4. Cannot usually provide information on incidence rates of disease.

5. Generally incomplete control of extraneous variables.

6. Choice of appropriate control group may be difficult.

7. Methodology may be hard to comprehend for nonepidemiologists, and correct interpretation of results may be difficult.

36. (A) Nonparametric tests are used when the data are not normally distributed. These tests include (1) the Wilcoxon signed-rank test (for comparing the difference between paired groups, as in the t-test for parametric data or normally distributed); (2) the Mann-Whitney U-test or the Wilcoxon ranks sum test (for comparing two sets of data that are derived from two different sets of subjects); and (3) the Kruskal-Wallis test (for comparing two or more independent groups, as in ANOVA for parametric data or normally distributed).

37. (B) See Table 17.7.

38. (A) Categorical data include (1) nominal data that describe data that can be in categories but have no particular order or magnitude differences (such as single ventricle and biventricular surgical strategies or antiarrhythmic agent for supraventricular tachycardia), and (2) ordinal data are data that can be allocated to an ordered set of categories (such as severity of AV valve regurgitation from mild to severe).

Numerical data include (1) discrete data that can only be certain whole numbers (such as number of reinterventions after Norwood procedure) and (2) continuous data that can be any numerical value (such as blood pressure measurements before and after ACE inhibitors).

39. (B) A cohort study (also termed follow-up, longitudinal, or prospective study) is a prospective observational study with study subjects (cohort) assigned to an exposure or condition category and then all followed for a defined observation period to see whether they develop disease. A historical cohort study, as the name implies, is a group of patients from the past and would not involve active enrollment of new study subjects.