- Start with exposed and non-exposed individuals
- Compare rates of disease in exposed and non-exposed
- Calculate Relative Risk or Odds Ratio
- Ideal when exposure of interest is rare

- Start with individuals with and without the disease
- Compare proportions of individuals with the exposure of interest with those without the exposure of interest
- Calculate the Odds Ratio only (cannot calculate RR)
- Ideal when the disease is rare

Alpha (α) Maximum P value to be considered statistically significant; the risk of committing a type I error

Alpha error Type I error

Alternative hypothesis Hypothesis considered an alternative to the null hypothesis; usually the alternate hypothesis is that there is an effect of the studied treatment on the measured variable of interest; also called test hypothesis

Beta (β) Risk of committing a type II error

Beta error Type II error

Null hypothesis Hypothesis that there is no effect of the studied treatment on the measured variable of interest

Power Probability of detecting a treatment effect the size of the treatment effect sought (i.e., obtaining a P value <α), given alpha, and the sample size of the clinical trial; power = 1 −β

P value Probability of obtaining results similar to those actually obtained, if the null hypothesis were true

Type I error Obtaining a statistically significant P value when, in fact, there is no effect of the studied treatment on the measured variable of interest; also called false positive

Type II error Not obtaining a statistically significant P value when, in fact, there is an effect of the treatment on the measured variable of interest that is as large as or larger than the effect the trial was designed to detect; also called false negative

Student's t-test Used to test whether the means of measurements from two groups are equal, assuming that the data are normally distributed and that the data from both groups have equal variance

Wilcoxon rank sum test (Mann-Whitney test) Used to test whether two sets of observations have the same distribution; similar in use to the t-test, but does not assume the data are normally distributed

Chi-square test Used with categoric variables (two or more discrete treatments with two or more discrete outcomes) to test the null hypothesis that there is no effect of treatment on outcome; assumes at least five expected observations of each combination of treatment and outcome under the null hypothesis

Fisher's exact test Used similar to chi-square test; may be used even when fewer than five observations are expected in one or more categories of treatment and outcome

One-way analysis of variance (ANOVA) Used to test the null hypothesis that three or more sets of continuous data have equal means, assuming the data are normally distributed and that the data from all groups have identical variances; may be regarded as a t-test for three or more groups

Kruskal-Wallis test Nonparametric test analogous to one-way ANOVA; no assumption is made regarding normality of the data; may be regarded as a Wilcoxon rank sum test for three or more groups

Bias Tendency to yield a result that lies to one side or the other of the true value; an error that is not centered around zero

Critically appraised topic (CAT) Summary of the results of an analysis of the evidence relevant to a particular clinical question

Likelihood ratio, negative (LR−) Probability that a negative test result would occur in a patient with the disease in question divided by the probability that a negative test result would occur in a patient without the disease in question (generally <1)

Likelihood ratio, positive (LR+) Probability that a positive test result would occur in a patient with the disease in question divided by the probability that a positive test result would occur in a patient without the disease in question (generally >1)

Negative predictive value Fraction of patients with a negative test result who, in fact, do not have the disease in question, depends on disease prevalence

Number needed to treat (NNT) Number of patients who must be treated with the better of two treatments so that one additional “good” outcome is obtained

Odds Probability of an outcome divided by the probability of the outcome not occurring

Odds ratio Odds of the outcome of interest occurring in one group of patients divided by the odds of the outcome of interest occurring in another group (case-control or cohort study)

Positive predictive value Fraction of patients with a positive test result who, in fact, do have the disease in question, depends on disease prevalence

Post-test odds Odds of the outcome of interest (e.g., particular disease being present), once a particular test result (positive or negative) is known to have occurred; equal to the pretest odds times the likelihood ratio for the test result obtained

Post-test probability Probability of the outcome of interest once a particular test result is known to have occurred

Pretest odds Odds of the outcome of interest before the test result is known

Pretest probability Probability of the outcome of interest before the test result is known

Relative risk Probability of an outcome in one group divided by the probability of the same outcome in another group (cohort study, a case-control cannot have a RR)

Reliability Degree with which a clinical trial or diagnostic test is likely to yield consistent results when repeated

Sensitivity Fraction of patients with the disease in question who have a positive test result

Specificity Fraction of patients without the disease in question who have a negative test result

Validity, external Degree to which the results from a study accurately reflect what would happen to similar patients who were not enrolled, including patients at other locations

Validity, internal Degree to which a study accurately measures the outcomes for the enrolled study population