Biomedical Literature Terms
NOTE: Images can be enlarged when clicked on.
Hierarchy of Study Designs
- The process by which any or all of the clinicians, patients or participants, outcome assessors, or statisticians were unaware of who received which study intervention.
Single Blind Study:
- A trial in which only one set of individuals (i.e. patients) were unaware of what the patients were receiving.
Double Blind Study:
- A trial in which two sets of individuals (i.e. patients and clinicians) were unaware of what the patients were receiving.
Triple Blind Study:
- A trial in which ALL individuals (i.e. patients, clinicians, and assessors) were unaware of what patients were receiving.
- When more than one placebo was used to help the treatments look the same in all groups.
- Important when there are differences in dosing and administration (i.e. the intervention is oral and the comparator is given via an IV infusion).
- All participants in the trial (clinicians, patients or participants, outcome assessors, and statisticians) were aware of who received which study intervention.
- Also known as an unblinded study.
- Outcomes that represent disease or symptoms
- Direct measures of how a person feels, functions, or survives and are expected to predict the effect of therapy
- Indirect measures of clinical endpoints that correlates with clinical outcomes
- Needs to be substantial evidence of casual relationship between change in the surrogate and change in clinical outcome
- Outcomes that capture the number of patients experiencing one or more of several outcomes
- Individual components must be clinically meaningful and of similar importance
- The results of composite endpoints and individual components should be similar.
- The degree to which a measurement truly assess what it is intended to measure
- There are 2 types of validity: internal and external
- The number of observations with a given characteristic divided by the total number of observations in a given group
- “Parts” divided by the “whole”
- Commonly reported as percentages
- Similar to proportions with two added properties:
- Rates are computed over a specific time-period (for example, per year)
- Rates typically use a multiplier (for example 1,000, 10,000 or 100,000) which is called the base
- Measures the number of new cases of a disease (or symptom or problem) that develop in a population at risk within a given period of time
- Measures the probability of having a disease at a point in time
- Reflects existing disease within a population
Performance Measures for Diagnostic Tests
- The ability of a diagnostic test to correctly identify individuals with disease
- The proportion of patients with the target disorder who have a positive test result
- The ability of a diagnostic test to correctly identify individuals without disease
- The proportion of patients without the target disorder who have a negative test result
Positive Predictive Value:
- The probability that a patient has the disease given that a positive test result was obtained
Negative Predictive Value:
- The probability that a patient does not have the disease given that a negative test result was obtained
Common Statistical Tests
Tests for Continuous Data:
- Student’s t-test
- Paired t-test
- > 2 samples with covariates
- Repeated measures ANOVA
- Multiple measures in the same individual but under different conditions or multiple time periods
Tests for Ordinal Data:
- Mann-Whitney U/Wilcoxon rank sum
- Wilcoxon signed rank
- Kruskal Wallis
Tests for Nominal Data:
- Chi-Square ()
- Fisher’s Exact
- 2 independent samples when the total sample size is < 20 or expected cell frequency for any cell is < 5
- Mantel-Haenszel (Regression)
- Research design whereby the researcher controls the treatment (independent variable) through randomization and determines its impact on the clinical outcome (dependent variable).
- An INTERVENTION is introduced.
Randomized Clinical Trial (RCT):
- Involves randomization of intervention(s) to two or more groups.
- The RCT is considered to be the gold standard in evaluating the safety and efficacy of an intervention.
- Can determine a cause-and-effect relationship between an item under investigation and an outcome.
- Studies that look like experimental studies but that LACK RANDOMIZATION.
Parallel Study Design:
- A study design that ensures each subject is randomized to either a treatment group or a control group only.
- A study where two or more groups receive different treatments and the outcomes are compared.
Crossover Study Design:
- A study design that ensures each subject receives all of the interventions based on a specified sequence.
- A study where each subject receives all study treatments, and endpoints during the various treatments are compared.
Factorial Study Design:
- Trials that are designed to evaluate multiple interventions in a single experiment.
- Factors to be studied can include multiple dose levels and multiple drug regimens.
- Trial that seeks to determine whether a new therapy is no worse than a standard therapy by some prespecified margin
- Typically have fewer patients than superiority trials.
- May be used when a placebo group is not ethically allowed.
- A prespecified amount of effect used to show that the test drug's treatment effect is not worse than the reference drug by more than this specific degree
- A study where the investigator analyzes naturally occurring events.
- An INTERVENTION is NOT introduced.
- Types of observational studies include: cohort, case control, cross sectional, and case report/case series.
- A study design whereby groups of similar groups of individuals (cohorts) that differ with respect to the factors being studied are recruited and followed FORWARD in time to determine how these factors affect the occurrence of the outcome of interest in the groups.
- A prospective cohort STARTS in the PRESENT and moves forward into the FUTURE.
- A type of cohort study that involves the use of previously collected (historical) data to identify exposure status and occurrence of outcome in the study groups.
- A retrospective cohort STARTS in the PAST and assesses the PRESENT.
- A study design that compares the frequency of exposure among cases that experience an outcome and controls who do not have the outcome.
- A retrospective trial design (often using medical charts or cases) strictly based on the observation of a group of people that have experienced a similar outcome. The study is used to determine the possible exposure these patients have had that would lead to this known outcome.
- Case-control studies START in the PRESENT and look back in the PAST.
- A trial design involving data collection only once in members of the study population that represents a snapshot in time. These members are not required to be studied all at once, but each member can be studied at a different time. The key is that the data collected represents a specific period in time.
- Study that examines population characteristics at a cross section (one point) in time.
- A brief report of clinical characteristics or course from a single clinical subject or event without a comparison.
- A descriptive study that consists of a group of patients who have been diagnosed with the same condition or are following similar procedures over a period of time.
- A grouping of records (case studies or case reports) that documents a practitioner's experiences, thoughts, or observations related to the care of multiple patients with similar medical situations.
- Quantitative synthesis of data derived from individual studies (usually three or more) identified through a systematic review process.
- Results of previously conducted similar clinical trials are combined, statistically analyzed, and new data are created for interpretation.
- Especially useful when previous studies are inconclusive or controversial.
- Useful when the sample size of multiple similar studies are too small to detect a statistically significant difference, but combining them will provide adequate sample size to meet a set power.
- A summary of previously conducted studies where the research to be included in the review is systematically identified; however, the results are not statistically combined as would occur with a quantitative systematic review or meta-analysis.
- Also called a qualitative systematic review.
- A review article that summarizes previously conducted research, but does not provide a description of the systematic methods used to identify the research included in the article.
- Tertiary resource that reviews a specific topic but differs from a systematic review in that the methodology is not based on a structured, predefined literature search strategy.
- Also called a narrative review.
Null Hypothesis (Superiority Study):
- The no-difference hypothesis
- Assumes equality among study treatments
- Probability that the result was due to chance
- Indicate statistical significance if it is less than a specified significance level (, alpha)
- Typically p<0.05 indicates statistical significance
- Probability of a Type I Error
Confidence interval (CI):
- Estimates the precision of the results in order to make generalizations to a larger population
- Usually reported as 95% CI
- The range of values within which we can be 95% sure that the true value for the whole population lies
- CI = 1-
- Probability of rejecting the null hypothesis given the null hypothesis is actually false
- Power = 1-(beta)
- Typically, = 0.2 or 0.1 (which translates to power levels of 80% and 90%, respectively)
- Setting an appropriate power and maintaining it throughout a study limit the risk of Type II Error
Type I Error:
- Occurs when a meaningful difference between groups is observed for an outcome when that difference actually does not, or would not, exist in the population
- False Positive
- Rejecting the null hypothesis when the null hypothesis is actually true
Type II Error:
- Occurs when a meaningful difference between groups is not observed for an outcome when that difference actually does, or would, exist in the population
- False Negative
- Failing to reject the null hypothesis when the null is not true
Covariates or Confounders:
- Differences between the two groups within a study that can directly alter the difference in effect between the intervention and control group
- If these are unbalanced between the treatment and comparison groups, can lead to selection bias
Relative Risk (RI):
- The probability of an event occurring in the treatment or exposure group versus the control group
- The likelihood of an event in the treated group vs. the control group AT THE END of the trial
- Generally expressed as a decimal, but can appear as a percentage
- RR = 1: no difference in risk between the two groups
- RR < 1: Fewer events are occurring in the treatment group compared to the control group
- RR > 1: More events are occurring in the treatment group compared to the control group
Relative Risk Reduction (RRR):
- Can be calculated when there is a DECREASED risk of an event occurring in the treatment or exposure group compared to the control group (RR<1)
Relative Risk Increase (RRI):
- Can be calculated when there is an INCREASED risk of an event occurring in the treatment or exposure group compared to the control group (RR>1)
- The difference in the rate of events for one arm of a trial compared to the other arm
Absolute Risk Reduction (ARR):
- The absolute arithmetic difference in event rates
- If the treatment or exposure group has a SMALLER proportion of patients having the outcome, there is a DECREASE in the risk of the outcome occurring, or an ARR.
- ARR = (% control group risk) – (% treatment group risk)
Absolute Risk Increase (ARI):
Number Needed to Treat (NNT):
- Represents the number of people who would need to be treated with the intervention for a certain period of time (e.g. one year) in order to achieve the desired outcome (e.g. prevent adverse event) in ONE patient.
- Number of patients who need to be treated to prevent 1 additional bad outcome
- Number of patients who must receive the experimental treatment to create 1 additional improved outcome in comparison with the control treatment.
- ANYTHING greater than a whole number must ALWAYS be ROUNDED UP!
- Three conditions MUST be met to calculate the NNT
- The result MUST be statistically significant
- The result MUST be clinically relevant
- The endpoint should be a yes/no response or nominal data
Number Needed to Harm (NNH):
- Represents the number of people who would need to be treated with the intervention for a certain period of time (e.g. one year) to see ONE patient harmed.
- Number of patients who, if they received the experimental treatment, would lead to 1 additional person being harmed compared with patients who receive the control treatment
- ANYTHING greater than a whole number must ALWAYS be ROUNDED DOWN!
- Three conditions MUST be met to calculate the NNH
- The result MUST be statistically significant
- The result MUST be clinically relevant
- The endpoint should be a yes/no response or nominal data
- The ratio of the odds of an event occurring in the treatment group compared to the odds of an event occurring in the control group
- It is a measure of the association between an exposure and an outcome
- OR = 1: no difference in odds between the two groups
- OR < 1: the odds of the outcome occurring with exposure compared to the control group are DECREASED
- OR > 1: The odds of the outcome occurring with exposure compared to the control group are INCREASED
- The chance of an unfavorable event occurring by a given point in time during the study in the treatment group compared to the comparator group
- Used in clinical trials with survival analyses
- Reported with Kaplan Meier Curves
- The likelihood of an event in the treated group vs. control group AT ANY GIVEN POINT in time during the trial
- HR = 1: no difference in odds between the two groups
- HR < 1: at any given time, relatively fewer patients in the treatment group have had an event compared to the control group
- HR > 1: at any given time, relatively more patients in the treatment group have had an event compared to the control group
This "Glossary of Common Terms" has been created from the following resources:
1. ACP Journal Club Archives. American College of Physicians website. http://acpjc.acponline.org/shared/ glossary.html#Theraupeutics1. Accessed April 1, 2020.
Copyright © 2020 American College of Physicians. Adapted with permission.
2. Aparasu RR, Bentley JP. eds. Principles of Research Design and Drug Literature Evaluation. 2nd ed. New York, NY: McGraw-Hill; 2020.
3. Malone PM, Malone MJ, Park SK. eds. Drug Information: A Guide for Pharmacists. 6th ed. New York, NY: McGraw-Hill; 2018.
4. Gabay M, ed. The Clinical Practice of Drug Information. Burlington, MA: Jones & Bartlett Learning; 2016.