| Adaptive design |
A trial that modifies itself based on accumulating data according to pre-specified rules |
| Alpha (\(\alpha\)) |
The maximum acceptable probability of Type I error; typically 2.5% or 5% |
| Active control |
The current standard-of-care treatment used as a comparator |
| Bayesian |
Statistical approach that directly estimates the probability a treatment works, given observed data; naturally updates as evidence accumulates |
| Bias (systematic error) |
A flaw causing treatment groups to differ in ways unrelated to treatment, distorting results |
| Biomarker |
A measurable biological indicator (e.g., genetic mutation, protein level) that can predict treatment response |
| Control arm |
The comparison group receiving placebo or standard of care |
| Effectiveness |
How well a treatment works in routine clinical practice |
| Efficacy |
The treatment’s ability to produce beneficial effects under controlled trial conditions |
| Endpoint |
The outcome measured to determine treatment effect (e.g., survival, tumor response, symptom improvement) |
| Estimand |
A precise definition of the treatment effect being estimated, including how to handle intercurrent events |
| External validity (generalizability) |
Whether conclusions apply to broader patient populations |
| Fixed design |
A traditional trial where sample size is set before the trial begins and cannot change |
| Frequentist |
Statistical approach that interprets probability as the frequency of outcomes over many repeated trials; focuses on controlling error rates |
| Intent-to-Treat (ITT) |
Analyzing all patients according to their original randomization, regardless of protocol adherence |
| Intercurrent events |
Events occurring after randomization that affect interpretation (e.g., discontinuation, rescue medication) |
| Interim analysis |
An analysis conducted partway through a trial, before all patients have completed |
| Internal validity |
Whether the trial’s conclusions are correct for the patients actually studied |
| Investigational arm |
The group receiving the new therapy being tested |
| Minimum clinically important difference (MCID) |
The smallest treatment effect worth detecting—below which approval would not change practice |
| Null hypothesis |
The default assumption that there is no treatment difference |
| p-value |
The probability of observing results as extreme as those seen, assuming the null hypothesis is true |
| Placebo |
An inactive substance given to the control group |
| Population |
The entire universe of patients with the condition of interest who might receive the drug if approved |
| Power |
The probability of correctly detecting a true treatment effect; typically set at 80% or 90% |
| Precision |
The degree of certainty in an estimate; higher precision means smaller standard error |
| Precision medicine |
Treatments targeted to patients with specific biological characteristics, rather than one-size-fits-all approaches |
| Probability |
A number between 0 and 1 (or 0%–100%) quantifying how likely something is to occur; 0 = impossible, 1 = certain |
| Random error |
Natural fluctuation in results due to studying a sample rather than the entire population |
| Randomization |
Using a chance mechanism to assign patients to treatment arms, eliminating systematic differences |
| Sample |
The subset of patients actually enrolled in the trial |
| Sample size |
The number of patients enrolled; balances Type I error, power, and detectable effect size |
| Standard error |
A measure of the expected magnitude of random error in an estimate |
| Statistical significance |
A result is “significant” when the p-value falls below the \(\alpha\) threshold, suggesting the effect is unlikely due to chance |
| Treatment arms |
The groups being compared in a trial |
| Treatment effect |
The true difference in outcomes between investigational and control treatments |
| Type I error (false positive) |
Concluding an ineffective drug works; regulators limit this to \(\leq 2.5\%\) (one-sided) or \(\leq 5\%\) (two-sided) |
| Type II error (false negative) |
Failing to detect that an effective drug works |
| Underpowered trial |
A trial with too few patients to reliably detect a true treatment effect |