Clinical trials

Core Concepts

• Definition: An experimental study design where participants are randomly assigned to an experimental group (receives intervention) or a control group (receives placebo or standard of care).
• “Gold Standard”: Considered the highest level of evidence for determining the efficacy and causality of a therapeutic or preventive intervention.
• Objective: To minimize bias when testing the effectiveness and safety of a new intervention.

Key Features

• Randomization:
  • Each participant has an equal chance of being assigned to any group.
  • Purpose: To minimize selection bias and control for confounding variables (both known and unknown) by distributing them evenly among the groups.
  • Allocation Concealment: The process of hiding the next allocated group from the person enrolling the participant, which prevents selection bias during recruitment.
• Control Group:
  • The group that receives a placebo, no treatment, or the current standard of care.
  • Purpose: Provides a baseline to compare the effect of the new intervention against.
• Blinding (Masking):
  • Concealing group allocation from individuals involved in the trial to prevent bias.
  • Single-blind: Only the participant is unaware of their assigned group.
  • Double-blind: Both the participant and the investigators/clinicians are unaware of the group assignments. This is the most common and preferred method.
  • Triple-blind: Participants, investigators, and data analysts are all unaware of group assignments.
  • Purpose: Reduces performance bias (systematic differences in care) and ascertainment/observer bias (distorted outcome assessment).

---

Trial Designs & Phases

• Common Designs:
  • Parallel Group: Each group receives a different treatment simultaneously (most common design).
  • Crossover: Each participant receives both treatments in a sequence, separated by a “washout” period. Participants serve as their own controls.
  • Factorial: Tests two or more interventions simultaneously using various combinations.

Clinical trials

Mnemonic

Does the drug SWIM? ⁣  “Is it Safe?” Assesses safety, toxicity, pharmacokinetics, and pharmacodynamics⁣  “Does it Work?” Assesses treatment efficacy, optimal dosing, and adverse effects.⁣  “Is it as good or better?” Compares the new treatment to the current standard of care (any Improvement?).⁣  “Can it stay?” Detects rare or long-term adverse effects (eg, black box warnings). Can result in treatment being withdrawn from Market.⁣ ⁣

• Phase 0 trial
  • Purpose: Exploratory phase with no therapeutic or diagnostic intent, aimed at gaining insight into pharmacokinetics and pharmacodynamics.
  • Only < ​1% of the therapeutic dose is used.
  • Safety and toxicity are not assessed.
  • Study population: Small sample consisting of either healthy individuals or a population with a disease of interest (∼ 10–15).
  • Study design: Open-label.
• Phase I trial
  • Purpose: Evaluation of pharmacodynamic and pharmacokinetic properties of the drug, safety, and toxicity.
  • Evaluation of the maximum tolerated dose.
  • Study population: Small number of healthy individuals or patients with a specific disease (∼ 15–30).
  • Study design: Open-label.
• Phase II trial
  • Purpose: Evaluation of efficacy, optimal dose range, and side effects (especially common and short-term side effects).
  • Study population: Moderate number of patients with a specific disease (∼ 10–100).
  • Study design: Randomized, Controlled, Anonymized.
• Phase III trial
  • Purpose: Final confirmation of safety and evaluation of efficacy against placebo or the current standard of care.
  • Study population: Randomized control trial with a large number of patients with a specific disease (∼ 100–1000).
  • Study design: Randomized Controlled Trial.
• Phase IV trial
  • Purpose: Postmarketing surveillance, comparing real-life efficacy to that described in research studies. Safety studies following approval (especially evaluation of rare and long-term side effects).
  • Study population: Large number of patients with a specific disease after drug approval.
  • Study design: Open-label.

---

Analysis & Interpretation

• Intention-to-Treat (ITT) Analysis:
  • All participants are analyzed in the groups to which they were originally randomized, regardless of whether they completed or adhered to the treatment protocol.
  • Benefit: Preserves the benefits of randomization and gives a more realistic measure of the treatment’s effectiveness in a real-world setting. Minimizes bias from attrition.
• Per-Protocol Analysis:
  • Only includes participants who strictly adhered to the protocol.
  • Risk: Breaks randomization and is more susceptible to bias, but may provide a better measure of the treatment’s effect under ideal conditions.

---

Potential Biases & Limitations

Selection Bias

Distortion during study enrollment or retention.

• Ascertainment/Sampling: Non-representative sample (e.g., volunteers only).
• Berkson: Hospitalized controls have more comorbidities than the general public.
• Healthy Worker Effect: Working cohort is inherently healthier than the general population.
• Attrition Bias: Differential dropout between arms. $\to$ Fix: Intention-to-Treat (ITT) analysis. c
• Susceptibility (Confounding by Indication): Sicker pts receive the experimental tx. $\to$ Fix: Randomization.

Measurement & Observation Bias

Distortion during data collection or evaluation.

• Recall Bias: Pts with disease recall exposures differently (classic in case-control).
• Observer/Detection: Unblinded evaluator scores groups differently. $\to$ Fix: Blinding.
• Hawthorne Effect: Pts change behavior because they are being observed. c
• Pygmalion Effect: Researcher’s expectancy alters participant performance.

Screening Biases

• Lead-Time Bias: c
  • Screening detects disease earlier $\to$ apparent ↑ survival, but biological lifespan is unchanged.
  • Fix: Measure cause-specific mortality rates, not survival time from diagnosis.
• Length-Time Bias:
  • Screening disproportionately detects indolent, slow-progressing cases with better baseline prognosis.

Confounding vs. Effect Modification

• Confounding:
  • Third variable distorts exposure-outcome link (not on causal pathway).
  • Stratified analysis: Association disappears or becomes equal across all strata.
  • Fix: Matching, restriction, randomization, stratification.
• Effect Modification:
  • Natural biological interaction (third variable modifies exposure effect).
  • Stratified analysis: Association differs significantly between strata. c
  • Note: Not a bias; must be reported, not controlled.

Pattern after stratification	Interpretation
Subgroup RRs are similar to each other, but differ from the crude RR	Classic confounding pattern
Subgroup RRs are meaningfully different from each other	Effect modification