Detailed STAT 509 Lesson 3: Overview of Clinical Trial Designs

Comprehensive vocabulary flashcards covering the foundation of clinical trial designs, phases, bias control, and study protocols based on STAT 509 Lesson 3.

Three classical experimental-design features

Control of the experimental process to reduce experimental error; replication to estimate response variability; and randomization.

Control in the corn example

Using the same planter, fertilizer, and weed-control methods in each test plot to control extraneous sources of variability.

Replication in the corn example

Planting multiple plots with each variety to provide repeated experimental units and information about variability.

Randomization in the corn example

Assigning seed variety randomly to test plots.

Clinical trial design vs. classical experiments

Investigators inherit design principles but cannot control as many variability sources as in laboratory or industrial experiments.

Human-response variability

Responses to medical treatment general vary more than responses in genetically identical plants/animals or controlled physical experiments.

Patient accrual and follow-up

The requirement of long recruitment and follow-up periods where subjects usually do not all enter on the same day.

Participant withdrawal

A feature where volunteers can choose to stop participating, not generally present in classical laboratory experiments.

Design and conduct versus analysis

Proper trial setup is more important than statistical choice; flaws here cannot be corrected later, whereas analysis can often be modified.

Two major shortcomings of poor trials

Inaccuracy (bias) and imprecision (large variability) in estimated treatment effects.

Advantage of proper design 1

Allows investigators to satisfy ethical constraints.

Advantage of proper design 2

Permits efficient use of scarce resources.

Advantage of proper design 3

Isolates the treatment effect of interest from confounders.

Advantage of proper design 4

Controls precision.

Advantage of proper design 5

Reduces selection bias and observer bias.

Advantage of proper design 6

Minimizes and quantifies random error or uncertainty.

Advantage of proper design 7

Simplifies and validates the analysis.

Advantage of proper design 8

Increases external validity of the trial.

Precision requirement

A design that separates treatment effects from person-to-person response variability.

Accuracy requirement

A design that minimizes bias to ensure treatment-effect estimates are correct.

Six general trial-design objectives - 1

Quantify and reduce errors due to chance.

Six general trial-design objectives - 2

Reduce or eliminate bias.

Six general trial-design objectives - 3

Yield clinically relevant estimates of effects and precision.

Six general trial-design objectives - 4

Be simple in design and analysis.

Six general trial-design objectives - 5

Provide a high degree of credibility, reproducibility, and external validity.

Six general trial-design objectives - 6

Influence future clinical practice.

Strength of clinical evidence

Depends on the control over bias and variability as well as the magnitude of the observed effect.

Broad clinical-study categories

Uncontrolled observations, observational comparative studies, and controlled clinical trials.

Case report

An uncontrolled observational report demonstrating that a clinical event is possible; used to generate hypotheses.

Case-report limitations

No control over treatment, endpoint ascertainment, confounders, or control group; descriptive only.

PPH case-report example

A patient on a specific anorexic drug developed primary pulmonary hypertension ($1-2$ per million Americans), raising a hypothesis.

Case series

A collection of related cases that cannot establish treatment efficacy.

Selection bias in case reports/series

Both are highly susceptible because patients are not selected through a controlled comparative design.

Laetrile case-series lesson

Belief in a cancer cure based on series was not supported by later rigorous controlled testing or reviews.

Database analysis

Secondary analysis of existing data that resembles a case series but may include a control group.

Bias in treatment comparisons from databases

Likely biased if treatment was chosen by physicians/patients rather than randomized in an open-label setting.

Best use of databases

To explore patterns and generate hypotheses using exploratory statistical analyses.

Interstitial cystitis database example

An NIH-sponsored 1990s analysis of more than $400$ individuals to find potentially effective treatment patterns.

Case-control study

Investigators select cases with disease and controls without, then retrospectively assess prior exposure.

Why relative risk cannot be calculated in case-control studies

Subjects are selected based on outcome status, so sampled disease proportions do not represent population risks.

Recall bias

Systematic differences in the accuracy or completeness of participants' recall of past exposures or events.

Prospective cohort study

Individuals are followed forward in time to determine who develops the outcome or case status.

Cohort risk-factor analysis

Factors are usually measured at baseline and related to subsequent outcomes over time.

Bias control in prospective cohorts

Standardized data collection and endpoint ascertainment can control some bias, but residual confounding may remain.

Prospective versus retrospective studies

Prospective studies have fewer design problems and less bias but require more time and money.

Cardiologist case-control example

$36$ cardiac valve patients matched to two controls by age ($\pm 5$ years) and BMI ($\pm 2.5$) to interview about diet-drug use.

Condition for convincing non-experimental evidence (Natural history)

The natural history of disease with standard therapy or no therapy is known.

Controlled clinical trial

Experimental design where treatments are assigned by design and administration is governed by protocol.

Experimental unit

The unit randomized to a treatment regimen and receiving the treatment directly.

Observational unit

The unit on which measurements are taken.

Community intervention trial

Communities are randomized to treatments (experimental unit), while outcomes are measured on individuals (observational unit).

Reproductive toxicology unit example

Female rodents receiving treatment are experimental units; their pups are observational units.

Factor

A variable controlled and varied during an experiment, such as treatment in a clinical trial.

Two-way factorial clinical trial

A trial studying combinations of levels of two factors, such as doses of two chemotherapeutic agents.

Incomplete factorial design

A factorial design that omits some treatment combinations, such as a double placebo, if they are inappropriate.

Parallel design

Patients are randomized to one treatment and remain on it throughout the trial.

Crossover design

Patients are randomized to a sequence of treatments and switch from one to another over separate periods.

Washout period

A period between crossover treatments used to reduce residual effects of the previous treatment.

Crossover advantage

Each patient serves as their own control, reducing person-to-person variability.

Carry-over effect

Residual effect of a treatment from an earlier period that influences response during a later period.

Selection bias

Systematic treatment-assignment differences occurring when the type of patient receiving one treatment differs from the other.

Blocking

A restriction of randomization designed to balance treatment assignments after a prescribed number of randomizations.

Stratification

Dividing participants into groups based on a characteristic and randomizing within those groups to control unwanted variation.

Age-stratification example

Creating strata for ages $18-30$, $31-50$, and $51-65$, then randomizing to treatments A and B within each group.

Block size 4 example

Ensures that after every 4, 8, or 12 patients in a stratum, equal numbers have been assigned to two treatments.

Placebo effect

Improvement caused by expectation of a positive response rather than active treatment.

True placebo

An inert treatment that mimics the route and appearance of the active treatment, such as a sugar pill.

Active control

An accepted therapy used as a comparison when a placebo is unethical for a serious disease.

Treatment masking / blinding

Keeping treatment identity hidden to preserve objectivity, especially for subjective outcomes.

Double-masked trial

Both investigators and patients are unaware of the treatment assignment.

Single-masked trial

Only patients are unaware of the treatment assignment.

Masking safety monitors

A committee may know groups as "A/B" until a trend requires treatment identity to be revealed.

Breaking the blind

The process of revealing treatment identity for an individual patient during an emergency.

Confounding

When the effect of other relevant factors is incorrectly attributed to the difference between study groups.

Calendar-time confounding example

Assigning the first $10$ patients to treatment on March 1 and the next $10$ to control on March 15.

Internal validity

The observed difference between study groups is real rather than caused by bias, chance, or confounding.

External validity

How well results from a human trial generalize to a broader population.

Large simple trial

Enrolls large numbers with simplified design to detect small treatment advantages across a large population.

Pragmatic trial

Emphasizes treatment effectiveness in ordinary practice outside academic medical centers.

Superiority trial

A trial designed to demonstrate that a new treatment is better than a control.

Noninferiority trial

Aims to show a new treatment is not worse than an accepted treatment by more than a prespecified margin.

Equivalence trial

Aims to show response lies within prespecified margins in both directions relative to the comparison treatment.

Phase 0

Preclinical testing in animals to obtain pharmacokinetic information.

Phase I

Small studies investigating dose levels, safe range, and side effects in a small number of volunteers.

Treatment mechanism (TM) trial

An early-stage study investigating pharmacokinetics / pharmacodynamics or drug bioavailability.

Dose-finding (DF) trial

An early study seeking a target such as the maximum tolerated dose or minimum effective dose.

Maximum tolerated dose (MTD)

The highest dose that can be tolerated within the study's safety framework.

Phase II

Investigates preliminary evidence of efficacy and safety, often in several hundred patients with the disease.

Safety and efficacy (SE) trial

A typical middle-stage development study focusing jointly on safety and efficacy evidence.

Phase III

Rigorous randomized trials with control groups and definitive endpoints, often involving thousands of patients.

Comparative treatment efficacy (CTE) trial

A Phase III-type trial using placebo or active controls to obtain valid estimates of outcome differences.

FDA pivotal-study standard

Approval hinges on well-controlled Phase III studies convincingly demonstrating safety and efficacy.

Phase IV

Post-approval expanded-safety research to detect rare side effects and interactions.

Expanded safety (ES) study

A large post-development safety study potentially involving more than $10,000$ patients.

Translational study

A study preceding large trials where the primary outcome is a biological measurement or target.

Prevention trial types

Tests whether an intervention prevents onset, progression, or additional episodes of established disease.

Gold standard

The reference diagnostic method assumed to be perfectly accurate for comparison in clinical trials.

Protocol

The document specifying the research plan and serves as the primary quality-control tool for a trial.

Manual of Operations (MOP)

A detailed document providing operational instructions beyond the protocol for data collection and procedures.

SPIRIT statement

An international initiative specifying minimal elements and a checklist for clinical-trial protocols.