Types of Clinical Trials - Park's Textbook of Preventive and Social Medicine

Broad Classification

A. TYPES OF RANDOMIZED CONTROLLED TRIALS

1. Clinical Trials

• Concerned with evaluating therapeutic agents, mainly drugs
• Examples: beta-blockers in reducing cardiovascular mortality post-MI; folate supplementation to prevent neural tube defects; aspirin on cardiovascular mortality; efficacy of tonsillectomy for recurrent throat infection; coronary bypass surgery for prevention of MI
• Not all clinical trials are susceptible to blinding (e.g., tonsillectomy - it is obvious who received surgery)
• Many ethical, administrative and technical problems are involved
• Should be conducted before any new therapy, procedure or service is introduced

2. Preventive Trials

• Trial of primary preventive measures - to prevent or eliminate disease on an experimental basis
• Most frequently: trials of vaccines and chemoprophylactic drugs
• Basic principles of experimental design are applicable
• May require application to groups of subjects rather than individuals
• Classic example: MRC UK trial (1946) of whooping cough vaccine - 3,801 vaccinated vs 3,757 unvaccinated children; attack rate 1.45 vs 6.72 per 1000 child-months (significant difference)
• Analysis must address: (a) benefit to community (b) risks involved (c) costs to health service

3. Risk Factor Trials

• A type of preventive trial - investigator intervenes to interrupt the usual sequence in disease development in individuals with a "risk factor"
• Involves risk factor modification
• Can be "single-factor" or "multi-factor" trials (both are complementary)
• Example: WHO trial on primary prevention of CHD using clofibrate to lower serum cholesterol - >15,000 men, 3 European centres (Edinburgh, Prague, Budapest), double-blind, mean observation 9.6 years. Result: significant reduction in non-fatal cardiac infarction BUT 25% more deaths in clofibrate group due to long-term toxic effects
• Other examples: Stanford Three Community Study, North Karelia Project (Finland), Oslo Study, MRFIT (USA)

4. Cessation Experiments

• Attempt to evaluate the termination of a habit or removal of a suspected agent causally related to disease
• If removal is followed by significant reduction in disease, hypothesis of cause is greatly strengthened
• Classic example: Cigarette smoking and lung cancer - one group continues smoking, other gives up; if incidence of lung cancer decreases in the study group, the causal relationship hypothesis is strengthened
• A large RCT has been conducted to study smoking cessation in primary prevention of CHD

5. Trial of Aetiological Agents

• One of the aims of experimental epidemiology is to confirm or refute an aetiological hypothesis
• Classic example: Retrolental fibroplasia (RLF) and oxygen - Kinsey and Hemphill trial in 18 US hospitals; premature babies ≤1500g randomized into: experimental group (50% oxygen for 28 days) vs control group (oxygen only for clinical emergency). Result confirmed oxygen as the aetiological agent in RLF
• Since most diseases are fatal, disabling or unpleasant, human experiments to confirm an aetiological hypothesis are rarely possible

6. Evaluation of Health Services

• RCTs extended to assess effectiveness and efficiency of health services
• Choices must be made between alternative policies of health care delivery (limited resources)
• Classic example: Controlled trials in chemotherapy of tuberculosis in India - demonstrated that "domiciliary treatment" of pulmonary TB was as effective as the more costly "hospital/sanatorium treatment"; ushered in the era of domiciliary treatment
• Another example: Multiphasic screening evaluated by RCT in South-East London - led to withholding of vast outlay of resources for a national programme
• Also called "health services research" studies (e.g., tasks traditionally performed by physicians shown to be performable by nurses and paramedical workers)

B. NON-RANDOMIZED (NON-EXPERIMENTAL) TRIALS

• Ethical/administrative/other reasons (e.g., smoking and lung cancer)
• Some preventive measures can only be applied community-wide (e.g., water fluoridation)
• Disease frequency is low with long natural history (e.g., cancer cervix) - follow-up of thousands needed for a decade or more

"Where the approach is sophisticated in randomized controlled trials, it is rather crude in non-randomized trials." - Park

1. Uncontrolled Trials

• Trials with no comparison group
• Use implied "historical controls" (experience of earlier untreated patients)
• Example: No RCTs of the Pap test when introduced in 1920s; today, evidence from >12 uncontrolled studies shows it reduces mortality from cervical cancer
• Useful initially to evaluate whether therapy has any value, determine appropriate dose, investigate adverse reactions

2. Natural Experiments

• Epidemiologist identifies "natural circumstances" that mimic an experiment
• People separate themselves into two groups "naturally" (e.g., smokers vs non-smokers)
• Other examples: (a) migrants (b) religious/social groups (c) atomic bombing of Japan (d) famines (e) earthquakes
• Classic example: John Snow's "grand experiment" on cholera - two randomly mixed populations alike in all respects except their water supply (Lambeth Co. water from upstream vs Southwark & Vauxhall Co. water from sewage-polluted basin) - proved cholera is water-borne disease

3. Before and After Comparison Studies

• Compare incidence of disease before and after introduction of a preventive measure
• "The experiment serves as its own control"
• Requirements: (a) data on incidence before and after must be available (b) only ONE factor introduced, others remain the same (c) same diagnostic criteria (d) preventive measure adopted over a wide area (e) reduction in incidence must be large (f) several trials needed before evaluation is conclusive
• Classic examples: Prevention of scurvy by James Lind (1750) with fresh fruit; John Snow's cholera studies (1854); prevention of polio by Salk and Sabin vaccines
• Example: Compulsory seat-belt legislation in Victoria, Australia 1971 - deaths fell from 564 to 464 (-17.7%); injuries fell from 14,620 to 12,454 (-14.8%)

• A "natural" control group is utilized to overcome limitations of the above
• If a preventive programme applies to an entire community, another similar community is selected as the control
• Example: Same seat-belt study compared Victoria (legislation introduced) vs other Australian states (no legislation) - Victoria showed -17.7% deaths while other states showed +0.2%; Victoria showed -14.8% injuries while other states showed +1.0%
• In evaluation of preventive measures, three questions considered: (a) How much benefit to community? (b) What are the risks to recipients? (c) What is the cost in money and manpower?

Source: Park's Textbook of Preventive and Social Medicine, Chapter on Epidemiology - Experimental Studies (pp. 93-99)

Types of Randomized Controlled Trials (RCTs)

Park's Textbook of Preventive and Social Medicine

1. Clinical Trials

• Evaluate therapeutic agents, mainly drugs
• Not all are susceptible to blinding
• Example: Beta-blockers reducing cardiovascular mortality post-MI; Efficacy of tonsillectomy for recurrent throat infection

2. Preventive Trials

• Trial of primary preventive measures - vaccines & chemoprophylactic drugs
• May be applied to groups rather than individuals
• Example: MRC UK (1946) whooping cough vaccine trial - attack rate 1.45 vs 6.72 per 1000 child-months in vaccinated vs unvaccinated

3. Risk Factor Trials

• Investigator interrupts the usual sequence of disease in persons with a known risk factor
• Can be single-factor or multi-factor
• Example: WHO clofibrate trial - >15,000 men, 3 European centres; reduced non-fatal MI but showed 25% more deaths in clofibrate group

4. Cessation Experiments

• Evaluates termination of a habit or removal of a suspected causal agent
• Positive result strengthens the hypothesis of causation
• Example: Smokers vs quitters - reduction in lung cancer incidence in quitters strengthens smoking-lung cancer causal hypothesis

5. Trial of Aetiological Agents

• Confirms or refutes an aetiological hypothesis
• Rarely possible as most diseases are fatal or disabling
• Example: Kinsey & Hemphill trial (18 US hospitals) - premature babies ≤1500g given 50% O₂ vs curtailed O₂ - confirmed oxygen as cause of Retrolental Fibroplasia (RLF)

6. Evaluation of Health Services

• Assesses effectiveness and efficiency of health service delivery between alternative policies
• Also called "health services research" studies
• Example: Domiciliary vs hospital treatment of pulmonary TB in India - domiciliary treatment found equally effective, ushered in new era of TB management

Source: Park's Textbook of Preventive and Social Medicine, pp. 93-98

Types of Clinical Trials

Park's Textbook of Preventive and Social Medicine

A. Randomized Controlled Trials (RCT) B. Non-Randomized (Non-Experimental) Trials

A. RANDOMIZED CONTROLLED TRIALS

1. Clinical Trials

• Evaluate therapeutic agents (mainly drugs)
• May not always be blinded (e.g., surgical trials)
• Example: Beta-blockers post-MI; Tonsillectomy for recurrent throat infections; Aspirin on cardiovascular mortality

2. Preventive Trials

• Trial of primary preventive measures - vaccines & chemoprophylaxis
• Applied to groups rather than individuals
• Example: MRC UK (1946) whooping cough vaccine - attack rate 1.45 vs 6.72 per 1000 child-months (vaccinated vs unvaccinated)

3. Risk Factor Trials

• Intervenes to interrupt usual disease sequence in high-risk individuals
• Single-factor or multi-factor
• Example: WHO clofibrate trial (>15,000 men, Edinburgh/Prague/Budapest) - reduced non-fatal MI but 25% more deaths in clofibrate group

4. Cessation Experiments

• Evaluates termination of a habit or removal of causal agent
• Positive result strengthens causal hypothesis
• Example: Smoking cessation trial - reduced lung cancer incidence in quitters confirms smoking-lung cancer causal link

5. Trial of Aetiological Agents

• Confirms or refutes an aetiological hypothesis
• Rarely possible as most diseases are fatal/disabling
• Example: Kinsey & Hemphill (18 US hospitals) - premature babies ≤1500g, 50% O₂ vs curtailed O₂ - confirmed oxygen as cause of Retrolental Fibroplasia (RLF)

6. Evaluation of Health Services

• Assesses effectiveness & efficiency of health care delivery
• Compares alternative policies; also called "health services research"
• Example: Domiciliary vs hospital/sanatorium treatment of pulmonary TB in India - domiciliary treatment equally effective, cheaper - ushered in new era

B. NON-RANDOMIZED (NON-EXPERIMENTAL) TRIALS

• Ethical reasons (e.g., smoking and lung cancer)
• Measures applicable only community-wide (e.g., water fluoridation)
• Low disease frequency with long natural history (e.g., cancer cervix)

1. Uncontrolled Trials

• No comparison group; uses implied historical controls
• Useful to evaluate initial value of therapy, determine dose, study adverse reactions
• Example: Pap smear introduced in 1920s without RCT - >12 uncontrolled studies later confirmed it reduces cervical cancer mortality

2. Natural Experiments

• Identifies natural circumstances that mimic an experiment
• Groups separate themselves naturally (e.g., smokers vs non-smokers)
• Example: John Snow's "Grand Experiment" - two populations differing only in water source (Lambeth Co. = clean upstream water vs Southwark & Vauxhall Co. = sewage-polluted water) - proved cholera is water-borne

3. Before & After Comparison Studies

• Compares disease incidence before vs after a preventive measure
• Experiment serves as its own control
• Example: Seat-belt legislation in Victoria, Australia (1971) - deaths fell 564 → 464 (-17.7%); injuries fell 14,620 → 12,454 (-14.8%)
• Also: James Lind (scurvy & fresh fruit, 1750); Salk & Sabin polio vaccines

• A natural control group (similar community without intervention) is used
• Overcomes limitations of uncontrolled before-after studies
• Example: Same seat-belt study compared Victoria (legislation) vs other Australian states (no legislation) - Victoria: -17.7% deaths; Other states: +0.2% deaths

Summary Table

Source: Park's Textbook of Preventive and Social Medicine, pp. 93-99

Steps Involved in a Randomized Controlled Trial (RCT)

Park's Textbook of Preventive and Social Medicine

"For new programmes or new therapies, the RCT is the No.1 method of evaluation."

• Park

STEP 1 - Drawing Up a Protocol

• One of the essential features of an RCT
• The protocol specifies:
  • Aims and objectives of the study
  • Questions to be answered
  • Criteria for selection of study and control groups
  • Size of the sample
  • Procedures for allocation of subjects into study and control groups
  • Treatments to be applied - when, where, how, and to what kind of patients
  • Standardization of working procedures and schedules
  • Responsibilities of parties involved, up to stage of evaluation
• Especially important when multiple centres are participating
• Once evolved, must be strictly adhered to throughout the study
• Aims at preventing bias and reducing sources of error

Preliminary test runs: Before a protocol is finalized, it may be tested in a small pilot study to check feasibility

STEP 2 - Selecting Reference and Experimental Populations

• The large population to which the findings of the trial will be applied
• E.g., all patients with a particular disease in the country

• The actual group from which study subjects are drawn
• Must be representative of the reference population
• Must satisfy eligibility criteria (inclusion and exclusion criteria)
• Must give informed consent

• Persons likely to benefit from the intervention
• Persons unlikely to be harmed
• Persons likely to comply with the regimen
• Persons likely to be available for follow-up

STEP 3 - Randomization

• The hallmark of an RCT
• Process by which each participant has an equal chance of being assigned to either study or control group
• Ensures comparability of the two groups at the start of the trial
• Controls for both known and unknown confounding variables

• Simple randomization - toss of a coin, random number tables
• Systematic randomization - every alternate subject to study group
• Stratified randomization - stratify by age/sex first, then randomize within strata
• Cluster randomization - entire communities/groups randomized

STEP 4 - Manipulation / Intervention

• The experimental group receives the new treatment/intervention
• The control group receives:
  • Placebo (inert substance resembling active treatment), OR
  • Standard/current treatment (active control), OR
  • No treatment (in some trials)

• Double-blind is the gold standard - eliminates observer bias and subject bias simultaneously
• Placebo effect: Patients may improve simply because they believe they are receiving treatment - blinding controls for this

STEP 5 - Follow-Up

• Both groups followed up for the same duration under the same conditions
• Must be regular and complete
• Losses to follow-up are a major problem - can introduce bias
• Both groups must be treated identically except for the intervention
• Data collected on: outcomes, side effects, compliance, withdrawals

• Non-compliance - subjects not taking allocated treatment
• Drop-outs - subjects withdrawing from the trial
• Cross-overs - subjects switching groups
• These are analyzed by Intention-to-Treat (ITT) analysis - analyze in the group originally allocated, regardless of what treatment they actually received

STEP 6 - Assessment of Outcome

• Compare outcomes in experimental vs control groups
• Outcome measures may include:
  • Incidence of disease
  • Mortality rate
  • Cure rate
  • Side effects/adverse reactions

• Calculate Relative Risk (RR), Odds Ratio (OR)
• Calculate p-value (statistical significance)
• Calculate 95% Confidence Interval

• Absolute Risk Reduction (ARR) = Risk in control - Risk in treatment group
• Relative Risk Reduction (RRR) = ARR / Risk in control group
• Number Needed to Treat (NNT) = 1/ARR

Summary Flowchart

Reference Population
        ↓
Experimental Population (eligibility criteria met)
        ↓
   RANDOMIZATION
    ↙         ↘
Study Group   Control Group
(Treatment)    (Placebo/Standard Rx)
    ↓               ↓
        FOLLOW-UP
        ↓
   OUTCOME ASSESSMENT
   (Compare results)

Key Advantages of RCT

• Controls for known and unknown confounders (via randomization)
• Blinding eliminates observer and subject bias
• Provides the strongest evidence for causation
• Results are quantifiable and statistically testable

Key Limitations

• Expensive and time-consuming
• Ethical constraints
• Not always feasible (rare diseases, long latency)
• Hawthorne effect - subjects may change behavior knowing they are being observed

Source: Park's Textbook of Preventive and Social Medicine, pp. 93-96