Lecture 16-19 Interventional Study Designs Flashcards
Key Differences Between Interventional and Observational Study Designs
Interventional: clinical trial/study, experimental study, human study, investigational study
- investigator selects “interventions” and allocates study subjects to forced intervention groups
- more “rigorous” in ability to show cause-and-effect
- can demonstrate causation
Phases of Interventional Studies
Pre-clinical= prior to human investigation (bench and animal research)
- trends moving from phase 1-4:
1. population size increases
2. longer duration
3. change in focus (safety-efficacy-back to safety)
4. phase 1 is only one with healthy subjects and phase 4 must be post approval (post marketing)
Phase 1
- new drug/device/procedure
- small sample size (20-80)
- healthy volunteers or used for the first time in humans to asses safety/toxicity/dosing/pharmokinetics in the diseased population
- short duration (few days-month)
- SAFETY primary purpose
Phase 2
- new drug/device/procedure, indication/study population
- larger sample size (100-300)
- use patients with condition of interest, expand on purpose of phase 1 (safety) but ALSO being assessing efficacy is diseased population
- short-medium duration (few months)
- have narrower inclusion criteria (tend to be restrictive to people w/disease of interest, no comorbidities and no other drugs)
Phase 3
- new drug/device/procedure, indication or study population
- even large sample size (1,000-3,000)
- used in patients w/condition of interest to continue determination of safety BUT primary purpose to assess efficacy
- longer duration (months-year)
- assess superiority, non-inferiority, equivalency
Phase 4
- post marketing/always after the drug/device/procedure approval
- long term effects in a large population of diseased patients (expanded use population)
- registries/surveys (collect info from patients on drugs, looking for temporality, documentation, etc.)
examples: Medwatch, FAERS, VAERS
Advantages of Interventional Trials
-cause precedes effect (shows causation)
- only design used by FDA for “approval” process on label
- vital steps in approval process, phase 1-3 need to show benefits in use to make a phase 4 study
Disadvantages of Interventional Trials
- expensive
- complex/takes time (development/approval/conductance)
- ethical considerations (never force people to smoke/drink) *usually the #1 reason observational studies are used instead
- generalizability (external validity)= greatest flaw, restrictive to co-morbidities/drugs use to isolate the effects of disease of interest (prove causation) but since restrictive it does not apply to general public (patients going to clinics with other co-morbidities/medications)
Explanatory Interventional Study
- Not the usual way to make clinical designs (rule book/restrictions must be followed)
- use placebos
example: patients get put on same dose medication but some people need more or less and with rule book you cannot change dose or adjust measures like in a real clinical setting (no flexibility) - less external validity , too restrictive
Pragmatic Interventional Studies
- use normal clinical setting where there are patients with multiple co-morbidities and doctors can prescribe individual dosing to patients (let in “regular” patients)
- never use placebo (compare real drug to real drug)
- researchers loose control over how physician prescribes medications and group allocation/restriction
- additional confounding with co-morbidities
- more flexible and more external validity
- less able to prove causation
Simple Study Design
- divides (randomizes) subjects exclusively into 2+ groups
- SINGLE randomization
- used to test SINGLE hypothesis
Factorial Study Design
- divides subjects into 2+ groups and then further subdivides each groups into 2+ subgroups
- used to test MULTIPLE hypotheses at the same time (useful for combination therapy)
- must increase sample size
- improves efficacy for answering clinical questions
- increases complexity (could be barrier to recruitment)
- increases risk of drop outs (due to increase complexity)
- may restrict generalizability of results
Parallel Study Design
- groups simultaneously and exclusively managed
- no switching of intervention groups after initial randomization
- *All simple and factorial study designs
- worry if people are equal in every aspect (want to limit differences between groups)
Cross-Over Study Design
- self control
- groups serve as their own control by crossing over from one intervention to the other during the study
- allows for small sample size (each patient contributes additional data)
- don’t way about differences between groups because the same people contribute to both groups
- between groups comparison= all group a vs all group b
- within groups comparison= person while in group a vs same person while in group b
Wash Out
- longest period you need for each group to get it all cleaned out (back to baseline)
- all effects eliminated
- want to avoid hangover (don’t want lingering effects from other therapy, want all effects gone before person starts other therapy)
Lead in phase
- practice period
- can also function like a wash out period
- let patients practice to see if they can accomplish everything required
- placebo will be taken
- at the end of lead in phase bring subjects back if they can complete and comply with the process at a certain level (usually 80-90%)
- drop them out of the study if don’t meet standards
Disadvantages of Cross Over Design
- only suitable for long term conditions which are not curable or which treatment provides short term relief
- duration of study for each subject is longer
- carry over effects (try to wash out)
- treatment by period interaction (differences in effects of treatments during different time periods)
- complexity in data analysis
Outcomes/Endpoints:
Primary
- most important key outcome(s)
- main research question used for developing/conducting study
- main purpose of the study
Outcomes/Endpoints
Secondary
- lesser importance yet still valuable
- possible for future hypothesis generation
- secondarily while I am at it I will ask about…..
- not primary purpose of the study
Outcomes/Endpoints
Composite
- combines multiple endpoints into a single outcome
- could be considered the primary outcomes and if so then secondary outcomes may be the individual elements from the composite outcome
example: main purpose of the study is to assess outcome of MI, stroke, or death (composite) - each one individually is secondary outcome
Outcomes/Endpoints
Patient Oriented Endpoints
- most clinically relevant
ex: death, stroke, MI, hospitalization, preventing dialysis - direct: mean things to patients, prevent cancer/death, prolong survival
- more useful and impactful
Outcomes/Endpoints
Surrogate Markers
- elements used in place of evaluating patient oriented (direct) endpoints
- not the only reason someone might have disease
ex: blood pressure, cholesterol, change in SCr levels - study can start off looking at surrogates and move into more clinically relevant direct outcomes
Non-Random Group Allocation
- subjects do NOT have an equal probability of being selected or assigned to each intervention group
- convenience sampling or non-probabilistic allocation
- can sound purposeful and pre-subscribed but it is NOT random
ex: patients attending morning clinic assigned to group 1 and patients attending afternoon clinic assigned group 2
ex: first 100 people admitted to the hospital
Random Group Allocation (most utilized)
- subjects do have an equal probability of being assigned to each intervention group
- goal is to make groups as equal as possible and takes allocation process away from research
- uses random number generating program
Randomization
- purpose: to make groups as equal as possible based on known and unknown important factors (confounders)
- attempts to reduce systematic differences (bias) between groups which could impact results/outcomes
- Equality of groups is not guaranteed
- ex)flip a coin get 5 heads in a row need tails so put next 3 people in tails
- documentation of equality of groups (effectiveness of randomization) reported in 1 of several locales:
1. p value shown table format
2. p value in text statement given in key
3. p value in text in article
Examples of Randomization:
Simple
-equal probability for allocation within one of the study groups
Examples of Randomization:
Blocked
-ensures balance within each intervention group
-when researchers want to assure that all groups are equal in size
Ex: 10 blocks, so every 10th person you can decided where they go (if most people in group b put them in group a)
Examples of Randomization:
Stratified
- ensures balance with known confounding variables
- can also pre-select levels to be balanced within each interfering factor (confounder)
ex: gender, age, disease severity, comorbidities
ex: separate heavy and light smokers
Masking:
single blind
study subjects are not informed which intervention intervention they are receiving (but researcher knows)
-usually used/okay if researcher is not interacting with subject so it is irrelevant because the researcher cannot influence the subject in any way
Masking:
double blind
- neither investigator nor study subjects are informed which intervention each subject is receiving
- some outside person will know but the researcher and subject will not so researcher does not influence subject if they interact
Masking:
open-label
-everyone knows which intervention each subject is receiving
- sometimes it doesn’t matter if you blind, just want to see effects of treatment or sometimes you can’t blind
ex: compare injection to inhaler or pill (you would need 2 placebos to blind)
Post-hoc Surveys
- used to assess adequacy of blinding
- have survey thats asks participants or researchers to guess what group they are in (if 97% know their group blinding did NOT work)
- can only know if blinding works if you ASK subjects
Forms of Blinding:
Placebo
- “dummy” therapy
- inert treatments made to look identical in all aspects to the active treatments
ex: dosage, frequency, monitoring, requirements all the same (need exactly the same conditions as if the real drug)
Double Dummy
- more than 1 placebo used
ex: pill vs IV or inhaler vs injection you must get 2 placebos to mimic both the pill and IV or injection and inhaler
Placebo Effect
- improvement in condition by power of suggestion and due to the care being provided
- can be as large as 30-50%
-counter to those who believe equipose can never be met because no benefit to placebo can say the placebo CAN be benificial
Hawthorne-Effect
- desire of the study subject to “please” investigators by reporting positive results (improvements) regardless of treatment allocation
- desire for positive outcomes to process
- want to be helpful (overly helpful/friendly/positive)
*lead in phase allows for this to kick in prior to actual data collection so can give subject new baseline if see they are overly positive during lead in phase
Post-hoc sub-group analysis
-not accepted as appropriate by most when it is not PROSPECTIVELY planned
- data dredging or fishing (trying to compare data in different ways to find a difference/similarity to discuss)
- reduced power and increases risk of type 2 error
-when authors prospectively say what they are comparing and it is pre planned it is okay but if not discussed in methods, it is fishing for significant data
Managing drop-outs/lost to followups:
Include them Anyway
- intent to treat= most conservative decision
ex: some people drop out so you can use the last observation carried forward LOCF (fill in last visit for every time they are missing so they do not improve or get worse they stay the same)
ex: convert all subsequent yet missed assessments for a subject to a null effect (no benefit)
Intent to Treat results in the following:
- preserves randomization process
- preserves baseline characteristics and group balance at baseline which controls for known and unknown confounders
- maintains statistical power (original sample size)
Managing drop-outs/lost to follow ups:
Ignore them
- include only compliant or completing subjects
- per-protocol or efficacy-analysis= compliance must be predefined (usually 80-90%)
ex: someone only complete 60% so they are not included
Managing drop-outs/lost to follow ups:
Treat them “as treated”
-ignores group assignments, allow subjects to switch groups and be evaluated in group they moved to, end in, or stay in the most
Assessing Adherence (Compliance)
- drug levels
- pill counts at each visit
- bottle counter tops
Methods of Improving Adherence (Compliance)
- frequent follow up visits and communication
- treatment alarms/notifications
- medication blister packs or dosage containers
*do everything necessary for them to not drop out
