CPPS 337 Flashcards
Objective
Every experiment starts with a research question, which is stated as the study objective - the objective is whatever is going to be assessed or measured
- primary (efficacy of drug) and secondary objectives (safety of drug)
Example:
- Experiment: test new drug for migraine prophylaxis (prevention)
- Primary Objective: does drug A reduce frequency of migraines
Outcomes
specific assessments designed to help inform a study objective
(outcomes get you to objective - help you answer the research question)
1 or 2 primary outcomes that decide the success of your study
Examples:
- Change from baseline to week 24 in monthly migraine days (primary)
- Number of patients with a migraine 1 day after their dose
of Drug A
Enrollment
Patients that are recruited for study
- inclusion criteria
- exclusion criterai
Intervention group
Gets treatment
Control group
Does not get treatment
Randomization
Random allocation of participants to treatment and control group
Stratification factors
Want to keep t and c groups balanced for baseline characteristics - helps randomization be balanced
- Tell computer to make groups with equal number of moderate/severe asthma, and high/low dose corticosteroids
Allocation concealment
Concealing which group the patient has been allocated to
- make the allocation process automated and conducted by a third party
- can use interactive voice/web response system
Blinding
Matching control to treatment
- can use doubly dummy for different dosage forms: One group gets tablet and placebo shot, other gets placebo tablet and shot
Parallel control group
treatment and control groups observed at same time
Historical control group
control group taken from old study
Active treatment
active established drug thats given to control group
- active controlled vs plavebo controlled trials
Prospective study
looks forward - watches for outcomes like the development of a disease
- relates the outcome to suspected risk or protection factors
- clinical trials (testing new drug) always prospective - go forward in time
Retrospective study
looks backward - outcome established at the start of the study; examines past exposures to suspected risk or protection factors
- figuring out about AIDS: looked at patient history and found virus)
Historical control
the outcomes of the intervention group are compared to results from a comparable group of patients whose data is retrieved from a database from the past
Reasons to choose a historical control
- there is no active control
- use of placebo considered unethical (patients would suffer)
- intervention is so good that all patients should be given access
Reasons not to use historical control
- results depend on choice of control
- subject to bias
Crossover designs
all participants are exposed to all controls and interventions in the trial - patients serve as their own control - takes individual variability out of the eqn
- start on a and switch to b after a period of time
- assuming order does not matter
- washout period between switch so no lingering effects of one drug (hard to do cuz some effect may stay)
Withdrawal studies
participants on a particular treatment are taken off therapy or have their dosage reduced
- objective is to asses response to discontinuation or dose reduction
- evaluating duration of benefit of an intervention already known to be useful
- checking for rebound effect (making condition worse after discontinuation)
Cohort study
observing a group/cohort over time
- participants selected based on characteristic/exposure that may cause disease
- incidence of the disease in the exposed individuals is compared with the incidence in those not exposed
- example: Framingham study (tracking cardiovascular outcomes and risk factors for CV disease)
Cohort studies advantages
- can study multiple outcomes for a given exposure
- good for investigating rare exposures
- can calculate rates of disease in exposed and unexposed individuals over time
Cohort studies disadvantages
- Large numbers of subjects are required to study rare exposures
Prospective Cohort Studies
- Expensive
- Long
- Maintaining follow-up is difficult
- Loss to follow-up or withdrawals
Retrospective Cohort Studies
- Susceptible to selection bias
- Susceptible to recall bias
- Less control over what information is available
Case-control studies
compare patients with a disease/outcome of interest (case) with patients who dont have it (control)
- look back retrospectively to see how often a risk factor was present in each group to determine the relationship between risk factor and disease
Difference between case-control studies and cohort studies
Case-control studies identify subjects by outcome status at the outset of the investigation
Advantages of case-control studies
- quick to conduct
- when a condition is uncommon - a lot of info can be revealed from few subjects
- relatively inexpensive
- can use existing records
Disadvantages of case-control studies
- recall bias
- hard to validate info
- selection of control group used to be hard
- individual pairing with control used to be hard
Case series
Bunch of case reports of patients who were given similar treatment
- when you see something odd - drug makes skin blue - look for other reports where this has happened, put them into series, publish that (b/c more evidence than case report)
- linking together case reports
case series disadvantage
no control group
reasons to use case series
Can provide:
- initial indication of treatment efficacy
- description of a novel clinical problem/complication.
- valuable in describing the natural history of a condition
- valuable in describing recovery and complication rates after a treatment/procedure
Characteristics of good case series
- Clearly defined question
- Well-described study population
- Well-described intervention
- validated outcome measures
- Appropriate statistical analyses
- Well-described results
- Discussion/conclusions supported by data
- Funding source acknowledged
median
middle number of a data set
- line up numbers from least to greatest - middle number with equal data points above and below
mode
most frequently reported data value in a data set
variance/ standard deviation
show the dispersion of data from the average value in a data set
skewness
deviations from a symmetrical bell curve - asymmetry from the mean
outliers
values much larger or smaller than others within a data set
validity (accuracy)
how close a measurement is to the true value (target/reference)
reliability (precision)
how close several measurements are to each other
how can you improve validity of an experiment
- mitigate bias
- use randomization
- blinding
how can you improve reliability of an experiment
- increase sample size
- control potential confounding variables
- increase number of measurements
internal validity
the extent to which a study’s design accurately measures the truth without influence from other variables
external validity
the extent to which the research results reflect what happens outside of a research setting (in the real world)
null hypothesis
states there is no difference between a treatment and control
alternative hypothesis
states there is a difference between a treatment and control
p value
denotes how probable it is that the results of an experiment are due to chance
- conventionally, p < 0.05 denotes significance (arbitrary cuz if studying effects of a drug on mortality, don’t want a 5% chance the person will die)
- if p is higher than 0.05 maybe you just don’t have a big enough sample size to discern statistically significant result)
- higher p values suggest weaker evidence against the null hypothesis
- lower p values suggest stronger evidence against the null hypothesis
inclusion criteria
who should you include in the study and why (based on research question)
exclusion criteria
who should you exclude in the study and why
- studying a drug that causes anxiety - exclude ppl with defects in amygdala
what if you include everyone and exclude no one
outliers skew results
what if you exclude certain people/ measurements but not others
introduces bias
confounding
extraneous variables that may influence the measurement of a relationship between other variables
nocebo effect
informing patient they will experience negative symptoms from a drug and they do
selection bias
issues in sampling
- studying anxiety and you recruit patients from a psychiatric hospital - more likely to display what you want them to display
recall bias
issues in remembering details
interviewer bias
issues in how a researcher may present themselves to an interviewee
- priming participants through leading questions, stereotyping (threat or boost effect on performance), influence by expectations or opinions
Hawthorne effect
people modify their own behavior because they know they are being observed/assessed
Pygmalion/Rosenthal effect
expectations may drive performance
- self fulfiiling prophecy
type I error
(false Positive)
inaccurately rejecting the null hypothesis when it is true
- related to the significance level (p)
type II error
(false Negative)
inaccurately accepting the null hypothesis when it is false
- depends on power (sample size)
selection/sampling bias
cherry picking people (need random selection instead)
statistically significant
gray area - causes rejection of null hypothesis
avoiding Type II error
increase statistical power
- increase sample size –> reduce variability (spread) –> reduce overlap
power
probability of correctly rejecting a false null hypothesis (avoiding type II error) - probability that sample means will be found statistically different when there is a true difference
- increasing sample size
- correctly reject the null hypothesis 80% of the time
- 80% of the curve is to the right of the cut off
confidence interval
if p value is 0.05, the confidence interval is 95%
- we are 95% sure that the true difference lies within the 95% confidence interval (confidence interval contains the true difference between the treatment groups)