RCTs and means Flashcards
Define RCT
A study in which participants are randomly allocated between a treatment intervention and a controlled group.
Confounder
A variable that influences both dependent and independent variable, causing a false association
RCTs ensure that confounding are distributed by chance.
Clinical Equipoise
Genuine uncertainty as to whether one treatment will be more beneficial than another when assigning patients to different treatments of a clinical trial.
It would be unethical to give patients a treatment that is harmful or inferior
Bias
A departure of results from the truth due to a systematic error in sampling/ testing by selecting/encourage one outcome.
It is independent from sample size and statistical significance.
Causes of bias
Instrument that measures outcome. May not represent true value.
Choice of participants.
- May not be representative of population
Influence of observers or researchers.
- May change measurements for different groups.
Selection bias
Occurs when study sample does not represent target population.
Examples: recruiting from volunteers who may have certain incentives, recruiting from inpatients who represent the more ‘ill’ population of a condition.
Observer bias
Systematic difference between the way information is collected and how the group is being studied.
Blinding
Concealing the group allocation from one or more individuals involved.
Pros of blinding
Reduces selection bias by preventing the different treatment or assessments of groups
Types of trials that can be blinded
Drug trial with placebo
Sham surgery (during everything you would do in a normal surgical intervention, without doing the actual therapeutic step) vs other intervention
Types of trials that cannot be blinded
Psychological interventions
Surgical vs non-surgical trial
Types of RCT
- Open
- Single-blinded
- Double blinded
- Triple blinded
Open
- Everyone in trial knowns
Single
- patients are blinded
Double
- patients and treating physicians blinded
Triple
- Patients, treating physicians and study investigators blinded
Type 1 and type 2 error
Type 1
- When there is an observed difference that has no true difference
- Significance level accepted at 5%.
Type 2
- When there is no observed difference but there is a true difference
Types of data
- Categorical
- Scale variables
- Qualitative
Categorical
- nominal, binary, dichotomous
- E.g sex, social class, ethnicity
Scale variables
- continuous, interval
- Age, height, weight
Qualitative
- Text, words
Data skew
When there is more count in the tail than expected in normal distribution.
Positive- tail on right
Negative- tail on left
Problems
- Mean is difference from median
- median is better representation of data than mean as mean is highly influenced by tail.
Prevalence
- Definition
- Equation
Frequency of cases in a given population at a designated time:
- diagnosed or diagnosed + receiving treatment/ management
Expressed as percentage/ proportion
Equation
- No. of people with diagnosis / total population.
- usually point prevalence, number affected in given time/ period.
Cross sectional study
- Definition
Measures prevalence and exposure of disease
Exposure
Proportion of people who undertake a health behaviour in specific populations.
Cohort studies
- Definition
Study where people who presently have a certain condition or receive a particular treatment are followed over time and compared with another group of people who are not affected by the condition.
Measures incidence and risk over a certain period of time.
Incidence
- Definition
- Equation
The number of onsets of disease or health event, in given period of a defined population.
Number of new events in population/ Average number of people exposed to risk during this period.
Relative risk
The incidence of disease/health event when exposed to a factor / incidence without exposure.
In a given time and population.
Risk ratio
Incidence of something with exposure/ incidence without exposure
Risk difference
Absolute difference between two risks/ incidence in at a specific time.
Calculated by subtracting risk (incidence) between two population
Strengths of cohort study [4]
Measures incidence and compares between exposed and unexposed groups.
Studies multiple effects of single exposure.
When prospective, can choose what variables to measure.
Prevents recall bias (when participants are unable to accurately recall events or experiences)
Limitations of cohort study [4]
- Takes time, might lose relevance
- Prospective study is expensive
- Loss of participants due to follow up
- Requires large population.
P value
Probability that difference observed could have occurred by chance if the groups compared were alike.
Lower p value= less likely to have occurred by chance.
Greater sample size= smaller P value, narrower confidence interval.
Confidence interval
Range of value with a given probability, that the true value of a variable is contained within.
Intervals become narrower as sample size increases.
- Required when you need to measure effect or give estimates.
Sensitivity
Proportion of subjects with condition who correctly test positive for the condition.
Specificity
Proportion of subjects without the condition who correctly test negative for the condition.
Positive predictive value
The probability that subjects with a positive test truly have the disease.
In 2x2 table= A/ A+B
As prevalence increases, PPV increases and NPV decreases
Negative predictive value
Probability that subjects with negative screening test do not have the disease.
In 2x2 table= D/C+D
As prevalence decreases, NPV increases, PPV decreases
Benefits of RCT (5 )
Reduces selection bias due to randomisation
- Confounders are equally distributed
when blinded
- Decreases observer/ research bias (measurements won’t be different if intervention allocation is known)
- Reduces performer bias (different treatment of group if allocation is known)
High internal validity
Components of sample size calculation
Statistical power
- Type 2 error
- 80-90% of correctly rejecting the null hypothesis and accepting it, when difference exists.
- <20% chance of accepting null hypothesis when there is true difference
Type 1 error
- <5% of falsely rejecting null hypothesis when there is no difference
Smallest effect of interest
- Minimally clinically relevant different
- Minimal difference between groups that is biologically plausible/ clinically relevant
Variance
- Variability of outcome measure (SD)