Intention to Treat Flashcards
What is the intention to treat principle?
Participants must be counted in the groups to which they were initially assigned, regardless of whether they were compliant with the intervention.
Includes those who withdraw.
Rationale behind intention to treat principle?
In an RCT one sets out to estimate broader effects of allocating intervention to a group and not the narrow effects of those who comply.
In reality, few people adhere to treatments strictly.
What does intention to treat principle give us?
Conservative estimate of treatment effect compared with what would be expected in ideal world of full compliance
Advantages of intention to treat principle
Makes results from RCT more generalisable and pragmatic
When might intention to treat principle not be suitable?
RCTs where investigation is of biological effects of drugs
Impact of ITT analysis in non inferiority trials?
Will favour equality due to dilution
When is ITT difficult to interpret?
If number of crossovers are high
Ways of dealing with missing data
Test the difference
Use available data
Imputation of data
Statistical methods
What is test the difference?
Test for difference between drop out group and available subject group.
Baseline characteristics if comparable can substantiate loss without further adjustment
Ways of using available data to deal with missing data
LOCF
Worst case scenario analysis
What does LOCF stand for?
Last observation carried forward
When is LOCF useful?
Only for longitudinal observations where continuous scales are used
Disadvantages of LOCF
Many psychiatric diseases have natural course of remission - LOCF does not allow for this’
‘Too good’ for controls and ‘too stringent’ for patient group
What is worst case scenario analysis?
Assumes that those lost to follow-up did not improve on trial intervention. Hence, data of treatment failure subjects used as proxy.
When is worst case scenario analysis used?
As part of sensitivity analysis
Types of imputations of data
Replacing missing values with mean of group
Hot deck imputation
What are some statistical methods to make up for missing data?
Multiple regression
Growth curve analysis
Random effects modelling
Multiple imputation
What is hot desk imputation?
Involves finding similar person among subjects and using that persons score
What is multiple regression?
Used to predict missing values based on several other variables in data set
What is growth curve analysis?
Line is fitted through existing data points for each subject. From 2 data points, we can derive parameters for straight line. For 3; quadratic curve. For 4; cubic curve.
What is needed for growth curve analysis
At least 2 follow-up visits
What is the definition of random effects modelling?
Derives from the fact that, when we usually fit a regression line to data, we assume that the parameters are fixed i.e. the same for all subjects
In random effects model, we allow them to differ for each person
Who defined random effects modelling?
Streiner
Implications of random effects modelling?
We do not expect curves for people who drop out to be same as for people who complete study
What is multiple imputation?
Sophisticated and accurate methods which use maximum likelihood and Monte Carlo procedures.
Recommendations if missing data are imputed?
Some sensitivity analysis is performed to ensure study conclusions are not misleading
Alternatives to ITT
Per-protocol analysis
Treatment-received analysis
What is per-protocol analysis?
Only patient who sufficiently complied with trials protocol are considered in final analysis.
Another name for per-protocol analysis?
On treatment analysis
Which types of RCTs use per-protocol analysis?
Biological explanatory RCTs
Secondary analysis in effectiveness trials
What is treatment-received analysis?
Analyse subjects according to actual treatment received - not allocation made.
Disadvantage of treatment-received analysis?
Effect of randomisation is compromised due to contamination of trial by one group of subjects moving to other group
ITT is the norm except when?
Overwhelming justification of different analysis policy
Advantages of ITT
Retains balance of prognostic factors arising from original random treatment allocation
Gives unbiased estimate of treatment effect
Admits non-compliance and protocol deviations, reflecting clinical life
Limitations of ITT
Estimate of treatment effect is conservative
In equivalence trials, ITT will favour equality of treatment
Interpretation is difficult if large proportion of participants cross over to opposite treatment arms
Why is estimate of treatment effect conservative in ITT?
Dilution effect of non-compliance
Requirements for ideal ITT
Full compliance with randomised treatment
No missing responses
Follow-up of all participants
