statistical adjustment in data analysis

Question 1

what do we mean by baseline covariate adjustment?

Answer 1

equalising the groups we compare in every possible way except in the one factor at issue. if this isnt achieved to begin with we must equalise them with analysis

Question 2

why adjust for baseline covariates

Answer 2

• Correct for imbalance
• Increase power
• Account for features of study design – randomisation
• Obtain treatment effect estimates that would be
  more closely relevant to individual patients

Question 3

does including covariates always increase power?

Answer 3

no, only does this if the covariates are related to the outcome.

Question 4

if covariates are related to the outcome, how does this increase power?

Answer 4

because thecovariates reduce the residual variance in the outcome and therefore increase the precision of the estimate

Question 5

how do including covariates affect continuous outcomes?

Answer 5

> doesnt change the estimated treatment effect
but does reduce the standard error .

therefore treatment effect is therefore more precise and study has more power to detect it

Question 6

how do including covariates affect binary or time-to-event outcomes?

Answer 6

> the estimated treatment effect moves further away from the null(i.e. larger effect size)
standard error however may increase

therefore the study has more power to detect an effect but the effect itself may be more precise

Question 7

what things can we adjust for?

Answer 7

• Prognostic factors
• Baseline measures
• Centres

Question 8

can we adjust for measures after randomisation

Answer 8

NO

Question 9

Why should the method of randomisation be taken into account in the statistical analysis in order to produce accurate and valid results.

Answer 9

well the randomisation method used influences the extent of balance.

therefore the randomisation method used should inform the statistical analysis plan. for example if used block randomisation its good to adjust for the blocking factor in the analysis. if stratification randomisation was used, useful to adjust for the stratification factor.

Question 10

adjusting for baseline measures

Answer 10

important because it produces only the change induced by treatment. takes into account differences in baseline values between the two groups

Question 11

define:
* Final value analysis
* Change value analysis
* Baseline adjustment analysis (ANCOVA)

which of these gives us the most precise estimate of the treatment effect

Answer 11

> first analysis method analysis the final value of the outcome variable ignoring baseline measures
second analysis - looks at the change in the outcome variable from baseline to final
involves adjusting the analysis for baseline values of the outcome variable and other relevant baseline covariates

last one gives us the most precise estimate of the treatment effect

Question 12

why adjust for centers?

Answer 12

n in one center may be more similar to those in another.

to adjust for this in the analysis you should have randomised by center

ignoring the center effect might influence the estimate of the treatment effect, p values and CI

Question 13

what ways can we adjust for center in the analyses

Answer 13

• Adjusting the centre as clustering effect
• Adjusting centre as fixed effect
• Adjusting centre as random effect

Question 14

essentially what is the difference between an unadjusted and adjusted analysis?

Answer 14

Unadjusted analyses asses populationaverage estimates of treatment effect

Adjusted analyses subgroup-specific estimates

Question 15

if we dont know the value of an adjusted variable what should we do?

Answer 15

use the the unadjusted analysis as the one for primary attention, the
adjusted analysis being supportive.