ipdmeta (version 2.4)

ipd.sep: IPD meta-analysis Subgroup Effect Power Estimator

Description

The function estimates the power of an IPD meta-analysis to detect a specified subgroup effect (covariate-treatment interaction) based on summary statistics.

Usage

ipd.sep(
	      effect, event0=NULL, event1=NULL, mean0=NULL, mean1=NULL, var0=NULL, var1=NULL, x0=NULL, x1=NULL, s20=NULL, s21=NULL, n0=NULL, n1=NULL,
	       data, alpha=.05
	       )

Arguments

effect

scalar, subgroup effect under alternative hypothesis

event0

vector, for binary outcome, events in group 0

event1

vector, for binary outcome, events in group 1

mean0

vector, for continuous outcome, mean in group 0

mean1

vector, for continuous outcome, mean in group 1

var0

vector, for contunuous outcome, sample variances for responses in group 0

var1

vector, for contunuous outcome, sample variances for responses in group 1

x0

vector of subgroup covariate means for group 0

x1

vector of subgroup covariate means for group 1

s20

vector of covariate sample variances for control group

s21

vector of covariate sample variances for treatment group

n0

vector of number of subjects for group 0

n1

vector of number of subjects for group 1

data

data frame containing the objects specified in response or covariate arguments

alpha

scalar significance level of Wald test (two-sided)

Value

A list with the following named components:

esimated.power

The estmated IPD meta-analysis interactive effect power

power.lower

Lower bound for level CI

power.upper

Upper bound for level CI

estimated.se

Estimated standard error of IPD meta-analysis interaction effect

se.lower

Lower bound for level CI

se.upper

Upper bound for level CI

sigma

The mean of the study residual variance

sigma0

Estimate of intercept random effect variance from simple RE meta-analysis with DSL estimator

sigma1

Estimate of treatment random effect variance simple RE meta-analysis with DSL estimator

Details

If a data frame is supplied, then the object indicated in each vector argument is looked for in data.

For a patient-level binary outcome, mean0, mean1, var0 and var1 should not be specified. Zero event counts will be corrected with a 0.5 factor. For a continuous response, event0 and event1 should not be specified.

For a covariate that is a mean proportion, such as proportion male, no sample variances need to be specified. If no values are given for the sample variances s20 and s21 it will be assumed that the covariate is a mean proportion and the sample variances will be determined from the proportions.

The SEP for the IPD meta-analysis is based on a generalized linear mixed model for the patient-level analysis. The model has intercept, treatment, covariate and interaction fixed effects and independent random effects for the baseline and treatment by study. Under this model, an estimator for the subgroup effect variance, that is, the variance for the estimate of the covariate-treatment interaction, for either an identity or logistic GLMM, can be obtained from the study sample statistics. This variance is then used to estimate the power of the IPD meta-analysis for a specified subgroup effect based on a two-sided Wald test.

Examples

data(poynard)

#AGE SEP FOR IPD META-ANALYSIS OF BETA-ANTAGONISTS TO PREVENT GI BLEEDING EVENTS

#ALTERNATIVE HYPOTHESIS FOR AGE-TREATMENT EFFECT
#WITH 10 YEARS CHANGE TO OR TREATMENT EFFECT exp(beta*10)
#EFFECT MODIFIER CHANGES TREATMENT EFFECT BY 30%

beta = log(1.3)/10

age.sep <- 

ipd.sep(
 effect=beta,
 event0=bleed0,
 event1=bleed1,
 n0=n0,
 n1=n1,
 x0=age0,
 x1=age1,
 s20=age.s20,
 s21=age.s21,
 data=poynard
)

age.sep

#GENDER SUBGROUP EFFECT; 30% OR CHANGE BY GENDER

beta <- log(1.3)

gender.sep <- 

ipd.sep(
 effect=beta,
 event0=bleed0,
 event1=bleed1,
 n0=n0,
 n1=n1,
 x0=male0,
 x1=male1,
 data=poynard
)

gender.sep