swPwr: Power of Stepped Wedge Cluster Randomized Trial (SW CRT)

Description

swPwr returns (two-sided) power for the treatment effect(s) from an immediate treatment effect (IT) model or an exposure time indicator (ETI) model (Kenny et al, 2022) for the specified SW CRT design using a linear models weighted least squares (WLS) approach. A cross-sectional or closed cohort sampling scheme can be specified. The response/outcome of interest can be binomial or Gaussian distributed and is assumed to come from a model with random intercepts, random treatment effects, random cluster-specific time effects, and, in the case of closed cohort sampling, an individual random effect. Variance components can be specified using either random effect variances (tau, eta, rho, gamma, and zeta) or icc, cac, iac (see details). If a cross-sectional sampling, random intercepts only model is used (i.e., eta, rho, gamma and zeta are 0 and n is constant over clusters and time), then the power calculation is comparable to the closed-form formula of [Hussey and Hughes, 2007]. See [Voldal et al., 2020] for more guidance. This function is also available as a Shiny app at https://swcrtdesign.shinyapps.io/stepped_wedge_power_calculation/.

Usage

swPwr(design, distn, n, mu0, mu1, H=NULL, sigma, tau=0, eta=0, rho=0, gamma=0, zeta=0,
icc=0, cac=1, iac=0, alpha=0.05, retDATA=FALSE, silent=FALSE)

Value

numeric (vector): swPwr returns the power of treatment effect(s), where the variance of treatment effect is computed by WLS.

numeric (list): swPwr returns all specified and computed items as objects of a list if retDATA = TRUE.

design: list: The stepped wedge design object as input.
distn: character: Distribution assumed (gaussian or binomial).
n: integer (scalar, vector, or matrix): Number of observations: (scalar) for all clusters and all time points; (vector) for each cluster at all time points; and (matrix) for each cluster at each time point, where rows correspond to clusters and columns correspond to time.
mu0: numeric (scalar): Mean outcome in the control group.
mu1: numeric (scalar): Mean outcome in intervention group(s). Note: treatment effect is difference in means $\theta = \mu_1 - \mu_0$.
H: numeric (vector): H specifies the desired linear combination of exposure time treatment effects for a ETI model-based estimate.
sigma: numeric (scalar): Standard deviation input. For binomial distribution, sigma = NA
tau: numeric (scalar): Standard deviation of random intercepts.
eta: numeric (scalar): Standard deviation of random treatment effects.
rho: numeric (scalar): Correlation between random intercepts and random treatment effects.
gamma: numeric (scalar): Standard deviation of random time effects.
zeta: numeric (scalar): Standard deviation of individual random effects for closed cohort sampling.
icc: numeric (scalar): Within-period intra-cluster correlation. Can be used with CAC and IAC instead of tau, eta, rho, gamma and zeta; see details.
cac: numeric (scalar): Cluster auto-correlation. Can be used with ICC and IAC instead of tau, eta, rho, gamma and zeta; see details.
iac: numeric (scalar): Individual auto-correlation for closed cohort sampling. Can be used with ICC and CAC instead of tau, eta, rho, gamma and zeta; see details.
alpha: numeric (scalar): Statistical significance level.
Xmat: numeric (matrix): Design matrix for this SW CRT design.
Wmat: numeric (matrix): Covariance matrix for this SW CRT design.
var: Variance-covariance matrix of treatment effect(s). Can be used to calculate power for constrasts other than control versus treatment
var.theta.WLS: numeric (scalar): Variance(s) of treatment effect(s) using weighted least squares (WLS) for this SW CRT design.
pwrWLS: numeric (scalar): Power of treatment effect ($\theta$) using weighted least squares (WLS) for this SW CRT design.

Arguments

design: list: A stepped wedge design object, typically from swDsn, that includes at least the following components: swDsn, swDsn.unique.clusters, clusters, n.clusters, total.time, nTxLev, TxLev. Fractional treatment effects specified in swDsn are used in the calculation.
distn: character: Distribution assumed (gaussian or binomial). Currently, 'Binomial' implies Bernoulli.
n: integer (scalar, vector, or matrix): Number of observations: (scalar) for all clusters and all time points that are not NA in design$swDsn; (vector) for each cluster at all time points that are not NA in design$swDsn; and (matrix) for each cluster at each time point, where rows correspond to clusters and columns correspond to time. n can also be used to specify a design with transition periods (a transition period is a period, typically, immediately after treatment is introduced were no data are collected). Simply define n as a matrix with a sample size of 0 during every transition period. This is equivalent to putting an NA in design$swDsn.
mu0: numeric (scalar): Mean outcome in the control group. See also documentation for H, below.
mu1: numeric (scalar or vector): Mean outcome in the treatment group(s). The treatment effect is the difference in means $\theta = \mu_1 - \mu_0$. If tx.effect.frac is specified in the call to swDsn then the assumed treatment effect at exposure time k is tx.effect.frac[k]*$\theta$. For an IT model with multiple treatment levels the length of mu1 should be equal to the number of treatment levels. See also documentation for H, below.
H: numeric (vector): If NULL, then swPwr assumes an immediate, constant treatment effect (IT) model. If not NULL, then an exposure time indicator (ETI) model is assumed and H should be a vector as long as the longest exposure time (in a classic stepped wedge design, the longest exposure time is the number of time periods minus one). H specifies the desired linear combination of exposure time treatment effects. For example, in a stepped wedge trial with 5 time periods and four exposure times, H = rep(.25,4) gives the average treatment effect over the four exposure times; H = c(0,0,.5,.5) ignores the first two periods after the intervention is introduced and averages the remaining periods. Typically, the sum of H is 1.0; if not, it is renormalized to sum to 1.0. H can only be specified when there is a single intervention level (i.e. design$swDsn includes only NA,0,1; see cautions in help for swDsn about including NA periods when H is used). mu0 and mu1 give expected values of the linear combination of exposure time treatment effects under the null and alternative hypotheses, respectively. Do not use an ETI model (non-null H) if you specified a fractional treatment effect in swDsn.
sigma: numeric (scalar): Standard deviation when assuming Gaussian distribution (distn=gaussian). For binomial distribution $\sigma^2$ is automatically set to $\bar{\mu}(1-\bar{\mu})$ where $\bar{\mu} = (\mu_1 + \mu_0)/2$
tau: numeric (scalar): Standard deviation of random intercepts.
eta: numeric (scalar): Standard deviation of random treatment effects.
rho: numeric (scalar): Correlation between random intercepts and random treatment effects.
gamma: numeric (scalar): Standard deviation of random time effects.
zeta: numeric (scalar): Standard deviation of individual effects for closed cohort sampling. Default is 0 which implies cross-sectional sampling. Values greater than 0 imply closed cohort sampling.
icc: numeric (scalar): Within-period intra-cluster correlation. Can be used with CAC instead of tau, eta, rho, and gamma; see details.
cac: numeric (scalar): Cluster auto-correlation. Can be used with ICC instead of tau, eta, rho, and gamma; see details.
iac: numeric (scalar): Individual auto-correlation for closed cohort sampling. Can be used with ICC and CAC instead of tau, eta, rho, gamma, and zeta; see details. Default is 0 which implies cross-sectional sampling. Values greater than 0 imply closed cohort sampling.
alpha: numeric (scalar): Two-sided statistical significance level.
retDATA: logical: if TRUE, all stored (input, intermediate, and output) values of swPwr are returned. Default value is FALSE.
silent: logical: if TRUE, hides most warning messages. Default value is FALSE.

Author

James P Hughes, Navneet R Hakhu, and Emily C Voldal

Details

The two-sided statistical power of treatment effect $(\theta = \mu_1 - \mu_0)$ is $$ Pwr(\theta) = \Phi( Z - z_{1 - \alpha /2} ) + \Phi( -Z - z_{1 - \alpha /2} ) $$ where $$ Z = \frac{ |\theta| }{ \sqrt{Var(\hat{\theta}_{WLS})} } $$ and $\Phi$ is the cumulative distribution function of the standard normal $N(0,1)$ distribution. If H is non-NULL then the $\mu$ are assumed to be equal to H$\delta$ where $\delta$ is a vector of exposure time treatment effects.

The value of zeta defines the samling scheme. When zeta = 0, cross-sectional sampling is assumed; if zeta > 0 then closed cohort sampling is assumed.

When eta (and rho) are 0, instead of using tau, eta, rho, gamma and zeta, the icc, cac and iac can be used to define the variability of the random effects. In this model, $$ ICC = \frac{\tau^2+\gamma^2}{\tau^2+\gamma^2+\zeta^2+\sigma^2} $$ $$ CAC = \frac{\tau^2}{\tau^2+\gamma^2} $$ $$ IAC = \frac{\zeta^2}{\zeta^2+\sigma^2}$$ Choose one parameterization or the other, do not mix parameterizations.

References

Hussey MA, Hughes JP. Design and analysis of stepped wedge cluster randomized trials. Contemporary Clinical Trials 2007;28:182-191.

Kenny A, Voldal E, Xia F, Heagerty PJ, Hughes JP. Analysis of stepped wedge cluster randomized trials in the presence of a time-varying treatment effect. Statistics in Medicine, in press, 2022.

Voldal EC, Hakhu NR, Xia F, Heagerty PJ, Hughes JP. swCRTdesign: An R package for stepped wedge trial design and analysis. Computer Methods and Programs in Biomedicine 2020;196:105514.

Examples

Run this code

# Example 1 (Random Intercepts Only, standard Stepped Wedge (SW) design)
swPwr.Ex1.RIO.std <- swPwr(swDsn(c(6,6,6,6)), distn="binomial",
n=120, mu0=0.05, mu1=0.035, tau=0.01, eta=0, rho=0, gamma=0, alpha=0.05, retDATA=FALSE)
swPwr.Ex1.RIO.std

# Example 2 (Random Intercepts Only, extended SW design)
swPwr.Ex1.RIO.extend <- swPwr(swDsn(c(6,6,6,6), extra.trt.time=3), distn="binomial",
n=120, mu0=0.05, mu1=0.035, tau=0.01, eta=0, rho=0, gamma=0,
alpha=0.05, retDATA=FALSE)
swPwr.Ex1.RIO.extend

# Example 3 (Independent Random Intercepts and Treatment effects, standard SW design)
swPwr.Ex1.IRIS <- swPwr(swDsn(c(6,6,6,6)), distn="binomial",
n=120, mu0=0.05, mu1=0.035, tau=0.01, eta=0.0045, rho=0, gamma=0,
alpha=0.05, retDATA=FALSE)
swPwr.Ex1.IRIS

# Example 4 (Correlated Random Intercepts and Slopes, standard SW design)
swPwr.Ex1.CRIS <- swPwr(swDsn(c(6,6,6,6)), distn="binomial",
n=120, mu0=0.05, mu1=0.035, tau=0.01, eta=0.0045, rho=0.4, gamma=0,
alpha=0.05, retDATA=FALSE)
swPwr.Ex1.CRIS

# Example 5 (Random time effect and correlated Random Intercepts and Slopes, standard SW design)
swPwr.Ex1.RTCRIS <- swPwr(swDsn(c(6,6,6,6)), distn="binomial",
n=120, mu0=0.05, mu1=0.035, tau=0.01, eta=0.0045, rho=0.4, gamma = 0.1,
alpha=0.05, retDATA=FALSE)
swPwr.Ex1.RTCRIS

#Example 6 (Sample size varying by cluster)
sample.size.vector <- c(35219,53535,63785,456132,128670,96673,
           51454,156667,127440,68615,56502,17719,
           75931,58655,52874,75936)
swPwr.Ex1.vector <- swPwr(swDsn(c(4,3,5,4)), distn="gaussian",
n=sample.size.vector, mu0=2.66, mu1=2.15,
sigma=sqrt(1/2.66), tau=0.31, eta=0.2, rho=0, gamma = 0.15,
alpha=0.05, retDATA=FALSE)
swPwr.Ex1.vector

#Example 7 (Sample size varying by cluster and time)
sample.size.matrix <- matrix(c(26, 493,  64,  45,  48, 231, 117,  17,  49,  36,  19,  77, 67, 590,
261, 212,  67, 318, 132,  58,  44,  57,  59,  78, 115, 532, 176, 199,  73, 293, 129,  79,  51,
62, 109,  94, 174, 785, 133,  79, 120, 305, 224,  99,  83,  79, 122, 122, 94, 961,  90, 131, 166,
352, 316,  59,  54, 131, 101, 133),nrow=12,ncol=5, byrow=FALSE)
swPwr.Ex1.matrix <- swPwr(swDsn(c(3,3,3,3)), distn="binomial",
n=sample.size.matrix, mu0=0.08, mu1=0.06, tau=0.017, eta=0.006, rho=-0.5, gamma = 0,
alpha=0.05, retDATA=FALSE)
swPwr.Ex1.matrix

#Example 8 (Using ICC and CAC)
swPwr.Ex1.icccac <- swPwr(swDsn(c(6,6,6,6)), distn="gaussian",
n=120, mu0=0.05, mu1=0.035, sigma=0.1, icc=0.02, cac=0.125, alpha=0.05, retDATA=FALSE)
swPwr.Ex1.icccac

# Example 9 (Random time effect, closed cohort sampling)
sample_size = matrix(c(rep(c(20,20,20,20,20,20,20,20,20,20,20,20,20,20),5),
                       rep(c( 0,20,20,20,20,20,20,20,20,20,20,20,20,20),5),
                       rep(c( 0, 0,20,20,20,20,20,20,20,20,20,20,20,20),5),
                       rep(c( 0, 0, 0,20,20,20,20,20,20,20,20,20,20,20),5),
                       rep(c( 0, 0, 0, 0,20,20,20,20,20,20,20,20,20,20),5)),
                       25,14,byrow=TRUE)
swPwr.Ex9 <- swPwr(design=swDsn(c(5,5,5,5,5),extra.ctl.time=3,extra.trt.time=5),
      distn="gaussian",n=sample_size,mu0=-0.4,mu1=0.4,
      sigma=1.0,tau=sqrt(.1316),gamma=sqrt(.1974),eta=0,rho=0,zeta=sqrt(2.5),
      silent=TRUE)
swPwr.Ex9

#Example 10 (Periods with no data, multiple treatment levels)
stdy <- swDsn(c(6,6,6,6),swBlk=matrix(c(0,1,2,2,2,2,
                                         NA,0,1,2,2,2,
                                         NA,NA,0,1,2,2,
                                         NA,NA,NA,0,1,2),4,6,byrow=TRUE))
swPwr.Ex10 <-swPwr(stdy, distn="binomial",n=120, mu0=0.05, mu1=c(0.035,0.03),
      tau=0.01, eta=0, rho=0, gamma=0, alpha=0.05, retDATA=TRUE, silent=TRUE)
swPwr.Ex10

Run the code above in your browser using DataLab