BuyseMultComp: Adjustment for Multiple Comparisons

In BuyseTest: Generalized Pairwise Comparisons

Adjustment for Multiple Comparisons

Description

Adjust p-values and confidence intervals estimated via GPC for multiple comparisons.

Usage

BuyseMultComp(
  object,
  cluster = NULL,
  linfct = NULL,
  rhs = NULL,
  endpoint = NULL,
  statistic = NULL,
  cumulative = TRUE,
  conf.level = NULL,
  band = TRUE,
  global = FALSE,
  alternative = NULL,
  transformation = NULL,
  ...
)

Arguments

object	A BuyseTest object or a list of BuyseTest objects. All objects should contain the same endpoints.
cluster	[character] name of the variable identifying the observations in the dataset used by each BuyseTest model. Only relevant when using a list of BuyseTest objects to correctly combine the influence functions. If NULL, then it is assumed that the BuyseTest objects correspond to different groups of individuals.
linfct	[numeric matrix] a contrast matrix of size the number of endpoints times the number of BuyseTest models.
rhs	[numeric vector] the values for which the test statistic should be tested against. Should have the same number of rows as linfct.
endpoint	[character or numeric vector] the endpoint(s) to be considered.
statistic	[character] the statistic summarizing the pairwise comparison: "netBenefit" displays the net benefit, as described in Buyse (2010) and Peron et al. (2016)), "winRatio" displays the win ratio, as described in Wang et al. (2016), "favorable" displays the proportion in favor of the treatment (also called Mann-Whitney parameter), as described in Fay et al. (2018). "unfavorable" displays the proportion in favor of the control. Default value read from BuyseTest.options().
cumulative	[logical] should the summary statistic be cumulated over endpoints? Otherwise display the contribution of each endpoint.
conf.level	[numeric] confidence level for the confidence intervals. Default value read from BuyseTest.options().
band	[logical] Should confidence intervals and p-values adjusted for multiple comparisons be computed.
global	[logical] Should global test (intersection of all null hypotheses) be made?
alternative	[character] the type of alternative hypothesis: "two.sided", "greater", or "less". Default value read from BuyseTest.options().
transformation	[logical] should the CI be computed on the logit scale / log scale for the net benefit / win ratio and backtransformed. Otherwise they are computed without any transformation. Default value read from BuyseTest.options(). Not relevant when using permutations or percentile bootstrap.
...	argument passsed to the function transformCIBP of the riskRegression package.

Details

Simulateneous confidence intervals and adjusted p-values are computed using a single-step max-test approach via the function transformCIBP of the riskRegression package. This corresponds to the single-step Dunnett described in Dmitrienko et al (2013) in table 2 and section 7.

References

Dmitrienko, A. and D'Agostino, R., Sr (2013), Traditional multiplicity adjustment methods in clinical trials. Statist. Med., 32: 5172-5218. https://doi.org/10.1002/sim.5990

Examples

#### simulate data ####
set.seed(10)
df.data <- simBuyseTest(1e2, n.strata = 3)

#### adjustment for all univariate analyses ####
ff1 <- treatment ~ TTE(eventtime, status = status, threshold = 0.1)
ff2 <- update(ff1, .~. + cont(score, threshold = 1))
BT2 <- BuyseTest(ff2, data= df.data, trace = FALSE)

## (require riskRegression >= 2021.10.04 to match)
confint(BT2, cumulative = FALSE) ## not adjusted
confintAdj <- BuyseMultComp(BT2, cumulative = FALSE, endpoint = 1:2) ## adjusted
confintAdj
cor(confintAdj$iid) ## correlation between test-statistic

#### 2- adjustment for multi-arm trial ####
## case where we have more than two treatment groups
## here strata will represent the treatment groups
df.data$strata <- as.character(df.data$strata)
df.data$id <- paste0("Id",1:NROW(df.data)) ## define id variable

BT1ba <- BuyseTest(strata ~ TTE(eventtime, status = status, threshold = 1),
                   data= df.data[strata %in% c("a","b"),], trace = FALSE)
BT1ca <- BuyseTest(strata ~ TTE(eventtime, status = status, threshold = 0.1),
                   data= df.data[strata %in% c("a","c"),], trace = FALSE)
BT1cb <- BuyseTest(strata ~ TTE(eventtime, status = status, threshold = 0.1),
                   data= df.data[strata %in% c("b","c"),], trace = FALSE)
rbind("b-a" = confint(BT1ba),
      "c-a" = confint(BT1ca),
      "c-b" = confint(BT1cb)) ## not adjusted
confintAdj <- BuyseMultComp(list("b-a" = BT1ba, "c-a" = BT1ca, "c-b" = BT1cb),
                            cluster = "id", global = TRUE)
confintAdj
dim(confintAdj$iid) ## number of subjects x number of analyses
cor(confintAdj$iid)

BuyseTest documentation built on March 31, 2023, 6:55 p.m.