In this vignette we illustrate how to simulate joint effects of covariates in a pediatric population between 2 to 6 years old. Since Age and Weight in kids are highly correlated, we will not simulate varying one covariate at a time rather we will incorporate a distribution of realistic Age/Weight pairs. This approach is recommended when a database of realistic covariates distribution is available.
knitr::opts_chunk$set( collapse = TRUE, message =FALSE, warning =FALSE, fig.width = 7, comment = "#>" ) if (capabilities("cairo")) { knitr::opts_chunk$set(dev.args = list(png = list(type = "cairo"))) } library(coveffectsplot) library(ggstance) library(gamlss.dist) library(tidyr) library(dplyr) library(ggplot2) library(mrgsolve) library(ggridges) library(table1) library(data.table) library(ggh4x) library(ggpmisc) library(quantreg) theme_set(theme_bw()) nsubj <- 1000
Here we have a simple one-compartment PK model with first-order absorption where clearance and volume are allometrically scaled. The reference subject is a 4 year old female with a weight of 15.9 kg.
First, we plot a typical PK profile with between subject variability (BSV).
pedpkmodelcov <- ' $PARAM @annotated KA : 0.5 : Absorption rate constant Ka (1/h) CL : 4 : Clearance CL (L/h) V : 10 : Central volume Vc (L) CLWT : 0.75 : Weight on CL (ref. 22.5 kg) VWT : 1 : Weight on V (ref. 22.5 kg) $PARAM @annotated // reference values for covariate WT : 15.8 : Weight (kg) SEX : 0 : Sex (0=Female, 1=Male) AGE : 4 : Age (years) $CMT GUT CENT $MAIN double CLi = CL * pow((WT/15.8), CLWT)*exp(ETA(1)); double Vi = V * pow((WT/15.8), VWT)*exp(ETA(2)); double KAi = KA; double Keli = CLi/Vi; $OMEGA 0.09 0.01 0.09 $ODE dxdt_GUT = -KAi*GUT; dxdt_CENT = KAi*GUT-Keli*CENT; $TABLE double CP = CENT/ Vi; $CAPTURE CP KAi CLi Vi WT SEX AGE ' pedmodsim <- mcode("pedpkmodelcov", pedpkmodelcov) partab <- setDT(pedmodsim@annot$data)[block=="PARAM", .(name, descr, unit)] partab <- merge(partab, melt(setDT(pedmodsim@param@data), meas=patterns("*"), var="name")) knitr::kable(partab) idata <- data.table( ID = 1:nsubj, WT = c(rep(15.8,nsubj/2), rep(16.2,nsubj/2)),#from Nhanes at 4 years female and male AGE = 4, SEX = c(rep(0,nsubj/2),rep(1,nsubj/2)) ) ev1 <- ev(time = 0, amt = 100, cmt = 1) data.dose <- ev(ev1) data.dose <- setDT(as.data.frame(data.dose)) data.all <- data.table(idata, data.dose) set.seed(678549) outputsim <- pedmodsim %>% data_set(data.all) %>% mrgsim(end = 24, delta = 0.25)%>% as.data.frame %>% as.data.table outputsim$SEX <- as.factor(outputsim$SEX) outputsim$SEX <- factor(outputsim$SEX, labels=c("Girls","Boys")) age.labs <- c("age: 4 years") names(age.labs) <- c("4") wt.labs <- c("weight: 15.8 kg","weight: 16.2 kg") names(wt.labs) <- c("15.8","16.2") p1 <- ggplot(data = outputsim[,], aes(time, CP, group = ID)) + geom_line(alpha = 0.2, size = 0.1) + facet_grid(AGE ~ WT+SEX, labeller = labeller(AGE = age.labs, WT = wt.labs) , switch ="y") + scale_y_log10(breaks =c(0.001,0.01,0.01,0.1,1,10), labels = c("0.001","0.01","0.01","0.1","1","10")) + labs(y = "Plasma~Concentrations\n(logarithmic scale)", x = "Time (h)") p1
Second, we compute the PK parameters AUC and Cmax, standardize and compute between subject variability ranges.
derive.exposure <- function(time, CP) { n <- length(time) x <- c( Cmax = max(CP), AUC = sum(diff(time) * (CP[-1] + CP[-n])) / 2 ) data.table(paramname=names(x), paramvalue=x) } refbsv <- outputsim[, derive.exposure(time, CP), by=.(ID, WT, SEX, AGE)] refbsv[, stdparamvalue := paramvalue/median(paramvalue), by=list(SEX,paramname)] bsvranges <- refbsv[,list( P05 = quantile(stdparamvalue, 0.05), P25 = quantile(stdparamvalue, 0.25), P50 = quantile(stdparamvalue, 0.5), P75 = quantile(stdparamvalue, 0.75), P95 = quantile(stdparamvalue, 0.95)), by = list(SEX,paramname)] bsvranges yvar_names <- c( 'AUC'="AUC", 'Cmax'="Cmax" ) p4 <- ggplot(refbsv[,], aes( x = stdparamvalue, y = paramname, fill = factor(..quantile..), height = ..ndensity..)) + facet_grid(paramname+AGE~WT+SEX , scales="free_y", labeller=labeller(paramname=yvar_names, .cols =label_both, AGE = label_both) ,switch="y")+ stat_density_ridges( geom="density_ridges_gradient", calc_ecdf=TRUE, quantile_lines=TRUE, rel_min_height=0.001, scale=0.9, quantiles=c(0.05, 0.25, 0.5, 0.75, 0.95)) + scale_fill_manual( name="Probability", values=c("white", "#FF000050", "#FF0000A0", "#FF0000A0", "#FF000050", "white"), labels = c("(0, 0.05]", "(0.05, 0.25]", "(0.25, 0.5]", "(0.5, 0.75]", "(0.75, 0.95]", "(0.95, 1]")) + theme_bw() + theme( legend.position = "none", axis.text.y = element_blank(), axis.ticks.y = element_blank(), axis.title.y = element_blank()) + labs(x="Standardized PK Parameters", y="") + scale_x_log10() + coord_cartesian(expand=FALSE) p4 # p1<- p1 +theme_bw(base_size=18) # p4 <- p4+ theme_bw(base_size=18)+ # theme(axis.text.y=element_blank(),axis.ticks.y = element_blank(), # legend.position = "none") # egg::ggarrange(p1 , p4,ncol=2)
The NHANES website provides a csv file containing the smoothed growth charts distribution parameters at specific ages for boys and girls. The gamlss.dist::rBCCG
function is used to show how we can use these parameters to generate a realistic pediatric Age/Weight/Sex distribution. The wtage
dataset is now part of the package.
wtage<- wtage[wtage$Agemos<=6*12,] # keeps only 2 to 6 years wtage[wtage$Agemos>=4*12-1&wtage$Agemos<=4*12 +1,] %>% group_by(Sex) %>% summarize(Median=median(M)) nweightsperage <- 50 # simulate 50 kid at each age/sex simwtageoutput <- data.frame(matrix(NA, nrow = nrow(wtage),ncol = nweightsperage)) names(simwtageoutput) <- paste0("Var", 1:nweightsperage) set.seed(209321) wtage <- as.data.frame(wtage) for (i in 1:nrow(wtage)) {# simpoints <- gamlss.dist::rBCCG(nweightsperage, mu = wtage[i,"M"], sigma = wtage[i,"S"], nu = wtage[i,"L"]) simwtageoutput[i, ] <- simpoints } simwtageoutput$Agemos <- wtage$Agemos simwtageoutput$AgeY <- wtage$Agemos/12 simwtageoutput$Sex <- ifelse( wtage$Sex==2,0,1)#recode girls to 0, boys to 1 simwtageoutput <- tidyr::gather(simwtageoutput,age,Weight, paste0("Var", 1:nweightsperage)) simwtageoutput$age <- NULL simwtageoutput$SEXLABEL <- factor(simwtageoutput$Sex,labels=c("Girls","Boys")) wtvsageplot<- ggplot(simwtageoutput,aes(AgeY,Weight,color=SEXLABEL))+ geom_point(alpha=0.2,size=1.5)+ facet_grid(~SEXLABEL)+ labs(y="Weight (kg)", x= "Age (years)",col="") wtvsageplot
The section above generated 4900 Age/Weight/Sex distribution values that we will use for the simulation. We will not remove the between subject variability to be able to appreciate the range of possible PK concentrations. We show a plot of the PK profiles and the normalized PK parameters versus Age and versus Weight. Since we are dealing with a distribution and not specific covariate values we split into quartile ranges of the covariate distribution.
idata <- as.data.frame(simwtageoutput) names(idata) <- c("Agemos","AGE","SEX","WT","SEXLABEL") ev1 <- ev(time=0,amt=100, cmt=1) data.dose <- ev(ev1) data.dose<-as.data.frame(data.dose) data.all<-merge(idata,data.dose) data.all$ID <- 1: nrow(data.all) outcovcomb<- pedmodsim %>% data_set(data.all) %>% mrgsim(end=24, delta=1) outcovcomb<-as.data.frame(outcovcomb) outcovcomb <- outcovcomb %>% arrange(ID,time,SEX,AGE,WT) outcovcomb$SEX <- as.factor(outcovcomb$SEX) outcovcomb$SEX <- factor(outcovcomb$SEX,labels=c("Girls","Boys")) f <- function(x, xcat, which, what, from, to, ...) { what <- sub("of ", "of\n", what) what <- sub("median ", "median\n", what) sprintf("%s %s [%s to %s[", which, what, signif_pad(from, 3, FALSE), signif_pad(to, 3, FALSE)) } p3 <- ggplot(data =outcovcomb , aes(time, CP, group = ID,color=SEX)) + geom_line(alpha = 0.1, size = 0.3) + facet_grid( table1::eqcut(AGE,2,f) ~ table1::eqcut(WT,4,f) ) + labs(y = "Plasma Concentrations", color = "Sex", x = "Time (h)")+ theme(strip.placement = "outside",legend.position =c(0.9,0.2), legend.background = element_blank())+ guides(colour=guide_legend(override.aes = list(alpha=1,size=0.5))) p3
out.df.multivariatecov <- as.data.frame(outcovcomb) %>% arrange(ID,time) %>% group_by(ID,SEX,AGE,WT)%>% summarise (Cmax = max(CP,na.rm = TRUE), AUC= sum(diff(time ) *na.omit(lead(CP) + CP)) / 2) out.df.multivariatecov.long <- out.df.multivariatecov %>% gather(paramname,paramvalue,Cmax,AUC) %>% group_by (paramname,SEX) %>% mutate(medparam = median(paramvalue), paramvalue = paramvalue / medparam) out.df.multivariatecov.long$SEXLABEL <- factor(out.df.multivariatecov.long$SEX, labels=c("Girls","Boys")) nca.summaries <- out.df.multivariatecov.long %>% mutate(REF = "All Subjects") nca.summaries$WTCAT3 <- table1::eqcut( nca.summaries$WT,3,varlabel = "Weight") nca.summaries$WTCAT4 <- table1::eqcut( nca.summaries$WT,4,varlabel = "Weight") nca.summaries$AGECAT4 <- table1::eqcut( nca.summaries$AGE,4,varlabel = "Age") nca.summaries$AGECAT4_label <- nca.summaries$AGECAT4 f <- function(x, xcat, which, what, from, to, ...) { what <- sub("of ", "of\n", what) what <- sub("median ", "median\n", what) sprintf("%s %s [%s to %s[", which, what, signif_pad(from, 3, FALSE), signif_pad(to, 3, FALSE)) } plot1 <- ggplot(nca.summaries %>% filter(WT<=25) %>% filter(WT>12) %>% gather(key,value,WT,AGE) %>% mutate(key = ifelse(key=="WT","Weight (kg)", "Age (years)")) , aes(x=value,y=paramvalue,fill= SEX))+ annotate("rect", ymin = 0.6, ymax = 1.4, xmin = -Inf, xmax = Inf, fill = "gray", alpha = 0.4) + geom_hline(data=data.frame (yintercept=1), aes(yintercept =yintercept ),size = 1)+ geom_point(alpha = 0.2, size = 1) + stat_quant_band(method = "rqss",formula = y ~ qss(x, lambda = 2), quantiles = c(0.05, 0.5, 0.95), col = "black",aes(fill=SEX))+ facet_nested(SEX~paramname+key,scales="free",switch = "both")+ #stat_quantile()+ #facet_grid(SEX~key,scales="free",switch = "both")+ theme_bw(base_size = 18)+ labs(x="",y="Standardized PK Exposure", fill ="Smoothed 90%\nquantreg intervals")+ theme(strip.placement = "outside",legend.position = "none",axis.title.x = element_blank())+ scale_y_log10( breaks =c(0.25,0.5,1,2))+ coord_cartesian(ylim=c(0.25,2)) plot2 <- ggplot(nca.summaries %>% filter(WT<=25) %>% filter(WT>12) %>% gather(key,value,WT,AGE) %>% mutate(key = ifelse(key=="WT","Weight (kg)", "Age (years)")) , aes(x=value,y=paramvalue,fill= SEX))+ annotate("rect", ymin = 0.6, ymax = 1.4, xmin = -Inf, xmax = Inf, fill = "gray", alpha = 0.4) + geom_hline(data=data.frame (yintercept=1), aes(yintercept =yintercept ),size = 1)+ geom_point(alpha = 0.2, size = 1) + stat_quant_band(method = "rqss",formula = y ~ qss(x, lambda = 2), quantiles = c(0.05, 0.5, 0.95), col = "black",aes(fill=SEX))+ facet_nested(paramname ~SEX+key,scales="free",switch = "both", strip = strip_split(c("left", "top","bottom")))+ theme_bw(base_size = 18)+ labs(x="",y="Standardized PK Exposure", fill ="Smoothed 90%\nquantreg intervals")+ theme(strip.placement = "outside",legend.position = "none",axis.title.x = element_blank())+ scale_y_log10( breaks =c(0.25,0.5,1,2))+ coord_cartesian(ylim=c(0.25,2)) plot2
nca.summaries.long <- gather(nca.summaries, covname, covvalue,REF,WTCAT3,WTCAT4,AGECAT4,SEX, factor_key = TRUE) nca.summaries.long$covvalue <- as.factor( nca.summaries.long$covvalue) nca.summaries.long$covvalue <- reorder(nca.summaries.long$covvalue,nca.summaries.long$paramvalue) ggridgesplot<- ggplot(nca.summaries.long, aes(x=paramvalue,y=covvalue,fill=factor(..quantile..),height=..ndensity..))+ facet_grid(covname~paramname,scales="free_y")+ annotate("rect", xmin = 0.8, xmax = 1.25, ymin = -Inf, ymax = Inf, fill = "gray", alpha = 0.4) + stat_density_ridges( geom = "density_ridges_gradient", calc_ecdf = TRUE, quantile_lines = TRUE, rel_min_height = 0.01,scale=0.9, quantiles = c(0.05,0.5, 0.95))+ scale_fill_manual( name = "Probability", values = c("white","#0000FFA0", "#0000FFA0", "white"), labels = c("(0, 0.05]", "(0.05, 0.5]","(0.5, 0.95]", "(0.95, 1]") )+ geom_vline(data=data.frame (xintercept=1), aes(xintercept =xintercept ),size = 1)+ theme_bw()+ labs(x="Effects Of Covariates on PK Parameter",y="") ggridgesplot
Similarly to previous sections, we prepare the data to use forest_plot
. We provide a two parameters plot illustrating some of the options.
coveffectsdatacovrep <- nca.summaries.long %>% dplyr::group_by(paramname,covname,covvalue) %>% dplyr::summarize( mid= median(paramvalue), lower= quantile(paramvalue,0.05), upper = quantile(paramvalue,0.95)) %>% dplyr::filter(!is.na(mid)) %>% dplyr::filter(covname !="WTCAT4") bsvranges <- refbsv[,list( P05 = quantile(stdparamvalue, 0.05), P25 = quantile(stdparamvalue, 0.25), P50 = quantile(stdparamvalue, 0.5), P75 = quantile(stdparamvalue, 0.75), P95 = quantile(stdparamvalue, 0.95)), by = list(paramname)] setkey(bsvranges, paramname) coveffectsdatacovrepbsv <- coveffectsdatacovrep[coveffectsdatacovrep$covname=="REF",] coveffectsdatacovrepbsv$covname <- "BSV" coveffectsdatacovrepbsv$covvalue <- "90% of patients" coveffectsdatacovrepbsv$label <- "90% of patients" coveffectsdatacovrepbsv$lower <- bsvranges$P05 coveffectsdatacovrepbsv$upper <- bsvranges$P95 coveffectsdatacovrepbsv2 <- coveffectsdatacovrep[coveffectsdatacovrep$covname=="REF",] coveffectsdatacovrepbsv2$covname <- "BSV" coveffectsdatacovrepbsv2$covvalue <- "50% of patients" coveffectsdatacovrepbsv2$label <- "50% of patients" coveffectsdatacovrepbsv2$lower <- bsvranges$P25 coveffectsdatacovrepbsv2$upper <- bsvranges$P75 coveffectsdatacovrepbsv <- coveffectsdatacovrep coveffectsdatacovrepbsv <- coveffectsdatacovrepbsv %>% mutate( label = covvalue, LABEL = paste0(format(round(mid,2), nsmall = 2), " [", format(round(lower,2), nsmall = 2), "-", format(round(upper,2), nsmall = 2), "]")) coveffectsdatacovrepbsv<- as.data.frame(coveffectsdatacovrepbsv) coveffectsdatacovrepbsv$label <- gsub(": ", ":\n", coveffectsdatacovrepbsv$label) coveffectsdatacovrepbsv$covname <-factor(as.factor(coveffectsdatacovrepbsv$covname ), levels = c("WTCAT3","AGECAT4","SEX","REF"), labels = c("Weight\nSplits","Age\nSplits","Sex\nSplits","All")) coveffectsdatacovrepbsv$label <- factor(coveffectsdatacovrepbsv$label) coveffectsdatacovrepbsv$label <- factor(coveffectsdatacovrepbsv$label, levels =c( "1st quartile of Age:\n[2.00,2.96)" , "2nd quartile of Age:\n[2.96,3.96)" , "3rd quartile of Age:\n[3.96,4.96)" , "4th quartile of Age:\n[4.96,5.96]" , "Boys", "Girls", "All Subjects" , "1st tertile of Weight:\n[9.40,14.6)" , "2nd tertile of Weight:\n[14.6,17.4)" , "3rd tertile of Weight:\n[17.4,38.2]" )) interval_bsv_text = "" ref_legend_text = "Reference (vertical line)\nClinically relevant limits\n(gray area)" interval_legend_text <- "Median (points)\n90% intervals (horizontal lines) of joint effects: covariate distributions, uncertainty\nand between subject variability" png("./Figure_7_4.png",width = 11 ,height = 7,units = "in",res=72) coveffectsplot::forest_plot(coveffectsdatacovrepbsv, ref_area = c(0.6, 1/0.6), x_range = c(0.4,2.2), strip_placement = "outside", base_size = 18, y_label_text_size = 10,x_label_text_size = 10, xlabel = "Fold Change Relative to Reference", ref_legend_text =ref_legend_text, area_legend_text =ref_legend_text , interval_legend_text = interval_legend_text, interval_bsv_text = interval_bsv_text, plot_title = "", legend_shape_reverse = TRUE, facet_formula = "covname~.", facet_switch = "both", table_facet_switch = "both", reserve_table_xaxis_label_space = TRUE, facet_scales = "free_y", facet_space = "free", paramname_shape = TRUE, table_position = "right", table_text_size= 4, plot_table_ratio = 5, show_table_facet_strip = "x", show_table_yaxis_tick_label = FALSE, logxscale = TRUE, major_x_ticks = c(0.5,0.6,1,1/0.6,1/0.5), major_x_labels = c("1/2","0.6","1","1.66","2"), return_list = FALSE) dev.off()
Any scripts or data that you put into this service are public.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.