data-raw/data.r
In chjackson/adverse: Data on adverse events of adjuvant breast cancer treatment
library(readxl)
library(tidyverse)
filter <- dplyr::filter
matches <- dplyr::matches
library(devtools)
load_all("../../gemtc/gemtc")
load_all("..")
# for(i in list.files("../data", full.names=TRUE, include.dirs=TRUE)) load(i)
bpaeraw <- read_excel("Bisphosphonates DOUBLE CHECKED MASTER 12-7-19.xlsx", range="A2:CWJ62")
header <- unlist(read_excel("Bisphosphonates DOUBLE CHECKED MASTER 12-7-19.xlsx", col_names=FALSE, range="A1:CWJ1"))
##' TIDY COLUMN NAMES
aenames <- as.character(na.omit(header[27:length(header)]))
## Fill in gaps in first row of column names by carrying forward
header <- data.frame(header=header, stringsAsFactors=FALSE) %>%
fill(header) %>% unlist
names(bpaeraw) <- gsub("^C1(.+)", "A1\\1", names(bpaeraw))
names(bpaeraw) <- gsub("^C2(.+)", "A2\\1", names(bpaeraw))
names(bpaeraw) <- gsub("^E1(.+)", "A3\\1", names(bpaeraw))
names(bpaeraw) <- gsub("^E2(.+)", "A4\\1", names(bpaeraw))
names(bpaeraw) <- gsub("^E3(.+)", "A5\\1", names(bpaeraw))
## Prepend the adverse event names to the arm numbers
fullnames <- paste(header, names(bpaeraw), sep=";")
bpaeraw <- rename(bpaeraw, study=`Trial name`, studyid=`Trial Code`)
fullnames <- gsub("__[0-9]+$", "", fullnames)
fullnames <- gsub("C1$", "A1", fullnames)
fullnames <- gsub("C2$", "A2", fullnames)
fullnames <- gsub("E1$", "A3", fullnames)
fullnames <- gsub("E2$", "A4", fullnames)
fullnames <- gsub("E3$", "A5", fullnames)
aeinds <- 26:ncol(bpaeraw)
names(bpaeraw)[aeinds] <- fullnames[aeinds]
## denominator put into control instead of experimental here
bpaeraw[bpaeraw$study=="REFORM","N Exp2"] <- bpaeraw[bpaeraw$study=="REFORM","N Control"]
bpaeraw[bpaeraw$study=="REFORM","N Control"] <- NA
bpaeraw[bpaeraw$study=="Pivot 2011","N Exp2"] <- bpaeraw[bpaeraw$study=="Pivot 2011","N Control"]
bpaeraw[bpaeraw$study=="Pivot 2011","N Control"] <- NA
## Remove duplicated study "NCT00172068"
bpaeraw <- bpaeraw[bpaeraw$studyid != 49,]
## This one has 56 vs 45 people in two experimental arms, rather than 127 in single experimental arm? Zoledronic acid in community vs hospital setting
#bpaeraw %>% select(`Trial name`)
bpaeraw <- bpaeraw %>% filter(study != "Wardley2005")
# Single arm study shouldn't be problem for the NMA. Include this for now
bpaeraw <- bpaeraw %>%
rename(reporting="Adverse reporting",
N1="N Control",
N2="N Control2",
N3="N Exp1",
N4="N Exp2",
N5="N Exp3") %>%
mutate(has_control = as.numeric(!is.na(`A1 coding`) | !is.na(`A2 coding`))) %>%
mutate(reporting = recode(reporting,
`1` = "Complete", `2` = "Table of highest", `3` = "Less")) %>%
mutate(reporting = factor(reporting, levels = c("Complete","Table of highest","Less","x"))) %>%
rename(meno = "Pre/post menopausal",
bonemeta = "bone metasteses") %>%
mutate(meno = recode(meno, `1` = "pre", `2` = "post", `3` = "mixed"),
bonemeta = recode(bonemeta, `1` = "all", `2` = "none", `3` = "mixed"),
meno = factor(meno, levels = c("mixed", "pre", "post")),
bonemeta = factor(bonemeta, levels = c("none", "mixed", "all"))
)
## Remove character notes in numeric cells
## AZURE. Spreadsheet reports "Myalgia" but AZURE reports "Myalgia/arthralgia"
## NCT00376740 classifies myalgia under arthralgia/joint pain
## Arbirtrarily assign these to myalgia, arthralgia/joint pain respectively
bpaeraw[39,"ARTHRALGIA/JOINT PAIN (ungraded or grade 1/2);A4...161"] <- NA
bpaeraw[,"ARTHRALGIA/JOINT PAIN (ungraded or grade 1/2);A4...161"] <-
as.numeric(bpaeraw$"ARTHRALGIA/JOINT PAIN (ungraded or grade 1/2);A4...161")
mcol <- grep("MYALGIA ungraded;A4",colnames(bpaeraw))
bpaeraw[26,mcol] <- NA
bpaeraw$"MYALGIA ungraded;A4...1733" <-
as.numeric(bpaeraw$"MYALGIA ungraded;A4...1733")
# Remove extraneous comma
acol <- grep("ANY ADVERSE EVENT;A1",colnames(bpaeraw))
bpaeraw[13,acol] <- gsub(",","",bpaeraw[13,acol])
bpaeraw$"ANY ADVERSE EVENT;A1...30" <- as.numeric(bpaeraw$"ANY ADVERSE EVENT;A1...30")
## Recode 2x2arm studies as two trials
bpaeraw$study[bpaeraw$studyid=="8.1"] <- "ABCSG12_B"
bpaeraw$study[bpaeraw$studyid=="30.1"] <- "Kristensen 2008_B"
use_data(bpaeraw, overwrite=TRUE)
## Note VonMoos, Pivot have same treatments in E1 and E2.
## Categories of AEs to be summed over, aggregating different grades
aecat <- scan("aecategs.txt", what="character", sep="\n") # one element per col of bpaeraw
aecategs <- unique(aecat)[15:143]
use_data(aecategs, overwrite=TRUE)
bpaeraw[1:2,1:20]
##' Reshape data into one row per arm and event
bpaelong <- bpaeraw %>%
gather(vname, x,
matches("^N[[:digit:]]"),
matches("A[[:digit:]]"),
matches(";A[[:digit:]]")
) %>%
select(study, studyid, reporting, meno, bonemeta, has_control, vname, x) %>%
extract(vname, "armno1", "^(?:A|N)([[:digit:]]).*", remove=FALSE) %>%
extract(vname, "armno2", ";A([[:digit:]]).*$", remove=FALSE) %>%
mutate(armno = ifelse(is.na(armno1), armno2, armno1)) %>%
select(-armno1, -armno2) %>%
mutate(armno = as.numeric(armno),
vname = gsub("N[[:digit:]]","N",vname),
vname = gsub("A[[:digit:]] (additional coding)", "\\1", vname),
vname = gsub("A[[:digit:]] (coding)", "\\1", vname),
vname = gsub(";A[[:digit:]]$","",vname)) %>%
mutate(aecat = aecat[match(vname, gsub(";A[0-9]$","",names(bpaeraw)))],
aecat = ifelse(is.na(aecat), vname, aecat)) %>%
mutate(vname = gsub("ungraded or grade 1/2", "grade 1/2", vname, ignore.case=TRUE))
bpaelong <- bpaelong %>%
mutate(grade = gsub(".+GRADE ([1-4]).*", "\\1", vname, ignore.case=TRUE)) %>%
mutate(grade = ifelse(grepl("ungraded", grade, ignore.case=TRUE), NA, grade)) %>%
mutate(grade = ifelse(vname %in% aecategs[-(1:3)], NA, grade)) %>%
mutate(grade = as.numeric(ifelse(grade %in% c(1,2,3,4), grade, NA)))
use_data(bpaelong, overwrite=TRUE)
## Sum AEs of same type within each arm
bpaelongsum <- bpaelong %>%
filter(!is.na(x)) %>%
group_by(study, studyid, reporting, meno, bonemeta, armno, aecat) %>%
summarise(
xser = sum(x[!is.na(grade) & grade>=3]),
xgraded = sum(x[!is.na(grade) & grade>=1]),
x = sum(x)
) %>%
rename(vname=aecat) %>%
ungroup()
use_data(bpaelongsum, overwrite=TRUE)
## Data with one row per arm, and cols for each aggregate AE type
bpae <- bpaelongsum %>%
select(-xser, -xgraded) %>%
spread(vname, x) %>%
filter(!is.na(coding), # remove empty arms
!is.na(N)) %>%
mutate(treatment = as.character(coding)) %>%
mutate(addtrt = as.character(`additional coding`)) %>%
replace_na(list(addtrt = "unknown"))
## Treatment description database
## Variables ending in 0 merge placebo and observation-only
bpcoding <- bpae %>%
select(treatment, addtrt) %>%
distinct %>%
extract(treatment,
c("drug", "dose", "delivery"),
"^([0-9])([0-9])([0-9])$", remove=FALSE) %>%
mutate(drugnameshort = drug,
drugname0 = recode(drug,
`1` = "Control",
`7` = "Denosumab",
`2` = "Zoledronic_acid",
`3` = "Pamidronate",
`4` = "Ibandronate",
`5` = "Risedronate",
`6` = "Clodronate"),
drugnameshort = recode(drugnameshort,
`1` = "Con",
`7` = "Den",
`2` = "Zol",
`3` = "Pam",
`4` = "Iba",
`5` = "Ris",
`6` = "Clo"),
drugname = ifelse(treatment %in% c("100","101"), "Placebo",
ifelse(treatment == "103", "Observation", drugname0)),
drugnameshort = ifelse(treatment %in% c("100","101"), "Pla",
ifelse(treatment == "103", "Obs", drugnameshort)),
delivery = recode(delivery,
`0` = "IV",
`1` = "Oral",
`2` = "IVAndOral",
`3` = "Observation")) %>%
## Two studies included arms only distinguished by using a non-AI hormone (141)
mutate(notai = ifelse((treatment == 103 & addtrt == 141)|
(treatment == 220 & addtrt == 141) , 1, 0)) %>%
## Consider BP + non-AI in same way as a different dose of the BP
mutate(dose = ifelse(notai, paste0(dose, "notai"), dose)) %>%
mutate(treatment = ifelse(notai, paste0(treatment, "notai"), treatment)) %>%
mutate(description = paste(drugname, dose, delivery, sep="_")) %>%
mutate(descriptionshort = paste(drugnameshort, dose, delivery, sep="_")) %>%
## don't include delivery method for placebo in "drugdose" and "drugdm"
mutate(drugdose = ifelse(drugname0 =="Control", drugname, description)) %>%
mutate(drugdm = paste(drugname, delivery, sep="_")) %>%
mutate(drugdm = ifelse(drugname=="Placebo", drugname, drugdm)) %>%
mutate(
drugdm = ifelse(drugdm=="Observation_Observation", "Observation", drugdm),
drugdm = ifelse(drugdm=="Observation_0notai_Observation", "Observation", drugdm),
description = ifelse(description=="Observation_0_Observation", "Observation", description),
description = ifelse(description=="Observation_0notai_Observation", "Observation_notai", description),
descriptionshort = ifelse(descriptionshort=="Obs_0_Observation", "Obs", descriptionshort),
descriptionshort = ifelse(descriptionshort=="Obs_0notai_Observation", "Obs_notai", descriptionshort)
) %>%
mutate(drugdm0 = ifelse(drugname0=="Control", drugname0, drugdm)) %>%
mutate(drugnitro = fct_collapse(drugname,
`Nitrogenous` = c("Zoledronic_acid",
"Pamidronate",
"Ibandronate",
"Risedronate"),
`Non_nitrogenous` = "Clodronate")) %>%
mutate(drugnitro = as.character(drugnitro)) %>%
mutate(drugbp = fct_collapse(drugnitro,
`Bisphosphonate` = c("Nitrogenous",
"Non_nitrogenous"))) %>%
mutate(drugbp = as.character(drugbp)) %>%
mutate(drugdose0 = ifelse(drugname0=="Control", drugname0, drugdose)) %>%
mutate(drugdm0 = ifelse(drugname0=="Control", drugname0, drugdm)) %>%
mutate(drugnitro0 = ifelse(drugname0=="Control", drugname0, drugnitro)) %>%
mutate(drugbp0 = ifelse(drugname0=="Control", drugname0, drugbp)) %>%
select(-addtrt) %>%
distinct
drugclasses <- c("treatment", "description", "descriptionshort", "dose", "delivery",
"drugdose","drugdm", "drugname", "drugnitro", "drugbp",
"drugdose0","drugdm0", "drugname0", "drugnitro0", "drugbp0", "drugnameshort")
bpcoding <- bpcoding %>%
select(drugclasses)
use_data(bpcoding, overwrite=TRUE)
addtrtcoding <- bpae %>%
select(addtrt) %>% distinct %>%
mutate(addtrtclass = fct_collapse(addtrt, ## TODO deduce from numbers
"chemoonly" = c("211", "311", "411", "711"),
"hormoneaionly" = c("131", "151", "161"),
"hormoneonly" = c("121", "141"),
"notrt" = c("111"),
"mixed" = c("120",
"212", "221", "261", "200", "222",
"300", "400", "522",
"600", "622", "262", "60", "70")
))
use_data(addtrtcoding, overwrite=TRUE)
bpae <- bpae %>%
mutate(treatment =
ifelse((addtrt == "141" & (treatment %in% c("103","220"))),
paste(treatment, "notai", sep=""), treatment))
aestr <- paste(aecategs, collapse="|")
bpae <- bpae %>%
left_join(bpcoding) %>%
left_join(addtrtcoding) %>%
select(study, studyid, reporting, meno, bonemeta, armno, drugclasses, N, matches(!!aestr), addtrt, addtrtclass)
bpae <- bpae %>%
mutate_at(vars(matches(!!aestr)), list(p = ~ . / N))
## Define the control arm for each study, for studies that have an observation/placebo control
bpaecon <- bpae %>%
mutate(armno2 = paste0("A", armno)) %>%
select(studyid, armno2, drugdose0) %>%
spread(armno2, drugdose0) %>%
select(studyid, A1:A5) %>%
mutate(control_arm = ifelse(is.na(A2), ifelse(is.na(A1), NA, 1), 2)) %>%
mutate(control_arm = ifelse(studyid==45, 1, control_arm)) %>% ## HOBOE has two control arms - select letrozole arm
select(studyid, control_arm)
bpae <- bpae %>% left_join(bpaecon, by="studyid")
## Get the event rate (and denominator) in the control arm for each study
control_outcomes <- bpae %>%
filter(armno == control_arm) %>%
select(studyid,
`N`,
`treatment`,
which(names(.) %in% aecategs),
which(names(.) %in% paste0(aecategs,"_p"))
) %>%
rename(ncon = "N",
trtcon = "treatment")
colnames(control_outcomes) <- gsub("_p", "_pcon", colnames(control_outcomes))
colnames(control_outcomes)[colnames(control_outcomes) %in% aecategs] <-
paste0(colnames(control_outcomes)[colnames(control_outcomes) %in% aecategs], "_rcon")
bpae <- bpae %>%
left_join(control_outcomes, "studyid")
use_data(bpae, overwrite=TRUE)
## Different classes of AEs based on clinical advice
## cat 1: finer category, cat2: broader category,
## cat 3: our choice
library(readxl)
library(dplyr)
setwd("../data-raw")
aetypes <- read_excel("AE TYPES 12 MAY 6pm.xlsx")[,c(1,3,4)]
names(aetypes) <- c("name", "cat1", "cat2")
aetypes$cat3 <- ifelse(aetypes$name %in% names(models_all), aetypes$name, aetypes$cat1)
## Significant events merged by hand include
aetypes$cat3[aetypes$name == "ARTHRALGIA/JOINT PAIN"] <- "ARTHRALGIA / JOINT PAIN"
aetypes$cat3[aetypes$name == "increased bone pain"] <- "PAIN IN EXTREMITY"
aetypes$cat3[aetypes$name == "stiffness"] <- "MYALGIA"
aetypes$cat3[aetypes$name == "NAUSEA"] <- "NAUSEA / VOMITING"
aetypes$cat3[aetypes$name == "CARDIAC EVENTS"] <- "CARDIOVASCULAR"
## For the moment:
## Keep back pain, seperate from pain in extremity
## diarrhoea separate from gastrointestinal
## chills, general flulike also separate from fever
## Though I'd merge all. Avoids having to describe and justify selection procedure
use_data(aetypes, overwrite=TRUE)
## Ultra tidy data with one row per arm and AE type. Use for ggplot2
props <- bpae %>%
gather(aetype, prop, which(names(.) %in% paste0(aecategs,"_p")))
pcon <- bpae %>%
gather(aetype, pcon, which(names(.) %in% paste0(aecategs,"_pcon")))
rcon <- bpae %>%
gather(aetype, rcon, which(names(.) %in% paste0(aecategs,"_rcon")))
bpaearmtype <- bpae %>%
gather(aetype, count, which(names(.) %in% aecategs)) %>%
mutate(prop = props$prop) %>%
mutate(pcon = pcon$pcon) %>%
mutate(rcon = rcon$rcon) %>%
mutate(aecateg = aetypes$cat3[match(aetype, aetypes$name)]) %>%
select(study, studyid, reporting, meno, bonemeta, armno,
drugclasses,
addtrt, addtrtclass,
N, aetype, aecateg, count, prop, pcon, rcon, ncon, trtcon) %>%
mutate(aetype = gsub("_p$","", aetype)) %>%
filter(!is.na(prop)) %>%
droplevels()
use_data(bpaearmtype, overwrite=TRUE)
## Data with one row per active treatment arm and AE type, with risk diff vs control
## TODO indicate whether control is obs only or placebo
bpaeriskdiff <- bpaearmtype %>%
filter(!is.na(pcon)) %>%
filter(armno >= 3) %>%
mutate(riskdiff = prop - pcon) %>%
mutate(rr = prop / pcon) %>%
mutate(or = plogis(prop) / plogis(pcon)) %>%
droplevels()
bpaeriskdiff[1:30,]
use_data(bpaeriskdiff, overwrite=TRUE)
## Plot colours for drug types
active_drugs <- unique(bpaearmtype$drugname)[!unique(bpaearmtype$drugname) %in%
c("Placebo","Observation")]
library(RColorBrewer)
drugcols <- brewer.pal(2 + length(active_drugs), "Paired")
drugcols <- c(drugcols[1], drugcols)
drugnames <- c("Placebo","Observation", active_drugs)
names(drugcols) <- drugnames
library(ggplot2)
drugScale <- scale_color_manual(breaks = drugnames, values = drugcols)
use_data(drugcols, overwrite=TRUE)
use_data(drugScale, overwrite=TRUE)
## Datasets for network meta-analysis
nmadata <- bpaearmtype %>%
rename("study"=study, "responders"="count", "sampleSize"="N") %>%
# filter(!is.na(aetype)) %>%
select(-pcon, -ncon, -rcon, -trtcon) %>%
droplevels() %>%
as.data.frame
# Remove studies for events which only have one arm reporting outcomes
narmdata <- nmadata %>%
group_by(aetype, study) %>%
summarise(narm = n()) %>%
ungroup()
nmadata <- nmadata %>%
left_join(narmdata, by=c("aetype","study")) %>%
filter(narm > 1)
setdiff(bpaearmtype$study, nmadata$study)
use_data(nmadata, overwrite=TRUE)
## Study-specific data.
## conte 1996 600 vs 622 ?
## These are both chemotherapy. Why coded 6?
## don't know whether control arm had additional hormone + trastuzumab. Unlikely if trial was meant to evaluate eff of pamidronate?
## Do we still need hormoneai stuff here now it's been included in bpcoding?
## Does this still need to be by aetype, or just by study?
studies <- nmadata %>%
select(aetype, aecateg, study, reporting, meno, bonemeta, addtrt, addtrtclass) %>%
unique %>%
filter(!(study=="HOBOE" & addtrtclass=="hormoneaionly")) %>%
filter(!(study=="ABCSG12" & addtrtclass=="hormoneaionly")) %>%
filter(!(study=="Conte 1996" & addtrt=="622")) %>%
mutate(chemo = as.numeric(addtrtclass %in% c("chemoonly", "mixed"))) %>%
mutate(addtrtclass = fct_collapse(addtrtclass,
"hormoneonly" = c("hormoneaionly","hormoneonly"))) %>%
rename(t = addtrtclass) %>%
cbind(model.matrix(~t-1, data=.))
use_data(studies, overwrite=TRUE)
treatments <- bpcoding %>%
mutate(id = treatment) %>%
as.data.frame
use_data(treatments, overwrite=TRUE)
chjackson/adverse documentation built on June 16, 2022, 4:53 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
library(readxl)
library(tidyverse)
filter <- dplyr::filter
matches <- dplyr::matches
library(devtools)
load_all("../../gemtc/gemtc")
load_all("..")
# for(i in list.files("../data", full.names=TRUE, include.dirs=TRUE)) load(i)
bpaeraw <- read_excel("Bisphosphonates DOUBLE CHECKED MASTER 12-7-19.xlsx", range="A2:CWJ62")
header <- unlist(read_excel("Bisphosphonates DOUBLE CHECKED MASTER 12-7-19.xlsx", col_names=FALSE, range="A1:CWJ1"))
##' TIDY COLUMN NAMES
aenames <- as.character(na.omit(header[27:length(header)]))
## Fill in gaps in first row of column names by carrying forward
header <- data.frame(header=header, stringsAsFactors=FALSE) %>%
fill(header) %>% unlist
names(bpaeraw) <- gsub("^C1(.+)", "A1\\1", names(bpaeraw))
names(bpaeraw) <- gsub("^C2(.+)", "A2\\1", names(bpaeraw))
names(bpaeraw) <- gsub("^E1(.+)", "A3\\1", names(bpaeraw))
names(bpaeraw) <- gsub("^E2(.+)", "A4\\1", names(bpaeraw))
names(bpaeraw) <- gsub("^E3(.+)", "A5\\1", names(bpaeraw))
## Prepend the adverse event names to the arm numbers
fullnames <- paste(header, names(bpaeraw), sep=";")
bpaeraw <- rename(bpaeraw, study=`Trial name`, studyid=`Trial Code`)
fullnames <- gsub("__[0-9]+$", "", fullnames)
fullnames <- gsub("C1$", "A1", fullnames)
fullnames <- gsub("C2$", "A2", fullnames)
fullnames <- gsub("E1$", "A3", fullnames)
fullnames <- gsub("E2$", "A4", fullnames)
fullnames <- gsub("E3$", "A5", fullnames)
aeinds <- 26:ncol(bpaeraw)
names(bpaeraw)[aeinds] <- fullnames[aeinds]
## denominator put into control instead of experimental here
bpaeraw[bpaeraw$study=="REFORM","N Exp2"] <- bpaeraw[bpaeraw$study=="REFORM","N Control"]
bpaeraw[bpaeraw$study=="REFORM","N Control"] <- NA
bpaeraw[bpaeraw$study=="Pivot 2011","N Exp2"] <- bpaeraw[bpaeraw$study=="Pivot 2011","N Control"]
bpaeraw[bpaeraw$study=="Pivot 2011","N Control"] <- NA
## Remove duplicated study "NCT00172068"
bpaeraw <- bpaeraw[bpaeraw$studyid != 49,]
## This one has 56 vs 45 people in two experimental arms, rather than 127 in single experimental arm? Zoledronic acid in community vs hospital setting
#bpaeraw %>% select(`Trial name`)
bpaeraw <- bpaeraw %>% filter(study != "Wardley2005")
# Single arm study shouldn't be problem for the NMA. Include this for now
bpaeraw <- bpaeraw %>%
rename(reporting="Adverse reporting",
N1="N Control",
N2="N Control2",
N3="N Exp1",
N4="N Exp2",
N5="N Exp3") %>%
mutate(has_control = as.numeric(!is.na(`A1 coding`) | !is.na(`A2 coding`))) %>%
mutate(reporting = recode(reporting,
`1` = "Complete", `2` = "Table of highest", `3` = "Less")) %>%
mutate(reporting = factor(reporting, levels = c("Complete","Table of highest","Less","x"))) %>%
rename(meno = "Pre/post menopausal",
bonemeta = "bone metasteses") %>%
mutate(meno = recode(meno, `1` = "pre", `2` = "post", `3` = "mixed"),
bonemeta = recode(bonemeta, `1` = "all", `2` = "none", `3` = "mixed"),
meno = factor(meno, levels = c("mixed", "pre", "post")),
bonemeta = factor(bonemeta, levels = c("none", "mixed", "all"))
)
## Remove character notes in numeric cells
## AZURE. Spreadsheet reports "Myalgia" but AZURE reports "Myalgia/arthralgia"
## NCT00376740 classifies myalgia under arthralgia/joint pain
## Arbirtrarily assign these to myalgia, arthralgia/joint pain respectively
bpaeraw[39,"ARTHRALGIA/JOINT PAIN (ungraded or grade 1/2);A4...161"] <- NA
bpaeraw[,"ARTHRALGIA/JOINT PAIN (ungraded or grade 1/2);A4...161"] <-
as.numeric(bpaeraw$"ARTHRALGIA/JOINT PAIN (ungraded or grade 1/2);A4...161")
mcol <- grep("MYALGIA ungraded;A4",colnames(bpaeraw))
bpaeraw[26,mcol] <- NA
bpaeraw$"MYALGIA ungraded;A4...1733" <-
as.numeric(bpaeraw$"MYALGIA ungraded;A4...1733")
# Remove extraneous comma
acol <- grep("ANY ADVERSE EVENT;A1",colnames(bpaeraw))
bpaeraw[13,acol] <- gsub(",","",bpaeraw[13,acol])
bpaeraw$"ANY ADVERSE EVENT;A1...30" <- as.numeric(bpaeraw$"ANY ADVERSE EVENT;A1...30")
## Recode 2x2arm studies as two trials
bpaeraw$study[bpaeraw$studyid=="8.1"] <- "ABCSG12_B"
bpaeraw$study[bpaeraw$studyid=="30.1"] <- "Kristensen 2008_B"
use_data(bpaeraw, overwrite=TRUE)
## Note VonMoos, Pivot have same treatments in E1 and E2.
## Categories of AEs to be summed over, aggregating different grades
aecat <- scan("aecategs.txt", what="character", sep="\n") # one element per col of bpaeraw
aecategs <- unique(aecat)[15:143]
use_data(aecategs, overwrite=TRUE)
bpaeraw[1:2,1:20]
##' Reshape data into one row per arm and event
bpaelong <- bpaeraw %>%
gather(vname, x,
matches("^N[[:digit:]]"),
matches("A[[:digit:]]"),
matches(";A[[:digit:]]")
) %>%
select(study, studyid, reporting, meno, bonemeta, has_control, vname, x) %>%
extract(vname, "armno1", "^(?:A|N)([[:digit:]]).*", remove=FALSE) %>%
extract(vname, "armno2", ";A([[:digit:]]).*$", remove=FALSE) %>%
mutate(armno = ifelse(is.na(armno1), armno2, armno1)) %>%
select(-armno1, -armno2) %>%
mutate(armno = as.numeric(armno),
vname = gsub("N[[:digit:]]","N",vname),
vname = gsub("A[[:digit:]] (additional coding)", "\\1", vname),
vname = gsub("A[[:digit:]] (coding)", "\\1", vname),
vname = gsub(";A[[:digit:]]$","",vname)) %>%
mutate(aecat = aecat[match(vname, gsub(";A[0-9]$","",names(bpaeraw)))],
aecat = ifelse(is.na(aecat), vname, aecat)) %>%
mutate(vname = gsub("ungraded or grade 1/2", "grade 1/2", vname, ignore.case=TRUE))
bpaelong <- bpaelong %>%
mutate(grade = gsub(".+GRADE ([1-4]).*", "\\1", vname, ignore.case=TRUE)) %>%
mutate(grade = ifelse(grepl("ungraded", grade, ignore.case=TRUE), NA, grade)) %>%
mutate(grade = ifelse(vname %in% aecategs[-(1:3)], NA, grade)) %>%
mutate(grade = as.numeric(ifelse(grade %in% c(1,2,3,4), grade, NA)))
use_data(bpaelong, overwrite=TRUE)
## Sum AEs of same type within each arm
bpaelongsum <- bpaelong %>%
filter(!is.na(x)) %>%
group_by(study, studyid, reporting, meno, bonemeta, armno, aecat) %>%
summarise(
xser = sum(x[!is.na(grade) & grade>=3]),
xgraded = sum(x[!is.na(grade) & grade>=1]),
x = sum(x)
) %>%
rename(vname=aecat) %>%
ungroup()
use_data(bpaelongsum, overwrite=TRUE)
## Data with one row per arm, and cols for each aggregate AE type
bpae <- bpaelongsum %>%
select(-xser, -xgraded) %>%
spread(vname, x) %>%
filter(!is.na(coding), # remove empty arms
!is.na(N)) %>%
mutate(treatment = as.character(coding)) %>%
mutate(addtrt = as.character(`additional coding`)) %>%
replace_na(list(addtrt = "unknown"))
## Treatment description database
## Variables ending in 0 merge placebo and observation-only
bpcoding <- bpae %>%
select(treatment, addtrt) %>%
distinct %>%
extract(treatment,
c("drug", "dose", "delivery"),
"^([0-9])([0-9])([0-9])$", remove=FALSE) %>%
mutate(drugnameshort = drug,
drugname0 = recode(drug,
`1` = "Control",
`7` = "Denosumab",
`2` = "Zoledronic_acid",
`3` = "Pamidronate",
`4` = "Ibandronate",
`5` = "Risedronate",
`6` = "Clodronate"),
drugnameshort = recode(drugnameshort,
`1` = "Con",
`7` = "Den",
`2` = "Zol",
`3` = "Pam",
`4` = "Iba",
`5` = "Ris",
`6` = "Clo"),
drugname = ifelse(treatment %in% c("100","101"), "Placebo",
ifelse(treatment == "103", "Observation", drugname0)),
drugnameshort = ifelse(treatment %in% c("100","101"), "Pla",
ifelse(treatment == "103", "Obs", drugnameshort)),
delivery = recode(delivery,
`0` = "IV",
`1` = "Oral",
`2` = "IVAndOral",
`3` = "Observation")) %>%
## Two studies included arms only distinguished by using a non-AI hormone (141)
mutate(notai = ifelse((treatment == 103 & addtrt == 141)|
(treatment == 220 & addtrt == 141) , 1, 0)) %>%
## Consider BP + non-AI in same way as a different dose of the BP
mutate(dose = ifelse(notai, paste0(dose, "notai"), dose)) %>%
mutate(treatment = ifelse(notai, paste0(treatment, "notai"), treatment)) %>%
mutate(description = paste(drugname, dose, delivery, sep="_")) %>%
mutate(descriptionshort = paste(drugnameshort, dose, delivery, sep="_")) %>%
## don't include delivery method for placebo in "drugdose" and "drugdm"
mutate(drugdose = ifelse(drugname0 =="Control", drugname, description)) %>%
mutate(drugdm = paste(drugname, delivery, sep="_")) %>%
mutate(drugdm = ifelse(drugname=="Placebo", drugname, drugdm)) %>%
mutate(
drugdm = ifelse(drugdm=="Observation_Observation", "Observation", drugdm),
drugdm = ifelse(drugdm=="Observation_0notai_Observation", "Observation", drugdm),
description = ifelse(description=="Observation_0_Observation", "Observation", description),
description = ifelse(description=="Observation_0notai_Observation", "Observation_notai", description),
descriptionshort = ifelse(descriptionshort=="Obs_0_Observation", "Obs", descriptionshort),
descriptionshort = ifelse(descriptionshort=="Obs_0notai_Observation", "Obs_notai", descriptionshort)
) %>%
mutate(drugdm0 = ifelse(drugname0=="Control", drugname0, drugdm)) %>%
mutate(drugnitro = fct_collapse(drugname,
`Nitrogenous` = c("Zoledronic_acid",
"Pamidronate",
"Ibandronate",
"Risedronate"),
`Non_nitrogenous` = "Clodronate")) %>%
mutate(drugnitro = as.character(drugnitro)) %>%
mutate(drugbp = fct_collapse(drugnitro,
`Bisphosphonate` = c("Nitrogenous",
"Non_nitrogenous"))) %>%
mutate(drugbp = as.character(drugbp)) %>%
mutate(drugdose0 = ifelse(drugname0=="Control", drugname0, drugdose)) %>%
mutate(drugdm0 = ifelse(drugname0=="Control", drugname0, drugdm)) %>%
mutate(drugnitro0 = ifelse(drugname0=="Control", drugname0, drugnitro)) %>%
mutate(drugbp0 = ifelse(drugname0=="Control", drugname0, drugbp)) %>%
select(-addtrt) %>%
distinct
drugclasses <- c("treatment", "description", "descriptionshort", "dose", "delivery",
"drugdose","drugdm", "drugname", "drugnitro", "drugbp",
"drugdose0","drugdm0", "drugname0", "drugnitro0", "drugbp0", "drugnameshort")
bpcoding <- bpcoding %>%
select(drugclasses)
use_data(bpcoding, overwrite=TRUE)
addtrtcoding <- bpae %>%
select(addtrt) %>% distinct %>%
mutate(addtrtclass = fct_collapse(addtrt, ## TODO deduce from numbers
"chemoonly" = c("211", "311", "411", "711"),
"hormoneaionly" = c("131", "151", "161"),
"hormoneonly" = c("121", "141"),
"notrt" = c("111"),
"mixed" = c("120",
"212", "221", "261", "200", "222",
"300", "400", "522",
"600", "622", "262", "60", "70")
))
use_data(addtrtcoding, overwrite=TRUE)
bpae <- bpae %>%
mutate(treatment =
ifelse((addtrt == "141" & (treatment %in% c("103","220"))),
paste(treatment, "notai", sep=""), treatment))
aestr <- paste(aecategs, collapse="|")
bpae <- bpae %>%
left_join(bpcoding) %>%
left_join(addtrtcoding) %>%
select(study, studyid, reporting, meno, bonemeta, armno, drugclasses, N, matches(!!aestr), addtrt, addtrtclass)
bpae <- bpae %>%
mutate_at(vars(matches(!!aestr)), list(p = ~ . / N))
## Define the control arm for each study, for studies that have an observation/placebo control
bpaecon <- bpae %>%
mutate(armno2 = paste0("A", armno)) %>%
select(studyid, armno2, drugdose0) %>%
spread(armno2, drugdose0) %>%
select(studyid, A1:A5) %>%
mutate(control_arm = ifelse(is.na(A2), ifelse(is.na(A1), NA, 1), 2)) %>%
mutate(control_arm = ifelse(studyid==45, 1, control_arm)) %>% ## HOBOE has two control arms - select letrozole arm
select(studyid, control_arm)
bpae <- bpae %>% left_join(bpaecon, by="studyid")
## Get the event rate (and denominator) in the control arm for each study
control_outcomes <- bpae %>%
filter(armno == control_arm) %>%
select(studyid,
`N`,
`treatment`,
which(names(.) %in% aecategs),
which(names(.) %in% paste0(aecategs,"_p"))
) %>%
rename(ncon = "N",
trtcon = "treatment")
colnames(control_outcomes) <- gsub("_p", "_pcon", colnames(control_outcomes))
colnames(control_outcomes)[colnames(control_outcomes) %in% aecategs] <-
paste0(colnames(control_outcomes)[colnames(control_outcomes) %in% aecategs], "_rcon")
bpae <- bpae %>%
left_join(control_outcomes, "studyid")
use_data(bpae, overwrite=TRUE)
## Different classes of AEs based on clinical advice
## cat 1: finer category, cat2: broader category,
## cat 3: our choice
library(readxl)
library(dplyr)
setwd("../data-raw")
aetypes <- read_excel("AE TYPES 12 MAY 6pm.xlsx")[,c(1,3,4)]
names(aetypes) <- c("name", "cat1", "cat2")
aetypes$cat3 <- ifelse(aetypes$name %in% names(models_all), aetypes$name, aetypes$cat1)
## Significant events merged by hand include
aetypes$cat3[aetypes$name == "ARTHRALGIA/JOINT PAIN"] <- "ARTHRALGIA / JOINT PAIN"
aetypes$cat3[aetypes$name == "increased bone pain"] <- "PAIN IN EXTREMITY"
aetypes$cat3[aetypes$name == "stiffness"] <- "MYALGIA"
aetypes$cat3[aetypes$name == "NAUSEA"] <- "NAUSEA / VOMITING"
aetypes$cat3[aetypes$name == "CARDIAC EVENTS"] <- "CARDIOVASCULAR"
## For the moment:
## Keep back pain, seperate from pain in extremity
## diarrhoea separate from gastrointestinal
## chills, general flulike also separate from fever
## Though I'd merge all. Avoids having to describe and justify selection procedure
use_data(aetypes, overwrite=TRUE)
## Ultra tidy data with one row per arm and AE type. Use for ggplot2
props <- bpae %>%
gather(aetype, prop, which(names(.) %in% paste0(aecategs,"_p")))
pcon <- bpae %>%
gather(aetype, pcon, which(names(.) %in% paste0(aecategs,"_pcon")))
rcon <- bpae %>%
gather(aetype, rcon, which(names(.) %in% paste0(aecategs,"_rcon")))
bpaearmtype <- bpae %>%
gather(aetype, count, which(names(.) %in% aecategs)) %>%
mutate(prop = props$prop) %>%
mutate(pcon = pcon$pcon) %>%
mutate(rcon = rcon$rcon) %>%
mutate(aecateg = aetypes$cat3[match(aetype, aetypes$name)]) %>%
select(study, studyid, reporting, meno, bonemeta, armno,
drugclasses,
addtrt, addtrtclass,
N, aetype, aecateg, count, prop, pcon, rcon, ncon, trtcon) %>%
mutate(aetype = gsub("_p$","", aetype)) %>%
filter(!is.na(prop)) %>%
droplevels()
use_data(bpaearmtype, overwrite=TRUE)
## Data with one row per active treatment arm and AE type, with risk diff vs control
## TODO indicate whether control is obs only or placebo
bpaeriskdiff <- bpaearmtype %>%
filter(!is.na(pcon)) %>%
filter(armno >= 3) %>%
mutate(riskdiff = prop - pcon) %>%
mutate(rr = prop / pcon) %>%
mutate(or = plogis(prop) / plogis(pcon)) %>%
droplevels()
bpaeriskdiff[1:30,]
use_data(bpaeriskdiff, overwrite=TRUE)
## Plot colours for drug types
active_drugs <- unique(bpaearmtype$drugname)[!unique(bpaearmtype$drugname) %in%
c("Placebo","Observation")]
library(RColorBrewer)
drugcols <- brewer.pal(2 + length(active_drugs), "Paired")
drugcols <- c(drugcols[1], drugcols)
drugnames <- c("Placebo","Observation", active_drugs)
names(drugcols) <- drugnames
library(ggplot2)
drugScale <- scale_color_manual(breaks = drugnames, values = drugcols)
use_data(drugcols, overwrite=TRUE)
use_data(drugScale, overwrite=TRUE)
## Datasets for network meta-analysis
nmadata <- bpaearmtype %>%
rename("study"=study, "responders"="count", "sampleSize"="N") %>%
# filter(!is.na(aetype)) %>%
select(-pcon, -ncon, -rcon, -trtcon) %>%
droplevels() %>%
as.data.frame
# Remove studies for events which only have one arm reporting outcomes
narmdata <- nmadata %>%
group_by(aetype, study) %>%
summarise(narm = n()) %>%
ungroup()
nmadata <- nmadata %>%
left_join(narmdata, by=c("aetype","study")) %>%
filter(narm > 1)
setdiff(bpaearmtype$study, nmadata$study)
use_data(nmadata, overwrite=TRUE)
## Study-specific data.
## conte 1996 600 vs 622 ?
## These are both chemotherapy. Why coded 6?
## don't know whether control arm had additional hormone + trastuzumab. Unlikely if trial was meant to evaluate eff of pamidronate?
## Do we still need hormoneai stuff here now it's been included in bpcoding?
## Does this still need to be by aetype, or just by study?
studies <- nmadata %>%
select(aetype, aecateg, study, reporting, meno, bonemeta, addtrt, addtrtclass) %>%
unique %>%
filter(!(study=="HOBOE" & addtrtclass=="hormoneaionly")) %>%
filter(!(study=="ABCSG12" & addtrtclass=="hormoneaionly")) %>%
filter(!(study=="Conte 1996" & addtrt=="622")) %>%
mutate(chemo = as.numeric(addtrtclass %in% c("chemoonly", "mixed"))) %>%
mutate(addtrtclass = fct_collapse(addtrtclass,
"hormoneonly" = c("hormoneaionly","hormoneonly"))) %>%
rename(t = addtrtclass) %>%
cbind(model.matrix(~t-1, data=.))
use_data(studies, overwrite=TRUE)
treatments <- bpcoding %>%
mutate(id = treatment) %>%
as.data.frame
use_data(treatments, overwrite=TRUE)
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.