R/format_cancer.R
In mightyphil2000/UkbCancerMortality: Format cancer data from UK Biobank for survival analyses
Defines functions
qc_check
cancer_age
cancer_behaviour
cancer_histology
find_max_date_diagnosis
length_followup2
length_followup
format_date_of_death
date_first_attended_assessment_centre
date_diagnosis2
number_deaths_cancer
format_cancer
Documented in
format_cancer
#' Format cancer mortality data
#'
#' Format cancer mortality dataset so that it can be used in survival analyses. The
#' function estimates the number of deaths due to the target cancer of interest and other causes, date of diagnosis, date of death, difference between latest and earlies dates of diagnosis for individuals with more than one diagnosis date, length of followup and survival time. There are two instances of primary cause of death in UK Biobank (f.40001.0.0 and f.40001.1.0). If the second instance is different to the first instance (ie an individual has more than one primary cause of death), the function checks to see if either of these instances match the indicated cancer. If there is a match, then the primary cause of death is set to that cancer. E.g. a case has prostate cancer (C61) and Pulmonary edema (J81) listed as primary causes of death. In this case, we would set the primary cause of death to prostate cancer. This is a rare occurence. I've only observed this example once for prostate cancer and 0 times for lung cancer, breast cancer and overall cancer. If a case has two primary causes of death and neither one matches the indicated cancer, then the function stops and returns a warning message.
#'
#' @param dat the cancer dataset
#' @param censor_date the date up until which the death registry data in UK Biobank is assumed to be complete. We assume that everyone not in the death registry at this date was still alive on this date. The default assumes a censoring date of "2018-02-06", based on the maximum value for the date of diagnosis f.40005.0.0 field in dataset 37205 on the rdsf /projects/MRC-IEU/research/data/ukbiobank/phenotypic/applications/15825/released/2019-05-02/data/derived/formats/r
#' @param cancer_name the name of the field/column in dat corresponding to case-control status for the cancer of interest. The field must take on values of 1 (control) or 2 (cases).
#' @param icd10 the icd10 code for the cancer of interest. Icd codes are hierarchically structured. Therefore, it is not necessary to include all the lowest level icd codes for the cancer of interest. Only the highest level code should be specified. For example, breast cancer corresponds to icd10 codes C50.0-C50.9. Therefore, to create a breast cancer dataset, you would only need to set icd10 to "C50" (ie icd10="C50"). It is only necessary to include multiple icd codes when the cancer is defined by more than one high level code. E.g. colorectal cancer is defined as icd10 codes C18, C19 and C20, in which case you would specify icd10=c("C18","C19","C20").
#'
#' @return data frame
#' @export
format_cancer<-function(dat=NULL,censor_date="2018-02-06",cancer_name=NULL,icd10=NULL){
# censor_date in yyyy-mm-dd
# Secondary cause of death f.40002
# Primary cause of death f.40001
# f.40000 date of death
dat<-dat[which(dat[,cancer_name] == 2),]
dat<-date_diagnosis2(dat=dat) #looks for earliest diagnosis date
dat<-cancer_histology(dat=dat) # histology at earliest diagnsosis
dat<-cancer_behaviour(dat=dat) # identify behaviour for earliest diagnosis date
dat<-cancer_age(dat=dat) #age at earliest diagnosis
dat<-format_date_of_death(dat=dat)
dat<-length_followup2(dat=dat,censor_date=censor_date)
# dat2<-dat
dat<-number_deaths_cancer(dat=dat,icd10=icd10,cancer_name=cancer_name)
# dat2<-dat
dat<-date_first_attended_assessment_centre(dat=dat)
names(dat)[names(dat) == "f.200.0.0"]<-"date_consent"
# seems f.200.0.0 and earliest date in f.53 are identical, at least in the lung cancer case-only dataset. makes sense they would be consented on the first visit
return(dat)
#this next section needs to be updated because currently it finds max date for any cancer diagnosis not max date for diagnosis of target cancer
# dat<-find_max_date_diagnosis(dat=dat)
# Diff<-as.numeric(dat$max_date_diagnosis - dat$date_diagnosis2)
# dat$date_diagnosis_range_days<-Diff
# dat$date_diagnosis_range_months<-round(Diff/(365.25/12),2)
# dat$date_diagnosis_range_years<-round(Diff/(365.25),2)
}
number_deaths_cancer<-function(dat=NULL,icd10=icd10,cancer_name=cancer_name){
Primary<-names(dat)[grep("f.40001",names(dat))]
Secondary<-names(dat)[grep("f.40002",names(dat))]
# length(which(!is.na(dat$date_of_death_40000 )))
Pos0<-which(!is.na(dat$f.40001.0.0 ))
Pos1<-which(!is.na(dat$f.40001.1.0 ))
# dat$f.40001.0.0[Pos1]
# dat$f.40001.1.0[Pos1]
# dat$f.40002.0.0[Pos1]
# dat$f.40002.1.0[Pos1]
# dat$f.40002.2.0[Pos1]
if(!all(Pos1 %in% Pos0)) stop("second instance of f.40001 primary cause of death is not nested within the first instance")
cancer_name2<-gsub("_"," ",cancer_name)
cancer_name2<-gsub("overall","",cancer_name2)
# cancer_name2<-gsub("cancer","",cancer_name2)
cancer_name2<-trimws(cancer_name2)
# cancer_name2<-paste0("Malignant neoplasm of ",cancer_name2)
dat$primary<-rep(NA,nrow(dat))
if(!all(dat$f.40001.1.0[Pos1] == dat$f.40001.0.0[Pos1] ))
{
if(length(icd10)>1) stop("more than icd10 code provided. this part of the script assumes there is only one icd10 code")
# position where first instance does not equal second instance of f.40001
Pos1_2<-Pos1[which(dat$f.40001.1.0[Pos1] != dat$f.40001.0.0[Pos1] )]
Dat2<-dat[Pos1_2,]
Dat1<-dat[1:nrow(dat) != Pos1_2,]
Pos1_3<-which(Dat2$f.40001.1.0 ==icd10 | Dat2$f.40001.0.0 ==icd10)
# if the target cancer is pan-cancer we used use all cancer codes
if(icd10 == "C"){
Pos1_3<-which(substring(Dat2$f.40001.1.0,1,1) ==icd10 | substring(Dat2$f.40001.0.0,1,1) ==icd10)
}
if(length(Pos1_3) == 2) stop("this part of the code assumes there is only one match to icd10")
if(sum(Pos1_3)==0){
stop("second instance of f.40001 primary cause of death is different to the first instance and neither corresponds to the indicated cancer of interest. Pos1_3.")
}
# if either the first or second instance match the cancer of interest, set primary cause of death to the cancer of interest. This means that when a case has two different primary causes of death, one of which matches the cancer of interest, we set primary cause of death to the cancer of interest. E.g. one case had prostate cancer (C61) and Pulmonary edema (J81) listed as primary causes of death. We set the primary cause of death for this case to prostate cancer. This is a rare occurence. I've only observed this example once for prostate cancer and 0 times for lung cancer, breast cancer and overall cancer
Dat2$primary[Pos1_3]<-cancer_name2
if(icd10 != "C"){
if(Dat2$f.40001.1.0 !=icd10 & Dat2$f.40001.0.0 !=icd10) stop("second instance of f.40001 primary cause of death is different to the first instance and neither corresponds to the indicated cancer of interest")
}
dat<-rbind(Dat1,Dat2)
}
dat$primary[grep("C",dat$f.40001.0.0)]<-"other malignant neoplasm"
dat$primary[unlist(lapply(icd10,FUN=function(x) grep(x,dat$f.40001.0.0)))]<-cancer_name2
if(any(duplicated(unlist(lapply(icd10,FUN=function(x) grep(x,dat$f.40001.0.0)))))) stop("unexpected duplicates present")
Pos<-which(is.na(dat$primary) & !is.na(dat$f.40001.0.0))
Pos2<-grep("I",dat$f.40001.0.0[Pos])
dat$primary[Pos[Pos2]] <-"Diseases of the circulatory system"
Pos<-which(is.na(dat$primary) & !is.na(dat$f.40001.0.0))
Pos2<-grep("J",dat$f.40001.0.0[Pos])
dat$primary[Pos[Pos2]] <-"Diseases of the respiratory system"
Pos<-which(is.na(dat$primary) & !is.na(dat$f.40001.0.0))
Pos2<-grep("K",dat$f.40001.0.0[Pos])
dat$primary[Pos[Pos2]] <-"Diseases of the digestive system"
Pos<-which(is.na(dat$primary) & !is.na(dat$f.40001.0.0))
Pos2<-grep("G",dat$f.40001.0.0[Pos])
dat$primary[Pos[Pos2]] <-"Diseases of the nervous system"
Pos<-which(is.na(dat$primary) & !is.na(dat$f.40001.0.0))
Pos2<-unique(unlist(lapply(c("D0","D1","D2","D3","D4"),FUN=function(x)
grep(x,dat$f.40001.0.0[Pos]))))
dat$primary[Pos[Pos2]] <-"In situ or benign neoplasms or neoplasms of unknown/uncertain behavior, polycythemia vera and myelodysplastic syndromes"
Pos<-which(is.na(dat$primary) & !is.na(dat$f.40001.0.0))
Pos2<-grep("F",dat$f.40001.0.0[Pos])
dat$primary[Pos[Pos2]] <-"Mental, Behavioral and Neurodevelopmental disorders"
Pos<-which(is.na(dat$primary) & !is.na(dat$f.40001.0.0))
Pos2<-unique(unlist(lapply(c("A","B"),FUN=function(x)
grep(x,dat$f.40001.0.0[Pos]))))
dat$primary[Pos[Pos2]] <-"Certain infectious and parasitic diseases"
Pos<-which(is.na(dat$primary) & !is.na(dat$f.40001.0.0))
dat$primary[Pos] <-"other diseases"
Names<-names(dat)[grep("f.40001",names(dat))]
Names2<-names(dat)[grep("f.40002",names(dat))]
Test<-dat[which( !is.na(dat$date_death) & is.na(dat$primary)),c("date_death","primary","date_diagnosis2",Names,Names2)]
Names<-c(Names,Names2)
if(!all(is.na(Test[,Names]))) stop("primary cause of death is NA but not all primary and secondary causes of death are NA")
dat$primary[which( !is.na(dat$date_death) & is.na(dat$primary))]<-"unknown cause of death"
dat$status <- NA
dat$status[!is.na(dat$primary)]<-2 #death from any cause
dat$status[is.na(dat$primary)]<-1 #censored
dat$status2 <- 1
dat$status2[which(dat$primary==cancer_name2)]<-2 #death from any cause
# dat$status2[which(dat$primary!=cancer_name2)]<-1 #censored
return(dat)
}
# this doesn't work because applies to any diagnosis in cancer registry
# date_diagnosis<-function(dat=NULL){
# # Pos<-which(!is.na(dat$f.40005.1.0))
# # length(Pos)
# Names<-names(dat)[grep("f.40005",names(dat))]
# N_diag<-lapply(1:nrow(dat),FUN=function(x)
# unique(as.numeric(dat[x,Names])))
# dat$N_diagnoses<-unlist(lapply(1:nrow(dat),FUN=function(x)
# length(unlist(N_diag[x])[!is.na(unlist(N_diag[x]))])))
# min_date_diagnosis<-unlist(lapply(1:nrow(dat),FUN=function(x)
# min(as.numeric(dat[x,Names]),na.rm=TRUE)))
# dat$date_diagnosis<-as.Date(min_date_diagnosis,origin="1970-01-01")
# return(dat)
# }
date_diagnosis2<-function(dat=NULL){
Pos1<-grep("date.diagnosis",names(dat))
Pos2<-grep("lung",names(dat))
Pos<-Pos1[Pos1 %in% Pos2]
Names<-names(dat)[Pos]
# N_diag<-lapply(1:nrow(dat),FUN=function(x)
# unique(as.numeric(dat[x,Names])))
# dat$N_diagnoses<-unlist(lapply(1:nrow(dat),FUN=function(x)
# length(unlist(N_diag[x])[!is.na(unlist(N_diag[x]))])))
# Are all ICD codes in ICD instances lung cancer codes? If all C34 (icd10) or 16 (icd19) then yes.
Names2<-c("lung.ICD9.1","lung.ICD9.2","lung.ICD10.1","lung.ICD10.2","lung.ICD10.3")
ICD<-unlist(lapply(1:nrow(dat),FUN=function(x)
unique(dat[x,Names2][!is.na(dat[x,Names2])])))
ICD[grep("C34",ICD,invert=TRUE)]
# Yes all instances correspond to lung cancer icd codes
min_date_diagnosis<-unlist(lapply(1:nrow(dat),FUN=function(x)
min(as.numeric(dat[x,Names]),na.rm=TRUE)))
dat$date_diagnosis2<-as.Date(min_date_diagnosis,origin="1970-01-01")
# dat[which(dat$N_diagnoses==2)[1:2],c(Names,Names2)]
return(dat)
}
date_first_attended_assessment_centre<-function(dat=NULL){
Pos<-grep("f.53",names(dat))
Names<-names(dat)[Pos]
Dates<-lapply(Names,FUN=function(x)
as.Date(dat[,x]))
dat[,Names] <- lapply(1:length(Dates),FUN=function(x)
as.Date(unlist(Dates[x]),origin="1970-01-01"))
min_date<-unlist(lapply(1:nrow(dat),FUN=function(x)
min(as.numeric(dat[x,Names]),na.rm=TRUE)))
dat$first_date_attend_centre<-as.Date(min_date,origin="1970-01-01")
# dat[which(dat$N_diagnoses==2)[1:2],c(Names,Names2)]
return(dat)
}
# bd[Pos[2],c("date_diagnosis",Names)]
format_date_of_death<-function(dat=NULL){
# Date of death f.40000 . two instances, the second is nested within, and is identical to, the first
Pos0<-which(!is.na(dat$f.40000.0.0))
names(bd)[grep("40000",names(bd))]
names(dat)[grep("f.4000",names(dat))]
Pos1<-which(!is.na(dat$f.40000.1.0))
if(!all(Pos1 %in% Pos0)) stop("second instance of date of death is not nested within the first instance")
if(!all(dat$f.40000.1.0[Pos1] == dat$f.40000.0.0[Pos1])) stop("second instance of date of death is different to the first instance")
dat$f.40000.0.0 <- as.Date(dat$f.40000.0.0)
dat$f.40000.1.0 <- as.Date(dat$f.40000.1.0)
names(dat)[names(dat)=="f.40000.0.0"]<-"date_death"
return(dat)
}
# dat=bd
# time from diagnosis to death or last known date assumed alive
# this function seems to contain an error, in that followup is measured as time to censor date defined as "2018-02-06".
length_followup<-function(dat=NULL,censor_date=NULL){
dat$survival_days<-as.numeric(dat$date_death-dat$date_diagnosis2)
dat$survival_months<-round(dat$survival_days/(365.25/12),2)
dat$survival_years<-round(dat$survival_days/365.25,2)
dat$follow_days<-as.numeric(as.Date(censor_date)-dat$date_diagnosis2)
dat$follow_months<-round(dat$follow_days/(365.25/12),2)
dat$follow_years<-round(dat$follow_days/365.25,2)
return(dat)
}
# corrected. Followup is time to death or censor date
length_followup2<-function(dat=NULL,censor_date=NULL){
dat$survival_days2<-as.numeric(dat$date_death-dat$date_diagnosis2)
dat$survival_months2<-round(dat$survival_days2/(365.25/12),2)
dat$survival_years2<-round(dat$survival_days2/365.25,2)
dat$follow_days2<-as.numeric(as.Date(censor_date)-dat$date_diagnosis2)
Pos<-which(!is.na(dat$survival_days2))
dat$follow_days2[Pos]<-dat$survival_days2[Pos]
dat$follow_months2<-round(dat$follow_days2/(365.25/12),2)
dat$follow_years2<-round(dat$follow_days2/365.25,2)
return(dat)
}
find_max_date_diagnosis<-function(dat=NULL){
Names<-names(dat)[grep("f.40005",names(dat))]
max_date_diagnosis<-unlist(lapply(1:nrow(dat),FUN=function(x)
max(as.numeric(dat[x,Names]),na.rm=TRUE)))
dat$max_date_diagnosis<-as.Date(max_date_diagnosis,origin="1970-01-01")
dat$max_date_diagnosis[which(as.numeric(dat$max_date_diagnosis) == -Inf)]<-NA
return(dat)
}
cancer_histology<-function(dat=NULL){
Pos1<-grep("histology",names(dat))
Pos2<-grep("lung",names(dat))
Pos<-Pos1[Pos1 %in% Pos2]
Names<-names(dat)[Pos]
# this lines are just useful for testing the function
N_diag<-lapply(1:nrow(dat),FUN=function(x)
unique(as.numeric(dat[x,Names])))
dat$N_diagnoses<-unlist(lapply(1:nrow(dat),FUN=function(x)
length(unlist(N_diag[x])[!is.na(unlist(N_diag[x]))])))
Histology<-lapply(1:nrow(dat),FUN=function(x)
dat[x,Names])
Pos1<-grep("date.diagnosis",names(dat))
Pos2<-grep("lung",names(dat))
Pos<-Pos1[Pos1 %in% Pos2]
Names2<-names(dat)[Pos]
min_date_diagnosis<-unlist(lapply(1:nrow(dat),FUN=function(x)
min(as.numeric(dat[x,Names2]),na.rm=TRUE)))
# which instance contains the earliest diagnosis.
Instance_min_date<-unlist(lapply(1:nrow(dat),FUN=function(x)
which(as.numeric(dat[x,Names2]) == min_date_diagnosis[x])))
Histology2<-as.numeric(unlist(lapply(1:nrow(dat),FUN=function(x)
unlist(Histology[x])[Instance_min_date[x]])))
dat$histology<-Histology2
His<-read.table("/data/ph14916/UkbCancerMortality/data/coding38.tsv",sep="\t",head=TRUE,stringsAsFactors=FALSE)
dat<-merge(dat,His[,c("coding","meaning")],by.x="histology",by.y="coding")
names(dat)[names(dat)=="histology"]<-"histology_code"
names(dat)[names(dat)=="meaning"]<-"histology"
return(dat)
}
cancer_behaviour<-function(dat=NULL){
Pos1<-grep("behaviour",names(dat))
Pos2<-grep("lung",names(dat))
Pos<-Pos1[Pos1 %in% Pos2]
Names<-names(dat)[Pos]
Behaviour<-lapply(1:nrow(dat),FUN=function(x)
dat[x,Names])
Pos1<-grep("date.diagnosis",names(dat))
Pos2<-grep("lung",names(dat))
Pos<-Pos1[Pos1 %in% Pos2]
Names2<-names(dat)[Pos]
min_date_diagnosis<-unlist(lapply(1:nrow(dat),FUN=function(x)
min(as.numeric(dat[x,Names2]),na.rm=TRUE)))
# which instance contains the earliest diagnosis.
Instance_min_date<-unlist(lapply(1:nrow(dat),FUN=function(x)
which(as.numeric(dat[x,Names2]) == min_date_diagnosis[x])))
Behaviour2<-unlist(lapply(1:nrow(dat),FUN=function(x)
unlist(Behaviour[x])[Instance_min_date[x]]))
dat$behaviour<-Behaviour2
return(dat)
}
cancer_age<-function(dat=NULL){
Pos1<-grep("age_diagnosis",names(dat))
Pos2<-grep("lung",names(dat))
Pos<-Pos1[Pos1 %in% Pos2]
Names<-names(dat)[Pos]
Age<-lapply(1:nrow(dat),FUN=function(x)
dat[x,Names])
Pos1<-grep("date.diagnosis",names(dat))
Pos2<-grep("lung",names(dat))
Pos<-Pos1[Pos1 %in% Pos2]
Names2<-names(dat)[Pos]
min_date_diagnosis<-unlist(lapply(1:nrow(dat),FUN=function(x)
min(as.numeric(dat[x,Names2]),na.rm=TRUE)))
# which instance contains the earliest diagnosis.
Instance_min_date<-unlist(lapply(1:nrow(dat),FUN=function(x)
which(as.numeric(dat[x,Names2]) == min_date_diagnosis[x])))
Age2<-unlist(lapply(1:nrow(dat),FUN=function(x)
unlist(Age[x])[Instance_min_date[x]]))
dat$age_diagnosis<-Age2
return(dat)
}
# cancer_histology<function(){
# # # histology
# table(df_split$lung.ICD9.histology.1)
# table(df_split$lung.ICD9.histology.2)
# table(df_split$lung.ICD10.histology.1)
# table(df_split$lung.ICD10.histology.2)
# table(df_split$lung.ICD10.histology.3)
# His<-read.table("/data/ph14916/UkbCancerMortality/data/coding38.tsv",sep="\t",head=TRUE,stringsAsFactors=FALSE)
# dim(His)
# # small cell lung carcinoma (SCLC, 15% of all lung cancers)
# # non-SCLC (NSCLC, 85% of all lung cancers).
# # NSCLCs are generally subcategorized into adenocarcinoma, squamous cell carcinoma (SqCC), and large cell carcinoma.
# His$meaning[grep("squamous",His$meaning)]
# head(His)
# His2<-His[,c("coding","meaning")]
# sort(His$meaning[grep("adenocarcinoma",His$meaning)])
# # Adenocarcinoma
# # df_split<-df_split[,grep("meaning",names(df_split),invert=TRUE)]
# # df_split$lung.ICD9.histology_meaning.1<-df_split$lung.ICD9.histology.1
# df_split.m<-merge(df_split,His2,by.x="lung.ICD9.histology.1",by.y="coding",all.x=TRUE)
# names(df_split.m)[names(df_split.m)=="meaning"]<-"lung.ICD9.histology_meaning.1"
# df_split.m<-merge(df_split.m,His2,by.x="lung.ICD9.histology.2",by.y="coding",all.x=TRUE)
# names(df_split.m)[names(df_split.m)=="meaning"]<-"lung.ICD9.histology_meaning.2"
# df_split.m<-merge(df_split.m,His2,by.x="lung.ICD10.histology.1",by.y="coding",all.x=TRUE)
# names(df_split.m)[names(df_split.m)=="meaning"]<-"lung.ICD10.histology_meaning.1"
# df_split.m<-merge(df_split.m,His2,by.x="lung.ICD10.histology.2",by.y="coding",all.x=TRUE)
# names(df_split.m)[names(df_split.m)=="meaning"]<-"lung.ICD10.histology_meaning.2"
# df_split.m<-merge(df_split.m,His2,by.x="lung.ICD10.histology.3",by.y="coding",all.x=TRUE)
# names(df_split.m)[names(df_split.m)=="meaning"]<-"lung.ICD10.histology_meaning.3"
# table(df_split.m$lung.ICD9.histology.1,df_split.m$lung.ICD9.histology_meaning.1)
# )
# table(df_split.m$lung.ICD10.histology_meaning.1,df_split.m$lung.ICD10.histology_meaning.2)
# Pos1<-which(df_split.m$lung.ICD10.histology_meaning.1=="Sarcoma, NOS")
# Pos2<-which(df_split.m$lung.ICD10.histology_meaning.2=="Small cell carcinoma, NOS")
# Pos<-Pos1[Pos1 %in% Pos2]
# name.pos<-grep("lung.ICD10",names(df_split.m))
# names(df_split.m)[name.pos]
# df_split.m[Pos1,c( "lung.ICD10.1",
# "lung.ICD10.2",
# "lung.ICD10.date.diagnosis.1",
# "lung.ICD10.date.diagnosis.2",
# "lung.ICD10.behaviour.1",
# "lung.ICD10.behaviour.2",
# "lung.ICD10.age_diagnosis.1",
# "lung.ICD10.age_diagnosis.2",
# "lung.ICD10.histology_meaning.1",
# "lung.ICD10.histology_meaning.2")]
# df_split.m$lung.ICD9.histology_meaning.2)
# df_split.m$lung.ICD9.histology_meaning.1)
# ),
# ,
# df_split.m$lung.ICD10.histology_meaning.3,
# collapse="; ")
# table(df_split.m$histology)
# # behaviour unknown or metstatic behavipour/site
# table(df_split.m$lung.ICD9.histology_meaning.1[df_split$lung.ICD9.behaviour.1=="NA"])
# Pos<-which(df_split$lung.ICD9.behaviour.1=="Malignant, primary site")
# Pos2<-which(df_split$lung.ICD9.behaviour.1=="NA")
# # histology missing
# df_split.m$lung.ICD9.histology_meaning.1[Pos]
# df_split.m$lung.ICD9.histology_meaning.1[Pos2]
# table(df_split.m$lung.ICD10.histology_meaning.1,)
# unique(df_split$lung.ICD10.behaviour.1)
# Pos<-which(df_split$lung.ICD10.behaviour.1=="Malignant, metastatic site")
# Pos2<-which(df_split$lung.ICD10.behaviour.1=="Uncertain whether benign or malignant")
# Pos3<-which(df_split$lung.ICD10.behaviour.1=="NA")
# df_split.m$lung.ICD10.histology_meaning.1[Pos]
# df_split.m$lung.ICD10.histology_meaning.1[Pos2]
# df_split.m$lung.ICD10.histology_meaning.1[Pos3]
# # missing histology
# head(df_split.m)
# head(His)
# df_split$lung.ICD9.histology_meaning.2<-df_split$lung.ICD9.histology.2
# df_split$lung.ICD10.histology_meaning.1<-df_split$lung.ICD10.histology.1
# df_split$lung.ICD10.histology_meaning.2<-df_split$lung.ICD10.histology.2
# df_split$lung.ICD10.histology_meaning.3<-df_split$lung.ICD10.histology.3
# names(df_split)
# print("generating histology flag")
# df_split$histology.flag <-
# ifelse(between(df_split$headneck.ICD9.histology.1, 8070, 8078) |
# between(df_split$headneck.ICD10.histology.1, 8070, 8078) |
# between(df_split$headneck.ICD10.histology.2, 8070, 8078) |
# between(df_split$headneck.ICD10.histology.3, 8070, 8078),
# 1,
# 0)
qc_check<-function(){
message("checking baseline date and definition of incident cases for possible issues")
# check date of consent same as first date attended assessment centre
if(!all(lun4$date_consent == lun4$first_date_attend_centre)) warning("consent date not always the same as first date of attendance at assessment centre")
# check that incident cases were diagnosed after UKB baseline date using alternative method for determining incident case
Pos1<-which(lun4$date_diagnosis2 > lun4$date_consent)
Pos2<-which(!is.na(lun4$incident_lung_cancer))
if(!all(Pos1==Pos2)) stop("possible conflict between date of diagnosis for incident cases and date of enrollment into UKB")
Pos3<-which(lun4$date_diagnosis2 <= lun4$date_consent)
Pos4<-which(is.na(lun4$incident_lung_cancer))
if(!all(Pos3==Pos4)) stop("possible conflict between date of diagnosis for incident cases and date of enrollment into UKB")
message("no issues found. nice")
}
mightyphil2000/UkbCancerMortality documentation built on Oct. 12, 2023, 4:40 a.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.
Defines functions qc_check cancer_age cancer_behaviour cancer_histology find_max_date_diagnosis length_followup2 length_followup format_date_of_death date_first_attended_assessment_centre date_diagnosis2 number_deaths_cancer format_cancer
Documented in format_cancer
#' Format cancer mortality data
#'
#' Format cancer mortality dataset so that it can be used in survival analyses. The
#' function estimates the number of deaths due to the target cancer of interest and other causes, date of diagnosis, date of death, difference between latest and earlies dates of diagnosis for individuals with more than one diagnosis date, length of followup and survival time. There are two instances of primary cause of death in UK Biobank (f.40001.0.0 and f.40001.1.0). If the second instance is different to the first instance (ie an individual has more than one primary cause of death), the function checks to see if either of these instances match the indicated cancer. If there is a match, then the primary cause of death is set to that cancer. E.g. a case has prostate cancer (C61) and Pulmonary edema (J81) listed as primary causes of death. In this case, we would set the primary cause of death to prostate cancer. This is a rare occurence. I've only observed this example once for prostate cancer and 0 times for lung cancer, breast cancer and overall cancer. If a case has two primary causes of death and neither one matches the indicated cancer, then the function stops and returns a warning message.
#'
#' @param dat the cancer dataset
#' @param censor_date the date up until which the death registry data in UK Biobank is assumed to be complete. We assume that everyone not in the death registry at this date was still alive on this date. The default assumes a censoring date of "2018-02-06", based on the maximum value for the date of diagnosis f.40005.0.0 field in dataset 37205 on the rdsf /projects/MRC-IEU/research/data/ukbiobank/phenotypic/applications/15825/released/2019-05-02/data/derived/formats/r
#' @param cancer_name the name of the field/column in dat corresponding to case-control status for the cancer of interest. The field must take on values of 1 (control) or 2 (cases).
#' @param icd10 the icd10 code for the cancer of interest. Icd codes are hierarchically structured. Therefore, it is not necessary to include all the lowest level icd codes for the cancer of interest. Only the highest level code should be specified. For example, breast cancer corresponds to icd10 codes C50.0-C50.9. Therefore, to create a breast cancer dataset, you would only need to set icd10 to "C50" (ie icd10="C50"). It is only necessary to include multiple icd codes when the cancer is defined by more than one high level code. E.g. colorectal cancer is defined as icd10 codes C18, C19 and C20, in which case you would specify icd10=c("C18","C19","C20").
#'
#' @return data frame
#' @export
format_cancer<-function(dat=NULL,censor_date="2018-02-06",cancer_name=NULL,icd10=NULL){
# censor_date in yyyy-mm-dd
# Secondary cause of death f.40002
# Primary cause of death f.40001
# f.40000 date of death
dat<-dat[which(dat[,cancer_name] == 2),]
dat<-date_diagnosis2(dat=dat) #looks for earliest diagnosis date
dat<-cancer_histology(dat=dat) # histology at earliest diagnsosis
dat<-cancer_behaviour(dat=dat) # identify behaviour for earliest diagnosis date
dat<-cancer_age(dat=dat) #age at earliest diagnosis
dat<-format_date_of_death(dat=dat)
dat<-length_followup2(dat=dat,censor_date=censor_date)
# dat2<-dat
dat<-number_deaths_cancer(dat=dat,icd10=icd10,cancer_name=cancer_name)
# dat2<-dat
dat<-date_first_attended_assessment_centre(dat=dat)
names(dat)[names(dat) == "f.200.0.0"]<-"date_consent"
# seems f.200.0.0 and earliest date in f.53 are identical, at least in the lung cancer case-only dataset. makes sense they would be consented on the first visit
return(dat)
#this next section needs to be updated because currently it finds max date for any cancer diagnosis not max date for diagnosis of target cancer
# dat<-find_max_date_diagnosis(dat=dat)
# Diff<-as.numeric(dat$max_date_diagnosis - dat$date_diagnosis2)
# dat$date_diagnosis_range_days<-Diff
# dat$date_diagnosis_range_months<-round(Diff/(365.25/12),2)
# dat$date_diagnosis_range_years<-round(Diff/(365.25),2)
}
number_deaths_cancer<-function(dat=NULL,icd10=icd10,cancer_name=cancer_name){
Primary<-names(dat)[grep("f.40001",names(dat))]
Secondary<-names(dat)[grep("f.40002",names(dat))]
# length(which(!is.na(dat$date_of_death_40000 )))
Pos0<-which(!is.na(dat$f.40001.0.0 ))
Pos1<-which(!is.na(dat$f.40001.1.0 ))
# dat$f.40001.0.0[Pos1]
# dat$f.40001.1.0[Pos1]
# dat$f.40002.0.0[Pos1]
# dat$f.40002.1.0[Pos1]
# dat$f.40002.2.0[Pos1]
if(!all(Pos1 %in% Pos0)) stop("second instance of f.40001 primary cause of death is not nested within the first instance")
cancer_name2<-gsub("_"," ",cancer_name)
cancer_name2<-gsub("overall","",cancer_name2)
# cancer_name2<-gsub("cancer","",cancer_name2)
cancer_name2<-trimws(cancer_name2)
# cancer_name2<-paste0("Malignant neoplasm of ",cancer_name2)
dat$primary<-rep(NA,nrow(dat))
if(!all(dat$f.40001.1.0[Pos1] == dat$f.40001.0.0[Pos1] ))
{
if(length(icd10)>1) stop("more than icd10 code provided. this part of the script assumes there is only one icd10 code")
# position where first instance does not equal second instance of f.40001
Pos1_2<-Pos1[which(dat$f.40001.1.0[Pos1] != dat$f.40001.0.0[Pos1] )]
Dat2<-dat[Pos1_2,]
Dat1<-dat[1:nrow(dat) != Pos1_2,]
Pos1_3<-which(Dat2$f.40001.1.0 ==icd10 | Dat2$f.40001.0.0 ==icd10)
# if the target cancer is pan-cancer we used use all cancer codes
if(icd10 == "C"){
Pos1_3<-which(substring(Dat2$f.40001.1.0,1,1) ==icd10 | substring(Dat2$f.40001.0.0,1,1) ==icd10)
}
if(length(Pos1_3) == 2) stop("this part of the code assumes there is only one match to icd10")
if(sum(Pos1_3)==0){
stop("second instance of f.40001 primary cause of death is different to the first instance and neither corresponds to the indicated cancer of interest. Pos1_3.")
}
# if either the first or second instance match the cancer of interest, set primary cause of death to the cancer of interest. This means that when a case has two different primary causes of death, one of which matches the cancer of interest, we set primary cause of death to the cancer of interest. E.g. one case had prostate cancer (C61) and Pulmonary edema (J81) listed as primary causes of death. We set the primary cause of death for this case to prostate cancer. This is a rare occurence. I've only observed this example once for prostate cancer and 0 times for lung cancer, breast cancer and overall cancer
Dat2$primary[Pos1_3]<-cancer_name2
if(icd10 != "C"){
if(Dat2$f.40001.1.0 !=icd10 & Dat2$f.40001.0.0 !=icd10) stop("second instance of f.40001 primary cause of death is different to the first instance and neither corresponds to the indicated cancer of interest")
}
dat<-rbind(Dat1,Dat2)
}
dat$primary[grep("C",dat$f.40001.0.0)]<-"other malignant neoplasm"
dat$primary[unlist(lapply(icd10,FUN=function(x) grep(x,dat$f.40001.0.0)))]<-cancer_name2
if(any(duplicated(unlist(lapply(icd10,FUN=function(x) grep(x,dat$f.40001.0.0)))))) stop("unexpected duplicates present")
Pos<-which(is.na(dat$primary) & !is.na(dat$f.40001.0.0))
Pos2<-grep("I",dat$f.40001.0.0[Pos])
dat$primary[Pos[Pos2]] <-"Diseases of the circulatory system"
Pos<-which(is.na(dat$primary) & !is.na(dat$f.40001.0.0))
Pos2<-grep("J",dat$f.40001.0.0[Pos])
dat$primary[Pos[Pos2]] <-"Diseases of the respiratory system"
Pos<-which(is.na(dat$primary) & !is.na(dat$f.40001.0.0))
Pos2<-grep("K",dat$f.40001.0.0[Pos])
dat$primary[Pos[Pos2]] <-"Diseases of the digestive system"
Pos<-which(is.na(dat$primary) & !is.na(dat$f.40001.0.0))
Pos2<-grep("G",dat$f.40001.0.0[Pos])
dat$primary[Pos[Pos2]] <-"Diseases of the nervous system"
Pos<-which(is.na(dat$primary) & !is.na(dat$f.40001.0.0))
Pos2<-unique(unlist(lapply(c("D0","D1","D2","D3","D4"),FUN=function(x)
grep(x,dat$f.40001.0.0[Pos]))))
dat$primary[Pos[Pos2]] <-"In situ or benign neoplasms or neoplasms of unknown/uncertain behavior, polycythemia vera and myelodysplastic syndromes"
Pos<-which(is.na(dat$primary) & !is.na(dat$f.40001.0.0))
Pos2<-grep("F",dat$f.40001.0.0[Pos])
dat$primary[Pos[Pos2]] <-"Mental, Behavioral and Neurodevelopmental disorders"
Pos<-which(is.na(dat$primary) & !is.na(dat$f.40001.0.0))
Pos2<-unique(unlist(lapply(c("A","B"),FUN=function(x)
grep(x,dat$f.40001.0.0[Pos]))))
dat$primary[Pos[Pos2]] <-"Certain infectious and parasitic diseases"
Pos<-which(is.na(dat$primary) & !is.na(dat$f.40001.0.0))
dat$primary[Pos] <-"other diseases"
Names<-names(dat)[grep("f.40001",names(dat))]
Names2<-names(dat)[grep("f.40002",names(dat))]
Test<-dat[which( !is.na(dat$date_death) & is.na(dat$primary)),c("date_death","primary","date_diagnosis2",Names,Names2)]
Names<-c(Names,Names2)
if(!all(is.na(Test[,Names]))) stop("primary cause of death is NA but not all primary and secondary causes of death are NA")
dat$primary[which( !is.na(dat$date_death) & is.na(dat$primary))]<-"unknown cause of death"
dat$status <- NA
dat$status[!is.na(dat$primary)]<-2 #death from any cause
dat$status[is.na(dat$primary)]<-1 #censored
dat$status2 <- 1
dat$status2[which(dat$primary==cancer_name2)]<-2 #death from any cause
# dat$status2[which(dat$primary!=cancer_name2)]<-1 #censored
return(dat)
}
# this doesn't work because applies to any diagnosis in cancer registry
# date_diagnosis<-function(dat=NULL){
# # Pos<-which(!is.na(dat$f.40005.1.0))
# # length(Pos)
# Names<-names(dat)[grep("f.40005",names(dat))]
# N_diag<-lapply(1:nrow(dat),FUN=function(x)
# unique(as.numeric(dat[x,Names])))
# dat$N_diagnoses<-unlist(lapply(1:nrow(dat),FUN=function(x)
# length(unlist(N_diag[x])[!is.na(unlist(N_diag[x]))])))
# min_date_diagnosis<-unlist(lapply(1:nrow(dat),FUN=function(x)
# min(as.numeric(dat[x,Names]),na.rm=TRUE)))
# dat$date_diagnosis<-as.Date(min_date_diagnosis,origin="1970-01-01")
# return(dat)
# }
date_diagnosis2<-function(dat=NULL){
Pos1<-grep("date.diagnosis",names(dat))
Pos2<-grep("lung",names(dat))
Pos<-Pos1[Pos1 %in% Pos2]
Names<-names(dat)[Pos]
# N_diag<-lapply(1:nrow(dat),FUN=function(x)
# unique(as.numeric(dat[x,Names])))
# dat$N_diagnoses<-unlist(lapply(1:nrow(dat),FUN=function(x)
# length(unlist(N_diag[x])[!is.na(unlist(N_diag[x]))])))
# Are all ICD codes in ICD instances lung cancer codes? If all C34 (icd10) or 16 (icd19) then yes.
Names2<-c("lung.ICD9.1","lung.ICD9.2","lung.ICD10.1","lung.ICD10.2","lung.ICD10.3")
ICD<-unlist(lapply(1:nrow(dat),FUN=function(x)
unique(dat[x,Names2][!is.na(dat[x,Names2])])))
ICD[grep("C34",ICD,invert=TRUE)]
# Yes all instances correspond to lung cancer icd codes
min_date_diagnosis<-unlist(lapply(1:nrow(dat),FUN=function(x)
min(as.numeric(dat[x,Names]),na.rm=TRUE)))
dat$date_diagnosis2<-as.Date(min_date_diagnosis,origin="1970-01-01")
# dat[which(dat$N_diagnoses==2)[1:2],c(Names,Names2)]
return(dat)
}
date_first_attended_assessment_centre<-function(dat=NULL){
Pos<-grep("f.53",names(dat))
Names<-names(dat)[Pos]
Dates<-lapply(Names,FUN=function(x)
as.Date(dat[,x]))
dat[,Names] <- lapply(1:length(Dates),FUN=function(x)
as.Date(unlist(Dates[x]),origin="1970-01-01"))
min_date<-unlist(lapply(1:nrow(dat),FUN=function(x)
min(as.numeric(dat[x,Names]),na.rm=TRUE)))
dat$first_date_attend_centre<-as.Date(min_date,origin="1970-01-01")
# dat[which(dat$N_diagnoses==2)[1:2],c(Names,Names2)]
return(dat)
}
# bd[Pos[2],c("date_diagnosis",Names)]
format_date_of_death<-function(dat=NULL){
# Date of death f.40000 . two instances, the second is nested within, and is identical to, the first
Pos0<-which(!is.na(dat$f.40000.0.0))
names(bd)[grep("40000",names(bd))]
names(dat)[grep("f.4000",names(dat))]
Pos1<-which(!is.na(dat$f.40000.1.0))
if(!all(Pos1 %in% Pos0)) stop("second instance of date of death is not nested within the first instance")
if(!all(dat$f.40000.1.0[Pos1] == dat$f.40000.0.0[Pos1])) stop("second instance of date of death is different to the first instance")
dat$f.40000.0.0 <- as.Date(dat$f.40000.0.0)
dat$f.40000.1.0 <- as.Date(dat$f.40000.1.0)
names(dat)[names(dat)=="f.40000.0.0"]<-"date_death"
return(dat)
}
# dat=bd
# time from diagnosis to death or last known date assumed alive
# this function seems to contain an error, in that followup is measured as time to censor date defined as "2018-02-06".
length_followup<-function(dat=NULL,censor_date=NULL){
dat$survival_days<-as.numeric(dat$date_death-dat$date_diagnosis2)
dat$survival_months<-round(dat$survival_days/(365.25/12),2)
dat$survival_years<-round(dat$survival_days/365.25,2)
dat$follow_days<-as.numeric(as.Date(censor_date)-dat$date_diagnosis2)
dat$follow_months<-round(dat$follow_days/(365.25/12),2)
dat$follow_years<-round(dat$follow_days/365.25,2)
return(dat)
}
# corrected. Followup is time to death or censor date
length_followup2<-function(dat=NULL,censor_date=NULL){
dat$survival_days2<-as.numeric(dat$date_death-dat$date_diagnosis2)
dat$survival_months2<-round(dat$survival_days2/(365.25/12),2)
dat$survival_years2<-round(dat$survival_days2/365.25,2)
dat$follow_days2<-as.numeric(as.Date(censor_date)-dat$date_diagnosis2)
Pos<-which(!is.na(dat$survival_days2))
dat$follow_days2[Pos]<-dat$survival_days2[Pos]
dat$follow_months2<-round(dat$follow_days2/(365.25/12),2)
dat$follow_years2<-round(dat$follow_days2/365.25,2)
return(dat)
}
find_max_date_diagnosis<-function(dat=NULL){
Names<-names(dat)[grep("f.40005",names(dat))]
max_date_diagnosis<-unlist(lapply(1:nrow(dat),FUN=function(x)
max(as.numeric(dat[x,Names]),na.rm=TRUE)))
dat$max_date_diagnosis<-as.Date(max_date_diagnosis,origin="1970-01-01")
dat$max_date_diagnosis[which(as.numeric(dat$max_date_diagnosis) == -Inf)]<-NA
return(dat)
}
cancer_histology<-function(dat=NULL){
Pos1<-grep("histology",names(dat))
Pos2<-grep("lung",names(dat))
Pos<-Pos1[Pos1 %in% Pos2]
Names<-names(dat)[Pos]
# this lines are just useful for testing the function
N_diag<-lapply(1:nrow(dat),FUN=function(x)
unique(as.numeric(dat[x,Names])))
dat$N_diagnoses<-unlist(lapply(1:nrow(dat),FUN=function(x)
length(unlist(N_diag[x])[!is.na(unlist(N_diag[x]))])))
Histology<-lapply(1:nrow(dat),FUN=function(x)
dat[x,Names])
Pos1<-grep("date.diagnosis",names(dat))
Pos2<-grep("lung",names(dat))
Pos<-Pos1[Pos1 %in% Pos2]
Names2<-names(dat)[Pos]
min_date_diagnosis<-unlist(lapply(1:nrow(dat),FUN=function(x)
min(as.numeric(dat[x,Names2]),na.rm=TRUE)))
# which instance contains the earliest diagnosis.
Instance_min_date<-unlist(lapply(1:nrow(dat),FUN=function(x)
which(as.numeric(dat[x,Names2]) == min_date_diagnosis[x])))
Histology2<-as.numeric(unlist(lapply(1:nrow(dat),FUN=function(x)
unlist(Histology[x])[Instance_min_date[x]])))
dat$histology<-Histology2
His<-read.table("/data/ph14916/UkbCancerMortality/data/coding38.tsv",sep="\t",head=TRUE,stringsAsFactors=FALSE)
dat<-merge(dat,His[,c("coding","meaning")],by.x="histology",by.y="coding")
names(dat)[names(dat)=="histology"]<-"histology_code"
names(dat)[names(dat)=="meaning"]<-"histology"
return(dat)
}
cancer_behaviour<-function(dat=NULL){
Pos1<-grep("behaviour",names(dat))
Pos2<-grep("lung",names(dat))
Pos<-Pos1[Pos1 %in% Pos2]
Names<-names(dat)[Pos]
Behaviour<-lapply(1:nrow(dat),FUN=function(x)
dat[x,Names])
Pos1<-grep("date.diagnosis",names(dat))
Pos2<-grep("lung",names(dat))
Pos<-Pos1[Pos1 %in% Pos2]
Names2<-names(dat)[Pos]
min_date_diagnosis<-unlist(lapply(1:nrow(dat),FUN=function(x)
min(as.numeric(dat[x,Names2]),na.rm=TRUE)))
# which instance contains the earliest diagnosis.
Instance_min_date<-unlist(lapply(1:nrow(dat),FUN=function(x)
which(as.numeric(dat[x,Names2]) == min_date_diagnosis[x])))
Behaviour2<-unlist(lapply(1:nrow(dat),FUN=function(x)
unlist(Behaviour[x])[Instance_min_date[x]]))
dat$behaviour<-Behaviour2
return(dat)
}
cancer_age<-function(dat=NULL){
Pos1<-grep("age_diagnosis",names(dat))
Pos2<-grep("lung",names(dat))
Pos<-Pos1[Pos1 %in% Pos2]
Names<-names(dat)[Pos]
Age<-lapply(1:nrow(dat),FUN=function(x)
dat[x,Names])
Pos1<-grep("date.diagnosis",names(dat))
Pos2<-grep("lung",names(dat))
Pos<-Pos1[Pos1 %in% Pos2]
Names2<-names(dat)[Pos]
min_date_diagnosis<-unlist(lapply(1:nrow(dat),FUN=function(x)
min(as.numeric(dat[x,Names2]),na.rm=TRUE)))
# which instance contains the earliest diagnosis.
Instance_min_date<-unlist(lapply(1:nrow(dat),FUN=function(x)
which(as.numeric(dat[x,Names2]) == min_date_diagnosis[x])))
Age2<-unlist(lapply(1:nrow(dat),FUN=function(x)
unlist(Age[x])[Instance_min_date[x]]))
dat$age_diagnosis<-Age2
return(dat)
}
# cancer_histology<function(){
# # # histology
# table(df_split$lung.ICD9.histology.1)
# table(df_split$lung.ICD9.histology.2)
# table(df_split$lung.ICD10.histology.1)
# table(df_split$lung.ICD10.histology.2)
# table(df_split$lung.ICD10.histology.3)
# His<-read.table("/data/ph14916/UkbCancerMortality/data/coding38.tsv",sep="\t",head=TRUE,stringsAsFactors=FALSE)
# dim(His)
# # small cell lung carcinoma (SCLC, 15% of all lung cancers)
# # non-SCLC (NSCLC, 85% of all lung cancers).
# # NSCLCs are generally subcategorized into adenocarcinoma, squamous cell carcinoma (SqCC), and large cell carcinoma.
# His$meaning[grep("squamous",His$meaning)]
# head(His)
# His2<-His[,c("coding","meaning")]
# sort(His$meaning[grep("adenocarcinoma",His$meaning)])
# # Adenocarcinoma
# # df_split<-df_split[,grep("meaning",names(df_split),invert=TRUE)]
# # df_split$lung.ICD9.histology_meaning.1<-df_split$lung.ICD9.histology.1
# df_split.m<-merge(df_split,His2,by.x="lung.ICD9.histology.1",by.y="coding",all.x=TRUE)
# names(df_split.m)[names(df_split.m)=="meaning"]<-"lung.ICD9.histology_meaning.1"
# df_split.m<-merge(df_split.m,His2,by.x="lung.ICD9.histology.2",by.y="coding",all.x=TRUE)
# names(df_split.m)[names(df_split.m)=="meaning"]<-"lung.ICD9.histology_meaning.2"
# df_split.m<-merge(df_split.m,His2,by.x="lung.ICD10.histology.1",by.y="coding",all.x=TRUE)
# names(df_split.m)[names(df_split.m)=="meaning"]<-"lung.ICD10.histology_meaning.1"
# df_split.m<-merge(df_split.m,His2,by.x="lung.ICD10.histology.2",by.y="coding",all.x=TRUE)
# names(df_split.m)[names(df_split.m)=="meaning"]<-"lung.ICD10.histology_meaning.2"
# df_split.m<-merge(df_split.m,His2,by.x="lung.ICD10.histology.3",by.y="coding",all.x=TRUE)
# names(df_split.m)[names(df_split.m)=="meaning"]<-"lung.ICD10.histology_meaning.3"
# table(df_split.m$lung.ICD9.histology.1,df_split.m$lung.ICD9.histology_meaning.1)
# )
# table(df_split.m$lung.ICD10.histology_meaning.1,df_split.m$lung.ICD10.histology_meaning.2)
# Pos1<-which(df_split.m$lung.ICD10.histology_meaning.1=="Sarcoma, NOS")
# Pos2<-which(df_split.m$lung.ICD10.histology_meaning.2=="Small cell carcinoma, NOS")
# Pos<-Pos1[Pos1 %in% Pos2]
# name.pos<-grep("lung.ICD10",names(df_split.m))
# names(df_split.m)[name.pos]
# df_split.m[Pos1,c( "lung.ICD10.1",
# "lung.ICD10.2",
# "lung.ICD10.date.diagnosis.1",
# "lung.ICD10.date.diagnosis.2",
# "lung.ICD10.behaviour.1",
# "lung.ICD10.behaviour.2",
# "lung.ICD10.age_diagnosis.1",
# "lung.ICD10.age_diagnosis.2",
# "lung.ICD10.histology_meaning.1",
# "lung.ICD10.histology_meaning.2")]
# df_split.m$lung.ICD9.histology_meaning.2)
# df_split.m$lung.ICD9.histology_meaning.1)
# ),
# ,
# df_split.m$lung.ICD10.histology_meaning.3,
# collapse="; ")
# table(df_split.m$histology)
# # behaviour unknown or metstatic behavipour/site
# table(df_split.m$lung.ICD9.histology_meaning.1[df_split$lung.ICD9.behaviour.1=="NA"])
# Pos<-which(df_split$lung.ICD9.behaviour.1=="Malignant, primary site")
# Pos2<-which(df_split$lung.ICD9.behaviour.1=="NA")
# # histology missing
# df_split.m$lung.ICD9.histology_meaning.1[Pos]
# df_split.m$lung.ICD9.histology_meaning.1[Pos2]
# table(df_split.m$lung.ICD10.histology_meaning.1,)
# unique(df_split$lung.ICD10.behaviour.1)
# Pos<-which(df_split$lung.ICD10.behaviour.1=="Malignant, metastatic site")
# Pos2<-which(df_split$lung.ICD10.behaviour.1=="Uncertain whether benign or malignant")
# Pos3<-which(df_split$lung.ICD10.behaviour.1=="NA")
# df_split.m$lung.ICD10.histology_meaning.1[Pos]
# df_split.m$lung.ICD10.histology_meaning.1[Pos2]
# df_split.m$lung.ICD10.histology_meaning.1[Pos3]
# # missing histology
# head(df_split.m)
# head(His)
# df_split$lung.ICD9.histology_meaning.2<-df_split$lung.ICD9.histology.2
# df_split$lung.ICD10.histology_meaning.1<-df_split$lung.ICD10.histology.1
# df_split$lung.ICD10.histology_meaning.2<-df_split$lung.ICD10.histology.2
# df_split$lung.ICD10.histology_meaning.3<-df_split$lung.ICD10.histology.3
# names(df_split)
# print("generating histology flag")
# df_split$histology.flag <-
# ifelse(between(df_split$headneck.ICD9.histology.1, 8070, 8078) |
# between(df_split$headneck.ICD10.histology.1, 8070, 8078) |
# between(df_split$headneck.ICD10.histology.2, 8070, 8078) |
# between(df_split$headneck.ICD10.histology.3, 8070, 8078),
# 1,
# 0)
qc_check<-function(){
message("checking baseline date and definition of incident cases for possible issues")
# check date of consent same as first date attended assessment centre
if(!all(lun4$date_consent == lun4$first_date_attend_centre)) warning("consent date not always the same as first date of attendance at assessment centre")
# check that incident cases were diagnosed after UKB baseline date using alternative method for determining incident case
Pos1<-which(lun4$date_diagnosis2 > lun4$date_consent)
Pos2<-which(!is.na(lun4$incident_lung_cancer))
if(!all(Pos1==Pos2)) stop("possible conflict between date of diagnosis for incident cases and date of enrollment into UKB")
Pos3<-which(lun4$date_diagnosis2 <= lun4$date_consent)
Pos4<-which(is.na(lun4$incident_lung_cancer))
if(!all(Pos3==Pos4)) stop("possible conflict between date of diagnosis for incident cases and date of enrollment into UKB")
message("no issues found. nice")
}
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.