SEERaBomb (version 2019.2)

canc2py: Converts canc made by mkSEER into a data.frame with py at risk after firsts

Description

Makes a data frame in the exernal second cancer data format expected by esd.

Usage

canc2py(canc,firstS,secondS)

Arguments

canc

output of mkSEER

firstS

Charcter vector of first cancers of interest.

secondS

Charcter vector of second cancers of interest.

Value

data.frame with columns: yrdx, agedx, sex, py at risk (in years), cancer1, and cancer2. Cases not ending in a second cancer have cancer2 set to "none".

Details

Use is currently limited to 2nd cancers in SEER since 1973, e.g. AML, since esd assumes its py column is in synch with times since diagnosis. This is not the case when e.g. MDS or CMML are in secondS.

See Also

SEERaBomb-package, esd,msd

Examples

# NOT RUN {
rm(list=ls()) 
library(SEERaBomb)
load("~/data/SEER/mrgd/cancDef.RData") 
secS=c("AML","MDS")
frstS=c("HL")
canc=canc%>%filter(cancer%in%union(frstS,secS))
load("~/data/SEER/mrgd/popsae.RData")
popsa=popsae%>%group_by(db,race,sex,age,year)%>%summarize(py=sum(py)) # sum on regs
pm=seerSet(canc,popsa,Sex="male",ageStart=0,ageEnd=100) 
pf=seerSet(canc,popsa,Sex="female",ageStart=0,ageEnd=100) 
pm=mk2D(pm,secondS=secS) 
pf=mk2D(pf,secondS=secS)
brks=c(0,1,5,10)
pm=csd(pm,brkst=brks,firstS=frstS) 
pf=csd(pf,brkst=brks,firstS=frstS)
S=rbind(pf$DF,pm$DF)%>%select(int,t,cancer1,cancer2,O,E) #merge sexes
S=S%>%group_by(int,cancer1,cancer2)%>%summarize(O=sum(O),E=sum(E),t=mean(t))
S=S%>%mutate(RR=O/E,rrL=qchisq(.025,2*O)/(2*E),rrU=qchisq(.975,2*O+2)/(2*E))
S$data="SEER"

ds=canc2py(canc,frstS,secS)
S2=esd(ds,pf$D,pm$D,brks)
S2=S2%>%group_by(int,cancer1,cancer2)%>%summarize(O=sum(O),E=sum(E),t=mean(t))
S2=S2%>%mutate(RR=O/E,rrL=qchisq(.025,2*O)/(2*E),rrU=qchisq(.975,2*O+2)/(2*E))
S2$data="SEER2"
d=rbind(S,S2)
d%>%filter(cancer1=="HL",cancer2=="AML")%>%arrange(int)
d%>%filter(cancer1=="HL",cancer2=="MDS")%>%arrange(int) #low because MDS not observable before 2001
yrcut=2001  # try to fix it like this
ds=ds%>%filter(yeaR+py>yrcut)
I=ds$yeaR<yrcut
ds$py[I]=(ds$py-(yrcut-ds$yeaR))[I]
S2=esd(ds,pf$D,pm$D,brks)
S2=S2%>%group_by(int,cancer1,cancer2)%>%summarize(O=sum(O),E=sum(E),t=mean(t))
S2=S2%>%mutate(RR=O/E,rrL=qchisq(.025,2*O)/(2*E),rrU=qchisq(.975,2*O+2)/(2*E))
S2$data="SEER2"
d=rbind(S,S2)
d%>%filter(cancer1=="HL",cancer2=="MDS")%>%arrange(int) 
#the problem is that esd assumes that py is synched with tsx. csd handles the delays correctly.
d%>%filter(cancer1=="HL",cancer2=="MDS")%>%group_by(data)%>%summarize(o=sum(O)) #cases were shifted
# }