coxph.risk: Absolute risk estimation with Cox proportional hazards models

In skoval/coxph.risk: Computes absolute risk for cohort data

Description

Method to estimate absolute risk in the presence of multiple competing events and with each event hazard specified by a Cox proportional hazards model.

Usage

coxph.risk(begin, end, newdata, coxph1, ..., na.action = na.exclude)

Arguments

begin	vector specifying the begin of the projection interval, [begin, end). If scalar is given, interval is repeated for all newdata
end	vector specifying the end of the projection interval. If scalar is given, interval is repeated for all newdata
newdata	data frame containing the risk profiles for the individualized prediction as in predict.glm
coxph1	a coxph.object for the primary event
...	additional coxph.objects for competing events
na.action	function for handling missing data among model variables

Value

A vector of the absolute risk of the primary event occurring within [begin, end) for each individual of newdata.

Author(s)

Stephanie Kovalchik <[e-mail]>

Examples

data(mgus)

# Mayo Clinic 20-35 year follow-up of patients with 
# monoclonal gammopathy of undetermined significance (MGUS)

# Hazard models of multiple myeloma, death, other plasma malignancy
# Time scale is days from MGUS diagnosis

myeloma.model <- Surv(time, status)~age+factor(sex)+alb+hgb+mspike
competing.model <- Surv(time, status)~age*factor(sex)

cox1 <- coxph(myeloma.model,data=mgus2,subset=event=="myeloma")
cox2 <- coxph(competing.model,data=mgus2,subset=event=="death")
cox3 <- coxph(competing.model,data=mgus2,subset=event=="other")

# Absolute risk predictions for multiple myeloma in 5 years
predict.data <-  mgus2[mgus2$event=="death",]

# ONLY COMPLETE CASES
predict.data <- predict.data[complete.cases(predict.data),]
risk <- coxph.risk(0, 5*365.25, newdata = predict.data,
     			     cox1, cox2, cox3)

summary(risk)

# RISK BY AGE AND GENDER AT MGUS DIAGNOSIS
cols <- c("dodgerblue","darkorchid")

plot(risk*1000~age, data = predict.data, 
	       ylab = "multiple myeloma 5-yr absolute risk (per 1000)",
	       las = 1, col = cols[predict.data$sex])

legend("topright", bty="n", levels(predict.data$sex), col= cols, pch=1)

skoval/coxph.risk documentation built on May 30, 2019, 1:07 a.m.