Scribd
Upload
54 views19 pages

T Merge - A Tool To Facilitate Creation of Multiple Time-Based Intervals Per Subject

The tmerge function in R makes it easier to create multiple start and stop intervals per subject for time-to-event analysis when there are time-dependent covariates, multiple outcomes, or multi-state models. It sequentially inserts new data into an existing dataset to create new intervals defined by variables added using tdc() for time-dependent covariates or event() for endpoints. Time-dependent covariates apply from the start of a new interval while events only occur at the end of an interval. The function handles gaps in intervals and carries forward last observed values.

Uploaded by

Hao Liu
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
54 views19 pages

T Merge - A Tool To Facilitate Creation of Multiple Time-Based Intervals Per Subject

The tmerge function in R makes it easier to create multiple start and stop intervals per subject for time-to-event analysis when there are time-dependent covariates, multiple outcomes, or multi-state models. It sequentially inserts new data into an existing dataset to create new intervals defined by variables added using tdc() for time-dependent covariates or event() for endpoints. Time-dependent covariates apply from the start of a new interval while events only occur at the end of an interval. The function handles gaps in intervals and carries forward last observed values.

Uploaded by

Hao Liu
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 19

tmerge

A Tool to Facilitate Creation of Multiple Time-Based Intervals


per Subject

Cynthia Crowson, Terry Therneau and Elizabeth Atkinson

SDSS
May 18, 2018

tmerge
Time-to-event analysis

I Often requires creation of multiple start/stop intervals per


subject
I time-dependent covariates
I multiple outcomes per subject
I multi-state models
I Deceptively simple task, easy to do incorrectly

0 1 2 3 4 5 6 7 8 9 10

Time since study entry, years

tmerge
Simple Example
I starting dataset has 1 observation per subject
I surgery is a time-dependent covariate

id age tm_surg futime event


1 1 40 5 10 0
2 2 20 8 20 1
3 3 50 NA 30 1

I need to separate time periods before and after surgery

id age tstart tstop death surgery


1 1 40 0 5 0 0
2 1 40 5 10 0 1
3 2 20 0 8 0 0
4 2 20 8 20 1 1
5 3 50 0 30 1 0
tmerge
tmerge function

I tmerge function in survival package in R makes this task


easier
I sequential insertion
I build the dataset one covariate or endpoint at a time
I each addition will be “slipped in” to the original data in the
same way that one would slide a new card into an existing deck
of cards

tmerge
tmerge function

I The basic form of the function call is

newdata <- tmerge(data1, data2, id,


newvar=tdc(time, value), ...)

I primary arguments:
I data1: baseline data to be retained in the analysis dataset
I data2: source for new data including events and time-dependent
covariates
I id: subject identifier used to merge the data together
I . . . : additional arguments that add variables to the dataset
I tstart, tstop: used to set the time range for each subject
I options

tmerge
tmerge function

I The key part of the call are the “. . . ” arguments, which each
can be one of 4 types:
I tdc() and cumtdc() add a time-dependent covariate
I event() and cumevent() add a new endpoint
I resulting dataset has 3 new variables (at least):
I id: identifier indicating which rows belong to the same subject
I tstart: start of the interval
I tstop: end of the interval

tmerge
Example

I dataset: d1

id age
1 1 40
2 2 20
3 3 50

I dataset: d2

id tm_surg futime event


1 1 5 10 0
2 2 8 20 1
3 3 NA 30 1

tmerge
Example: step 1 - create start/stop time

step1 <- tmerge(data1=d1, data2=d2, id=id,


death=event(futime, event))
step1

id age tstart tstop death


1 1 40 0 10 0
2 2 20 0 20 1
3 3 50 0 30 1

tmerge
Example: step 2 - create time-dependent covariate
step2 <- tmerge(data1=step1, data2=d2,id=id,
surgery=tdc(tm_surg))
step2

id age tstart tstop death surgery


1 1 40 0 5 0 0
2 1 40 5 10 0 1
3 2 20 0 8 0 0
4 2 20 8 20 1 1
5 3 50 0 30 1 0

Note: this can also be done in just one step:

tmerge(data1=d1, data2=d2, id=id,


death=event(futime, event),
surgery=tdc(tm_surg))
tmerge
How covariates differ from events

I time-dependent covariates
I apply from the start of a new interval
I persist for all remaining intervals unless subsequently changed
I events
I occur at the end of an interval
I only occur once
I in time-to-event analyses, time intervals are open on the left
and closed on the right, i.e., (tstart, tstop].

0 1 2 3 4 5 6 7 8 9 10

Time since study entry, years

tmerge
Example: Multiple Events

I Here is the data for the first 4 subjects

id treat sex age futime etime1 etime2 etime3


1 1 1 2 12 414 219 373 NA
2 2 0 1 15 439 8 26 152
3 3 1 1 19 382 NA NA NA
4 4 1 1 12 388 NA NA NA

tmerge
Example: Multiple Events
dim(cgd0)

[1] 128 20

newcgd <- tmerge(data1=cgd0[, 1:13], data2=cgd0,


id=id, tstop=futime)
newcgd <- tmerge(newcgd, cgd0, id=id,
infect = event(etime1))
newcgd <- tmerge(newcgd, cgd0, id=id,
infect = event(etime2))
newcgd <- tmerge(newcgd, cgd0, id=id,
infect = event(etime3))
dim(newcgd)

[1] 196 16
tmerge
Example: Multiple Events

id treat sex age futime tstart tstop infect


1 1 1 2 12 414 0 219 1
2 1 1 2 12 414 219 373 1
3 1 1 2 12 414 373 414 0
4 2 0 1 15 439 0 8 1
5 2 0 1 15 439 8 26 1
6 2 0 1 15 439 26 152 1
7 2 0 1 15 439 152 439 0

tmerge
tcount attribute

attr(newcgd, "tcount")

early late gap within boundary lead trail tied


infect 0 0 0 44 0 0 0 0
infect 0 0 0 16 0 0 1 0
infect 0 0 0 8 0 0 0 0

tmerge
Example: Continuous values that change over time
I pbc data set contains baseline data and follow-up status for
312 subjects with primary biliary cirrhosis (one obs per person)
I pbcseq data set contains 1945 repeated laboratory values

pbc2 <- tmerge(pbc, pbc, id=id,


death = event(time, status)) #set range
pbc2 <- tmerge(pbc2, pbcseq, id=id,
ascites = tdc(day, ascites),
bili = tdc(day, bili),
albumin = tdc(day, albumin))

id tstart tstop death ascites bili albumin


1 1 0 192 0 1 14.5 2.60
2 1 192 400 2 1 21.3 2.94
3 2 0 182 0 0 1.1 4.14
4 2 182 365 0 0 0.8 3.60
5 2 365 768 0 0
tmerge
1.0 3.55
Example: Continuous values that change over time

early late gap within boundary lead trail tied


death 0 0 0 0 0 0 312 0
ascites 0 131 0 1442 0 312 0 0
bili 0 138 0 53 1442 312 0 0
albumin 0 138 0 0 1495 312 0 0

I Missing values in time or value from data2 are ignored


I Consequence: “last value carried forward”
I Default can be changed by adding
options=list(na.rm=FALSE) to the second call
I Any tdc calls with a missing time are still ignored, independent
of the na.rm value, since we would not know where to insert
them.

tmerge
Time delay
I For any data set containing constructed time-dependent
covariates, it is a good idea to re-run the analyses after adding
a 7-14 day lag to key variables.
I One reason is to check for cases of reverse causality. A covariate
measured soon before death may not be a predictor of death but
rather is simply a marker for an event that is already in progress.
I Even more subtle biases can occur via coding errors.
I When the results show a substantial change, understanding
why this occurred is an critical step.

0 1 2 3 4 5 6 7 8 9 10

Time since study entry, years

pbc2a <- tmerge(pbc2, pbcseq, id=id,


ascites = tdc(day,
tmerge ascites),
Gaps
I Time dependent covariates that occur before the start of a
subject’s follow-up interval or during a gap in time do not
generate a new interval split, but they do set the value of that
covariate for future times.
I Rationale: during a subject’s time within the county we would
like the variable “prior diagnosis of diabetes” to be accurate,
even if that diagnosis occurred during a non-resident period
I Events that occur in a gap are not counted.
I For events outside of the timeline, we have no way to know who
the appropriate comparison group is, and so must ignore those
events.

0 1 2 3 4 5 6 7 8 9 10

Time since study entry, years


tmerge
Summary

I tmerge is a simple to use, flexible tool to create multiple


start/stop intervals per subject
I time-dependent covariates - both binary and continuous
I multiple outcomes per subject
I allows for gaps in time
I data checks can help avoid errors
I tcount attribute
I use of delay for time-dependent covariates

tmerge

You might also like