simulated HCV data {mspath}R Documentation

HCV Progression after liver transplant

Description

Simulated data for individuals with liver transplants, giving covariates, time since transplant, and observed liver state. Cases generally have repeated biopsies.

Format

A data frame. sim1 has 21 cases with 46 observations; sim2 has 335 cases with 789 records, and sim3 has 1,340 cases with 3,356 records.

id
The patient id.
fib
Fibrosis state from 1 through 5, using the Metavir scale (see the reference by Bedosa et al. We actually use the Metavir value +1 because the algorithm expects 1 as the first state, not 0.). 1 means a healthy liver, with higher numbers representing increasing liver damage. 5 means cirrhosis.
time
Years since transplant.
k
(sim1 only) Standard Normal covariate intended to vary between cases but not across time within a case. The intention was not realized: it variest within cases as well.
x1
(sim1 only) Standard Normal covariate, drawn independently for each time of each case.
C1
(sim2 and sim3) Normally distributed continuous variable.
C2
(sim2 and sim3) Exponentially distributed continuous variable.
D3
(sim2 and sim3) binary variable
C4
(sim2 and sim3) age, from a uniform initial age and time of observation.
K5
(sim2 and sim3) Normally distributed continuous variable, constant within case.
K6
(sim2 and sim3) Binary variable, constant within case.

Details

The datasets include a time 0 observation of fibrosis stage 1 (uninfected) for each case. Note the times are time of observation, not time of entry into a state. Fibrosis is observed with error; errors are likely to produce a reading lower than the actual score.

As required by mspath, observations are sorted by id and then time.

It is unusual to know the exact time of infection and have repeated observations per case. We were interested in a study of transplant patients that had both features.

The observed fibrosis stage was generated by simulating the mspath model, using do.what=10. As an example, after generating the cases, covariates and observation times for sim1 the outcomes came from

data(q2, e2, sim1)
cons1 <- list(k=c(1, 1, 2, 2), x1=rep(3, 4))
init1 <- c(seq(-1, by=.2, length=4),# intercepts
              0.5, 0.8, #k
              -1.0, # x1
              -0.5, .7, # history
              .2, .05 # misclassification
           )

r <- mspath(fib~time, q2, misc=SIMPLE, e2,
            inits=init1,
            subject=id,
            covariates=~k+x1,
            constraint=cons1,
            econstraint=c(1, 2, 1, 2, 1),
            pathvars=c("LN(TSO)", "TIS"),
            pathoffset=0.5,
            pathconstraint=list("LN(TSO)"=rep(1, 4), "TIS"=rep(2, 4)),
            data=sim1,
            isexact=TRUE,
            fixedpars=seq(along=init1),
            stepnumerator=1,
            stepdenominator=1,
            do.what=10)
The final sim1 was extracted from the return value r.

Source

See the reference by Berenguer et al. for a published study using a dataset in the same format as this simulated one. That study did not use multi-state path methods.

References

Bedosa P, Pynard T. An algorithm for the grading of activity in chronic hepatitis C. The METAVIR Cooperative Study Group. Hepatology 1996. 24:289-93.

Berenguer M, Ferrell L, Watson J, et al. HCV-related fibrosis progression following liver transplantation: increase in recent years. J Hepatology, 32:673-84, 2000.

Examples

library(mspath)
data(q2, e2, sim2)
 r <- mspath(fib~time, misc=TRUE, ematrix=e2, qmatrix=q2, inits=rep(.5, 9), subject=id,
             data=sim2, stepnumerator=1, stepdenominator=1, initprobs=c(1.0, 0, 0, 0, 0),
             do.what=0)

[Package mspath version 0.9-9 Index]