jointmeta2: Function to pool joint model fits in two stage MA
In mesudell/joineRmeta: Joint Modelling for Meta-Analytic (Multi-Study) Data

Description Usage Arguments Details Value References See Also Examples

This function takes joint model fits from either joint or jointModel and pools the information from the fits in the second stage of a two stage meta-analysis (MA).

1 2	jointmeta2(fits, SE = NULL, longpar = NULL, survpar = NULL, assoc = TRUE, studynames = NULL)

`fits`	a list of joint modelling fits. These fits should all be of the same type, with the same model specification.
`SE`	a list of the results from `jointSE`. Only to be supplied if the model fits supplied in `fits` are all fitted using the `joineR` package.
`longpar`	a vector of character strings of parameters from the longitudinal sub-model for which meta-analyses should be performed
`survpar`	a vector of character strings of parameters from the survival sub-model for which meta-analyses should be performed
`assoc`	a logical indicating whether a meta-analysis should be performed for the association parameter(s)
`studynames`	a vector of character strings containing the names for the studies present in the dataset that the joint models were fitted to. These character strings if supplied are used to label the meta-analyses performed by the function

The joint model fits modelled using the joineR package link the sub-models using shared zero mean random effects (see Henderson et al (2000)). However the joint model fits modelled using the JM package link the sub-models using sharing structures that involve both the fixed and random effects. If a parameter specified in survpar is also present in the fixed effects of the longitudinal sub-model, a direct effect of the parameter on the risk of an event can be extracted from the survival sub-model, as well as the overall effect resulting from the sum of fixed effect in the survival sub-model, and the presence of the parameter in the longitudinal sub-model, present in the sharing structure of the joint model. As such, if a parameter specified in survpar is also present as a fixed effect in the longitudinal sub-model, and the fixed and random effects make up the sharing structure linking the sub-models, the overall parameter effect is found by β_2 + (α * β_1), where α is the association parameter, β_2 is the coefficient for the parameter in question from the survival sub-model, and β_1 is the coefficient for the parameter in question from the longitudinal sub-model. For more information about overall effects versus direct effects see Ibrahim et al (2010), Rizopoulos (2012) and Gould et al (2015). Because both a direct and an overall effect of the survival parameters can be extracted from the model, both are present in the results if the joint models supplied in the fits are fitted using the JM package.

This function returns a list of results for the two stage MA. These results are split by the type of parameter being pooled. If the names of longitudinal parameters were supplied to longpar then an element named longMA will be present in the results. If the names of survival parameters were supplied to survpar then if the supplied joint model fits were fitted using the joint function from the joineR package, an element named survMA.direct will be present in the results. If the supplied joint model fits were fitted using the jointModel function from the JM package, two elements named survMA.direct and survMA.overall will be present. If assoc = TRUE then an element labelled assocMA will be present in the results.

Each element of each of these components of the results (longMA, survMA.direct, assocMA...) is of class metagen, and is the result of using the metagen function on the results of joint models fitted to multiple studies in the dataset. This method pools the supplied information in fixed and random MA using inverse variance weighting. Forest plots can be produced for these results simply by applying the function forest to the objects of class metagen / meta supplied in the results.

Ibrahim et al (2010) Basic Concepts and Methods for Joint Models of Longitudinal and Survival Data. JOURNAL OF CLINICAL ONCOLOGY 28 (10): 2796-2801

Rizopoulos (2012) Joint Models for Longitudinal and Time-to-Event Data With Applications in R. Chapman and Hall/CRC Biostatistics Series

Henderson et al (2000) Joint modelling of longitudinal measurements and event time data. Biostatistics, 1,4, pp. 465–480

Gould et al (2015) Joint modeling of survival and longitudinal non-survival data: current methods and issues. Report of the DIA Bayesian joint modeling working group. Statistics in Medicine 34(14): 2181–2195. doi:10.1002/sim.6141.

joint, jointModel, jointSE, metagen

joineRmodels <- joineRfits[c("joineRfit1", "joineRfit2", "joineRfit3")]
joineRmodelsSE <- joineRfits[c("joineRfit1SE", "joineRfit2SE",
                               "joineRfit3SE")]

MAjoineRfits <- jointmeta2(fits = joineRmodels, SE = joineRmodelsSE,
                          longpar = c("time", "treat1"),
                          survpar = "treat1", assoc = TRUE,
                          studynames = c("Study 1", "Study 2", "Study 3"))