Package {landmaRk}

Title:	Time-to-Event Landmark Analysis using an Array of Longitudinal and Survival Sub-Models
Version:	0.1.1
Description:	Provides a modular end-to-end framework for dynamic risk prediction based on time-to-event and longitudinal data. This allows flexible specifications for the longitudinal and survival sub-models. The 'landmaRk' package enables reproducible benchmarks of different model choices, including cross-validation to assess out-of-sample predictive performance. Methods are described in Velasco-Pardo, Constantine-Cooke, Lees and Vallejos (2026, manuscript under preparation) 'Landmarking with Latent Class Mixed Models for Dynamic Prediction of Time-to-event Data with Heterogeneous Biomarker Trajectories'.
License:	GPL (≥ 3)
BugReports:	https://github.com/VallejosGroup/landmaRk/issues
Encoding:	UTF-8
RoxygenNote:	7.3.3
Depends:	R (≥ 4.1.0), survival
Imports:	dplyr, ggplot2, lme4, Matrix, methods, pec, riskRegression, rlang, lcmm (≥ 2.2.2), doParallel, foreach, utils, timeROC, prodlim
Suggests:	JMbayes2, knitr, rmarkdown, testthat (≥ 3.0.0), xml2, tidyverse, spelling, joineR, survminer, withr
Config/testthat/edition:	3
Config/testthat/parallel:	true
LazyData:	true
VignetteBuilder:	knitr
URL:	https://vallejosgroup.github.io/landmaRk/, https://github.com/VallejosGroup/landmaRk
Language:	en-GB
NeedsCompilation:	no
Packaged:	2026-05-18 09:12:58 UTC; vvelasco
Author:	Victor Velasco-Pardo [aut, cre], Nathan Constantine-Cooke [aut], Catalina Vallejos [aut], Charlie Lees [aut]
Maintainer:	Victor Velasco-Pardo <vvelasco@ed.ac.uk>
Repository:	CRAN
Date/Publication:	2026-05-27 10:00:16 UTC

landmaRk

Description

Provides a modular end-to-end framework for dynamic risk prediction based on time-to-event and longitudinal data. This allows flexible specifications for the longitudinal and survival sub-models. The 'landmaRk' package enables reproducible benchmarks of different model choices, including cross-validation to assess out-of-sample predictive performance. Methods are described in Velasco-Pardo, Constantine-Cooke, Lees and Vallejos (2026, manuscript under preparation) 'Landmarking with Latent Class Mixed Models for Dynamic Prediction of Time-to-event Data with Heterogeneous Biomarker Trajectories'.

Author(s)

Maintainer: Victor Velasco-Pardo vvelasco@ed.ac.uk (ORCID)

Authors:

• Nathan Constantine-Cooke nathan.constantine-cooke@ed.ac.uk (ORCID)

• Catalina Vallejos catalina.vallejos@ed.ac.uk (ORCID)

• Charlie Lees charlie.lees@ed.ac.uk (ORCID)

See Also

Useful links:

• https://vallejosgroup.github.io/landmaRk/

• https://github.com/VallejosGroup/landmaRk

• Report bugs at https://github.com/VallejosGroup/landmaRk/issues

Creates an S4 class for a landmarking analysis

Description

Creates an S4 class for a landmarking analysis

Usage

LandmarkAnalysis(
  data_static,
  data_dynamic,
  event_indicator,
  ids,
  event_time,
  times,
  measurements,
  censor_at_landmark = TRUE,
  K = 1
)

Arguments

Value

An object of class LandmarkAnalysis

S4 class for performing a landmarking analysis

Description

S4 class for performing a landmarking analysis

Slots

landmarks

A numeric vector of landmark times.

data_static

A data frame containing subject ids, static covariates,

data_dynamic

Data frame in long format with subject ids, measurement values, measurement times and measurement name.

event_indicator

Name of the column indicating event or censoring.

ids

Name of the column indicating subject ids.

times

Name of the column indicating observation time in data_dynamic.

measurements

Name of the column indicating measurement values in data_dynamic.

censor_at_landmark

Boolean indicating whether to fit a single longitudinal model to the complete dataset (FALSE) or to censor observations at the landmark time prior to fitting the longitudinal model, iterating through landmark times (TRUE; default)

event_time

Name of the column indicating time of the event/censoring.

risk_sets

List of indices.

longitudinal_fits

List of model fits for the specified landmark times and biomarkers.

longitudinal_predictions

List of model predictions for the specified landmark times and biomarkers.

longitudinal_predictions_test

List of out-of-sample predictions for the specified landmark times and biomarkers.

survival_datasets

List of survival dataframes used in the survival submodel.

survival_datasets_test

List of survival dataframes used for out-of-sample predictions with the survival submodel.

survival_fits

List of survival model fits at each of the specified landmark times.

survival_predictions

List of time-to-event predictions for the specified landmark times and prediction horizons.

survival_predictions_test

List of out-of-sample predictions for the time-to-event outcome if K > 1.

K

Number of cross-validation folds (1 by default)

cv_folds

Data frame associating individuals to cross-validation folds

Checks convergence of lcmm models

Description

Checks convergence of lcmm models

Usage

check_lcmm_convergence(x)

Arguments

Value

No return value. Issues a message if all lcmm models converged, or a warning for each model that did not converge.

Compute the list of individuals at risk at landmark times

Description

Compute the list of individuals at risk at landmark times

Usage

compute_risk_sets(x, landmarks, .warn_when_less_than = 0, ...)

Arguments

Details

A risk set describes all subjects still at risk (i.e., not experienced the event of interest or censored) at a given time. In LandmarkAnalysis, risk sets define which subjects should be included in the longitudinal and survival sub-models for each landmark time.

The risk sets are stored in the risk_sets slot of the LandmarkAnalysis object, where each risk set is a list of indices corresponding to the subjects at risk at the respective landmark time.

Value

An object of class LandmarkAnalysis including desired risk sets for the specified landmark times.

Compute the list of individuals at risk at landmark times

Description

Compute the list of individuals at risk at landmark times

Usage

## S4 method for signature 'LandmarkAnalysis'
compute_risk_sets(x, landmarks, .warn_when_less_than = 0, ...)

Arguments

Details

A risk set describes all subjects still at risk (i.e., not experienced the event of interest or censored) at a given time. In LandmarkAnalysis, risk sets define which subjects should be included in the longitudinal and survival sub-models for each landmark time.

The risk sets are stored in the risk_sets slot of the LandmarkAnalysis object, where each risk set is a list of indices corresponding to the subjects at risk at the respective landmark time.

Value

An object of class LandmarkAnalysis, including desired risk sets for the relevant landmark times.

Dose calibration of anti-epileptic drugs data

Description

The SANAD (Standard and New Anti-epileptic Drugs) study (Marson et al., 2007) is a randomized control trial of standard and new anti-epileptic drugs, comparing effects on longer term clinical outcomes. The data consists of longitudinal measurements of calibrated dose for the groups randomized to a standard drug (CBZ) and a new drug (LTG). The objective of the analysis is to investigate the effect of drug titration on the relative effects of LTG and CBZ on treatment failure (withdrawal of the randomized drug). There are several baseline covariates available, and also data on the time to withdrawal from randomized drug. This version of the data has been modified from the joineR package.

Usage

data(epileptic)

Format

This is a data frame in the unbalanced format, that is, with one row per observation. The data consists of columns for patient identifier, time of measurement, calibrated dose, baseline covariates, and survival data. The column names are identified as follows:

id

integer: patient identifier.

time

integer: timing of clinic visit at which dose recorded (days).

with.time

integer: time of drug withdrawal/maximum follow up time (days).

with.status

censoring indicator (1 = withdrawal of randomized drug and 0 = not withdrawn from randomized drug/lost to follow up).

dose

numeric: calibrated dose.

treat

factor: randomized treatment (CBZ or LTG).

age

numeric: age of patient at randomization (years).

gender

factor: gender of patient. F = female, M = male.

learn.dis

factor: learning disability.

Source

SANAD Trial Group, University of Liverpool. Data modified from the joineR package.

References

Marson AG, Appleton R, Baker GA, et al. A randomised controlled trial examining longer-term outcomes of standard versus new antiepileptic drugs. The SANAD Trial. Health Tech Assess. 2007; 11(37).

Marson AG, Al-Kharusi AM, Alwaidh M, et al. The SANAD study of effectiveness of carbamazepine, gabapentin, lamotrigine, oxcarbazepine, or topiramate for treatment of partial epilepsy: an unblinded randomised controlled trial. Lancet. 2007; 365: 2007-2013.

Williamson PR, Kolamunnage-Dona R, Philipson P, Marson AG. Joint modelling of longitudinal and competing risks data. Stats Med. 2008; 27(30): 6426-6438.

See Also

joineR::epileptic, which this dataset was modified from.

Fits the specified longitudinal model for time-varying covariates up to the landmark times

Description

Fits the specified longitudinal model for time-varying covariates up to the landmark times

Usage

fit_longitudinal(
  x,
  landmarks,
  method,
  formula,
  dynamic_covariates,
  validation_fold = 0,
  cores = getOption("Ncpus", 1L),
  .warn_when_prop_few_obs = 0.25,
  ...
)

Arguments

Details

Parallel processing

As the longitudinal model for each landmark time is independent of the longitudinal models for other landmark times, parallel processing can be used to vastly speed up computation. However, due to issues with parallel processing in R, currently only Unix-like operating systems are supported by landmaRk.

Value

An object of class LandmarkAnalysis.

See Also

lcmm::hlme() and lme4::lmer() for additional arguments.

Fits the specified longitudinal model for time-varying covariates up to the landmark times

Description

Fits the specified longitudinal model for time-varying covariates up to the landmark times

Usage

## S4 method for signature 'LandmarkAnalysis'
fit_longitudinal(
  x,
  landmarks,
  method,
  formula,
  dynamic_covariates,
  validation_fold = 0,
  cores = getOption("Ncpus", 1L),
  .warn_when_prop_few_obs = 0.25,
  ...
)

Arguments

Details

Parallel processing

As the longitudinal model for each landmark time is independent of the longitudinal models for other landmark times, parallel processing can be used to vastly speed up computation. However, due to issues with parallel processing in R, currently only Unix-like operating systems are supported by landmaRk.

Value

An object of class LandmarkAnalysis.

See Also

lcmm::hlme() and lme4::lmer() for additional arguments.

Fits the specified survival model at the landmark times and up to the horizon times specified by the user

Description

Fits the specified survival model at the landmark times and up to the horizon times specified by the user

Usage

fit_survival(
  x,
  formula,
  landmarks,
  horizons,
  method,
  dynamic_covariates = c(),
  include_clusters = FALSE,
  censor_at_horizon = FALSE,
  validation_fold = 0
)

Arguments

Details

Mathematical formulation

This function estimates the conditional probability of survival to horizon s+w, conditioned on having survived to the landmark time, s, that is

\pi_i(s+w \vert s) = P(T_i > s+w \vert T_i \ge s, \bar{x}_i(s)),

where i denotes an individual's index, T_i is the time to event outcome for individual i and \bar{x}_i(s) are the covariates observed for individual i, including the observed history of dynamic covariates.

Value

An object of class LandmarkAnalysis.

Fits the specified survival model at the landmark times and up to the horizon times specified by the user

Description

Fits the specified survival model at the landmark times and up to the horizon times specified by the user

Usage

## S4 method for signature 'LandmarkAnalysis'
fit_survival(
  x,
  formula,
  landmarks,
  horizons,
  method,
  dynamic_covariates = c(),
  include_clusters = FALSE,
  censor_at_horizon = FALSE,
  validation_fold = 0
)

Arguments

Details

Mathematical formulation

This function estimates the conditional probability of survival to horizon s+w, conditioned on having survived to the landmark time, s, that is

\pi_i(s+w \vert s) = P(T_i > s+w \vert T_i \ge s, \bar{x}_i(s)),

where i denotes an individual's index, T_i is the time to event outcome for individual i and \bar{x}_i(s) are the covariates observed for individual i, including the observed history of dynamic covariates.

Value

An object of class LandmarkAnalysis.

Performance metrics

Description

Computes concordance index (c-index) and Brier scores at the specified landmark times and prediction horizons.

Usage

performance_metrics(
  x,
  landmarks,
  horizons,
  c_index = TRUE,
  brier = TRUE,
  auc_t = FALSE,
  train = TRUE,
  h_times = c()
)

Arguments

Value

Data frame with performance metrics across the specified landmark times and prediction horizons.

Performance metrics

Description

Computes concordance index (c-index) and Brier scores at the specified landmark times and prediction horizons.

Usage

## S4 method for signature 'LandmarkAnalysis'
performance_metrics(
  x,
  landmarks,
  horizons,
  c_index = TRUE,
  brier = TRUE,
  auc_t = FALSE,
  train = TRUE,
  h_times = c()
)

Arguments

Value

Data frame with performance metrics across the specified landmark times and prediction horizons.

Plot longitudinal observations and predicted survival curve for one individual

Description

Produces a single-panel plot with a common time axis. The left of the landmark dashed line shows the individual's observed longitudinal measurements; the right shows their predicted survival curve. The summary value at the landmark that feeds into the survival sub-model is highlighted.

Usage

## S4 method for signature 'LandmarkAnalysis'
plot(x, id, landmark, dynamic_covariate, horizon = NULL, train = TRUE, ...)

Arguments

Value

A ggplot object.

Make predictions for time-varying covariates at specified landmark times

Description

Make predictions for time-varying covariates at specified landmark times

Usage

predict_longitudinal(
  x,
  landmarks,
  method,
  dynamic_covariates,
  validation_fold = 0,
  ...
)

Arguments

Value

An object of class LandmarkAnalysis.

Make predictions for time-varying covariates at specified landmark times

Description

Make predictions for time-varying covariates at specified landmark times

Usage

## S4 method for signature 'LandmarkAnalysis'
predict_longitudinal(
  x,
  landmarks,
  method,
  dynamic_covariates,
  validation_fold = 0,
  ...
)

Arguments

Value

An object of class LandmarkAnalysis.

Make predictions for time-to-event outcomes at specified horizon times

Description

Make predictions for time-to-event outcomes at specified horizon times

Usage

predict_survival(
  x,
  landmarks,
  horizons,
  method = "survfit",
  dynamic_covariates = c(),
  include_clusters = FALSE,
  censor_at_horizon = FALSE,
  validation_fold = 0,
  ...
)

Arguments

Value

An object of class LandmarkAnalysis.

Make predictions for time-to-event outcomes at specified horizon times

Description

Make predictions for time-to-event outcomes at specified horizon times

Usage

## S4 method for signature 'LandmarkAnalysis'
predict_survival(
  x,
  landmarks,
  horizons,
  method = "survfit",
  dynamic_covariates = c(),
  include_clusters = FALSE,
  censor_at_horizon = FALSE,
  validation_fold = 0,
  ...
)

Arguments

Value

An object of class LandmarkAnalysis.

Prunes a landmark time from a LandmarkAnalysis, removing the risk set, longitudinal submodel and survival submodel from the object.

Description

Prunes a landmark time from a LandmarkAnalysis, removing the risk set, longitudinal submodel and survival submodel from the object.

Usage

prune(x, ...)

Arguments

Value

An object of class LandmarkAnalysis.

Prunes a landmark time from a LandmarkAnalysis, removing the risk set, longitudinal submodel and survival submodel from the object.

Description

Prunes a landmark time from a LandmarkAnalysis, removing the risk set, longitudinal submodel and survival submodel from the object.

Usage

## S4 method for signature 'LandmarkAnalysis'
prune(x, landmark = NULL, ...)

Arguments

Value

An object of class LandmarkAnalysis.

Prune a set of individuals from a risk set

Description

Prune a set of individuals from a risk set

Usage

prune_risk_sets(x, landmark, individuals)

Arguments

Value

An object of class LandmarkAnalysis after having pruned the individuals indicated in individuals from the risk set at landmark time landmark.

Prune a set of individuals from a risk set

Description

Prune a set of individuals from a risk set

Usage

## S4 method for signature 'LandmarkAnalysis'
prune_risk_sets(x, landmark, individuals)

Arguments

Value

An object of class LandmarkAnalysis after having pruned the individuals indicated in individuals from the risk set at landmark time landmark.

Displays an object of class "LandmarkAnalysis"

Description

Displays an object of class "LandmarkAnalysis"

Usage

## S4 method for signature 'LandmarkAnalysis'
show(object)

Arguments

Value

No return value(prints a summary to the console).

Examples

data(epileptic)
epileptic_dfs <- split_wide_df(
  epileptic,
  ids = "id", times = "time",
  static = c("with.time", "with.status", "treat", "age", "gender", "learn.dis"),
  dynamic = c("dose"),
  measurement_name = "value"
)
x <- LandmarkAnalysis(
  data_static = epileptic_dfs$df_static,
  data_dynamic = epileptic_dfs$df_dynamic,
  event_indicator = "with.status",
  ids = "id", event_time = "with.time",
  times = "time", measurements = "value"
) |>
  compute_risk_sets(365.25)
show(x)

Split a wide dataframe containing static and dynamic covariates and splits in into a dataframe with the static covariates and a list of dataframes, each associated to a dynamic covariate.

Description

Split a wide dataframe containing static and dynamic covariates and splits in into a dataframe with the static covariates and a list of dataframes, each associated to a dynamic covariate.

Usage

split_wide_df(df, ids, times, static, dynamic, measurement_name)

Arguments

Value

A data frame with the static covariates, and a list of data frames, one per dynamic covariate.

Summarise a LandmarkAnalysis object

Description

Summarise a LandmarkAnalysis object

Usage

## S4 method for signature 'LandmarkAnalysis'
summary(
  object,
  type = c("longitudinal", "survival"),
  landmark,
  horizon = NULL,
  dynamic_covariate = NULL
)

Arguments

Value

A summary of the desired submodel, printed to the console. Returns NULL invisibly.

Examples

data(epileptic)
epileptic_dfs <- split_wide_df(
  epileptic,
  ids = "id", times = "time",
  static = c("with.time", "with.status", "treat", "age", "gender", "learn.dis"),
  dynamic = c("dose"),
  measurement_name = "value"
)
x <- LandmarkAnalysis(
  data_static = epileptic_dfs$df_static,
  data_dynamic = epileptic_dfs$df_dynamic,
  event_indicator = "with.status",
  ids = "id", event_time = "with.time",
  times = "time", measurements = "value"
) |>
  compute_risk_sets(365.25) |>
  fit_survival(
    formula = survival::Surv(event_time, event_status) ~ treat + age,
    landmarks = 365.25,
    horizons = 2 * 365.25,
    method = "coxph"
  )
summary(x, type = "survival", landmark = 365.25, horizon = 2 * 365.25)

# Full pipeline including longitudinal submodel
x2 <- LandmarkAnalysis(
  data_static = epileptic_dfs$df_static,
  data_dynamic = epileptic_dfs$df_dynamic,
  event_indicator = "with.status",
  ids = "id", event_time = "with.time",
  times = "time", measurements = "value"
) |>
  compute_risk_sets(365.25) |>
  fit_longitudinal(
    landmarks = 365.25,
    method = "lme4",
    formula = value ~ treat + age + gender + learn.dis + (time | id),
    dynamic_covariates = c("dose")
  ) |>
  predict_longitudinal(
    landmarks = 365.25,
    method = "lme4",
    allow.new.levels = TRUE,
    dynamic_covariates = c("dose")
  ) |>
  fit_survival(
    formula = survival::Surv(event_time, event_status) ~
      treat + age + gender + learn.dis + dose,
    landmarks = 365.25,
    horizons = 2 * 365.25,
    method = "coxph",
    dynamic_covariates = c("dose")
  ) |>
  predict_survival(landmarks = 365.25, horizons = 2 * 365.25)
summary(x2, type = "longitudinal", landmark = 365.25, dynamic_covariate = "dose")
summary(x2, type = "survival", landmark = 365.25, horizon = 2 * 365.25)