Package {TTE}

Type:	Package
Title:	Design and Analysis Tools for Target Trial Emulation
Version:	1.1.1
Date:	2026-07-23
Maintainer:	Hisashi Noma <noma@ism.ac.jp>
Description:	Design and analysis tools for target trial emulation using longitudinal observational data. Functions are provided for checking person-period data, expanding longitudinal data into sequentially nested trials, estimating inverse probability weights for intention-to-treat and per-protocol analyses, and assessing weight distributions and covariate balance. Additional functions fit weighted pooled discrete-time outcome models, obtain standardized risks and treatment contrasts, and estimate weighted Kaplan-Meier and Aalen-Johansen curves. Two worked examples based on fully synthetic data illustrate an active-comparator new-user study comparing sodium-glucose cotransporter 2 inhibitors with dipeptidyl peptidase-4 inhibitors and an analysis of sequentially nested trials comparing angiotensin receptor blocker and calcium channel blocker strategies.
Depends:	R (≥ 4.1.0)
Imports:	graphics, nnet, sandwich, stats, utils
Suggests:	testthat (≥ 3.0.0)
License:	GPL-3
Encoding:	UTF-8
Config/testthat/edition:	3
NeedsCompilation:	no
Packaged:	2026-07-23 07:39:04 UTC; nomah
Author:	Hisashi Noma [aut, cre]
Repository:	CRAN
Date/Publication:	2026-08-02 16:30:13 UTC

Design and Analysis Tools for Target Trial Emulation

Description

Design and analysis tools for target trial emulation using longitudinal observational data.

Details

The main workflow comprises the following steps:

• check person-period data with check_tte();

• expand longitudinal data into sequentially nested trials with seqdesign_tte();

• estimate and combine inverse probability weights with est_wt() and combine_wt();

• assess covariate balance and diagnose weight distributions with balance_wt() and diagnose_wt();

• fit weighted pooled discrete-time outcome models with discsurvreg();

• obtain standardized risks and treatment contrasts with std_tte(); and

• estimate and plot weighted Kaplan-Meier and Aalen-Johansen curves with curve_tte().

The package includes two pairs of fully synthetic datasets:

• SGLT2 and SGLT2_baseline; and

• ARB and ARB_baseline.

No records from the motivating studies are included.

Author(s)

Hisashi Noma

References

Cashin, A. G., Hansford, H. J., Hernán, M. A., Swanson, S. A., Lee, H., Jones, M. D., Dahabreh, I. J., Dickerman, B. A., Egger, M., García-Albéniz, X., Golub, R. M., Islam, N., Lodi, S., Moreno-Betancur, M., Pearson, S. A., Schneeweiss, S., Sharp, M. K., Sterne, J. A. C., Stuart, E. A., and McAuley, J. H. (2025). Transparent reporting of observational studies emulating a target trial: the TARGET Statement. BMJ 390, e087179.

Hernán, M. A., Dahabreh, I. J., Dickerman, B. A., and Swanson, S. A. (2025). The target trial framework for causal inference from observational data: why and when is it helpful? Annals of Internal Medicine 178, 402–407.

Hernán, M. A., Wang, W., and Leaf, D. E. (2022). Target trial emulation: a framework for causal inference from observational data. JAMA 328, 2446–2447.

Noma, H., Goto, A., Sugimoto, T., Sunada, H., Oda, F., Maeda, M., and Fukuda, H. (2026). Real-world effectiveness of SGLT2 inhibitors in adults aged 75 years or older: a target trial emulation. Age and Ageing, in press.

Noma, H., Kurita, N., Fukuma, S., Fujisawa, T., Oda, F., Maeda, M., and Fukuda, H. (2026). Heart failure and renal outcomes with angiotensin receptor blockers compared with calcium channel blockers in patients with chronic kidney disease: a target trial emulation. Heart. doi:10.1136/heartjnl-2026-328193.

Examples

library(TTE)
data(SGLT2)

fit <- discsurvreg(
  Y_death ~ A + splines::ns(time, df = 3) + trial_period,
  data = SGLT2,
  id = id,
  weights = w_itt,
  var_method = "standard"
)
fit

Synthetic ARB versus CCB Sequential-Trial Example

Description

Fully synthetic baseline and person-month data representing sequentially nested new-user trials comparing angiotensin receptor blocker (ARB) and calcium channel blocker (CCB) strategies among people with chronic kidney disease. Heart failure hospitalization is the primary endpoint, and death is a competing event.

Usage

data(ARB_baseline)
data(ARB)

Format

ARB_baseline is a data frame with 900 person-trial entries and the following variables:

id

Original individual identifier.

trial

Identifier for the sequential trial in which the individual entered.

A

Binary treatment-strategy indicator: 0 for CCB and 1 for ARB.

treatment

Treatment-strategy label: "CCB" or "ARB".

age

Age at trial baseline, in years.

female

Indicator for female sex: 0 for no and 1 for yes.

bmi

Body mass index at trial baseline, in kg/m^2.

sbp

Systolic blood pressure at trial baseline, in mm Hg.

dbp

Diastolic blood pressure at trial baseline, in mm Hg.

egfr

Estimated glomerular filtration rate at trial baseline, in mL/min/1.73 m^2.

proteinuria

Indicator for proteinuria at trial baseline: 0 for no and 1 for yes.

diabetes

Indicator for diabetes at trial baseline: 0 for no and 1 for yes.

prior_heart_failure

Indicator for a history of heart failure at trial baseline: 0 for no and 1 for yes.

prior_stroke

Indicator for a history of stroke at trial baseline: 0 for no and 1 for yes.

trial_period

Calendar-period category for the trial baseline.

ps_oracle

Probability of ARB initiation generated by the simulation treatment model.

w_iptw

Stabilized baseline inverse probability of treatment weight.

ARB is a person-month data frame containing the variables in ARB_baseline, except ps_oracle, together with the following variables:

time

Zero-based follow-up-month index; time = 0 denotes the first month after trial baseline.

Y_hf

Indicator of heart failure hospitalization during the interval: 0 for no and 1 for yes.

Y_death

Indicator of death during the interval: 0 for no and 1 for yes.

event_code

Competing-event code: 0 for no event, 1 for heart failure hospitalization, and 2 for death.

stay_ltfu

Interval-specific indicator of remaining uncensored with respect to loss to follow-up: 0 for no and 1 for yes.

stay_adherent

Interval-specific indicator of adherence to the baseline treatment strategy: 0 for no and 1 for yes.

pp_at_risk

Indicator that the person-trial remains in the per-protocol risk set: 0 for no and 1 for yes.

w_itt

Combined analysis weight for the intention-to-treat analysis.

w_pp

Combined analysis weight for the per-protocol analysis.

In ARB, sbp, dbp, and egfr are updated over follow-up; the other clinical covariates are carried forward from trial baseline.

Details

The datasets were independently simulated and include no participant records from the motivating study. They are intended solely for education, software testing, and reproducible examples; numerical results obtained from them have no clinical interpretation.

References

Noma, H., Kurita, N., Fukuma, S., Fujisawa, T., Oda, F., Maeda, M., and Fukuda, H. (2026). Heart failure and renal outcomes with angiotensin receptor blockers compared with calcium channel blockers in patients with chronic kidney disease: a target trial emulation. Heart. doi:10.1136/heartjnl-2026-328193.

Examples

data(ARB_baseline)
data(ARB)
head(ARB_baseline)
table(ARB$event_code)

Synthetic SGLT2i versus DPP-4i Target-Trial Example

Description

Fully synthetic baseline and person-month data representing an active-comparator new-user target trial emulation comparing sodium-glucose cotransporter 2 inhibitors (SGLT2i) with dipeptidyl peptidase-4 inhibitors (DPP-4i) among older adults with type 2 diabetes. The primary endpoint is all-cause death after a one-month induction period.

Usage

data(SGLT2_baseline)
data(SGLT2)

Format

SGLT2_baseline is a data frame with 700 trial entries and the following variables:

id

Original individual identifier.

trial

Trial identifier.

A

Binary treatment-strategy indicator: 0 for DPP-4i and 1 for SGLT2i.

treatment

Treatment-strategy label: "DPP-4i" or "SGLT2i".

age

Age at trial baseline, in years.

female

Indicator for female sex: 0 for no and 1 for yes.

bmi

Body mass index at trial baseline, in kg/m^2.

hba1c

Glycated hemoglobin at trial baseline, in percent.

egfr

Estimated glomerular filtration rate at trial baseline, in mL/min/1.73 m^2.

proteinuria

Indicator for proteinuria at trial baseline: 0 for no and 1 for yes.

prior_heart_failure

Indicator for a history of heart failure at trial baseline: 0 for no and 1 for yes.

prior_stroke

Indicator for a history of stroke at trial baseline: 0 for no and 1 for yes.

recent_hospitalization

Indicator for a recent hospitalization at trial baseline: 0 for no and 1 for yes.

trial_period

Calendar-period category for the trial baseline.

ps_oracle

Probability of SGLT2i initiation generated by the simulation treatment model.

w_iptw

Stabilized baseline inverse probability of treatment weight.

SGLT2 is a person-month data frame containing the variables in SGLT2_baseline, except ps_oracle, together with the following variables:

time

Zero-based follow-up-month index; time = 0 denotes the first month after trial baseline.

Y_death

Indicator of death during the interval: 0 for no and 1 for yes.

event_code

Event code: 0 for no event and 1 for death.

stay_ltfu

Interval-specific indicator of remaining uncensored with respect to loss to follow-up: 0 for no and 1 for yes.

stay_adherent

Interval-specific indicator of adherence to the baseline treatment strategy: 0 for no and 1 for yes.

pp_at_risk

Indicator that the trial entry remains in the per-protocol risk set: 0 for no and 1 for yes.

w_itt

Combined analysis weight for the intention-to-treat analysis.

w_pp

Combined analysis weight for the per-protocol analysis.

The baseline covariates are repeated across follow-up months in SGLT2.

Details

The datasets were independently simulated and include no participant records from the motivating study. They are intended solely for education, software testing, and reproducible examples; numerical results obtained from them have no clinical interpretation.

References

Noma, H., Goto, A., Sugimoto, T., Sunada, H., Oda, F., Maeda, M., and Fukuda, H. (2026). Real-world effectiveness of SGLT2 inhibitors in adults aged 75 years or older: a target trial emulation. Age and Ageing, in press.

Examples

data(SGLT2_baseline)
data(SGLT2)
head(SGLT2_baseline)
head(SGLT2)

Assess Covariate Balance Before and After Weighting

Description

Computes unweighted and weighted means, standard deviations, and standardized mean differences using externally supplied analysis weights.

Usage

balance_wt(formula, data, weights = NULL, absolute = TRUE)

Arguments

Value

An object of class balance_wt containing unweighted and weighted covariate summaries and standardized mean differences. A summary() method provides a formatted balance table.

See Also

est_wt, diagnose_wt

Examples

data(SGLT2_baseline)
wt <- est_wt(
  A ~ age + female + bmi + hba1c + egfr + prior_heart_failure +
    prior_stroke + recent_hospitalization + trial_period,
  data = SGLT2_baseline, type = "treatment"
)
bal <- balance_wt(
  A ~ age + female + bmi + hba1c + egfr + prior_heart_failure +
    prior_stroke + recent_hospitalization + trial_period,
  data = SGLT2_baseline, weights = wt
)
summary(bal)

Individual-Cluster Bootstrap

Description

Resamples original individuals while retaining all person-period rows and repeated trial entries belonging to each sampled individual.

Usage

boot_tte(data, id, statistic, R = 500L,
  conf_level = 0.95, seed = NULL, ...)

Arguments

Details

Sampling is performed at the level of the original individual. Consequently, all follow-up rows and all sequential-trial entries associated with a sampled individual are retained together.

Value

An object of class boot_tte containing the estimate from the original data, bootstrap replicates, and percentile confidence intervals.

Examples

data(SGLT2)
small <- subset(SGLT2, id <= 80 & time < 24)
stat_fun <- function(d) {
  fit <- discsurvreg(
    Y_death ~ A + time, data = d,
    id = id, weights = w_itt,
    var_method = "standard"
  )
  c(log_hr = unname(coef(fit)["A"]))
}
b <- boot_tte(small, id = id, statistic = stat_fun, R = 10, seed = 1)
b

Check a Target-Trial Person-Period Dataset

Description

Detects duplicate intervals, time-ordering problems, rows occurring after an event, within-trial treatment changes, and invalid analysis weights.

Usage

check_tte(data, id, time, trial = NULL, event = NULL,
  treatment = NULL, weights = NULL, expected_start = 0,
  allow_gaps = FALSE)

Arguments

Value

An object of class check_tte containing the detected data-quality problems and the rows or person-trials affected by each problem.

See Also

seqdesign_tte

Examples

data(ARB)
chk <- check_tte(
  ARB,
  id = id, trial = trial, time = time,
  event = event_code, treatment = A, weights = w_itt
)
chk

Combine Inverse Probability Weight Components

Description

Multiplies treatment, loss-to-follow-up, and adherence weight components and optionally truncates and normalizes the resulting analysis weights.

Usage

combine_wt(..., truncate = c(0.01, 0.99),
  normalize = c("none", "mean1", "sum_n"))

Arguments

Value

An object of class combine_wt containing the combined weights and metadata describing truncation and normalization. Use weights() to extract the numeric weight vector.

See Also

est_wt, diagnose_wt

Examples

data(ARB)
w <- combine_wt(
  ARB$w_iptw,
  ARB$w_itt / ARB$w_iptw,
  normalize = "mean1"
)
summary(weights(w))

Estimate Weighted Kaplan-Meier or Aalen-Johansen Curves

Description

Estimates weighted survival or cumulative-incidence curves from person-period risk sets. Curves are represented as right-continuous step functions.

Usage

curve_tte(data, time, event, treatment, weights = NULL,
  type = c("km", "aj"), cause = 1, id = NULL, trial = NULL,
  labels = NULL, bootstrap = 0L, conf_level = 0.95,
  seed = NULL)

Arguments

Details

The time variable indexes follow-up intervals, whereas summary and plotting times are expressed as elapsed follow-up time. For example, interval indices 0, ..., 59 represent the 60 intervals ending at elapsed times 1, ..., 60. Thus, summary(x, time = 60) reports the estimate after 60 follow-up intervals.

Value

An object of class curve_tte containing treatment-specific curve estimates and, when requested, bootstrap confidence intervals. A summary() method returns estimates at specified elapsed follow-up times.

See Also

std_tte, boot_tte

Examples

data(SGLT2)
km <- curve_tte(
  SGLT2, time = time, event = Y_death,
  treatment = treatment, weights = w_itt,
  type = "km", id = id, trial = trial
)
summary(km, time = 60)


data(ARB)
aj <- curve_tte(
  ARB, time = time, event = event_code,
  treatment = treatment, weights = w_itt,
  type = "aj", cause = 1, id = id, trial = trial
)
summary(aj, time = 60)

Diagnose Inverse Probability Weights

Description

Summarizes weight distributions and effective sample sizes. When analysis data are supplied, treatment-group and risk-set diagnostics are also produced.

Usage

diagnose_wt(weights, data = NULL, treatment = NULL, time = NULL,
  id = NULL, trial = NULL, extreme = c(0.01, 0.99))

Arguments

Value

An object of class diagnose_wt containing distributional summaries, effective sample sizes, and any requested treatment-group or risk-set diagnostics.

See Also

est_wt, combine_wt, balance_wt

Examples

data(ARB)
diag <- diagnose_wt(
  w_itt, data = ARB,
  treatment = A, time = time, id = id, trial = trial
)
diag

Fit a Weighted Pooled Discrete-Time Outcome Model

Description

Fits a weighted pooled generalized linear model and computes cluster-robust inference with clustering at the level of the original individual.

Usage

discsurvreg(formula, data, id, weights = NULL,
  family = stats::quasibinomial(link = "cloglog"),
  var_method = c("standard", "MBN"), eform = TRUE,
  conf_level = 0.95)

Arguments

Value

An object of class discsurvreg containing the fitted glm object, the cluster-robust covariance matrix, model coefficients, and analysis settings. Methods for coef(), vcov(), and confint() are available.

References

Morel, J. G., Bokossa, M. C., and Neerchal, N. K. (2003). Small sample correction for the variance of GEE estimators. Biometrical Journal 45, 395–409. doi:10.1002/bimj.200390021.

See Also

std_tte

Examples

data(SGLT2)
fit <- discsurvreg(
  Y_death ~ A + splines::ns(time, df = 3) + trial_period,
  data = SGLT2,
  id = id, weights = w_itt,
  family = stats::quasibinomial(link = "cloglog"),
  var_method = "standard"
)
fit
confint(fit, parm = "A", eform = TRUE)

Estimate Inverse Probability Weights

Description

Estimates baseline treatment weights or cumulative censoring and adherence weights while retaining intermediate predicted probabilities and interval-specific weight factors.

Usage

est_wt(formula, data, numerator = NULL,
  type = c("treatment", "censoring", "adherence"),
  id = NULL, trial = NULL, time = NULL, cumulative = NULL,
  lag = NULL, stabilize = TRUE, truncate = c(0.01, 0.99),
  eps = 1e-06)

Arguments

Details

For type = "treatment", the function estimates a baseline inverse probability of treatment weight. For type = "censoring" or type = "adherence", interval-specific factors may be multiplied within person-trial to obtain cumulative weights. Weight truncation is an analysis choice; the default applies truncation at the 1st and 99th percentiles.

Value

An object of class est_wt containing the estimated weight vector, predicted numerator and denominator probabilities, interval-specific factors, and settings used in weight construction. The numeric weights can be extracted with weights(), and a summary() method provides diagnostics.

See Also

combine_wt, diagnose_wt, balance_wt

Examples

data(SGLT2_baseline)
wt <- est_wt(
  A ~ age + female + bmi + hba1c + egfr +
    prior_heart_failure + prior_stroke +
    recent_hospitalization + trial_period,
  data = SGLT2_baseline,
  type = "treatment"
)
summary(wt)


data(ARB)
wc <- est_wt(
  stay_ltfu ~ A + time + I(time^2) + age + female + egfr +
    prior_heart_failure,
  numerator = stay_ltfu ~ A + time + I(time^2),
  data = ARB,
  type = "censoring", id = id, trial = trial, time = time,
  cumulative = TRUE, lag = 1
)
summary(wc)

Expand Longitudinal Data into Sequentially Nested Trials

Description

Creates a trial entry at each eligible treatment decision time and expands each entry into follow-up person-period records.

Usage

seqdesign_tte(data, id, calendar_time, eligible, treatment,
  outcome, censor = NULL, max_follow = Inf, induction = 0L,
  keep = names(data))

Arguments

Details

An individual may enter more than one sequential trial when eligible at multiple treatment decision times. Follow-up for each trial entry is obtained from subsequent rows of the original longitudinal data and ends according to the supplied outcome, censoring, and maximum-follow-up definitions.

Value

A person-period data frame with added trial, time, A, and Y variables, together with the retained source variables and indicators used to define follow-up.

See Also

check_tte

Examples

data(SGLT2)
source_data <- subset(SGLT2, id <= 8)
source_data$eligible <- source_data$time == 0
expanded <- seqdesign_tte(
  source_data, id = id, calendar_time = time,
  eligible = eligible, treatment = A, outcome = Y_death,
  max_follow = 24, induction = 1
)
head(expanded)

Standardize a Discrete-Time Model to Marginal Risks

Description

Creates counterfactual copies of a target population, predicts interval-specific event probabilities, and obtains standardized survival, risk, cumulative incidence, risk differences, risk ratios, and time-specific numbers needed to treat or harm.

Usage

std_tte(fit, data, treatment, time, times,
  values = NULL, labels = NULL, competing_fit = NULL,
  target_weights = NULL)

Arguments

Details

The values supplied in times index follow-up intervals, whereas summary horizons are expressed as elapsed follow-up time. For example, times = 0:59 represents the 60 intervals ending at elapsed times 1 through 60, so summary(x, horizon = 60) reports standardized estimates after 60 follow-up intervals.

When competing_fit is omitted, risks are obtained from the fitted event model. When a competing-event model is supplied, the function combines the event-specific probabilities to obtain standardized cumulative incidence.

Value

An object of class std_tte containing strategy-specific standardized curves and treatment contrasts. A summary() method returns estimates at a specified elapsed follow-up horizon.

See Also

discsurvreg, curve_tte

Examples

data(SGLT2)
data(SGLT2_baseline)
fit <- discsurvreg(
  Y_death ~ A + splines::ns(time, df = 3) + trial_period,
  data = SGLT2, id = id, weights = w_itt,
  var_method = "standard"
)
std <- std_tte(
  fit, data = SGLT2_baseline,
  treatment = A, time = time, times = 0:59,
  labels = c("DPP-4i", "SGLT2i")
)
summary(std, horizon = 60)

# Competing-risk example
data(ARB)
data(ARB_baseline)
fit_hf <- discsurvreg(
  Y_hf ~ A + splines::ns(time, df = 3) + trial_period,
  data = ARB, id = id, weights = w_itt,
  var_method = "standard"
)
fit_death <- discsurvreg(
  Y_death ~ A + splines::ns(time, df = 3) + trial_period,
  data = ARB, id = id, weights = w_itt,
  var_method = "standard"
)
std_hf <- std_tte(
  fit_hf, data = ARB_baseline,
  treatment = A, time = time, times = 0:59,
  competing_fit = fit_death, labels = c("CCB", "ARB")
)
summary(std_hf, horizon = 60)