gcemod

Generalized competing event (GCE) modeling of omega+ — the ratio of the hazard for an event of interest to the hazard for a competing event — on the cause-specific (Cox) and subdistribution (Fine-Gray) scales, with confidence intervals and p-values from the Lunn-McNeil stacked-data approach.

What it does

1. Estimate covariate effects on omega+ for Cox and Fine-Gray models, with 95% CIs, Wald p-values, and an omnibus test from the Lunn-McNeil stacked-data approach (gcecox(), gcefg(), lunnmcneil()).
2. Build a risk score from the model linear predictor, and predict it and the absolute omega+ for new subjects (predict()).
3. Find cutpoints that maximize the omega+ difference between groups (gce_cutpoints(method = "optimal")).
4. Alligator plots — cumulative incidence of the event of interest and the competing event within risk groups (gce_alligator()).
5. Calibration plots — predicted vs. observed omega+ by rank, with a user-specified number of groups (gce_calibration()).
6. Compare effects across events — primary, competing, total (composite), and relative (omega+) hazard ratios in one table (gce_tidy(fit, effects = TRUE)).
7. Score new subjects relative to the average subject, and expose the covariate means/SDs behind the risk score (gce_riskscore(), gce_scaling()).
8. Cumulative-incidence-ratio GCE model — covariate effects on rho(t) = F1(t)/F2(t), the odds that a realized event is the event of interest, by pseudo-observation regression (gcecif()). Returns a gcemod object, so it works with gce_tidy(), gce_scaling(), and gce_riskscore() like the gcecox() / gcefg() fits.

Per-subject omega = omega+/(1+omega+) is kept as a descriptive quantity (predict(fit, type = "share")) so model-predicted values can be compared with observed ones. Discrimination is available via gce_performance() (cause-specific C-index).

Install and check (from source)

# one time: build the bundled example datasets
source("data-raw/make_sample.R")      # writes data/hn.rda and data/prostate.rda

# regenerate man/ and NAMESPACE from the roxygen headers
devtools::document()

devtools::check()

Quick start

library(gcemod)
data(hn)   # head-and-neck cohort: status 1 = recurrence, 2 = death w/o recurrence

fit <- gcecox(Surv(time, status) ~ age + smoker + t_cat + n_cat + p16,
              data = hn, t = 5)             # or gcefg(...) / gcecif(...)
fit                                         # short display
summary(fit)                                # coefficient table, CIs, omnibus test
coef(fit); confint(fit, exponentiate = TRUE)

predict(fit, newdata = hn[1:3, ], type = "ratio", times = c(2, 5))

gce_tidy(fit, effects = TRUE)   # primary / competing / total / relative HRs
cuts <- gce_cutpoints(fit, groups = 3)
cuts
plot(gce_alligator(fit, groups = cuts))
plot(gce_calibration(fit, which = "omegaplus", groups = 5))
plot(gce_omegatime(fit, which = "omegaplus", groups = cuts))

cif <- gcecif(Surv(time, status) ~ age + smoker + p16, data = hn, t = 5)
summary(cif)                                # cumulative-incidence-ratio scale

The matrix interface of earlier versions, gcecox(Time, Ind, Cov, t = ), remains available.

See vignette("gcemod") for the full workflow.

References

Lunn M, McNeil D (1995) Applying Cox regression to competing risks. Biometrics 51:524-32.

Carmona R, et al. (2014) Validated competing event model for the stage I-II endometrial cancer population. Int J Radiat Oncol Biol Phys 89:888-98.

Carmona R, et al. (2016) Improved method to stratify elderly patients with cancer at risk for competing events. J Clin Oncol 34:1270-77.

Mell LK, et al. (2019) Nomogram to predict the benefit of intensive treatment for locoregionally advanced head and neck cancer. Clin Cancer Res 25:7078-7088.

Zakeri K, et al. (2020) Predictive classifier for intensive treatment of head and neck cancer. Cancer 126:5263-5273.

Mell LK, et al. (2024) Effects of androgen deprivation therapy on prostate cancer outcomes according to competing event risk. Eur Urol 85:373-381.