Abstract
Inverse probability censoring weighting (IPCW) technique is often used to adjust for right censoring or recover information for censored individuals in survival analysis and in multi-state modeling. A simple IPCW (SIPCW) technique which does not consider the intermediate states, has been proposed for analyzing multi-state data. However, our simulation studies show that the SIPCW technique may lead to biased estimates when being applied in complex multi-state models. We thereby propose a model-specific, state-dependent adaptive IPCW (AIPCW) technique for estimating transition probabilities in multi-state models. Intensive simulation results verified that the proposed AIPCW technique improves the accuracy of transition probability estimates compared to the SIPCW technique and leads to asymptotic unbiased estimates. We applied the proposed technique to a real-world hematopoietic stem cell transplant (HSCT) data to assess the acute and chronic graft-versus-host disease (GVHD) effects on disease relapse rates and mortality rates.
Keywords
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsReferences
Aalen, O. (1978). Nonparametric inference for a family of counting processes. The Annals of Statistics, 6, 701–726.
Aalen, O., & Johansen, S. (1978). An empirical transition matrix for non-homogeneous Markov chains based on censored observations. Scandinavian Journal of Statistics, 5, 141–150.
Andersen, P. K., Borgan, O., Gill, R. D., & Keiding, N. (1993). Statistical models based on counting processes. New York: NY Springer.
Bie, O., Borgan, Ø., & Liestøl, K. (1987). Confidence intervals and confidence bands for the cumulative hazard rate function and their small sample properties. Scandinavian Journal of Statistics, 14, 221–233.
Borgan, Ø., & Liestøl, K. (1990). A note on confidence intervals and bands for the survival function based on transformations. Scandinavian Journal of Statistics, 17, 35–41.
Datta, S., & Satten, G. A. (2001). Validity of the Aalen–Johansen estimators of stage occupation probabilities and Nelson–Aalen estimators of integrated transition hazards for non-Markov models. Statistics & Probability Letters, 55(4), 403–411.
Datta, S., & Satten, G. A. (2002). Estimation of integrated transition hazards and stage occupation probabilities for non-Markov system under stage dependent censoring. Biometrics, 58, 792–802.
Fine, J. P., & Gray, R. J. (1999). A proportional hazards model for the subdistribution of a competing risk. Journal of the American Statistical Association, 94(446), 496–509.
He, P., Eriksson, F., Scheike, T. H., & Zhang, M.-J. (2015). A proportional hazards regression model for the subdistribution with covariates-adjusted censoring weight for competing risks data. Scandinavian Journal of Statistics, 43, 103–122.
Klein, J. P., Szydlo, R. M., Craddock, C., & Goldman, J. M. (2000). Estimation of current leukaemia-free survival following donor lymphocyte infusion therapy for patients with leukaemia who relapse after allografting: Application of a multistate model. Statistics in Medicine, 19(21), 3005–3016.
Luger, S. M., Ringdén, O., Zhang, M.-J., Pérez, W. S., Bishop, M. R., Bornhauser, M., et al. (2012). Similar outcomes using myeloablative vs reduced-intensity allogeneic transplant preparative regimens for AML or MDS. Bone Marrow Transplantation, 47(2), 203–211.
Nelson, W. (1969). Hazard plotting for incomplete failure data. Journal of Quality Technology, 1(1), 27–52.
Nelson, W. (1972). Theory and applications of hazard plotting for censored failure data. Technometrics, 14(4), 945–966.
Robins, J. M., & Rotnitzky, A. (1992). Recovery of information and adjustment for dependent censoring using surrogate markers. In AIDS epidemiology (pp. 297–331). Boston: Birkhäuser.
Scheike, T., & Zhang, M. J. (2007). Direct modelling of regression effects for transition probabilities in multistate models. Scandinavian Journal of Statistics, 34(1), 17–32.
Scheike, T. H., Zhang, M. J., & Gerds, T. A. (2008). Predicting cumulative incidence probability by direct binomial regression. Biometrika, 95(1), 205–220.
Zhang, X., Zhang, M. J., & Fine, J. P. (2009). A mass redistribution algorithm for right-censored and left-truncated time to event data. Journal of Statistical Planning and Inference, 139(9), 3329–3339.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Appendices
Appendix 1
Large sample property of IPCW-based estimator has been studied extensively [8, 9, 16]. We present a brief derivation for the variance estimations. Let \(T_{i,123}=(T_{i,123}^\ast \wedge C_i)\), \(N_i^{C_{00}}(t)=I(T_{i,123} \le t, \varDelta _{C_{00},i}=1)\), \(Y_i^{C_{00}}(t)=I(T_{i,123}\ge t)\), and \(Y_{\bullet }^{C_{00}}(t)=\sum _i Y_i^{C_{00}}(t)\). Under regularity condition [9], we have that
Thus, variance of \(\left \{\widehat P_{0k}^{\mathrm {AIPCW}}(0,t) - P_{0k}(0,t)\right \}\), for k = 0, 1, 2, can be estimated by
where
For the variance of \(P_{03}^{\mathrm {AIPCW}}(0,t)\), since
where \(M_i^{C_{33}}(u)\) can be estimated by
Then, variance of \(\left \{ \widehat {P}_{03}^{\mathrm {AIPCW}}(0,t)- P_{03}(0,t) \right \}\) can be estimated by
where
Similarly for k = 4, 5, variance of \(\left \{\widehat {P}_{0k}^{\mathrm {AIPCW}}(0,t)- P_{0k}(0,t) \right \}\) can be estimated by
where
The derived moment-type variance estimates have two parts. For example, there are two parts in \(\widehat {W}_{i,00}(t)=\widehat {W}_{i1,00}(t)+\widehat {W}_{i2,00}(t) \) for estimating variance of \(P_{00}^{\mathrm {AIPCW}}(0,t)\). The second part
is needed due to the estimation of the censoring distributions. In other words, when the censoring distributions are known, \(\widehat {W}_{i2,00}(t)\) becomes zero, and \(\widehat {W}_{i,00}(t)\) reduces to
In most situations, \(\widehat {W}_{i2,00}(t)\) is negligible. The variance estimates estimated from \(\widehat {W}_{i,00}(t)\) are close to that estimated from \(\widehat {W}_{i1,00}(t)\) alone, but not necessarily larger, because the covariance of W i1,00(t) and W i2,00(t) are not necessarily positive. Similar arguments are true for the variances of transition probabilities to other states. We presented both variance estimates in the simulation studies.
Appendix 2
For a four-level twelve-state model, we here only present the variance estimates of P 09(t), P 0,10(t), P 0,11(t), one can refer to Appendix 1 for the variance of other transition probabilities. We have
where \(M_i^{C_{99,s}}(u)\), for s = 3, 4 can be estimated by
Thus, the variance of \(\sqrt {n}\left \{\widehat {P}_{09}^{\mathrm {AIPCW}}(0,t)- P_{09}(0,t) \right \}\) can be estimated by
where
Similarly, for k = 10, 11, the variance of \(\sqrt {n}\left \{\widehat {P}_{0k}^{\mathrm {AIPCW}}(0,t)- P_{0k}(0,t) \right \}\) can be estimated by:
where
Rights and permissions
Copyright information
© 2020 Springer Nature Switzerland AG
About this chapter
Cite this chapter
Zhang, Y., Zhang, MJ. (2020). Inference of Transition Probabilities in Multi-State Models Using Adaptive Inverse Probability Censoring Weighting Technique. In: Zhao, Y., Chen, DG. (eds) Statistical Modeling in Biomedical Research. Emerging Topics in Statistics and Biostatistics . Springer, Cham. https://doi.org/10.1007/978-3-030-33416-1_19
Download citation
DOI: https://doi.org/10.1007/978-3-030-33416-1_19
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-33415-4
Online ISBN: 978-3-030-33416-1
eBook Packages: Mathematics and StatisticsMathematics and Statistics (R0)