On estimating the time to statistical cure.

BACKGROUND: The mortality risk among cancer patients measured from the time of diagnosis is often elevated in comparison to the general population. However, for some cancer types, the patient mortality risk will over time reach the same level as the general population mortality risk. The time point at which the mortality risk reaches the same level as the general population is called the cure point and is of great interest to patients, clinicians, and health care planners. In previous studies, estimation of the cure point has been handled in an ad hoc fashion, often without considerations about margins of clinical relevance. METHODS: We review existing methods for estimating the cure point and discuss new clinically relevant measures for quantifying the mortality difference between cancer patients and the general population, which can be used for cure point estimation. The performance of the methods is assessed in a simulation study and the methods are illustrated on survival data from Danish colon cancer patients. RESULTS: The simulations revealed that the bias of the estimated cure point depends on the measure chosen for quantifying the excess mortality, the chosen margin of clinical relevance, and the applied estimation procedure. These choices are interdependent as the choice of mortality measure depends both on the ability to define a margin of clinical relevance and the ability to accurately compute the mortality measure. The analysis of cancer survival data demonstrates the importance of considering the confidence interval of the estimated cure point, as these may be wide in some scenarios limiting the applicability of the estimated cure point. CONCLUSIONS: Although cure points are appealing in a clinical context and has widespread applicability, estimation remains a difficult task. The estimation relies on a number of choices, each associated with pitfalls that the practitioner should be aware of.

Risk of death, relapse or progression, and loss of life expectancy at different progression-free survival milestones in primary central nervous system lymphoma.

In this study, we analyzed the evolution of the prognosis of primary central nervous system lymphoma (PCNSL) patients as they reach selected progression-free survival (PFS) milestones after high-dose methotrexate (HD-MTX)-based therapy. In total, 258 and 146 patients were included from Denmark and British Columbia, respectively. All patients were diagnosed during 2000-2017. The 5-year PFS was 27% (95% CI 23; 32); however, for patients reaching 5 years of PFS, this increased to 71% (95% CI 57; 86). Within the first 5 years after diagnosis, patients lost 2.0 years (95% CI 1.8; 2.2) when compared to a similar background population. This reduced to 0.5 years (95% CI 0.2; 0.9) for patients reaching 5 years of PFS. Treatment with rituximab was associated with improved outcomes. The prognosis of patients with PCNSL treated with HD-MTX-based regimens in this cohort is poor, although it improves as patients survive without progression/relapse. However, survival does not conclusively normalize to that of a similar background population.

Estimating the loss of lifetime function using flexible parametric relative survival models.

BACKGROUND: Within cancer care, dynamic evaluations of the loss in expectation of life provides useful information to patients as well as physicians. The loss of lifetime function yields the conditional loss in expectation of life given survival up to a specific time point. Due to the inevitable censoring in time-to-event data, loss of lifetime estimation requires extrapolation of both the patient and general population survival function. In this context, the accuracy of different extrapolation approaches has not previously been evaluated. METHODS: The loss of lifetime function was computed by decomposing the all-cause survival function using the relative and general population survival function. To allow extrapolation, the relative survival function was fitted using existing parametric relative survival models. In addition, we introduced a novel mixture cure model suitable for extrapolation. The accuracy of the estimated loss of lifetime function using various extrapolation approaches was assessed in a simulation study and by data from the Danish Cancer Registry where complete follow-up was available. In addition, we illustrated the proposed methodology by analyzing recent data from the Danish Lymphoma Registry. RESULTS: No uniformly superior extrapolation method was found, but flexible parametric mixture cure models and flexible parametric relative survival models seemed to be suitable in various scenarios. CONCLUSION: Using extrapolation to estimate the loss of lifetime function requires careful consideration of the relative survival function outside the available follow-up period. We propose extensive sensitivity analyses when estimating the loss of lifetime function.

Optimizing Outcome Prediction in Diffuse Large B-Cell Lymphoma by Use of Machine Learning and Nationwide Lymphoma Registries: A Nordic Lymphoma Group Study.

PURPOSE: Prognostic models for diffuse large B-cell lymphoma (DLBCL), such as the International Prognostic Index (IPI) are widely used in clinical practice. The models are typically developed with simplicity in mind and thus do not exploit the full potential of detailed clinical data. This study investigated whether nationwide lymphoma registries containing clinical data and machine learning techniques could prove to be useful for building modern prognostic tools. PATIENTS AND METHODS: This study was based on nationwide lymphoma registries from Denmark and Sweden, which include large amounts of clinicopathologic data. Using the Danish DLBCL cohort, a stacking approach was used to build a new prognostic model that leverages the strengths of different survival models. To compare the performance of the stacking approach with established prognostic models, cross-validation was used to estimate the concordance index (C-index), time-varying area under the curve, and integrated Brier score. Finally, the generalizability was tested by applying the new model to the Swedish cohort. RESULTS: In total, 2,759 and 2,414 patients were included from the Danish and Swedish cohorts, respectively. In the Danish cohort, the stacking approach led to the lowest integrated Brier score, indicating that the survival curves obtained from the stacking model fitted the observed survival the best. The C-index and time-varying area under the curve indicated that the stacked model (C-index: Denmark [DK], 0.756; Sweden [SE], 0.744) had good discriminative capabilities compared with the other considered prognostic models (IPI: DK, 0.662; SE, 0.661; and National Comprehensive Cancer Network-IPI: DK, 0.681; SE, 0.681). Furthermore, these results were reproducible in the independent Swedish cohort. CONCLUSION: A new prognostic model based on machine learning techniques was developed and was shown to significantly outperform established prognostic indices for DLBCL. The model is available at https://lymphomapredictor.org .