ABSTRACT
Background: This study aimed to assess the performance of currently available risk calculators in a cohort of patients with malignant peripheral nerve sheath tumors (MPNST) and to create an MPNST-specific prognostic model including type-specific predictors for overall survival (OS). Methods: This is a retrospective multicenter cohort study of patients with MPNST from 11 secondary or tertiary centers in The Netherlands, Italy and the United States of America. All patients diagnosed with primary MPNST who underwent macroscopically complete surgical resection from 2000 to 2019 were included in this study. A multivariable Cox proportional hazard model for OS was estimated with prespecified predictors (age, grade, size, NF-1 status, triton status, depth, tumor location, and surgical margin). Model performance was assessed for the Sarculator and PERSARC calculators by examining discrimination (C-index) and calibration (calibration plots and observed-expected statistic; O/E-statistic). Internal-external cross-validation by different regions was performed to evaluate the generalizability of the model. Results: A total of 507 patients with primary MPNSTs were included from 11 centers in 7 regions. During follow-up (median 8.7 years), 211 patients died. The C-index was 0.60 (95% CI 0.53-0.67) for both Sarculator and PERSARC. The MPNST-specific model had a pooled C-index of 0.69 (95%CI 0.65-0.73) at validation, with adequate discrimination and calibration across regions. Conclusions: The MPNST-specific MONACO model can be used to predict 3-, 5-, and 10-year OS in patients with primary MPNST who underwent macroscopically complete surgical resection. Further validation may refine the model to inform patients and physicians on prognosis and support them in shared decision-making.
ABSTRACT
The use of individual participant data (IPD) from multiple studies is an increasingly popular approach when developing a multivariable risk prediction model. Corresponding datasets, however, typically differ in important aspects, such as baseline risk. This has driven the adoption of meta-analytical approaches for appropriately dealing with heterogeneity between study populations. Although these approaches provide an averaged prediction model across all studies, little guidance exists about how to apply or validate this model to new individuals or study populations outside the derivation data. We consider several approaches to develop a multivariable logistic regression model from an IPD meta-analysis (IPD-MA) with potential between-study heterogeneity. We also propose strategies for choosing a valid model intercept for when the model is to be validated or applied to new individuals or study populations. These strategies can be implemented by the IPD-MA developers or future model validators. Finally, we show how model generalizability can be evaluated when external validation data are lacking using internal-external cross-validation and extend our framework to count and time-to-event data. In an empirical evaluation, our results show how stratified estimation allows study-specific model intercepts, which can then inform the intercept to be used when applying the model in practice, even to a population not represented by included studies. In summary, our framework allows the development (through stratified estimation), implementation in new individuals (through focused intercept choice), and evaluation (through internal-external validation) of a single, integrated prediction model from an IPD-MA in order to achieve improved model performance and generalizability.
Subject(s)
Clinical Trials as Topic/methods , Data Interpretation, Statistical , Models, Statistical , Forecasting/methods , Humans , Meta-Analysis as Topic , Venous Thrombosis/diagnosisABSTRACT
A two-stage, numerical deconvolution approach was employed to develop level A in vitro-in vivo correlations using data for three formulations of an extended-release oral dosage form. The in vitro dissolution data for all formulations exhibited near-complete dissolution within the time frame of the test. The pharmacokinetic concentration-time profiles for 16 subjects in a cross-over study demonstrated notably limited bioavailability for the slowest formulation. These data were used as the basis for the IVIVC model development. Two models were identified that satisfied the nominal requirements for a conclusive internal predictability of the IVIVC, provided that all three formulations were used as internal datasets. These were a simple linear model with absorption cut-off and a piecewise-linear variable absorption scale model. A subsequent cross-validation of the models' robustness indicated that neither model predicted satisfactorily the pharmacokinetic characteristics of all formulations in a conclusive manner. The piecewise-linear variable absorption scale model provided the most accurate results, particularly with respect to the prediction of the slowest formulation's pharmacokinetic metrics. But this latter model also involved additional free parameters compared with the simple linear model with absorption cut-off. It is argued that more complex IVIVC models with extra parameterization require comprehensive validation to ascertain the accuracy and robustness of the model. In order to achieve this, it is necessary to ensure a complete suite of supporting datasets for internal and external validation, irrespective of the mathematical approach used subsequently to develop the IVIVC.