Reports de recerca
http://hdl.handle.net/2117/3389
Sat, 22 Oct 2016 07:33:17 GMT2016-10-22T07:33:17ZExtension of the asymptotic relative efficiency method to select the primary endpoint in a randomized clinical trial
http://hdl.handle.net/2117/26456
Extension of the asymptotic relative efficiency method to select the primary endpoint in a randomized clinical trial
Plana-Ripoll, Oleguer; Gómez Melis, Guadalupe
We extend the ARE method proposed in Gómez and Lagakos (2013) devised to decide which primary endpoint to choose when comparing two treatments in a randomized clinical trial. The ARE method is
based on the Asymptotic Relative Efficiency (ARE) between two logrank tests to compare two treatments: one is based on a relevant endpoint E1 while the other is based on a composite endpoint E* = E1 ¿ E2, where E2 is an additional endpoint. The ARE depends, besides some intuitive parameters, on the joint law of the times T1 and T2 from randomization to E1 and E2, respectively. Gómez and Lagakos (2013) characterize this joint law by means of Frank’s copula. In our work, several families of copulas can be chosen for the bivariate survival function of (T1, T2) so that different dependence struc- tures between T1 and T2 are feasible. We motivate the problem and show how to apply the method through a real cardiovascular clinical trial. We explore the influence of the
copula chosen into the ARE value by means of a simulation study. We conclude that the recommendation on whether or not to use
the composite endpoint as the primary endpoint for the investigation is, almost always, independent of the copula chosen.
Fri, 20 Feb 2015 18:40:25 GMThttp://hdl.handle.net/2117/264562015-02-20T18:40:25ZPlana-Ripoll, OleguerGómez Melis, GuadalupeWe extend the ARE method proposed in Gómez and Lagakos (2013) devised to decide which primary endpoint to choose when comparing two treatments in a randomized clinical trial. The ARE method is
based on the Asymptotic Relative Efficiency (ARE) between two logrank tests to compare two treatments: one is based on a relevant endpoint E1 while the other is based on a composite endpoint E* = E1 ¿ E2, where E2 is an additional endpoint. The ARE depends, besides some intuitive parameters, on the joint law of the times T1 and T2 from randomization to E1 and E2, respectively. Gómez and Lagakos (2013) characterize this joint law by means of Frank’s copula. In our work, several families of copulas can be chosen for the bivariate survival function of (T1, T2) so that different dependence struc- tures between T1 and T2 are feasible. We motivate the problem and show how to apply the method through a real cardiovascular clinical trial. We explore the influence of the
copula chosen into the ARE value by means of a simulation study. We conclude that the recommendation on whether or not to use
the composite endpoint as the primary endpoint for the investigation is, almost always, independent of the copula chosen.Review of multivariate survival data
http://hdl.handle.net/2117/22543
Review of multivariate survival data
Gómez Melis, Guadalupe; Calle Rosingana, M. Luz; Serrat Piè, Carles; Espinal Berenguer, Anna
This paper reviews some of the main contributions in the area of multivariate survival data and proposes some possible extensions. In particular, we have concentrated our search and study on those papers that are relevant to the situation where two (or more) consecutive variables are followed until a common day of analysis and subject to informative censoring.
The paper reviews bivariate nonparametric approaches and extend some of them to the case of two nonconsecutive times. We introduce the notation and construct the likelihood for the general problem of more than two consecutive survival times. We formulate the time dependencies and trends via a Bayesian approach. Finally, three regression models for multivariate survival times are discussed together with the differences among them which will be useful when the main interest is on the effect of covariates on the risk of failure.
Document de recerca publicat per la UPC. Departament d'Estadística i Investigació operativa
Mon, 07 Apr 2014 14:05:21 GMThttp://hdl.handle.net/2117/225432014-04-07T14:05:21ZGómez Melis, GuadalupeCalle Rosingana, M. LuzSerrat Piè, CarlesEspinal Berenguer, AnnaThis paper reviews some of the main contributions in the area of multivariate survival data and proposes some possible extensions. In particular, we have concentrated our search and study on those papers that are relevant to the situation where two (or more) consecutive variables are followed until a common day of analysis and subject to informative censoring.
The paper reviews bivariate nonparametric approaches and extend some of them to the case of two nonconsecutive times. We introduce the notation and construct the likelihood for the general problem of more than two consecutive survival times. We formulate the time dependencies and trends via a Bayesian approach. Finally, three regression models for multivariate survival times are discussed together with the differences among them which will be useful when the main interest is on the effect of covariates on the risk of failure.Recommendations to choose the primary endpoint in cardiovascular clinical trials
http://hdl.handle.net/2117/21995
Recommendations to choose the primary endpoint in cardiovascular clinical trials
Gómez Melis, Guadalupe; Gómez Mateu, Moisés; Dafni, Urania
Background – A composite endpoint is often used as the primary endpoint to assess the efficacy of a new treatment in randomized clinical trials (RCT). In cardiovascular trials, the often rare event of the relevant primary endpoint
(individual or composite), such as cardiovascular death (CV death),
Myocardial Infarction (MI), or both, is combined with a more common
secondary endpoint, such as target lesion revascularization, with the aim to
increase the statistical power of the study.
Methods – Gómez and Lagakos developed statistical methodology to be used
at the design stage of a RCT for deciding whether to expand a study
relevant primary endpoint e1 to e*, the composite of e1 and a secondary
endpoint e2. The method uses the asymptotic relative efficiency of the
logrank test for comparing treatment groups based on e1 versus the logrank
test based on e*. The method is used to assess, in the cardiovascular
research area, the characteristics of the candidate individual endpoints that
should govern the choice of using a composite endpoint as the primary
endpoint in a clinical trial.
Results and conclusions – A set of recommendations is provided based on
the reported values of the frequencies of observing each candidate endpoint
as well as on the magnitude of the effect of treatment as expressed by the hazard ratio, supported by cardiovascular RCTs published in 2008.
Tue, 11 Mar 2014 14:33:24 GMThttp://hdl.handle.net/2117/219952014-03-11T14:33:24ZGómez Melis, GuadalupeGómez Mateu, MoisésDafni, UraniaBackground – A composite endpoint is often used as the primary endpoint to assess the efficacy of a new treatment in randomized clinical trials (RCT). In cardiovascular trials, the often rare event of the relevant primary endpoint
(individual or composite), such as cardiovascular death (CV death),
Myocardial Infarction (MI), or both, is combined with a more common
secondary endpoint, such as target lesion revascularization, with the aim to
increase the statistical power of the study.
Methods – Gómez and Lagakos developed statistical methodology to be used
at the design stage of a RCT for deciding whether to expand a study
relevant primary endpoint e1 to e*, the composite of e1 and a secondary
endpoint e2. The method uses the asymptotic relative efficiency of the
logrank test for comparing treatment groups based on e1 versus the logrank
test based on e*. The method is used to assess, in the cardiovascular
research area, the characteristics of the candidate individual endpoints that
should govern the choice of using a composite endpoint as the primary
endpoint in a clinical trial.
Results and conclusions – A set of recommendations is provided based on
the reported values of the frequencies of observing each candidate endpoint
as well as on the magnitude of the effect of treatment as expressed by the hazard ratio, supported by cardiovascular RCTs published in 2008.Statistics for spatial functional data
http://hdl.handle.net/2117/2446
Statistics for spatial functional data
Delicado Useros, Pedro Francisco; Giraldo, Ramón; Comas, Carles; Mateu, Jorge
Functional Data Analysis is a relatively new branch in Statistics. Experiments where a complete function is observed for each individual give rise to functional data. In this work we focus on the case of functional data presenting spatial dependence. The three classic types of spatial data structures (geostatistical data, point patterns and areal data)
can be combined with functional data as it is shown in the examples of each situation provided here. We also review some contributions in the literature on spatial functional data.
Mon, 15 Dec 2008 17:04:59 GMThttp://hdl.handle.net/2117/24462008-12-15T17:04:59ZDelicado Useros, Pedro FranciscoGiraldo, RamónComas, CarlesMateu, JorgeFunctional Data Analysis is a relatively new branch in Statistics. Experiments where a complete function is observed for each individual give rise to functional data. In this work we focus on the case of functional data presenting spatial dependence. The three classic types of spatial data structures (geostatistical data, point patterns and areal data)
can be combined with functional data as it is shown in the examples of each situation provided here. We also review some contributions in the literature on spatial functional data.The role of survival functions in competing risks
http://hdl.handle.net/2117/2202
The role of survival functions in competing risks
Porta Bleda, Núria; Gómez Melis, Guadalupe; Calle Rosingana, M. Luz
Competing risks data usually arises in studies in which the failure of an individual may be classified into one of k (k > 1) mutually exclusive causes of failure. When competing risks are present, there are two main differences with classical survival analysis: (i) survival functions are not mainly used to describe cause-specific failures and, (ii) classical estimation techniques may provide biased results.
The main goal of this paper is to review, clarify and present the formulation of a competing risks model and the basic nonparametric estimation methods. We show why the use of survival functions in the competing risks framework may mislead the user, and we illustrate the presented methodologies by developing two examples from real data. The methods presented here can be implemented with several statistical packages, including R, SPSS and SAS: we give some highlights on how to perform
a competing risks analysis with these software packages.
Fri, 01 Aug 2008 14:33:24 GMThttp://hdl.handle.net/2117/22022008-08-01T14:33:24ZPorta Bleda, NúriaGómez Melis, GuadalupeCalle Rosingana, M. LuzCompeting risks data usually arises in studies in which the failure of an individual may be classified into one of k (k > 1) mutually exclusive causes of failure. When competing risks are present, there are two main differences with classical survival analysis: (i) survival functions are not mainly used to describe cause-specific failures and, (ii) classical estimation techniques may provide biased results.
The main goal of this paper is to review, clarify and present the formulation of a competing risks model and the basic nonparametric estimation methods. We show why the use of survival functions in the competing risks framework may mislead the user, and we illustrate the presented methodologies by developing two examples from real data. The methods presented here can be implemented with several statistical packages, including R, SPSS and SAS: we give some highlights on how to perform
a competing risks analysis with these software packages.Competing risks methods
http://hdl.handle.net/2117/2201
Competing risks methods
Porta Bleda, Núria; Gómez Melis, Guadalupe; Calle Rosingana, M. Luz; Malats i Riera, Núria
Competing risks data usually arises in studies in which the failure of an individual may be classified into one of k (k > 1) mutually exclusive causes of failure. When competing risks are present, classical survival analysis techniques may not be appropriate to use. The main goal of this paper is to review the specific methods to deal with competing risks. To this aim, we first focus on how to specify a competing risks model, which is the structure of observed data in this framework, and
how components of the model are estimated from a given random sample. In addition, we discuss how to correctly interpret probabilities in the presence of competing risks, and regression models are
considered in detail. To conclude, we illustrate the problem with data from a bladder cancer study.
Fri, 01 Aug 2008 14:25:34 GMThttp://hdl.handle.net/2117/22012008-08-01T14:25:34ZPorta Bleda, NúriaGómez Melis, GuadalupeCalle Rosingana, M. LuzMalats i Riera, NúriaCompeting risks data usually arises in studies in which the failure of an individual may be classified into one of k (k > 1) mutually exclusive causes of failure. When competing risks are present, classical survival analysis techniques may not be appropriate to use. The main goal of this paper is to review the specific methods to deal with competing risks. To this aim, we first focus on how to specify a competing risks model, which is the structure of observed data in this framework, and
how components of the model are estimated from a given random sample. In addition, we discuss how to correctly interpret probabilities in the presence of competing risks, and regression models are
considered in detail. To conclude, we illustrate the problem with data from a bladder cancer study.