東華大學圖書館 |

Parametric and Semi-Parametric Approaches to Estimation of Survival Distributions of Treatment Strategies in Sequential Multiple Assignment Randomized Trials.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Parametric and Semi-Parametric Approaches to Estimation of Survival Distributions of Treatment Strategies in Sequential Multiple Assignment Randomized Trials./
作者:	Gu, Wenjuan.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2018,
面頁冊數:	107 p.
附註:	Source: Dissertations Abstracts International, Volume: 79-09, Section: B.
Contained By:	Dissertations Abstracts International79-09B.
標題:	Biostatistics. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10743753
ISBN:	9780355621679

Parametric and Semi-Parametric Approaches to Estimation of Survival Distributions of Treatment Strategies in Sequential Multiple Assignment Randomized Trials.
Gu, Wenjuan.

Parametric and Semi-Parametric Approaches to Estimation of Survival Distributions of Treatment Strategies in Sequential Multiple Assignment Randomized Trials. - Ann Arbor : ProQuest Dissertations & Theses, 2018 - 107 p.

Source: Dissertations Abstracts International, Volume: 79-09, Section: B.

Thesis (Ph.D.)--The George Washington University, 2018.

This item must not be sold to any third party vendors.

Background: Dynamic treatment regimens (DTRs) are a sequence of treatments tailored to individual patients to achieve optimal outcomes. DTRs are also known as adaptive treatment strategies/policies. They are sequences of decision rules that adjust dynamically to the time-varying patients conditions at multiple stages. DTRs have gaind popularity in chronic conditions such as cancer and HIV. In this dissertation we are particualarly interested in one type of DTR: Sequential Multiple Assignment Randomized trial (SMART), in which patients are sequentially randomized to multiple treatments with the assignment dependent on responses in previous stages. Because of this relationship between treatment stages, it's of interest to estimate the overall outcome of the entire treatment strategy instead of each single stage. Nonparametric inverse probability weighting methods have been developed and studied in recent years. These methods are unbiased and consistent but not as efficient as their parametric counterparts. We propose parametric and semiparametric methods to estimate survival probabilities for two-stage two-treatment SMARTs. Purpose: To develop parsimonious and efficient methods for estimation of survival probabilities in SMART. When the data distribution is correctly specified, parametric and nonparametric models not only improve efficiency of estimation but also facilitate understanding and help generalization. We'd like to study appropriate models for various types of data and explore application of our methods to identify optimal treatment strategies. Methods: We developed two parametric and two semiparametric approaches. Let Ajbk be a stragegy in which patients receive Aj first, then at time τ the responses are evaluated, responders then receive Bk and nonresponders are not further treated; where j = 1,2 and k = 1,2. We define ω as the proportion of deaths before τ. Among those who survived past τ, we define p as the probability of responding. The probability of death at time t (t ≥ τ) for strategy Ajbk is: FAjbk( t) = ω + (1 - ω){pF AjBk(t - τ) + (1 - p)FAjN(t - τ)} Here FAjBk is the probability of death for patients who received AjBk and FAjN is the probability of death for those received Aj and didn't respond. To estimate F AjBk and FAjN we used exponential, Weibull and Cox proportional hazard models (CoxI). We developed a second Cox model, which we call the CoxII. Insead of using ω to estimate the death rate before , we used a time-dependent indicators Z(t) for each treatment assignment. The variance-covariance matrix for FAjbk(t) is derived using maximum likelihood approaches and -method. We also derived sandwich estimator of variance for the exponential model. The estimated variances from exponential, Weibull, CoxI and CoxII models are compared to the variance of estimator from inverse probability weighting approach (IPW). Relative efficiency(RE), which is variance of our estimator devided by variance of the IPW estimator, is used to measure approvement in efficiency. RE greater than one means our estimator is more efficient. We used the CoxII model as an example to briefly explore the application of our methods in identification of optimal strategy based on patient characteristics. We added one binary variable X for patient characteristics and the interaction of X and treatment in the CoxII model. The difference given X = 0 is denoted as Δ0 and Δ1 given X = 1, the null hypothesis is H0 : Δ0 = Δ1. We studied three scenarios: null interaction, quantitative interaction and qualitative interaction. For each of the three scenarios the relationship between Δ 0 and Δ1 are Δ0 = Δ1, Δ 0 ≠ Δ1 with equal signs and Δ0 ≠ = Δ1 with different signs. Wald test was used to test for interaction. We conducted simulations and analyzed real data to compare our methods to IPW. Results: We simulated data from exponential distribution and compared variance of the exponential estimator to variance of the IPW estimator. We conducted simulation studies for τ = 0 and τ ≥ 0 and set the probability of response p to be 0.1, 0.5 and 0.9 for each τ. We calculated relative efficiency from 6 months to five years in the increment of 0.5 years. Regardless of prespecified values of τ and p, the exponential model is always more efficient and the efficiency improves from 50% to 4 folds depending on the time points of evaluation. For each scenario we calculated both the model based variance and the sandwich variance, as expected the two are always close to each other with the sandwich variance being slightly larger. To further study the sandwich estimator under misspecification of data distribution, we vi simulated data from distributions of lognormal, Weibull with shape parameter γ = 1:05 and Weibull with γ = 2. In each simulation we estimated FAjbk using the exponential model and calculated both the model based and sandwich variances. (Abstract shortened by ProQuest.).

ISBN: 9780355621679Subjects--Topical Terms:

1002712
Biostatistics.

Parametric and Semi-Parametric Approaches to Estimation of Survival Distributions of Treatment Strategies in Sequential Multiple Assignment Randomized Trials.
LDR:06143nmm a2200313 4500 001 2207913
005 20190929184010.5
008 201008s2018 ||||||||||||||||| ||eng d
020 $a 9780355621679
035 $a (MiAaPQ)AAI10743753
035 $a (MiAaPQ)gwu:13980
035 $a AAI10743753
040 $a MiAaPQ $c MiAaPQ
100 1 $a Gu, Wenjuan. $3 3434913
245 1 0 $a Parametric and Semi-Parametric Approaches to Estimation of Survival Distributions of Treatment Strategies in Sequential Multiple Assignment Randomized Trials.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2018
300 $a 107 p.
500 $a Source: Dissertations Abstracts International, Volume: 79-09, Section: B.
500 $a Publisher info.: Dissertation/Thesis.
500 $a Advisor: Follmann, Dean;Younes, Naji.
502 $a Thesis (Ph.D.)--The George Washington University, 2018.
506 $a This item must not be sold to any third party vendors.
520 $a Background: Dynamic treatment regimens (DTRs) are a sequence of treatments tailored to individual patients to achieve optimal outcomes. DTRs are also known as adaptive treatment strategies/policies. They are sequences of decision rules that adjust dynamically to the time-varying patients conditions at multiple stages. DTRs have gaind popularity in chronic conditions such as cancer and HIV. In this dissertation we are particualarly interested in one type of DTR: Sequential Multiple Assignment Randomized trial (SMART), in which patients are sequentially randomized to multiple treatments with the assignment dependent on responses in previous stages. Because of this relationship between treatment stages, it's of interest to estimate the overall outcome of the entire treatment strategy instead of each single stage. Nonparametric inverse probability weighting methods have been developed and studied in recent years. These methods are unbiased and consistent but not as efficient as their parametric counterparts. We propose parametric and semiparametric methods to estimate survival probabilities for two-stage two-treatment SMARTs. Purpose: To develop parsimonious and efficient methods for estimation of survival probabilities in SMART. When the data distribution is correctly specified, parametric and nonparametric models not only improve efficiency of estimation but also facilitate understanding and help generalization. We'd like to study appropriate models for various types of data and explore application of our methods to identify optimal treatment strategies. Methods: We developed two parametric and two semiparametric approaches. Let Ajbk be a stragegy in which patients receive Aj first, then at time τ the responses are evaluated, responders then receive Bk and nonresponders are not further treated; where j = 1,2 and k = 1,2. We define ω as the proportion of deaths before τ. Among those who survived past τ, we define p as the probability of responding. The probability of death at time t (t ≥ τ) for strategy Ajbk is: FAjbk( t) = ω + (1 - ω){pF AjBk(t - τ) + (1 - p)FAjN(t - τ)} Here FAjBk is the probability of death for patients who received AjBk and FAjN is the probability of death for those received Aj and didn't respond. To estimate F AjBk and FAjN we used exponential, Weibull and Cox proportional hazard models (CoxI). We developed a second Cox model, which we call the CoxII. Insead of using ω to estimate the death rate before , we used a time-dependent indicators Z(t) for each treatment assignment. The variance-covariance matrix for FAjbk(t) is derived using maximum likelihood approaches and -method. We also derived sandwich estimator of variance for the exponential model. The estimated variances from exponential, Weibull, CoxI and CoxII models are compared to the variance of estimator from inverse probability weighting approach (IPW). Relative efficiency(RE), which is variance of our estimator devided by variance of the IPW estimator, is used to measure approvement in efficiency. RE greater than one means our estimator is more efficient. We used the CoxII model as an example to briefly explore the application of our methods in identification of optimal strategy based on patient characteristics. We added one binary variable X for patient characteristics and the interaction of X and treatment in the CoxII model. The difference given X = 0 is denoted as Δ0 and Δ1 given X = 1, the null hypothesis is H0 : Δ0 = Δ1. We studied three scenarios: null interaction, quantitative interaction and qualitative interaction. For each of the three scenarios the relationship between Δ 0 and Δ1 are Δ0 = Δ1, Δ 0 ≠ Δ1 with equal signs and Δ0 ≠ = Δ1 with different signs. Wald test was used to test for interaction. We conducted simulations and analyzed real data to compare our methods to IPW. Results: We simulated data from exponential distribution and compared variance of the exponential estimator to variance of the IPW estimator. We conducted simulation studies for τ = 0 and τ ≥ 0 and set the probability of response p to be 0.1, 0.5 and 0.9 for each τ. We calculated relative efficiency from 6 months to five years in the increment of 0.5 years. Regardless of prespecified values of τ and p, the exponential model is always more efficient and the efficiency improves from 50% to 4 folds depending on the time points of evaluation. For each scenario we calculated both the model based variance and the sandwich variance, as expected the two are always close to each other with the sandwich variance being slightly larger. To further study the sandwich estimator under misspecification of data distribution, we vi simulated data from distributions of lognormal, Weibull with shape parameter γ = 1:05 and Weibull with γ = 2. In each simulation we estimated FAjbk using the exponential model and calculated both the model based and sandwich variances. (Abstract shortened by ProQuest.).
590 $a School code: 0075.
650 4 $a Biostatistics. $3 1002712
690 $a 0308
710 2 $a The George Washington University. $b Biostatistics. $3 1285320
773 0 $t Dissertations Abstracts International $g 79-09B.
790 $a 0075
791 $a Ph.D.
792 $a 2018
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10743753