The Application of Generalized Propensity Score Approaches on Dichotomous, Continuous, and Ordinal Treatment to Analyze the Effect of Total Radiation Dose on Cervical Esophageal Cancer Survival Public

Abstract

Introduction:The purpose of this paper is to investigate the impact of total radiation dose level on overall survival for cervical esophageal cancer (CEC) patients receiving chemoradiotherapy. As CEC is rare, there is little data currently available on the optimal radiation dosage for prolonging survival.

Methods:A retrospective cohort survival analysis was conducted using 2014 data from the National Cancer Database. Univariate, multivariable, and stratified Cox proportional hazards models are used to evaluate the impact of total radiation dose on overall survival. Cox-PH models using propensity score weighting and covariate adjustment are also used. For covariate adjustment, we evaluate two cases of the generalized propensity score: treating dose as a continuous variable and as an ordinal variable with four treatment groups: >4500-5040 cGy, >5040-5940 cGy, >5940-6400 cGy, and >6400-7020 cGy. We further use univariate and multivariable logistic regression models to investigate characteristics associated with a patient receiving a total radiation dose in the highest dose group.

Results:The multivariable and ordinal generalized propensity score Cox-PH models both show improved survival for patients in the highest total radiation dose group compared to patients in the lowest dose group (Hazard Ratios: 0.68 [0.50-0.90] and 0.732 [0.550, 0.973], respectively).The weighted propensity score model reveals patients in the highest dose groups have superior survival compared to patients in the three lower dose groups combined (HR: 0.71 [0.56-0.92]).The continuous generalized propensity score model shows that a general increase in total radiation dose is associated with improved survival (HR: 0.983 [0.969, 0.998]). Further, logistic regression shows that facility location, urban/rural residence, and regional treatment modality are associated with receiving a total radiation dose in the highest dose group. Stratified analysis showed that higher dose may be more beneficial to patients receiving IMRT radiation modality or patients with an AJCC clinical stage group of 3 or 4.

Conclusion: Total radiation dose above 6400 cGy appears to improve overall survival in CEC patients receiving chemoradiotherapy. Generalized propensity score methods are useful extensions of the traditional propensity score for evaluating treatment as a continuous or ordinal variable, but more research in proper assessment covariate balance is needed.

Table of Contents

I. Introduction.................................................................1

II. Background.................................................................1

III. Methods....................................................................4

A. Inclusion and Exclusion Criteria........................................4

B. Outcome, Cohorts, and Variables of Interest......................5

C. Statistical Analysis.......................................................6

i. Descriptive Statistics.....................................................6

ii. Univariate Analysis........................................................6

iii. Multivariable Analysis....................................................7

iv. Stratified Analysis........................................................8

D. Traditional Propensity Score Analysis...............................9

i. Overview.....................................................................9

ii. Matching...................................................................10

iii. Inverse Probability of Treatment Weighting......................12

E. Generalized Propensity Score Analysis.............................13

i. Overview....................................................................13

ii. Continuous Treatment Variable.......................................13

iii. Ordinal Treatment Variable............................................14

iv. Balance Checking........................................................16

IV. Results.....................................................................19

A. Statistical Analysis......................................................19

i. Descriptive Statistics....................................................19

ii. Univariate Analysis.......................................................20

iii. Multivariable Analysis...................................................21

iv. Stratified Analysis.......................................................23

B. Traditional Propensity Score Analysis..............................24

i. Matching....................................................................24

ii. Inverse Probability of Treatment Weighting.......................24

C. Generalized Propensity Score Analysis.............................24

i. Continuous Treatment Variable........................................24

ii. Ordinal Treatment Variable.............................................25

iii. Balance Checking........................................................25

V. Discussion..................................................................25

VI. Conclusion................................................................29

VII. References..............................................................30

VIII. Appendix: Tables and Figures.....................................34

About this Master's Thesis

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Rollins School of Public Health
Department	Biostatistics and Bioinformatics
Subfield / Discipline	Biostatistics - MPH & MSPH
Degree	MSPH
Submission	Master's Thesis
Language	English
Research Field	Statistics Health Sciences, Oncology
Mot-clé	generalized propensity score cervical esophageal cancer survival analysis propensity score
Committee Chair / Thesis Advisor	Liu, Yuan, Emory University
Committee Members	Kutner, Michael, Emory University
Partnering Agencies	Hospital or other health care provider

Dernière modification

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	The Application of Generalized Propensity Score Approaches on Dichotomous, Continuous, and Ordinal Treatment to Analyze the Effect of Total Radiation Dose on Cervical Esophageal Cancer Survival ()	2018-08-28 15:40:07 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions