Development and Validation of an Algorithm to Map the Prostate Cancer Index to the Patient Oriented Prostate Utility Scale (PORPUS)
CUA Online Library. Bremner K. 06/22/13; 31416; UP-37
Disclosure(s): none
CLICK HERE TO LOGIN
REGULAR CONTENT
REGULAR CONTENT
Login now to access Regular content available to all registered users.
Abstract
Rate & Comment (0)
Introduction and Objectives: The Prostate Cancer Index (PCI) is frequently used to measure quality of life (QOL) in prostate cancer (PC) patients. It provide 6 subscale scores: urinary, sexual, and bowel function and bother. The Patient Oriented Prostate Utility Scale (PORPUS-U) measures utility, a global measure of QOL, on a scale where 0 = dead and 1 = full health, for use in population surveys, clinics, clinical trials, and decision analyses. Our objective was to develop a function to predict PORPUS-U utilities from PCI scores.
Methods: We used patient-level data from two previous studies which administered the PORPUS-U and PCI concurrently. Study 1 included 248 PC outpatients interviewed 3 times within 12 months. Study 2 included 676 community-dwelling PC survivors surveyed by mail. Study 2 data were used to fit three linear regression models, which were validated using study 1 data (3 time periods). One model used original PORPUS-U scores, and two used log-transformed PORPUS-U scores, one with a hierarchy constraint and one without. All were tested with and without age as a covariate. Models were selected using stepwise selection and 5-fold cross validation. The predictive abilities of the models were assessed.
Results: The best-fitting model used the log-transformed PORPUS-U with no hierarchy constraint, without age. The R-squared was 0.72. The root mean squared error ranged from 0.041 to 0.061 for the 3 validation datasets. The overall mean predicted and observed utilities were similar (eg., 0.956 vs 0.955) but predicted utilities slightly overestimated the lowest 5% of observed PORPUS-U scores and underestimated the highest observed scores.
Conclusions: Our algorithm can estimate PORPUS-U utility scores from PCI scores, thus supplementing descriptive QOL with utility scores for a variety of populations and applications.
Methods: We used patient-level data from two previous studies which administered the PORPUS-U and PCI concurrently. Study 1 included 248 PC outpatients interviewed 3 times within 12 months. Study 2 included 676 community-dwelling PC survivors surveyed by mail. Study 2 data were used to fit three linear regression models, which were validated using study 1 data (3 time periods). One model used original PORPUS-U scores, and two used log-transformed PORPUS-U scores, one with a hierarchy constraint and one without. All were tested with and without age as a covariate. Models were selected using stepwise selection and 5-fold cross validation. The predictive abilities of the models were assessed.
Results: The best-fitting model used the log-transformed PORPUS-U with no hierarchy constraint, without age. The R-squared was 0.72. The root mean squared error ranged from 0.041 to 0.061 for the 3 validation datasets. The overall mean predicted and observed utilities were similar (eg., 0.956 vs 0.955) but predicted utilities slightly overestimated the lowest 5% of observed PORPUS-U scores and underestimated the highest observed scores.
Conclusions: Our algorithm can estimate PORPUS-U utility scores from PCI scores, thus supplementing descriptive QOL with utility scores for a variety of populations and applications.
Introduction and Objectives: The Prostate Cancer Index (PCI) is frequently used to measure quality of life (QOL) in prostate cancer (PC) patients. It provide 6 subscale scores: urinary, sexual, and bowel function and bother. The Patient Oriented Prostate Utility Scale (PORPUS-U) measures utility, a global measure of QOL, on a scale where 0 = dead and 1 = full health, for use in population surveys, clinics, clinical trials, and decision analyses. Our objective was to develop a function to predict PORPUS-U utilities from PCI scores.
Methods: We used patient-level data from two previous studies which administered the PORPUS-U and PCI concurrently. Study 1 included 248 PC outpatients interviewed 3 times within 12 months. Study 2 included 676 community-dwelling PC survivors surveyed by mail. Study 2 data were used to fit three linear regression models, which were validated using study 1 data (3 time periods). One model used original PORPUS-U scores, and two used log-transformed PORPUS-U scores, one with a hierarchy constraint and one without. All were tested with and without age as a covariate. Models were selected using stepwise selection and 5-fold cross validation. The predictive abilities of the models were assessed.
Results: The best-fitting model used the log-transformed PORPUS-U with no hierarchy constraint, without age. The R-squared was 0.72. The root mean squared error ranged from 0.041 to 0.061 for the 3 validation datasets. The overall mean predicted and observed utilities were similar (eg., 0.956 vs 0.955) but predicted utilities slightly overestimated the lowest 5% of observed PORPUS-U scores and underestimated the highest observed scores.
Conclusions: Our algorithm can estimate PORPUS-U utility scores from PCI scores, thus supplementing descriptive QOL with utility scores for a variety of populations and applications.
Methods: We used patient-level data from two previous studies which administered the PORPUS-U and PCI concurrently. Study 1 included 248 PC outpatients interviewed 3 times within 12 months. Study 2 included 676 community-dwelling PC survivors surveyed by mail. Study 2 data were used to fit three linear regression models, which were validated using study 1 data (3 time periods). One model used original PORPUS-U scores, and two used log-transformed PORPUS-U scores, one with a hierarchy constraint and one without. All were tested with and without age as a covariate. Models were selected using stepwise selection and 5-fold cross validation. The predictive abilities of the models were assessed.
Results: The best-fitting model used the log-transformed PORPUS-U with no hierarchy constraint, without age. The R-squared was 0.72. The root mean squared error ranged from 0.041 to 0.061 for the 3 validation datasets. The overall mean predicted and observed utilities were similar (eg., 0.956 vs 0.955) but predicted utilities slightly overestimated the lowest 5% of observed PORPUS-U scores and underestimated the highest observed scores.
Conclusions: Our algorithm can estimate PORPUS-U utility scores from PCI scores, thus supplementing descriptive QOL with utility scores for a variety of populations and applications.
Code of conduct/disclaimer available in General Terms & Conditions
{{ help_message }}
{{filter}}