Sandhya C Nair1, Paco M J Welsing2, Anne Karien C A Marijnissen2, Paulina Sijtsma3, Johannes W J Bijlsma2, Jacob M van Laar2, Floris P J G Lafeber2, G Ardine de Wit4. 1. Department of Rheumatology and Clinical Immunology, University Medical Centre Utrecht, S.C.Nair@umcutrecht.nl. 2. Department of Rheumatology and Clinical Immunology, University Medical Centre Utrecht. 3. Diakonessenhuis Utrecht. 4. Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht and National Institute of Public Health and the Environment, Bilthoven, The Netherlands.
Abstract
OBJECTIVE: Treatment in general is mostly directly aimed at disease activity, and measures such as the DAS28 might therefore present important additional information. Our aim was to develop and validate a model that uses a combination of disease activity (DAS28) and HAQs to estimate EuroQoL 5-dimension scale (EQ5D) utilities. METHODS: Longitudinal data from a cohort study in RA patients from the Utrecht Rheumatoid Arthritis Cohort study Group (Stichting Reumaonderzoek Utrecht) who started treatment with a biologic drug were used for mapping and validation. All 702 observations, including DAS28, HAQ and EQ5D assessed at the same time points, were used. The observations were randomly divided into a subset for development of the model (n = 428 observations) and a subset for validation (n = 274). A stepwise multivariable regression analysis was used to test the association of DAS28 (components) and HAQ (domains) with EQ5D. Model performance was assessed using the explained variance (R(2)) and root mean square errors. Observed and predicted utility scores were compared to check for under- or overestimation of the scores. Finally, the performance of the model was compared with published mapping models. RESULTS: Lower DAS28 score and HAQ items dressing and grooming, arising, eating, walking and activities were associated with higher EQ5D scores. The final model had an explained variance of 0.35 and a lower root mean square error as compared with other models tested. The agreement between predicted and observed scores was fair. CONCLUSION: HAQ components estimate EQ5D better than total HAQ. Adding DAS28 to HAQ components does not result in better utility estimations.
OBJECTIVE: Treatment in general is mostly directly aimed at disease activity, and measures such as the DAS28 might therefore present important additional information. Our aim was to develop and validate a model that uses a combination of disease activity (DAS28) and HAQs to estimate EuroQoL 5-dimension scale (EQ5D) utilities. METHODS: Longitudinal data from a cohort study in RApatients from the Utrecht Rheumatoid Arthritis Cohort study Group (Stichting Reumaonderzoek Utrecht) who started treatment with a biologic drug were used for mapping and validation. All 702 observations, including DAS28, HAQ and EQ5D assessed at the same time points, were used. The observations were randomly divided into a subset for development of the model (n = 428 observations) and a subset for validation (n = 274). A stepwise multivariable regression analysis was used to test the association of DAS28 (components) and HAQ (domains) with EQ5D. Model performance was assessed using the explained variance (R(2)) and root mean square errors. Observed and predicted utility scores were compared to check for under- or overestimation of the scores. Finally, the performance of the model was compared with published mapping models. RESULTS: Lower DAS28 score and HAQ items dressing and grooming, arising, eating, walking and activities were associated with higher EQ5D scores. The final model had an explained variance of 0.35 and a lower root mean square error as compared with other models tested. The agreement between predicted and observed scores was fair. CONCLUSION: HAQ components estimate EQ5D better than total HAQ. Adding DAS28 to HAQ components does not result in better utility estimations.
Authors: Junzeng Fu; Johannes C Schoeman; Amy C Harms; Herman A van Wietmarschen; Rob J Vreeken; Ruud Berger; Bart V J Cuppen; Floris P J G Lafeber; Jan van der Greef; Thomas Hankemeier Journal: Anal Bioanal Chem Date: 2016-07-12 Impact factor: 4.142