60 - Concern About Addiction is Associated with Lower Quality of Life in Patients with Osteoarthritis: An Observational Data Analysis
Has Audio
Louis P Garrison Jr1, Patricia Schepman2, Andrew B Bushmakin2, Rebecca L Robinson3, Leslie Tive2, Jerry Hall3, Mendy Dzingina2, James Jackson4, Mia Berry4, Joseph C Cappelleri2
1University of Washington, Seattle, WA, USA. 2Pfizer Inc, New York, NY, USA. 3Eli Lilly and Company, Indianapolis, IN, USA. 4Adelphi Real World, Bollington, United Kingdom
Purpose Osteoarthritis (OA) pain is one of the most common and economically burdensome conditions in the United States, affecting approximately 20% of adults and resulting in high healthcare costs and lost work productivity. Clinical guidelines recommend a multimodal approach to treating OA, combining physical therapies with pharmacological intervention, such as acetaminophen, NSAIDs, opioids, and other medicines. According to a prior US treatment preference study of a hypothetical pharmacological treatment that would prevent OA from worsening, patients with OA would be willing to accept some degree of risk for adverse events (Fraenkel et al., 2014). In a more recent study of OA patient preferences, Turk et al. (2020) showed that control of OA pain and symptoms and reduced treatment-related risk of physical dependency would be the two most important attributes of a new medicine for adult patients with moderate to severe OA and inadequate response to pain treatment. Several different measurement instruments could be helpful in weighing these impacts on patient quality of life (QoL). One of the most widely used disease-specific measure of OA symptoms is the Western Ontario and McMaster Universities of Osteoarthritis Index (WOMAC). While the WOMAC is commonly used in clinical studies, it is not suitable for direct use in conventional economic evaluation because WOMAC scores provide neither a cardinal nor a preference-based index scale. And generic instruments used to measure patient QoL for economic analysis, such as the EQ-5D, are infrequently included in clinical studies. Therefore, economic evaluations sometimes rely on a mapping from WOMAC to predict the EQ-5D. Several studies, including Cappelleri et al. (2016), have a demonstrated consistent statistical relationship between the two with demonstrated goodness of fit. In the current research, we aim to evaluate the relationship between self-reported concerns about becoming addicted to a medicine (for this condition, opioids) and individual patient QoL measured alternatively by (a) the EQ-5D-5L Index score and (b) the EQ-5D Visual Analogue Scale (VAS) in patients with OA. Methods This unique, non-trial observational study used patient-level cross-sectional surveydata collected from February-May 2017 from the US Adelphi Disease Specific Programme (DSP), which provides a holistic assessment for illnesses by gathering descriptive data on how diseases are managed in clinical practices based on both physician and patient perspectives. The Adelphi DSP for OA selected 153 physicians (primary care, rheumatology, and orthopedic surgeons) identified from publicly available lists of healthcare professionals. Physicians completed an on-line survey and an electronic patient record form collecting de-identified data (including patient demographics, medical history, treatment patterns) on their next 9 adult (18 years+) patients with OA. Each patient was invited to complete a self-completion survey relevant to the disease area. The question of interest for this analysis was about a “concerns of medication addiction” as reflected in the following Likert-scale question based on the level of agreement (from completely disagree [1] to completely agree[5]) with the statement “I am concerned about becoming addicted to my medicine” (CAA). A set of ordinary least squares (OLS) regressions using QoL measures (EQ-5D Index score and EQ-5D VAS) as outcomes and CAA as a continuous predictor were estimated, including models with CAA as a categorical predictor as a sensitivity analysis. The relationship between EQ-5D utility score as a predictor and EQ-5D VAS as the outcome was also studied. Finally, treating the EQ-5D VAS as the more general indicator of QoL, an OLS regression with the EQ-5D VAS as an outcome and with the EQ-5D-5L Index score and the CAA as two independent continuous predictors was estimated in this sample. Correlations between the measures were also assessed. Results A total of 866 OA patients completed the survey with the majority being female (61.2%), white (77.7%) and with mean age of 64.2 years (Standard deviation 11.7). 835 patients completed the single item: ‘I am concerned about becoming addicted to my medicine’. The responses were well distributed with sizable representation for each category: about one-third of the patients responded that they “agree” (18%) or “completely agree” (11%), while 27% responded “completely disagree” and 20% “disagree”. The relationship between CAA as a continuous predictor and the EQ-5D Index score revealed that a one-category increase in CAA score is associated with a 0.029 reduction in the EQ-5D Index score, equivalent to 0.14 in terms of the standardized effect sizes (ES), which can be interpreted “trivial-to-small” effect. The difference in means between the lowest category (“Completely disagree”) and the highest category (“Completely agree”) corresponds to value of 0.11 (p <0.0001) in the EQ-5D Index score (a “medium” 0.57 ES). Correlation between CAA and the EQ-5D Index score is 0.19 (p-value<0.0001). The relationship between CAA and the EQ-5D VAS showed that a one-category increase in CAA score was associated with a 2.6 points reduction in the EQ-5D VAS (0.15 ES). The difference in means between the lowest and the highest category is 10.5 (p <0.0001) representing “medium” ES of 0.59. Correlation between CAA and the EQ-5D VAS is 0.20 (p-value<0.0001). Using CAA as a categorical predictor indicated that a linear approximation is appropriate in both models. A significant and robust relationship between EQ-5D VAS as an outcome and EQ-5D Index score as a predictor was observed (slope 60.7; p value<0.0001). Correlation between EQ-5D Index score and the EQ-5D VAS is substantial 0.69 (p-value<0.0001). Using EQ-5D Index score as a categorical predictor indicated that a linear approximation is appropriate. When both EQ-5D Index score and CAA scores where used simultaneously as predictors of EQ-5D VAS, the effect of CAA (after adjusting for EQ-5D utility) was still significant (slope -0.97, p=0.0071). In this case, the difference between the CAA lowest and highest categories is 3.89 and the associated effect size is 0.22, which would be regarded as “small” statistically. This is equivalent to -0.039 on a utility scale of 0-1.0, which would be regarded as significant in utility and economic terms. Conclusions This study found that patients with a diagnosis of OA who have concerns about medication addiction—as reflected in self-reported concern about addiction—have significantly and meaningfully different EQ-5D utility and EQ-5D VAS scores compared with patients who do not have this concern. Furthermore, concern about addiction has an additional negative impact—of potential clinical and economic importance—that is not fully captured in EQ-5D utility. Health technology assessment authorities who rely on the EQ-5D may underestimate the value of products that reduce concerns about opioid dependency. It is well-accepted that the EQ-5D works better for some diseases than others. One work-around or adjustment that is sometimes used is to include a “bolt-on” question for patients in trials. In this case, it would be worthwhile to consider a bolt-on question for inclusion (after successful psychometric validation) about these concerns in an assessment of the impact of new interventions on OA patients. To our knowledge, this study is the first to use rigorous methodologies to estimate the disutility impact of concern about opioid addiction on patient quality of life in OA. Further research is needed to evaluate direction and magnitude of effect by severity for this disease.