Socioeconomic vulnerability and osteoporosis treatment disparities during COVID-19 lockdown among U.S. medicare enrollees who initiated romosozumab

Osteoporosis is a major public health issue that is expanding as the US population grows older [1]. Several medications are available for osteoporosis prevention and treatment, including romosozumab, the newest therapy approved by the Food and Drug Administration (FDA) in April 2019. Romosozumab is a monoclonal antibody that inhibits sclerostin, a molecule that predominately represses Wnt-mediated bone formation. Romosozumab, which has been shown to reduce fracture risk, is indicated for postmenopausal women at high risk of fracture and is administered subcutaneously in a healthcare setting once a month for 12 doses [2]. Despite the availability of romosozumab and several many other bone-protective osteoporosis treatments, premature discontinuation remains a significant issue in osteoporosis management. For example, one observational study showed that after 12 months, the discontinuation probability was highest among patients on every three-month injectable ibandronate (69.1%), followed by those taking daily injections of teriparatide (67.1%), yearly zoledronic acid (59.2%), and twice yearly denosumab (48.8%) [3]. Poor adherence to osteoporosis treatment leads to suboptimal outcomes, including increased accelerated bone loss and a higher risk of fractures [4‐6].

The reasons for non-adherence to anti-osteoporosis treatment are multifaceted. The asymptomatic nature of osteoporosis reduces the perceived benefit of the treatment, which may lead to non-adherence [7]. Additionally, factors such as the method of administration, dosing frequency, medication costs, perceived side effects, comorbidities, and socioeconomic status can all influence adherence to anti-osteoporosis treatment [8‐10]. Socioeconomic status as well as other social determinants of health may have strong influence on adherence to osteoporosis medications [11]. Devold et al. [12] showed a higher adherence to oral alendronate among patients with higher household income and higher education level. However, there is limited evidence regarding the association between area-level social determinants of health and adherence to osteoporosis treatment.

Romosozumab has the highest dosing frequency (once monthly) among commonly used injectable osteoporosis medications in the U.S. that require in-office visits for each injection. This makes romosozumab adherence especially vulnerable to disruptions in healthcare delivery caused by public health emergencies such as the COVID-19 pandemic, which led to widespread social distancing, restrictions on face-to-face visits, and clinic closures. Furthermore, evidence indicates that healthcare access among populations with lower socioeconomic status (SES) faced greater disruptions during COVID-19, including difficulties in receiving treatment for ongoing conditions [13‐15]. In this study, we aim to evaluate the association between area-level Social Determinants of Health (SDOH) and romosozumab treatment discontinuation, as well as the effects of the COVID-19 lockdown on osteoporosis treatment among Medicare-enrolled women who initiated romosozumab. Findings from this study will not only inform future policies to support equitable osteoporosis care during health crises but will also benefit our preparedness to maintain essential treatment adherence in similar pandemics or emergencies in the future.

There were 6,777 new users of romosozumab identified between April 1, 2019 and Sep. 30, 2020 (see supplementary Figure S1 for study diagram). Table 1 presents the baseline characteristics of romosozumab new users in each period. The mean age of romosozumab new users was 76.3 years (SD: 6.8). Patients from counties with the lowest SES level consisted of about 40% of all romosozumab new users. About 60% of romosozumab new users had been treated for osteoporosis, while 43.9% had a history of fracture.

Among 6,777 romosozumab new users, 2,937 (43.3%) discontinued romosozumab within 1 year after initiation. As shown in Table 2, individuals in the lowest SES group (SES tertile 3, most vulnerable) showed shorter persistence time and higher discontinuation rates. The 25th percentile persistence time (Kaplan–Meier analysis) was 2.9 months and 4.9 months in Period 2 and 3 subcohorts, respectively — significantly shorter than those of the highest SES group (3.5 months and 7.3 months, Peto’s test p = 0.040 and 0.002 for Period 2 and 3 subcohorts, respectively). No statistically significant difference was found in persistence across SES tertiles in Period 1 subcohort (Peto’s test p = 0.84).

The results of crude, DcRS-adjusted, and matching weight-adjusted hazard ratios are shown in Table 2. After adjusting for DcRS, individuals in the highest SES tertile (tertile 1, least vulnerable) had a significantly lower risk of discontinuing romosozumab in the Period 3 subcohort (HR = 0.78, 95% CI: 0.66–0.92), compared to those in the lowest SES tertile. Matching weight-adjusted analyses yielded consistent results (HR = 0.76, 95% CI: 0.63–0.92). No statistically significant differences in discontinuation risk were observed across SES tertiles in the Period 1 and Period 2 subcohorts in adjusted models. Sensitivity analyses using a 60-day grace period for discontinuation (Table S1) produced qualitatively similar findings.

Detailed information and evaluation of DcRS and matching-weights are shown in supplementary tables and figures (Table S2, S3. Figure S2, S3). For DcRS estimation, the results showed that older age, higher Charlson comorbidity score, more prescription drugs used, and dual eligible for Medicaid were positively correlated with romosozumab discontinuation in at least 2 subcohorts with statistical significance (Table S1). For matching-weight analysis, as shown in Supplementary Figure S4, all covariates were well balanced after weighting with SMD not greater than 0.1, indicating good balance between groups.

In the secondary analysis examining treatment patterns following romosozumab discontinuation, among 2,937 romosozumab initiators who met premature discontinuation criteria, 1,816 (61.8%) restarted treatment within 6 months. Of those 1,816 patients, 661 (36.4%) switched to other medications (denosumab (70.5%), bisphosphonates (27.8%), PTH analogues (1.7%)). The median treatment gap duration (95% CI) was 93 days (86, 99) for high SES tertile, 98 days (92, 108) for medium tertile, and 98 days (94, 104) for low SES tertile (p = 0.030). At 6 months, the cumulative incidence (95% CI) for switching to other osteoporosis treatments was 27.5% [24.2%, 30.9%] for high SES, 22.8% [20.1%, 25.6%] for medium, and 22.7% [20.3%, 25.1%] for low SES (p = 0.035). For resuming romosozumab, the cumulative incidence was 44.3% [40.5%, 48.1%] for high SES, 41.0% [37.7%, 44.2%] for medium, and 43.8% [40.9%, 46.6%] for low SES (p = 0.55). (Fig. 1).

Bild vergrößern

Figure 2 showed shows the factors correlated with restarting osteoporosis treatment after romosozumab discontinuation in the multivariable Cox model. Factors collected during romosozumab treatment were are listed in Table S4. Higher Charlson comorbidity score, history of COPD, myocardial infarction or stroke, or dementia were negatively associated with treatment resumption, while longer romosozumab treatment duration before discontinuation, DXA scan before romosozumab initiation, previous use of anti-osteoporosis treatment, history of fracture, and use of telemedicine during romosozumab treatment were positively associated with osteoporosis treatment. Table 3 shows the HR (95% CI) of osteoporosis treatment restart, switch, and resumption of romosozumab use by SES tertiles before and after lockdown. Romosozumab users who lived in the highest SES level counties were more likely to switch osteoporosis treatment during COVID lockdown compared with those living in lower SES level counties (HR = 1.30, 95% CI: 1.06—1.60), while resumption of romosozumab did not vary by SES, suggesting the higher likelihood of restarting osteoporosis treatment among those romosozumab users who lived in the highest SES level counties (HR = 1.14, 95% CI: 1.01–1.30) was primarily driven by treatment switching.

Bild vergrößern

In this observational study using national Medicare data, our results showed a lower risk of discontinuation of romosozumab for patients from counties of higher SES level (lower SES vulnerability) during lockdown (with a hazard decrease of more than 20%) after adjusting for individual factors. Moreover, those who lived in the highest SES level (least SES vulnerable) counties were more likely to switch to other osteoporosis treatment after the discontinuation of romosozuamb during COVID lockdown compared with those living in counties with lower SES level (more vulnerable), contributing to a shorter osteoporosis treatment gap duration and a lower risk of osteoporosis treatment discontinuation.

There is evidence indicating that women from high-poverty or rural communities are at a higher risk of lower bone mineral density, osteoporosis, and osteoporotic fractures [29‐31]. Therefore, maintaining adequate osteoporosis treatment for this patient population is crucial, as these patients are vulnerable for osteoporosis and fracture. Area social determinants of health, often referred to as"community vital signs,"encompass social and environmental factors that influence patient health. Recognizing and understanding these factors can enhance our insight into the drivers of treatment adherence and health outcomes, thereby guiding clinical recommendations and improving community services [32]. During the COVID pandemic, physician factors, such as closed physician’s offices and availability of appointments, were identified as major contributors to forgone care during the pandemic among Medicare beneficiaries [33]. This may exacerbate the existing barriers and disparities in health care accessibility as residents in low-income communities report a lack of specialty care compared to those with higher incomes [34]. Our study highlights the need to prioritize measures aimed at maintaining access to medical care, especially for SES vulnerable communities, during future pandemics that create barriers to healthcare access. Moreover, our results showed that patients living in counties with highest SES (least SES vulnerable) were more likely to switch their osteoporosis treatment during the lockdown compared to those in more vulnerable counties. This suggests that these patients may have had better support and healthcare access, enabling them to switch treatments during the lockdown to avoid long delays and treatment discontinuation. In addition to better healthcare accessibility in less vulnerable counties, future studies should evaluate the role of social support, as it has been shown to be associated with health-promoting behaviors such as increased adherence to medical treatments and help-seeking behavior [35, 36].

In our primary analysis, we applied two methods to control for individual comorbidities and other factors, including dual Medicaid eligibility as an indicator of individual socioeconomic status. While Arbogast and Ray [37] found that disease risk score-based and propensity score-based approaches had comparable performance in controlling confounding, Stürmer et al. [26] suggested that, in smaller studies, differences in effect estimates between analytic strategies become more pronounced. Given that our analysis involved three small subcohorts, using different approaches to ensure robust results was essential. Our findings from the DcRS and matching weight-adjusted approaches were consistent, except in the period 1 subcohort (Table 2). This discrepancy may be due to the smaller sample size in this subcohort. Although all covariates were balanced after weighting, we prefer not to draw conclusive inferences from this subcohort. Further studies are needed to evaluate area socioeconomic vulnerability and treatment discontinuation in the early post-approval stage of new medications.

The factors potentially associated with the termination or resumption of osteoporosis treatment after discontinuing romosozumab (Fig. 2) in our study align with previous literature [21, 38]. Notably, we found that the use of telemedicine during romosozumab treatment was positively correlated with an increased probability of resuming osteoporosis treatment after discontinuation. While prior evidence has shown that telehealth interventions, including telephone counseling, can improve adherence to osteoporosis treatment, the findings are mixed [39, 40]. The rapid expansion of telehealth during the COVID-19 pandemic thus represents an opportunity for improved medication management. However, evidence on the impact of telehealth on osteoporosis treatment adherence during the pandemic remains limited. An observational study focusing on denosumab adherence during the COVID-19 lockdown found that adherent patients used more telemedicine compared to those who failed to adhere, although the difference was not significant [41]. Our results suggest that patients with access to telehealth may be more willing to seek help and more likely to make informed treatment decisions, which could contribute to the resumption of osteoporosis treatment. However, there are concerns that telemedicine may exacerbate health inequities. Evidence indicates that during the pandemic, minority patients and those living in low SES regions were less likely to use telemedicine [42, 43]. The role of telemedicine in managing chronic diseases such as osteoporosis, especially when in-person visits are compromised, warrants further investigation.

This study has several limitations. First, the claims data did not include bone mineral density measures, which may influence the risk of treatment non-adherence. Second, our data cannot identify primary non-compliance (i.e., drugs prescribed but never administered) due to the lack of prescribing data. In the future, linking claims data with electronic health records may help address this limitation. Third, we used two methods to evaluate the association between socioeconomic vulnerability and romosozumab treatment discontinuation, aiming to address confounding by individual characteristics. However, we did not make direct comparisons between subcohorts due to a significant risk of unmeasured confounding. Factors contributing to this risk include patient disease severity, overall health status, and external factors such as marketing exposure and prescriber preference, etc. [44] These influences may vary between patients who initiated treatment shortly after approval and those who began treatment several months later. Additionally, unmeasured confounding may exist between patients who initiated treatment earlier within each six-month period versus those who started later. Fourth, patients entering the secondary analysis were required to have a 60‐day gap after the last romosozumab injection to meet discontinuation criteria, which could introduce a brief period of immortal time. Although our design allowed us to capture early switching events as many patients in our cohort switched treatment to denosumab, future studies focusing on treatment sequence should consider alternative analytic strategies to fully characterize the full spectrum and timing of treatment switching. Fifth, we accounted for both events that occurred during romosozumab treatment and baseline factors for treatment resumption/termination, but only baseline factors for discontinuation. As shown by Yun et al. [21], events that occur closer to discontinuation, such as more office visits, side effects, or cost, may have a greater impact. This warrants further investigation using appropriate designs to address the dynamic risk of discontinuation. Finally, we only included denosumab, bisphosphonates, and PTH analogues as potential options for switching from romosozumab after discontinuation. Other treatments, such as selective estrogen receptor modulators (SERMs), were not included. Although SERMs may benefit patients who are intolerant to other therapies or at risk for atypical femoral fractures or osteonecrosis of the jaw, evidence shows that raloxifene has a weaker fracture-prevention effect than the osteoporosis medications included in this study, particularly for non-vertebral fractures [45]. Future studies should explore the potential role of SERMs following or in combination with romosozumab among female patients.

The COVID-19 pandemic exacerbated many pre-existing health disparities [14]. Our study found that Medicare beneficiaries who initiated romosozumab treatment and resided in counties with higher SES level were less likely to discontinue the treatment during the lockdown, after adjusting for individual comorbidities. Furthermore, if they did discontinue, they were more likely to switch to alternative osteoporosis therapies, leading to shorter treatment gaps and better osteoporosis treatment persistence. To maintain adherence to long-term therapies like osteoporosis treatment during future public health emergencies, it is essential to address community-level factors. Further study should focus on community level intervention that could provide timely support to vulnerable patients, particularly those in low SES areas.