Patterns and predictors of osteoporosis medication discontinuation and switching among Medicare beneficiaries

Adherence to oral bisphosphonate therapy has been reported to be suboptimal, with about 50% of patients who initiate treatment stopping within one year [1–3]. This poor adherence is thought to increase the risk of consequent morbidity (e.g., fractures) and associated costs [4]. Adherence to osteoporosis medications has been evaluated in many studies in the United States (US) [5–10], and certain factors inversely related to adherence have been identified, including drug related side effects, having multiple comorbid conditions, inconvenient dosing, higher cost, and hospitalization during the one year period before the initiation of osteoporosis medications. However, our understanding of factors related to poor adherence may be incomplete due to methods used to define discontinuation. Once patients discontinue their specific osteoporosis medication, some will switch to another anti-osteoporotic treatment. However, most previous research did not address patients’ switching agents, either to a different agent within the same therapeutic class, or to an agent in a different therapeutic class. One of the few analyses addressing osteoporosis medication discontinuation reported that 30% of patients resumed their medication within six months of discontinuation, and 48% resumed their medication within one year [4].

Hence, the commonly used methods of measuring adherence, which classify patients as discontinued after a gap of at least 90 days even if they later restart the same therapy [11, 12], or exclude patients if they switch to another anti-osteoporosis medication or different dosage [13, 14], may underestimate overall treatment adherence and hinder understanding of the utilization of bisphosphonate therapy. For a study of the efficacy, effectiveness, and/or safety of a specific type, route, or dose of bisphosphonate (e.g., alendronate, risedronate, ibandronate), a narrow definition of discontinuation may be appropriate. However, research on factors associated more generally with adherence, adverse effects or effectiveness of being treated with any anti-osteoporosis medication may benefit from a broader definition. In addition, factors precipitating the discontinuation of anti-osteoporosis medications have not been studied in-depth. In the osteoporosis field, these issues are of key importance given the growing interest in a ‘drug holiday’ [15, 16] during which time patients are instructed to cease bisphosphonate therapy for a time (e.g., 1–3 years), making additional work on methods for discontinuation of anti-osteoporosis drugs important and timely. Accordingly, we conducted a study to describe discontinuation and switching patterns of anti-osteoporosis medications in the national US Medicare population, to evaluate factors associated with discontinuation using a variety of definitions, and to identify factors precipitating discontinuation of all anti-osteoporosis therapy.

We identified 30,990 Medicare beneficiaries in the 5% sample who initiated a bisphosphonate during 2007–2009 (Figure 3). The number of patients initiating each agent was 11,421 for generic alendronate, 6,173 for risedronate, 4,394 for oral ibandronate, 5,251 for branded alendronate without vitamin D, 1,794 for IV zoledronic acid, 1,532 for branded alendronate with vitamin D, and 425 for IV ibandronate. Overall, branded alendronate without vitamin D was the most commonly prescribed medication in 2007, whereas generic alendronate was the most commonly used drug in 2009.

Of the 30,990 new users of bisphosphonates identified in 2007–2009, 21,452 were eligible for inclusion in analyses of medication status at the end of 12 months of follow-up. (i.e., they had at least 12 months of coverage after the earliest treatment initiation date) (Figure 1). Of these new users followed for 12 months, 44% continued the same therapy, 36% totally discontinued without switching or restarting after a treatment gap, 8% restarted the same drug after a gap of greater than 90 days after the end of the days’ supply, and 11% switched to another bisphosphonate or non- bisphosphonate drug. By the end of 12 months of follow-up, most of beneficiaries who restarted the same drug after a treatment gap continued taking the anti-osteoporosis medications until the end of follow-up, whereas 17% of beneficiaries who switched to another drug discontinued all anti-osteoporosis medications. Of 4,738 eligible new users followed for 30 months, 19% continued the same therapy, 45% totally discontinued, 12% restarted the same drug after a gap, and 24% switched to another osteoporosis drug without a gap. By the end of 30 months of follow-up, 30% beneficiaries who switched to another drug discontinued all anti-osteoporosis medications (Additional file 1).

We found large temporal changes in the pattern of use of several types of bisphosphonates in 2007–2009 among Medicare beneficiaries. Our analysis of factors associated with different definitions of discontinuation found factors measured at baseline were, at most, weakly associated with discontinuation and differed little according to the definitions of discontinuation. However, in the case-crossover analysis evaluating the effect of factors measured immediately before discontinuation, a DXA test, having upper gastrointestinal problems, having more physician visits, higher Charlson comorbidity scores, and higher out-of-pocket drug payments, were associated with total discontinuation of bisphosphonate therapy after one year of high adherence, whereas a higher total number of all types medications dispensed was associated inversely with discontinuation.

Starting in 2008, we found the major increases in bisphosphonate utilization were for generic alendronate and IV zoledronic acid, and the major decreases were for branded alendronate products and risedronate. Our findings are consistent with a report from Medco Health Solutions, Inc (a healthcare company) indicating that branded alendronate lost 84% of its market share from retail stores and 94% of its market share from mail pharmacies during the first 30 days after the launch of generic products [27]. Several factors could explain the increase in IV/injection bisphosphonates utilization that we observed during the 2007–2009 time period. More aggressive marketing and perceptions of potentially better adherence with IV compared to oral bisphosphonates could lead to greater prescribing of IV therapies. Parenteral administration also may minimize certain side effects, such as gastrointestinal upset, and avoids the complex instructions required for proper administration of oral bisphosphonates [28].

We also found that switching within the bisphosphonates class, switching out of the bisphosphonate treatment class, and having treatment gaps were common among new bisphosphonate users. For our purposes, discontinuation was defined in two alternative ways: first, as total discontinuation of any osteoporosis medications; second, as total discontinuation, having a therapy gap then restarting the same bisphosphonate, or switching to another therapy during follow-up. Compared to the strict definition (I) of discontinuation, an additional 17-29% of people were classified as discontinued due to switching to other types of bisphosphonate or to non-bisphosphonate osteoporosis medications or due to a gap in therapy with a particular type of bisphosphonate.

Our findings are consistent with those of prior studies showing that people are prone to switch or restart bisphosphonates after discontinuation [4, 29]. Using 1996–2002 Pharmaceutical Assistance Contract for the Elderly (PACE) of Pennsylvania data, Brookhart et al. reported that among patients who stopped bisphosphonate therapy for at least 60 days, an estimated 30% restarted treatment within six months [4]. Their results are higher than the percentage restarting in our study, possibly because the prior analysis examined the therapeutic class of all bisphosphonates combined, while our analysis examined each form of bisphosphonate separately. Brookhart et al. also defined a gap as 60 days, whereas we used a gap length of 90 days. Another study, in a commercial health plan population during the period of 2006–2008, reported substantial discontinuation, restarting, and switching among new users of various anti-osteoporosis medications [29]. For example, 52-67% of new users of specific types of bisphosphonates discontinued their original medication within 12 months, 13-22% restarted their original medication after a greater than 90-day treatment gap, and 0.4-9.6% switched to another drug. Therefore, examining patterns of restarting and switching in future osteoporosis studies may generate a more complete definition and assessment of adherence, particularly in studies that include time periods when a major new anti-osteoporosis medication, such as generic alendronate, has been introduced.

Regardless of length of follow-up or the definition of discontinuation, the factors that we assessed at baseline displayed generally weak associations or no relation with discontinuation. With area under the receiver operator curve as low as 0.58-0.60 in both multivariable logistic models, these results are consistent with those of prior studies reporting that baseline characteristics, measured only at treatment initiation, yielded a poor ability to discriminate osteoporosis patients with high adherence versus low adherence [30].

Our case-crossover analysis was designed to identify factors emerging during follow-up that lead to discontinuation among adherent bisphosphonate users. The results indicated that events occurring during follow-up are likely to be more important as predictors of adherence and persistence than are characteristics measured at therapy initiation. Specifically, we identified six important factors that were positively associated with discontinuation – DXA, more than four ambulatory physician visits, hospitalization, upper gastrointestinal disorders, higher Charlson comorbidity index, and higher out-of-pocket drug payments.

DXA is the gold standard for diagnosing osteoporosis, and it may play a role in monitoring response to osteoporosis treatment [31]. Our finding that after one year of high adherence, having a DXA was significantly associated with discontinuation, suggests that physicians and patients may consider discontinuing osteoporosis medications if DXA test results indicate a bone mineral density improvement treatment [32] or that patients may discontinue the therapy due to a lack of improvement in bone mineral density [33]. Unfortunately Medicare data do not include DXA test results, which would be required to confirm this interpretation. Our observation that having more than four ambulatory physician visits was strongly associated with discontinuation contrasts with another study that reported that physician visits were related to patients’ reinitiating therapy [4]. The discordant results may reflect differences in study populations: our case-crossover analysis was limited to people who had been adherent for at least one year and then experienced discontinuation, whereas the latter study was limited to new users. Also, it is possible that patients with more diseases tend to see physicians more often and stop taking drugs with fewer perceived beneficial effects [34, 35].

Our study confirms results from other investigations reporting that upper gastrointestinal conditions are significantly associated with bisphosphonate discontinuation [1, 11]. Finally, our findings indicate that patients who have more comorbid conditions, pay more out-of-pocket for drugs, or stay in the hospital are much more likely to discontinue drug therapy. These associations are consistent with the phenomenon called the “sick stopper”, wherein patients who are getting sicker and spending more healthcare dollars maybe more likely to discontinue treatments for non-symptomatic conditions, such as osteoporosis [35].

A major strength of our study was its assessment of factors possibly associated with discontinuation both at baseline and, in the case-crossover analysis, during follow-up. Other strengths include the study’s large size, generalizability to the entire US traditional fee-for-service Medicare population with Part D coverage, the use of prescription claims to define exposures instead of self-reported drug usage, and our sensitivity analyses. The methodological aspects to our work should grow in importance given increasing interest on drug holidays from bisphosphonate exposure. We varied the definition of discontinuation across a range of possible options and examined factors associated with discontinuation, which is expected to be useful for future analyses that examine new reasons for patients to stop therapy such as a drug holidays [15, 16].

The main limitation was the study’s reliance on claims data to assess utilization and possible reasons for discontinuation. Claims data do not includes baseline or follow-up bone mineral density measures, which could have influenced the likelihood of adherence. Our case-crossover analysis was limited to patients who had been adherent to bisphosphonates for 12 months, whereas the cohort analysis used new users, therefore, the characteristics of subjects discontinuing included in the case-crossover analysis were not comparable to those totally discontinued in the cohort analysis. There also could be misclassification of safety outcomes without well validated claims-based algorithms (e.g., osteonecrosis of the jaw, atypical fracture, and upper gastrointestinal symptoms). The possibility that some patients may have filled their anti-osteoporosis prescriptions without submitting a claim to Medicare is another potential limitation. However, we recently reported that this may occur for a maximum of 15% of generic prescriptions, using alendronate as an example, available through low-cost pharmacy programs [36].

Finally, our analyses included patients who had primary adherence (i.e., they filled a prescription for a bisphosphonate), but not patients with primary non-adherence (i.e., had anti-osteoporosis prescriptions but never have it filled or administrated), and our case-crossover analysis was further restricted to patients who were adherent for 12 months after initiation of therapy. Thus, the results may not be generalizable to all Medicare patients prescribed and taking bisphosphonate drugs.

Generic alendronate and IV zoledronic use among Medicare beneficiaries increased substantially during 2007–2009. Among new bisphosphonate users, switching among bisphosphonates, switching out of the bisphosphonate therapeutic class, and having treatment gaps were common. Factors measured at therapy initiation were at most weakly associated with discontinuation, regardless of varying the definition of discontinuation and the length of follow-up. In contrast, several factors measured during follow-up appeared to precipitate discontinuation within 12 months of initiating bisphosphonate therapy including DXA, more than six ambulatory physician visits, hospitalization, upper gastrointestinal disorders, higher Charlson comorbidity score, and higher out-of-pocket drug payments. Further studies are needed to characterize medication discontinuation more carefully with consideration of time-varying factors.