Gaps in Treatment Among Users of Osteoporosis Medications: The Dynamics of Noncompliance

Article Outline

• Abstract
• Methods
  • Data
• Statistical Methods
• Results
• Discussion
• References
• Copyright

Abstract 

Purpose

Cyclical patterns of compliance have been observed with many health-related activities such as dieting and exercise. It is not known whether such patterns of compliance exist among users of chronic medications. We sought to estimate the percentage of patients who restart osteoporosis therapy after a prolonged lapse in medication use and to identify the factors associated with a return to compliance.

Methods

We studied 26,636 new users of an osteoporosis medication (alendronate, calcitonin, estrogen, raloxifene, or risedronate) who were age 65 or older and had an extended lapse in refill compliance, defined as a period of at least 60 days after the completion of one prescription in which no refill for any osteoporosis medication was obtained. Survival curves were used to estimate the length of time until therapy is resumed. We estimated the association between patient characteristics and the rate of resuming treatment using Cox proportional hazards analysis. We then conducted a case crossover analysis to examine whether certain events occurring during follow-up triggered a return to refill compliance.

Results

Of patients who stopped therapy for at least 60 days, an estimated 30% restarted treatment within 6 months, and 50% restarted within 2 years. Among patients who had at least 6 months of continuous use before their interruption in treatment (n=5863), 42% restarted therapy within 6 months and 59% within 2 years. Younger patients, women, and those with a history of a fracture were more likely to return after a break in medication use. Recent hip fractures, discharges from nursing homes, and bone mineral density testing also predicted a return to treatment.

Conclusion

Extended gaps in treatment are common among users of osteoporosis medications. Because the effectiveness of these drugs used in an interrupted way is unknown, compliance interventions should emphasize the need for continuous medication use. Further research is needed to understand why patients often go for months without refilling prescriptions and also whether similar utilization patterns exist for other chronic medications.

Keywords: Osteoporosis, Medication compliance, Adherence, Persistence, Drug holiday

A substantial body of research has found that patient compliance with long-term therapeutic regimens is poor.¹ Sub-optimal compliance with effective treatments for chronic disease is thought to be responsible for considerable morbidity and associated costs.², ³ If unrecognized, poor compliance can be mistaken for treatment failure and can lead to inappropriate dosage increases or unnecessary treatment changes.⁴ As the public health, clinical, and economic consequences of poor medication compliance have become better understood, there has been increasing interest in the development of interventions to promote patient compliance with effective therapeutic regimens.⁵, ⁶, ⁷ A thorough understanding of common patterns of drug use is a necessary first step in the design of an effective intervention to improve use of prescription medications.

In many studies of long-term medication use, non-compliance is either explicitly or implicitly treated as an end-point. These analyses commonly report the percentage of patients still compliant with a therapeutic regimen at various time points or conduct analyses examining associations between patient characteristics and the probability of stopping therapy.⁸, ⁹, ¹⁰, ¹¹, ¹², ¹³, ¹⁴ While these analytic approaches have led to important characterizations of the problem of non-compliance, they are not designed to reveal dynamic patterns of medication use, such as patients stopping and then restarting therapy.

Cyclical patterns of compliance have been observed with other health-related activities; well-known examples include dieting and exercise. In controlled settings in which medication use is electronically monitored, compliance has been found to be erratic.¹⁵ In such contexts, the term “drug holiday” has been used to describe an interruption in medication use lasting more than 3 days.¹⁵, ¹⁶ However, relatively little is known about the extent to which patients in routine care stop and restart drug therapies for chronic diseases.

In a previous study of compliance among new users of osteoporosis medications, we found that 1 year after initiation of therapy, 45% of patients had a period of ≥120 days in which their prescription was not refilled.⁸ The present study was conducted within this same cohort to understand whether or not patients who discontinue osteoporosis medication for an extended period ultimately resume medication use.

Back to Article Outline

Methods 

Data 

We studied Medicare beneficiaries aged ≥65 years who were concurrently enrolled in the Pharmaceutical Assistance Contract for the Elderly (PACE) of Pennsylvania. To be eligible for PACE, a participant’s annual income must be ≤$17,700. The program reimburses the cost of all prescription medications with a co-payment of ≤$9. There are no prescribing restrictions on medications used for osteoporosis. Medication information from PACE includes the drug name, dosage, number of pills dispensed, and days supplied. The study database also includes Medicare information on all inpatient and outpatient encounters, including diagnoses, procedures, and tests ordered.

Our study was conducted within an existing retrospective cohort of new users of osteoporosis medication who initiated treatment between 1996 and 2002.⁸ The medications that we studied were bisphosphonates, calcitonin, estrogen therapy (except vaginal creams), and raloxifene. Teriparatide was not included because it only became available in the last year of the study. Vitamin D preparations were also excluded because they are often used for other bone diseases. To reduce the likelihood that patients were obtaining medicines through other programs, we required patients to have filled at least one prescription through PACE in each of the two 6-month intervals preceding the filling of their initial osteoporosis prescription. Patients were followed until they became ineligible for PACE, died, or reached the administrative end of follow-up, which occurred on December 31, 2002.

Follow-up time was broken into discrete 60-day intervals. Within each 60-day period for each patient, we computed the proportion of days covered by osteoporosis medications using the dates on which prescriptions were filled and the “days supply” field in the pharmacy claim. If a prescription was refilled before an existing one was completed, the new prescription was assumed to start on the day the previous one should have ended. Within each 60-day interval, the proportion of days covered was calculated by summing the number of days in each interval covered by a prescription for an osteoporosis medication and dividing by the number of days in the interval in which the subject was not hospitalized or in a nursing home.

We restricted the present analysis to the members of the cohort who experienced an extended lapse in refill compliance. This occurred when a patient experienced a 60-day interval in which no days were covered by an osteoporosis medication. Among these patients, we defined the index date as the start of the 60-day period in which no days were covered by an osteoporosis medication. A return to medication use was defined by the filling of any prescription medication used to treat osteoporosis. This allowed patients to switch treatments and still be considered compliant.

The study investigators have Data Use Agreements in place with Center for Medicare and Medicaid Services and PACE. The Partners Healthcare Institutional Review Board approved this research.

Back to Article Outline

Statistical Methods 

We used the Kaplan-Meier method¹⁷ to estimate the survival distribution of the time until a patient returned to refill a medication used for treating osteoporosis and Cox proportional hazards models¹⁸ to estimate the hazard of resuming therapy using baseline variables defined during the year before initiation. For both analyses, subjects were censored by loss of PACE eligibility, death, or the end of follow-up. Baseline variables included in the Cox model were age, sex, number of co-morbid conditions, number of medications used, and history of the following in the year before initiation of an osteoporosis drug: a fracture, a nursing home visit, an acute care hospitalization, or a bone mineral density test.

We then performed a case-crossover analysis¹⁹, ²⁰ to determine if particular events occurring during the follow-up period triggered transitions back to compliance. The case-crossover approach stratifies the analysis across individuals, so that each patient serves as his or her own control (Figure 1). This design removes the confounding effects of all patient-level variables that are constant across time. The events that we considered were the occurrence of a new fracture, a bone mineral density test, acute care hospitalization, or a nursing home stay. The case-crossover analysis was implemented by comparing the frequency of events in the 60 days immediately before a transition in medication use (the hazard period) with the frequency of events in the period 120 days to 60 days before the transition occurred (the control period). This analysis was necessarily restricted to subjects who had at least 2 consecutive 60-day periods with no days covered by an osteoporosis medication. Because estrogen therapy may be used for indications other than osteoporosis, we repeated all analyses, excluding patients who started on estrogen. We conducted an additional sensitivity analysis in which “stopping” was defined to be 120 days not covered by medication, rather than 60 days. In a final sensitivity analysis, patients were censored by nursing home admission. All statistical analyses were performed in SAS 9.1.²¹

• 
  • View Large Image
  • Download to PowerPoint

• Figure 1. 

  Diagram of case-crossover analysis of events predicting a return to medication use. The shaded area represents time when patient is on medication.

Back to Article Outline

Results 

We identified 40,002 patients who initiated a medication used for osteoporosis during 1996-2002; this cohort has been described in previous work.⁸ Of these new users, 26,636 (67%) had at least one 60-day period in which no days were covered by a prescription for any osteoporosis medication. The characteristics of this sample are given in Table 1. These patients had an average age of 82 years and were predominately white and female, had an average of 3.5 co-morbid conditions, and used 9 different medications. Slightly over 50% had an acute care hospitalization, and 20% had a nursing home stay during the year before initiation. Approximately 25% had a fracture, and a similar proportion underwent bone mineral density testing during the year before the start of treatment. Among these patients, 11,857 (45%) ultimately returned to medication use, 2994 (11%) were censored by death, 3908 (15%) were censored by loss of PACE eligibility, and 7877 (30%) reached the end of follow-up without returning to regular use.

Table 1. Baseline Characteristics of Subjects in Sample

	n (%) or Mean (±SD)
n	26,636
Female sex	25,660(96.3)
Age, years	81.8(±6.8)
White race	25,550(95.9)
Number of co-morbid conditions	3.5(±3.1)
Number of different medications	9.1(±5.4)
Number of physician visits	10.1(±6.9)
Time on medication before interruption (in months)	9.6(±12.9)
Acute care hospitalization	14,178(53.2)
Nursing home residence	5271(19.8)
Fracture of the hip, wrist, radius, or spine	6559(24.6)
Bone mineral density testing	6338(23.8)
Starting medications
Bisphosphonate	11,459(43.0)
Calcitonin	8691(32.6)
Estrogen	4054(15.2)
Raloxifene	1431(5.4)
Bisphosphonate and Calcitonin	608(2.3)
Other combinations	393(1.5)

In Figure 2, we present Kaplan-Meier estimates of the survival function for time until a return to therapy. An estimated 30% of the population returned to regular use within 6 months after initially discontinuing medication and 50% by 2 years. In Figure 3, we present Kaplan-Meier estimates stratified by length of time on medication before the interruption in treatment. Patients with a longer history of medication use returned to treatment at a higher rate. Six months after an interruption in treatment, 42% of patients with 6 months or more of regular use returned to refill a prescription, and 59% returned by 2 years.

• 
  • View Large Image
  • Download to PowerPoint

• Figure 2. 

  Kaplan-Meier estimate of the cumulative probability of returning to treatment.

• 
  • View Large Image
  • Download to PowerPoint

• Figure 3. 

  Kaplan-Meier estimate of the cumulative probability of returning to treatment (stratified by length of time on medication before interruption).

Table 2 presents the results of the Cox proportional hazards model examining the association between patient characteristics and the rate at which medication use was resumed. Younger female patients with a history of fracture and few co-morbid conditions were the most likely to return to treatment. Consistent with the Kaplan-Meier analysis, increasing the length of time spent on a medication before the interruption increased the probability of an ultimate return to compliance, with each additional 60-day period spent on medication increasing the probability of return by 3%. Patients who initiated estrogen therapy were the least likely to return to treatment after an interruption. Patients who started on calcitonin or a combination therapy were more likely to return than patients on a bisphosphonate. In Table 3, we present the results of the case-crossover analysis. A fracture or the use of a bone mineral density test was associated with a higher likelihood that a patient would re-start therapy in the subsequent 60-day period. Conversely, nursing home stays were associated with a lower likelihood of returning to therapy.

Table 2. Cox Proportional Hazards Analysis of Baseline Variables Associated with a Return to Medication Use

Variables Assessed at Initiation of Therapy	Hazard Ratio	95% LCL	95% UCL
Age in decades	0.66	0.65	0.68
History of fracture in past year	1.18	1.12	1.23
History of bone mineral density testing in past year	1.02	0.97	1.06
History of a nursing home visit in past year	1.04	0.98	1.10
History of hospitalization in past year	1.01	0.97	1.05
Female sex	1.63	1.45	1.84
Number of physician visits (per 5 visits)	1.04	1.03	1.06
Number of comorbid conditions (per 5 conditions)	0.71	0.69	0.74
Number of other prescription medications (per 5 medications)	1.01	0.99	1.03
Length of time spent on osteoporosis medication	1.03	1.02	1.03
Type of osteoporosis medication used⁎
Calcitonin	1.22	1.16	1.27
Estrogen	0.67	0.63	0.71
Raloxifene	0.93	0.85	1.01
Bisphosphonate + Calcitonin	1.20	1.06	1.35
Other combination therapy	1.17	1.01	1.34

LCL = lower confidence limit; UCL = upper confidence limit.

Bisphosphonate is the reference category.

Table 3. Results from Case-crossover Analysis of Events Predicting a Return to Medication Use

To assess whether treatment interruptions were due to medication switching, in Table 4 we present a cross-tabulation of drugs used before and after an observed interruption in treatment. For the majority (68%) of the 11,884 patients who were observed to return to treatment, the agent used after an interruption was the same as the agent used before the interruption. Patients on combination therapies often returned to use a single agent.

Table 4. Cross-tabulation of the Drugs Used Before and After the Break in Therapy

In the sensitivity analysis in which estrogen users were excluded, the results were almost identical. For example, we estimated that 51% of patients in this population return to treatment by 2 years (compared with 50%). When we re-defined “stopping” to be 120 days without a refill, many people still returned to treatment, but the estimated probability of return was substantially smaller (38% by 2 years). When patients were censored by nursing home admission, slightly more were estimated to return to use (56% by 2 years), but the results from the regressions were largely unchanged.

Back to Article Outline

Discussion 

Within a population of moderate- to low-income older Americans, we found that long breaks in treatment were common among new users of osteoporosis medication. We previously reported that within 2 years of treatment initiation, a majority of new users experienced a period of at least 120 days after the completion of a prescription in which no refill was obtained.⁸ In the present study, we found that many of the same individuals that stopped refilling medication ultimately returned to treatment. Although many previous studies have reported that patient compliance with therapeutic regimens is poor, to our knowledge this is one of the few reports suggesting that many patients return to refill prescriptions after long periods of apparent nonuse.

The process of patients returning to therapy is associated with a variety of subject characteristics and events. Female patients and those with a history of fracture were the most likely to return to refill prescriptions. Patients with a prior fracture have a strong indication for osteoporosis treatment and possibly viewed themselves at particular risk. However, older patients and those with many co-morbid conditions were less likely to return to treatment. For this frail group, it is possible that preventive therapies were a lower priority than treatment for other more acute medical conditions. Several factors during the period of medication nonuse were strong predictors of returning to therapy, including a new fracture and bone mineral density testing. Similar to a fracture during the baseline period, the occurrence of a fracture after discontinuation presents a compelling reason for a patient to return to therapy and puts the patient in contact with the health care system. Use of bone mineral density testing was possibly a marker for concern about osteoporosis by the physician or the patient.

Recent nursing home stays were negatively associated with a return to therapy. This finding is consistent with other studies that have observed that few patients receive osteoporosis treatment while in nursing homes.²², ²³ The low use of osteoporosis treatment in nursing homes may reflect a reluctance of nursing home staff to use nonessential treatments or possibly a concern about the use of bisphosphonates in patients who may be supine most of the day. However, it underscores the fact that many patients at high risk of fracture are likely to have stopped treatment in this environment.

We have conceptualized compliance as a dichotomous state, but medication use is complex. Patients who appear to be completely noncompliant may be taking their medications continuously but infrequently. This is supported by the observations that many people who appear to have stopped return immediately and that fewer patients return when a break in treatment is defined to be 120 days. Nevertheless, the strong association between events such as an incident fracture and a new refill strongly suggests that compliance behavior can change abruptly. For some patients, however, the behavior change may be from “under use” to “regular use”, rather than from “non-use” to “use.”

We acknowledge other limitations of our study related to our use of administrative health care utilization data. First, we are unable to assess the reasons for starting or stopping a medication. Thus, some of the lapses in treatment that we have observed could be appropriate and physician-directed. For example, the interruption could represent a switch in therapy because of an adverse event or concern about one. Or it is possible that a physician concerned about a potential adverse reaction might suspend treatment rather than immediately switch a patient to a new drug. Second, we do not capture data on IV bisphosphonates or over-the-counter calcium and vitamin D. Patients who are temporarily switched to one of these therapies would be assumed to have experienced a gap in treatment. Finally, it is possible that some of the interruptions in use that we have observed could have resulted from patients developing stockpiles at home (eg, from physician samples). Future research in this area should focus on developing a greater understanding of the reasons why patients often go for extended periods of time between refills.

For those involved with efforts to improve patient compliance with medication, our results yield insight into the process of compliance, identify particular patient groups most likely to return from temporary interruptions in therapy, and also point to events that predict transitions in medication use. These findings also have specific relevance to physicians treating patients for osteoporosis; namely, that prolonged breaks from osteoporosis medications are probably common. Because the effectiveness of osteoporosis medications used in an interrupted way is unknown, physicians should stress to their patients the importance of continuous medication use. When persistently low bone mineral density test results are observed, physicians should attempt to determine whether the patient has been consistently using their prescribed medication before changing treatment or dosing.

Although our study points out an underappreciated problem with long-term medication use, it also gives some reasons for optimism: patients frequently return to refill prescriptions after long episodes of apparent noncompliance. Cross-sectional estimates of compliance based on refill data may paint an overly pessimistic picture of long-term medication use. These estimates will miss the potentially numerous patients who have not had their prescription refilled recently but who will ultimately return to regular use. Although it is clear that long-term compliance with osteoporosis treatment is poor by any measure, our study suggests the possibility that there are many patients who, if properly motivated, could become more regular medication users.

Back to Article Outline

References 

1. Haynes RB, Taylor WD, Sackett DL. Compliance in Health Care. Baltimore, MD: The Johns Hopkins University Press; 1979;
   • View In Article
2. Osterberg L, Blaschke T. Adherence to medication. N Engl J Med. 2005;353:487–497
   • View In Article
   • CrossRef
3. Hughes DA, Bagust A, Haycox A, Walley T. The impact of non-compliance on the cost-effectiveness of pharmaceuticals: a review of the literature. Health Econ. 2001;10:601–615
   • View In Article
   • MEDLINE
   • CrossRef
4. Stephenson J. Noncompliance may cause half of antihypertensive drug “failures”. JAMA. 1999;282:313–314
   • View In Article
   • MEDLINE
   • CrossRef
5. McDonald HP, Garg AX, Haynes RB. Interventions to enhance patient adherence to medication prescriptions: scientific review. [see comment] [erratum appears in JAMA. 2003 Jun 25;289:3242] JAMA. 2002;288:2868–2879
   • View In Article
   • MEDLINE
   • CrossRef
6. Haynes RB, McDonald H, Garg AX, Montague P. Interventions for helping patients to follow prescriptions for medications [update of Cochrane Database Syst Rev. 2000(2):CD000011; PMID: 10796686]. Cochrane Database Syst Rev. 2002(2):CD000011.
   • View In Article
7. Sabate E. Adherence to long-term therapies: evidence for action. World Health Organization [Web report]. 2005;Available at: http://www.who.int/chronic_conditions/en/adherence_report.pdf. Accessed August 10
   • View In Article
8. Solomon DH, Avorn J, Katz JN, et al. Compliance with osteoporosis medications. Arch Intern Med. 2005;165:2414–2419
   • View In Article
   • MEDLINE
   • CrossRef
9. Clowes JA, Peel NF, Eastell R. The impact of monitoring on adherence and persistence with antiresorptive treatment for postmenopausal osteoporosis: a randomized controlled trial. J Clin Endocrinol Metab. 2004;89:1117–1123
   • View In Article
   • MEDLINE
   • CrossRef
10. Ellis JJ, Erickson SR, Stevenson JG, Bernstein SJ, Stiles RA, Fendrick AM. Suboptimal statin adherence and discontinuation in primary and secondary prevention populations. J Gen Intern Med. 2004;19:638–645
    • View In Article
    • MEDLINE
    • CrossRef
11. Yang CC, Jick SS, Testa MA. Discontinuation and switching of therapy after initiation of lipid-lowering drugs: the effects of comorbidities and patient characteristics. Br J Clin Pharmacol. 2003;56:84–91
    • View In Article
    • MEDLINE
    • CrossRef
12. Grant RW, O’Leary KM, Weilburg JB, Singer DE, Meigs JB. Impact of concurrent medication use on statin adherence and refill persistence. Arch Intern Med. 2004;164:2343–2348
    • View In Article
    • MEDLINE
    • CrossRef
13. Benner JS, Glynn RJ, Mogun H, Neumann PJ, Weinstein MC, Avorn J. Long-term persistence in use of statin therapy in elderly patients [see comment]. JAMA. 2002;288:455–461
    • View In Article
    • MEDLINE
    • CrossRef
14. Chapman RH, Benner JS, Petrilla AA, et al. Predictors of adherence with antihypertensive and lipid-lowering therapy. Arch Intern Med. 2005;165:1147–1152
    • View In Article
    • MEDLINE
    • CrossRef
15. Urquhart J. Erratic patient compliance with prescribed drug regimens: target for drug delivery systems. Clin Pharmacol Ther. 2000;67:331–334
    • View In Article
    • MEDLINE
    • CrossRef
16. Urquhart J. The electronic medication event monitor (Lessons for pharmacotherapy). Clin Pharmacokinet. 1997;32:345–356
    • View In Article
    • MEDLINE
    • CrossRef
17. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc. 1958;53:457–481
    • View In Article
18. Cox DR. Regression models and lifetables. J R Stat Soc Ser B. 1972;187–220
    • View In Article
19. Maclure M. The case-crossover design: a method for studying transient effects on the risk of acute events. Am J Epidemiol. 1991;133:144–153
    • View In Article
    • MEDLINE
20. Wang PS, Schneeweiss S, Glynn RJ, Mogun H, Avorn J. Use of the case-crossover design to study prolonged drug exposures and insidious outcomes. Ann Epidemiol. 2004;14:296–303
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (152 KB)
    • CrossRef
21. SAS, Version 9.1 [computer program]. Cary, NC: SAS Institute Inc.; 2003.
    • View In Article
22. Jachna CM, Shireman TI, Whittle J, Ellerbeck EF, Rigler SK. Differing patterns of antiresorptive pharmacotherapy in nursing facility residents and community dwellers. J Am Geriatr Soc. 2005;53:1275–1281
    • View In Article
    • MEDLINE
    • CrossRef
23. Gupta G, Aronow WS. Underuse of procedures for diagnosing osteoporosis and of therapies for osteoporosis in older nursing home residents. J Am Med Dir Assoc. 2003;4:200–202
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (81 KB)
    • CrossRef