The Association of Pharmacy Fill Synchronization With Breast Cancer Endocrine Therapy Adherence

Abstract

BACKGROUND:

One-third to one-half of patients prescribed adjuvant endocrine therapy are nonadherent during the recommended 5-year endocrine therapy course. This study investigated whether poor pharmacy synchronization of medication fills (requiring refills on different days) acts as a barrier to adherence.

METHODS:

A cohort of older women with stage 0 to III endocrine receptor–positive breast cancer in 2011 was identified from the Surveillance, Epidemiology, and End Result–Medicare claims-linked cancer registry. Women with endocrine therapy and at least 1 other medication fill were identified, and the 3-month synchronization of their fills was calculated as 1 minus the quotient of the number of pharmacy visits and the number of filled medications. Regression models were used to examine the association between synchronization (in quartiles adjusted for the number of medications) and adherence to endocrine therapy (defined as a medication possession ratio > 80%) over the subsequent year.

RESULTS:

During the 3 months after the first endocrine therapy prescription, the study cohort of 3212 women had a mean of 8.6 pharmacy visits (standard deviation, 4.7) with a mean synchronization of 0.3 (standard deviation, 0.2). Those in the third (odds ratio, 1.29; 95% confidence interval, 1.04–1.59) and fourth (most) medication number–adjusted synchronization quartiles (odds ratio, 1.49; 95% confidence interval, 1.19–1.86) were more likely to be adherent than those in the least. Multivariate model predictions showed that the proportion of patients who were adherent over 1 year varied from 68.9% in the least synchronized quartile to 76.6% in the most synchronized one.

CONCLUSIONS:

Prescription refill synchronization is strongly associated with adherence to endocrine therapy. Efforts to improve adherence should address this.

INTRODUCTION

As more oral oncologic agents have become available, it has become clear that many patients skip pills or stop taking them before the planned course has been completed. Research now spanning more than a decade has shown the extent of this problem in breast cancer.¹ One-third to one-half of patients prescribed adjuvant endocrine therapy (AET) with tamoxifen or an aromatase inhibitor either discontinue early or skip a substantial number of pills.^1–5

Studies in several chronic conditions outside cancer have begun to show that health care delivery systems may play an important role in nonadherence. Two such studies in primary care–focused conditions focused on the process of receiving and filling prescriptions, and these suggested that poor synchronization of medication fills, so that prescriptions are picked up on multiple days, may be a particularly underappreciated barrier to adherence.^6,7 For example, even with adjustments for other factors, patients in one study with the poorest synchronization of prescription fills visits had a 10% to 15% lower probability of adherence to lipid-lowering medications.⁶

There have been few published interventions focused on improving adherence to oral oncologic agents, and the results of most have been disappointing.⁸ Hypothesizing that synchronization might be a new target for adherence interventions in oncologic care, we examined the association of synchronization with AET adherence in a cohort of older patients with breast cancer.

MATERIALS AND METHODS

Data Source and Study Sample

The data source for the study was the Surveillance, Epidemiology, and End Results (SEER)–Medicare database. Detailed sociodemographic and clinical information, including the extent of disease at presentation, was obtained from the SEER program. Information on subjects’ cancer treatment was obtained from linked Medicare claims files. These data were further supplemented with geocoded county-level census data linked to patient information by SEER.

We included female subjects with a 2011 first incident breast cancer who had been continuously enrolled in Medicare A and B for the year before their diagnosis (to allow an assessment of comorbidity) and in Part D (pharmaceutical coverage) for 1 year after their diagnosis (Table 1). To examine the synchronization of the overall medication regimen among users of AET for breast cancer, subjects were also required to have stage 0 to III estrogen- or progesterone-positive disease, to have filled at least 1 prescription for tamoxifen or an aromatase inhibitor in the year after their diagnosis, and to have filled at least 1 other medication during the 3 months after the first endocrine therapy prescription.

TABLE 1.

Derivation of the Study Cohort

	No.
66- to 89-y-old female with first primary breast cancer identified in SEER in 2011 with Medicare A, B, and D insurance for 1 y before diagnosis	6059
Estrogen receptor– and/or progesterone receptor–positive	4906
Received first tamoxifen or an aromatase inhibitor within 12 mo of diagnosis	3496
Alive and enrolled in Medicare until end of study (15 mo after first HT or December 31, 2012)	3387
More than 1 unique drug during baseline period	3227
No missing data for key variables	3212

Abbreviation: SEER, Surveillance, Epidemiology, and End Results.

Medication Synchronization

To compute synchronization of prescription fills, we adapted the measure proposed by Choudhry et al.⁶ Specifically, we used the 3-month prescription record dating from the first endocrine therapy fill to calculate synchronization of all prescription fills from the Medicare prescription drug event file of drugs reimbursed by Medicare Part D. We defined synchronization over the 3-month period as 1 minus the quotient of the number of pharmacy visits at which any prescription was filled and the number of unique filled medication prescriptions. Because a complex relationship between the number of medications and synchronization has been reported with the use of this measure,^6,9,10 we stratified subjects into 3 groups based on the number of unique prescription medications that they filled (2–4, 5–8, and >8). We then assigned subjects in each group to quartiles based on synchronization within their group. Those in each quartile were combined with patients in the same quartile of the other groups (eg, those patients whose synchronization was in the lowest quartile of those taking 2–4 medications were combined with the patients in the lowest quartile of the 5–8 and >8 groups) to create medication number–adjusted synchronization quartiles (Supporting Fig. 1). During the same 3-month period, we also assessed the number of total medication fills of 30 or 90 days’ supply.

Patient Characteristics

Age, race (white, black/African American, Asian, or other), and ethnicity (Hispanic or non-Hispanic) were defined from SEER-Medicare information. Information about recipients of a low-income Medicare D subsidy, which included all dual-eligible Medicare-Medicaid recipients, was also obtained from Medicare files.¹¹

Comorbidity was calculated for the year before breast cancer surgery with a breast cancer–specific algorithm¹² for outpatient and inpatient diagnoses in Medicare files. Other breast cancer treatments (radiation therapy, cytotoxic therapy,¹³ and targeted chemotherapy^14,15) were assessed from Medicare claims. The SEER region was examined, and linked census information was used for the urban/rural status of the zip code of residence based on census definitions.

Adherence

To avoid measuring the exposure and outcome in the same time period, adherence to AET was calculated during the 1-year period that immediately followed the initial 3-month (baseline) period. With the AET dispensing date and the days of supply from the prescription drug event file, a supply diary was created that recorded whether a subject had AET available for each day. Medications were tracked from the initial prescription fill to account for any available pills stockpiled when patients filled prescriptions early. If a subject had an overlapping supply of more than 1 AET medication during the 12-month period (eg, she filled a prescription for letrozole before the date by which she should have used the entire supply of anastrozole pills dispensed), we considered the new medication to start on the day after the end date for the initial medication.¹⁶ Following prior literature, we considered individuals possessing enough medication for at least 80% of days to be adherent.^3,17–19

Statistical Analysis

Baseline demographics were described with means and proportions in summary statistics. On the basis of bivariate results, Hispanic ethnicity, chemotherapy receipt, and number of physician visits were not included in further analyses, and urban/rural and comorbidity categories were collapsed. Hypothesizing that greater synchronization would be associated with greater adherence, we then examined the association of synchronization with adherence in adjusted logistic regression models with the remaining variables.

In sensitivity analyses, we examined several alternative specifications for study variables and models. We repeated the models with the originally published definition of synchronization, which did not fully account for the complex correlation between synchronization and the number of prescription drugs taken.⁶ We repeated the models as well with an alternative synchronization measure that examined synchronization of only drugs that were likely to be chronic as identified from the Wolters Kluwer Medi-Span Electronic Drug File²⁰ and with all days on which subjects were hospitalized removed from adherence calculations. Finally, on the basis of evidence showing that users of initial 90-day prescriptions have higher adherence to cardiovascular medications,²¹ we also examined whether the results were changed by the inclusion of the patients’ use of 90-day fills in adherence models.

Given the inherent difficulty in interpreting the underlying coefficients from the synchronization model, we also calculated model-predicted adherence for subjects in the lowest and highest medication-stratified synchronization groups. We computed the predicted adherence at the individual level by setting the relevant variable to new values while holding all other variables constant at their original (individual) levels. Model-predicted adherence was calculated for each subject in the sample and then averaged across the total population. The predicted mean adherence for each quartile was also estimated.

RESULTS

The study cohort included 3212 women treated with endocrine therapy for breast cancer (Table 2). More than 31% were 70 years old or younger, and 38% had stage II or III disease. During the 3 months after the first endocrine therapy prescription, the mean number of unique medications filled by subjects was 7.0 (standard deviation [SD], 3.8). The mean number of prescription fills was 13.6 (SD, 9.3), and the mean number of pharmacy visits was 8.6 (SD, 4.7). The mean 3-month synchronization was 0.30 (SD, 0.2).

TABLE 2.

Characteristics of the Study Cohort (n = 3212)

Characteristic	No.	%
Age
66–70 y	1006	31.1
71–75 y	861	26.8
76–80 y	657	20.5
81–85 y	481	15.0
86–90 y	207	6.4
Race (SEER)
White	2823	87.9
Black	204	6.4
Other	167	5.2
Unknown	18	0.6
Hispanic ethnicity	213	6.6
Low-income subsidy (including dual eligibility for Medicare and Medicaid)	883	27.5
Comorbidity (≥1)	1083	53.4
No. of medications, mean (SD)	7 (3.8)
No. of pharmacy visits, mean (SD)	8.6 (4.7)
Stage
0	256	8.0
I	1663	51.8
II	972	30.3
III	247	7.7
Unknown	74	2.3
Chemotherapy	523	16.3
Aromatase inhibitor	2632	81.9
Rural/urban countya
Metropolitan area, >1 million	1656	51.6
Metropolitan area, <1 million	936	29.1
Micropolitan (urban cluster of 10,000–49,999)	194	6.0
Rural	426	13.3

Abbreviations: SD, standard deviation; SEER, Surveillance, Epidemiology, and End Results.

^aBased on census definitions for the zip code of residence.²²

During the 12-month follow-up, 72.8% of the study cohort had more than 80% adherence to endocrine therapy. As shown in Table 3, compared with those in the first (lowest) synchronization group, those in the third (odds ratio, 1.29; 95% confidence interval, 1.04–1.59) and fourth (highest) groups (odds ratio, 1.49; 95% confidence interval, 1.19–1.86) were more likely to be adherent. Race other than white or black was associated with higher adherence, but there was no difference between white and black women. Other demographic variables, the receipt of a low-income subsidy, and the cancer stage were not associated with adherence, but those with 1 or more comorbidities had lower adherence than those without comorbidities. Residence in a city of less than 1 million residents was associated with a lower probability of adherence than residence in a larger city.

TABLE 3.

Association of Medication Regimen Synchronization With Adjuvant Endocrine Therapy Adherence

	Odds Ratioa	99% CI
Medication-adjusted synchronization quartiles
1	1	Reference
2	1.22	0.97–1.53
3	1.30	1.05–1.60
4	1.49	1.19–1.87
Age
66–70 y	1	Reference
71–75 y	0.93	0.75–1.16
76–80 y	0.85	0.68–1.07
81–85 y	0.75	0.59–0.97
86–90 y	0.80	0.57–1.14
Race (SEER)
White	1	Reference
Black	0.99	0.71–1.39
Other	1.84	1.16–2.92
Unknown	0.75	0.28–2.06
Low-income subsidy (including dual eligibility for Medicare and Medicaid)	1.06	0.87–1.29
Comorbidity (≥1)	0.82	0.69–0.96
Stage
0	1	Reference
I	1.00	0.72–1.38
II	1.21	0.86–1.72
III	0.96	0.63–1.46
Unknown	0.78	0.43–1.40
Tamoxifen use (vs aromatase inhibitor)	0.87	0.70–1.10
Rural/urban county
Metropolitan area, >1 million	1	Reference
Metropolitan area, <1 million	0.76	0.62–0.94
Micropolitan (urban cluster of 10,000–49,999)	1.07	0.73–1.55
Rural	1.13	0.83–1.52

Abbreviations: CI, confidence interval; SEER, Surveillance, Epidemiology, and End Results.

^aThe odds ratios represent the odds of adherence, which is defined as possession of ≥80% of the expected medication as described in the text. Odds ratios have also been adjusted for the SEER site.

Prescription fill synchronization is strongly associated with adherence to endocrine therapy. Efforts to improve adherence should address poor prescription synchronization.

See also pages 000–000.

The association of synchronization with adherence persisted in sensitivity analyses that used actual quartiles of the synchronization measure with a separate adjustment for the number of medications, that used only chronic/maintenance medications for the synchronization measure, and that excluded hospital days from adherence calculations (Supporting Tables 1–3). Thirty-eight percent of the cohort subjects had only 30-day prescriptions (Supporting Fig. 2), and 9.6% had only 90-day prescriptions. In a sensitivity analysis that included the percentage of 90-day prescriptions, higher use of 90-day prescription fills was associated with higher adherence (odds ratio, 1.008 for each percentage point increase; 95% confidence interval, 1.005–1.011), but the association of synchronization with adherence persisted (Supporting Table 4).

To put our results into perspective, we used parameter estimates from our multivariate models to examine how adherence varied with certain characteristics. Results from these analyses revealed that an increase in synchronization from the lowest group to the highest group led to an increase in the proportion of patients who were adherent (from 68.9% to 76.6%). In a similar model that instead predicted mean adherence, the mean predicted adherence was 77.1% for quartile 1 but was 81.0% for quartile 4.

DISCUSSION

In this cohort of women with stage 0 to III breast cancer newly started on AET, greater synchronization of total medication prescription fills was associated with a higher likelihood of endocrine therapy adherence. The effect was substantial, with a nearly 8% absolute difference (77% vs 69%) in the proportion of patients who were adherent when those in the highest synchronization group were compared with those in the lowest. Sociodemographic and economic characteristics, including age and receipt of a low-income subsidy, were not associated with adherence in this cohort. Greater use of 90-day prescription fills was associated with higher adherence but did not mediate the association of adherence with synchronization.

Our study contributes to the literature by showing several potentially remediable health delivery–related factors that play a role in nonadherence to AET. Prior studies of cancer medication delivery have primarily focused on costs.^23,24 With the advent of generic alternatives to aromatase inhibitors, cost barriers actually appear to be decreasing. Several studies in the 2000s showed that out-of-pocket costs were a barrier to adherence, but recent evidence suggests that adherence improved when manufacturers released generic aromatase inhibitors in 2011.^25,26 Our findings regarding both medication synchronization and use of 90-day prescriptions suggest that efforts to address other factors that are related to health delivery could also help improve cancer care.

Our results are also notable given the current literature regarding the other most commonly cited barriers to adherence.^23,24 According to a number of studies,^1,27–32 side effects play a large role in nonadherence to endocrine therapy. There are now a number of guideline-recommended treatments for those side effects,^33,34 but studies of these treatments have not yet reported improvements in adherence. Differences between physicians and patients in perceptions of medicines may also pose important barriers to adherence: patients in 1 large AET study reported that an important barrier to continuing AET was “I was not sure if it (the medication) was helping,”³⁵ and a meta-analysis of 5 randomized studies showed that educational interventions to address such issues had little impact on adherence.³⁶ Our study, therefore, adds a potential new target to otherwise disappointing efforts to improve adherence to oral cancer medications.

A few studies outside oncology have examined whether improving synchronization can improve adherence. Two studies in regional pharmacy chains^37,38 recently showed that patients who enrolled in a refill synchronization program improved their odds of adherence by more than 3 times in comparison with controls who did not choose to enroll. A synchronization study at a national chain that used a stronger wait-list design found smaller effects (a 7.5% improvement in adherence in the intervention group vs a 4.7% improvement in the wait-list group),³⁹ and this suggests that further research is needed. To our knowledge, our study extends that work into oncologic care for the first time. However, because even small improvements in adherence could have large clinical significance and because there is minimal corresponding risk, our study suggests that oncology teams might wish to aid patients in seeking out pharmacy synchronization programs. In 2015, 23 pharmacy chains and approximately 2000 independent pharmacies offered some synchronization programs.⁴⁰ Unlike many other oral cancer agents that are delivered through specialty pharmacies, oral endocrine therapy is primarily provided by commercial pharmacies,⁴¹ so oncologists and/or primary care physicians could recommend these currently available programs to patients.

Several limitations merit comment. Given the observational study design, we could not account for all variables that might affect the relationship between synchronization and adherence or the possibility that those who are more likely to adhere are also more likely to arrange more synchronized prescription fills. However, the evidence from cohort studies^37,38 and one controlled trial³⁹ outside cancer suggest that synchronizing medications is likely to be an effective intervention. There are few reported descriptions of synchronization estimates in the literature, so we were unable to compare the findings for our cohort with those for other cohorts; future studies should examine the range of synchronization across cohorts. Other factors that may add to or reduce the complexity of interacting with the pharmacy, including mail-order options and side effects of medications, also could not be fully addressed. However, only 9% of Medicare patients used mail order during our study period,⁴² and one more recent study of other chronic medications found that synchronization was strongly associated with adherence even among mail-order users.⁷

In summary, better synchronization of medication fills is strongly associated with better adherence to oral endocrine therapies for breast cancer. Our results provide evidence that, even with the motivation of preventing a cancer recurrence, modifiable health delivery factors remain a barrier to appropriate medication use. Our study also provides an important target for future efforts to improve adherence to oral medications in patients with cancer. With the growth in oral anti-oncologic agents, interventions that simplify patients’ experiences are needed to foster optimal cancer outcomes.