Impact of pharmacy channel on adherence to oral oncolytics

Oral drug therapy use for patients with cancer has increased sharply over the last few years and this trend is expected to increase over the next several decades [1]. Experts estimate that oral medications comprise 25-35% of the total cancer drugs in development [2‐4]. In 2016, a report from IMS concluded that approximately 40% of total U.S. drug expenditures on targeted therapies were for oral formulations, up from 26% in 2010 [5].

Oral therapy provides potential advantages to patients. Oral agents may have different side effect profiles than their intravenous (IV) counterparts and patients may tolerate them better [6]. Patients often prefer oral agents to IV therapy because they offer more convenience as treatment interferes less with work and social activities, eliminating the need to travel to and from infusion clinics [7, 8]. As oncologists become more sensitive to patient preferences and quality of life, treatment options that offer more convenience and flexibility are likely to be used more often [9].

Despite the benefits of oral therapy, concerns have been raised regarding the risks of nonadherence as the responsibility of managing regimens and monitoring toxicities are centered more directly on the patient [1, 2, 10, 11]. One concern is under-dosing as a result of either nonadherence or dose reductions [12]. Under-dosing may lower drug plasma levels, thereby increasing the risk of cancer relapse or progression [13, 14]. Nonadherence may even cause clinicians to incorrectly attribute a patient’s poor response to an absence of drug activity, which may lead to unnecessary diagnostic tests, hospitalizations, and changes in the dose or regimen [9, 15]. Nonadherence is associated with an increased use of medical services [16].

Prior research has produced important findings regarding factors which adversely affect adherence in patients receiving oral oncolytics, including a lack of understanding of proper treatment administration, older age, complex dosing regimens, increased drug costs, and side effects [17‐19]. However, there is limited data concerning the benefits of customized therapy management programs and whether specialty pharmacies can play a key role by working with physicians to ensure adherence and persistence in cancer patients. The purpose of this study was to examine this issue using a sample of patients newly treated with common oral oncolytics including erlotinib, capecitabine, and imatinib from the Medco research database. Erlotinib is approved for advanced non-small cell lung cancer (NSCLC) and advanced pancreatic cancer. Imatinib is approved for adult cancer indications including Philadelphia positive (Ph+) chronic myeloid leukemia, Ph + acute lymphoblastic leukemia, and unresectable/metastatic gastrointestinal stromal tumors. Capecitabine is indicated for colorectal and metastatic breast cancers.

We used data from the Medco research database to estimate the impact of pharmacy channel on adherence and persistence among patients receiving erlotinib, imatinib, or capecitabine during 2007–2011. Our results indicate that filling prescriptions in a SP channel may lead to greater adherence compared with TR. This finding is consistent with previous research examining SPs [22‐25]. In addition, SP patients were less likely to abandon their initial prescription and remained on treatment longer versus TR. Prior studies have focused on adherence in cancer patients receiving oral therapies [11, 13, 26‐31]. However, there are limited data examining the impact of SP channel on adherence. The overall mean proportion of days covered (adherence) in the current study was 79% for the TR channel (90%, 83%, and 71% for TR patients receiving erlotinib, imatinib, and capecitabine, respectively) and 86% for SP (92%, 87%, and 81% for SP patients receiving erlotinib, imatinib, and capecitabine, respectively). Prior retrospective studies examining adherence to imatinib and utilizing prescription claims data have reported mean medication possession ratios between 73%–90% [28, 29, 31]. Mean MPRs and adherence rates for imatinib estimated as part of this current study were within the reported range of these prior studies. Among studies of adherence to breast cancer medications utilizing prescription refills, adherence rates (proportion of days covered ≥80%) over 1–2 years varied significantly, from 62%–89% [32]. Many published rates in breast cancer are based on studies assessing adherence to tamoxifen regimens, which may differ significantly from typically shorter oral chemotherapy regimens [26, 32]. A few studies assessing capecitabine adherence in patients with breast, colorectal, and GIST tumors were identified [11, 30, 33, 34]. One study was a randomized controlled trial of patients with breast cancer where adherence was calculated as the number of doses taken (as measured by a microelectronic monitoring system) divided by the number of planned doses. Average adherence of 78% was reported across all cycles [34]. Other studies assessing capecitabine used patient self-reports and noted adherence rates of 76.7% [11] and 91% [30] which were higher relative to the current study. A possible explanation for this is that Bhattacharya et al. and Winterhalder et al. relied on patient self-reports, which have a tendency to overstate compliance. In addition, the retrospective database design of the current study allowed for an observational period spanning multiple years whereas these prior studies examining capecitabine had follow-up periods of only 4–6 months [30]. The study by Walter et al., reported capecitabine adherence rates of only 61% with microelectronic monitoring system, which was more in line with the current study, compared to 99% with patient self-report. Relatively few studies examining adherence to erlotinib were identified. One study from the Netherlands utilizing patient self-reports recorded a mean MPR of 96.8% for erlotinib [35] which was comparable to the mean MPRs of 90% + reported in this study. Rates of adherence may vary across different studies due to factors such as study design, cancer type, patients’ overall disease burden, or treatment received. As part of this study, we also performed analyses in a subset of erlotinib and capecitabine patients with a cancer diagnosis recorded in the medical claims. Among the erlotinib subset, 86% had lung cancer and 14% had pancreatic cancer. Forty-seven percent of the capecitabine subgroup had breast cancer and 53% had colorectal cancer. The mean proportion of days covered within both tumor and oral oncolytic subgroups were consistent with results of the full patient sample stratified only by oncolytic received.

Our results suggest that customized therapy management programs instituted at SPs can have a positive influence on adherence and persistence in cancer patients prescribed oral oncolytics. SP therapy management programs provide additional services to patients which are likely to contribute to improved levels of adherence compared with TR. For example, when patients start new prescriptions in the SP channel, pharmacists or nurses counsel patients by telephone, often verifying that the appropriate dose has been prescribed, explaining adverse effects, potential drug interactions, dosing requirements, instructions for use, and storage requirements. Patients might also be contacted prior to each refill to monitor outcomes and encourage adherence. Some SPs even offer 24-h nursing support for patients throughout their therapy as well as home visits prior to dispensing the medication. Many SPs also have therapeutic centers dedicated to oncology with staff specializing in this area.

The level of cost sharing was also a significant factor in determining adherence to oral medication. Patients with higher copays were more likely to have lower rates of adherence and higher rates of abandonment of the index oncolytic. Many prior studies have demonstrated a clear relationship between higher copays and decreased adherence [36‐38]. Other unmeasured factors such as total household income or the proportion of out-of-pocket expenses in relation to income may have also played a role in affecting adherence. These variables were not available in the database.

This study used data from a healthcare claims database. Claims data are primarily collected for reimbursement purposes rather than research. The classification of patients into study groups and the measures derived from the data relied on inferences based upon the information appearing on the claims. Thus, there is the potential for coding inaccuracies leading to the misclassification of certain events or measures of interest. For example, the abandonment measure was inferred from the information appearing on the claims. Claim reversals were examined as an indication that the patient was having difficulty filling the prescription. The accuracy of our algorithm in identifying true instances of abandonment is not known. It is possible that some reversed claims were excluded from our dataset if the reversal occurred prior to the claims being loaded into the database. Consequentially, the abandonment rate might have been underestimated. Indeed, a prior study using nationally representative pharmacy data reported first fill abandonment rates of 6.2%, 13.5%, and 12.8% for capecitabine, imatinib, and erlotinib, respectively; rates which are higher compared to our study [39]. Finally, our adherence measure was not an actual measure of medication consumption. We assumed that patients actually took the medication after filling prescriptions.

There were differences observed in the baseline characteristics of patients filling prescriptions in the SP channel and those utilizing TR which could have influenced results. Therefore, multivariate analyses controlling for age, gender, geographic region of residence, index oncolytic, average days of supply, and copay were conducted to confirm unadjusted results showing that SP patients have higher rates of adherence (unadjusted OR = 1.95) and lower rates of abandonment (unadjusted OR = 0.40) compared with TR. In multivariate models, SP channel compared with TR was associated with significantly lower rates of abandonment (adjusted OR = 0.44) and higher rates of adherence (adjusted OR = 1.99) affirming the unadjusted results.

Our analyses were also limited by the variables available in the database. Data on overall income was not available. Hence, we could not explore the burden of patients’ out-of-pocket expenses relative to their overall income, which could have had a greater effect on adherence compared with their actual expense. We did not know whether patients were receiving assistance with copays from patient programs which could have also influenced results. A minimum post-eligibility period was not used for analyses to avoid bias from excluding patients with shorter follow-up periods due to death. We assumed death was the most likely reason for health coverage termination and that bias due to censored data would be minimal. Mortality data were not available to confirm this assumption. For most patients, we did not have access to their medical claims data. Therefore, it was inferred that patients were taking their oral chemotherapy to treat advanced cancer. An analysis was conducted to assess patient subgroups with medical claims containing a cancer diagnosis indicated for treatment with an oncolytic. Subgroup analyses were consistent with overall results. We also assumed differences in comorbidity or the severity of the patients’ cancer would be negligible across pharmacy channel cohorts and would not significantly influence abandonment or adherence metrics. Finally, it should be noted that some TR pharmacies also have adherence and specialty clinical programs; the degree to which these programs were implemented was not captured in the database. Therefore, it is difficult to determine what effect this may have had on results.

The rapid growth in the development of specialty pharmaceuticals including oral oncolytics has increasingly led to the transfer of responsibility for obtaining and administering complex and costly medications from healthcare providers and hospitals to patients in the community [40]. These changes have driven new models of care including the rise of customized therapy management programs with the aim of improving otherwise unfavorable medication use behaviors, mitigating waste, and improving outcomes [40]. The published evidence, although limited, does suggest that the majority of patient support and therapy management programs are having a positive impact [40].

Patients filling prescriptions in the SP channel were more likely to achieve adherence compared with patients using TR. Pharmacy channel also appears to be influential on abandonment, with lower rates observed among SP patients compared with TR.