Lung Cancer: Copayments and Behavior Following Erlotinib Formulary Tier Change

Lung cancer is the second-most common form of cancer in both men and women¹ and is the leading cause of cancer-related mortality in the United States.² It was estimated that more than 219,440 new cases of lung disease were diagnosed in 2009, with 159,690 deaths.¹ Of the lung cancers, approximately 80% are non—small cell lung cancer (NSCLC) types.³ For patients with advanced NSCLC (stages IIIb and IV) for whom radiation and/or surgical resection are no longer a viable option, first-line treatment involves intravenous chemotherapy agents and may be combined with radiotherapy and/or other targeted biologic agents.⁴ However, approximately 30% to 40% of patients with advanced NSCLC fail to respond adequately to first-line chemotherapy and are candidates for second-line therapy.⁵ For patients who have locally advanced or metastatic NSCLC and have had failure of at least 1 prior chemotherapy regimen, erlotinib (Tarceva; Genentech, Inc, and OSI Pharmaceuticals, Inc) was approved in November 2004 as monotherapy.⁶ Erlotinib is an oral epidermal growth factor receptor tyrosine-kinase inhibitor (TKI), and in clinical trials, resulted in a 2-month increase in survival for lung cancer patients.⁷

With the use of self-administered oral chemotherapy drugs, concern has been raised regarding patient adherence and persistence.⁸ In particular, underdosing due to either nonadherence or dose reductions/delays may result in lower plasma drug levels and subsequently increased risk of cancer relapse and progression.^9,10 Furthermore, nonadherence has been associated with higher medical service utilization.¹¹ Factors that have been found to modify medication adherence include patient characteristics (female sex¹² and younger age^13-16) and comorbid conditions (increased depression symptoms¹³ or higher Charlson Comorbidity Index score¹⁵), medication dosing12 and type,14 clinician drug choice and patient education,¹⁴ and financial aspects of the healthcare system^12,14,16 including higher out-of-pocket expenses^13,15 and copayment.¹⁷ However, it is unclear whether these factors affect patients’ adherence to treatment for diseases such as lung cancer with high and rapid mortality as much as they influence medication use for the treatment of chronic diseases.

Most oral chemotherapy adherence research has been conducted on either breast cancer patients^15,16 or chronic myeloid leukemia patients.^10-12,18 In one study of NSCLC patients, Greer and colleagues¹⁹ identified smoking behavior and anxiety as predictors of infusion-based chemotherapy regime dose delays/reductions; for the 14% of their patients who were also receiving an oral epidermal growth factor receptor TKI, adherence data were not presented. Furthermore, Ramsey and colleagues²⁰ noted in their evaluation of lung cancer chemotherapy use within a large US insurance claims database that erlotinib prescriptions were increasingly filled and that erlotinib was the second most commonly used second- and third-line treatment for refractory lung cancer. These researchers also evaluated cost utilization and grouped oral erlotinib medication costs with other outpatient pharmacy costs. Erlotinib therapy resulted in pharmacy costs of $3888 and $1290 per-patient-per-month for second- and third-line outpatient therapy, respectively.

During the November 2004—December 2006 time frame of the study by Ramsey and colleagues,²⁰ erlotinib was classified as a tier 3 drug on the prescription drug list for healthcare plan insurance reimbursement. However, this healthcare plan moved erlotinib to tier 2 starting on September 1, 2007.

This natural experiment provided the opportunity to examine whether the high copayments and other aspects of financial burden^12-16 that have been associated with adherence in other medication studies also affect adherence with erlotinib for patients with NSLC. The purpose of this retrospective study was to examine the impact of the erlotinib tier change on patient copayments and its impact on treatment behavior among patients with lung cancer.

PATIENTS AND METHODSStudy Design and Eligibility Criteria

This retrospective analysis used medical and pharmacy claims to identify patients from a large US healthcare claims database affi liated with OptumInsight who had evidence of lung cancer and had fi lled prescriptions for erlotinib. Medical and pharmacy claims, as well as enrollment information from May 1, 2004, to March 30, 2009, were used. A variable follow-up period (until the earliest of 1 year following index date, disenrollment, death, or March 30, 2009) was used to assess copayments, discontinuation, and adherence with erlotinib. The pharmacy benefit change that moved erlotinib from tier 3 to tier 2 occurred on September 1, 2007. A 6-month baseline period (baseline period) before the first filled prescription date (index date) was used to identify patient characteristics. The follow-up period was from the index date until either disenrollment from the health plan or March 30, 2009, whichever occurred earlier, and was used to assess erlotinib use and outcomes (

Figure 1

).

Eligibility Criteria

Eligible patients had erlotinib prescriptions filled from November 1, 2004, through December 31, 2008. Patients were also required to have continuous eligibility in a commercial healthcare plan with both medical and pharmacy benefits during the baseline period through the follow-up period. Patients were 18 years or older with no evidence of erlotinib use during the baseline period. Diagnostic evidence of lung cancer in the medical claims was based on International Classification of Diseases, Ninth Revision, Clinical Modification codes occurring at any time during the study period. Patients were further required to have either at least 2 qualifying medical claims for lung cancer with codes of 162.2-162.9 (malignant neoplasm of bronchus and lung) or 163.x (malignant neoplasm of pleura) in the primary position on separate service dates within the study period, or at least 1 qualifying medical claim for lung cancer with codes of 162.2-162.9 or 163.x in the primary position plus at least 1 qualifying medical claim for lung cancer with codes 231.2-231.9 (carcinoma in situ of bronchus and lung—other specified site, including pleura or unspecified site) in the primary position. All qualifying claims were required to occur on separate service dates within the study period. Patients were required to have evidence of chemotherapy use during the baseline period based on appropriate Healthcare Common Procedure Code System chemotherapy codes (

eAppendix,

available at www.ajpblive.com). Laboratory claims either at laboratory sites or for laboratory services were excluded.

Treatment Cohorts

Patients were categorized into 3 analytic cohorts based on the period of time when they initiated and ended erlotinib use. Patients who started (index date) and ended use of erlotinib before September 1, 2007, were designated as pre-tier patients. Patients with index dates occurring after September 1, 2007, were designated as post-tier patients. Patients starting use of erlotinib before September 1, 2007, and continuing erlotinib use beyond this same date were designated as crossover patients.

Outcomes Measures

Patient demographics including age, sex, geographic location, and enrollment dates were used. Clinical characteristics were evaluated and included the following: Charlson Comorbidity Index score; diagnosis of additional disease and tumor types at baseline and follow-up; use of radiation, surgery, or other chemotherapy at baseline and follow-up; and mortality.

Erlotinib copayment was defined as the mean monthly prescription copayment cost (per 30 days) and was calculated as the total out-of-pocket cost divided by total number of 30-day prescriptions (accounting for mail order). For prescription counts, total number of prescriptions was counted, with prescriptions filled by mail order credited as multiple prescriptions based on days of supply.

Erlotinib adherence was defi ned as the total days of supply of all prescription fi lls between the first and last fill divided by the number of days between the first and last fill; adherence values of >80% were used as a cutoff value (commonly used in the literature^21,22). This definition excluded patients with just 1 prescription fill. Erlotinib persistence (or discontinuation) was defined as a 60-day gap in therapy between the run-out date of the medication fill and the subsequent refill.

Statistical Analysis

Descriptive statistics were generated for baseline patient characteristics stratified by tier-change cohort. Comparisons of baseline characteristics among tier-change cohorts were assessed using the t test for continuous variables and the χ2 test for categorical variables. Erlotinib mean copayments and mean adherence rates in the pre-tier and post-tier cohorts were compared by using t tests; Wilcoxon rank-sum tests were used to compare median copayments. Paired t tests were used to compare mean copayments, and Wilcoxon signed-rank tests were used to compare median copayments for the crossover cohort before and after the tier change. Rates of treatment discontinuation were compared by using χ2 tests.

Due to small sample size and difficulty in determining intake of oral doses before and after tier change, the differences within the crossover cohort for erlotinib use were not determined. The dichotomous measure of adherence (>80%) to treatment was assessed using multivariate logistic regression; persistence was assessed using Cox proportional hazards modeling. Both compared pretier and post-tier cohorts on risk of treatment use (either adherence or discontinuation) and were adjusted for age, sex, Charlson Comorbidity Index score, and prior surgery or radiation. For the primary analysis, the primary independent variables of interest were the cohort (pre-tier and post-tier) and patient copayment for erlotinib.

Sensitivity Analysis

For sensitivity analysis, the sample was restricted to patients with similar follow-up times and included patients with 1 or more prescriptions. Prescription use (an alternative adherence measure) was defined as total supply divided by length of follow-up with adherence values of >80% used as a cutoff. Multivariate logistic regression was used to assess prescription use, and treatment discontinuation was assessed using Cox proportional hazards modeling. Both models compared cohorts on risk of treatment utilization (use or discontinuation) and were adjusted for age, sex, Charlson Comorbidity Index score, and prior surgery or radiation. For the sensitivity analysis, the independent variable of interest was the cohort (pre-tier and post-tier).

RESULTSPopulation Demographics

We identified 1460 patients with lung cancer who had filled at least 1 erlotinib prescription from November 1, 2004, through December 31, 2008. Of these, 949 patients received erlotinib before the tier change (pre-tier cohort), 410 received erlotinib after the tier change (post-tier cohort), and 101 patients received erlotinib both before and after the tier change (crossover cohort).

As shown in

Table 1

, patients in the post-tier cohort were slightly older compared with the pre-tier cohort patients (60.91 years vs 59.49 years; P = .0212), although the difference in distribution of age groups was not significant. Sex distribution was similar between the pre-tier and post-tier cohorts. Patients in the post-tier cohort had significantly fewer other tumor types identified; across all cohorts, the most common type of other tumor was connective tissue/breast cancer (13.6%). By design, there were significant differences between the cohorts for index date. The post-tier group had signifi cantly fewer claims-based deaths compared with pre-tier cohort; however, this may be due to the longer follow-up period in the pre-tier cohort. During the baseline period, patients in the post-tier cohort had signifi cantly fewer other chemotherapies and radiation and more surgeries compared with the pre-tier cohort patients. Although post-tier patients had a significantly lower mean Charlson Comorbidity Index score (7.29 vs 7.58; P = .0008), the difference of 0.29 was considered clinically insignificant.

Copayments

Erlotinib mean (

Figure 2A

) and median (

Figure 2B

) copayment levels were compared across years of erlotinib initiation. The overall difference between the mean copayment of post-tier patients and the mean copayment of pre-tier patients was not statistically signifi cant, although the overall median copayment was signifi cantly lower for post-tier patients than for pre-tier patients. For post-tier patients who initiated erlotinib in 2007, both mean and median copayments were signifi cantly lower compared with those for pre-tier patients. Median, but not mean, copayment differed for post- and pre-tier patients initiating erlotinib in 2004, 2005, or 2006. For patients in the crossover cohort, mean (P <.05) and median (P <.001) copayments were significantly lower after the tier change compared with before the tier change.

Prescription Counts

Erlotinib mean (

Figure 3A

) and median (

Figure 3B

) prescription counts were compared across years and by year of erlotinib initiation. The average number of erlotinib prescriptions appeared to vary across the pretier and post-tier cohorts when stratified by the year of erlotinib initiation, although differences were not statistically significant when all years were combined. The mean number of prescriptions was 3.37 for the pre-tier cohort and 3.41 for the post-tier cohort (median of 2 prescriptions for both the pre-tier and post-tier cohorts) when all years were combined. Among post-tier cohort patients who initiated erlotinib therapy in 2007, significantly more prescriptions were fi lled (3.41 prescriptions) compared with those for pre-tier cohort patients who initiated erlotinib in 2007 (2.39 prescriptions; P = .0416).

Primary Adherence and Persistence Analysis

Pre-tier and post-tier patients had similar rates of adherence to erlotinib (84.45% and 83.75%, respectively; P = .7917;

Table 2

). After multivariate adjustments (

Table 3

), compared with patients with copayments of <$30, patients with copayments of $45 to <$65 had significantly decreased odds of adherence (odds ratio [OR] 0.54, 95% confidence interval [CI] 0.30-0.96). Differences between the lowest-copayment group and patients with copayments of $30 to $45 (OR 0.67, 95% CI 0.0.40-1.12) or >$65 (OR 0.52, 95% CI 0.26-1.06) were not statistically significant.

For erlotinib persistence (representing drug discontinuation), pre-tier patients were more likely than post-tier patients to discontinue erlotinib (44.78% vs 37.07%; P = .008) (Table 3). In multivariate analysis, post-tier patients were 21% less likely to discontinue treatment than pretier patients (hazard radio 0.79, 95% CI 0.65-0.95).

Sensitivity Analysis

Sensitivity analysis of patients’ prescription use found that 41.94% of the pre-tier cohort and 47.56% of the post-tier cohort were adherent using this alternative definition (

Table 4

), with no statistically significant difference in adherence between pre-tier and post-tier cohorts (P = . 0551). After multivariate adjustment, the logistic regression model showed no statistically significant difference in adherence between pre-tier and post-tier cohorts (P = .0815;

Table 5

). Erlotinib copayments and patient total out-of- pocket expenses had a significant effect on prescription use and persistence (both P <.001; Table 5).

In the sensitivity analysis of erlotinib use, 44.76% of the pre-tier cohort and 37.07% of the post-tier cohort discontinued treatment (Table 4) when the sample was restricted to patients with similar follow-up times and 1 or more prescriptions. After multivariate analysis, the effect of treatment cohort was no longer statistically significant (P = .059; Table 5).

DISCUSSION

Our results show that for 2007, when erlotinib therapy was initiated both mean and median copayments were significantly lower for patients who started erlotinib after the change from tier 3 to tier 2; this fi nding is further supported in the crossover cohort. Although median copayments had increased from 2005 to 2007, these rates were lower compared with erlotinib copayments in 2004. Following the insurance formulary change from tier 3 to tier 2, significantly more prescriptions were filled for erlotinib on average. Although the formulary change was not associated with signifi cant differences in adherence, patients with copayments of <$30 were more likely to be adherent to erlotinib compared with patients with copayments of $45 to $60. Furthermore, post-tier patients were 21% less likely to discontinue treatment than pre-tier patients. Of note, we used the same large US insurance claims database used by Ramsey and colleagues.²⁰

This was the first study of erlotinib to evaluate the effect of copayments on treatment behaviors. Previous studies of erlotinib did not separate copayments from either other pharmacy claim expenses²⁰ or other oral anticancer treatments.²³ Our data show that copayments were significantly lower in 2007 (mean $48 vs $121; median $25 vs $43) following the formulary change.

Notably, our adherence rates were higher (83%-85%) when patients with only 1 prescription fill were removed and adherence was defined by the proportion of days covered between the first and last erlotinib fill. Comparisons with other erlotinib studies are limited as these are either currently ongoing²⁴ or did not separate TKI treatments and only looked at infusion chemotherapy dosing adherence.¹⁹ Our adherence rates were higher than those reported for real-world adherence to oral breast cancer drugs (50%-80% of patients when the definition included patients with at least 1 prescription or microelectronic monitoring system was used).²⁵ However, the definition of adherence varies across studies. Other researchers have used alternative adherence definitions,²⁶ including number of patient prescriptions filled divided by number of physician prescriptions written¹⁰ and days of supply divided by the length of the study period.^11,12

In our sensitivity analysis, erlotinib prescription use rates varied from 41% to 48% when defi ned by >80% adherence and including patients with 1 or more prescriptions. Our lower prescription use rates may be due to prescribing over multiple years, since adherence rates tend to drop over a longer duration of prescribing.^27,28 Furthermore, the inclusion of patients who had only 1 prescription filled may have resulted in lower adherence rates.

Our study found that 37% to 45% of patients discontinued erlotinib therapy and the difference between the pre-tier and post-tier cohorts was significant. Those rates are higher than the 12.8% discontinuation rate reported by Streeter and colleagues.²³ In addition, our discontinuation rates are higher than the 15% to 32% rate reported across 3 studies evaluating tamoxifen discontinuation in breast cancer patients.^28-30 In general, Streeter and colleagues²³ reported increased abandonment of oral anticancer drugs with increasing cost sharing, increasing multiple prescription needs, and decreasing income. With tamoxifen therapy, women with a nonpositive belief in treatment, in the youngest or oldest age groups, previous use of antidepressant drugs, and side effects (depression, nausea, visual problems, vaginal bleeding) were more likely to discontinue therapy.^29-31

In our study, formulary change was not associated with significant differences in adherence, whereas Madden and colleagues³² identified cost-related nonadherence when evaluating the impact of implementation of Medicare Part D. In a study by St. Charles and associates,¹⁷ higher copayments were associated with lower adherence, similar to our findings.

Nonadherence and nonpersistence with oral chemotherapy medication may result in safety and efficacy concerns.³³ Furthermore, use of the agent as second- or third-line chemotherapy may have infl uenced the numbers. With approval of erlotinib as first-line maintenance treatment,³⁴ additional changes in use and associated costs may occur in the future.

Limitations

Claims data are collected for reimbursement purposes rather than for research; therefore, claims may not include all relevant clinical information needed. For example, differences in the underlying severity of illness of patients using erlotinib in the pre-tier and post-tier cohorts may have changed over time and may not have been adequately captured in claims data. We were limited in how long we could follow the post-tier cohort; thus, potential length of follow-up varied significantly between pre-tier and post-tier cohorts. We did not investigate the relative burden of patients’ out-of-pocket expenditures in relation to overall income or wealth, which may be a more significant driver of patients’ adherence than the actual amount of the expenditures. The proportion of patients in this study receiving assistance with copayments from patient assistance programs was unknown. This is an important limitation because it is unknown whether such assistance resulted in unobserved use of erlotinib (ie, patients who appeared to have discontinued erlotinib were actually receiving erlotinib through the patient assistance program). The average change in copayment was not very large; it is possible that for a condition with a high risk of fatality, a larger copayment/patient out-of-pocket burden would be required to affect patient adherence. Also, the use of erlotinib as second-line versus third-line therapy may influence pharmacy-related costs.²⁰ The data used for this study come from a commercial managed care population; therefore, results of this study may not be applicable to patients who were uninsured or receiving Medicare or Medicaid.

CONCLUSIONS

Although moving erlotinib from tier 3 to tier 2 resulted in a decrease in mean copayments, median copayments had also risen signifi cantly in the 4 years preceding the tier change; thus, the median copayment following the tier change dropped only to the level that it had been 3 to 4 years before the tier change. The relative position of erlotinib on the formulary (tier 3 vs tier 2) was not associated with statistically signifcant differences in adherence; however, absolute copayment level was associated with adherence. Following the formulary change from tier 3 to tier 2 for erlotinib, patients with lung cancer had lower copayments with evidence of higher treatment persistence.