GSK Medicine: Study Number: Title: Rationale: Study Period: Objectives: Indication: Study Investigators/Centers: Research Methods: Data Source

Transcription

1 The study listed may include approved and non-approved uses, formulations or treatment regimens. The results reported in any single study may not reflect the overall results obtained on studies of a product. Before prescribing any product mentioned in this Register, healthcare professionals should consult prescribing information for the product approved in their country. GSK Medicine: Lapatinib Study Number: Title: Retrospective Study of the Prevalence, Predictors, and Consequences of Non-Adherence with Lapatinib in Women with Metastatic Breast Cancer Who Were Previously Treated with Trastuzumab Rationale: Adherence with prescribed treatment in typical clinical practice is often poorer than adherence in controlled trials, as subjects are typically monitored less frequently. And, although oral anti-cancer therapies, such as lapatinib, have advantages in terms of convenience and savings in administration costs compared with medications that must be infused, little is know about adherence with lapatinib treatment in clinical practice. Study Period: 14 Jun 2011 to 30 Dec 2011 Objectives: The objective of this study is to examine the prevalence, predictors, and consequences of non-adherence with lapatinib in women with metastatic breast cancer who were previously treated with trastuzumab. All analyses will be considered hypothesis generating. Indication: metastatic breast cancer Study Investigators/Centers: Policy Analysis, Inc Research Methods: A retrospective observational cohort study of women with metastatic breast cancer wtih data from January 2000 to March 2010 Data Source: Data for this project were obtained from the Thomson MedStat MarketScan Commercial Claims and Encounters (CCAE) Database and the Medicare Supplemental and Coordination of Benefits (MDCR) Database. These databases contain information on the health insurance claims of employees of large, self-insured corporations and their dependents, along with a few commercial health plans, and for Medicare-eligible persons (mainly retirees) who are also covered by self-insured employers. In total, over the period of study, the two databases contain healthcare claims data for over 50 million persons, with an average of three to five years of follow-up. Study Design: Retrospective, observational cohort study Study Population: Study subjects consist of adult women with metastatic breast cancer who received lapatinib and who have a history of prior treatment with trastuzumab and who meet pre-specified enrollment criteria in one of the health plans included in the Thomson MedStat MarketScan Commercial Claims and Encounters (CCAE) Database and the Medicare Supplemental and Coordination of Benefits (MDCR) Database. Inclusion Criteria included female gender, age 18 years or older, at least twelve months of continuous healthplan eligibility prior to index date (date of first lapatinib claim = index date ), at least three months of continuous healthplan eligibility from index date to end of study period, one or more claims for lapatinib during the study period, one or more claims for trastuzumab during the study period, one or more claims with a diagnosis of breast cancer during the study period (International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code 174.xx: Malignant neoplasms 238.3x, 239.3x, V10), one or more claims with a diagnosis of distant metastases during the study period; distant metastases will be identified based on the following ICD-9-CM code: 155.xx, 162.xx, 170.xx, 191.xx, 192.xx, 196.xx, and secondary and unspecified malignant neoplasm of lymph nodes; ICD-9-CM (197.xx, 198.xx, 199.xx) Exclusion Criteria included male sex, age less than 18 years as of index date, less than twelve months of continuous eligibility prior to index date, less than three months from index date to end of study period, no claims for trastuzumab prior to first claim for lapatinib, no claims for trastuzumab after first claim for distant metastases, less than two claims occurring on different days ( claim days ) for lapatinib, and missing or ambiguous data on claims or enrollment information required to calculate subject. Study Exposures, Outcomes: Adherence to lapatinib was assessed using information on therapy-days supplied from outpatient pharmacy claims. Measures of adherence to lapatinib included the medication possession ratio (MPR), time to treatment interruption, time to treatment discontinuation, and time to end of continuous treatment. MPR for lapatinib was defined as the maximum of 1.0 (100%) and the ratio of total days of lapatinib supplied divided by the lapatinib treatment period. The lapatinib treatment period was defined as the number of days from the index date to the last day during the follow-up period with supply on hand. The last day with supply on hand was calculated as the date of the last claim for lapatinib plus the number of 1

2 days supplied on that claim. Nonadherence was defined as 100 x (1 MPR). Subjects were alternatively classified as nonadherent based on nonadherence of 5% or more (MPR<95%) and nonadherence of 20% or more (MPR<80%). A lapatinib treatment interruption was defined as a failure, during the lapatinib treatment period, to refill a prescription for lapatinib within 30 days of the end of supply from prior prescriptions. Supply of lapatinib on hand was calculated as the sum of the number of days supplied from prior claims minus the number of days elapsed since the index date. Time to lapatinib discontinuation was calculated as the time between the index date and the last day with supply of lapatinib on hand. Time to end of continuous lapatinib therapy was calculated as the time between the index date and the minimum of the time to treatment interruption or the time to discontinuation. Time to treatment interruption was considered censored at the end of the treatment period for subjects with no gap of 30 days during the treatment period. Time to treatment discontinuation and time to end of continuous lapatinib therapy were considered censored for subjects with supply of medication on hand less than 30 days from the end of the study period. Because discontinuation may be due to disease progression, whereas treatment interruption is not likely due to disease progression, we report these separately. We report time to end of continuous therapy (the combination of treatment interruption and discontinuation) as this measure is comparable to time to discontinuation as reported in prior studies of adherence. Measures of utilization and costs were calculated separately for the pre-index period, the lapatinib treatment period, and the post-lapatinib treatment period. Measures of health-care utilization included the numbers of physicians office or outpatient hospital visits, numbers of hospitalizations, and numbers of inpatient days. Measures of health-care costs included the cost of lapatinib, trastuzumab, chemotherapy (CT), CT administration, all other costs, and total costs. To control for duration of each period (the lapatinib treatment period and post-lapatinib treatment period could differ across subjects), measure of health care utilization and cost were calculated on monthly basis by dividing total utilization and costs during the period by the number of months of observation during the period. Time to change in treatment was defined as discontinuation of lapatinib or receipt of new hormonal, chemotherapy or anti-her2 therapy. New hormonal therapy, chemotherapy and anti-her2 therapies was defined as any hormonal therapy, chemotherapy or anti-her2 therapy received during the follow-up period that was not received during the peri-index period (defined as the period beginning 7 days prior to the index prescription date for lapatinib and ending 30 days after the index prescription date). For subjects with no change in treatment, censoring time was the time between index day and end of study. Data Analysis Methods: Time to dose-reduction, treatment interruption, treatment discontinuation, end of continuous treatment, change in treatment, and end of continuous enrollment were analyzed using Kaplan Meier methods. A variety of analyses were undertaken to identify predictors of nonadherence with lapatinib. In these analyses, nonadherence was analyzed as a continuous variable using ordinary least squares regression with log(100 x (1-MPR)+1) as the dependent variable and as a binary variable using logistic regression with subjects classified as non-adherent based on nonadherence of 5% or more (MPR<95%) and nonadherence of 20% or more (MPR<80%). Covariates included in these models were population (commercial or Medicare), age (nested within population), region, plan type, comorbidity index, taxane and anthracycline use pre-index, number of physician's office or hospital outpatient visits pre-index, total health-care costs pre-index, anti-cancer therapies received during the peri-index period (none, capecitabine only, trastuzumab only, hormonal therapy only, taxanes only, capecitabine and hormonal therapy, capecitabine and trastuzumab, trastuzumab and hormonal therapy, and other), subject contribution on index prescription (0, >$0 to <$50, $50 to <$100, and $100), and daily dosage of lapatinib on index prescription (<1250 mg vs mg). We also conducted similar analyses of predictors of treatment interruption using Cox proportion hazard regression. Models were generated alternatively using all covariates and stepwise selection. Analyses of the association between nonadherence and health-care utilization and costs were conducted on the numbers of physicians office and outpatient hospital visits and total health-care costs excluding the costs of lapatinib. These variables were selected based on a power analysis suggesting that power to detect an association would be greatest for these measures. The analysis of costs excluding lapatinib costs is reasonable because it is elementary that lapatinib costs will increase with better adherence. The question of interest was whether there might be some savings in other areas that might offset the increased costs of medication. An analysis of the association between nonadherence and total costs including the costs of lapatinib was included for completeness. Analyses of the association between nonadherence and these measures were conducted using generalized linear regression (GLM) models. For visits, log link and negative binomial error terms were employed. For costs, log link and gamma error terms were used. These analyses were conducted with nonadherence (100 x (1-MPR)) first included as continuous variable, then as a categorical variable (MPR<95% vs. 95%; MPR<80% vs. 80%; MPR<80%, 80% MPR<80%, vs. MPR 95%). 2

3 Limitations: First, this analysis used health insurance claims data which are subject to coding errors and incomplete claims history. Claims data lack information on laboratory data, clinical and radiological examinations, and cancer disease staging. Mortality information is unavailable from the database. Although the database contains fields for subject contribution, the completeness of this data for claim as oral oncology drugs such as lapatinib that may be received through specialty pharmacies is uncertain. In examining predictors of adherence, we only examined baseline covariates and did not examine post-index factors that might be affected by treatment such as response to treatment or adverse events. While it likely is not feasible to identify response in a claims database, it might be feasible to identify adverse events. This analysis of the association between adverse events and adherence is a potential area for future research. Also, this analysis focused only on subjects receiving lapatinib after trastuzumabthe MarketScan database is based on a large convenience sample. Because the sample is not random, it may contain biases or fail to generalize well to other populations. Although the database is geographically diverse it is not geographically representative of the United States population. Further, the data come mostly from large employers; subjects employed at medium and small firms or covered by Medicaid are not represented. Study Results: A total of 666 met all inclusion criteria (572 Commercial, 94 Medicare). Fourteen percent of subjects had a lapatinib dose reduction. Median time to first dose reduction was not reached. Concomitant therapy with a taxane was a predictor of nonadherence (odds ratio for MPR<80%=10.30, 95%CI 3.32 to 31.92, p<0.001). In generalized linear regression models, there was a statistically significant association between nonadherence with lapatinib and greater physicians' office / hospital outpatient visits (exponent of coefficient on 100-MPR [range: 0 to 100] = 1.256, 95%CI: to 1.540, p=0.028). In the multivariate GLM regression model predicting the log of 100 x (1-MPR) and logistic regression models of MPR<95% and MPR<80%, there were statistically significant associations between receipt of concomitant taxanes and receipt of concomitant trastuzumab and hormonal therapy and worse adherence. In Cox regression analysis, there was a statistically significant association between receipt of concomitant taxanes and reduced time to treatment interruption (Hazard Ratio (HR)=3.98, 95%C: 1.93, 8.22, p<.001). There was a statistically significant association between subject contribution on lapatinib index prescription >$0 to <$50 (vs. $0) and reduced adherence in all models. The multiple linear regression on the log of 100 x (1-MPR) had a very low R-squared (0.05), suggesting that the models explained relatively little of the variability in nonadherence. In multivariate GLM regression models, there was a statistically significant association between nonadherence to lapatinib (measured as 1-MPR) and increased physician s office / outpatient visits after controlling for baseline characteristics. Each 1% absolute decline in MPR (expressed as a percentage) was associated with a 0.2% increase in the number of visits. There were no statistically significant associations between nonadherence and total health care costs (p=0.870) or total health care costs excluding the costs of lapatinib (p=0.146). When binary and categorical measures of adherence were entered as predictors of visits and costs, there no statistically significant associations between measures of adherence and number of visits or costs. Demographics/Baseline Characteristics Commercial (N=572) Medicare (n=94) Combined (N=666) Age in years, mean(sd) 52 (8) 72 (6) 54 (10) Geographic region, N % Northeast 56 (9.8) 13 (13.8) 69 (10.4) North central 169 (29.5) 38 (40.4) 207 (31.1) South 262 (45.8) 34 (36.2) 296 (44.4) West 81 (14.2) 9 (9.6) 90 (13.5) Unknown 4 (0.7) 0 (0) 4 (0.6) Insurance type, n(%) HMO 92 (16.3) 3 (3.2) 92 (16.3) PPO 354 (62.9) 43 (46.2) 397 (60.5) 3

4 POS 72 (12.8) 6 (6.5) 78 (11.9) Indemnity 32 (5.7) 40 (43.0) 72 (11.0) Other 13 (2.3) 1 (1.1) 14 (2.1) Comorbidity Index, mean (SD) 6.53 (0.85) 7.07 (1.26) 6.61 (0.94) Time since last trastuzumab claim in days, mean (SD) 108 (176) 112 (185) 109 (177) Anti-cancer therapy during pre-index period, N (%)* Anastrozole 54 (9.4) 11 (11.7) 65 (9.8) Exemestane 39 (6.8) 11 (11.7) 50 (7.5) Fulvestrant 52 (9.1) 14 (14.9) 66 (9.9) Letrozole 56 (9.8) 15 (16.0) 71 (10.7) Tamoxifen 34 (5.9) 3 (3.2) 37 (5.6) Chemotherapy Taxanes 269 (47.0) 44 (46.8) 313 (47.0) Docetaxel 109 (19.1) 14 (14.9) 123 (18.5) Paclitaxel 186 (32.5) 33 (35.1) 219 (32.9) Anthracyclines (AC) 51 (8.9) 7 (7.4) 58 (8.7) Capecitabine 144 (25.2) 24 (25.5) 168 (25.2) Carboplatin 93 (16.3) 13 (13.8) 106 (15.9) Gemcitabine 79 (13.8) 15 (16.0) 94 (14.1) Vinorelbine 124 (21.7) 28 (29.8) 152 (22.8) Supportive medications during pre-index period, N % Intravenous bisphosphonates 256 (44.8) 41 (43.6) 297 (44.6) Oral bisphosphonates 8 (1.4) 5 (5.3) 13 (2) Colony stimulating factors 164 (28.7) 25 (26.6) 189 (28.4) Erythropoietic stimulating factors 206 (36) 43 (45.7) 249 (37.4) Anti-cancer drugs received during the peri-index period Combined (N=666) Number of anti-cancer drugs received, n(%) 0 94 (14.1) 4

5 1 432 (64.9) (17.9) 3 19 (2.9) 4+ 2 (0.3) Anti-cancer drugs received (any combination), n(%) Capecitabine 417 (62.6) Paclitaxel 27 (4.1) Docetaxel 8 (1.2) Anthracyclines 4 (0.6) Vinorelbine 24 (3.6) Gemcitabine 18 (2.7) Any hormonal 68 (10.2) Trastuzumab 142 (21.3) Other chemotherapy 21 (3.2) Specific Anti-cancer drug combinations received, n(%) None 94 (14.1) Capecitabine 328 (49.2) Trastuzumab 49 (7.4) Capecitabine+trastuzumab 45 (6.8) Capecitabine+hormone therapy 21 (3.2) Hormone therapy 20 (3.0) Taxane 13 (2.0) Trastuzumab + hormone 12 (1.8) Capecitabine+trastuzumab+vinorelbine 6 (0.9) Vinorelbine 8 (1.2) Gemcitabine 8 (1.2) Trastuzumab+taxane 9 (1.4) Other 53 (8.0) 5

6 Combined (N=666) 12 Months Pre- Index Period, mean (SD) Lapatinib Treatment Period, mean (SD) Post Lapatinib Treatment Period, mean (SD) Follow up Period, mean (SD) Number of physicians' office or outpatient hospital visits 6.09 (3.04) 5.45 (3.49) 4.55 (3.90) 5.22 (3.12) Number of hospitalizations 0.05 (0.08) 0.09 (0.19) 0.15 (0.32) 0.10 (0.17) Number of inpatient days 0.24 (0.52) 0.53 (1.39) 0.96 (2.52) 0.64 (1.30) Costs, US$ Lapatinib 0 (0) 2,097 (877) 0 (0) 1,469 (1,031) Trastuzumab 2,762 (2,020) 722 (1,731) 1,166 (2,134) 924 (1,556) Chemotherapy 1,355 (1,898) 1,496 (1,698) 1,437 (2,729) 1,543 (1,805) CT Administration 508 (447) 189 (298) 326 (516) 249 (343) Other 5,446 (4,680) 5,564 (7,198) 6,880 (9,726) 5,973 (5,765) Total 10,071 (6,056) 10,067(7,695) 9,809 (10,853) 10,158 (6,468) Total excluding lapatinib 10,071 (6,056) 7,970 (7,734) 9,809 (10,853) 8,689 (6,526) Lapatinib Treatment and Post-lapatinib Treatment Periods Combined Medication Possession Ratio Results, Combined (N=666) Number of lapatinib claims, mean (SD) 7.1 (5.4) MPR, mean percentage (SD) 87% (19%) MPR, median % 96% 6

7 Median time to discontinuation of lapatinib, in months (95% CI) 9.1 ( ) Median time to end of continuous lapatinib treatment, in months (95% CI) 5.9 ( ) Median time to change in ACT (discontinuation or receipt of new anti-cancer agent) 8.0 ( ) Conclusion: In women with MBC previously-treated with trastuzumab, more than half were highly adherent to lapatinib treatment in typical clinical practice. Further research is needed to confirm these findings and identify additional predictors of nonadherence so that clinicians may develop strategies to enhance adherence to lapatinib. 7