Study design
This retrospective cohort study utilized a commercially available Japanese administrative healthcare database (MinaCare Co., Ltd., Tokyo, Japan) to investigate the adherence of patients to fixed-combination and unfixed-combination therapies for glaucoma treatment after switching from a monotherapy. The study consisted of a 12-month pre-index period, an index date (defined as a first prescription day of either when a single eye drop was first switched to a fixed-combination therapy [fixed-combination cohort] or when the second eye drop was added to a single eye drop initially [unfixed-combination cohort]), and a 12-month post-index period. Patients without claims for 12 months after the index date and those who claimed the third drug in addition to a fixed-combination therapy or in addition to an unfixed-combination therapy indicated withdrawal. Approval for this research by an ethical review committee and informed consent of each subject were not required because studies using only unlinkably anonymized data are outside the scope of “Ethical Guidelines for Medical and Health Research Involving Human Subjects”23 set by the Japanese government.
Data source and study cohort
The MinaCare database includes anonymized data on both health checkup and medical/pharmacy claims of workers and their family members in a wide range of age groups below the age of 75 years.24, 25 Maintained by MinaCare Co., Ltd., this database is updated periodically with newly available data obtained from the employment-based health insurance groups. It includes the data of 6.3 million uniquely identified individuals as of April 2017 and has recorded the information of 1.8 million patients from April 2016 to March 2017, covering 1.7% of the Japanese population under the age of 75 years.26 Furthermore, it is generally consistent with two national databases and is useful as it has low selection bias and large sample size with wide age distribution; however, it only targets national-wide big corporations, such as manufacturing, food, information transportation, and energy industries, and does not include individuals in the primary industry (agriculture, fisheries, forestry, etc.) or those who are self-employed.24
Patients were extracted from the database with the ICD10 diagnosis codes H401 (normal tension glaucoma, primary open-angle glaucoma, and open-angle glaucoma) and H409 (unspecified glaucoma) and with prescription claims for the eye drops to treat glaucoma between April 1, 2011 and March 31, 2016 We included those who had received monotherapy for 1 year or longer without surgery including laser treatment and then switched to a fixed-combination therapy as second-line treatment (the fixed-combination cohort) or those who received an additional eye drop, such as timolol as an addition to latanoprost or vice versa, as second-line treatment (the unfixed-combination cohort). The first prescription date of the second-line therapy was defined as the index date. Patients without prescription of a fixed or unfixed-combination therapy, with history of surgery or laser surgery for glaucoma, or with pre-index monotherapy for <12 months were excluded.
Study endpoints
The primary endpoint was medication adherence, which was assessed by the proportion of days covered (PDC) with medication over the 12-month post-index period. The a priori definition of adherence cutpoint was 80%. Meanwhile, the secondary endpoints were related to medication persistence, defined as the act of continuing the use of claimed eye drops, and included persistence rate and its duration, a distribution of patients according to the type of claimed eye drops at the index date and switching patterns of eye drops during the 12-month post-index period. In addition, the risk factors for PDC < 80% were analyzed.
Assessments
Subject background characteristics included sex, age, body weight, height, body mass index, smoking status/history, health examination results, comorbidity, residence region, the number of eye drop bottles consumed, and eye drop types. These characteristics were summarized in all patients and in each cohort. Moreover, residence regions were classified into eight regions from Hokkaido to Kyushu including Okinawa,27 two areas (East Japan vs. West Japan),28 and city sizes (big cities including Tokyo and 20 ordinance-designated cities vs. the others).29 We also identified the PDC over 12 months after the index date in all patients and each cohort, the risk factors for PDC < 80%, and the persistence rate.
To address the possible variations in the dosage and prescribed period reported on a claim for glaucoma medication, we made adjustments based on the unit volume of the formulation. For example, if the dosage was “7.5” and the period was “1” in a claim, we converted them to “2.5 mL (approved bottle) × 3 bottles” prescribed for 3 months because one bottle (2.5 mL for once-daily formulations or 5 mL twice-or-more-daily formulations) is enough to apply one eye for more than 30 days but is correspondent to 30 days dispensed. The general recommendation of the use-by period of opened eye drop bottles is generally 4 weeks in Japan. Accordingly, the patients were permitted a 30-day grace period to obtain the next prescription. In addition, if one prescription consists of ≥7 bottles, the patient was considered receiving treatment of both eyes. Therefore, a PDC value was calculated using the following formula:
PDC (%) = (total prescription days during the 365-day assessment/365 days) × 100
In this study, the PDC is calculated according to the period only prescribed with two-drug fixed-combination drops or two-drug unfixed-combination therapies, regardless of switching drug class [e.g., PG+BB to PG+ carbonic anhydrase inhibitor (CAI)].
Persistence refers to the act of continuing index therapy (i.e., a fixed-combination therapy or an unfixed-combination therapy). The definition of “persistence” is the duration in days from the index date (the first prescription date of the index therapy) to the last prescription date + prescription days or to the discontinuation date allowing a 30-day grace period. In the case of one drug addition to the combination of two drugs (e.g., PG+BB to PG+BB+AA), it was considered as “discontinued/withdrawn.” Likewise, in a subset analysis using specific drug classes, no prescription record for more than 30 days or switched or changed prescriptions were treated as “discontinued.” The cumulative discontinuation rate was analyzed based on the time to discontinuation and was estimated using the Kaplan–Meier method. Patients who continued the treatment for 12 months were treated as censored at 12 months. The persistence rate at the evaluated time point was calculated as follows:
Persistence rate (%) = 100% − cumulative discontinuation rate
Statistical analysis
We presented summary statistics (mean and standard deviation [SD] or a number and a proportion of patients) for the patients’ background characteristics and PDC. Using t-test, we compared the mean PDC values between two cohorts. Furthermore, proportions of patients with or without ≥80% PDC between two cohorts were compared using Pearson's χ2 tests. Bonferroni correction was used to adjust for multiplicity (m = 9, where m is the total number of statistical tests in this study). A p-value =< 0.00556 (0.05 / 9) was considered statistically significant for comparisons between the groups. We also explored the risk factors for nonadherence to the medications (PDC < 80%) by using a univariate logistic regression model. Subsequently, multivariate analyses were performed for those with P < 0.1, followed by a stepwise (forward and backward) logistic regression model. The odds ratios (OR) and Wald 95% confidence intervals (CIs) were presented. The subject persistence rates at 6 and 12 months with a pointwise 95% CI were estimated using the Kaplan–Meier method, and the differences between the two treatment cohorts were tested by log-rank test. All the statistical analyses were performed using the SAS version 9.4 (SAS Institute Inc., Cary, NC, USA).