Study design and data source
This observational, retrospective cohort study extracted patient data from the Japanese NDB for HIV-positive patients who received treatment between April 2009 and March 201914. The NDB is the largest nationwide cross-sectional database in Japan, and it contains comprehensive health insurance claims records from the National Health Insurance system of Japan for direct primary care delivered as inpatient care. The NDB has been used to supported clinical studies15,16. The NDB includes data for diagnoses, age, sex, dates of outpatient services, dates of admission and discharge, procedures undertaken, diagnoses, drug prescriptions, and health checkup data. Patients treated from April 2009 to March 2019 with at least one diagnosis and any treatment processed in the claims data were enrolled. All included diagnoses were categorized according to the “The International Classification of Disease, 10th Revision, Clinical Modification” (ICD-10) diagnostic system. Patients with a diagnostic code of HIV-2 infection were excluded.
All insurance claims data are deidentified by the Ministry of Health, Labour and Welfare, and the ministry’s guidelines on information security were followed in the study. To ensure patient privacy, inspection by and permission from the Ministry for publication is required before submission of the draft manuscript.
Study population
HIV-positive patients in the database were identified by the presence of at least one record of the International Classification of Diseases 10th Revision (ICD-10) codes B20‒24, including: HIV disease resulting in infectious and parasitic diseases (B20), malignant neoplasms (B21), other specified diseases (B22), other conditions (B23), and unspecified HIV disease (B24). To avoid including doubtful HIV-positive patients (i.e., poorly recorded or intentionally recorded for a claim), patients were required to have at least one prescription record of ART. The corresponding ICD-10 codes are listed in Supplemental Table 1. ART was defined as a prescription for any of the following antiretroviral drugs: NRTIs, NNRTIs, PIs, INSTIs, or EIs.
Data of patients meeting these criteria (n = 28,089) during the study period were extracted from the database. Patients who had received first prescriptions for an ART regimen between 2011 and 2019 (n = 16,069) were included in the present analysis.
Outcomes
Primary outcomes were switch rates and time-to-switch associated with individual ART regimens by anchor drug classes. Secondary outcomes were switch rates and time-to-switch associated with anchor drug class-based ART regimens by type of backbone drug, characteristics of patients who experienced one anchor drug switch in each anchor drug class-based ART regimen, and common switching patterns of anchor drug classes.
Definitions
Data extracted from the NDB included patients’ demographic characteristics (age, sex), and clinical characteristics, including year of first ART record in the database, prescription records of anchor drugs and backbone drugs, comorbidities, hospitalization history, and AIDS-defining illnesses. In this study, age was defined as the age of the patient at time of the first ART regimen.
Anchor drugs and backbone drugs
According to treatment guideline an ART regimen generally consists of two nucleoside reverse transcriptase inhibitors (NRTIs) administered in combination with a third active ART drug from one of three drug classes: INSTI, NNRTI, or PI with a pharmacokinetic enhancer (also known as a booster)1,2. In our study we referred to nucleoside reverse transcriptase inhibitors (NRTIs) as backbone drugs; and anchor drug is the third active ART drug.
The anchor drugs of the ART regimens were identified using receipt codes and were classified into three anchor drug classes according to the anti-HIV drug classification available in Japan3: (1) NNRTIs, (2) PIs, or (3) INSTIs.
The backbone drugs of the ART regimens were identified using receipt codes and classified into four categories: (1) tenofovir disoproxil fumarate (TDF); (2) abacavir (ABC); (3) tenofovir alafenamide fumarate (TAF); and (4) others.
ART regimen switch and time-to-switch
An ART regimen switch involved only anchor drug classes and was defined on the basis of a switch in the specific anchor drug class used in the ART regimen. The time-to-switch of an ART regimen was defined as the period from the date of the first record of anchor drug class in the ART regimen (defined as the first regimen) recorded within patient data (index date) to the date of switching to another anchor drug class in the subsequent ART regimen (defined as the second regimen) during the study period. The date of an anchor drug class switch was defined as the date of prescription of the new anchor drug class after the termination of the preceding (first) ART regimen. A regimen was considered discontinued when no initiation of any new anchor drug class was identified after termination of the preceding (first) ART regimen. A change of anchor drug within the same anchor drug class was not considered a switch. Patients with multiple prescription records of anchor drug class on the index date, patient who switched ART in the month following initial ART prescription and had more than one ART within the same month of the switch were also excluded.
AIDS-defining illnesses
AIDS-defining illnesses were identified by the presence of any of the following records prior to index date: HIV non-tuberculous mycobacteria, HIV cytomegalovirus infection, HIV candidiasis, HIV Pneumocystis carinii pneumonia, HIV Kaposi’s sarcoma, HIV Burkitt’s lymphoma, HIV non-Hodgkin’s lymphoma, HIV encephalopathy, HIV-associated dementia, slim disease, acquired immune deficiency syndrome, AIDS, neonatal HIV infection, and AIDS-related complex. The corresponding ICD-10 codes are listed in Supplemental Table 1.
Since there was no disease severity information in the NDB database, patients were sub-grouped according to status of AIDS diseases at the time of the first ART regimen. Patients with diagnosis of AIDS disease after initiation of 1st ART regimen were defined as no AIDS.
Comorbidities
Comorbidities were identified if any ICD-10-coded chronic illnesses were present prior to the index date, including: HIV-related diseases, hypertension, dyslipidemia, hepatitis B/C coinfection, diabetes mellitus, bone disorder, vascular disease, psychiatric disorders, kidney disease, and malignancy. Corresponding ICD-10 codes are listed in Supplemental Table 1.
History of hospitalization
A history of hospitalization was identified if a record of hospitalization was present before the ART regimen was prescribed.
Statistical analysis
The proportions of anchor drug class-based and backbone-based drugs prescribed on the index date were obtained by year. Demographic and clinical characteristics of all patients on ART regimens were analyzed descriptively according to the anchor drug class prescribed on the index date. The median time-to-switch and switch rates according to anchor drug class prescribed on the index date and those stratified by the backbone drugs were estimated using Kaplan–Meier analysis. To estimate 95% confidence interval (CIs), the Brookmeyer and Crowley method was used for the median number of days, and the Greenwood method was used for switch rates. Log-rank tests were used to compare switch rates between the respective drug classes in each year. The Bonferroni method was performed to adjust p-values on multiple comparison. Discontinuation or continuation of the regimen to the end of the study period was censored.
The demographic and clinical characteristics of patients who switched anchor drug classes in their ART regimens were analyzed descriptively according to the anchor drug class prescribed on the index date. Timings of < 1 and ≥ 1 year were analyzed descriptively for patients who switched anchor drug classes in their ART regimens according to the anchor drug class prescribed on the index date and the corresponding 95% CI using Wilson scores.
Confounding factors for switching ART regimens and factors interacting with the anchor drug class were selected a priori based on previous studies. Time-to-switch was an objective variable and the anchor drug class, risk factors, and interaction term between anchor drug classes and each risk factor were included as explanatory variables in this model. The hazard ratio (HR) of each anchor drug class was calculated after adjusting for remaining variables and stratified by interaction factors to estimate the risk of switching anchor drug classes from the ART regimen prescribed on the index date. All statistical analyses were performed in the R 4.0.3 environment (R Core Team, 2020). All remaining statistical tests were two-sided, and p < 0.05 indicated statistical significance.
Ethics statement
This study was performed in compliance with national regulations and institutional policies. The study protocol was approved by the Institutional Review Board (IRB) of Juntendo University Hospital (IRB #20-036). Because all patient data in the NDB were deidentified, informed consent was waived by the Institutional Review Board (IRB) of Juntendo University Hospital (IRB #20-036).