How are new drugs disseminated in Japan? Analysis using the National Database of Health Insurance Claims of Japan

Abstract Drug lag refers to the difference in the time of a new drug's approval in different countries; the dissemination of the new drug after approval within the countries is another problem. We examined the nationwide dissemination of 11 cancer drugs approved in Japan between 2011 and 2015 using the National Database of Health Insurance Claims data. We extracted data on the number of cancer drug prescriptions from 47 prefectures and associated demographic information, such as age and sex. Eight diabetes drugs were also examined for comparison. We observed a lag between the marketing approval date of the drugs and their first use. To further explore the rise and pattern of each drug’s dissemination, we analyzed the trend of the cumulative number and total of new prescriptions for each prefecture. The results showed that the first month of new cancer drug prescriptions varied across prefectures. On average, they lagged by up to 2 months in the slowest prefectures, whereas the variation was almost nonexistent for diabetes drugs. The patterns of dissemination varied more among cancer drugs across the seven Japanese geographical regions. After the initial prescription, the number of prescriptions showed a steep rise for most cancer drugs, whereas the increase was gradual for diabetes drugs. In conclusion, the dissemination of cancer drugs had a greater lag time than that of diabetes drugs. Further research is needed to explore the causative factors to ensure that all effective drugs are equally accessible for those who need them.

approved in Japan between 2011 and 2015 using the National Database of Health Insurance Claims data. We extracted data on the number of cancer drug prescriptions from 47 prefectures and associated demographic information, such as age and sex.
Eight diabetes drugs were also examined for comparison. We observed a lag between the marketing approval date of the drugs and their first use. To further explore the rise and pattern of each drug's dissemination, we analyzed the trend of the cumulative number and total of new prescriptions for each prefecture. The results showed that the first month of new cancer drug prescriptions varied across prefectures. On average, they lagged by up to 2 months in the slowest prefectures, whereas the variation was almost nonexistent for diabetes drugs. The patterns of dissemination varied more among cancer drugs across the seven Japanese geographical regions. After the initial prescription, the number of prescriptions showed a steep rise for most cancer drugs, whereas the increase was gradual for diabetes drugs. In conclusion, the dissemination of cancer drugs had a greater lag time than that of diabetes drugs. Further research is needed to explore the causative factors to ensure that all effective drugs are equally accessible for those who need them.

K E Y W O R D S
cancer drug, diabetes drug, drug dissemination, drug lag, national database reduction in reviewing time. 2,[5][6][7] However, drug approval is just a prerequisite for use in clinical practice. Little is known about how quickly drugs are included in frontline clinical practice, and geographical disparities. As equitable health-care delivery becomes a priority, vis-à-vis legislative efforts to ensure easy access to medicines, it is now imperative that all effective drugs are equally accessible for those who need them.
Therefore, this study aimed to investigate the nationwide dissemination of selected newly approved cancer drugs between 2011 and 2015 using the National Database of Health Insurance Claims (NDB). We used some diabetes drugs newly approved within the same period as the comparison group to illuminate the situation further.

| National database
We obtained data from the longitudinal database of electronic health insurance claims submitted from all insurance entities in Japan (NDB). In Japan, 99% of hospitals submit insurance claims electronically, which are accumulated in the NDB. 8 These data contain information on all medical services covered by public health insurance, including diagnostic tests, surgery, and prescription drugs.
It also includes demographic information such as the prefecture of patient residence, age, and gender. A de-identified unique number is allocated to each patient, allowing users to follow all medical procedures a patient has undertaken in chronological order within the defined period. We used NDB data between January 2010 and December 2017. The 2010 data was used to ensure that no drugs were prescribed before the first commercial sale.
This study protocol was approved by the institutional review board of the National Cancer Center, Japan (approval number: 2017-265).

| Drugs and variables
We extracted prescription data from the NDB on 47 patient residence prefectures and associated demographic information such as age and sex. The targeted drugs were either cancer or diabetes drugs approved between January 2011 and December 2015. Information on the first month of prescription and changes in prescription volume were collected on the 23 cancer drugs and eight newly approved diabetes drugs. However, because the NDB Reporting Guidelines prevented us from reporting numbers less than 10 for privacy protection, any drugs with fewer than 10 prescriptions were excluded.

| Data analysis
We sought to determine whether there was a lag time between when the drugs first became available on the market and their prescription in practice from the perspective of geographical dissemination. To do this, we determined the month in which a drug was first prescribed. For each prefecture, we calculated the difference between the month the first prescription was observed anywhere and the month the first prescription appeared in each prefecture. We calculated the average lag time for all the analyzed cancer drugs to characterize the prefecture's new drug uptake pattern.
Next, we examined the uptake of newly approved drugs and their geographic variation by the trend of the cumulative number of new prescriptions for each prefecture. For this analysis, we chose the drug with the largest number of prescriptions for cancer patients (degarelix acetate) and for diabetes patients (ipragliflozin).
Since degarelix was only used for men, nivolumab, which had the largest number of prescription among those used for both males and females cancer patients were also analyzed.
To look at the trends in new drug uptake, we also analyzed the number of prescriptions for both cancer and diabetes drugs. The analysis included all patients administered newly approved cancer drugs and several diabetes drugs between January 2011 and December 2015.
A concave curve illustrates a rapid increase of new prescriptions in the early phase, while a convex curve shows slow uptake and later acceleration. As all sodium-glucose cotransporter 2 (SGLT2) and glucagon-like peptide-1 drugs can be used interchangeably, we presented the aggregated result of all included diabetes drugs.
All NDB data extraction and processing were undertaken using Oracle SQL*plus (Oracle Corporation). Aggregated results were analyzed using Python 3.5.1.

| RE SULTS
The data of 346,729 patients with cancer and 1,436,990 patients with diabetes were analyzed. Of the patients with cancer, 211,876 (61.1%) were men, and the mean age was 68.6 ± 12.7 years. Similarly, 886,143 (61.6%) of the patients with diabetes were men, and the mean age was 60.0 ± 12.7 years. Sodium-glucose cotransporter 2 was prescribed to 1,345,144 patients (mean age, 59.5 ± 12.5 years), and 62.2% of them were men. Glucagon-like peptide 1 was pre-

| First month of prescription
For all drugs, the earliest prescriptions were observed in the same month as the drug approval. The slowest cancer drug uptake was for axitinib and nivolumab, which took 12 months and 8 months, respectively. The slowest for a diabetes drug, dulaglutide, was 2 months.
The average lag time in prefectures ranged from 0 to 2.3 months.
The lag time for each prefecture was plotted against the prefecture population for cancer drugs ( Figure 1A) and diabetes drugs ( Figure 1B). Regarding the 21 cancer drugs, the gap tended to diminish as the population increased ( Figure 1A). The graph also showed that, especially for cancer drugs, the average lag time varied even among prefectures with similar population sizes. For the eight diabetes drugs, the gap was very small or almost nonexistent, regardless of the population size ( Figure 1B). Figure 2A,B show the dissemination of the cancer drugs nivolumab and degarelix acetate in seven administrative regions in Japan by the cumulative percentage of new prescriptions in proportion to all prescriptions up to 36 months. We used 36 months as it was observed to best illustrate the trend for most drugs. Figure 2C shows the same graphs for ipragliflozin, the first SGLT2 inhibitor sold. However, when we compared the dissemination patterns of the graphs between cancer and diabetes drugs, there was more variation for cancer drugs, with similar trends across the region. Furthermore, densely populated areas such as the Kanto or Kinki areas had less variation in dissemination. Figure 2D shows the same curve for all diabetes drugs, as most diabetes drugs shared the same effect. Figure 2C shows that there was little variance in diabetes drug dissemination across the regions. The trend was even more apparent in Figure 2D after the diabetes drugs were grouped.

| Trends in number of prescriptions
The monthly prescriptions for cancer and diabetes drugs are shown in Figure 3. While a steep rise of new prescriptions is seen in the initial phase for cancer drugs, there is a gradual rise in prescriptions over 20 months for diabetes drugs after 12 months.

| DISCUSS ION
Our study shows that the dissemination of cancer drugs lags more than that of diabetes drugs and has greater geographic variation. This is the first study in Japan that investigated the geographic variation in the dissemination of new drugs. In contrast, prior studies that examined the dissemination of new cancer drugs were from other countries, and they focused on practice (eg, practice size and teaching status) and drug characteristics (eg, indication of the drugs and promotion by the manufacturer). [11][12][13] Geographic variations are important from the perspective of access to quality care. 14-16 Such variation suggests that location and the number of medical providers affect patients' access to cancer drugs. Prior studies have shown that the geographic variation is associated with access to optimal treatment and consequently mortality [17][18][19] ; therefore diminishing the gap in accessibility becomes pertinent. privacy policy of the NDB. More detailed information, such as per capita provider volumes and distance to the capital city in a smaller area unit, would have enabled more detailed analysis.
The drug approval lag between Japan and other countries has improved, but drug dissemination patterns remain relatively unknown. Our study showed that the dissemination of cancer drugs appeared to lag more than for diabetes drugs. This finding is worth attention because cancer drugs have fewer alternatives than diabetes drugs. Further research is needed to ascertain the factors that contribute to lags in drug dissemination to ensure that effective drugs are equally accessible for those who need them.

ACK N OWLED G M ENT
This study was funded by the National Cancer Center Research and Development Grant (31-A-21).

D I SCLOS U R E
The authors have no conflict of interest.

E TH I C A L A PPROVA L
This study protocol was approved by the institutional review board of the National Cancer Center, Japan (approval number: 2017-265).