Data validation
We chose rare diseases from the ICD10-based Standard Disease-Code Master through a semiautomatic process to follow NDB regulations. In Figure 3, we roughly compare the prevalence distribution in NDB with that in Orphanet to validate the disease selection. They show the same trend, indicating that most rare diseases affect only several persons per 100,000. Moreover, Figure 4 demonstrates that more than 60 percent of the diseases in NDB have equivalent terms in other rare disease databases. Therefore, it is our informed opinion that the selected diseases can mostly be recognized as rare.
Prevalence
To reveal the difference between the prevalence of each disease in NDB and that in Orphanet, we compared them by discretizing the actual prevalence value in NDB. Figure 5 shows that the difference of about 80 percent of the diseases is only within one category. It is also revealed that most diseases belong to less than 1–9/100,000. If we pay attention to the Japanese data, a difference in one category means practically less than about one error for 1,000 Japanese patients. Accordingly, the analysis implies only a slight discrepancy between NDB data and Orphanet data.
If we focus on the relatively small difference, given that the categories of Orphanet tend to be higher, there seem to be two possible explanations. First, as Orphanet mainly contains prevalence in European countries and most diseases with a categorical difference are genetic, it could reflect an ethnic group difference. The other is that rare diseases in many Japanese patients remain undetected. The latter reason can be supported by the fact that there are much fewer genetic tests available in Japan than there are in the West [30]. On the other hand, if we look closely at the relatively significant difference, 11 diseases indicate a higher category in Orphanet by more than four categories. This tendency might be due to excessive regional accumulation, such as in the cases of schistosomiasis and Cooley anemia.
As for the comparison between Japanese prevalence in Orphanet and NDB, Figure 6 shows the same results as aforementioned. As most diseases are within one category’s difference, the Orphanet data are consistent with the NDB data.
Natural history
According to Figure 8 and Figure 9, high-density areas seem to accord with the Orphanet categories, predefined on the basis of age distribution patterns. Therefore, the natural history of most diseases in NDB is consistent with that in Orphanet in general.
However, some diseases indicate different natural histories in Japan. This could happen because of the unique therapeutic environment available in each country; clinical variability caused by ethnic factors, such as genetic variation and diet; and artifacts accompanied by time-series clustering. To focus on the apparent distinction of natural history, we calculated the number of diseases with more than two categorical differences, excluding those with the “All ages” onset category, “normal life expectancy” death category, and “any age” death category (Table 1, Table 2).
Table 1
The differences in the average age of onset in NDB and Orphanet
Case 1. Average age of onset is later in NDB than that in Orphanet |
| NDB | Orphanet | Number of diseases |
| Acquired | Antenatal, Neonatal, Infancy | 21 |
| Elderly | Antenatal, Neonatal, Infancy, Childhood | 80 |
Case 2. Average age of onset is earlier in NDB than that in Orphanet |
| NDB | Orphanet | Number of diseases |
| Congenital | Adolescent, Adult, Elderly | 4 |
Table 2
Differences in the average age of death in NDB and Orphanet
Case 3. Average age of death is later in NDB than that in Orphanet |
| NDB | Orphanet | Number of diseases |
| Childhood | stillbirth, infantile | 1 |
| Acquired | infantile, adolescent, late childhood | 2 |
| Elderly | infantile, stillbirth, early childhood, adolescent | 8 |
Case 4. Average age of death is earlier in NDB than that in Orphanet |
| NDB | Orphanet | Number of diseases |
| Congenital | adult, elderly | 4 |
| Childhood | elderly | 1 |
Tables 1 and 2 show the disagreement between NDB-based age distribution patterns and Orphanet categories of the average age of onset and the average age of death, respectively. As some diseases have multiple categories in Orphanet, we defined coincidence between two databases as the condition for more than one Orphanet category to be included in the range of the NDB category.
In Case 1, Japanese patients tend to present symptoms later than those in the Orphanet data do. About two-thirds of the diseases are genetic, and only 30% are covered by the Nanbyo system for children. Moreover, as mentioned above, genetic tests of numerous diseases are not available in Japan. This seems to relate to the delayed detection of Japanese patients. On the other hand, Case 2 indicates that the average onset age of the four diseases is earlier in NDB than it is in Orphanet. All diseases, except encephalitis lethargica, are categorized as cancer, and the number of patients demonstrates a bimodal distribution. As our time-series clustering could not catch bimodality, we observed this difference.
Case 3
implies that Japanese patients live longer than those in the Orphanet data do. This seems due to the bimodal distribution of diseases that can be secondary, such as biliary atresia. Conversely, in Case 4, five diseases in all, Bloom syndrome, achondroplasia, 4p deletion syndrome, Dent disease, and hereditary pancreatitis, show earlier death. Although it is difficult to determine the reason, this may be due to the variation of their severity or phenotypes and their therapeutic environment.
Japanese policy on rare diseases
Although Japan has provided medical expense subsidies for Nanbyo patients, fairness between rare diseases has always been controversial. That is to say, the increasing number of rare diseases casts doubt on the legitimacy of a dividing line between Nanbyo and non-Nanbyo rare diseases although several diseases are added to Nanbyo almost every year. In fact, there are some procedures and obstacles to having a disease covered by the Nanbyo system. First, an official research group focusing on a specific disease is required to be established. Its role is to set diagnostic criteria or a guideline and reveal its prevalence, mainly through mail questionnaire surveys of medical institutions all over the country. Then, the advisory board of each Nanbyo system discusses whether it should be added to the system on the basis of its report.
Consequently, only a few diseases pass after receiving public comments every year. A serious problem, in particular, among various concerns, is that as research groups can only focus on limited rare diseases, it depends on chance or the degree of disease recognition at best, whether a rare disease is designated as Nanbyo and its patients can receive medical support. Accordingly, ultra-rare diseases in particular have hardly any opportunities to be examined.
As aforementioned, one of the requirements for designation is that the number of patients with a specific disease be less than 120,000, which indicates 0.1% of the Japanese population. In light of this regulation, Figure 2 suggests that approximately 4,000 diseases are supposed to be covered, which means newly adding 3,000 diseases. Moreover, even if the Japanese government comprehensively adopts the Orphanet criteria to meet global standards, Figure 4 implies that about 800 diseases are left outside the coverage of the Nanbyo systems. Strictly speaking, as Nanbyo is not always included in Orphanet, 1,085 non-Nanbyo diseases recognized as rare afflict patients as per Orphanet. Therefore, despite the unique circumstances surrounding Japan’s healthcare system, it is evident that Nanbyo does not cover enough rare diseases.
To ensure fairness, Japanese policies need to promote active comprehensive research to detect rare-disease patients, referring to the medical databases in the world, and become more flexible in providing medical expense subsidies. In addition, even if a specific disease is inside the system, patients afflicted by it need to be strictly examined for eligibility for medical support based on objective criteria, such as phenotypes, disease severity, medical costs, and quality of life, although this does not easily seem acceptable to patients who have already been designated as eligible. Accordingly, these enable a significant turnover of patients eligible for the support and sustainability of Nanbyo systems.
Limitations
As NDB was not originally constructed for analysis, it presents several limitations in its use. First, as diseases in the NDB data are standardized to meet the procedural needs of clinical practice in Japan, our NDB data could inevitably be incomplete according to the definition of rare diseases in the world. The shortage of rare diseases is mainly because they are too rare in Japan to be added to the code master. In addition, as we semi-automatically chose them from 23,939 diseases, we could have overlooked or excessively detected diseases. Therefore, it is still challenging to thoroughly elucidate the circumstances of rare-disease patients in Japan.
Second, the system of health insurance claims and procedures complicated the interpretation of the NDB data. Some patients are exempt from any medical expense, such as welfare recipients, atomic bomb survivors, and people under the severe condition of insanity. It is sometimes challenging for persons with rare, severe diseases to work. In such cases, the NDB data do not contain their information. Moreover, it is said that 80% of diseases labeled “other diseases” in the code master are wrongly allocated although there are correct categories for them [31]. Consequently, the number of patients seems underestimated. On the other hand, its overestimation can occur because doctors often make a temporary diagnosis stored in the NDB to prescribe common treatments such as analgesics under health insurance. However, this is primarily true of common diseases; hence, this artifact has little effect on the interpretation of our data. Another confusing factor is that both disease and subtype are coded in the master. For example, as Niemann–Pick disease and Niemann–Pick disease type A are on the list, it is nearly impossible to know whether patients are counted conflatingly or exclusively, which depends on doctors or clinical coders.
The last limitation is that in calculating prevalence, we might have underestimated it because we adopted the entire population of Japan as a denominator instead of the total population in the NDB data, assuming that almost all the people receive medical treatment under health insurance more than once a year.