Childhood cancer mortality in Japan, 1980–2013

This updated analysis provides the latest information on trends in mortality from
childhood cancer in Japan. We found that dramatic declines in mortality occurred during
1980–2003 for boys and 1980–2001 for girls. In the last 5 years, there have been modest
but statistically insignificant reductions in mortality for both sexes.

The declines in childhood cancer mortality in the 1980s and 1990s are most likely
due to improved survival. Early diagnosis and better therapies for childhood cancer
over recent decades have greatly improved the prognosis of pediatric cancer patients
14]. The 5-year survival rate reached 79 % for children with cancer diagnosed between
1998 and 2000, based on a population-based report from Hiroshima, Japan 15]. The contribution of changes in cancer incidence to mortality decline remains unclear
because of the absence of incidence data in Japan. A population-based study in Osaka,
Japan that found a decline in the incidence rate of all cancers combined in the 1990s
suggested that the constant decline in death from childhood cancer during 1973–2001
was primarily due to improved survival between the 1970s and 1980s and reduced incidence
after the 1990s 16].

The impressive decrease in leukemia mortality is consistent with a substantial increase
in survival from this disease, particularly in patients with acute lymphoblastic leukemia.
Standardization of the applied protocol and the expanded use of chemotherapeutic agents
and combination regimens have improved the treatment of childhood leukemia 17]–19]. In the latest report on pediatric hematological malignancies from the Japan Society
of Pediatric Hematology/Oncology, the 5-year overall survival rate for patients diagnosed
between 2006 and 2010 reached 88.7 % for acute lymphoblastic leukemia and 75.2 % for
acute myeloid leukemia 4].

Malignant CNS tumors continue to be the second largest contributor to cancer-related
mortality in Japanese children. We found a slight but statistically significant increase
in mortality of childhood CNS tumors in Japan. This increase in mortality implies
a modest increase in incidence of CNS tumors. A recent study of mortality of CNS tumors
by subtype between 1993 and 2013 indicates that mortality rises for the tumors of
C71.7 (brain stem) and C71.5 (cerebral ventricle). Since patients with tumors in these
sites have a poor prognosis, the trends of cancer incidence and mortality for these
sites are thought to be similar. Therefore, the study author speculates that tumors
occurring in these sites are increasing 20]. Although this explanation is reasonable, since childhood CNS tumors are rare and
even rarer by subtype, whether there is a true increase in brain stem or cerebral
ventricle tumors in Japan awaits confirmation. More importantly, our results highlight
the necessity of establishing an active, nationwide, childhood cancer registry to
provide reliable incidence information for Japan.

The lack of decline in the mortality rate of CNS tumors in recent years has also been
reported in other countries. For example, a population-based study in Australia found
a stable trend in mortality between 1998 and 2008. The author attributed this finding
to fluctuations in CNS cancer incidence and stable survival rates 21].

Stable trends were observed in mortality of all cancers combined in recent years in
Japan. Accompanying the rapid reduction in the number of deaths due to leukemia, the
proportion of deaths from CNS tumors increased over the studied period. In 1980, leukemia
accounted for nearly 52 % of childhood cancer deaths; by 2013, it accounted for 34 %.
In contrast, CNS tumors’ share of all cancer deaths increased from about 9 % in 1980
to 29 % in 2013. With this recent shift in disease distribution, the number of deaths
from leukemia has become similar to that from CNS tumors in Japan. This means that
trends in mortality of CNS tumors tend to contribute more to the changes in mortality
of all cancers combined. The recent statistically insignificant drop in mortality
from all cancers combined is due in part to the rise in death from CNS tumors during
the study period.

Our estimates of childhood cancer rates and trends for Japan are similar to those
of four comparison countries, but some differences exist. For example, mortality from
CNS tumors was lower in Japan than elsewhere. That difference might be explained by
the relatively low disease incidence rate in Japan since treatment achievements from
using the same regimens are similar in Japan and other developed countries. Study
of the Osaka Cancer Registry showed an incidence of CNS tumors in children of about
two per 100,000 population between 1988 and 2001 16]. Even lower rates resulted from a study of childhood cancer incidence from 1993 to
2001 based on 15 population-based cancer registries in Japan: 1.5 per 100,000 population
for boys and 1.4 for girls 22]. In the same period, a study from the Surveillance, Epidemiology, and End Results
Program reported a higher childhood cancer incidence rate in the US, over three per
100,000 population 23].

Data consistency and accuracy should be considered when comparing Japan with other
countries. Our study used the WHO mortality database, an established, international
death file, to evaluate cancer trends. According to Mathers et al., for developed
countries the quality of cause-of-death information is good and the coverage of populations
is complete 11]. Hence, we consider our comparison between Japan and the other four countries in
this study to be reasonable and reliable.

The statistical methods used to evaluate changes in mortality rates may have influenced
the observed patterns. In this study, we described trend characteristics using joinpoint
analysis, a method used extensively to describe mortality trends in epidemiological
cancer studies. Joinpoint analysis allows identification of points of change and determination
of the trends between joinpoints, considering variations within the relevant period.
In addition, we calculated both APC and AAPC, which may provide complete characterization
of the trend for the whole study period.

It is important to address the limitations when explaining and comparing results involving
mortality statistics. First, as described above, the quality of data differs among
countries in the WHO mortality database, and thus comparison is limited to those countries
with good completeness and accuracy of death certification. Second, the rarity of
childhood cancer deaths limits the analysis of changes in cancer mortality by subtype.
Finally, no known study has measured the accuracy of cancer death certificate diagnoses
for children with cancer. Although a study of all age groups reported 65 % agreement
of cancer diagnoses between death certificates and hospital records, whether this
is also true for pediatric cancer is unknown 24], 25].