Age-period-cohort analysis of stroke mortality attributable to high systolic blood pressure in China and Japan

Stroke is a principal cause of mortality in China and Japan. High systolic blood pressure (SBP) was considered a chief risk factor for stroke mortality. Herein, we evaluated temporal trends of high SBP-attributable stroke mortality in China and Japan between 1990 and 2017. Data on stroke mortality were retrieved from the Global Burden of Disease Study 2017 (GBD 2017). Using the age-period-cohort method, we computed overall net drifts, local drifts, longitudinal age curves, and cohort/period rate ratios (RRs) for high SBP-attributable stroke mortality. The age-standardized mortality rates (ASMRs) displayed decreasing trends for high SBP-attributable stroke mortality. The annual net drift values were − 1.4% and − 3.5% in Chinese men and women versus − 3.1% and − 4.9% in Japanese men and women. The local drift values in both countries were < 0 among all age groups but were lower in women than in men. The longitudinal age curves showed a greater high SBP-attributable stroke mortality in men than in women across all age groups. Similar decreasing patterns were shown in the period and cohort RRs in both sexes with women having a quicker decline than men. In China and Japan, the ASMRs, as well as the period and cohort RRs of high SBP-attributable stroke mortality, decreased between 1990 and 2017 in both sexes and across all age groups. Yet, the prevalence of high SBP remained worrisome in both countries. Thus, SBP control should be encouraged to prevent stroke mortality.

Statistical analyses. In this study, an APC framework was utilized to appraise the roles of age, period, and cohort in high SBP-attributable stroke mortality. The age effect represented the sociodemographic and biological factors affecting aging. The period effect reflected over-time changes affecting high SBP-attributable stroke mortality among all age groups such as applying preventive measures or providing new medical procedures. The cohort effect reflected the variations in high SBP-attributable stroke mortality across generations due to certain risk factors shared by each generation [22][23][24][25] . APC analysis has the following functions: (1) the net drift which assesses the annual percentage change via calculating the overall log-linear trend by cohort and period, (2) the local drifts which assesses the annual percentage changes for each age group via calculating the log-linear trend by cohort and period, (3) the longitudinal age curves which shows the fitted longitudinal age-specific rates in the reference cohort adjusted for period deviations, and (4) the cohort/period rate ratios (RR) that represents the cohort/ period relative risk, compared to the reference one, after adjustment for age and nonlinear cohort and period effects.
The study of the GBD 2017 recorded the date of death, location, sex, age (progressive 5-year age groups), year (consecutive 5-year periods from 1990), disease causes, risk factors, metric (number and rate of cases), and value (upper and lower). The SBP-attributable stroke mortality was measured based on the following four components: the number of stroke deaths, the exposure levels for SBP, the relative risk of stroke mortality due to SBP, and the counterfactual level of SBP exposure. Rates of SBP-attributable stroke mortality were calculated as the total number of stroke deaths multiplied by the population attributable fraction (PAF) for stroke-SBP pair in a given year, location, sex, and age; where the PAF identifies what percentage of stroke mortality could be avoided in a certain year if the exposure to SBP in the past were increased to the counterfactual level of the theoretical minimum risk exposure level (SBP ≤ 110-115 mmHg) 26 . The sex-specific high SBP-attributable stroke mortality rates in both countries were age-standardized according to the GBD 2013 global age-standard population and the ASMRs were calculated by the calculation formulas 18,27 . Since the period and age intervals should be fixed and equal in the APC tool, all individuals aged 80 years or older could not be analyzed in this study because they were recorded as one group only in the GBD database. SBP-attributable stroke mortality is rare below the age of 25 years, people aged < 25 years were also not included in the analyses. Central age groups, cohort, and period were defined as the reference.
The estimable functions were used to perform the APC analysis. APC Web Tool (Biostatistics Branch, National Cancer Institute, Bethesda, MD https:// analy sisto ols. nci. nih. gov/ apc/) was used to get the estimable parameters 28,29 .
We compared overall net drifts, local drifts, longitudinal age curves, and cohort/period rate ratios (RRs) for high SBP-attributable stroke mortality in men and women of China and Japan. Wald chi-square tests were used to investigate the statistical significance of the estimable functions. To assess the statistical significance of the slopes for the period/cohort RRs, the interaction between sex and calendar year/birth cohort was assessed using general linear models. All statistical analyses were performed using the SAS Version 9.4 software (SAS Institute Inc, Cary, NC), and all statistical tests were two-tailed.
Ethical statement. This study used deidentified publicly available data from the Global Burden of Disease Study 2017 repository. Thus, ethical approval and an ethical statement from an institutional review board or ethics committee were not required for the secondary analysis of data for China and Japan.

Results
Between 1990 and 2017, in China, the high SBP-attributable stroke mortality rate per 100,000 population declined from 164.7 to 108.7 in men and from 129.1 to 55.5 in women. In Japan, the corresponding rates were 63.7 and 24.7 in men and 35.9 and 8.9 in women, respectively. The overall declining ASMR trends in China were faster than before the year 2000 ( Fig. 1).
Throughout the period between 1990 and 2017, the annual net drift values were − 1.4% in Chinese men, − 3.5% in Chinese women, − 3.1% in Japanese men, and − 4.9% in Japanese women (p < 0.01 for all). In both countries, the local drift values were < 0 among all age groups and were lower among women compared to men. Yet, the sex-specific local drift curves for high SBP-attributable stroke mortality showed certain trends. In China, the trends were increasing in 45 years old among Chinese men and 55-79 years old among Chinese women. In Japan, these trends increased in both sexes before 45 years old but decreased in older ages (Figs. 2, 3). www.nature.com/scientificreports/ The peak mortality rates per 100,000 person-years reached 708.0 in Chinese men, 362.6 in Chinese women, 132.3 in Japanese men, and 39.8 in Japanese women. The high SBP-attributable stroke risk, in both countries, was higher among men than among women across all age groups (Fig. 4).
The period RRs among men and women showed declining trends in both countries (except for Chinese men in 1990-2000), with a relatively more rapid decline in women compared with men and in Japanese people compared with Chinese people. Despite Chinese people had lower high SBP-attributable stroke mortality than Japanese people during the 1990s, RRs crossed in the year 2000 before decreasing at a slower pace among Chinese people in comparison with Japanese people. Alike, the Chinese and Japanese cohort RRs crossed in the year 1950, with greater high SBP-attributable stroke mortality rates before then among Japanese people, but lower rates afterward in comparison with Chinese people. The cohort RRs in both countries showed declining trends among men and women with a more significant decline in women than men across all birth cohorts (Figs. 5, 6). The cohort and period RRs in both sexes were statistically significant whether in China or Japan (p < 0.01) in addition to the net and local drifts except for the local drift in Japanese men (Table 1).

Discussion
This study indicated that, between 1990 and 2017, the ASMRs and cohort/period RRs for high SBP-attributable stroke mortality showed declining trends among Chinese and Japanese men and women from all age groups. Yet, high SBP-attributable stroke mortality longitudinal age curves grew rapidly among Chinese men and women compared with their Japanese counterparts.
The declining trends of high SBP-attributable stroke mortality in China and Japan reflect the improvement of high SBP control in both countries. Such control could not have been achieved in both countries without the adherence of their people to healthy behaviors such as physical activity and salt restriction in addition to the significant signs of progress in health indicators such as healthcare coverage and advancement of the clinical diagnosis and treatment technology. These factors might have contributed together to the SBP control and consequently the declining trends of stroke in China and Japan 30-36 .
Of note, this study showed that high SBP-attributable stroke mortality was higher among men than women and older than younger people. Age is, by far, the topmost demographic risk factor for stroke. A prospective cohort study on a half million adults in China showed that SBP increased linearly with age in both sexes and was log-linearly associated with stroke risk. Every 10 mmHg higher of usual SBP was associated with a 30% higher www.nature.com/scientificreports/ risk of ischemic stroke and 68% increased risk of intracerebral hemorrhage, with men more vulnerable than women at every age group 37 .
It is worth pointing out that this study had two major strengths, First, it is the first study to investigate the impacts of age-period-cohort on the temporal trends of stroke mortality attributable to high SBP in China and Japan. Second, GBD 2017 provided estimates of cause-specific and age-and sex-specific mortality rates that were internally consistent, thus, the possibility of misclassification bias was unlikely. However, some limitations should be addressed. First, because the GBD data was generated by systematic reviewing and synthesizing of published and unpublished data, there are many uncertainty issues related to measurement errors, missing data, and systematic bias of the countries' prior estimates of the risk factors and outcomes pooled in the GBD 38 which could add bias to the cross-national comparisons. This is truly right for countries with poor vital registration systems. However, we compared stroke mortality, in Japan and China, where the death certificate diagnosis of stroke was highly valid due to the widespread use of CT and MRI in both countries since the 1980s 39,40 . Also, the GBD data estimated SBP for the year, sex, and age group, and was a more comprehensive system of risk analysis than any previous global or national analysis which could assess the changes in SBP burden 41 . Second, since period and  www.nature.com/scientificreports/ age intervals have to be fixed and equal in the APC tool, those who were aged ≥ 80 years were excluded from this study because they were merged in one group only in the GBD database. However, the stroke mortality rate among people aged ≥ 80 years was shown to decline in a similar pattern 42 . Another, we did not include people aged ≤ 25 years as stroke mortality attributable to high SBP is considered negligible in this age group. Third, we could not stratify our results by stroke subtype because of the absence of data on the Japanese stroke subtype. Fourth, one main limitation of the APC analyses is that it could be conducted only on the pair of a risk factor and an outcome (high SBP and stroke ASMR in this research). Many other risk factors, such as alcohol intake, tobacco smoking, and even clinical advancement contribute to the ASMR of stroke and may substantially affect high SBP-attributable stroke ASMR because these factors are related to both the blood pressure and the stroke mortality, adding to the above-mentioned uncertainties.
In conclusion, the ASMRs, as well as the period and cohort RRs of high SBP-attributable stroke mortality, declined in China and Japan in both sexes and across all age groups during the period between 1990 to 2017. However, the proportion of the high SBP population remained significant in both countries. Thus, policymakers, especially in China, should prioritize hypertension control approaches to reduce the future risk of stroke mortality.