As illustrated in the previous chapters, the literature suggests that the Japanese population has unhealthy consumption patterns. In particular, the consumption of salt, alcohol and combustible tobacco is too high, while the intake of dietary fiber is notably low. Our data analysis confirms these findings and reveals distinctive consumption patterns in Japan that diverge significantly from nutritional guidelines.
Table 1 illustrates these disparities, comparing the baseline model values of fruits, vegetables, fiber, salt, cholesterol, fat, alcohol, and combustible tobacco with counterfactual model values of those substances. As described in the methodological and data chapter, the baseline model values are the actual consumption derived from survey data. The counterfactual model values consist of recommended consumption amounts for dietary risk factors and considered limits for the consumption of alcohol and combustible tobacco.
We observe a national dietary behavior that closely aligns with the recommended intake levels of fruit and vegetables but deviates markedly in other crucial areas. Specifically, the Japanese diet is characterized by a significantly higher intake of salt, fat, and cholesterol and a markedly lower consumption of dietary fiber than recommended. This trend shows negligible differences between genders, indicating that both men and women share dietary habits in these respects. Salt illustrates this well: while the recommended limit of salt consumption is 5 grams per day, women consume around 9 grams and men around 11 grams.
Table 1
Baseline and Counterfactual Model Values
|
Males
|
Females
|
Risk Factors
|
Baseline
|
Counterfactual
|
Baseline
|
Counterfactual
|
Fruit (grams/day)*
|
224
|
200
|
230
|
200
|
% with < 1 fruit per day
|
71
|
0
|
60
|
0
|
Vegetable (grams/day)**
|
276
|
200
|
268
|
200
|
% with < 1 vegetable per day
|
7
|
0
|
8
|
0
|
Fiber (grams/day)
|
19
|
25
|
17
|
25
|
Salt (grams/day)
|
11
|
5
|
9
|
5
|
Total fat (% total energy)
|
28
|
22.5
|
29
|
22.5
|
Cholesterol (milligrams/day)
|
371
|
300
|
317
|
300
|
% low alcohol consumers***
|
39
|
99
|
69
|
99
|
Alcohol (grams/day)
|
24
|
10
|
12
|
10
|
% Former smoker
|
12
|
18
|
3
|
5
|
% Current smoker
|
28
|
22
|
8
|
6
|
*Of those consuming more than 1 fruit portion (100g) per day |
** Of those consuming more than 1 vegetable portion (70g) per day |
*** Consuming < 1 gram of alcohol per day |
Source: Own calculations, based on the Japan National Health and Nutrition Survey (2019).
Beyond dietary behavior, Table 1 also illustrates the consumption of the harmful substances alcohol and combustible tobacco. For both products, the consumption is considerable and exceeds the counterfactual model values by far. Interestingly, whereas dietary behavior was similar between men and women, the consumption of alcohol and tobacco exhibits a stark gender disparity. Men in Japan consume alcohol and tobacco at rates far exceeding the recommended limits, whereas women's consumption patterns are closer to the guidelines. For instance, 31% of Japanese women consume alcohol, with an average of 12 grams of alcohol per day, whereas 61% of men consume alcohol, with almost 25 grams per day. That is, more than twice as much.
The gender divergence in alcohol and tobacco consumption is not only a matter of lifestyle choice but also impacts health outcomes, as shown in Fig. 1. The excessive and risky consumption patterns observed among men have profound implications for public health, with the potential to reduce disease incidence significantly. With PRIME we estimate the number of avertable NCD incidences for 2019 if individuals had adhered to recommended dietary and lifestyle guidelines. We find that approximately 346,000 disease cases among men and 218,000 cases among women could have been prevented.
Furthermore, the age-related analysis of avertable disease cases reveals another layer of gender differentiation. For men, the incidences of preventable diseases sharply increase between the ages of 40–44 and 65–69 years, thereafter remaining at varying high levels. In contrast, women experience a moderate increase, with incidences plateauing between the ages of 65–69 and 80–84, followed by an abrupt spike beyond the age of 85. The differences in age-related patterns between men and women suggest that lifestyle choices, such as higher alcohol consumption and smoking behavior among men, contribute to earlier incidences of NCDs in men.
In Fig. 1, we introduced the concept of avertable incidence, which we further dissect in Fig. 2 by presenting avertable incidences by disease and risk factor. This breakdown offers insights into the changes needed to improve health outcomes in Japan over the long term. The largest potential reductions in disease incidence are associated with cerebrovascular disease, diseases of the colorectum and ischemic heart disease, accounting for preventable cases of 152,000, 138,000, and 115,000, respectively. The distribution of risk factors contributing to these diseases is notably heterogeneous, indicating that interventions across various areas are necessary to enhance public health. Simultaneously, many areas present opportunities for significant health improvements through targeted measures. The sectors of dietary fiber, fruit and vegetable intake, smoking, alcohol, and salt consumption each contribute 15 to 20 percent to the potential reduction in disease incidence. This diverse contribution of risk factors underscores the complexity of public health challenges in Japan and highlights the multifaceted approach required to address them effectively.
While avertable incidence provides a measure of the potential reduction in disease and subsequent mortality rates, the pragmatic aspect of cost savings associated with treating these diseases is equally pertinent for governments. Given that costs for various diseases can differ substantially, a detailed examination of the cost savings achievable through modifications in risk factors is crucial.
This analysis is presented in Table 2. Results for avertable costs through decreased alcohol consumption exclude diabetes and ischemic heart disease. This is further discussed in the limitations sub-section. According to our counterfactual scenario, which posits a shift towards healthier living, overall savings of $35.1 billion USD could be realized, as indicated in column (1) and (2). Of these savings, $17.9 billion could be saved through reduced salt intake, $6.5 billion through reduced alcohol consumption, $6.0 billion through increased consumption of fruits and vegetables, and $3.4 billion from changes in smoking habits.
We assume that HTPs are associated with a 30% lower health risk than cigarettes, which is a conservative assumption in the context of the current body of literature. Assuming that the lack of combustion in HTPs leads to even less emitted toxicants and a likely risk reduction level of up to 90%, this would increase savings. In such a case, we would expect the current savings of $3.4 billion to increase three times. If we projected the savings of former and current smokers into the future, this would increase economic savings by a factor of ten. Such high long-term savings of ten times the 2019 savings can be explained by a large number of middle-aged current smokers that are expected to cause smoking-attributable costs in the future. Those future costs for potentially 10 to 20 years can be reduced if they switched to HTPs today.
To put our estimated cost savings for 2019 into perspective, we estimate the total attributable costs following a population-attributable fraction approach. Such an approach includes all costs attributable to harmful consumption, even if consumers have already adopted healthier behaviors but their health is not yet equivalent to that of non-risky consumers. In the case of tobacco, total attributable costs include costs from current and former smokers. 75% of total costs can be attributed to current and 25% to former smokers. Thus, only the 75%-portion of costs can actually be reached by policy interventions and be reduced.
We find that the unhealthy consumption habits in Japan in 2019 can be associated with $343.2 billion in total costs, as can be seen in Table 2 in column (3) and (4). Of which $113 billion can be attributed to salt, $80.3 billion to alcohol and $66 billion to current and former smokers. Consequently, we infer that at least 10% of the costs attributable to unhealthy consumption could have been averted in 2019 if individuals had adhered to healthier lifestyles.
The significant financial impact of unhealthy consumption underscores the complex relationship between health behavior, disease incidence, and social and economic costs. It highlights the importance of considering both the health and economic benefits of risk factor modification. This dual focus not only supports the goal of improving public health but also aligns with governmental objectives of reducing healthcare expenditures. By targeting these key areas—alcohol and salt consumption as well as smoking—policymakers can devise strategies that not only enhance the population's health but also generate substantial economic savings, thereby creating a win-win scenario for both public health and the economy.
Sensitivity Analysis – Monte Carlo Simulation
The Monte Carlo simulation, used in this study as a sensitivity analysis for the PRIME model, is a statistical technique that explores variance in model outcomes by drawing random samples from probability distributions of uncertainty sources. This approach assesses the robustness of the model's results considering the uncertainties in underlying data. By incorporating confidence intervals for the lower and upper bounds of relative risks, the simulation provides a detailed evaluation of the dispersion of outcomes across 5,000 iterations. This offers a more comprehensive perspective on the potential impacts of behavioral risk factors on the prevention or delay of disease incidences than a single analysis might.
The Monte Carlo results for this study, assessing the avertable incidences associated with excessive consumption of salt, smoking, alcohol, fats, and insufficient intake of fiber, fruits, and vegetables in Japan in 2019, are shown in Table 3. Notably, the results illustrate that dietary factors, specifically fruits and vegetables, have the most significant effect with an average of 151,000 averted incidences, followed by salt, fiber, alcohol, smoking, and fats.
However, it should be noted that the averted incidences for fruits and vegetables in Japan might be overstated due to the PRIME model's inability to aggregate daily consumption of fruits and vegetables. This limitation overlooks the compensatory health benefits of high vegetable intake when fruit consumption is low, potentially skewing incidence rates and associated healthcare costs, especially in populations like Japan's, where fruit and vegetable intake significantly surpasses recommended levels. This must be considered when interpreting dietary behaviors' impact on public health outcomes.
The average averted incidences for salt intake stand at 120,000, highlighting the importance of salt reduction in the diet. Smoking and alcohol, well-recognized health risk behaviors, show averages of 79,000 and 85,000 averted incidences respectively, underscoring the impact of substance use on public health. Fiber shows a significant preventative potential with 110,000 incidences averted on average, while reduced consumption of fats is associated with the least impact, showing a mean of 23,000 averted incidences. Overall, sensitivity analysis results show a total mean of 564,000 averted incidences across all risk factors, underscoring the potential for public health interventions to lessen the impact of NCDs. By transitioning towards a healthier population, between 461,000 (at the 2.5th percentile) and 663,000 (at the 97.5th percentile) incidences could have been averted. This range highlights the profound effect that comprehensive health promotion and disease prevention strategies can have, illustrating a scenario where a shift in population health behaviors towards recommended guidelines dramatically reduces the burden of disease.
Table 3
Incidences averted or delayed by behavioral risk factor, in thousands
Risk factor
|
2.5th percentile
|
Mean
|
97.5th percentile
|
Salt
|
52
|
120
|
180
|
Fruits & Vegetables
|
112
|
151
|
193
|
Alcohol
|
28
|
85
|
137
|
Smoking
|
60
|
79
|
96
|
Fiber
|
69
|
110
|
151
|
Fat
|
16
|
23
|
30
|
Total*
|
461
|
564
|
663
|
Source: Own calculations, based on the PRIME |
*Risk Factors and Total are simulated separately within Monte Carlo simulation, thus values do not necessarily add up |
Limitations
Despite its strengths, the PRIME model, designed to evaluate changes in risk factors for non-communicable diseases, faces several limitations that affect its utility in public health strategy formulation. Unlike tools geared towards forecasting, PRIME is tailored to retrospective analysis, relying on fixed relative risks, which may not fully encapsulate the evolving nature of risk factor-disease relationships over time. This inherent design choice restricts its application mainly to historical data assessment rather than predicting future trends in NCD incidence.
The PRIME model is supposed to incorporate mortality data as input, but in this paper, we consider incidence data. This shift towards preventable disease incidences emphasizes the importance of healthier consumption patterns, underlining how such changes can lead to decreased healthcare costs associated with treating NCDs. This perspective not only broadens the understanding of disease impact but also paves the way for more effective prevention strategies in public health policy. However, the model's dependence on literature-based relative risks introduces potential biases, especially since some of these risks are more aligned with mortality than with the likelihood of contracting a disease. Nevertheless, the relative risks in the PRIME model are not consistently solely based on mortality, but also on contracting a disease.
Furthermore, the PRIME model's limitation in aggregating daily fruit and vegetable consumption fails to account for the nuanced dietary patterns observed in some populations. By not considering the combined daily intake of these foods, the model overlooks how a high-vegetable diet might offset the lower consumption of fruits, a factor that could significantly alter disease incidence rates and the associated healthcare costs projected by the model. Consequently, this oversight likely leads to an overestimation of avertable incidences and the related healthcare costs in the model's results. This overestimation is particularly critical in contexts like Japan's, where the consumption levels of vegetables mainly surpass standard dietary recommendations, while fruit consumption falls short of expectations. This potentially leads to discrepancies in the model's predictions regarding health outcomes and costs.
The reliance on incomplete Global Burden of Disease (GBD) incidence data further complicates the model's accuracy. For example, the absence of data on hypertensive disease in Japan may lead to an underestimation of disease burden attributable to risk factors. Additionally, while the model's use of incidence data over mortality offers a broader view of disease impact, the application of mortality-based relative risks could bias the results.
The direct financial savings estimated in this paper must also be interpreted with caution, as they are potentially underestimated in the model due to the lack of disaggregated cost data for some individual NCDs in PRIME. Moreover, the estimates of indirect costs represent only rough approximations of their potential scale, as they are derived from the indirect cost fractions of total costs found in the related literature. For tobacco and alcohol consumption, studies specifically for Japan are available, reflecting country-specific characteristics [17, 18], whereas for the dietary factors studies from other countries had to be used. To estimate the impact of salt intake, a study from Brazil provided a useful framework [19], while for fruit and vegetable intake, fat and fiber, we referred to a study on the cost of obesity and overweight for Korea [20]. Both studies cannot capture country-specific aspects, however they both use a cost-of-illness approach, connecting to our methodology of linking risk factors to disease incidences. Furthermore, the similarity in risk factors considered in these studies results in a significant overlap in the types of diseases analyzed to capture the total cost. In addition, our total cost estimates for the baseline scenario may be overestimated due to the challenge of disentangling and precisely matching disease incidences to a specific consumption good, particularly given the complex nature of interaction effects and estimation methods in PRIME.
Another important point is the nuanced impact of low amounts of alcohol on health, particularly in the context of NCDs such as diabetes and stroke. In the model, relative risks for diabetes and partially for stroke are less than one, indicating that alcohol consumption is associated with a lower risk of these diseases than no consumption. Consequently, this results in negative avertable incidences for alcohol consumption in Japan in 2019, since low amounts of alcohol are deemed healthy, and the considered counterfactual scenario represents less alcohol consumption than the baseline scenario. These resulting negative avertable incidences also lead to negative financial health savings. The scientific literature further complicates the narrative by presenting a dual-faced impact of alcohol consumption on the risk of developing NCDs. While certain studies advocate for the protective effect of moderate alcohol consumption against diseases like type 2 diabetes and stroke, others caution against the harmful effects of excessive intake [39]. A particularly notable finding is the variability in alcohol's impact among the Asian population, attributed to genetic factors that affect alcohol metabolism. The presence of specific genetic variants, particularly involving the ALDH2 and ADH1B genes, may alter the risk dynamics for NCDs within this demographic. This genetic predisposition suggests a differential effect of alcohol that could elevate the risk for certain diseases, underscoring the necessity for a tailored approach in public health recommendations and the critical need for further research to delineate alcohol's health implications more clearly [42, 43]. Given these insights, the alcohol-related results of the PRIME Model warrant cautious interpretation. The genetic underpinnings, particularly those affecting populations differently, as seen in the Asian context, highlight the complexity of formulating universal guidelines on alcohol consumption.
In summary, while PRIME serves as a critical tool for quantifying the impact of lifestyle changes on NCD mortality, offering invaluable insights for evidence-based public health strategies worldwide, its limitations underscore the need for cautious interpretation of its outputs. These include its design focus on retrospective analysis, reliance on fixed relative risks, the methodological oversight of reduced tobacco consumption or switching to less harmful alternatives, and the challenges posed by incomplete data and the model's specific assumptions about diet. Acknowledging these constraints is crucial for leveraging PRIME effectively within the broader context of public health policy and prevention strategies.