Cardiovascular disease (CVD) is the leading cause of morbidity and mortality in industrialised countries, causing high health care utilisation costs. Elevated systolic blood pressure (SBP), total cholesterol (TC) and smoking are the best known major risk factors of CVD (Vartiainen et al., 2010). SBP (>115 mmHg) is the single most important cause responsible for >60% of strokes and 50% of coronary heart disease events, whereas corresponding figures for elevated TC (>3.8 mmol/l) are around 20% and >50%, respectively (World Health Report, 2002).

The population-wide reduction in dietary salt and replacement of dietary saturated fat (SFA) with polyunsaturated fat (PUFA) are essential in reducing CVD incidence. Their importance is mediated at least in part by the effect of salt on blood pressure (BP) and dietary fat quality on total and low-density lipoprotein cholesterol. The risk of CVD events increases progressively with increasing BP and cholesterol. The major portion of CVD deaths attributable to BP and cholesterol occurs in the upper range of ‘normal’ BP and TC levels (that is, BP >130/80 mmHg and TC >3.8 mmol/l; He and MacGregor, 2009).

In Finland, around 75% of the observed decline in coronary mortality in middle-aged men can be explained by decreases in BP, TC and smoking (Vartiainen et al., 2010). Even though public health interventions aimed at these risk factors have dramatically decreased CVD disability and deaths in developed countries (He and MacGregor, 2009), CVD is still the most common cause of death, and stroke and acute myocardial infarction (MI) are major causes of disability. Despite marked changes in diet since the 1970s, average dietary intake of salt (8.4 g per day and 6.1 g per day for men and women, respectively) and SFA (12.7 E% per day for both sexes) are still above recommended levels (<5 g per day and <10 E%), with relatively low PUFA intake (World Health Organisation, 2007; Paturi et al., 2008). Therefore, modest reduction of dietary salt and improvement of dietary fat quality, with even small shifts in the population distributions of SBP and TC towards lower values, may improve the health economic consequences of CVD substantially.

No detailed analyses on the economic consequences of reducing dietary salt and improving dietary fat quality intake have been carried out in the adult Finnish population. This study aims to predict the potential health economic consequences of moderate reduction of dietary salt combined with replacing some dietary SFA by PUFA. Owing to the data limitations, the analysis was restricted to Finns aged 30–74 years.

Subjects and methods

The year 2007 was selected as the baseline year for the CVD risk factors. A previously developed state transition Markov cohort model (Figure 1) for coronary heart disease (Martikainen et al., 2007; Peura et al., 2008) was enhanced by adding stroke and population components (Mar et al., 2008). The new model was applied to estimate the incidence and prevalence of CVD events and mortality, impact on quality of life and associated costs in the adult Finnish population. First, the most current available national risk factor survey data (Peltonen et al., 2008) and CVD risk functions (Vartiainen et al., 2007) were incorporated into the Markov model. Second, an additional model component (Mar et al., 2008) was built into the Markov model to estimate the total number of CVD patients in the Finnish population in 2007, and to make predictions for the years 2010–2030. Population estimates for adults aged 30 in year 2007 and population projections for years 2010–2030 were obtained from the official Finnish statistics. To predict the effects of risk-reducing interventions, expected numbers of events were extrapolated for the years 2010–2030 using the corresponding annual extrapolated populations from Finnish statistics. All analyses were implemented in Microsoft Excel.

Figure 1
figure 1

Simplified structure of state transition model. The model included five health states and death, which was modelled as an absorbing terminal state. The ‘healthy’ state includes subjects without a history of myocardial infarction (MI) or stroke, and they have an annual probability of experiencing an initial non-fatal CVD event (MI or stroke), fatal event (non-CVD death or CVD death) or no CVD event.

Event probabilities

For each 1-year cycle, the hypothetical cohort of men or women without established CVD were at risk of having of an acute non-fatal CVD event, a fatal CVD event, a fatal non-CVD event, or they might survive to the next year without suffering any event. The annual probabilities for CVD events were determined by logistic risk functions based on the FINRISK study (Vartiainen et al., 2007). After each cycle, CVD risk factors were updated based on the Finnish age- and sex-specific risk factor profile data from the FINRISK 2007 survey (Peltonen et al., 2008). As the FINRISK risk functions are developed to predict total CVD risk, information from the National Cardiovascular Disease Register (Laatikainen et al., 2004) was used to determine the age- and gender-based case fatality. Annual risks of non-CVD deaths were estimated from the causes of death register by subtracting fraction of deaths due to CVDs from the total mortality. A more detailed description of the methods used to estimate the event probabilities is given in appendix.

The model was run until all subjects in the cohort entered into terminal states or until 75 years of age was reached. Because the FINRISK 2007 survey (Peltonen et al., 2008) and its ancillary study FINDIET 2007 survey (Paturi et al., 2008) did not include information about the levels of CVD risk factors for people aged 75 years and older, the target population was restricted to Finnish adults aged 30–74 to ensure the reliability of the model outcomes.

Health care resource use and costs

Costs, estimated from a societal perspective, included direct cost of prevention, morbidity, rehabilitation and production losses due to non-fatal CVD events. All costs were adjusted to the 2009 price level using the official health care price index determined by the Association of Finnish Local and Regional authorities. All costs were also predicted as future financial streams over time to reflect the actual amount that the reduction of SFA and salt intake can be expected to save in a given year (Mauskopf et al., 2007; Marshall et al., 2008).

Annual age- and sex-specific costs (including costs of hospitalisation, outpatient care and prescribed outpatient medicines) associated with CVD events were obtained from reports based on the national discharge and prescription registries (Häkkinen et al., 2007; Meretoja et al., 2010). The model assumes that, after hospital discharge, subjects received a prescription for chronic coronary heart disease, dyslipidaemia or hypertension. Average drug treatment costs were weighted by the proportion of subjects receiving reimbursement for those medicines.

Costs of CVD rehabilitation were estimated by weighting the average rehabilitation costs obtained from the national rehabilitation registry at the Social Insurance Institution of Finland by the proportion (30–40%) of CVD patients receiving rehabilitation. Productivity losses due to non-fatal CVD events were estimated by multiplying the average length of sickness leave due to MI (4.78 months; Vehviläinen et al., 2004) and stroke (8.11 months; Vohlonen et al., 2002) by the average monthly income in Finland including payroll tax. Productivity losses were applied to patients below the average age of retirement, assumed to be 65 years. All unit cost estimates are summarised in Table 1.

Table 1 Parameters used in the modela

Quality-adjusted life years

Age- and sex-specific quality of life estimates measured by generic EQ-5D instrument were used represent the average quality of life in the Finnish population aged 30–75 (Saarni et al., 2006). The disutility weights due to initial or subsequent CVD events were determined from previously published studies (Saarni et al., 2006; Lindgren et al., 2007; Soini et al., 2010).

Estimating the effectiveness of reduced salt and SFA intake on the number of CVD events and other outcomes

The reduction of 1 g of salt per day was assumed to change SBP approximately by −1.185 mm Hg among hypertensive subjects and by −0.595 mm Hg among normotensive subjects. These changes were estimated based on a dose–response relationship between the changes in dietary salt intake, 24-h urinary sodium and BP (He and MacGregor, 2003). These effect estimates were weighted by the age- and sex-specific proportions of hypertensive subjects (SBP >130) in the Finnish population (Peltonen et al., 2008). The effect of salt reduction on SBP was assumed to be linear over the range of 0–3 g per day (He and MacGregor, 2003).

Effects on serum TC of replacing SFA with PUFA were obtained from a meta-analysis of metabolic ward studies (Clarke et al., 1997). Every 1% increase in energy (E%) from PUFA intake was expected to lead to a −0.026 mmol/l change in TC and 0.005 mmol/l increase in high-density lipoprotein. Alternatively, the predicted changes in serum cholesterol were estimated by the Keys equation (Keys et al., 1957), where 1 E% increase from PUFA was expected lead to around −0.043 mmol/l reduction in TC. The effect of changing BP and cholesterol levels on CVD events and deaths was then estimated using the FINRISK risk functions (Vartiainen et al., 2007). To simplify the analysis, the full effect of dietary salt and SFA reduction was assumed to be achieved immediately and to persist during the whole time horizon.

Finally, the proportion of subjects receiving antihypertensive medication was assumed to change as a function of SBP. On the basis of an estimated logistic regression model, the reduction of 1 mmHg in SBP reduced the probability of receiving antihypertensive medication with adjustment for age, sex and body mass index, by 1.5% in men and 2.8% in women (data on file).

Model calibration and sensitivity analyses

The model was calibrated to reproduce the number of key cardiovascular events among the 30–74 years old population in year 2007 in Finland. In addition, uncertainty associated with the model parameters was handled by determining probability distributions for all relevant parameters and then conducting probabilistic sensitivity analysis with 1000 iteration rounds (Briggs et al., 2002). The selection of proper probability distributions for the model parameters has been described in details in technical appendix. The results of the probabilistic sensitivity analysis were used to determine 95% credibility intervals for the model outcomes and the probability of savings as a function of amount of savings. One-way sensitivity analyses were used to test the robustness of model assumptions.


The model was calibrated by comparing the model results to the actual number of patient in the different health states in the year 2007. For example, according to the FINRISK 2007 study, the prevalence of people with history of stroke and MI is between 59 900–68 800 and 53 800–63 300 individuals among the Finnish population aged 30–74 years old, respectively (Peltonen et al., 2008). The calibrated model estimated that the number of people with history of stroke is around 61 100 and the number of people with history of MI is 58 300. Thus, on average, the calibrated model was considered to be sufficiently accurate to reproduce the baseline population for the year 2007.

Over 20 years, a population-wide intervention could potentially prevent 8000–13 000 CVD cases among the Finnish adults aged 30–74 compared with the status quo scenario, where the CVD risk factors stay at the year 2007 levels over time. In addition, the reduced incidence of CVD cases could yield an additional 26 000–45 000 quality-adjusted life years and save 150–225 million Euros over the same time period. Detailed results of base case and additional one-way sensitivity analyses are presented in Table 2. Owing to the linearity assumption of effects, the approximations of outcomes with different reduction levels are easily extrapolated based on the results presented in Table 2. For example, reducing the daily intake of both dietary salt and SFA by 0.5 U would lead to around 74–113 million Euros savings during the next 20 years.

Table 2 Predicted cumulative health and economic changes compared the situation in year 2007 between the years 2010 and 2030

To take the uncertainty of model parameters into account, the probability of obtained cumulative savings was estimated. Figure 2 shows that in the base case scenario there is >95% chance that the cumulative total savings with and without productivity losses are at least 130 and 115 million Euros between the years 2010 and 2030, respectively. In addition, sensitivity analyses showed that the independent effect of salt reduction in the prevention of cardiovascular events is larger than the effect of replacing SFA by PUFA (Table 2). The reduction of dietary salt is explained by >99% of changes in the incidence of CVD events and their consequences. However, when the Keys equation was used to predict the changes in TC, the reduction of dietary salt explained only around 62% the changes in the number of CVD cases. Thus, the interpretation of results needs to be carried out with care, as the results are sensitive to the selection of effect size estimates.

Figure 2
figure 2

Probability of cumulative savings (with and without productivity losses) given a dietary salt and saturated fat reduction of 1 g and 1 E% per day (base case scenario) among Finnish adults aged 30 to 74 between years 2010and 2030. PL=productivity losses included.


Main findings of the study

This study demonstrated that even a modest reduction of dietary salt and replacement of dietary SFA with PUFA would have a great impact on the incidence of CVD events among the population aged 30–74 years. The lower number of CVD events would result in substantial gains in quality-adjusted life years and cost savings.

The results of this study agree with previous studies (Selmer et al., 2000; Murray et al., 2003; Bibbins-Domingo et al., 2010; Smith-Spangler et al., 2010) that have demonstrated the potential benefits of salt reduction to reduce the burden of CVD. However, none of these previous studies have directly estimated the joint effects of reducing both salt and SFA intake simultaneously, considered here.

Strengths of the study

A strength of this study is that cardiovascular risk factors, dietary habits and nutrient intake of the Finnish population aged 30–74 years were obtained from a well-characterised and representative sample of Finnish adults (Paturi et al., 2008; Peltonen et al., 2008), and the cost and quality of life estimates were obtained from representative studies.

In the analysis, the independent role of dietary salt reduction of 1 g was larger than the role of a 1 E% change of SFA to PUFA. This is because both the effect sizes obtained from the literature differ and also the β-coefficients of FINRISK functions for the association of SBP with both CVD events and deaths are more significant than for the corresponding association of serum TC with CVD cases. This is in part because serum TC is not included in the FINRISK function that predicts the risk of stroke. In contrast, SBP is a major risk factor for both stroke and MI. The analyses also assume that dietary fat quality mediates its effects on CVD risk solely via effects on lipids. However, dietary fat quality may influence CVD risk by mechanisms other than cholesterol (Erkkilä et al., 2008), and the relatively strong inverse associations of PUFA intake and the PUFA/SFA ratio with coronary heart disease or CVD risk from some prospective cohort studies (Hu et al., 1997; Laaksonen et al., 2005; Erkkilä et al., 2008) also implies that dietary fat quality impact cardiovascular risk more than would be predicted by effects on the lipid profile alone.

This study focused on replacement of SFA with PUFA. Because replacement of SFA with monounsaturated fat yields similar effects on the lipid profile (Clarke et al., 1997), results would have been similar if monounsaturated or other unsaturated fat had been used in place of PUFA.

Limitations of the study

As the results of this study are based on statistical computer modelling requiring several assumptions, the results should be interpreted cautiously. This study did not specify any particular intervention aiming to reduce the dietary salt and SFA intake from daily food. However, the modest reductions presented here could be achieved by voluntary actions of the food industry aimed at a reducing the use of salt and SFA in processed food, a major source of dietary salt and SFA (Paturi et al., 2008). There are already examples about successful strategies that have managed to reduce the intake of salt and SFA in cooperation with the food industry (Prättäla, 2003; Laatikainen et al., 2006; Food Standard Agency, 2008; Pietinen et al., 2008). These interventions, for example, have managed to reduce the intake of salt by 10–25% (Laatikainen et al., 2006; Food Standard Agency, 2008).

We assumed no changes in consumer behaviour and persistence associated with these modifications, and dramatic changes in consumer behaviour would be unlikely, given that a reduction of dietary salt, for example, by 10–15% is possible without detection (Girgis et al., 2003; Ruusunen and Puolanne, 2005). Moreover, after an adaptation period people appear to prefer food with less salt (Blais et al., 1986).

To ensure the robustness of the study results, the study population was restricted to people aged 30–74 years. However, >50% of CVD cases occurs among the population aged 75 in Finland, so an important segment of the Finnish population could not be modelled. The impact in the whole population is, thus, clearly underestimated. However, at the moment there are no epidemiological data available to inform the association between CVD risk factors and the risk of CVD events among the elderly in Finland. The reduction of dietary salt and SFA is likely to benefit also subjects aged <30 years, which will potentially decrease the future prevalence of hypertension and hyperlipidaemia in older age groups.

Moreover, there were no relevant estimates available for the proportion of patients provided assisted care or care in institutions such as nursing homes after non-fatal stroke or MI. Therefore, the cost of institutionalisation, which is especially important following stroke, has not been considered in the analysis. Also, potential changes in the number of patients with heart failure, end-stage renal disease or dementia (Kivipelto et al., 2006; Solomon et al., 2010) were not considered in the analysis due to the lack of data.

This analysis assumes that CVD risk factors stay at the same levels as they were in year 2007. However, if the declining trend in the levels of CVD risk factors seen in 1972–2007 (Vartiainen et al., 2010) will continue unabated, the burden of CVD may not increase as much as presented in this study. On the other hand, the current epidemic of obesity and type II diabetes may have a negative impact on the trends seen in CVD incidence over the past several decades (Poirier et al., 2006). Finally, the assumed linear relationships between salt and SFA intakes and SBP and TC may not hold, and there is evidence that reductions in BP could be even greater at lower levels of salt intake (Sacks et al., 2001). An underestimate of the true amount of future cost savings and improved health at the population level may thus have resulted from these modelling assumptions.


Our findings suggest that even a modest reduction of dietary salt and replacement of SFA content with PUFA in food products can substantially reduce the total burden of CVDs in the adult Finnish population, with large cost savings from the societal point of view. The aging of the Finnish population may reinforce this impact.