Estimating the health and economic effects of the voluntary sodium reduction targets in Brazil: microsimulation analysis

IMPACT_NCD-BR is a discrete-time stochastic, dynamic, microsimulation model based on the simulated adult life course of a close-to-reality open cohort of synthetic individuals under different policy scenarios, considering the population heterogeneity and the lag times between exposures and outcomes. The data sources to the model are presented in Table 1, and the model inputs, structure, and key assumptions are detailed in Supplementary Materials Appendix.

The model is designed first to run the no intervention scenario, simulating the life courses of the synthetic individuals and recording sodium consumption, systolic blood pressure, first cardiovascular episode, and death (from CVD or any other cause). Then, it simulates the life courses of the same synthetic population under the intervention scenario (in this case, food reformulation through the national voluntary sodium targets) and records its outcomes.

The close-to-reality population built in the model mirrors the Brazilian national population structure by age and sex [18]. It includes data from the National Health Survey (Pesquisa Nacional de Saúde—PNS 2013) [22] and the National Household Budget Survey (Pesquisa de Orçamentos Familiares—POF 2008–2009) [20, 21] regarding sodium and systolic blood pressure (SBP) exposure.

Figure 1 shows a simplified structure of the IMPACT_NCD-BR model. The model tracks individual-level sodium consumption, considering the different sources of dietary sodium, its impact on SBP, and the consequent risk of developing CHD, stroke, and death from these or any other cause. IMPACT_NCD-BR is calibrated to forecasts of CHD, stroke, and any-other-cause mortality for the whole Brazilian population from 2013 to 2032. The results are presented for adults aged 30 to 79 years within the simulation horizon of 20 years.

For this, the model draws the traits of the synthetic individuals from conditional distributions, projects the sodium intake into the future, and uses the projections to evolve the traits of the synthetic individuals over time. We used PNS 2013 for the SBP projections (assuming SBP remains constant over time for all age and sex groups) and POF 2008–2009 for the sodium intake projections [20‐22].

Excess dietary sodium consumption has been linked to an increased risk of CVD [28]. For CVD, the excess risk appears to be mainly mediated through the deleterious effect of excess sodium consumption on blood pressure (BP) [3]. Our methods for evaluating the causality of sodium reduction on BP and BP reduction on CVD have been previously described [3].

Brazil’s Ministry of Health has set voluntary reduction targets with food industries, from 2011 to 2017, based on upper bound sodium concentration targets for 34 food categories [11, 29]. In addition, publicly available data from the Household Budget Survey (POF 2008–2009) were used to map these 34 food categories through household food acquisition [20] and a 24-h recall dietary questionnaire [21]. These data enabled the model to estimate the potential impact of the modelled policies on every synthetic individual based on their age, sex, and sodium consumption in the no intervention baseline scenario. The model then used the estimated reduction in sodium consumption of the synthetic individuals to calculate the effect upon their SBP using a published meta-regression equation [3].

We assumed that reformulation of food products would adjust sodium content to targets until 2017 and that this would lead to an immediate change in sodium intake in synthetic individuals according to the level of reformulation. We also assumed that the reformulated products would, after that, sustain their sodium concentration.

Although changes in sodium intake influence SBP within weeks [3, 30], we conservatively assumed a median duration of 5 years from a change in SBP to the health outcomes of CVD cases and CVD deaths.

We have considered changes in food composition from the voluntary sodium targets using data from official national food labelling surveys in 2013 and 2017: the baseline targets and the most recent documented official monitoring, respectively [12]. In addition, changes in sodium intake were modelled using microdata from the 2002–2003 and 2008–2009 Brazilian Household Budget Survey. We assumed that the average food consumption and use of table salt in the population remained constant from 2011 to 2017 and that sodium content was reduced for the targeted food categories only by industries that have voluntarily committed to the national sodium targets (which correspond to a 70% market share in the country).

We used the publicly available data from the Household Budget Surveys (POF) of 2002–2003 [31] and 2008–2009 [20] in order to estimate the changes in sources of dietary sodium (non-industrial and industrial sources) during the period between both surveys. We projected the continuity of the replacement of discretionary salt (table salt) by processed and ultra-processed foods (representing the other sodium sources), assuming that the replacement would continue at the same rate in the future.

The model generates the total numbers of relevant events and reported case-years (CHD and stroke) and deaths prevented or postponed (CHD or stroke or other) for each scenario. We present the results for Brazilian adults aged 30 to 79 years from 2011 to 2032 (simulation horizon of 20 years), rounded to 2 significant digits. For simplicity, we used the term “cases prevented or postponed” to express case-years prevented or postponed.

The CHD and stroke hospitalisation costs to the Brazilian National Health System (SUS—Sistema Único de Saúde) were obtained from the publicly available tables from the Hospital Information System—SIH-SUS [26]. There were no readily available Brazil-specific data for the informal care costs; therefore, we used estimates for these costs from a study on European Union (EU) countries [27]. Cost savings to the health system and the population were estimated considering the CHD and stroke cases prevented or postponed and the average costs for a person-year living with these diseases. Costs were collected in Brazilian Reals (R$) and subsequently converted to US dollars (US$), at an exchange rate of US$1 = R$3.876, current as of December 31, 2018, as reported by the Central Bank of Brazil, and all costs and were discounted at a 3% annual rate.

We performed probabilistic sensitivity analyses using a second-order Monte Carlo approach for estimating the uncertainty of different model parameters and population heterogeneity to be propagated to the outputs [32]. Uncertainty was based on sources of the sampling errors of baseline sodium intake, baseline SBP, and the relative risk of CHD and stroke based on SBP, the uncertainties around the lowest sodium and SBP exposures below which no excess risk is observed, around the effect of sodium on SBP, around the lag between SBP changes and CVD risk changes, around the true incidence of CHD and stroke, and the uncertainty of mortality forecasts. The output distributions produced through the Monte Carlo approach were summarised by reporting the medians and 95% uncertainty intervals (UIs) using the 2.5% and 97.5% percentiles of the output distribution.

The modelled results suggest that Brazil’s recent voluntary sodium reduction policies have led to a 0.1 g/day of sodium decline in sodium consumption (0.25 g/day salt) from 2011 to 2017 (Fig. 2). If that 0.1g/day sodium reduction continues, it could reduce median SBP from 127.7 mmHg (95% UI: 127.6 to 127.8) in the baseline scenario to 127.5 mmHg (95% UI: 127.4 to 127.6) in the reformulation scenario by 2032. Furthermore, it could prevent or postpone approximately 2500 CVD deaths (95% UI: −900 to 11,000) and some 180,000 CVD cases (95% UI: 45,000 to 430,000), as well as 12,000 deaths from other hypertension-related causes (95% UI: 3600 to 22,000), from 2013 to 2032 (Table 2).

The health benefits of sodium reduction, especially in reducing CHD, were larger in men than women, reflecting men’s higher sodium consumption and higher CHD and stroke burden. In total, during the 20-year time period from 2013 to 2032, the voluntary sodium targets could prevent or postpone approximately 110,000 CVD cases (95% UI: 28,000 to 260,000) in men and some 70,000 cases in women (95% UI: 16,000 to 170,000), and approximately 1400 fewer CVD deaths in men (95% UI: −500 to 4100) and 1200 fewer deaths in women (−500 to 4800).

Almost two-thirds of the CVD cases and CVD deaths that would be prevented or postponed by this policy between 2013 and 2032 would be in people aged 50 to 69 years (Appendix).

From the public healthcare perspective, the voluntary targets for sodium could result in a net saving of approximately US$ 290 million (95% UI: US$ 71 to 690 million) in cumulative hospitalisation, primary health, outpatient, pharmaceutical, and informal care costs, over the 20 years. Most of the savings would be related to CHD (~75%) rather than stroke.

The US$ 290 million estimated savings through the continuity of the voluntary sodium targets would come from (Table 3):

The IMPACT_NCD microsimulation framework has been extensively validated and replicated in different countries [13, 15]. This is the first IMPACT family microsimulation model adapted to a Latin American country and represents a big step forward for using models to inform policy in the region. The study’s strengths also include the use of representative population data and suitable effect sizes from meta-analyses.

This study also has potential limitations. First, the model’s effect estimates are based on interventional and prospective observational studies and may retain possible biases and confounding factors. However, the etiological effects of dietary changes were estimated from meta-analyses with confirmatory validity analyses, including from randomised clinical trials. Second, the estimated benefits from the model may be conservative and underestimate the full health and economic health gains from sodium reformulation, as (1) the counterfactual scenario assumed that the participation of processed and ultra-processed foods in sodium intake would not change into the future; (2) the model only evaluated diseases mediated through BP, while decreasing sodium consumption could have beneficial effects upon other health burdens such as gastric cancer [42]; (3) it was assumed a median duration of 5 years from changes in SBP to CVD risk; (4) only the first cardiovascular episode was considered. Therefore, only primary prevention benefits were quantified; and (5) food reformulation by industries might additionally increase potassium intake through substitution of NaCl with KCl [43], which potential beneficial effect was not included in the model. Third, informal costs were based on evidence from European countries and, therefore, might not represent Brazil’s informal care costs. Finally, medical costs from the private sector, which covers 30% of the Brazilian population, were not included in the economic estimates, which will therefore be conservative.