Table 1 Summary of how each intervention was modelled and evidence of effect from literature

Broad strategy 1: A research-based strategy that generates evidence that would be used to promote consumption of indigenous foods found to have high nutritional value

We searched PubMed and Scopus electronic databases with an aim to identify evidence of the effect of consumption of healthy indigenous foods on either energy intake, body weight or body mass index (SF Table 3). Details of the search process and results are provided in the SF. A total of eight studies were included for full text review [46,47,48,49,50,51,52,53]. Apart from two studies [52, 53], the rest did not provide empirical evidence on the effect size of increased consumption of healthy indigenous foods on either energy intake, body weight or body mass index (SF Table 4). The two studies reported findings from patients who had overweight or obesity. In one study, patients were randomized into 2 groups: group A (n = 142) followed a basic traditional Chinese diet (BTCD), and group B (n = 142) followed a Western standard diet (WSD). After 6 weeks of treatment in the BCTD group, BMI decreased by 0.46 kg/m², versus 0.28 kg/m² in the WSD [53]. In an earlier trail, patients with overweight were divided into two groups: group A undergoing a 1200-kcal basic traditional Chinese diet (BTCD), and group B undergoing a 1200-kcal standard western diet. On 6 weeks after treatment, patients in group A lost more weight (0.37 +/- 0.52) kg than group B (0.26 +/- 0.79) kg [52].

The evidence from the two studies was not used as it reported on patients who had overweight or obesity, rather than the whole population.

We built two intervention scenarios based on data on food consumption levels in Sub-Sahara Africa [54] and Kenyan literature on supermarket food purchase and its effect on BMI [40, 55]. More information on this is provided in the SF.

For ease in reporting results, we refer to these two intervention scenarios as ‘interventions’.

First intervention scenario: Change in consumption levels related to supermarket food purchase

We modelled an intervention scenario where we assumed that if people did not purchase foods from supermarket, this would increase consumption of indigenous foods and their BMI would change by -1.8 (SE 0.24) kg/m² [40]. In the cross sectional Kenyan study, the authors found that 53% of their sampled population lived in households that purchased food in supermarkets [40]. We considered the identified studies relevant since supermarket purchase is seen to contribute to a shift from indigenous foods to processed foods and drinks, often provided in larger packaging sizes and accompanied by promotional campaigns [40, 41]. We considered that promotion of consumption of indigenous foods would see people purchase more of these foods instead of the processed foods found in the supermarket.

Effect size= -1.8 (SE 0.24) kg/m² [40].

In absence of data from other sources on number of people in Kenya purchasing foods from supermarkets, we used the 53% estimate from the primary study [40] to scale effect size to reflect the target population for this intervention.

Notably, in Kenya, supermarket chains initially set up in the big cities are now expanding to the smaller towns as evidenced by the 3 towns included in this study [40].This has led to urbanization of rural areas and the expansion of the peri-urban territories [56].

Modelled uncertainty distribution: normal

The modelled effect size change in BMI is presented in SF Table 2.

We model the intervention effect as starting after 5 years (i.e., from year 6 as illustrated in Fig. 1) and increasing linearly to achieve a full effect from the 15th year.

Second intervention scenario: Change in national consumption levels back to the 1975 average levels of intake

A scenario where the government funded research on indigenous foods, findings resulted in promotion of the production and consumption of healthy indigenous foods leading to a change in actual food consumption for the population (population reverts to the 1975 levels of food consumption of 2,079 kcal/person/day [54]). The 1975 consumption would be expected to reflect a higher consumption of indigenous foods.

This is an idealistic hypothetical scenario with estimated timelines.

Model assumptions are explained further in the SF and SF Tables 5 and 6 provide additional data that inform this scenario. This assumption is supported by the recommended energy intake requirements for healthy adults (2,300 kcal/day for men and 1,900 kcal/day for women) [57]. However, since the data available does not report sex specific consumption levels [54], we modelled the reported average consumption levels as applying to both male and female.

A return to the 1975 consumption levels in our scenario is on the assumption that only individuals who have overweight and obesity would reduce intake hence not a return to high undernutrition rates (SF Table 6).

In the model, the new consumption levels (1975 levels) are compared to the 2015 levels (taken to represent estimated 2019 levels) [54].

1975 levels: 2,079 kcal/person/day

2015 levels: 2,360 kcal/person/day

Effect size = -281 kcal/person/day.

The 281 kcal/person/day reduction in consumption, translates to a shift in the BMI distribution for the population. This is about 1,176 kJ/day, which equates to an average body weight change of 12.5 kg based on Sacks et. al. [43] who estimate that a change in net energy intake of 94 kJ per day (95% CI 88.2 to 99.8) corresponds to a 1 kg change in body weight [21].

The modelled effect size change in BMI is presented in SF Table 2. The resulting BMI change is sex and age specific due to the average height measures by age and sex for the Kenya population (SF Table 1).

As we did for the previous scenario, we model the intervention effect as starting from year 6 (Fig. 1) and increasing linearly to achieve a full effect from the 15th year.

Broad strategy 2: Interventions that Kenya could implement towards creation of healthy food environment

Intervention: Tax on sugar-sweetened beverages (SSBs)

The specification of the tax in this study is similar to that modelled for the Australian population by Veerman et al [59]. Definitions for SSB, ‘own-price’ and ‘cross-price’ elasticities are provided in the SF.

The price elasticity data were based on 2015 updated values of a systematic review and meta-analysis of studies in the Unites States of America, Mexico, Brazil, and France [60]. We did not find evidence on cross price elasticity from Kenya or similar setting. However using available evidence, in our sensitivity analysis, we assessed the impact that ‘cross-price elasticities’ (SF Table 7) would have on the outcomes [60].

We assess a 20% valoric tax on SSBs. In the model, the tax leads to higher prices of SSBs and a decrease in the purchases (via price elasticity) and consumption of SSBs and thus a lower total energy consumption. This translates into a reduction in BMI across the Kenyan adult population.

We used data from Euromonitor International and Global Dietary Database to derive estimates of current dietary intake data for Kenya (baseline data before tax) (SF Table 8) [61, 62]. The costs per unit price of beverages was sourced from Euromonitor International. SF Table 9 provides the baseline consumption levels with trend applied for 15 years and levels after the intervention in kJ/day/person.

Changes in quantity purchased were assumed to lead to changes in what was consumed, with no compensatory changes in physical activity levels.

To estimate how changes in price resulting from the tax would lead to changes in food purchases in the Kenyan adult population, we used global estimates of the ‘own-price’ elasticity for soft drinks (mean − 1.198, 95% CI -1.340 to -1.057) [60]. In the model, we applied a normal distribution to the price elasticity values for uncertainty.

We determined estimate changes in mean daily energy intake for each age and sex group based on the estimated changes in purchases and the average energy density of relevant products.

The effect size in sex, age specific change in BMI is presented in SF Table 2.

Intervention effect starts from 1st year of implementation.

Intervention: Mandatory kilojoule labelling on food served in formal sector restaurants

We provide a background description of this intervention, and the literature search process in the SF.

We identified the latest evidence as a meta-analysis where authors explored the effect of mandatory calorie exposure on both the retailers (41 studies) and consumers (186 studies) [58].

We model an effect of kilojoule labelling on consumer consumption as being 27.21 fewer calories per meal [58].

Effect size = − 27.21 fewer calories per meal.

We adjusted the intervention effect modelled to reflect current national estimate of two meals per week consumed outside home [10]. Further, we applied a scaling effect of 0.5 in our model to account for the proportion of food eaten in the formal sector restaurants in Kenya where enforcement is more feasible. Another scaling effect of 0.5 was applied as the evidence of effect is from the US where this regulation is already in place. The propensity of Kenya consumers to select food based on kilojoule information and attitudes towards reading kilojoule labels on menus may differ from that seen in the United States.

The effect size in sex, age specific change in BMI is presented in SF Table 2.

Intervention effect starts from 1st year of implementation.