Designing low-cost “heart healthy bread”: optimization using linear programing and 15-country comparison

Potential ingredients considered for bread designs

Potential ingredients not included

Plausible healthy bread ingredients from a CVD perspective identified in the literature included: soy protein [40], sourdough [41], and high phenolic olive oil [42]. But in this initial work we did not include these ingredients on the grounds of potentially relatively lower consumer acceptability.

Constraints and optimization approach used for the linear programing

The simplex algorithm was used to solve these linear programing problems (see Briend et al. [23] for a detailed description of the linear programing). The problem of multiple objectives with priority levels is addressed with lexicographic goal programing, which involves a sequence of linear programing steps [43]. Most of the scenarios were modeled in Microsoft Excel 2010 (Excel Solver, Simplex method), but where there was a high level of complexity with the food combination options the R programing language was used (version 2.15.0, lpSolve package).

Nutrient and price data for the HHBs

The nutrient data for the HHBs was sourced from a New Zealand food composition database [44]. But, in some cases we had to use other sources such as a study on levels of alpha-linolenic acid covering nuts and seeds [45] and from the nutrient panel on the packaging (i.e., for the salt substitute). For the price data for HHB ingredients, the cheapest retail source for the ingredients were used, with these typically being supermarkets and online stores in New Zealand.

Country comparisons of commercial breads

For country comparisons we considered just OECD countries and we conducted internet searches for those which had online documentation of nutrient profiles on commercial breads. In particular we used the EuroFIR database [46] which had 15 freely accessible nutrient databases. Ultimately, the only OECD countries with such databases were: Canada, Czech Republic, Denmark, Finland, Greece, Iceland, Italy, the Netherlands, New Zealand, Slovakia, Sweden, Switzerland, United States and the United Kingdom. For these countries the nutrient data on the most standard pre-packaged white bread was obtained. Only six of these countries had nutrient data on any breads containing seeds or nuts (Denmark, Finland, Iceland, Netherlands, New Zealand and Switzerland). Where there were multiple such breads, the one with the highest level of PUFA was selected as a proxy for the highest level of seeds/nuts.

Bread price information was sourced from online supermarkets in each country. At first, an attempt to identifying the largest supermarket chain in each country was made. Failing this, a search of the online supermarkets in the capital city was performed. Out of the breads marketed in these online supermarkets, the lowest priced most standard white pre-packaged loaf was selected for price information. For breads with seeds/nuts, the price data was from the one with the highest seed/nut level (based on the ingredient listing). For all the prices we calculated cost per 100 g and cost per 1000 kJ in US dollars using OECD exchange rates and purchasing power parity adjustments [47]. For specific nutrient and price methods details for each country see Additional file 1.

Estimating mass production costs of HHBs

To estimate the cost of mass production of HHB loafs, publicly-available data for the cost-structure data of bread manufacture from the Canadian baking industry was used. Such data are not readily available in other countries (including New Zealand), presumably due to reasons of commercial sensitivity. This Canadian cost-structure covered: “materials and supplies”, “energy, water and vehicle fuel” and “production worker wages” [48]. We applied this to the cost of the cheapest supermarket white bread from New Zealand (NZ$1.00 or US$0.68 for a 600 g loaf) and assumed at this price it was basically “cost price” i.e., with no profit. Using recipes for a white loaf of bread [49], we then estimated how much more cheaply the ingredients for this commercial bread might typically cost the baking industry, compared to ingredient prices in the supermarket. This scalar was then applied to all the HHB ingredients and varying levels of profit margin were also considered (for the manufacturer and retailer). We assumed HHBs would be mass produced by existing bread manufacturers with existing plant and machinery and distribution networks. Further details on the methods are shown with the tabulated results.

Optimized HHB ingredient selection

The optimized HHB$1.5 loaf was characterized by having some wholemeal flour but the cost constraint resulted in only trivial levels of added seeds (Table 2, Fig. 1). The optimized HHB$3 loaf (Fig. 1) had up to the maximum weight of seeds permitted (150 g dry ingredients per loaf, 12.6 % of the final loaf weight). But linseed dominated over walnut and other seeds due to it being a low-cost source of ALA.

	HHB$1.5		HHB$3
	Quantity of dry ingredients (g)	Percentage of dry ingredients	Quantity of dry ingredients (g)	Percentage of dry ingredients
Wholemeal flour	334	48 %	464	58 %
White flour	334	48 %	155	19 %
Yeast	12	2 %	16	2 %
Normal salt	8	1 %	7	1 %
Salt substitute	8	1 %	9	1 %
Nuts and seeds
Linseeds	4	1 %	132	16 %
Walnuts	0	0 %	18	2 %
Sunflower seeds	0	0 %	1	0 %
Pumpkin seeds	0	0 %	0	0 %
Sesame seeds	0	0 %	0	0 %
Totals
All dry ingredients	700	100 %	800	100 %
All nuts and seeds	4	1 %	150	19 %

Heart health and other nutrients in the loafs

In indicative rankings, the HHB$1.5 loaf appeared to be superior to the commercial white loafs from the 15 high-income countries in three out of the eight heart health nutrient categories (Table 3, Fig. 2). That is, lower in sodium, and higher in potassium and the ratio of potassium to sodium. It ranked relatively highly in terms of lower saturated fatty acids (SFA) and a high PUFA/SFA ratio.

Nutrient/cost component	White breads (n = 15 countries)		HHB$1.5	Rank of HHB$1.5a	Breads with seeds/nuts (n = 6 countries)		HHB$3	Rank of HHB$3a
	Median	IQR			Median	IQR
Eight heart health relevant characteristics b
Alpha-linolenic acid (ALA)	0.10	0.05 – 0.17	0.08	6/8	1.19	0.64 – 1.44	2.62	1/3
Total polyunsaturated fatty acids (PUFA) (g)	0.66	0.51 – 1.33	0.52	10=/14	4.52	2.98 – 5.73	4.19	4/6
Total saturated fatty acids (SFA) (g)	0.40	0.29 – 0.70	0.14	2/13	1.25	0.89 – 1.83	0.62	5/6
Ratio PUFA/SFA	2.1	1.7 – 2.5	3.7	3/13	4.4	3.0 – 4.7	6.8	1/6
Sodium (mg)	498	457 – 528	309	1/15	464	437 – 472	231	1/6
Potassium (mg)	130	115 – 157	561	1/15	220	195 – 277	647	1/6
Ratio of potassium to sodium	0.3	0.2 – 0.3	1.8	1/15	0.5	0.4 – 0.6	2.8	1/6
Dietary fiber (g)	3.1	2.6 – 4.3	4.8	4/15	5.2	4.3 – 5.4	8.1	1/6
Other micronutrients
Calcium (mg)	97	54 – 110	23	14/15	40	37 – 43	51	2/6
Folate (mcg) (including synthetic forms in some fortified white breads)	30	27 – 41	65	3/12	52	44 – 57	83	2/6
Iodide (mcg)	3.8	1.2 – 6.0	40.0	1/8	5.3	3.1 – 16.2	29.6	2/6
Iron (mg)	1.6	1.1 – 2.5	2.2	5/14	1.7	1.4 – 2.7	2.8	2/6
Selenium (mcg)	5.8	2.6 – 14.3	3.0	8/10	5.2	4.1 – 8.0	6.4	3/6
Zinc (mg)	0.8	0.7 – 1.0	1.4	1=/13	1.4	1.1 – 1.9	1.9	3/6
Selected other macronutrients
Dietary energy (kJ)	1047	1026 – 1169	951	2/14 (if low is best)	1188	1168 – 1291	1036	2/5 (if low is best)
Protein (g)	8.4	8.3 – 9.1	7.4	14/15	10.6	9.5 – 11.6	8.8	6/6
Sugars (total) (g)	3.5	1.7 – 4.8	0.5	2/10	2.2	1.45 – 3.4	0.6	2/5
Cost (relevant to food security) c
Cost per 100 g of loaf (US$)	0.28	0.21 – 0.34	0.10	2/15	0.73	0.64 – 0.95	0.17	1/5
Cost per 1000 kJ of food energy (US$)	0.17	0.12 – 0.20	0.11	4=/14	0.34	0.33 – 0.38	0.20	2/5

^aIndicative rank in terms of the best values from a health perspective and food security perspective (some of these ranks could change with a larger sample of comparison breads). From an energy security perspective, low cost per kJ is best. But in terms of weight control, low kJ per 100 g (low energy density) is probably best. Of note is that the nutrient data in some of the national databases had some gaps e.g., for ALA and iodide (both only 8 of the 15 databases for white bread)

^bSome of these nutrient characteristics overlap to some extent (e.g., the two ratios reflect the amounts of the two relevant components)

^cHHB costing based on supermarket prices for ingredients and no other production costs. See Table 4 for more details on possible mass production costs

For specific nutrient and price methods details for each country see Supplementary Information

IQR interquartile range

The HHB$3 loaf appeared to be nutritionally superior to the commercial loafs with seeds/nuts from six countries in six out of the eight heart health categories (Table 3, Fig. 2). While it had the best PUFA/SFA ratio, it did not have the highest levels of PUFA and lowest levels of SFA. It also appeared superior to the HHB$1.5 loaf in all eight heart health categories except for low saturated fat. Some of the differences were relatively large, such as the ALA level in the HHB$3 loaf being 26 times the median for the white loafs and 2.2 times the median for the commercial high seed/nut loafs.

In terms of other nutrients, the HHB$1.5 bread appeared to perform better than most commercial white breads (e.g., highest in iodide and zinc) (Table 3). But it was competing with some breads which were fortified with calcium, folate and iron. However, relative to the commercial seed breads, the HHB$3 loaf was in the top two rankings for: calcium, folate, iodide, iron, low sugars, and low energy density.

In terms of value-per-weight, a commercial white loaf from the UK was slightly cheaper than the HHB$1.50 loaf (at US$0.07 vs US$0.10 per 100 g loaf, Table 3). The HHB$1.5 loaf also performed relatively well in terms of price per kJ of food energy (fourth equal). But compared to the other seed/nut loafs, the HHB$3 loaf was by far the best value-per-weight at US$0.17 vs a mean of US$0.85 per 100 g.

Possible mass production costs of HHB loafs

Where commercial manufacturers/retailers produced and sold HHB loafs (e.g., as per contracts for a government voucher system for provision of HHB) a profit margin would be relevant (Table 4). For example, a 25 % profit margin shared by producers and retailers would push up the price to: US$1.24 for the HHB$1.5 loaf and US$2.19 for the HHB$3 loaf (Table 4). Based on these estimates and a 25 % profit mark-up, the HHB$1.5 loaf at US$0.12 per 100 g would remain well under half that of the median for the commercial white loafs from the 15 high-income countries in Table 3. Likewise the HHB$3 loaf at this mark-up would still be the lowest cost per 100 g out of the seed/nut loafs in Table 3.

Main findings and interpretation

This proof-of-concept work generated two potential HHB designs which were generally superior to commercial breads from 15 high-income (OECD) countries from a nutrient/CVD prevention perspective (i.e., in terms of sodium, potassium, fatty acid ratios and dietary fiber). The higher levels in the HHB$3 bread were often substantial e.g., 26 times the median ALA level of the white loafs. Furthermore, in principle the higher fiber levels in the HHBs could contribute to non-CVD benefits such as colorectal cancer prevention [31]. Similarly, the walnut in the HHB$3 loaf could potentially contribute in a small way to cancer prevention [38, 39], but obviously a diet designed to minimize cancer risk has many other aspects such as being high in fruit and vegetables.

Of relevance to food security, the HHB$1.5 loaf was better value-per-weight than all but one of the commercial white loafs from the 15 countries. The HHB$3 also outperformed all the seed/nut loafs at a fifth the cost per 100 g of the average. Furthermore, the relatively higher level of dietary fiber in the HHB loafs might mean they also provide better value in terms of inducing satiety. Such a satiety benefit has been reported in a study on wholegrain vs refined bread [50], but not in one of nut-enriched bread vs nut-free bread [51].

The mass production cost estimates (Table 4) suggest it would be feasible to produce the HHB$3 loaf in a place like a hospital kitchen at around US$1.75 per loaf. The voucher costs for government for this type of loaf might be US$2.19 per loaf when assuming a 25 % mark-up by the contracted manufacturer and retailers (or US$2.63 if a 50 % mark-up is assumed). At two loafs a week per person (e.g., for a patient at high risk of CVD) this would be US$274 per year in voucher costs for a 100 % subsidized loaf supply (albeit not including the voucher delivery and other administration costs). Whether this is justified from a funder perspective would depend on many issues such as the priority given to protecting public health and the health economics (e.g., the cost-effectiveness of such a program should ideally be modeled before any implementation).

Study strengths and limitations

This study was, to our knowledge, the first to study the issue of the optimized design of bread from a heart health and a low-cost perspective using linear programing. We consider that the constraints imposed, and the typical ingredients used, will help ensure that such designs have at least some acceptability in high-income countries where there is often a majority of adults who currently consume at least a light-grained bread (e.g., as per New Zealand [52]). Indeed, some countries currently have seed/nut breads sold commercially at up to 17 % seeds by weight (for Finland, see Table 1).

Nevertheless, this was still just a proof-of-concept study which only considered two potential designs of HHB. Indeed, the full “design-universe” of such potential breads is large and might include other relatively healthy additional ingredients such as: soy, sourdough, dried fruit, high phenolic olive oil, and extra fiber sources (e.g., wheat bran and rye flour), (see the Methods ). A version particularly suitable for women of child-bearing age could also include added folic acid for preventing neural tube defects.

A further limitation was the relatively simplistic nature of our cost estimates for mass production of the two HHB designs. Nevertheless, it is plausible that the cost-structure of Canadian bread manufacture is not markedly different from other high-income countries. Our costing estimates also didn’t consider administrative costs associated with setting up and running any HHB voucher system (e.g., for people with CVD risk factors). It also did not consider the environmental impact, as per Swedish work on emissions of oxides of carbon, sulfur and nitrogen from bread production [53]. Finally, we did not explore various other complexities with bringing any HHB to market. For example, consumer taste testing would be required (see below), and packaging would need to be designed. For example, packaging would probably need to display potassium levels to help people with renal disease or on particular medication such as diuretics to avoid excessive intake.

Implications for further research and policy

Consumer testing prior to mass production of any HHB could be done by a commercial bread manufacturer as part of their routine product development processes. Similarly, an institution such as a public hospital could “market test” the acceptability of various HHBs on staff in a hospital cafeteria. The latter could be done in conjunction with health orientated non-governmental organizations which exist in many countries and sometimes have “endorsement label programs” for heart healthy foods (e.g., as in New Zealand [54]). If there were concerns about public acceptability of bread design, then a more incremental approach could involve slowly phasing down the levels of white flour and sodium while phasing up the levels of linseed and salt substitute in the HHBs promoted for mass consumption. Final bread products could also be considered in terms of the health claims that could be justifiably made on HHB packaging or in marketing about these being “functional foods”. Many countries have rules around such claims with one example being the European Food Safety Authority’s process for European Union countries [55].

Epidemiological and economic modeling should also attempt to quantify the effectiveness (in quality-adjusted life-years gained) and cost-effectiveness of a government-run voucher system for a HHB version for those at increased CVD risk (and even for those at increased risk of cancer given the extra fiber and seeds in the HHB$3 loaf).

Ethical review

This was not relevant for this modeling study which did not involve human subjects.

Consent to publish

Not applicable.

Availability of data and materials

All supporting data is in Additional file 1.