An updated estimate of benzoate intakes from non-alcoholic beverages in Canada and the United States

ABSTRACT

In 2017, the results of a comprehensive assessment of intake for benzoic acid and its salts from non-alcoholic beverages were published for four regions (Brazil, Canada, Mexico, and the United States [U.S.]). These regions were among those identified as having the most prevalent use of benzoates in beverages globally. The results of the 2017 study did not indicate a safety concern relative to the acceptable daily intake (ADI) established for benzoates (5 mg kg body weight^-1 day⁻¹, as benzoic acid), and supported maintaining the Codex maximum benzoate level in water-based beverages (250 mg kg⁻¹). Since this time, population-specific food consumption data have been released for public use for Canada, and updated beverage consumption data have become available for the U.S. To ensure estimated intakes remain relevant, these consumption data were incorporated with previously collected brand-specific benzoate use level and market volume data for beverages. Dietary exposure to benzoates from non-alcoholic beverages was assessed using statistical modelling, either probabilistic (non-brand loyal; considering the full distribution of use levels) or deterministic (brand loyal; assuming all regular carbonated soft drinks, the brand loyal beverage type, contain benzoates at the maximum use level, and all other beverage types in which benzoates are used contain benzoates at the market-weighted average use level). In both models, estimated daily intakes at the mean and 95^th percentile were below the ADI (≤76% of the ADI) in all Canadian and U.S. population groups with a statistically reliable population size. The findings from updated Canadian and U.S. consumption data continue to support the Codex maximum benzoate level in water-based flavoured drinks at 250 mg kg⁻¹.

KEYWORDS:

• Benzoate
• benzoic acid
• exposure assessment
• non-alcoholic beverages
• water-based flavoured drinks
• additive
• CCHS-Nutrition
• NHANES
• ADI

Introduction

Benzoates (benzoic acid and its salts, INS 210–213) are a class of preservatives used in a range of foods and beverages as noted in the Codex General Standard for Food Additives (GSFA) (CAC 2019). They are most effective in acidic foods and beverages , such as soft drinks (WHO 2000). Several scientific bodies have conducted toxicological evaluations of benzoic acid and its salts and have established a group acceptable daily intake (ADI) of 0 to 5 mg kg body weight^-1 day⁻¹, expressed as benzoic acid equivalents (JECFA 1974a, 1974b, 1997; SCF 1996, 2002; EFSA 2016). The ADI is based on the highest dose evaluated in a multi-generational reproductive rat study (Kieckebusch and Lang 1960). Two scientific investigations have since suggested that the default toxicokinetic uncertainty factor for interspecies differences could be reduced, which would result in a higher ADI for benzoates (Hoffman and Hanneman 2017; Zu et al. 2017). These and other recently completed safety studies on benzoic acid will soon be considered by the Joint FAO/WHO Expert Committee on Food Additives (JECFA) in its re-evaluation of benzoates Summer 2021 (JECFA 2020).

In 2015, the JECFA identified water-based flavoured drinks (GSFA category 14.1.4) as the primary source of dietary exposure to benzoates across a range of countries based on data submitted by industry and the literature (JECFA 2015). JECFA conducted an exposure assessment of benzoates from non-alcoholic ‘soft’ beverages using summary statistics of consumption (Chronic Individual Food Consumption Data – Summary statistics [CIFOCOss] or summary consumption figures from data submitted by industry) combined with average typical use levels. The use of summary statistics rather than individual-based data is noted as a potential source of overestimation of this assessment. There were some exceedances of the ADI at the 95th percentile (97.5th percentile for South Africa) in younger subpopulation groups among consumers only. Published concentrations of benzoates measured analytically were 63 to 259 mg L⁻¹ in non-alcoholic beverages, while industry reported averages were 83 to 209 mg L⁻¹ in water-based flavoured drinks. Consequently, the Codex Committee on Food Additives (CCFA) established a maximum level of 250 mg kg⁻¹ for water-based flavoured drinks (GSFA category 14.1.4), similar to the above mentioned levels from JECFA’s assessment (CCFA 2016; CAC 2019).

Martyn et al. (2017) conducted a more refined dietary exposure assessment of benzoates from water-based flavoured drinks (GSFA category 14.1.4) in four jurisdictions (Brazil, Canada, Mexico, and the United States [U.S.]) using individual-based consumption data in combination with brand-specific use level data (as benzoic acid) and market volume data. These four regions were selected as they were identified as having the most prevalent use of benzoates in beverages globally by virtue of higher maximum permitted regulatory limits and higher consumption of water-based beverages (Singh et al. 2015; Popkin and Hawke 2016; Beverage Digest & Zenith International 2020). Dietary exposure was modelled using probabilistic (non-brand loyal) and deterministic (brand loyal) approaches. Food consumption data were obtained from national dietary surveys, namely the Inquérito Nacional de Alimentação (INA) of the 2008–09 Pesquisa de Orcamenos Familiares (POF – Brazil), the 2012 Encuesta Nacional de Salud y Nutrición (ENSANUT – Mexico), and the 2011–2012 U.S. National Health and Nutrition Examination Survey (NHANES); the latter was also used as a surrogate for Canada.

In the 2017 assessments, benzoate consumer-only exposure estimates were significantly lower than the group ADI for benzoates at the mean (14 to 40% of the ADI) in all four jurisdictions (Martyn et al. 2017). Likewise, estimated intakes were generally below the ADI at the 95^th percentile (36 to 110% of the ADI). The ADI was only slightly exceeded in Canadian and Mexican ‘toddlers and young children’ at the 95^th percentile in the brand loyal scenario, in which regular carbonated soft drinks (RegCSDs) were identified as the brand loyal beverage type simulated to contain the Codex maximum limit (RegCSDs with pH ≤ 3.5; 250 mg kg⁻¹) or the maximum industry-reported use level (RegCSDs with pH > 3.5; 438 mg kg⁻¹ for Canada, 299 mg kg⁻¹ for Mexico). A higher use level is generally needed under less acidic conditions to achieve the desired technological function. Given the conservative nature of the models and that the ADI relates to lifetime use, Martyn et al. (2017) concluded that the results rendered do not indicate a safety concern in these high benzoate intake jurisdictions (considered a proxy for global estimates), supporting the maintenance of the Codex maximum benzoate use level of 250 mg kg⁻¹ in GSFA category 14.1.4, and the use of a higher benzoate use level (up to 438 mg kg⁻¹) in beverages with pH > 3.5.

Since the 2017 publication, the 2015 Canadian Community Health Survey (CCHS) – Nutrition has been made publicly available, and food consumption data from the 2015–2016 cycle of the U.S. NHANES have been released. In view of these updates, it is important to ensure the dietary exposure assessments for these jurisdictions are kept up-to-date and relevant.

Materials and methods

Food consumption data

Canada

The public use microdata file (PUMF) for the 2015 CCHS-Nutrition was released in December of 2018 (Statistics Canada 2017a, 2018a, 2018b). This is the second cross-sectional dietary survey providing detailed information on foods and nutrient supplements consumed by Canadian individuals at national and provincial levels (Statistics Canada 2017b, 2017c).

The 2015 CCHS-Nutrition includes two components: a 24-hour dietary recall (conducted on two non-consecutive days) and a general health questionnaire. Individuals aged 1 year and over were sampled. Of the 23,584 individuals selected to participate in the survey, 20,487 individuals completed the first interview (86.9% response rate), and a subsample of 7,608 individuals completed the second interview (Statistics Canada 2017b).

The 24-hour dietary recall component used a variation of the Automated Multiple-Pass Method (AMPM) developed by the United States Department of Agriculture (USDA), adapted to reflect foods available in Canada, differences in food preparation in Canada, and ethnic foods consumed in Canada. The method consisted of five steps – (1) Quick list, (2) Forgotten foods, (3) Time and occasion, (4) Detail cycle, and (5) Final review – to keep respondents interested and engaged. The amount of each food that the respondent reported consuming was standardised and converted to a gram amount. Initial sizing was performed on items reported using standard quantitative sizes (millilitres, grams, ounces, etc.); this was followed by the sizing of items reported using non-standard qualitative sizes (bowls, glasses, mugs, etc.) using Food Model Booklet standard sizes. Ingredients in recipes were sized according to Health Canada’s stock recipe ingredient amounts using USDA standards, where available; volumetric sizes were converted to grams using Health Canada’s reference file of food densities.

The Health component consisted of 19 ‘modules’, which included information on nutrient supplement consumption, socio-economic status, anthropometric measurements, and other health parameters. Anthropometric measurements were taken for individuals over 2 years of age and were measured by the interviewer or were self-reported.

For the purposes of the current analysis, data for individuals with two days of consumption were utilised to allow a comparison of chronic intakes to the ADI established for benzoates. To increase the reliability of intake estimates expressed on a body weight basis, only individuals with measured body weights were included in the assessment. Therefore, 5,601 individuals 2 years of age and above meeting the above criteria were included in the assessment using appropriate sample weights.

United States

Food consumption data from the 2015–2016 release of the U.S. NHANES were made publicly available in July of 2018 (CDC 2018a, 2020a; USDA 2019). NHANES are conducted as continuous, annual surveys, and are released in 2-year cycles. For the 2015–2016 NHANES, 15,327 individuals were selected for the sample, 9,971 individuals were interviewed (61.3%), and 9,544 individuals were examined (58.7%). Dietary survey data for NHANES are collected from individuals and households via 24-hour dietary recalls administered on two non-consecutive days – Day 1 (in-person) and Day 2 (by telephone) – throughout all four seasons of the year.

In addition to collecting information on the types and quantities of foods being consumed, the 2015–2016 NHANES collected socio-economic, physiological, and demographic information from individual participants in the survey, such as sex, age, body weight, and other variables (such as height and race-ethnicity) that may be useful in characterising consumption. The inclusion of this information allows for further assessment of food consumption by specific subpopulation groups of interest within the total population. The primary sample design for the 2015–2016 NHANES included an oversample of the same subgroups as the 2011–2012 NHANES (used in Martyn et al. 2017), though the threshold for non-Hispanic ‘white or other’ low income persons was changed from ≤ 130 to ≤ 185% according to the updated Department of Health and Human Services poverty guidelines (CDC 2020b). Sample weights were incorporated according to analytic guidelines for the 2015–2016 NHANES (CDC 2018b) to allow estimates from these subgroups to be combined to obtain national estimates that reflect the relative proportions of these subgroups in the population as a whole (CDC 2018a, 2020a; USDA 2019).

Use level data

Focusing specifically on Canada, benzoates are Class 2 preservatives permitted for use in unstandardised foods at a maximum use level of 0.1% (expressed as benzoic acid), which includes unstandardised non-alcoholic water-based beverages (Health Canada 2020). In the U.S., benzoates are Generally Recognised as Safe (GRAS) for use at levels not exceeding 0.1% (expressed as benzoic acid) in accordance with Good Manufacturing Practice (USFDA 2019). To ensure consistent food categorisation across Canada and the U.S., the food categorisation system established under the Codex GSFA was utilised to organise food codes and data on the use of benzoates in non-alcoholic water-based beverages in both regions. Although the broader GSFA category corresponds to non-alcoholic (‘soft’) beverages (GSFA category 14.1) and its subcategories, particular focus was given to GSFA category 14.1.4 ‘water-based flavoured drinks, including “sport”, “energy”, or “electrolyte” drinks and particulated drinks’ (CAC 2019). As previously noted, benzoate-containing beverages were limited to those in the 14.1.4 category only (Martyn et al. 2017). This category was further stratified according to conventional beverage types as defined by industry, and aligns with beverage types identified to contain benzoates in Martyn et al. (2017), namely:

• Regular carbonated soft drinks (RegCSDs).

• Low- and no-calorie sweetened carbonated soft drinks (CSDs).

• Flavoured water.

• Juice drinks (i.e., less than 100% juice; previously denoted as ‘Fruit-juice based drinks’ in Martyn et al. 2017).

• Energy drinks.

• Sports drinks.

• Ready-to-drink (RTD) tea.

• RTD coffee.

Canadian- and U.S.-specific market-weighted average use levels and maximum use levels for benzoates (expressed on a benzoic acid basis) for the above beverage types, summarised in Table 1, were previously determined in Martyn et al. (2017) using brand-specific use level data and market-volume share information. Briefly, brand-specific use data were provided by industry in 2016 (complemented by other market data sources), specifying (1) levels of addition of benzoates to non-alcoholic beverages and (2) information on use patterns (i.e., presence or absence of benzoates) in each beverage type and brand identified. The pH of beverages at the brand level was also annotated. Volume data (in millions of gallons) for top 2015 contributing brands within each beverage type were used to calculate market share, and were purchased from the Beverage Marketing Corporation (BMC 2020) and Global Data PLC (previously Canadean Ltd) (Global Data PLC 2020).

Table 1. Percent benzoate-containing beverages, weighted average use level of benzoates (as benzoic acid), and highest reported use level of benzoates (as benzoic acid) in Canada and the United States by beverage type within the GSFA 14.1.4 ‘Water-based flavoured drinks’ category (Martyn et al. 2017), as utilised in the exposure assessment models

GSFA 14.1.4 ‘Water-based flavoured drinks’ category, beverage types^a	Canada			United States
	% benzoate-containing beverages^b	Benzoate use level (mg kg^-1)		% benzoate-containing beverages^b	Benzoate use level (mg kg^-1)
		Ave.^c	Max.		Ave.^c	Max.
Low- and no-calorie sweetened CSDs	97.2	174	690^d	100.0	197	400^d
Regular CSDs	38.3	79	438^d	50.0	115	428^d
Flavoured water drinks	0.0	0	0	2.3	4	197
Juice drinks^e	0.0	0	0	0.0	0	0
Energy drinks	64.3	109	429	65.1	111	431
Sports drinks	0.0	0	0	0.0	0	0
RTD teas	61.0	56	91	44.3	67	196^d
RTD coffees	0.0	0	0	0.0	0	0

Ave., average; CSDs, carbonated soft drinks; GSFA, Codex General Standard for Food Additives; Max., maximum; RTD, ready-to-drink

^aBeverage types within the Codex GSFA 14.1.4 ‘Water-based flavoured drinks’ category were determined based on conventional beverage products as defined by industry.

^bPercent of benzoate-containing beverages in corresponding beverage types was calculated based on presence of benzoates as noted by participating beverage manufacturers and/or label verifications. These values have been weighted according to the market volume share of the available brands in the specific jurisdiction.

^cMarket volume-weighted average use level of benzoates in beverage types.

^dUse level reported for product with pH > 3.5.

^eIncluding concentrates, products with less than 100% juice; previously ‘Fruit-juice based drinks’ (Martyn et al. 2017).

Exposure assessment methodology

Food code selection and categorisation

All food codes that correlated with water-based flavoured drinks beverage types were identified in the 2015 CCHS-Nutrition and the 2015–2016 U.S. NHANES. Food codes for the 2015 CCHS-Nutrition analysis were selected from the Food and Ingredients Details (FID) file, capturing consumption at the basic food level and at the ingredient level in recipes. Food codes selected for the 2015–2016 NHANES analysis are the same as those selected in the previous 2011–2012 NHANES analysis (Martyn et al. 2017), accounting for any new or updated food codes. All food codes selected for benzoate-containing beverages were representative of ready-to-drink beverage formats; therefore, no dilution or reconstitution factors were applied in the current assessment.

Exposure assessment models

Applying the same approach as Martyn et al. (2017), two exposure assessment models were utilised:

• Model 1: Non-brand loyal probabilistic: A distribution of available use levels was established using market-share volume of individual brands. This distribution was applied to each food code within the pre-assigned beverage type (see Table 1). In practical terms, the chance that any reported use level was encountered by a consumer was directly related to the market-volume share of the corresponding brand within the specific beverage type grouping.

• Model 2: Brand loyal distributional: ‘Brand loyal’ consumers are assumed to habitually consume a beverage type containing benzoates at the maximum use level, as well as consuming other beverage types containing benzoates at market-weighted average use levels. The ‘brand loyal’ beverage type was determined to be RegCSDs, as these were the top contributing beverage type to total mean daily intakes both in Martyn et al. (2017) and in the current assessment. The Codex maximum use level of 250 mg kg⁻¹ (as benzoic acid) was applied broadly to RegCSDs with pH < 3.5, and the highest industry-reported use level was applied to RegCSDs with pH > 3.5 (i.e., 438 mg kg⁻¹ for Canada and 428 mg kg⁻¹ for the U.S.; see Table 1). The market-weighted average use level was applied to all other beverage types (see Table 1).

Data analysis

Statistical analyses and data management were conducted using DaDiet software (Dazult Ltd. 2018). As mentioned above, consumption data were available in both datasets at the level of the individual consumption event over the two recall days.

The same age groups as examined in Martyn et al. (2017) were also investigated here, which match those reported in the 2015 JECFA exposure assessment, namely ‘toddlers and young children’, ages 1 to 7 years; ‘other children, including adolescents’, ages 8 to 17 years; ‘adults’ ages 18 years and over’; and the total population (all age and gender groups combined). As noted above, because measured body weights in the 2015 CCHS-Nutrition were available only for respondents 2 years of age and above, 2 years of age was the youngest age to be considered in this dataset.

The percent of consuming individuals was determined for each beverage type that falls within the GSFA 14.1.4 ‘Water-based flavoured drinks’ category by subpopulation group in Canada and the U.S. (see Table 2). These estimates represent the proportion of the population that consumed one or more of the specified beverage types during the recall period.

Table 2. Percent consumers^a of individual water-based flavoured drink (GSFA category 14.1.4) beverage types by subpopulation group in Canada (2015 CCHS-Nutrition) and the United States (2015–2016 NHANES)

Subpopulation group	n	Percent consumers (%)^b
		Low- and no-calorie sweetened CSDs	Regular CSDs	Flavoured water drinks	Juice drinks^c	Energy drinks	Sports drinks	RTD teas	RTD coffees	Any water-based flavoured drink
Canada, 2015 CCHS-Nutrition (Statistics Canada 2018a)
Toddlers and young children (2–7 yr^d)	618	1.0	7.6	0.3	22.1	0	1.0	1.7	n.a.	29.7
Other children and adolescents (8–17 yr)	1,707	5.0	30.6	2.4	27.0	0.6	4.2	8.0	n.a.	58.5
Adults (18+ yr)	3,276	10.2	22.4	1.3	9.6	1.1	1.1	3.3	n.a.	41.5
Total population (2+ yr^d)	5,601	8.5	22.7	1.4	13.8	0.9	1.7	4.0	n.a.	43.6
United States, 2015–2016 NHANES (CDC 2018a)
Toddlers and young children (1–7 yr)	977	1.7	20.5	21.2	28.9	<0.1	5.4	7.7	0.5	59.6
Other children and adolescents (8–17 yr)	1,319	2.3	49.3	19.9	25.7	0.7	11.5	16.6	1.8	82.4
Adults (18+ yr)	4,363	15.5	35.7	10.0	11.1	2.8	5.5	23.1	3.4	70.9
Total population (all ages, including infants)	6,968	12.3	35.8	12.2	14.6	2.2	6.3	20.6	2.9	70.6

CSDs, carbonated soft drinks; GSFA, Codex General Standard for Food Additives; n, number of subjects (unweighted); n.a., no food code available; RTD, ready-to-drink; yr, years

^aConsumers were identified as individuals reporting consumption of the specified beverage type on any of the two dietary survey recall days.

^bWeighted values.

^cIncluding concentrates, products with less than 100% juice; previously ‘Fruit-juice based drinks’ (Martyn et al. 2017).

^dBody weight data only available for individuals ≥ 2 years of age.

As per the approach taken previously (Martyn et al. 2017), the probabilistic assessments were conducted iteratively using a Monte Carlo simulation, with a target of more than 100,000 person days per survey (Boon and Van Klaveren 2003; Boon et al. 2004; EFSA 2012). This model combined the full range of industry-reported use levels, weighted according to market volume data for the use of benzoates in each beverage type, with daily consumption values for the corresponding beverage type. For example, benzoate use levels of up to 690 mg kg⁻¹ were used for low- and no-calorie sweetened CSDs in the Canadian assessment. A full distribution of intakes was calculated from the simulation (100,818 person days for Canada and 112,432 person days for the U.S.), from which mean and high percentile estimates were determined.

As described above, RegCSDs were identified as the brand loyal beverage type in the brand loyal model, to which the Codex maximum use level (i.e., 250 mg kg⁻¹) was applied for RegCSDs with pH ≤ 3.5 and the maximum industry reported use level (438 mg kg⁻¹ in Canada and 428 mg kg⁻¹ in the U.S.) was applied for RegCSDs with pH > 3.5. For the remaining beverage types containing benzoates, the average weighted use level was multiplied by the daily individual consumption level. Estimates for the daily intake of benzoates represent projected two-day averages for each individual from each of the dietary survey recall days. Mean and 95^th percentile intake estimates were determined from this distribution.

For both jurisdictions, dietary exposure to benzoates was calculated for all relevant subpopulation groups using both models, considering only those beverage types reported to contain benzoates. The latter allowed the determination of the highest potential exposure in consumers of benzoates. Results are presented on a body weight basis (mg kg body weight^-1 day⁻¹) and as a percentage of the ADI on a benzoic acid basis (% ADI) for the general population and for benzoate consumers (Tables 3 and 4). ‘General population’ values represent intakes for all survey respondents irrespective of whether they consumed benzoate-containing water-based flavoured drinks in either model. ‘Benzoate consumer’ values represent calculated mean and high-level exposure by only those individuals who consumed a benzoate-containing beverage during the survey recall period. In the probabilistic model, a ‘consumer’ represents any individual who was simulated to consume a beverage containing benzoates during the iterations (i.e., intake > 0 mg kg body weight^-1 day⁻¹). For the subpopulation groups evaluated, 5.4 to 25.1% of Canadian individuals and 18.0 to 49.2% of U.S. individuals were consequently identified as benzoate consumers in the probabilistic model (Table 3). For the brand loyal model, consumers of beverage types in which benzoates were reported to be used according to the jurisdiction (see Table 1) were identified as ‘benzoate consumers’, i.e., 100% presence of benzoates was assumed for each beverage type reported to contain benzoates. For the subpopulation groups evaluated, 10.0 to 39.5% of Canadian individuals and 41.6 to 68.1% of U.S. individuals were consequently identified as benzoate consumers in the brand loyal model (see Table 4). The percent consumers is lower in the probabilistic model compared to the brand loyal model in both jurisdictions, likely because the probabilistic model used the full range of use levels including 0 mg kg⁻¹ (notably, brands identified as having the highest market volume, and consequently a greater weight in the probabilistic assessment, were benzoate free), and RegCSDs (the brand loyal beverage type) would have a lower proportion of products containing benzoates (38.3 to 50.0%; Table 1) in the probabilistic model.

Table 3. Estimated daily exposure to benzoates from water-based flavoured drinks (limited to only those reported to contain benzoates) in Canada (2015 CCHS-Nutrition) and in the United States (2015–2016 NHANES) by subpopulation group for the general population and for benzoate consumers only, and comparison to the ADI of benzoates (5 mg kg bw^-1 day-1 as benzoic acid) – non-brand loyal probabilistic model

Subpopulation group	n	General population exposure					Benzoate consumer only exposure^a
		mg kg bw^-1 day^-1			%ADI		mg kg bw^-1 day^-1						%ADI
		Mean	SD	P95	Mean	P95	% cons	Mean	SD		P95		Mean		P95
Canada, 2015 CCHS-Nutrition (Statistics Canada 2018a)
Toddlers and young children (2–7 yr^b)	618	0.048	0.268	0.201	1	4	5.4	0.880		0.770	2.169			18	43
Other children and adolescents (8–17 yr)	1,707	0.174	0.447	0.977	3	20	25.1	0.693		0.662	2.178			14	44
Adults (18+ yr)	3,276	0.153	0.412	0.891	3	18	23.7	0.645		0.632	1.844			13	37
Total population (2+ yr^b)	5,601	0.148	0.410	0.882	3	18	22.4	0.660		0.642	1.898			13	38
United States, 2015–2016 NHANES (CDC 2018a)
Toddlers and young children (1–7 yr)	977	0.217	0.655	1.476	4	30	18.0	1.206		1.090	3.110			24	62
Other children and adolescents (8–17 yr)	1,319	0.410	0.749	1.933	8	39	40.5	1.011		0.880	2.742			20	55
Adults (18+ yr)	4,363	0.506	0.883	2.199	10	44	49.2	1.030		1.023	2.920			21	58
Total population (all ages, including infants)	6,968	0.461	0.848	2.103	9	42	44.6	1.034		1.009	2.902			21	58

%cons, percent consumers (weighted) within specified age cohort; ADI, acceptable daily intake; bw, body weight; CCHS, Canadian Community Health Survey; GSFA, Codex General Standard for Food Additives; n, number of subjects (unweighted); NHANES, United States National Health and Nutrition Examination Survey; P95, 95^th percentile; SD, standard deviation; yr, years

^a‘Benzoate consumers’ were identified as individuals who consumed any water-based flavoured drink (GSFA category 14.1.4) during the two dietary survey recall days which were simulated to contain benzoates during the iterations.

^bBody weight data only available for individuals ≥ 2 years of age.

Table 4. Estimated daily exposure to benzoates from water-based flavoured drinks (limited only to those reported to contain benzoates) in Canada (2015 CCHS-Nutrition) and in the United States (2015–2016 NHANES) by subpopulation group for the general population and benzoate consumers only, and comparison to the ADI of benzoates (5 mg kg bw^-1 day-1 as benzoic acid) – brand loyal distributional model

Subpopulation group	n	General population exposure					Benzoate consumer only exposure^a
		mg kg bw^-1 day^-1			%ADI		%cons	mg kg bw^-1 day^-1			%ADI
		Mean	SD	P95	Mean	P95		Mean	SD	P95	Mean	P95
Canada, 2015 CCHS-Nutrition (Statistics Canada 2018a)
Toddlers and young children (2–7 yr^b)	618	0.149	0.709	0.893*	3	18*	10.0	1.489	1.740	6.108*	30	122*
Other children and adolescents (8–17 yr)	1,707	0.482	0.906	2.364	10	47	39.5	1.220	1.085	2.928	24	59
Adults (18+ yr)	3,276	0.307	0.674	1.585	6	32	34.1	0.901	0.894	2.600	18	52
Total population (2+ yr^b)	5,601	0.326	0.730	1.672	7	33	33.1	0.985	0.980	2.699	20	54
United States, 2015–2016 NHANES (CDC 2018a)
Toddlers and young children (1–7 yr)	977	0.436	1.019	2.707	9	54	41.6	1.050	1.363	3.693	21	74
Other children and adolescents (8–17 yr)	1,319	0.833	1.185	3.187	17	64	68.1	1.222	1.260	3.779	24	76
Adults (18+ yr)	4,363	0.765	1.140	2.958	15	59	64.6	1.185	1.231	3.500	24	70
Total population (all ages, including infants)	6,968	0.736	1.136	2.942	15	59	62.3	1.182	1.243	3.586	24	72

%cons, percent consumers (weighted) within specified age cohort; ADI, acceptable daily intake; bw, body weight; CCHS, Canadian Community Health Survey; GSFA, Codex general standard for food additives; n, number of subjects (unweighted); NHANES, United States National Health and Nutrition Examination Survey; P95, 95^th percentile; RegCSDs, regular carbonated soft drinks; SD, standard deviation; yr, years

^a‘Benzoate consumers’ were identified as individuals who consumed any water-based flavoured drink (GSFA category 14.1.4) containing benzoates during the two dietary survey recall days. The benzoate use level applied was the market volume-weighted average for each respective beverage type, except for RegCSDs. For RegCSDs the Codex maximum level (250 mg kg^-1, as benzoic acid) was applied to RegCSDs with pH ≤ 3.5 and the highest reported use level (438 mg kg^-1 for Canada and 428 mg kg^-1 for the United States, as benzoic acid) was applied for RegCSDs with pH > 3.5.

^bBody weight data only available for individuals over 2 years of age.

*Intake estimates (i.e., 0.893 or 6.108 mg kg bw^-1 day^-1 at the 95^th percentile for Canadian ‘toddlers and young children’ in the general population or who are benzoate consumers only, respectively, and 18 or 122% of the ADI for same) are not be statistically reliable as the sample size does not meet the minimum reporting requirements (mean n < 30; 95^th percentile n < 160).

Dietary exposure to benzoates was also calculated considering all GSFA category 14.1.4 water-based flavoured drinks and all GSFA category 14.1 non-alcoholic (‘soft’) beverages (excluding natural and bottled water, yet including juices, nectars, and hot coffee and tea beverages in addition to water-based flavoured drinks) with results presented in Tables S1 and S2, respectively. In these scenarios, consumers were defined as individuals who reported consuming any beverage from GSFA category 14.1.4 or GSFA category 14.1 (excluding ‘water only’ consumers) over the course of the dietary survey period, regardless of benzoate use (i.e., ‘beverage consumers’ only). For subpopulation groups evaluated, 29.7 to 58.5% of Canadian individuals and 59.6 to 82.4% of U.S. individuals were identified as consumers of any beverage from GSFA category 14.1.4 in the probabilistic and brand loyal models, whereas 78.0 to 97.8% of Canadian individuals and 82.6 to 96.4% of U.S. individuals were identified as consumers of any beverage from GSFA category 14.1 (excluding ‘water only’ consumers) in the probabilistic and brand loyal models.

Results

Benzoates in water-based flavoured drinks

The percent of benzoate-containing beverages within each beverage type of water-based flavoured drinks (GSFA category 14.1.4) is provided in Table 1 (replicated from Martyn et al. 2017). As there were no reported benzoate-containing beverages for GSFA categories 14.1.2, 14.1.3, and 14.1.5, beverage types within these categories (e.g., fruit juices, fruit nectars, vegetable nectars, and hot coffees and teas) were excluded from Table 1.

For beverage types in GSFA category 14.1.4, the use pattern of benzoates was similar in both jurisdictions. Low- and no-calorie sweetened CSDs was the beverage type with the highest proportion of products containing benzoates (97.2 to 100%), followed by energy drinks (64.3 to 65.1%), RTD teas (44.3 to 61.0%) and RegCSDs (38.3 to 50.0%). Reported benzoate use in the flavoured water beverage type (e.g., vitamin water) was minimal in the U.S. market (2.3%) and non-existent in Canada. Additionally, there was no reported use of benzoates in the following beverage types (from GSFA category 14.1.4) in either jurisdiction: juice drinks (including concentrates), sports drinks, and RTD coffees.

Consumption of water-based flavoured drinks

Table 2 presents the percent of each cohort who reported consumption of at least one of the GSFA category 14.1.4 ‘water-based flavoured drinks’ beverage types during the dietary survey period. The Canadian population was identified as having a smaller proportion of consumers of any type of water-based flavoured drink (29.7 to 58.5%) compared to the U.S. population (59.6 to 82.4%). This trend was also noted for individual beverage types such as RegCSDs (7.6 to 30.6% versus 20.5 to 49.3%, respectively), flavoured water drinks (0.3 to 2.4% versus 10.0 to 21.2%, respectively), sports drinks (1.0 to 4.2% versus 5.4 to 11.5%, respectively), and RTD teas (1.7 to 8.0% versus 7.7 to 23.1%, respectively). Details of the findings for each country are presented in Table 2. A comparison to the previous analysis (Martyn et al. 2017) is discussed in detail below.

Canada

The Canadian population in the 2015 CCHS-Nutrition was identified as having a smaller proportion of consumers of any type of water-based flavoured drink (29.7 to 58.5%) compared to the U.S. population in the 2011–2012 NHANES, the latter previously used as a surrogate dataset for this cohort (69.0 to 83.8%; Martyn et al. 2017). The same trend was noted when examining individual beverage types within the broader GSFA 14.1.4. category. The percent consumers of low- and no-calorie sweetened CSDs was relatively low in the Canadian population using the 2015 CCHS-Nutrition, ranging from 1.0% ('toddlers and young children') to 10.2% ('adults'). RegCSDs were identified as the beverage type with the highest percent consumers among the total Canadian population ages 2 years and older (22.7%), with ‘other children and adolescents’ reaching 30.6%. The percent consumers of juice drinks (including concentrates) was 2- to 3-times higher in ‘toddlers and young children’ (22.1%) and ‘other children and adolescents’ (27.0%), respectively, compared to ‘adults’ (9.6%). A relatively small proportion of the Canadian population was identified as consumers of flavoured water drinks (0.3 to 2.4%), energy drinks (0 to 1.1%), and sports drinks (1.0 to 4.2%). The percent consumers of RTD teas was 8.0% in ‘other children and adolescents’, but relatively low in all other population groups (1.7 to 3.3%). No food codes for RTD coffees were available from the 2015 CCHS-Nutrition; as a result, this beverage type could not be assessed in the Canadian population (note: there was no benzoate use reported in this beverage type for the Canadian market either; see Table 1).

United States

The reported consumers of any type of water-based flavoured drink decreased to 59.6% in the 2015–2016 NHANES release compared to 69.0% in the 2011–2012 NHANES release among ‘toddlers and young children’. Percent consumers among ‘other children and adolescents’ (82.4% compared to 83.8%, respectively) and ‘adults’ (70.9% compared to 68.5%, respectively) however were equivalent. For individual beverage types, the percent consumers in the 2015–2016 NHANES release were generally lower than values in the 2011–2012 NHANES release, with increases noted only for RTD teas and RTD coffee beverage types.

The percent consumers of low- and no-calorie sweetened CSDs in the 2015–2016 NHANES ranged from 1.7% ('toddlers and young children') to 15.5% ('adults'), approximately half those in the 2011–2012 NHANES (from 3.8 to 21.2%; Martyn et al. 2017). Similarly, comparing the 2015–2016 NHANES to the 2011–2012 NHANES, the percent consumers of RegCSDs was lower for younger age groups up to 17 years of age (20.5 to 49.3% compared to 33.7 to 55.9%, respectively). The values were however similar for ‘adults’ (35.7% compared to 39.6%, respectively). Also, the percent consumers of juice drinks (11.1 to 28.9% compared to 13.8 to 37.1%, respectively) and sports drinks (5.4 to 11.5% compared to 7.6 to 15.2%, respectively) were lower amongst all age groups. The percent consumers for flavoured water drinks (~10.0 to 22.4%) were similar for both survey cycles amongst all age groups. Although the percent consumers of energy drinks were also similar amongst individuals under 17 years of age (<1.0%), there was an increase among ‘adults’, at 2.8% compared to 0.9%. The percent consumers of RTD teas increased, with a range of 7.7 to 23.1% in the 2015–2016 NHANES compared to a range of <1.0% previously. While there were no RTD coffee consumers identified in the 2011–2012 NHANES (due to absence of food codes), 0.5 to 3.4% RTD coffee consumers are reported in the more recent 2015–2016 NHANES dataset (food codes representing RTD coffees were first introduced in the 2013–2014 NHANES cycle release).

Dietary exposure to benzoates

Probabilistic model

Benzoate consumers probabilistic modelling intake estimates are presented in Table 3. Probabilistic modelling of benzoate intake estimates for consumers of, more broadly, any beverage type from GSFA category 14.1 (excluding water) and from GSFA category 14.1.4 (i.e., these would include beverage types with no reported benzoate use in addition to those with reported benzoate use) can be found in Table S1 of the supplementary material.

Canada

In the non-brand loyal scenario, mean exposure to benzoates was notably lower than the ADI (5 mg kg body weight^-1 day⁻¹) in the general population (1 to 3% of the ADI) and in the benzoate consuming population (13 to 18% of the ADI) (Table 3). At the 95^th percentile, intakes were still well below the ADI in the general population (4 to 20% of the ADI) and in the benzoate consuming population (37 to 44% of the ADI), even though the full range of use levels were utilised in this model (up to 690 mg kg⁻¹). The updated exposure estimates for the Canadian population (derived using Canada-specific food consumption data) are slightly lower than estimates previously reported for Canada that leveraged food consumption data from the 2011–2012 U.S. NHANES as a surrogate. For example, the benzoate consuming population at the 95^th percentile in the previous assessment had estimated intakes ranging from 45 to 70% of the ADI (Martyn et al. 2017).

United States

Benzoate intakes by the general population ranged from 4 to 10% of the ADI at the mean and 30 to 44% of the ADI at the 95^th percentile (Table 3). When focusing on only those individuals simulated to consume benzoate-containing beverages, intakes were higher, with 20 to 24% of the ADI for the mean and 55 to 62% of the ADI for the 95^th percentile. These 2015–2016 NHANES estimates however are lower than the 2011–2012 NHANES estimates among benzoate consumers, which were at 21 to 28% of the ADI for the mean and 56 to 79% of the ADI for the 95^th percentile (Martyn et al. 2017). For the 95^th percentile benzoate only consumers for the total population (all ages) in the non-brand loyal model, the slight decrease in estimated intakes observed between then and now was found to be due to a reduction in the consumption volume of low- and no-calorie sweetened CSDs (by 299 mL day⁻¹) and flavoured water drinks (by 153 mL day⁻¹).

Brand loyal model

Benzoate consumers brand loyal modelling intake estimates are presented in Table 4. Brand loyal modelling of benzoate intake estimates for consumers of, more broadly, any beverage type from GSFA category 14.1 (excluding water) and from GSFA category 14.1.4 (i.e., these would include beverage types with no reported benzoate use in addition to those with reported benzoate use) can be found in Table S2 of the supplementary material.

Canada

Mean exposure to benzoates was higher in the brand loyal scenario compared to the probabilistic model, at 3 to 10% of the ADI for the general population and at 18 to 30% of the ADI among benzoate consumers (Table 4). Intakes remained below the ADI at the 95^th percentile in the general population (32 to 47% of the ADI) and among benzoate consumers (52 to 59% of the ADI) who had a statistically reliable sample size. This pattern is generally similar to what was reported in Martyn et al. (2017) when 2011–2012 NHANES consumption data were used to model benzoate intakes in the Canadian population, with the exception of the exceedance of the ADI previously reported in benzoate-consuming ‘toddlers and young children’ at the 95^th percentile (104% of the ADI). In the current assessment, the 95^th percentile intake estimate in brand loyal benzoate-consuming Canadian ‘toddlers and young children’ was derived using the 2015 CCHS-Nutrition and was considered not statistically reliable – i.e., less than 160 individuals from this population group were identified as benzoate consumers (n [weighted] = 77 consumers) (CDC 2013). It was determined that the exceedance was caused by four individuals in this population group who consumed a particularly high volume of RegCSDs (the brand loyal beverage type), of up to 673 mL day⁻¹, equivalent to a total amount of up to 1.3 L over the two survey days. As noted in the methods, the brand loyal model assumes 100% presence of benzoates at the Codex maximum level (250 mg kg⁻¹) in RegCSDs with pH ≤ 3.5 and at the maximum industry reported use level (438 mg kg⁻¹) in RegCSDs with pH > 3.5. The combination of (1) the high consumption volumes for the brand loyal RegCSD beverage type in these four individuals, (2) the application of maximum use levels to the brand loyal RegCSDs beverage type, and (3) the assumption that 100% of RegCSDs contain benzoates, resulted in the 95^th percentile estimated intake among ‘toddlers and young children’ exceeding the ADI. Nevertheless, it is not possible to make conclusions for Canadian high-level benzoate consumers from this subpopulation group in the absence of a sufficiently large consumer base to produce a statistically reliable estimate at the 95^th percentile.

United States

Benzoate exposure by the general population ranged from 9 to 17% of the ADI at the mean and 54 to 64% of the ADI at the 95^th percentile (Table 4). When focusing on only benzoate consumers, exposure levels were expectedly higher, at 21 to 24% of the ADI at the mean and 70 to 76% of the ADI at the 95^th percentile. When compared to the results from the 2017 publication, the estimated intakes for benzoate consumers from the 2015-2016 NHANES are notably lower than the 2011–2012 NHANES estimates, previously reported at 27 to 31% of the ADI at the mean and 80 to 98% of the ADI at the 95^th percentile (Martyn et al. 2017). Similar to the non-brand loyal model, the decrease in intake observed at the 95^th percentile in benzoate consumers for the total population (all ages) in the brand loyal model when using the 2015–2016 NHANES dataset was due to decreased consumption (volume wise) of low- and no-calorie sweetened CSDs and flavoured water drinks.

Discussion

The aim of this analysis was to provide updated intake estimates for exposure to benzoates in Canada and the U.S. based on more recently available consumption data for these two regions. The other two regions evaluated in the 2017 publication – Brazil and Mexico – did not have updated consumption datasets and therefore were not included here. It is notable that the U.S., Mexico, and Brazil have been identified as the highest consuming beverage markets in the world (Beverage Digest & Zenith International 2020), consistent with other reports (Singh et al. 2015, 2019; Popkin and Hawke, 2016). Therefore, updated estimated daily intakes of benzoates from beverages presented for Canada (2015 CCHS-Nutrition) and the U.S. (2015-2016 NHANES) together with results from Martyn et al. (2017) for Brazil (INA from the 2008–2009 POF) and Mexico (2012 ENSANUT) (Martyn et al. 2017), would be a ceiling of potential global intakes such that if no safety concern is noted with these datasets (particularly estimated intakes for the U.S., Mexico, and Brazil) then no safety concern should exist across any geographies worldwide.

With respect to updates to estimated intakes, although the previous analysis showed a slight exceedance of the ADI amongst Canadian ‘toddlers and young children’ at the 95^th percentile in brand loyal benzoate consumers using consumption data from the 2011–2012 NHANES as a surrogate (104% of the ADI; Martyn et al. 2017), the current analysis (in which the 2015 CCHS-Nutrition dataset was used) could not generate a statistically reliable estimate at the 95^th percentile for this age group (n [weighted] = 77 consumers; ≥ 160 individuals are required for statistical reliability at the 95^th percentile when data are normally distributed [CDC 2013]). In all other cases, across age groups and both models, mean and 95^th percentile estimates fell well below the ADI.

In reviewing recent publications for other jurisdictions (Korea and Iran), exposure estimates for benzoates (based on analytic data) from the consumption of non-alcoholic beverages tend to be generally below the ADI. Notably, the Korean publication shows how estimating intakes based on only the non-alcoholic beverage category tends to be higher than when all foods and beverages are considered together.

Shin et al. (2017) estimated daily intakes for benzoic acid from processed foods in the Korean diet using the 2012 Korea National Health and Nutrition Examination Survey (KNHANES) consumption dataset (Shin et al. 2017). Benzoic acid was measured in 33.3 to 52.9% of samples analysed across non-alcoholic beverage types (fruit and vegetable beverages, carbonated beverages, beverage base, and other beverages), with a maximum reported concentration of 473.34 mg kg⁻¹ in carbonated beverages. While 90^th percentile dietary intakes of benzoic acid for individual non-alcoholic beverage types were higher than the 90^th percentile estimated daily intake for foods in the consuming group, there was no ADI exceedance for this heavy consuming group when considering dietary intake from all processed foods (26.1% of the ADI). These results demonstrate that cumulative exposures from various food sources in addition to beverages may serve to counter-intuitively dilute overall intake estimates due to an expanded consumer-base. In recent evaluations for ‘direct addition’ ingredients other than preservatives, a strict focus on the water-based flavoured drinks beverage category in exposure modelling was shown to generate conservative intake estimates when compared to assessments that more broadly integrate foods and beverages (Tran et al. 2020a; Tran et al. 2020b [in press]). Thus, a ‘beverage’ focus to ingredient intake assessments – as was conducted here and previously (Martyn et al. 2017) – may very well be a more conservative approach in evaluating beverage additive provisions compared to the traditional ‘food and beverage’ approach to modelling exposures, although this would depend on the pattern of use in non-beverage categories.

Two other studies evaluated exposure to sodium benzoate in the Iranian population from analytical levels measured in orange juice (Akbari-adergani et al. 2018) and various food and beverage items including carbonated soft drinks (Chalshetori et al. 2018). Akbari-adergani et al. (2018) measured sodium benzoate in 9 out of 10 orange juice brands, where mean concentrations were reported to range between 12.23 ± 1.50 and 56.80 ± 2.13 mg kg⁻¹. The resulting estimated daily intake of sodium benzoate based on per capita orange juice consumption was 1.11 mg kg body weight^-1 day⁻¹ on a whole sodium benzoate basis, and thus significantly less than the ADI for benzoic acid (19% of the ADI, or ~ 0.94 mg benzoic acid kg body weight^-1 day⁻¹). Chaleshtori et al. (2018) measured sodium benzoate in all carbonated soft drink samples (n = 19), where mean and maximum reported concentrations were 61.75 ± 55.53 and 130 mg kg⁻¹, respectively. Estimated daily intakes from the consumption of carbonated soft drinks based on average and maximum concentrations, assuming a consumption of 144 mL day⁻¹ and a 60 kg body weight, are well below the ADI (after correcting for the 1,000x discrepancy in Chaleshtori et al.’s (2018) estimated intakes in which units reported in the publication were mg kg body weight^-1 day⁻¹ yet should have been expressed as µg kg body weight^-1 day⁻¹). Collectively, these studies and analyses consistently show estimated intakes for benzoates from beverages to be well below the ADI in Iran.

Non-alcoholic beverages have consistently been identified as the top contributing food category to the dietary intakes of benzoates from all foods (JECFA 2015; EFSA 2016). As such, the analyses conducted herein are considered suitably conservative. The original four markets (U.S., Brazil, Mexico and Canada – using U.S. data as a surrogate to Canadian dietary exposure) were selected based not only on higher volumes of beverages consumed (Singh et al. 2015; Popkin and Hawkes 2016; Beverage Digest & Zenith International 2020), but also based on markets with higher regulatory permissible benzoate levels in beverages and the existence of a robust set of individual consumption survey data (Martyn et al. 2017). Only U.S. and Canadian intake estimates were updated in the present analyses as both these jurisdictions had more recent consumption data available for integration (i.e., 2015 CCHS-Nutrition and 2015–2016 NHANES data instead of the 2011–2012 NHANES dataset). Updated intake estimates are more reflective of current exposure patterns keeping the analyses relevant. The use of refined exposure assessment methodology, with a focus on the beverage category (a primary contributor to dietary intakes of benzoic acid), facilitated a more precise (due to brand-specific market volume weighted use levels coupled with individual consumption datasets that are nationally representative) yet conservative estimate (especially for the brand loyal scenario in which 100% presence of benzoic acid was assumed at the maximum use level of 250 mg kg⁻¹ for the brand loyal RegCSDs beverage type with pH < 3.5, and up to 438 mg kg⁻¹ for RegCSDs with pH > 3.5) of benzoate intakes for these high consuming markets. Taken together, the latter model identifies a possible ceiling of potential intakes across geographies worldwide such that if no concern is noted with this analysis then no concern should be noted for any other geography at similar maximum use levels.

Sources of uncertainty associated with consumption data and use level data utilised in the assessments, as well as probabilistic and brand loyal modelling, are summarised in Table 5, along with their influence on dietary intake estimates reported herein. These were identified and qualified in accordance with criteria specified in Codex and EFSA (CAC 2003; EFSA 2006).

Table 5. Qualitative evaluation of the influence of uncertainties on estimates of exposure to benzoates from non-alcoholic beverages calculated in Canada and the United States^a

Source of uncertainty	Direction and magnitude^b
Consumption data
Mis-/under-reporting of beverage consumption in dietary surveys	±
Use of data from short-term surveys to estimate chronic exposures	+
Mis-categorisation of beverage types	±
Beverage consumption data not considered at brand level^c	±
Rounding errors in food consumption data	±
Use level data
Use levels were provided by members of the International Council of Beverages Associations only (representing 73 to 83% of the market), no account for use levels by other industry members	±
Reported use level data were employed as opposed to analytically measured values (accounts for current formulation)	±
Lack of benzoate usage data for smaller brands not identified to hold a significant market share	±
Rounding errors, incorrect entries for concentration data	±
Probabilistic model
No consideration for brand loyal consumers	±
Probability of benzoate use levels being encountered for each consumer was based on market share (by volume) serving as a proxy to actual consumer behaviour	±
Brand loyal model
Codex ML of 250 mg kg^-1 (or max. reported level of 438 mg kg^-1 for Canada and 428 mg kg^-1 for the United States for beverages with pH > 3.5, where possible) applied to RegCSDs, the highest contributing beverage type	+
100% presence assumed for each beverage type in which benzoates were reported to be used – irrespective of whether these reported values accounted for a large proportion of the market (or not)	+

CCHS, Canadian Community Health Survey; EFSA, European Food Safety Authority; max., maximum; ML, maximum level; NHANES, United States National Health and Nutrition Examination Survey; RegCSDs, regular carbonated soft drinks

^aConducted in accordance with the criteria specified in Codex and EFSA (CAC 2003; EFSA 2006).

^b+, ++, +++ are uncertainties likely to cause small, medium or large over-estimates of exposure, respectively, and -, – -, – - – are uncertainties likely to cause small, medium or large under-estimates of exposure, respectively.

^cIn the 2015–2016 NHANES dataset, some food codes representative of flavoured water drinks, juice drinks, energy drinks, and sports drinks were associated with a brand. However, as not all food codes were associated with a brand, and some of the other water-based flavoured drink beverage types were not associated with food codes containing any information on branding (e.g., carbonated soft drinks), consumption data at the brand-level could not be considered in the current assessment. In the 2015 CCHS-Nutrition, food code descriptions do not contain brand level information.

Limitations with these surveys are similar to those for all dietary surveys where food consumption data are collected at the level of the individual. A large source of error relates to dietary information reported by survey participants, which can lead to under- or overestimations of intake. Although there is a potential for under-reporting amounts consumed of foods perceived as less health such as sugary drinks (Briefel et al. 1997; Krebs-Smith et al. 2000; Han and Powell 2013; Statistics Canada 2018c), it is worthwhile noting high consumption volumes of RegCSDs were reported for Canadian ‘toddlers and young children’, suggesting these are not under-reported. In both surveys, food consumption data were collected using 24-hour recalls on two non-consecutive days, used to estimate ‘chronic’ dietary intakes to benzoates. However, two recall days is usually not sufficient to capture day-to-day variation for the consumption of foods (especially foods that are not routinely consumed) – a greater number of recall days provides a more accurate representation of ‘usual’ intakes (Palaniappan et al. 2003). It has been suggested by others that short-duration surveys provide an overestimate of long-term high levels of exposure in consumers (WHO 2009). Thus, mean – not the 95^th percentile – dietary exposure is typically compared with a chronic (long-term) health-based guidance value such as the ADI (WHO 2009).

Although brand-specific use level data and market volume data collected approximately five years ago (2016 and 2015, respectively) may be viewed as another limitation and source of uncertainty, these data nevertheless remain relevant to this evaluation as these were collected in the same timeframe as food consumption data available from the 2015 CCHS-Nutrition and the 2015–2016 NHANES datasets. Additionally, as previously noted in Martyn et al. (2017), use level data were primarily provided by relevant International Council of Beverages Associations (ICBA) members at the time of data collection. Use level data collected by the ICBA represent around 80% of the market share in Canada and the U.S. In cases where market volume data were available but associated use level data were unavailable for a particular branded beverage product, use levels for beverage products within a specific beverage type were re-weighted to exclude brands with ‘missing’ use level data and could be a source of either under-estimation or over-estimation. However, reported use level data capture most major brands and should be considered adequately representative of most water-based flavoured drinks in Canadian and U.S. commerce. Finally, the pairing of use level data to market share volume data at the brand level provides more accurate, precise, and robust estimates of the potential benzoate exposure from beverages in Canada and the U.S.

It is commonly accepted that the refined probabilistic model characterises the full distribution of consumer exposures by inputting all available concentration data, which then provides more realistic consumer practice simulations. A source of underestimation of this model is that it does not consider the brand loyal consumer exposure scenario. Although the European Food Safety Authority (EFSA) previously noted the existence of consumer brand loyalty to non-alcoholic beverages in their evaluation of benzoates (EFSA 2016), it should be acknowledged that the brand loyal consumer may be brand loyal to a beverage that contains low levels of benzoates (e.g., benzoate level below the Codex GSFA maximum level for RegCSDs with pH ≤ 3.5 or maximum level reported by industry for RegCSDs with pH > 3.5) or that does not contain benzoates. As noted earlier, RegCSDs (the brand loyal beverage type) contain benzoates in up to half of the marketed products, with brands with the highest market volume (i.e., carrying the most weight) being benzoate free. In fact, the market volume weighted average use level was determined to be well below the Codex GSFA maximum level of 250 mg kg⁻¹ in both jurisdictions (79 and 115 mg kg⁻¹ in Canada and the U.S., respectively – see Table 1). Thus, the brand loyal model is conservative as all RegCSDs (the brand loyal category) were assumed to contain benzoates at the Codex GSFA maximum level (250 mg kg^-1, as benzoic acid) for beverages with pH ≤ 3.5 and at the maximum industry reported use level (438 mg kg⁻¹ for Canada and 428 mg kg⁻¹ for the U.S., as benzoic acid) for beverages with a pH > 3.5 when, in reality, only 38.3% of RegCSDs contain benzoates in Canada and 50.0% in the U.S (see Table 1). Regardless, and as previously noted by Martyn et al. (2017), the models were designed using suitably conservative assumptions to avoid the risk of potential underestimation of exposure to benzoates in Canada and in the U.S., while providing more realistic yet sufficiently conservative estimates of benzoate exposure in consumers of water-based flavoured drinks in these jurisdictions.

Summary and conclusions

In summary, this analysis provides up-to-date exposure estimates of benzoates among Canadian and American consumers of water-based flavoured drinks, which are complementary to estimates reported in Martyn et al. (2017) in these populations and in other regions identified as having the most prevalent use of benzoates in beverages globally (Brazil and Mexico). The dietary exposure assessments utilised brand-specific use level and market volume data previously determined by Martyn et al. (2017) in combination with population-specific food consumption data for Canada (2015 CCHS-Nutrition) and updated beverage consumption data for the U.S. (2015–2016 NHANES). Dietary exposure to benzoates in water-based flavoured drinks was evaluated using probabilistic (non-brand loyal) and deterministic (brand loyal) statistical models. Specifically, the non-brand loyal model considered the full distribution of use levels, whereas the brand loyal model assumed the brand loyal beverage type (identified as RegCSDs) contained benzoates at the maximum use level (250 mg kg⁻¹ for RegCSDs with pH < 3.5 and up to 438 mg kg⁻¹ for RegCSDs with pH > 3.5), and all other beverage types in which benzoates are used, contained benzoates at the market-weighted average use level. Assessments using either model retained conservative assumptions (to avoid underestimation) while also driving towards more robust, precise, and realistic estimates of exposure to benzoates among the consuming population.

There were no exceedances of the ADI established for benzoates (5 mg kg body weight^-1 day⁻¹, as benzoic acid) in Canadian and U.S. population groups that possessed a statistically reliable population size in either model (up to 76% of the ADI at the 95^th percentile). Taken together, the findings continue to support the safety of benzoate use in water-based flavoured drinks at current use levels (i.e., 250 mg kg⁻¹ for most beverages, and up to 438 mg kg⁻¹ for beverages with pH > 3.5) in markets with prevalent use of this additive in beverages.

Acknowledgments

Branded-use level information on benzoates was provided by International Council of Beverages Associations (ICBA) members including member companies from the American Beverage Association (ABA), the Associacao Brasileira das industrias de Refrigerantes e de Bebidos nao Alcoolicas (ABIR), the Asociación Nacional de Productores de Refrescos y Aguas Carbonatadas, A.C. (ANPRAC), and the Canadian Beverage Association (CBA). Important contributions in the implementation of the 2015 CCHS-Nutrition dataset for the purposes of the analysis was provided by Dr. Shafagh Fallah, of Intertek Scientific & Regulatory Consultancy. This research was funded by the ABA.

Disclosure statement

No potential conflict of interest was reported by the authors. Maia M. Jack is an employee of the American Beverage Association.

Supplementary material

Supplemental data for this article can be accessed on the publisher’s website.

Additional information

Funding

This work was supported by the American Beverage Association.