Understanding consumption-related sucralose emissions — A conceptual approach combining substance-flow analysis with sampling analysis

Understanding consumption-related sucralose emissions — A conceptual approach combining substance-flow analysis with sampling analysis

Science of the Total Environment 408 (2010) 3261–3269 Contents lists available at ScienceDirect Science of the Total Environment j o u r n a l h o m...

1MB Sizes 7 Downloads 24 Views

Science of the Total Environment 408 (2010) 3261–3269

Contents lists available at ScienceDirect

Science of the Total Environment j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / s c i t o t e n v

Understanding consumption-related sucralose emissions — A conceptual approach combining substance-flow analysis with sampling analysis Tina-Simone Schmid Neset a,⁎, Heinz Singer b, Philipp Longrée b, Hans-Peter Bader b, Ruth Scheidegger b, Anita Wittmer b, Jafet Clas Martin Andersson b a b

Department of Water and Environmental Studies, Linköping University, SE-58183 Linköping, Sweden Eawag, Swiss Federal Institute of Aquatic Science and Technology, Ueberlandstrasse 133, CH-8600 Dübendorf, Switzerland

a r t i c l e

i n f o

Article history: Received 18 December 2009 Received in revised form 30 March 2010 Accepted 1 April 2010 Available online 6 May 2010 Keywords: Substance-flow modelling Sucralose Substance-flow analysis Sampling High resolution mass spectrometry

a b s t r a c t This paper explores the potential of combining substance-flow modelling with water and wastewater sampling to trace consumption-related substances emitted through the urban wastewater. The method is exemplified on sucralose. Sucralose is a chemical sweetener that is 600 times sweeter than sucrose and has been on the European market since 2004. As a food additive, sucralose has recently increased in usage in a number of foods, such as soft drinks, dairy products, candy and several dietary products. In a field campaign, sucralose concentrations were measured in the inflow and outflow of the local wastewater treatment plant in Linköping, Sweden, as well as upstream and downstream of the receiving stream and in Lake Roxen. This allows the loads emitted from the city to be estimated. A method consisting of solid-phase extraction followed by liquid chromatography and high resolution mass spectrometry was used to quantify the sucralose in the collected surface and wastewater samples. To identify and quantify the sucralose sources, a consumption analysis of households including small business enterprises was conducted as well as an estimation of the emissions from the local food industry. The application of a simple model including uncertainty and sensitivity analysis indicates that at present not one large source but rather several small sources contribute to the load coming from households, small business enterprises and industry. This is in contrast to the consumption pattern seen two years earlier, which was dominated by one product. The inflow to the wastewater treatment plant decreased significantly from other measurements made two years earlier. The study shows that the combination of substance-flow modelling with the analysis of the loads to the receiving waters helps us to understand consumption-related emissions. © 2010 Elsevier B.V. All rights reserved.

1. Introduction New substances are continuously being introduced to the food processing sector. While testing for human toxicity before introduction is routine, harmful effects on the environment are not always on the agenda. Monitoring programs are designed to detect long-term changes in the environmental concentrations. Such concentration measurements can also be used to estimate the total loads impacting the environment. However, it is in most cases impossible to identify and rank the sources of pollution from environmental measurements alone. This paper presents a study that tries to connect substance-flow modelling and water sampling and analysis for a substance that was introduced in European food products in 2004 (Loos et al. 2009). Correlating the consumption of products with the relevant emissions is complementary to pure environmental monitoring, since it tries to trace the pollution pathways back to the sources. Identifying the ⁎ Corresponding author. Tel.: + 46 13 282288. E-mail address: [email protected] (T.-S.S. Neset). 0048-9697/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.scitotenv.2010.04.003

sources is necessary in order to design possible measures to reduce the pollution and ongoing accumulation of substances of concern in the environment and especially in the aquatic environment. The sweetener sucralose was discovered over 30 years ago and has been in industrial use since the early 1990s. Grice and Goldsmith (2000) pointed out its low toxicity and safety for human consumption in their overview of studies of its toxicity, which has been most relevant for its introduction in the food industry. Its stability in heat and over a wide range of pH conditions make it advantageous for industrial food processing. Being over 600 times sweeter than sugar, this sweetener has been increasingly used in food products and beverages. Today it is used in more than 4000 different food products worldwide and has been approved as a food additive in more than 40 countries (Brorström-Lundén et al. 2008a). In Sweden, more than 100 products containing sucralose are available, and even more via international internet sales. Due to the limited knowledge about its health effects, a maximum daily intake of 15 mg/kg body weight is recommended by the European Union Scientific Committee. (Brorström-Lundén et al. 2008a).


T.-S.S. Neset et al. / Science of the Total Environment 408 (2010) 3261–3269

Sucralose is a tri-chlorinated disaccharide, which is thermally stable and is discharged unchanged after consumption via the excreta. Its high persistence in the aquatic environment gives reason to fear that it can influence aquatic ecosystems (Labare and Alexander, 1993; Lubick, 2008; Loos et al. 2009; Scheurer et al. 2009). The objective of this study is to apply a combination of substanceflow modelling and sample analysis in order to establish the current extent of sucralose emissions from consumption. This study is delimited by the city of Linköping in south-eastern Sweden.

3. Theory and method In this study we combined a consumption analysis of sucralose with concentration measurements at selected sites. This substanceflow analysis gives a first insight into and an overview of sucralose sources and pathways. This is crucial for a discussion of possible measures to reduce pollutant concentrations. Concentration measurements alone only permit the sinks and the loads they receive to be quantified. 3.1. Substance-flow analysis

2. State of the art Baccini and Brunner (1991) introduced regional substance-flow analysis to quantify substance flows in the anthroposphere. This “classical” approach required complete datasets mostly acquired through huge measurement campaigns. Later, the substance-flow analysis was extended by modelling concepts in order to achieve results for flows of systems where only scarce and/or imprecise data where obtainable by Baccini and Bader (1996). Huang et al. (2007) and Chèvre et al. (in press) used measurement results to calibrate their models. Sucralose was chosen as an example by the following reasons: i) sucralose has become a substance of concern see below, ii) the amount used is relatively small compared to other substances but shows a very dynamic behaviour with respect to the main food producers. Recent screening studies across Europe have detected varying concentrations of up to 1000 ng/l of sucralose in aquatic recipients (Brorström-Lundén et al., 2008a, Green et al., 2008, Loos et al., 2009). The highest concentrations were found in the UK, Belgium, Netherlands, France, Switzerland, Spain, Italy, Norway and Sweden. Buerge et al. (2009) and Scheurer et al. (2009) analysed artificial sweeteners in wastewater and lake samples from Germany and Switzerland. Based on their measured concentrations, they estimated daily sucralose wastewater loads per capita of 0.14–0.23 mg/(Cap d) for Germany and 1.5 ± 0.6 mg/(Cap d) for Switzerland respectively. A Swedish screening program including several wastewater treatment plants (WWTPs) in 2007 (Brorström-Lundén et al. 2008a) found large emissions from WWTPs and estimated total emissions from all Swedish WWTPs of approximately 5.5–7 tons of sucralose annually, corresponding to 1.7–2.1 mg/(Cap d). A significant accumulation in aquatic recipients was also noted and the non-degradability of sucralose means that continuous emissions could potentially present a serious risk to aquatic ecosystems. Sucralose seems to be persistent in water. Labare and Alexander (1993) detected that it was slowly metabolised in their environmental samples. Previous studies have shown no bioaccumulation in biota (Brorström-Lundén et al. 2008b), which might be due to its high solubility in the water phase. Nevertheless, sucralose-enriched water could interfere with organism behaviour (Reinders et al., 2006). However, an initial report of the Department of Applied Environmental Science at Stockholm University (Adolfsson-Erici et al., 2009) indicates significantly increasing mortality related to sucralose concentrations in juvenile gammarideans after long-term exposure to 0.5–500 μg/l sucralose. A large number of human toxicology studies of sucralose (e.g. Grice and Goldsmith, 2000, McLean Baird et al., 2000) indicated an increased risk of migraine (Bigal and Krymachantowski, 2006), and some animal testing particularly indicated genotoxicity (Sasaki et al., 2002; Abou-Donia and El-Masry, 2008). Besides many studies about effects of sucralose on the environment and toxicity of sucralose hardly any studies exist tracing back the sucralose loads to its sources. In this study measurement—results from wastewater samples were used in the model to estimate gaps in the consumption patterns.

The method of substance-flow analysis (SFA), as defined in Baccini and Brunner (1991) and extended in Baccini and Bader (1996), is applied to analyse the flows of sucralose through the urban system of Linköping. This method provides a systematic description of substance flows through a defined system. In the past two decades it has been applied to many problems in different fields. For references, see Neset et al. (2008), Schaffner et al. (2009) or Chèvre et al. (in press). The procedure consists of four steps 1. system analysis, 2. model approach, 3. data acquisition and calibrations, and 4. simulations, including sensitivity and uncertainty analysis. 3.1.1. System analysis The system border is the city of Linköping, Sweden, including its receiving waters. The time frame is one day since we are interested in daily flows. The time frame is the same for the measurements made in the waste water and the receiving lake. Sucralose is found in food, beverages, food related articles and medical products. Accordingly possible sources for sucralose in wastewater are industry, households and small business enterprises (households&SBE) all related to the above listed products in a wider sense. Therefore a detailed survey about products containing sucralose was carried out as follows. First, a literature search was conducted and an initial product survey was based on information from a Swedish Consumers Organisation (Sveriges Konsumenter i Samverkan, 2009). Second, producers of food containing sucralose were interviewed on the basis of this information. Data on sucralose concentration and sales were collected. Third, supermarkets and other shops were screened for product groups containing artificial sweeteners such as sugar-free or low-calorie soft drinks, yoghurts, chocolates, chewing gum, sweets, cough drops, syrups and others. Fourth, a detailed internet search was conducted, focusing on online product sales, and last but not least, experts on sucralose were contacted. According to the system knowledge obtained, the most relevant balance volumes for sucralose are households&SBE, food industries, combined sewers, wastewater treatment plants (WWTP) and surface waters. The corresponding flows are the inputs to households&SBE through food products containing sucralose, the wastewater flows and effluent flows from WWTPs. The inputs were grouped into six groups: Coca Cola light, yoghurt and sour milk, ketchup, chocolate, water (drinking and cooking) and others. The input from tap-water was calculated via the concentration measured in the freshwater plant. The group “others” contains all products that were available in stores or online in January 2009 but for which no sales data were obtainable as well as sucralose containing products used or processed in small business enterprises such as bakeries and restaurants. Products containing sucralose are among others artificial sweeteners, drinks (soft drinks, syrups and juices), chewing gum, sweets, dietary food, medical products and protein food. About 100 products were identified belonging to one of these seven subgroups. The system derived from this information is shown in Fig. 1 3.1.2. Model approach If data were available for all inputs of the above system, it would be adequately described by a classical stationary input–output model.

T.-S.S. Neset et al. / Science of the Total Environment 408 (2010) 3261–3269


Fig. 1. System analysis for sucralose flows in Linköping city.

However, it turned out (see Section 3.1.3 below) that not enough data was obtainable for all the food products included in the input “Others” in order to calibrate the model. A modified stationary input–output model was consequently chosen to represent the current system knowledge. This means that data for the input “WWTP input” is given instead of for the input “Others”. The selected approach is a stationary one. This is a good first approximation since we are interested in daily averaged flow patterns and the transport time of wastewater from households&SBE or industries to surface water is only several hours. A dynamic description would be needed to simulate hourly concentration fluctuations, for example, which is not the focus of this study. The parameters of the modified input–output model are the inputs to households&SBE except “Others”, the “WWTP input”, the “Industrial wastewater” and the transfer coefficients of the balance volumes. (Balance volumes: boxes in Fig. 1.) 3.1.3. Data collection and calibration of the model Data on consumption was estimated for January 2009 on the basis of the data for the whole of 2008 and January 2009 respectively. Sampling was conducted in January 2009. 3.2. Measurement campaign The water flow rates were measured at intervals of 6 min at the inlet and the outlet of the wastewater treatment plant and added up to 6-h flows. Since the flow meter at the inlet was not working properly, the 6-h flows at the inlet were calculated from the outlet data, taking into account the hydraulic retention time of the plant, which is about 1 h. Samples were taken from the wastewater treatment plant and from natural waters upstream and downstream of the plant, see Fig. 2. Inflowing wastewaters were sampled for two 24 h cycles (one weekday, and one weekend). An ISCO 6712 sampler was used for timeproportional sampling of the WWTP influent. For each 24-h cycle, four 6 h time-proportional composite samples were collected by taking 35 ml samples from the influent every 5 min. Immediately after sampling, all samples were stored at −20 °C in the freezer and sent the next day packed with cooling elements to the laboratory via express mail.

3.3. Sample treatment and analysis After unfreezing at room temperature, the samples were adjusted to pH 6.8–7 with formic acid and ammonia (both obtained from Merck; Darmstadt, Germany) and filtered with a glass fibre filter (GF/F; 0.7 µm, Whatman; Brentford, U.K.) to determine the dissolved sucralose concentration. Due to an octanol–water partition coefficient (log Kow) of −0.5, sucralose stays predominantly in the water phase. The influent and effluent samples of the waste water treatment plant were diluted at 1:20 and the industrial waste water at 1:100 with water. Surface and drinking water samples were measured undiluted. To correct for analyte losses and matrix interferences during enrichment and analysis, 10 ng d6-sucralose (Toronto Research Chemicals, Toronto, Canada) was added to each sample as an internal standard. The analyses were performed by fully automated online solid-phase extraction coupled directly to liquid chromatography followed by electrospray ionisation (ESI) with detection by high resolution mass spectrometry (SPE–LC–HRMS). The apparatus consisted of a tri-directional autosampler equipped with a 20 ml injection loop (HTC PAL, CTC Analytics, Switzerland), two HPLC gradient pumps (Rheos 2000, Flux Instruments, Switzerland), a loading pump (Thermo Scientific, San Jose, USA), and a LTQ-Orbitrap mass spectrometer (Thermo Scientific, San Jose, USA). The software and hardware setup of the online-SPE–LC–MS/MS approach is described in detail by Stoob et al. (2005). For analysis, 20 ml of the samples were enriched by the loading pump with a flow rate of 2 ml/min on a mixed-layer SPE cartridge containing Oasis HLB in the front layer (Waters, Rupperswil, Switzerland), and a mixture of Strata XCW, Strata XAW (Phenomenex, Munich, Germany) and Isolute ENV + (Separtis GmbH, Grellingen, Switzerland) in the back layer. The cartridge was conditioned with water containing 2.5 mM ammonium acetate (eluents obtained from Acros Organics; Geel, Belgium in HPLC-grade purity). The analytes were eluted from the SPE cartridge in the load direction with methanol (solvent B) by the gradient HPLC pump. Water containing 0.1% formic acid (solvent A) was added to this eluate in a dynamic mixing chamber to form the LC gradient. The chromatographic separation was achieved at a flow of 400 µl/min using a reversed-phase column (Waters, T3 Atlantis, 150 × 3 mm, 3 µm). The LC run started at 10% of solvent B for 4 min, increasing solvent B up to 90% for 13 min; it was then held for 3 min, returning back to 10% of solvent B in 0.5 min and re-equilibrated for 4.5 min. The


T.-S.S. Neset et al. / Science of the Total Environment 408 (2010) 3261–3269

Fig. 2. Map of sampling sites: waste water treatment plant (1a), drinking water plants Berggarden (2a) and Råberga (2b), Stångå upstream WWTP (1b), Stångå mouth (3), Svartå mouth (4), Motala Ström (5), Roxen (6), Stångå Sturefors (7). Map from Sverigekartan ©Lantmäteriverket,Gävle 2010. With kind permission, MEDGIV-2010-25004.

total run time was 25 min. The cartridge was cleaned after elution by flushing with acetonitrile. Sucralose and its internal standard sucralose-d6 were detected after ionisation in negative mode (ESI−) with the LTQ-Orbitrap running in full-scan acquisition mode (scan range of 200–500 m/z) at a resolution of 30,000 (at 400 m/z). As sucralose and sucralose-d6 form significant amounts of adducts with formic acid (FA) in the ESI spray, the exact mass of the FA adduct ions were extracted from the full-scan chromatograms with a mass accuracy of 3 ppm to quantify the sucralose in the samples. MSMS fragments from the FA adducts were additionally analysed in the Orbitrap at a resolution of 7500. For confirmation purposes, the isotope pattern of the parent masses, the intensity ratio of the product ions and the retention time (12.5 min) were checked against standard solutions (see Table 1). Using sucralose-d6 as an internal standard, a calibration curve was constructed from extracted nanopure water standards spiked at sucralose concentrations of 1, 2, 5, 10, 25, 50, 100, 200, 400, 600, 800 and 1000 ng/L. The recovery of spiked sucralose amounts was checked

Table 1 Analysed parent and product ions of sucralose and sucralose-d6. Compound










441.0128b 443.0098c 441.0128b 447.0504b 449.0475c 447.0504b

30,000 30,000 – 30,000 30,000 –

– – 395.0073 – – 401.0449

– – 7500 – – 7500

– – 35% – – 35%

R — resolving power at 400 mlz (full width at half maximum). CID — collision induced dissociation. a Formic acid adduct. b Monoisotopic mass (A). c Isotopic mass (A+2).

for each sample matrix. The recoveries for all matrices were in the range of 80–120%. The limit of quantification (LOQ) in influent, effluent, surface and drinking water was 100, 50, and 10 ng/L respectively. The precision of the method was determined by replicate analysis (n = 7) of aliquots from spiked samples (spike level 100 ng/L). The relative standard deviation of the average concentration was less than 7% for all sample matrices. The flow rates measured and the concentrations obtained through the lab analysis are shown in Table 2. The difference of the 24 h loads between weekdays and weekends is not significant. Some 70–75% of the total daily load enters the WWTP between noon and midnight, corresponding to the times of highest food and drink consumption. 3.4. Consumption analysis To obtain data for the input flows, we contacted known and potential food processing companies in Sweden and importers for information on sucralose content and sales data for the year 2008. Sales statistics were only obtainable for some products, and merely on a national scale. Therefore the data for whole of Sweden was downscaled to Linköping according to the population (population of Sweden: 9 mio; population of Linköping connected to WWTPs: 140,000). This procedure can be expected to involve some uncertainty. However, sales data for the whole of Sweden and the Linköping area were obtained from one company for one product, showing that the uncertainty was less than 10%. For the products included in “Others”, no sales data were obtainable from the producers due to business confidentiality. A private company (Nielsen, The Nielsen Company Sweden, 2009) collects sales data for many products and carefully selected sales channels in Sweden. On the basis of these data, Nielsen provides statistical estimations for product sales data on a regional scale. These data are commercially available. In order to quantify the contribution of the identified products in “Others”, the sucralose concentration for

T.-S.S. Neset et al. / Science of the Total Environment 408 (2010) 3261–3269 Table 2 Results of the measurement campaign of Eawag and Linköping University in January 2009 (flow rates and concentrations, resulting sucralose loads) and concentrations of the measurement campaign of IVL in 2007, Brorström-Lundén et al. (2008a). Samples taken during measurement campaign of Eawag and Linköping University

Flow rate Sucralose conc. [m3/time unit]a [ng/l]

Weekday (Tuesday noon to Wednesday noon) Inflow WWTP 20.01.2009 12,490 12:00–18:00 Inflow WWTP 20.01.2009 12,880 18:00–24:00 Inflow WWTP 21.01.2009 6650 00:00–06:00 Inflow WWTP 21.01.2009 10,180 06:00–12:00 Total inflow 20./21.01.2009 41,600 12:00–12:00 39,960 Effluent 21.01.2009 00:00–24:00b Weekend (Saturday noon to Sunday Inflow WWTP 24.01.2009 12:00–18:00 Inflow WWTP 24.01.2009 18:00–24:00 Inflow WWTP 25.01.2009 00:00–06:00 Inflow WWTP 25.01.2009 06:00–12:00 Total inflow 24./25.01.2009 12:00–12:00 Effluent 25.01.2009 00:00–24:00b Drinking water Bergårdens 19.01.2009 14:00 raw water Clean water Drinking water Råberga 19.01.2009 14:00 raw water Clean water Industry waste water 20./21.01.2009 09:00– 09:00 (part separately measured)

Sucralose load [mg/time unit]a








19,400 103,000



noon) 16,150












43,210 40,510

102,600 2300



7.2 b10


b10 17,000


Measurements performed by IVL 19. June 2007 (Brorström-Lundén et al. (2008a)) Influent to WWTP 39,554 m3/dc 3530, 5270, 5990 195,000 mg/d Linköping 3 c 4920, 5110, 5440 204,000 mg/d Effluent to WWTP 39,554 m /d Linköping a

Time unit: is according to the information in the first row e.g. 6 h, 24 h. b A delay of 12 h was assumed to account for the retention time of substances in the WWTP. Therefore the shown sucralose effluent corresponds to the total inflow. c The hourly flow rate data for 19 June 2007 were obtained from the WWTP operators in Linköping.

each product would have to be measured in addition to such sales data. For the first overview presented here, very detailed information about the individual contributions to “Others” was not required. For plausibility considerations (see Results and discussion section), we quantified the sucralose load for one product. In conclusion, we have data for all input flows to “Household&SBE” except “Others”. Nevertheless, a complete dataset for the model could be obtained by using input data for “WWTP input” instead of “Others”. The “Others” flow is thus a result of the simulation rather than part of the input data as explained in the “Model approach” section above. For the industrial effluent, discharge concentrations were measured and water discharge flows were provided. Data for the transfer coefficients of the input/output ratios for households&SBE, sewers and WWTPs


were obtained from the literature and our own estimations. Table 3 shows all the data used.

4. Results and discussion Fig. 3 shows the simulated sucralose flow for a typical weekday and a weekend day in winter 2009. Industrial emissions to WWTPs include losses caused by food production as well as the contribution from the water used. The total load (based on measured concentrations and flow rates) to the WWTP is 103 g/d, equivalent to 0.7 mg/(Cap d). A total of

Table 3 Data for the model: input flows and transfer coefficients. “Tnormal” means truncated normal distribution. Mean On the market in Jan2009 Coca Cola light

Probability Source distribution

Stopped 0 July 2008 105 mg/l

Concentration of sucralose Yes Yoghurt and sour milk, national sales data Concentration of sucralose Ketchup sales, Yes national sales data Concentration of sucralose Chocolate sales, Yes national sales data Concentration of sucralose Others: Drinks Concentration of sucralose Diet food Concentration of sucralose Chewing gum and sweets Concentration of sucralose Protein food (fitness) Concentration of sucralose Medical products Concentration of sucralose Industry waste water (only weekdays) (part separately measured) Concentration of sucralose Input household water Concentration of sucralose Transfer coefficient HH to sewer Transfer coefficient CS to surface water Transfer coefficient WWTP to surface water

Dye et al. (2007)

400,000 kg Jan. 2009


Pers. comm. food industry

100 mg/kg


94,000 kg/y


200 mg/kg


21,800 kg/y


200 mg/kg


Pers. comm. industry Pers. comm. industry Pers. comm. industry Pers. comm. industry Pers. comm. industry

? b300 mg/kg

food food food food food

Undén and Knutsson (2007)

? b800 mg/kg

Undén and Knutsson (2007)

? b3000 mg/kg

Undén and Knutsson (2007)

? ∼500 mg/kg ? ∼4000 mg/l 1265 m3/d


See Table 2

17,000 ng/l


See Table 2

25,200 m3/d



7.2 ng/l


See Table 2



Loos et al. (2009)






Brorström-Lundén et al. 2008a, Buerge et al. (2009), Scheurer et al. 2009


T.-S.S. Neset et al. / Science of the Total Environment 408 (2010) 3261–3269

Fig. 3. Sucralose flows in [g/d] through Linköping city for a) weekday and b) weekend day.

21% (0.15 mg/(Cap d) originates from the food industry and 79% from households&SBE. This shows that the contribution from households&SBE is significantly larger than that of the local food industry. Yoghurt and sour milk contributed 21% (0.13 mg/Cap d) to the household&SBE load and others 79% (0.5 mg/(Cap d)). The loads from ketchup, chocolate and water are 1–2 orders of magnitude smaller and of minor importance. The load from Coca Cola light is zero, since it was removed from the market in summer 2008. It is still offered on the Internet and there might still have been some minor stocks left in January 2009. However, a survey of supermarkets, restaurants, canteens and private households showed that the storage time of soft drinks is less than two months in summer. The reasons are the expiry date of less than six months and the increased demand in summer. We therefore ignored the Coca Cola light loads in such stocks. The total loads on weekdays and on weekend days are almost the same. The reduced load from industry is apparently compensated by a higher consumption of sucralose by households&SBE on weekends. In the simulations for the weekend day, we assumed the same yoghurt

and sour milk consumption as on weekdays. This assumption is supported by a small survey in several households and the fact that yoghurt is a daily food, and not a luxury food, which is mainly consumed on weekends. We suppose that the higher load from households&SBE on weekends is due to a higher consumption of sweets, candies, etc. on weekends.

4.1. Uncertainty and sensitivity analysis A crucial question arising from the results seen in Fig. 3 is: what is the certainty that the flow through the input “Others” is high compared to the other input flows. This question can only be answered by a careful uncertainty analysis. There are three main sources of uncertainty for the data collected, namely: 1) uncertainties caused by the measurement procedure, 2) uncertainties due to downscaling of national Swedish sales data to Linköping, and 3) uncertainties due to using yearly or monthly

T.-S.S. Neset et al. / Science of the Total Environment 408 (2010) 3261–3269

averages for daily consumption and using one day measurements. We identified the following quantitative uncertainties: 4.1.1. Measured data Flow rates to and from WWTPs: the standard deviation (STDV) of the measurement procedure caused by the flow meter is estimated to be 15% according to the WWTP staff. A time series for the years 2002 to 2008 reveals a regular weekly pattern. In particular, it can be seen that the individual weekdays differ less than 10–20% from each other. The same is true for weekend days. This uncertainty is taken into account since the measurement campaign was run on only a single day. Concentration of sucralose: according to the sampling procedure and the chemical analytical methods, it is know from experience that the STDV is within 20% (see Sample treatment and analysis section). 4.1.2. Sales data We assume that the data directly obtained from producers are quite accurate. According to the discussion in the “Consumption analysis” section, a STDV of 15% for the downscaling from the whole of Sweden to Linköping is adequate. The sales data obtained for yoghurt and ketchup show a seasonal variation depending on the product. In summer, for instance, sales data for yoghurt are above and in winter below the yearly average. For ketchup, the variation is much smaller. Since our measurement campaign took place in the last third of January, we used the January sales data for yoghurt and the annual sales data for ketchup. The sales data for yoghurt for January 2009 was about 15% below the yearly average. By using this average value for daily consumption, we assigned a STDV of 15%. The probability distributions for the data in Table 3 are assumed to be normal, lognormal and uniform, depending on the following assumptions: a lognormal distribution was assumed for the concentrations since it is known from chemical measurements that concentrations usually show a lognormal distribution. For all sales data, a truncated normal distribution was assumed, as is common for fluctuating quantities, which can only be in the positive range. Here, a truncated normal distribution means truncated by zero to allow only positive sales values. A uniform distribution was assumed due to the small range and the limited information on the transfer coefficients. Table 3 shows the probability distribution for all the data used.


Fig. 4 shows the calculated probability density distribution for some flows on weekdays, using a Monte Carlo simulation with a sample size of 100,000. The input “Others” shows a slightly asymmetrical normal distribution between 25 and 125 g/d, according to the uncertainty of the data. With a probability of 95%, “Others” is greater than 43 g/d, with 90% exceeding 48 g/d and 80% exceeding 55 g/day, which proves that this flow cannot be neglected. An important question is: which data uncertainty is most responsible for the large uncertainty of “Others”. An “uncertaintyrelated” sensitivity analysis (see also Schaffner et al. (2009)) shows the following: the uncertainties of the WWTP input data and the data for wastewater from “Industry” account for 88% of the uncertainty of “Others”. Another 7% of the uncertainty is explained by the uncertainty of the sales data and the concentration of “Yoghurt and sour milk”. The rest comes from the other data uncertainties. 4.2. Plausibility considerations The simulation shows that the load attributed to “Others” is 69 g/d or 0.5 mg/(Cap d) on weekdays and 90 g/d or 0.64 mg/(Cap d) per day on weekends respectively. The crucial question is whether this load can be explained by consumption of the products identified in the group “Others”. Table 4 shows the sucralose content in some products or subgroups of “Others” as well as the ‘fictive’ daily consumption per capita and product that is equivalent to the daily per capita load of ‘Others’. As can be seen, even a moderate consumption of these products is sufficient to result in the measured load. This analysis shows that the number of products equivalent to the daily load of “Others” is quite small. In order to get a feeling for the contribution of products in “Others”, we quantified the sucralose load of one product, namely a medical chewing gum designed to replace smoking. We ordered sales data for the southeast region for the year 2008 from Nielsen: 553,000 pieces of chewing gums through nonpharmacy sales channels. This relates to a population of 1,123,400 and a sucralose concentration of 5 mg/piece. Downscaled for Linköping on a daily basis results in a contribution of about 1 g/day to the load to WWTPs. This shows that this specific product alone contributes 1.4% of the daily load to ‘Others’.

Fig. 4. Probability density for sucralose input of “Yoghurt and sour milk”, “Others”, “Industry waste water” and “WWTP outflow” in g/d for Linköping. The values of the x-axis represent sucralose loads in g/d for Linköping. The units of the y-axis are such that the area under the curve is equal to one.


T.-S.S. Neset et al. / Science of the Total Environment 408 (2010) 3261–3269


Content of sucralose as given by producers

Product equivalent to daily per capita “Others” load

years 2007 and 2009. In contrast, the concentration in Lake Roxen did not vary from 2007 to 2009. In fact, Brorström-Lundén et al. (2008a) found 52 ng/l in 2007 whereas this study found 50 ng/l in January 2009. This might be due to the fact that the sample site in Lake Roxen is separated from the main stream of effluent of the WWTP.

Chewing gum Medical chewing gum Medical mouth rinse Antihistamine solution Soft drinks Energy bar

1–2 mg/piece 4–5 mg/piece

0.25–0.5 piece 0.1–0.125 piece

4.4. Comparison with other countries

0.4–0.9 mg/ml

0.5–1 ml

4 mg/ml

0.125 ml

50–150 mg/l 3–15 mg/bar (100–400 mg/kg)

0.003–0.01 l 0.03–0.17 bar

Table 4 Sucralose content in selected products and daily product and person equivalent to the load of “Others”.

4.3. Comparison with the past and expectation for the future Comparing the measurements for 2009 with those of BrorströmLundén et al. (2008a) in Table 2 shows that the load to WWTPs decreased from 195 g/d in 2007 to about 103 g/d in January 2009. What is the reason for this significant decrease? The same consumption analysis as above for January 2009 was performed for 2007 using sales data for 2007. The main difference to 2009 is the large contribution from Coca Cola light. Since no sales data for Coca Cola were obtainable for Sweden, we used the estimations of Dye et al. (2007, 2008) made for Norwegian consumption. Their yearly per capita consumption based on sales data was 4.92 l. The measured concentration was 105 mg/l. This corresponds to a consumption of 1.4 mg/(Cap d) equivalent to a load of 198 g/d for the whole of Linköping. The results are presented in Table 5. For January 2009, the largest loads could not be attributed to a few individual products, but to the many products included in “Others”. This is significantly different from 2007/2008, when the estimated load from Coca Cola light was dominant. This reduction of the sucralose load from households&SBE in Linköping from 225 g/d in 2007 to 88 g/d in 2009 is very significant. In comparison the measured decrease in the input to the WWTP decreased from 195 g/d to 103 g/d. Table 5 also shows what can be expected for 2010, namely a further reduction due to the announcement by the main producers of yoghurt and ketchup to remove sucralose from their products in spring 2009. However, there still remains a significant load caused by the many products contained in “Others”. In conclusion, some main producers in Sweden have removed sucralose from their products, resulting in a strong decrease of the sucralose load. However, imported products are not affected by this decision. Whether the load in 2010 will be lower or higher compared with 2009 is difficult to predict. We have observed in the last three months that food producers in Switzerland, Austria, Italy and Norway have added sucralose to soft drinks and chewing gum. Based on this observation, we do not expect a strong reduction of the remaining sucralose load of “Others”. What is certain is that the load of sucralose in wastewater is very dynamic and sensitive to the decisions of the food producers. This is clearly demonstrated by comparing the sucralose concentrations in the WWTP inflows and outflows for the

Sucralose concentrations in wastewater have been measured in several other countries. These concentrations are not directly comparable, unlike the estimated daily loads per capita. (The reason is that concentrations depend on the number of households connected to a WWTP and on the volume of the receiving waters.) Table 6 shows the estimated daily loads for Switzerland, Germany and Sweden. Table 6 shows that the sucralose load in the WWTPs and the receiving water varies from about 0.2 mg/(Cap d) to 2 mg/(Cap d) depending on the country. A possible reason for this difference could be that the German food industry prefers other artificial sweeteners. The sucralose load is currently significantly smaller than that of other artificial sweeteners. An important question is whether there is a potential for producers to increase the use of sucralose in food and drink. On the basis of the sweetness factor, total artificial sweeteners correspond to about 5% of sugar consumption, of which sucralose accounts for about 20%. These considerations show that there is indeed some potential for producers to use more sucralose. 5. Conclusions Environmental monitoring often focuses on measuring concentrations in various environmental compartments. This is of course important, since these concentrations are crucial for organisms. However, the sources of pollution can then only be supposed rather than identified. This can be done by a combination of substance-flow analysis and monitoring programs. The substance-flow analysis assigns the use of substances in daily consumption to the final loads to the environment. This identification of the entire pathways and origins of the substances is needed to design possible measures to reduce the loads. Such a combination of substance-flow analysis and measurement results is the contribution of this study. For the example of sucralose, the contribution of some food products was quantified. The contribution of the sum of the products was quantified in the absence of consumption data by applying a simple model and using uncertainty and sensitivity analysis. In order to further reduce the part of the sucralose load not currently assigned to single products this analysis shows that for the product groups in Table 4 consumption data would be needed. Although no decisive ecotoxicological tests are available to date, a number of studies point out the potential risk of spreading a Table 6 Comparison of estimated daily load to WWTP for artificial sweeteners for several countries ± x.y means STDV. Substance

Table 5 Estimated daily contributions to household emissions for Linköping on a weekday g/sucralose/d.

Coca Cola light Yoghurt Ketchup Chocolate Others Freshwater Total



2010 (expected)

198 25 0.7 0.19 ? 0.18 N 225

0 18 0.8 0.19 69 0.18 88.2

0 0 0 0.19 ? 0.18 ?

Acesulfame Cyclamate Saccharin Sucralose Total Sugar

Switzerland 2008 [mg/(Cap d)]

Germany 2009 [mg/(Cap d)]

Sweden 2007 [mg/(Cap d)]

Sweden 2009 [mg/(Cap d)]

Buerge et al. (2009)

Scheurer et al. 2009

BrorströmLundén et al. (2008a)

This study

10 ± 3.4 11 ± 6.7 3.9 ± 1.7 1.5 ± 0.6 37 130–140 [g/ (Cap d)]

8–9.3 32.3–39 7.3–9.3 0.14–0.23 47

1.7–2.1 ?

0.76 ± 0.1 110–120 [g/ (Cap d)]

T.-S.S. Neset et al. / Science of the Total Environment 408 (2010) 3261–3269

persistent organochlorides in high concentrations to the aquatic environment. As this study proposes, new substances that bear a potential environmental risk should be monitored in a similar way, namely by monitoring the inputs to the environment, the environment itself (e.g. lakes) as well as the inputs to consumption. Acknowledgements We would like to thank the following people: M. Bader and L. Gradin for their helpful data inquiries and discussions, a number of experts in the field of sucralose who helped us with their expertise, C. Johansson at Tekniska Verken AB, Linköping, for support with our sampling and C. Dye for helpful discussions. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.scitotenv.2010.04.003. References Abou-Donia MB, El-Masry EM. Splenda alters gut microflora and increases intestinal P-glycoprotein and cytochrome P-450 in male rats. J. Toxicol. Environ. Health, Part A 2008;71:1415–29. Adolfsson-Erici M, Eriksson Wiklund A-K, Alsberg T, Breitholtz M, Ek C, Minten J. Undersökning av det syntetiska sötningsmedlet sukralos med avseende på eventuella ekotoxikologiska effekter. Institutionen för tillämpad miljövetenskap, ITM. Sweden: Stockholms universitet; 2009. Baccini P, Bader H-P. Regionaler Stoffhaushalt. Heidelberg, Germany: Spektrum Akademischer Verlag; 1996. Baccini P, Brunner PH. Metabolism of the Anthroposphere. Berlin/New York: Springer; 1991. Bigal ME, Krymachantowski AV. Migraine triggered by sucralose — a case report. headache. J. Head Face Pain 2006;46(3):515–7. Brorström-Lundén E, Svenson A, Viktor T, Woldegiorgis A, Remberger M, Kaj L, et al. Measurements of sucralose in the Swedish Screening Program 2007 — PART I; sucralose in surface waters and STP samples. IVL Report B1769; 2008a. Brorström-Lundén E, Svenson A, Viktor T, Woldegiorgis A, Remberger M, Kaj L, et al. Measurement of sucralose in the Swedish Screening Program 2007 — Part II. Sucralose in Biota samples and regional STP samples; 2008b. Buerge IJ, Buser H-R, Kahle M, Müller MD, Poigner T. Ubiquitous occurrence of the artificial sweetener acesulfame in the aquatic environment: an ideal chemical marker of domestic wastewater in groundwater. Environ Sci Technol 2009;43(12):4381–5.


Chèvre, N., Guignard, C., Rossi, L., Pfeifer, H.-R., Bader, H.-P., Scheidegger, R.. Substance flow analysis as a tool for urban water management: the case of copper in Lausanne, Switzerland. (submitted Apr. 2009). Water Science and Technology in press. Dye,C; Kylin, H., Schlabach, M.. Sucralose — a new persistent pollutant in the environment? NILU PP 08/2007. Dye C, Kylin H, Schlabach M. Sucralose — a new persistent pollutant in the environment? NILU Presentation; 2008. Green N, Schlabach M, Bakke T, Brevik EM, Dye C, Herzke D, et al. Screening of selected metals and new organic contaminants 2007. Report 1013/2008. Oslo, Norway; 2008. Grice HC, Goldsmith LA. Sucralose — an overview of the toxicity data. Food Chem Toxicol 2000;38(Suppl. 2):1–6. Huang D-B, Bader H-P, Scheidegger R, Schertenleib R, Gujer W. Confronting limitations: new solutions required in urban water management of a Chinese mega-city. J. Environ. Manage. 2007;84(1):49–61. Labare MP, Alexander Martin. Biodegradation of sucralose in samples of natural environments. Environ. Toxicol. Chem. 1993;12(5):797–804, doi:10.1897/1552-8618 (1993)12[797:BOSACC]2.0.CO;2. Loos R, Gawlik BM, Boettcher K, Locoro G, Contini S, Bidoglio G. Sucralose screening in European surface waters using a solid-phase extraction–liquid chromatography–triple quadrupole mass spectrometry method. J Chromatogr 2009;1216:1126–31. Lubick N. Artificial sweetener persists in the environment. Environ. Sci. Technol. 2008;42(9):3125. McLean Baird I, Shepard NW, Merritt RJ, Hildick-Smith G. Repeated dose study of sucralose tolerance in human subjects. Food Chem Toxicol 2000;38(Suppl. 2):123–9. Neset T-SS, Bader H-P, Scheidegger R, Lohm U. The flow of phosphorous in food production and consumption Linköping, Sweden 1870–2000. Sci Total Environ 2008;396:111–20. Reinders A, Sivitz AB, Hsi A, Grof CPL, Perroux JM. Sugarcane ShSUT1: analysis of sucrose transport activity and inhibition by sucralose. Plant Cell Environ 2006;29:1871–80. Sasaki YF, Kawaguchi S, Kamaya A, Ohshita M, Kabasawa K, Iwama K, et al. The comet assay with 8 mouse organs: results with 39 currently used food additives. Mutat. Res./Genet. Toxicol. Environ. Mutagen. 2002;519(1–2):103–19. Schaffner M, Bader HP, Scheidegger R. Modelling the contribution of point-source and nonpoint-source pollution to the Thachin River. Sci Total Environ 2009;407:4902–15. Scheurer M, Brauch HJ, Lange FT. Analysis and occurrence of seven artificial sweeteners in German waste water and surface water and in soil aquifer treatment (SAT). Anal Bioanal Chem 2009;394(6):1585–94. Stoob K, Singer HP, Goetz CW, Ruff M, Mueller SR. Fully automated online solid phase extraction coupled directly to liquid chromatography–tandem mass spectrometry — quantification of sulfonamide antibiotics, neutral and acidic pesticides at low concentrations in surface waters. J Chromatogr 2005;1097(1–2):138–47. Sveriges Konsumenter i samverkan. Produkter med sucralose. (Swedish Consumers Organization. Products containing sucralose). http://www.konsumentsamverkan. se/11verk/kampanj/lightprd/listaforetag_sukr.htm2009 20090930. Nielsen (The Nielsen Company Sweden). Maria Skolgata 83, Stockholm, Sweden. www. se.nielsen.com2009. Undén E, Knutsson P. Socker i ny sockerfri version — sukralos godkänt trots otillräcklig forskning. Göteborg University (A sugar-free sweetener — sucralose approved despite insufficient research); 2007.