European Food Research and Technology

, Volume 228, Issue 5, pp 679–684

Release of bisphenol A from polycarbonate baby bottles: water hardness as the most relevant factor


  • Sandra Biedermann-Brem
    • Official Food Control Authority of the Canton of Zürich
    • Official Food Control Authority of the Canton of Zürich
Original Paper

DOI: 10.1007/s00217-008-0978-8

Cite this article as:
Biedermann-Brem, S. & Grob, K. Eur Food Res Technol (2009) 228: 679. doi:10.1007/s00217-008-0978-8


The concentrations of bisphenol A (BPA) in the content of polycarbonate baby bottles reported by scientific literature were almost always clearly below 1 μg/l, but in a German consumer journal they reached 157 μg/l. These high values were interpreted as a result of microwave heating, but here they are shown to be the result of testing with tap water. Since BPA is primarily released by degradation of the polycarbonate, rather than by migration from the polymer, testing with food simulants (distilled water or distilled water/ethanol) is not appropriate. Degassing of tap water during boiling causes the pH to increase and the water to become more aggressive. BPA concentrations may reach 50 μg/l if a polycarbonate bottle is sterilized by boiling water in it (well feasible only by means of microwave heating) and this same water is used to prepare a beverage. Increased concentrations are also observed when boiling-hot beverages with a high pH are filled into the bottle, such as boiled plain water or tea. Respecting simple rules, the BPA concentrations can be kept below 0.5 μg/l.


Baby bottlesBisphenol APolycarbonateHard tap waterMicrowave heating


The release of bisphenol A (BPA) from polycarbonate baby bottles was a subject of debate regarding its safety as well as the exposure levels. There is uncertainty in the toxicological evaluation, as expressed by the NTP report [1] and the report of the Canadian authorities [2], the latter proposing to ban the use of polycarbonate baby bottles in the mean time. The European Food Safety Authority (EFSA) concluded that these concerns are not sufficiently supported and adheres to a tolerable daily intake (TDI) of 0.05 mg/kg body weight (bw), derived from weak estrogenic activity with reproductive and development toxicity [3, 4]. In a draft, the US Food and Drug Administration (FDA) arrives at the essentially same conclusion [5]. This TDI is higher than the dose for which some effects have been observed in animal tests [1], but it is debated whether these effects could be a threat to human health.

Baby bottles are the most critical source of BPA, since exposure of babies (in terms of amount per body weight) tends to be the highest, BPA might be conjugated to the glucuronide more slowly than in adults (resulting in higher plasma levels of BPA in the active form) and the exposure occurs at a possibly critical development stage of the organism.

The EU-Directive 2002/72/EC set a specific migration limit (SML) of 3 mg/kg food. In 2002, the EU Scientific Committee on Food (SCF) reacted to new findings by increasing the uncertainty factor to 500 and lowering the TDI to 0.01 mg/kg bw as a precautionary temporary measure, which prompted the European Commission to lower the SML to 0.6 mg/kg (Directive 2004/19EC). The opinions of the EFSA [3, 4] would allow to increase the SML again, but this did not happen so far.

For babies, the maximum tolerable concentration in foods and beverages must be calculated differently, since the consumption related to the body weight is higher. Assuming a daily consumption of 800 ml from baby bottles and a body weight of 4 kg, the TDI of 0.05 mg/kg bw would be reached with a BPA concentration of 250 μg/l.

There were discrepancies also in the data on exposure (concentrations in the bottle content). For new bottles as well as those somewhat aged in the laboratory, the release of BPA under mostly rather drastic conditions (such as 1 h at 100 °C) was typically in the range of 0.1–1 μg/l [69]. Le et al. [10] found values reaching from the same range up to 7.7 μg/l, but this highest value involved heating to 100 °C for 24 h, which does not reflect realistic treatment. Several food control authorities in Europe investigated the release of BPA from baby bottles and the large majority of the results was again below 1 μg/l (mostly not published; exception [11]). However, there were also a few far higher results, and the question was whether these are relevant.

In 2003, Brede et al. [8] showed that the release of BPA into hot water was around 0.2 μg/l, when the bottles were new and increased to 6–8 μg/l after repeated washing in a dish washer. The highest concentrations reached 16 μg/l. These results were disturbing, since they did not conform to usual migration behavior, where a decrease is observed after continued use, and since the increase could not be explained, this left the possibility that sometimes still higher concentrations would be encountered. We observed that aging increases the wettability of the bottle wall [12]. Since drying in the washing machine causes dissolved salts in the remaining water to reconcentrate on the bottle wall and to be baked onto the polycarbonate at elevated temperature, these may promote the degradation of the polymer and the release of BPA. The amount of water adhering to the bottle surface is a function of the wettability. The contribution of BPA released during drying of the bottle is particularly high when alkali material is deposited, such as washing solution, if rinsing at the end of the washing cycle fails.

Also in 2003, the German consumer journal Oeko-Test [13] published concentrations reaching 157 μg/l. These concentrations far exceeded those reported previously and were ascribed to the microwave heating. Investigations revealed that the data was obtained with tap water from Berlin, heated to a temperature close to the boiling point for 2 h [14].

Probably as a response to the high BPA concentrations published by Oeko-Test, Ehlert et al. [9] published work sponsored by PlasticsEurope focusing on microwave heating. Heating to the boiling point for 5 min resulted in BPA concentrations between <0.1 and 0.7 μg/l, i.e. in the same low range as reported in most literature. It was concluded that electromagnetic radiation had no effect on the BPA release. It was also shown that there was no correlation with the concentration of the residual BPA in the polycarbonate (1.4–35.3 mg/kg). This left the high concentrations published by Oeko-Test alone and unexplained.

Basically it was known for a long time [15] (and shown again in [12]) that BPA migration in the proper sense, i.e. diffusion from the polymer into the beverage, is a negligible contribution, and most of the BPA is from the degradation of the polymer; it is release rather than migration. For this reason exposure should not be assessed by conventional migration testing, but by investigation of the conditions causing depolymerization of the polycarbonate. In fact virtually all data published in literature was obtained using distilled or otherwise demineralized water, sometimes including 10% ethanol, whereas the Oeko-Test data was obtained with tap water.

Since polycarbonate is labile in alkali environment and tap water tends to have an increased pH upon boiling owing to evaporation of carbon dioxide and shift of the equilibrium of carbonic acid/bicarbonate/carbonate towards the carbonate, it appeared important to revisit the release of BPA using tap water. This enabled to reproduce the high concentrations published by Oeko-Test, to identify the scenarios of preparing beverages in baby bottles which could result in elevated BPA concentrations and to conclude in recommendations enabling to keep BPA release at a low level.



Polycarbonate baby bottles of type 1 were from Dr. Brown’s (St. Louis, USA; 240 ml), bottles 2 from Avent (Suffolk, UK; 260 ml). HPLC involved a SpectraSystem P4000 (Thermo Separation Products, San Jose, USA) with a fluorescence detector FL3000. Methanol (HPLC) was from Baker (Deventer, The Netherlands). Deionized water was produced by a Purelab ultra system (ELGA, Labtec Services AG, Wohlen, Switzerland). The kitchen microwave oven Microwave 2200 M was from König (Zurich, Switzerland) and included a turning table.

Water 1 from Küsnacht had a hardness of 35°f (French degrees) and a pH 7.7. Water 2 was from a fountain at the laboratory with a hardness of 22°f and a pH 7.4. Water 3 was from a fountain on Forch: 37°f, pH 7.5.

Baby bottles were filled with 200 ml water and heated in a microwave oven to the temperature and during the time of interest. A cap was loosely screwed on top. During boiling, the heating power was reduced as soon as the first vapor bubbles were formed.

Bisphenol A was analyzed by the method described in [12]. Briefly, 200–500 μl of water was injected into a 250 × 4.5 mm i.d. column packed with Spherisorb ODS-2, 5 μm (Grom, Herrenberg-Kayh, Germany). BPA was eluted isocratically with 70% methanol/water at 0.75 ml/min. Fluorescence was detected at 226/296 nm. The detection limit was at 0.5 μg/l, determined by interfering peaks of which it was not investigated whether or not they represented BPA. The measuring uncertainty was below 20%.


The series of experiments reported in Table 1 reflect the key findings. The new bottles 1 and 2 were filled with 200 ml deionized water (pH 5.0) and heated in the microwave oven for 5 min, with gentle bubbling after the boiling point was reached. No BPA was detected above the detection limit of 0.5 μg/l, which is in line with previous findings.
Table 1

Sequence of experiments with baby bottles 1 and 2 showing that elevated release of BPA is due to elevated temperature combined with the use of hard water and an increased pH as a result of the evaporation of carbon dioxide


Bottle 1

Bottle 2

BPA (µg/l)


BPA (µg/l)


1. New bottle, deionized water, 5 min boiling





2. Water 1, 5 min boiling





3. …same water boiling another 5 min





4. Standing overnight




5. Water replaced by deionized water, 5 min boiling




6. Water replaced by deionized water, 5 min boiling




7. Filled with boiling deionized water




8. Water replaced by deionized water of 50 °C

< 0.5



9. Bottle cleaned by brush, deionized water, 5 min boiling




10. Cleaned in hot water/vinegar; deionized water, 5 min boiling



The experiment was repeated using rather hard (35°f) drinking water 1. After 5 min of boiling and cooling to about 50 °C in the bottle, 2.9 and 2.5 μg/l BPA were detected for bottles 1 and 2, respectively, and the pH of the water had slightly increased from 7.4 to 7.9 (line 2). There was well visible precipitation of limescale on the bottle wall (Fig. 1), with some more precipitated to the bottom of the flask. The same water was boiled another time for 5 min, causing the BPA content to increase to 10.6 and 7.5 μg/l, respectively, and the pH to increase to 8.1. The increased pH caused the BPA release to be accelerated. The water in bottle 1 was allowed to stand overnight in the bottle at ambient temperature, which had no substantial effect. This experiment shows that an increased pH resulting from the use of hard water strongly increases the liberation of BPA from the polycarbonate. There were no significant differences between the bottles from the two different manufacturers.
Fig. 1

Limescale on the internal bottle wall: new baby bottle heated 5 min with tap water in a microwave oven. Black paper inside the bottle to improve the visibility

The water in the bottle with limescale on the wall was replaced by deionized water and again boiled for 5 min. The pH rose from 5.4 to 9, and 18 μg/l of BPA was released. Repeating the experiment with new deionized water reproduced the result. The presence of lime and the absence of bicarbonate buffering the system caused a rapid increase of the pH, which in turn caused more BPA to be liberated.

Just filling boiling water into the bottle and allowing it to cool to ambient temperature merely released 3 μg/l BPA and the pH only rose to 8.3 (line 7). Deionized water heated to only about 50 °C for 5 min and allowed to cool to ambient temperature inside the bottle (line 8) liberated BPA below the detection limit of 0.5 μg/l. The pH remained at 5.8.

Finally the bottle was cleaned with water using a brush, which resulted in the complete removal of the limescale as far as visible by eye. Boiling deionized water in the bottles for 5 min (repeating experiments 5 and 6) merely formed 1.5 μg/l BPA and the pH hardly increased. Cleaning by sterilization in boiling water containing 3% vinegar in a pan completely removed the limescale: boiling deionized water for 5 min inside the bottle by microwaves no longer released a detectable amount of BPA.

Table 2 presents further results using bottles of type 1. Experiment 1 aimed at distinguishing whether it is the limescale adhering to the bottle wall or the water at elevated pH that degraded the polycarbonate. Water 2 (22°f) was boiled in a pan for some 10 min, during which much limescale precipitated and the pH increased to 9. This water was filtered (a blank checked absence of BPA in the filter) and boiled in the baby bottle for another 5 min by microwaves. No further precipitation of limescale was visible. After cooling, 64 μg/l BPA were measured in the water, indicating that it is the alkali water that hydrolyzes the polycarbonate.
Table 2

Experiments with bottles of type 1: see text


BPA (µg/l)


1. Boiled water 2 (10 min), 5 min boiling



2. 60 min boiling with water 3



3. Filled with well boiled tea from water 2, wrapped in towel



4. Laying bottle, 5 min boiling in pan with 1.5 l water 2



5. Standing filled bottle, 5 min in pan with boiling water (1.5 l)



Experiment 2 checked whether the high concentrations reported by “Oeko-Test” [13] (88, 114 and 157 μg/l after 2 h of boiling) could be reproduced. Moderately hard water 2 (similar to that in Berlin) was boiled in the bottle during 1 h. Indeed, 86 μg/l were found and the pH had reached 9.5, confirming that well degassed tap water may liberate BPA at such high concentrations. It also shows, however, that the 50 μg/l assumed by the EFSA as a worst case exposure is not easily exceeded under realistic conditions.

The third experiment simulated a scenario of filling hot tea into a bottle and keeping this tea warm as long as possible, as it may occur when tea is taken onto a trip or saved for later in the day: water 2 was boiled in a pan and filled boiling-hot into the bottle. The bottle was wrapped into a towel to keep it warm. Next morning, a BPA concentration of 6 μg/l was measured. With milk type beverages the release is likely to be far lower, since the proteins buffer a neutral pH.

Experiments 4 and 5 intended to catch scenarios with heating in a pan. Simulating sterilization, the bottle was laid and covered by 1.5 l water 2, which was boiled for 5 min. This same water was assumed to be used for preparing a beverage. It contained 0.8 μg/l BPA. Considering that the inside as well as the outside of the bottle was exposed, but also that the amount of water was 7.5 times larger than in the experiments with microwave heating inside the bottle, a fair agreement with the previous result was noted.

A bottle filled with water of about 95 °C and kept standing in a pan filled with boiling water for 5 min resulted in a relatively low release of BPA (0.8 μg/l) and indicates that boiling in a pan does not result in as high BPA release as may result from microwave heating. This is not because microwaves would damage the bottle, but, firstly, because with microwave heating the BPA is released into a relatively small amount of water and, secondly, because of a more intense removal of carbon dioxide: microwaves enable real boiling of water inside the bottle; the bubbles of water vapor promote degassing which results in an according strong increase of the pH.

Figure 2 shows the effect of temperature for a 5 min heating in the microwave oven. Experiments were performed with new bottles 2 (different for each measurement), with both fresh (hard) water 3 (37°f) and the same water after boiling in a pan for 10 min and removal of the precipitated scale by filtration. The pH of the fresh water remained between 7.4 and 7.9, because only boiling efficiently removes the carbon dioxide. BPA concentrations remained correspondingly low: they increased from <0.1 to 0.6 μg/l at 95 °C. The boiled water with a pH of 9.5 released far more BPA: 2 μg/l at 50 °C and 33 μg/l at 95 °C.
Fig. 2

Fresh or boiled water 3 heated in new bottles at the temperatures indicated during 5 min

Figure 3 shows the release of BPA as a function of time during which water was boiled in bottles 2 by the help of microwaves. Again experiments were performed with fresh water 3 and with the same water previously boiled in a pan for 10 min (pH 9.5). Using fresh water, the initial pH went from initial 7.9 to 7.5 after 5 min and to 8.3 after 10 min. Accordingly, during the first 5 min of boiling, only 1.5 μg/l BPA was released, whereas it was 23 μg/l during the second 5 min. It shows that the release increases exponentially, as the proceeding stripping of carbon dioxide increases the pH.
Fig. 3

Release of BPA from new bottles in which either fresh or previously boiled water was gently boiled in a microwave oven

Using previously boiled water with a pH of 9.5, BPA was released at far higher and fairly constant rate. After 5 min (recommended for sterilization), the BPA concentration was 36 μg/l and it reached 137 μg/l after 10 min.


Distinction between migration and chemical release

It was disturbing to see many results reported in scientific literature showing BPA concentrations in the content of polycarbonate baby bottles essentially all being clearly below 1 μg/l, whereas the German consumer journal “Oeko-Test” [13] reported well 100 times higher values. “Oeko-Test” suggested this being due to the use of microwave heating, but the work of Ehlert et al. [9] confirmed that microwave heating (5 min boiling in the bottle) generally leaves the BPA contents below 1 μg/l.

The case points out the importance of distinguishing between migration of residual BPA from the polymer and release of BPA by degradation of the polymer. Data in scientific literature was obtained with media suitable for determining migration, i.e. the food simulants A or C (distilled water or 10% ethanol/distilled water). However, as shown by [12], migration of BPA from polycarbonate is low—usually much lower than release by degradation. Degradation cannot be investigated by migration testing, since the official simulants are not designed to mimic the aggressiveness of foods and beverages. In particular, there is no simulant testing aggressiveness of an alkali medium.

The experiments confirmed that polycarbonate is labile in alkali media. Upon boiling tap water, the degassing of carbon dioxide causes the pH to increase and form a far more aggressive medium than demineralized water. It is the high pH combined with a high temperature that determines the BPA concentrations in the filling of polycarbonate bottles.

The experiments reported by “Oeko-Test” were performed with tap water (of intermediate hardness), and the above experiments confirm that in this case tests with tap water yield more pertinent data for risk assessment. It was the hard water rather than the microwave heating which made the difference. However, microwave heating is indirectly responsible for the relatively high BPA concentrations, since it is the only practical way of boiling water directly inside the bottle. Heating a bottle filled with water in a pan with boiling water is slow and does not result in efficient degassing of carbon dioxide, since there are no bubbles of vapor formed.

Alkalinity of tap water increases slowly: it takes time to remove the carbon dioxide (also depending on the intensity of boiling), and initially the precipitation of carbonate hinders a rapid increase of the pH. Since boiling in a baby bottle with smooth plastic walls needs rather gentle heating to avoid bumping by superheating, it takes some 10 min for the pH to reach values approaching 9. For these reasons BPA release is not linearly depending on heating time, but increases exponentially.

The experiments did not show a dependence of the pH increase and BPA release on the hardness of water. Using harder water caused more precipitation, but precipitated carbonate is no longer pH-active. It would also be erroneous to believe that water decalcified in household installations would reduce BPA release: ion exchangers usually exchange calcium against sodium, which means not only that such decalcification leaves the relevant bicarbonate/carbonate in the water, but also that the carbonate formed upon stripping carbon dioxide is no longer precipitated. In fact, the pH increases more rapidly when the water is decalcified by the normal ion exchangers.

BPA in beverages administered by baby bottles

To estimate realistic exposure to BPA, the laboratory findings must be translated to potential practices in households. The following scenarios should be considered:
  1. 1.

    The baby bottle is sterilized by boiling water inside the bottle (by means of microwaves, as this is not practical in a pan) and, as this water is sterilized at the same time, it is used for preparing the beverage. Sterilization is often recommended to involve 5 min of heating, which in our experiments resulted in the release of 3–10 μg/l BPA. However, in household work this heating may easily be overdone and concentrations will be far higher then.

  2. 2.

    Water is boiled outside the bottle and hot water or a beverage like tea is poured into the bottle. It may even be kept warm by insulating the bottle for later use. Using water heated in a pan for 10 min, a BPA concentration of 6 μg/l was measured. There is no relevant release of BPA into milk or analogous beverages, since the pH is buffered near neutral.

  3. 3.

    Highest BPA concentrations (possibly exceeding 100 μg/l) are obtained by boiling water inside the bottle which has already been boiled before. This might occur when tea is heated again for re-sterilization, but this scenario is unlikely to regularly occur.

  4. 4.

    Drying a polycarbonate bottle with hard water or residues of washing solution at a high temperature in a dishwasher may liberate BPA on the bottle wall which is then redissolved in the beverage filled into the bottle [12]. Such release seldom exceeds 10 μl/l.

  5. 5.

    Preparing a drink according to the usual recommendations, i.e. by boiling the water in a pan, then adding milk powder and adequately cooled water to the bottle results in BPA release below 0.5 μg/l.


It is concluded that BPA concentrations in beverages administered by polycarbonate baby bottles are unlikely to exceed the 50 μg/l taken into consideration for risk evaluation by the EFSA [3].


A release of 50 μg/l BPA was considered safe by the EFSA, but not sufficiently safe by the Canadian authorities. There is an easy way to reliably keep the BPA concentration about 100 times lower and eliminate the concerns of the Canadian authorities:
  1. 1.

    Do not boil water to be used for preparing drinks in a polycarbonate bottle.

  2. 2.

    Avoid filling hot water or tea into a polycarbonate bottle (not critical for milk and similar drinks).

  3. 3.

    Rinse a polycarbonate bottle washed in a dishwasher with some water, possibly of that sterilized for preparing a beverage, before filling in a beverage.


Sterilizing baby bottles in boiling water in a pan or by boiling water inside the bottle using microwave heating does not transfer BPA into the beverage if the water used for sterilization is poured out. If some vinegar or citric acid (lemon juice) is added to the water used for boiling the bottle, the precipitation of the limescale can be avoided, but precipitation of limescale is not relevant for the release of BPA if the above rules are respected.


The results have theoretical and practical relevance. Firstly they demonstrate that release by chemical attack must be clearly distinguished from migration as a diffusion phenomenon. The simulants for migration testing are not designed for determining release by degradation and the scenarios potentially resulting in high release must be investigated case by case. Secondly, parents being uncertain about the safety of BPA released from polycarbonate bottles can strongly reduce this release by respecting simple rules.

Copyright information

© Springer-Verlag 2008