Concerns about the safety of chemicals have prompted the chemical industry to undertake a major international testing programme. But these studies may do little to allay fears about endocrine-disrupting effects, which seem to exist in a parallel universe outside the currently accepted norms of toxicology.
Endocrine disruptors work by interfering with hormone signalling systems and can produce effects at very small doses - either by activating hormone receptors or by interfering with normal metabolism and binding of hormones. Endocrine theory predicts that they may have effects at low doses, but that these may disappear at higher concentrations.
Such low-dose effects exist but their significance is uncertain, a high-powered US National Institute of Health (NIH) panel concluded in 2001.1 Attempts to reproduce such findings have often failed, and sufficient doubt exists in the whole area to prevent them having regulatory repercussions. However, this uneasy situation cannot last indefinitely. Will low-dose effects turn out to be a mirage - or will they succumb to reason and understanding?
No such resolution appears close at hand. Low-dose effects are a paradox. They are effects which occur at chemical doses below the level at which effects have previously been observed. And so they bring the conventional toxicological concept of a 'no observed effect level' into question and strike at the heart of toxicology.
Can any chemical be shown to be safe in such a logical minefield? The billion-dollar oestrogenic plastics intermediate bisphenol A (BPA) has been first to run the gauntlet. It has been in the firing line ever since independent US researchers, led by Professor Fred vom Saal, showed in 1997 that very low doses fed to pregnant mice resulted in male offspring with enlarged prostate glands (ENDS Report 266, pp 8-9 ).
The chemical industry countered the study by repeating it twice on a larger scale and under the rigours of "good laboratory practice" required for regulatory chemical testing. They found no effects.
Several assessments have since taken a lead from these industry-sponsored studies. The conclusion of the NIH study was that low-dose effects of BPA had "not been conclusively established as a general or reproducible finding." But its unease on the issue was enough for it to recommend revisiting chemical testing procedures to consider changing endpoints, dose selection, animal strains and other important specifications (ENDS Report 316, pp 3-4 ).
Professor vom Saal has continued to pursue the low-dose hypothesis - that very small amounts of endocrine-disrupting compounds cause effects which are often reversed or undetectable at higher doses. At a scientific meeting hosted by the Japanese Environment Agency last November he said that no fewer than 30 recent studies had reported low-dose effects from BPA alone.
"There has been a profound paradigm shift in toxicology," he told ENDS. "Since the low-dose panel report, so many people are studying low-dose effects."
Vom Saal's work suggests that foetal exposure to low doses of BPA may cause enlarged prostate glands in male mice as well as early puberty in females and an increase in size in both sexes. New studies point to a gamut of other effects, many of which suggest that exposure to BPA or other environmental oestrogens may be linked to modern human ills such as prostate and breast cancer, early puberty, behavioural problems in children and falling male fertility.
Food can linings, baby feeding bottles made of polycarbonate, coated wine storage vats and plastic sealants used in dentistry are the most widely recognised sources of human exposure to BPA.
In April 2001, the Food Standards Agency completed a study of BPA in canned food (ENDS Report 212, pp 8-9 ). It found that most foods contained levels of around 30-40 micrograms per kilogram. For average consumers, this translates into an exposure of 0.36-0.38 µg/kg bodyweight per day for adults and 0.83-0.87µg/kg/day for infants.
The Agency concluded that these doses were well below the tolerable daily intake of 50µg/kg/day and therefore safe. Professor vom Saal reported effects on the mouse prostate at only 2µg/kg/day, leaving little margin of safety for infants or children exposed in the womb. The findings have led WWF to call for a precautionary ban on BPA in sensitive applications like can linings and baby feeding bottles (ENDS Report 314 pp 10-11 ).
A recent study by German scientists found that BPA is present not only in human blood but also in foetal blood and the placenta.2 For toxicologists the finding is remarkable because studies have shown that the vast majority of BPA fed to animals by mouth is rapidly metabolised in the liver and then excreted by the kidneys.
Nevertheless, the German study found that an average of 3.1µg/l of BPA was present in blood plasma from a sample of pregnant women in Berlin and 2.3µg/l in their foetuses. The placenta also contained BPA levels averaging 12.7µg/kg.
The scientists concluded that BPA exposures were "within a range typical of those used in recent animal studies [that] were shown to be toxic to the reproductive organs of male and female offspring."
Bisphenol A 'safe for food use'
In response to the controversy over BPA, the EU Scientific Committee for Food (SCF) reviewed the toxicological evidence in May 2002.3 It recommended a temporary tolerable daily intake of 10µg/kg/day - a factor of five lower than that previously used by the Food Standards Agency - based on a weight of evidence approach.
The figure was derived from industry-sponsored toxicity studies which found that animals fed BPA doses of 50mg/kg/day showed weight loss and produced smaller offspring - gross effects and at much higher doses than those found in many studies by independent researchers.
The SCF considered that the worst case human and infant BPA intakes were 0.48µg/kg/day and 1.6µg/kg/day respectively - still comfortably within tolerable levels.
On low-dose effects, the SCF concluded that "the implications of these conflicting observations for human risk assessment are difficult to assess at present. The findings...are considered insufficiently robust for derivation of a tolerable daily intake, but remain as current uncertainties."
The SCF is not the only scientific body to feel that the evidence for low-dose effects is too flimsy a foundation for regulation.
Bisphenol A is currently undergoing health and environmental risk assessments led by the UK under EU legislation on existing substances. At an evaluation meeting in 2001, technical experts decided that there was generally no concern over consumers' exposure to the compound. However, despite assurances by the assessment's authors - the Environment Agency and Health and Safety Executive - the experts did not feel comfortable in ignoring reported low-dose effects on development. The final compromise was to appoint a steering committee to commission further studies.
According to David Thomas of the BPA Industry Group, part of the European chemical industry body CEFIC, the steering committee has taken the "unusual step" of commissioning three-generation exposure studies on mice to complement data already available for rats. The contract has been given to a testing laboratory in the US, and the timetable is the end of 2004, although this "was not cast in stone".
The risk assessment also highlighted potential low-dose effects of BPA on fish and invertebrates. While traditional studies would suggest a predicted no effect concentration of around 16µg/l, there was a suggestion of effects on sperm production in minnows at only 1µg/l and of reproductive effects at even lower levels in snails.
Technical experts decided that further work was needed to investigate both effects. Industry is funding studies aimed at confirming or refuting these findings before the end of 2004 when the risk assessment will be revised.
Low-dose studies have therefore had a very significant impact on the BPA risk assessment process, despite being regarded as insufficiently reliable to be taken into account in the regulatory process so far.
The range of endocrine-disrupting compounds for which low-dose effects have been reported is growing wider. The list now includes the herbicide atrazine, surfactant intermediates nonyl and octyl phenols, the pesticide hexachlorobenzene, phytoestrogens, and herbicide mixtures that include 2-4D, dicamba and mecoprop.
Studies by Tyrone Hayes from the University of California at Berkeley and colleagues have shown that low doses of atrazine feminise developing male frogs of two species - the African clawed frog and the North American leopard frog. The work has been taken by many to be a possible explanation for widespread declines in amphibian species.
Atrazine is one of the most frequently used pesticides in the US and levels in surface water commonly reach up to 25µg/l. It is also recognised to be an endocrine-disrupting compound because of its effects on sex hormone synthesis.
Dr Hayes reports that retarded sexual development and intersexuality seen in atrazine-treated laboratory animals can be found in wild frogs in many locations across the US. The study also suggests that atrazine is more potent at a concentration of 0.1µg/l than at 25µg/l - suggesting a U-shaped dose response curve believed to be typical of many endocrine disruptors.
The health and environmental impacts of atrazine are currently being reassessed by the Environment Protection Agency. Syngenta, a major manufacturer, has funded many studies to support atrazine - including some to counter Dr Hayes' findings through the Atrazine Endocrine Ecological Risk Assessment Panel EcoRisk.
One such study repeated Dr Hayes' work on the African frog but with more animals and more dose levels. It found that atrazine had no effect on development at exposures between 1µg/l and 25µg/l. Researchers sponsored by Syngenta have also claimed that studies on native African frog populations have failed to uncover any evidence of intersexuality despite exposure to atrazine.
The failure of large, industry-funded studies to repeat low-dose findings made by independent scientists is a central issue - and it is not just Professor vom Saal's or Dr Hayes' work which has proved not to be repeatable.
An early study by Professor Richard Sharpe which showed that low doses of phthalates and nonyl phenol affected fertility in male rats could not be repeated (ENDS Report 268, pp 26-29 ). And very recently, John Ashby of Syngenta reported that he had been unable to reproduce low-dose effects of BPA reported by Japanese studies under Motoharu Sakaue.
The Japanese studies suggested that low doses of BPA fed to young male rats reduced sperm production, and that the results had been repeated in the same laboratory. But at a conference in Yokohama last November, Dr Ashby revealed that four attempts to reproduce the findings had failed - using the same strain of mice and three different diet formulations.
One difference that did emerge between the results was that the rats in the different laboratories produced different amounts of sperm, even in untreated control animals. Similarly, when industry scientists attempted to repeat Professor vom Saal's work on mouse prostate glands the animals in their experiments were larger than those in the original tests.
It is a point that Dr Ashby made in his presentation at Yokohama. "It would be profitable to understand and then control the several possible sources of control-variability for the major endpoints studied in endocrine disruption," he recommended. "At present, this endeavour is being conducted ad hoc within a series of non-replicated and often inconclusive chemical toxicity evaluations."
Professor vom Saal has a different perspective of the lack of repeatability. He maintains that there are difficulties in carrying out sensitive low-dose studies under the kind of standard protocols which industry good laboratory practice demands.
The sheer size of industry studies may reduce rather than increase the reliability of the results, he maintains: "If you have multiple investigators, you need to look at the impact of that on variability of the results. It means that you increase variation in the results, making it harder to find effects. We use smaller studies evaluated by a single highly trained person. But if you are using 8,000 animals it is difficult to do that with just one person."
Another factor likely to increase the variation and noise in the results is using more than one male to father the second generation in multigenerational studies, he maintains. Other researchers have argued against the practice on the grounds that it produces false positive results.
Professor vom Saal has been highly critical of industry scientists' attempts to reproduce his own work, and maintains that the mice used in their studies were not only larger but obese. This meant that the prostate glands of exposed animals were already "maximally enlarged" and could not respond to BPA treatment, he told ENDS.
"Instead of reporting the interesting finding that there were conditions that had made these animals obese, they specifically lied about it. That's what you get when you are trying to protect a product rather than being interested in the science."
Despite the acrimony of the dispute, there is clearly some common ground. Both industry and independent scientists recognise that uncontrolled factors in the experiments are affecting the results. True repeatability will not be achievable until these are uncovered and standardised.
Professor Wade Welshons, a colleague of Professor vom Saal at the University of Missouri, has been studying animal diet as one such potential factor. Oestrogenic compounds produced by plants and fungi are recognised as significant influences.
In in vitro endocrine studies it is well recognised that contaminants in laboratory reagents can be oestrogenic and disrupt the experiments, he says. "There are standard methods for detecting these kind of effects in in vitro studies but not for experiments involving live animals."
John Sumpter of Brunel University, author of many ground-breaking reports on endocrine effects on fish, commented: "The message on low-dose effects is that there is no coherent message. We are no further forward than we were a few years ago."
"Various groups are reporting low-dose effects, some allied to Fred [vom Saal], but not all. Ethinyl oestradiol [the synthetic oestrogen used in birth control pills] has very low-dose effects on fish, everyone agrees with that. But there is uncertainty over weak oestrogens like phthalates, parabens, alkyl phenols and so on - but effect concentrations are getting lower."
He thought that the work by vom Saal and Sakaue on BPA were the crucial issue. "I still don't think there are very many well designed studies that would convince one that low-dose effects occur. Many studies show effects but they are not convincing - do they find effects or is it noise? Are they repeatable?"
"When very large studies are conducted - usually by industry because academics can't afford to do them - they don't show reproducible effects."
But he dismissed the idea that industry generally is biased: "When I hear John Ashby I am persuaded. He is not saying low-dose effects don't occur, he is saying we can't reproduce them in large studies. His opinion is that if these effects are so specific to one strain and one lab, is it something we should regulate on? I think probably no.
"One can always find a reason why experiments didn't repeat in another lab - that can go on for eternity. In the end you have to take a weight of evidence view.
"But low-dose data have shown that we have little understanding of many issues and that the range of endpoints considered in toxicological studies is inadequate.
"If a chemical is causing, say, testicular cancer in 1% of animals - that would be highly significant. What are the chances of detecting such an effect? No chance, because you would need 500 animals to do it.
"As [US researchers] have pointed out, our tests are designed to reduce the number of animals used, but they are losing statistical power. By repeating tests on animals in the medium dose range we are finding effects which keep drawing down no observed effect levels - and therefore safe doses. So out of these arguments on experimental design is coming quite a lot of good."
One area where more consensus is emerging is in interactive effects. This area of research got off to a false start in 1997 when a paper reporting enormous synergistic effects between oestrogens was withdrawn when the findings proved not to be reproducible (ENDS Report 268, pp 26-29 ).
But important advances have been made by researchers led by Andreas Kortenkamp of the University of London who have introduced pharmacological concepts on how additive effects might be calculated.4Dr Kortenkamp used a simple cell assay - the yeast oestrogen screen (YES) - which responds to oestrogens by producing a red colour of varying intensity. He tested mixtures of xenoestrogenic compounds such as phytoestrogens, polychlorinated biphenyls, bisphenol A and benzophenone.
Using a carefully constructed mixture of eight oestrogenic compounds, he was able to show that the most appropriate model for calculating combined oestrogenic effects in this instance was the concentration addition method. This applies when the components each act in a similar way and the components then contribute to the overall effect in proportion to their concentration.
By contrast, methods which did not apply included independent action - where components act in different ways and do not produce additive results - and effect summation, where the total effect of the mixture is simply the sum of the observable effects (rather than concentrations) of its parts.
A feature of the concentration addition method, Dr Kortenkamp notes, is that additive effects occur between the compounds present, even when each is present at a concentration below its no-effect level. Merely summing effects in such circumstances would greatly underestimate the effect of a mixture.
"The assertion that individual oestrogenic chemicals pose no harm because they are present at low, ineffective levels in humans or in wildlife may be irrelevant when dealing with mixed exposures," Dr Kortenkamp concludes. "Our results highlight the limitations of the traditional focus on the effects of single agents. Hazard assessments that ignore the possibility of joint action of oestrogenic chemicals will almost certainly lead to significant underestimations of risk."
Although there are many reasons for believing that humans and animals will behave in more complex ways than this experimental system, the results do suggest that the newly emerging discipline of endocrine toxicology will not be able to ignore cocktail effects in the way that mainstream toxicologists have largely done for many years.
In a chemical environment that appears to consist of many weak or very weak oestrogenic compounds, these results may prove to be extremely significant. Everyday exposures to chemicals such as BPA in baby milk or canned foods, traces of nonyl phenol in food from other plastics, pesticide residues, and phthalates or parabens in cosmetics could add up to a doses that cause harmful effects in infants and produce life-long impacts on the hormonal sensitivity of organs like the breast or prostate.
Kortenkamp's conclusions have been reinforced by studies on fish commissioned by the Environment Agency. The latest research report from Professor Sumpter of Brunel University and Dr Charles Tyler from Exeter notes that laboratory experiments using mixtures of oestrogens suggested additive effects, even at levels below no-effect concentrations.
The studies used mixtures of natural oestrogen, the synthetic oestrogen ethinyl oestradiol and alkyl phenols. All three are found in rivers due to sewage discharges, the use of the contraceptive pill and as breakdown products of alkyl phenol ethoxylate detergents.
WWF is the only environmental group to have engaged seriously in the endocrine debate. It is also the only body to have come forward with any suggestions as to how endocrine disruptors might be regulated to help reduce exposures.
In a recent discussion paper, the group proposes a classification system for endocrine disruptors and incorporating them in the European Commission's proposed REACH authorisation scheme for regulating chemicals (ENDS Report 313, pp 26-30 ).
The classification scheme would be akin to those for carcinogens and reproductive toxicants. Endocrine disruptors in the top two tiers - those known to cause adverse effects in whole organisms through endocrine-mediated effects, or those "strongly suspected" of doing so - should be subject to authorisation, WWF suggests.
The proposal would herald a more precautionary approach to the use of such compounds, the group believes. Chemicals of such high concern should be authorised for use only if there is no predicted human or environmental exposure, an overriding need for the compound and no safer alternatives, and if measures are taken to reduce risks.
A new definition of "adverse effects" comes along with these proposals. The group has been dissatisfied with the conservative approach adopted by industry and some scientific groups to date. It wants to see all changes in endpoints likely to be under endocrine control brought into the definition.
The definition would include such effects as increases in the weight of the uterus in female animals - a common response to oestrogens not currently considered adverse. It would also include population-level effects like vitellogenesis - the production of female egg proteins by male fish - which has been shown to be virtually ubiquitous in most rivers receiving sewage effluents.