A Scientific uncertainty

The biggest causal issue in PFAS torts is a conceptually simple one of scientific uncertainty. Whilst this will be especially problematic in personal injury claims, it will also arise in respect of, for example, damage to crops or cattle, and scientific knowledge could also be a limiting factor in determining whether the presence of PFAS in the body, in soil or in water ‘counts’ as damage for the purposes of tort.^[33] Only relatively few of the thousands of chemicals in the PFAS category have been studied,^[34] and many of those studies are incomplete or inconclusive.^[35] PFAS were first discovered by accident in 1938, but far newer substances have been, and are being, developed.^[36] Two of the formerly most common PFAS (PFOA and PFOS) are no longer used, and their replacements are less well understood. Whilst PFAS share certain qualities, some argue that the different substances are distinct and should be regu172lated differently.^[37] No doubt their distinctiveness, and the ability to analogise between different substances, will be hard fought in tort.

To establish but for causation, claimants need to show both that the level and mechanism of exposure to the substance at issue are capable of causing the particular disease they are suffering from (‘general’ or ‘generic’ causation), and that their disease was indeed caused in that way (‘specific’ or ‘individual’ causation).^[38]

In Holmes v Poeton, the claimant argued that negligent workplace exposure to tricholorethylene (TCE)^[39] in the cleaning and degreasing of engineering machinery had caused his Parkinson’s disease. The Court of Appeal decision brings out some generally interesting issues around causation and disease. Stuart-Smith LJ described generic causation as ‘the hypothesis that exposure to TCE can cause (or materially contribute to the causing of) Parkinson’s disease’. He said that the evidence relied on by the judge to establish generic causation was ‘weak’,^[40] since it established only that TCE is ‘a risk factor for Parkinson’s disease and that there is a plausible mechanism based on the rodent studies for a finding that TCE may cause or materially contribute to the development of Parkinson’s disease’, concluding that ‘the evidence to prove generic causation is lacking.’^[41] In the absence of proof of generic causation, the claimant will generally not be able to establish specific causation, that is that their harm was caused by the defendant’s negligence,^[42] absent possibly some exceptional situation, such as an acute reaction to massive exposure.

We return to Holmes for what it tells us about material contributions to damage in the following subsection. First though, the scientific uncertainties identified by the Court of Appeal are likely to be a persistent feature of PFAS claims. The absence of sufficient epidemiological evidence to support a general causal link will be a feature of many substances in the PFAS category. The discipline of epidemiology studies the distribution and determinants of disease in human populations,^[43] so that, for example, a key piece of evidence in the link between smoking and lung cancer 173was the study of disease in large numbers of smokers and non-smokers.^[44] The novelty of some of the individual PFAS chemicals precludes long-term epidemiological studies; even if there may be some shorter-term effects, the studies take time. Even for more established substances, the complexity of PFAS exposures over a lifetime (PFAS are after all distinguished for their ubiquity and their persistence) can make observational studies difficult. And as suggested above, the extent to which it is legitimate to extrapolate knowledge from one PFAS chemical to another is also likely to be hard fought. And finally, when chronic exposure has small – but real, it causes some disease – effects, as is often the case for chronic chemical exposure, large sample sizes (and significant resources) are needed to detect it, especially if there are many cases of the disease in the general population. A significant ‘background incidence’ of any disease associated with PFAS creates a challenge for researchers (and claimants), since it implies other possible causes, in many cases poorly understood. For example, kidney cancer is one of the diseases associated with PFAS, but the UK National Health Service (NHS) tells us that causes of kidney cancer include smoking, obesity, high blood pressure and genetics.^[45] By contrast, for example, we know that mesothelioma is always caused by asbestos, that it is a ‘signature disease’ of asbestos.

I do not wish to be too negative. Epidemiological evidence is, and will be, available, not least because funding for epidemiological research was part of the settlement in one of the early US cases.^[46] I do, however, want to make the point that compelling epidemiological evidence on many PFAS is likely to be hard to come by.

In the absence of epidemiological evidence, the claimant may turn to evidence from animal testing. In Holmes, this also fell short. In particular, the ‘inability to extrapolate the rodent studies to the impact of TCE exposure in humans’ was fatal, ‘not least because of the great disparity in relative exposure levels’.^[47] This, again, is a commonplace problem in efforts to understand the human health impacts of novel 174chemicals. Using animal (or in vitro) experiments to establish contribution to risk of disease, or causation of disease, in humans requires contested extrapolations across species and across dosages. The most widespread human impacts are likely to arise in respect of chronic rather than acute exposure, that is low levels of exposure across many years, whilst for practical reasons, testing is often at high doses in short time periods. This is a hard task for regulators working at the population level, but in tort, even with respect to general causation, we will be asking specifically whether a particular, probably low level, chronic exposure is capable of causing a particular disease.

The scientific evidence by which general causation is established will develop. If we review the English case law on asbestos, for example, where long latency periods mean we are concerned with defendants’ breaches of duty decades earlier, we are presented with an idealised story of steadily and inexorably improving knowledge: eventually, the contestation and denial around PFAS may even fall out of view.^[48] Courts may moreover be willing to sidestep bespoke assessment of general causation if they can rely on the determination of an authoritative source – the World Health Organization, national governments, the EU – that a substance is capable of causing the relevant disease.^[49]

If general causation can be established, however, the claimant is then likely to face further difficulties in establishing specific causation. In particular, whilst epidemiological evidence can establish general causation in appropriate cases, its role in individual causation is more problematic. I suggested above that epidemiological evidence can be hard to come by. It is also often the best evidence of disease causation that we have, so if claimants cannot use it to establish causation, they face systemic barriers to recovery.

Epidemiological evidence operates at the population level and does not explain the aetiology of disease at the individual level, only likelihoods and risks. In Sienkiewicz, Lord Rodger explores the distinction between ‘claimant A, who proves on the balance of probability that a defendant probably injured him, and claimant B, who proves on the balance of probability that a defendant actually injured him’; the ‘claimant who seeks to prove his case on the balance of probability in reliance entirely on statistical evidence will inevitably fail, since he is able to do no more than 175prove on the balance of probability that the defendant probably injured him.’^[50] These comments are made in the case of Sienkiewicz v Greif,^[51] a Supreme Court mesothelioma case in which the claimants had used the decision in Fairchild to establish causation.^[52] We return to Fairchild below, but essentially it allows a claimant to succeed if they can establish that the defendant’s negligence increased their risk of contracting mesothelioma, which risk eventuated; Fairchild applies when it is impossible to prove that the defendant caused or materially contributed to the damage. The defendant employers in Sienkiewicz argued that in their case, proof was not impossible. They said that the very low level of workplace exposure relative to environmental exposure meant that it was more likely than not that they did not cause the damage (that it was caused by environmental rather than occupational exposure). Thus, Fairchild could not apply.

The Supreme Court, however, held that the impossibility of proving causation survived this statistical evidence on the dominance of environmental over workplace asbestos exposure. Epidemiological evidence in this case did not allow the defendant to escape liability. Whether epidemiological evidence could ever suffice to prove causation was not strictly decided, but the decision points towards a conclusion that ‘statistical’ evidence alone will never satisfactorily establish individual causation. Lord Dyson makes the important point that this ‘is of particular relevance in relation to causation in toxic torts. It is often the basic impossibility of proving individual causation which distinguishes toxic tort cases from ordinary personal injury cases.’^[53]

The role of epidemiology in English law is not categorically decided by Sienkiewicz: the members of the Supreme Court say slightly different things on the subject, the discussion of epidemiology is obiter,^[54] and even taking the case at its strongest, epidemiology may be part of the evidential picture on specific or individual causation. Part of the discomfort with epidemiological evidence in English law (aside of course its departure from the individuals before the court) may be its interaction with the ‘balance of probabilities’ standard of proof. The assumption is sometimes 176made that if epidemiological evidence is capable of establishing that the defendant’s negligence (or product or activity, depending on the tort) caused the damage, tort becomes a statistical and probabilistic endeavour. And so, if the defendant doubled the claimant’s risk (if there was a 51 % likelihood that it caused the damage), the causal link is necessarily taken to have been proven (it is more likely than not), and vice versa.^[55] Doubling the risk may be sufficient in some jurisdictions,^[56] but although there are some suggestive cases,^[57] it probably is not (especially post-Sienkiewicz) in English law. The key point of those who argue that courts should use epidemiological expert evidence with more confidence (and indeed of other scholars of causation)^[58] is that the balance of probabilities is not a question of numbers.^[59] The balance of probabilities is about whether the court is convinced that it is more likely than not that the defendant caused the damage.^[60]

Further, we might note that suspicion of statistics beyond the courts is long-standing, even in now classic cases (like the pivotal 19th century John Snow cholera 177studies).^[61] Epidemiological evidence may be more sophisticated than the Supreme Court in Sienkiewicz would suggest. In particular, epidemiology is not a simple matter of statistics, but uses other evidence, including, for example, biological evidence and a critical use of assumptions, in drawing causal inferences.^[62] The quality of epidemiological studies will vary, and even with excellent studies, some will be more fit to answer the questions being asked in a tort trial than others. We might imagine a careful use of expert epidemiological evidence in which highly relevant studies (for example, reflecting the claimant’s exposure pathways or pre-dispositions), showing strong correlations^[63] might convince a court that in this case, it is more likely than not that the relevant PFAS exposure caused this claimant’s damage.

In tort generally, establishing causation of disease and ill-health faces problems that are not present with respect to traumatic injury, since the causal links between injuries such as broken bones, or even acute toxic effects, are far more intuitive.^[64] Some of PFAS’ special qualities might exacerbate the challenges. The persistence of PFAS in the environment means that their impact may far outlast those who produced, used and disposed of them, and also creates challenges for gathering evidence, exacerbating the challenges discussed here.^[65] The claimant is unlikely to have detailed records of activities and exposure prior to discovery in litigation. But 178the uncertainties discussed in this section are not unique to PFAS. Our difficulties in tort are part of a much broader, shockingly poor understanding of the potential impacts of chemicals. Notwithstanding the advances of REACH’s ‘no-data, no market’ approach, the burden of establishing the need for regulation is (simplifying) basically on regulators and governments, and on civil society to convince them. When tort is used as a last resort, after the fact of injury, it reflects pervasive societal uncertainties and inequalities.

B Multiple possible causes

There are many ways in which multiple causal factors can arise to complicate a single case. I will not try to categorise and label them, largely because scholars and courts use the various labels in different ways that cannot always be helpfully mapped onto each other, and nor will I consider whether there might be more appropriate alternatives to the language of causation. Instead, in this section I work through some of the possibilities with respect to PFAS.

First, we can imagine a number of manufacturers and users of PFAS introducing PFAS into water, soil or the claimant’s body. The claimant may not be able to establish in this case that but for any one of them (but for the defendant’s negligence, activity or defective product), there would have been no damage.

The courts have been dealing with multiple polluters since at least the mid-19th century. The New Zealand Supreme Court in the recent climate tort case of Smith v Fonterra observed that: ‘As a consequence of the long, global industrial revolution, the common law had to deal with new, widespread risk and damage caused by air and water pollution and the escape of biohazards’.^[66] The orthodox position has long been that the claimant may establish causation by establishing that the defendant made a material contribution to the damage. The New Zealand Supreme Court confirms that the 19th century cases continue to apply, and rejects the defendants’ argument that the ‘material contribution’ approach does not apply if some of the multiple polluters are not before the court.^[67] This was a strike out decision, and the New Zealand Supreme Court explicitly leaves for trial issues associated with ‘newer technologies and newer harms ([greenhouse gases], rather than sewage and other water pollution)’.^[68] The magnitude of the complexities associated with climate change is even greater than for PFAS contamination cases.

179The UK Supreme Court has also recently reaffirmed the continued legitimacy and generosity of this industrial revolution case law. Financial Conduct Authority v Arch Insurance (FCA)^[69] was a case about insurance for business interruptions during COVID lockdowns. Simplifying a little, coverage was for the effects on the business of the occurrence of a notifiable disease within a specified distance of the business. Many such cases of notifiable disease (COVID-19) occurred. The problem was that these ‘local’ cases could not be said to be a but for cause of the lockdown, and the resulting effects on the business. Whilst it is evidently not a tort case, the Court used tort cases and tort scholarship to resolve this issue.^[70] The Supreme Court referred to the industrial revolution pollution cases:

... where multiple polluters discharge hazardous waste into a river ... each individual contribution is reasonably capable of being regarded as a cause of the harm that occurs, even though it was neither necessary nor sufficient to cause the harm by itself.^[71]

Whilst in some cases, the defendant’s contribution may be necessary (without it, no damage) or sufficient (without the others, damage anyway), FCA is wholly explicit – the defendant can be held to have materially contributed to the damage even if their contribution was neither necessary nor sufficient.

We can easily imagine the significance of this line of case law in PFAS contamination cases: defendants will not be allowed to point to existing pollution or to other sources of PFAS and say the contamination of the defendant’s property or person is not their doing. And whilst the old cases are public and private nuisance cases, ‘material contribution’ can also apply more broadly. In its emphatic confirmation that ‘but for’ causation does not exhaust liability, the Supreme Court in FCA uses examples beyond nuisance, including:

20 individuals who all combine to push a bus over a cliff. Assume it is shown that only, say, 13 or 14 people would have been needed to bring about that result. It could not then be said that the participation of any given individual was either necessary or sufficient to cause the destruction of the bus. Yet it seems appropriate to describe each person’s involvement as a cause of the loss. Treating the ‘but for’ test as a minimum threshold which must always be crossed if X is to be regarded as a cause of Y would again lead to the absurd conclusion that no one’s actions caused the bus to be destroyed.^[72]

180There is also long-standing case law on the application of the ‘material contribution’ test to appropriate personal injury cases. In the leading case of Bonnington Castings v Wardlaw,^[73] dust was produced in the workplace, some through the negligence of the employer and some without negligence – ‘innocent’ and ‘negligent’ dust – which combined to cause personal injury. The defendant was liable for a negligent material contribution to the plaintiff’s disease.

The disease in this case was pneumonoconiosis, a divisible injury (where every bit of exposure makes the disease worse). The extent of the defendant’s liability was not explored in Bonnington, but subsequent cases have clarified that with respect to a divisible injury, the defendant’s liability is proportionate to their contribution.^[74] I discussed the scientific uncertainty in Holmes v Poeton above. In terms of legal doctrine, the defendants had attempted to distinguish Bonnington (applied by the High Court) on the basis that Mr Holmes’ Parkinson’s disease is an indivisible disease (ie one either has the disease or does not) and that ‘material contribution’ cannot apply to a case of indivisible rather than divisible injury.

The Court of Appeal in Holmes explores Bonnington and later cases in depth. It confirms that Bonnington can indeed apply to an indivisible disease, and ‘the fact that the disease actually at stake in Bonnington was divisible does not assist’ the defendant.^[75]Bonnington is about ‘a partial cause of the whole injury, as distinct from being a cause of part of the injury’.^[76] Citing Lord Rodger,^[77] the Court of Appeal holds that Bonnington is ‘classic authority’ for material contribution as a more generous alternative to the but for test. The Court of Appeal in Holmes also cites Lord Rodger for the proposition that it is ‘hard – and settled law – that a defendant is held liable in solidum, even though all that can be shown is that he made a material, say 5 %, contribution to the claimant’s indivisible injury.’^[78] So if the injury is indivisible, the defendant carries the entirety of the damage (subject to seeking contribution from others). PFAS, ‘forever chemicals’, may well outlast some of those who produce, use or dispose of them. This distinction between a divisible or an indivisible injury becomes decisive as to whether claimants or any remaining defendants bear the risk of absent wrongdoers.

181PFAS defendants may well have an incentive to press the ‘immateriality’ of their own contribution. The threshold that must be reached for a contribution to be ‘material’ has never been precisely pinned down. In Bonnington itself, the Court took a clearly generous approach, only excluding de minimis contribution;^[79] similarly in the quotation from Lord Rodger above, we would certainly not be excluding anything above a 5 % contribution.

In terms of the factual situation, the High Court had found that Mr Holmes’ Parkinson’s disease was caused not either by genetic or by workplace exposure, but that there is a ‘complex interplay between genetic and environmental factors and that both factors work together to produce the condition’.^[80] ‘Material contribution’ can apply if some of the contributing factors are non-negligent, or contingent in some way. So genetic pre-dispositions of the claimant to a cancer associated with PFAS exposure, or other sources of exposure, would not in themselves prevent the application of the ‘material contribution’ test (although of course it may not be satisfied). There is, for example, evidence that certain PFAS can initiate prostate cancer in individuals where a first-degree family member has prostate cancer.^[81]

In these cases, we see a defendant whose negligence (or activity or defective product) can be said to have contributed to the harm. A quite different case arises where it is impossible for the claimant to prove that the defendant’s negligence either caused or materially contributed to the harm. In the cases above, the defendant was causally involved in a way that makes it appropriate to allow causation: here the claimant cannot establish this, since something else entirely may have caused the damage. This is certainly plausible for PFAS. It might be, for example, that genetic susceptibility and PFAS operate independently, so that one or the other caused the disease; or it might be that a low level of PFAS exposure causes (or may cause) the disease, so that any one of a number of exposures could have caused the disease.

This brings us to the famous/notorious Fairchild exception in English law. In Fairchild, a number of claimants, all of whom died of mesothelioma, were negligently exposed to asbestos by more than one employer. Mesothelioma is a signature disease, caused only by asbestos. The parties accepted that the greater the quantity of asbestos dust and fibres a person inhales, the greater the risk of developing mesothelioma; at the time, the scientific evidence was that the disease could have been caused by many asbestos fibres, just a few fibres or a single fibre,^[82] and the way in which the impact of different fibres might interact was not known. The claimants 182could not therefore prove on the balance of probabilities that any single defendant or combination of defendants had either caused or materially contributed to the damage. The claimants lost for that reason before the Court of Appeal.

The House of Lords held, however, that a sufficient causal link exists if the claimant can establish that the defendant materially contributed to the risk faced by the claimant, if that risk subsequently materialises. Fairchild applies only in cases where it is impossible, because of the limits of science, to prove causation. Subsequent case law has added both refinements and ‘wrinkles’ to the Fairchild exception. First, it is clear that Fairchild’s innovation is in causation, not in the definition of damage: the damage is the mesothelioma, not the risk.^[83] Secondly, Fairchild can apply when there is a single tortious possibility, alongside other non-tortious potential causes,^[84] including if some of the asbestos exposure is down to the contributory negligence of the claimant.^[85] Further, as with material contribution to damage, the ‘material’ contribution to risk by the defendant does not have to be significant, just more than de minimis.^[86] Thirdly, in an unorthodox approach to compensation for an indivisible injury, the Supreme Court in Barker held that if Fairchild applies, defendants are responsible for damages proportionate to the risk imposed by the tort (rather than full damages).^[87] That has been overturned by legislation in respect of mesothelioma,^[88] but continues to apply in respect of other diseases, and explains why the claimant in Holmes had not claimed under Fairchild.^[89] Again, given the persistence of PFAS in the environment, and the likely latency of disease, some potential defendants are very likely to be missing by the time of a successful tort claim; proportionate liability places that risk on the claimant rather than the defendant.

The application of Fairchild, even if environmental exposure to asbestos is a possible alternative to the defendant’s negligence, is particularly pertinent for PFAS, given the ubiquity of PFAS in the environment. PFAS are, however, even more complicated than the mesothelioma at stake in Sienkiewicz, where this proposition was established.^[90] All of the possible causes or agents of the mesothelioma were asbestos 183exposure, whilst the bodily harms at stake with respect to PFAS (kidney or testicular cancer, immune system problems, etc) have many possible causes other than PFAS. If these factors can be shown to interact, the ‘material contribution to damage’ line of case law may be an option; if they are, or could be, independent, that is not possible.

Whether Fairchild could be applied to the multiple possible causes of disease associated with PFAS is open to question. The ‘single agent’ (all the potential agents of the damage are the same – asbestos dust) notion was understood to be an important (albeit pragmatic rather than principled) limitation on the application of Fairchild in its early years. The main confounding factor that has arisen in occupational disease cases has been the claimant’s smoking. Lord Hoffmann in Barker says explicitly that:

I do not think that the exception applies when the claimant suffers lung cancer which may have been caused by exposure to asbestos or some other carcinogenic matter, but may also have been caused by smoking and it cannot be proved which is more likely to have been the causative agent.^[91]

In Heneghan v Manchester Dry Docks Ltd,^[92] the possible causes of the claimant’s lung cancer included not only the negligent exposure to asbestos by the six defendant employers and some other employers not before the court, but also the claimant’s smoking. The Court of Appeal applied Fairchild without engaging with the single agent criterion. The defendants, however, had conceded liability, that asbestos had caused the lung cancer, and Fairchild was used to decide which defendants should pay for what. But although the smoking was not therefore discussed by the Court of Appeal, the upshot of Heneghan was to apply Fairchild to a case where there was more than one potential agent of the disease.

If that is followed generally, it could of course be helpful for claimants in PFAS cases. It is however challenging for tort law. As Sarah Green observes, employers and others create risks all of the time, whenever they do anything that is capable of leading to an untoward outcome, and so Heneghan is potentially an extension without limit.^[93] She argues that the ‘single agent’ constraint on Fairchild ensures that the 184very risk to which the defendant exposed the claimant materialises.^[94] We might add that it takes us beyond any sense that Fairchild was designed to avoid a number of negligent employers effectively hiding behind each other’s wrongdoing to avoid liability,^[95] although the same applies of course to the environmental exposure in Sienkiewicz or the contributory negligence in Barker. It is at least open to question whether Heneghan would be applied to the diseases currently associated with PFAS, given the factual agreement that the Court of Appeal was dealing with. Perhaps the earlier case of Novartis v Cookson^[96] provides some middle ground. The claimant alleged that his bladder cancer had been caused by negligent workplace exposure to aromatic amines. The point was not taken by the parties, but the Court of Appeal suggests that Fairchild can apply in this case because the amines in cigarette smoke ‘operate in the same way’ as the amines in occupational exposure.^[97]

Whether seeking to apply Heneghan or the more measured flexibility in Novartis, we might bear in mind the Supreme Court’s cautious approach to Fairchild: Lord Hoffmann, who gave an important speech in Fairchild, has, for example, suggested that it teaches a ‘lesson of caution ... in relation to future invitations to depart from conventional principles of causation.’^[98] Fairchild is important, however, not only for the scope of the exception it creates, but also for indicating that the courts recognise that sometimes a strict application of the rules on causation would lead to injustice. When courts intervene in evidential or conceptual aspects of causation out of fairness for claimants, there may be further complexities and disagreement around the scope of responsibility for damages: Fairchild, Barker and the Compensation Act 2003 are evidence of that.^[99] These questions in some cases will simply push the need for courts to grapple with uncertainty elsewhere in the analysis, although the English courts seem largely to have taken a pragmatic approach to allocating responsibility between defendants.^[100]