FEDERAL COURT OF AUSTRALIA

McNickle v Huntsman Chemical Company Australia Pty Ltd (Initial Trial) [2024] FCA 807

ORDERS

THE COURT ORDERS THAT:

1.    Pursuant to s 33ZF of the Federal Court of Australia Act 1976 (Cth) (FCA Act) (and notwithstanding that no application has been made by the representative party in accordance with s 33K(1) of the FCA Act) leave be granted for the group membership to be amended nunc pro tunc so that it is defined as it is in the last version of the statement of claim filed prior to the fixing of the date for opt out (being the second further amended statement of claim filed on 19 October 2020).

2.    Pursuant to ss 33ZF and 37P(2) of the FCA Act, the following question, common to the claim of the applicant and all group members (Central Common Question) be answered separately and before any other question or issue in the proceeding:

3.    The Central Common Question be answered as follows:

4.    The proceeding be dismissed upon finalisation of any issues relating to costs.

5.    Any further evidence as to costs (as contemplated in the reasons (at [1202])) and any submissions as to costs be filed by 12 noon on 30 July 2024.

6.    For the purposes of s 33ZB of the FCA Act, Orders 1 to 3 above affect (and hence bind) all current group members in the class action (being those named in the second further amended statement of claim who did not opt out).

7.    The proceeding be adjourned for the parties to be heard on the appropriate costs orders in conformity with these reasons at 10:15am on 31 July 2024.

AND THE COURT NOTES THAT:

8.    The leave granted after the date of opt out had passed to amend the statement of claim so as to file the third and fourth statement of claim did not include leave being granted on application made by the representative party to alter the description of the group pursuant to s 33K(1) of the FCA Act.

Note:    Entry of orders is dealt with in Rule 39.32 of the Federal Court Rules 2011.

REASONS FOR JUDGMENT

LEE J:

A    INTRODUCTION

1    Mr Kelvin McNickle lived what might be thought by a young lad to be an Australian idyll. Born in September 1982, he grew up on a rural property on the beautiful Mid North Coast of New South Wales. His father operated a vegetation management business, and, during school holidays, Mr McNickle assisted his dad clearing vegetation.

2    But this was no passing stage. After completing school, Mr McNickle began working full-time for his father. No doubt it was hard work. Almost every working day, Mr McNickle and his father sprayed a weedkiller called “Roundup” on weeds and other unwanted growth using a knapsack sprayer or a drenching unit. As might be expected, drips or leaks of the Roundup soaked through his clothes and onto his skin. Mist or spray would go all over the place and get into his eyes and onto his face and fumes were inhaled.

3    Some years later, after marrying, Mr and Mrs McNickle left Coffs Harbour for Darwin. Mr McNickle is an apparently devout man. The Apostle Paul said: “let him do hard work, doing with his hands what is good work” (Ephesians 4:28). Apparently motivated by the perceived moral value of good and hard manual work, Mr McNickle volunteered to maintain the lawns and gardens of the Kingdom Hall of Jehovah’s Witnesses. His evidence is that in doing so he used Roundup.

4    In May 2018, Mr McNickle was diagnosed with a type of cancer known as non-Hodgkin lymphoma (NHL). Just over a year later, he was informed by a haematologist that he was in ongoing remission. Unfortunately, about six weeks prior to the commencement of the initial trial, his NHL recurred. As a result, Mr McNickle was unable to attend the hearing.

5    I have introduced Mr McNickle, who is the lead applicant in this class action. But this judgment is not, at least directly, about him. This is not the occasion to deal with questions of individual causation.

6    Rather, when this initial trial was fixed, it was contemplated I would determine a set of common questions relating to Mr McNickle’s contention that during a period from July 1976 until an end date defined at the time I made the orders (Relevant Period), glyphosate, glyphosate-based formulations (GBFs), and Roundup Herbicide and Roundup Biactive (Roundup Products) (which have glyphosate as a component), were carcinogenic to humans. The respondents (collectively, Monsanto) reject this contention completely.

7    The common questions were identified in orders made by consent following a case management hearing in April 2023. I set out the terms of those questions in Section J below and why I do not propose to answer them in the form initially proposed.

8    As it happened, issues were refined during the initial trial. In the end, for reasons that will be explained, the determinative question is whether Mr McNickle has discharged his legal onus of proving on the evidence adduced that the use of and/or exposure to Roundup Products can increase an individual’s risk of developing NHL or cause an individual to develop NHL. I will call this, unimaginatively, the central issue.

9    There is no need for me to explain the issues in this proceeding any further. This is a case which has been mired in interlocutory skirmishes and, save for matters which I will mention shortly, I do not propose to rehearse that background here. There have been several judgments of the Court which identify the central issue, the utility of determining it initially, and set out what might be described as the superstructure of the proceeding: see McNickle v Huntsman Chemical Company Australia Pty Ltd (Common Questions) [2023] FCA 662; McNickle v Huntsman Chemical Company Australia Pty Ltd (Additional Expert Conclave) [2022] FCA 1596; McNickle v Huntsman Chemical Company Australia Pty Ltd (Hearing Vacation) [2022] FCA 133; McNickle v Huntsman Chemical Company Australia Pty Ltd (Assessors) [2021] FCA 780; (2021) 285 FCR 244; McNickle v Huntsman Chemical Company Australia Pty Ltd (Expert Evidence) [2021] FCA 370.

10    I will return to the common questions and the utility of answering them below, but at the outset, it is worth saying something about the proper approach to fact finding and the legal method.

B    PROPER APPROACH TO FACT FINDING AND THE LEGAL METHOD

B.1    Relevant Procedural Background and the Role of the Assessor

11    Enough has been said to indicate that this case presented significant challenges in fact finding. In the light of the central issue, it is an exercise in considerable understatement to describe the expert opinion evidence and underlying scientific material as dense, complex, and voluminous.

12    As I recorded in McNickle v Huntsman Chemical Company Australia Pty Ltd (Expert Evidence) (at [4]), when this matter first came before me for case management, I indicated my preliminary view was that the Court should adopt the process of appointing referees in a number of scientific disciplines to inquire into, and to report upon, questions concerning the characteristics of the Roundup Products and their alleged carcinogenic effects.

13    The parties vigorously opposed this proposed course.

14    An array of arguments were marshalled and then deployed to oppose referees, including that to proceed along these lines would be constitutionally invalid (an argument wholly devoid of merit and decisively rejected in CPB Contractors Pty Limited v Celsus Pty Limited (formerly known as SA Health Partnership Nominees Pty Ltd) (No 2) [2018] FCA 2112; (2018) 268 FCR 590).

15    With that said, there was one contention made by Mr McNickle that persuaded me (reluctantly) to adopt the course of allowing expert evidence to be deployed in a more traditional way. Mr McNickle’s argument was that prevailing scientific opinion concerning the alleged carcinogenic effects of Roundup had been skewed or somehow manipulated by Monsanto. I concluded that the risk of referees assessing material said to be infected by this allegation of scientific manipulation created a prospect of complicating the inquiry process. Further, the allegation was one of some seriousness and is of a character that meant it should be resolved in open court.

16    Shortly after rejecting the process of references, and faced with what I knew was going to be a raft of highly complex opinion evidence, I noted in McNickle v Huntsman Chemical Company Australia Pty Ltd (Assessors) (at 246 [7]) that almost 100 years ago, Dean Roscoe Pound in his famous work The Spirit of the Common Law (Marshall Jones Company, 1921) (at 214–215) suggested that steps be taken to equip courts with research staffs comparable to those employed by administrative agencies in order to provide judges with specialised investigations, information, and advice. Yet, despite what some have described as centuries of legal history dotted with proof to the contrary, the traditional common law confidence that courts can handle any dispute without assistance has long persisted: see Beuscher, J H “The Use of Experts by the Courts” (1941) 54(7) Harvard Law Review 1105.

17    I also observed that in saying this, there have been attempts by courts to provide, in different ways, expert aid to determination. There are three obvious examples: (1) special juries; (2) referees (properly seen as a version of a special jury); and (3) assessors.

18    Having determined that there was to be no form of special jury, I went on to explain (at 247–253 [11]–[38]) the long history of appointment of assessors tracing back to the sixteenth century; the basis of the Court’s power to order an assessor; and the scope of the role of an assessor. In this last regard, I referred to the judgment of J Forrest J in Matthews v SPI Electricity Pty Ltd (No 32) [2013] VSC 630 and relevantly noted (at 252 [37]) that the role of any assessor would be to assist the judge, but any decision would be that of the tribunal of fact, being the judge alone. Conscious of the need to avoid any procedural unfairness and define the role of the assessor precisely, I explained, with the consent of the parties, that if an assessor was appointed:

(1)    the assessor would sit with me during the concurrent evidence sessions, and could question the experts (or counsel) but such questioning would be limited to clarification of the evidence: that is, where the evidence given is perceived by the assessor or me to be ambiguous, unclear or incomplete;

(2)    I would consult with the assessor if I found a point of evidence unclear;

(3)    I would consult with the assessor in chambers on matters raised by the experts in their oral evidence and in their individual and joint reports, including advice as to any questions the assessor thinks should be asked;

(4)    I would seek the guidance of the assessor on scientific and technical matters upon which I lack the requisite knowledge to understand without qualified assistance;

(5)    if the assessor raised a theory or opinion that had not previously been identified by the parties, I would disclose this fact and discuss it with counsel; and

(6)    the assessor would, from time to time, provide me with advice on matters over which there is dispute between the experts but that such advice is not binding and the determination of a particular issue rests with me.

19    After receiving suggestions from the parties, on 16 August 2021, I made an order pursuant to s 23 and/or s 33ZF of the Federal Court of Australia Act 1976 (Cth) (FCA Act) that Professor Sir John Stewart Savill BA, MBChB, PhD, FRCP, FRCPE, FRCSEd(Hon), FRCPCH(Hon), FASN, FRSE, FMedSci, FAHMS, FRS be appointed as Assessor. In dealing with the Assessor, care has been taken not to depart from the six points identified above and, more generally, to ensure the findings I will make are mine and mine alone, notwithstanding the assistance received from the Assessor in understanding the complex evidence adduced in this case. For completeness, in my dealings with the Assessor, I record he did not raise a theory or opinion privately that had not previously been identified by the parties through their retained experts (see [18(5)] above).

20    Further, and again with the express consent of the parties, when I had completed an initial draft of these reasons without the assistance of the Assessor, I provided a copy of the draft reasons to the Assessor in order to ensure, as it was explained to the parties, that I had not committed any scientific “solecisms” caused by any fundamental misapprehension of the scientific evidence. I only adopted this course after obtaining confirmation by both parties that they regarded any communication between my chambers and the Assessor as being equivalent to an “in chambers” communication and hence wholly confidential, and that they disavowed seeking a copy of any communication of any type between my chambers and the Assessor, including on any appeal.

21    I should record that these reasons would have been significantly impoverished, and the delivery of judgment delayed, without the assistance of Sir John. It would have been much more difficult to identify and then understand the nuances of the opinion evidence without an Assessor, particularly one like Sir John with his happy facility for explaining complex scientific concepts in a way that can be comprehended by an interested layman. On behalf of myself and the Court, I express my appreciation to Sir John for his great assistance.

22    But this was not all the assistance I received.

23    As is well-known, Australia’s version of concurrent evidence typically involves two interrelated processes. First, there is a pre-trial joint expert conferral or conclave phase during which the parties’ experts meet to clarify the areas of agreement and/or disagreement to produce a joint report. The second part of the process is the giving of concurrent evidence.

24    In the present case, the first of these processes became particularly important and I appointed an experienced facilitator and independent barrister, Mr Edward Cowpe, who maintained the integrity of the conclave process and generally ensured the conclave process proceeded effectively and the various joint reports emerging from any conclave were expressed clearly, admissibly, and in a form that would best assist the Court. I am also grateful for his assistance.

B.2    Expert Assistance Generally

25    One of the reasons I was reluctant to abandon a reference and proceed in the time-honoured way in this case is because of my concerns in navigating potentially polarised expert views. In McNickle v Huntsman Chemical Company Australia Pty Ltd (Expert Evidence), I observed (at [1]) that controversy as to how opinion evidence is often deployed by litigants is not new, and some best placed to judge have expressed their criticism with some force. As Lord Woolf MR observed in the report, Access to Justice, Interim Report to the Lord Chancellor on the Civil Justice System in England and Wales (HMSO, London, 1995) (at 183):

Expert witnesses used to be genuinely independent experts. Men of outstanding eminence in their field. Today they are, in practice, hired guns. There is a new breed of litigation hangers-on, whose main expertise is to craft reports which will conceal anything that might be to the disadvantage of their clients.

26    This might be thought as putting the point too highly in this country, but reflecting similar concerns, in Australia there has been extensive law reform discussion concerning the role, deficiencies and remedial responses to perceived difficulties with expert evidence. Most relevantly for present purposes, in the Victorian Law Reform Commission Civil Justice Review (2008) (at 484), reference was made by the Victorian Law Reform Commission (VLRC) to the notion of “adversarial bias”. That conception falls into three varieties:

(1)    deliberate partisanship – an expert who deliberately tailors evidence to support the retaining client;

(2)    unconscious partisanship – the expert does not intentionally mislead the Court, but is influenced by the situation to give evidence in a way that supports the retaining client; and

(3)    selection bias – litigants choose as their expert witnesses persons whose views are known to support their case.

27    Given the evident polarisation of scientific views and the possibility of unconscious partisanship and selection bias, from the moment it became apparent there was to be no special jury, I explained I considered it was likely I would be best assisted if the parties could seek out and retain experts in various disciplines who might be described as “cleanskins”: that is, experts who were not wedded to any particular concluded view and who could approach the questions posed informed by the Expert Evidence Practice Note (GPN-EXPT) (Expert Evidence Practice Note) (see also Harmonised Expert Witness Practice Note (Expert Witness Code of Conduct)) and without feeling the need to justify earlier expressed concluded views.

28    My suggestion was met with a mixed response. Indeed, the implicit (and sometimes explicit) suggestion made in Mr McNickle’s submissions was that my suggestions or concerns in this regard were wrongheaded. I will deal with each of the experts below in detail, but it is worthwhile making four related and general points concerning the criticism that entreating the parties to engage experts who did not hold fixed views was somehow heterodox.

29    First, I accept we are dealing with abstruse areas of specialised knowledge. The relatively limited pool of qualified persons to give opinion evidence and the polarisation of scientific view as to the central issue (and matters adjectival to the central issue) would have perhaps caused some difficulty in identifying appropriate experts; a fortiori experts who had not already formed views.

30    Secondly, notwithstanding the first point, I did not come down in the last shower: like any judge who has been an experienced barrister, I understand the forensic attractions that lead to what the VLRC described as selection bias. I hope I am not being unfair in remarking that my distinct impression throughout the whole of the interlocutory stages of this case was that Mr McNickle’s legal team, on one level understandably, resisted any independent referee process and eschewed engaging “cleanskins” in some areas of specialised knowledge because they felt more comfortable in the traditional approach of fixing upon experts they considered would likely reflect their case theory.

31    Thirdly, in the end, the role of the advocates acting for Mr McNickle was to persuade. As a general proposition, it stands to reason that the more objective, dispassionate, and disinterested an expert is, the more the opinions of the expert are likely to be perceived by the tribunal of fact as being: (a) unaffected by unconscious bias; and (b) persuasive. Of course, I stress this does not mean one assesses the admissibility and then reliability of an expert opinion by reference to some a priori view that it would be better to get a “cleanskin”. Disputed questions of fact must be decided according to the evidence that the parties adduce, not according to some speculation about what other evidence might possibly have been led: Australian Securities and Investments Commission v Hellicar [2012] HCA 17; (2012) 247 CLR 345 (at 412–413 [165]–[167] per French CJ, Gummow, Hayne, Crennan, Kiefel and Bell JJ).

32    Fourthly, Mr McNickle in his submissions repeatedly made the point that the mere fact that an expert witness has a pre-existing opinion on a key issue is an infirm basis for rejecting the expert’s opinion evidence “or placing less weight on it”. He cited Born Brands Pty Ltd v Nine Network Australia Pty Ltd [2013] NSWSC 1646 (and on appeal Born Brands Pty Ltd v Nine Network Australia Pty Ltd [2014] NSWCA 369; (2014) 88 NSWLR 421), to assert a pre-existing opinion does not mean an expert is not open-minded or biased but can rather indicate an expert has come to a considered opinion after years of relevant experience. In other words, preconceived principles can inform opinion evidence, if the principles are scientifically based.

33    Of course, Born Brands was dealing with admissibility and not weight: two different concepts. But I accept the underlying point made by Mr McNickle has real merit. The mere fact that the opinion an expert has expressed in this case as to whether, for example, glyphosate is carcinogenic, is the same as the opinion previously expressed on the same issue in other cases against Monsanto in the United States, is neither a complete nor satisfactory basis for the Court to reduce the weight to be afforded to the expert’s opinion evidence; let alone reject its admission. But depending upon the witness and the characteristics of the evidence, it can be part of the constellation of matters that may be relevant to assessing the persuasiveness of the evidence (depending upon how the evidence of the witness is challenged and whether procedural fairness has been provided to the witness to respond to whether the witness is approaching matters with an open mind or is fixed upon defending a pre-existing opinion).

34    In the end, each witness and the opinions expressed by the witness must be judged on their overall substantive merits, not by reference to the preconceived notion that it might have been thought more compelling to adduce evidence from an expert who first formed opinions through the prism of adherence to the Expert Witness Code of Conduct and approached the task without having formed pre-existing views. Recognising the merit of Mr McNickle’s submissions, I have been careful to approach consideration of the expert evidence in this way.

35    As it happened, there was extensive cross-examination by each party alleging partisanship by some experts. Having foreshadowed this as a potential issue early on, I facilitated the experts engaging usefully with each other under the supervision of an independent barrister and settled the questions to be asked of them following extensive argument. I will deal below with the competing contentions of unconscious bias and partisanship; but I will do so as is appropriate: not by reference to sweeping generalisations; but rather informed by my impressions of the relevant expert’s overall demeanour and credit and, most importantly, the perceived underlying scientific merit of the evidence given by the expert.

B.3    Fact Finding Generally in this Case

36    No lay witnesses were called. As noted above, the spectre of an attack on Monsanto for manipulating the science (decisive in my reasoning to not appoint referees) proved a damp squib, and although it was raised by one expert witness called by Mr McNickle directly, minimal evidence was adduced at this hearing directed to this allegation.

37    Leaving aside agreed facts, all the evidence was either expert or documentary. In respect of the latter, what might be described as a tsunami of scientific articles, studies and other materials was placed before me. I deal below with how, by the consent of the parties, I have approached the admissibility and weight to be given to the representations contained in that material.

38    But it must be emphasised at the outset that the Court’s task in determining the central issue is not an exercise of choice between the evidence of the respective parties’ expert witnesses.

39    It is trite to observe that in accordance with s 140(1) of the Evidence Act 1995 (Cth) (EA), the relevant standard of proof for the Court to apply is on the balance of probabilities, and that the applicant bears the onus of establishing that the evidence satisfies that test.

40    I will return to differences between the approach of both science and the law to causation enquiry below, but as I explained in some detail in Transport Workers’ Union of Australia v Qantas Airways Limited [2021] FCA 873; (2021) 308 IR 244 (at 324–325 [284]–[288]), when the law requires proof of any fact, the tribunal of fact must feel actual persuasion as to its occurrence or existence before it can be found. A party bearing the onus will not succeed unless the whole of the evidence establishes a reasonable satisfaction on the preponderance of probabilities such as to sustain the relevant issue: Axon v Axon (1937) 59 CLR 395 (at 403 per Dixon J). In this way, the facts proved must form a reasonable basis for a definite conclusion affirmatively drawn of the truth of which the tribunal of fact may reasonably be satisfied: Jones v Dunkel (1959) 101 CLR 298 (at 305).

41    As I have also noted, despite the criticisms referred to in Transport Workers’ Union of Australia v Qantas (at 325 [286]), this approach is well entrenched and unquestionably represents the current state of the law. This was reinforced by Kiefel CJ, Gageler and Jagot JJ in GLJ v The Trustees of the Roman Catholic Church for the Diocese of Lismore [2023] HCA 32; (2023) 97 ALJR 857, where their Honours observed (at 875 [60]):

[60]    … “[t]o satisfy an onus of proof on the balance of probabilities is not simply a matter of asking whether the evidence supporting that conclusion has greater weight than any opposing evidence ... It is perfectly possible for there to be a scrap of evidence that favours one contention, and no countervailing evidence, but for the judge to not regard the scrap of evidence as enough to persuade him or her that the contention is correct.” The evidence must “give rise to a reasonable and definite inference” to enable a factual finding to be made; mere conjecture based on "conflicting inferences of equal degrees of probability" is insufficient. As Dixon CJ said in Jones v Dunkel, the law:

does not authorise a court to choose between guesses, where the possibilities are not unlimited, on the ground that one guess seems more likely than another or the others. The facts proved must form a reasonable basis for a definite conclusion affirmatively drawn of the truth of which the tribunal of fact may reasonably be satisfied.

(Footnotes omitted)

42    Before passing from these self-directions, it is useful to make two further points.

43    First, in final submissions, Monsanto asserted that the seriousness of Mr McNickle’s allegation, being an allegation that Roundup Products cause NHL, must be considered in deciding whether Mr McNickle has discharged the onus of proof (referring to s 140(2) EA and Briginshaw v Briginshaw (1938) 60 CLR 336 (at 361–362 per Dixon J)).

44    With respect, at best, this is a distraction. Section 140(2) needs to be faithfully applied, but in considering “the gravity of the matters alleged” in s 140(2)(c), the primary focus is upon the nature of the factual allegations in the case (although the consequences of any finding is also a relevant matter). This makes sense when one considers the focus on the gravity of the finding is linked to the notion that the Court takes into account the inherent unlikelihood of alleged conduct, and common law principles concerning weighing evidence: see Qantas Airways Limited v Gama [2008] FCAFC 69; (2008) 167 FCR 537 (at 576 [137]–[138] per Branson J). There is no suggestion of knowing wrongdoing or of other intuitively unlikely conduct; here I am dealing with a question of general legal causation based upon the scientific evidence adduced and Monsanto’s submission is less than helpful when the rationale of Briginshaw is understood.

45    Secondly, on the one hand, Mr McNickle asserts the evidence of general factual causation is pellucid; and on the other hand, Monsanto denies its existence completely. Often these submissions elided the concepts of legal and scientific causation, but it suffices to note for present purposes that the principal case theories adopted are polar opposites. But, as Besanko J observed in Roberts-Smith v Fairfax Media Publications Pty Limited (No 41) [2023] FCA 555 (at [117]–[118]), there is no doubt that a court is not bound to accept the case of one or other of the parties.

46    In this regard, his Honour referred to the “great clarity” of the speech of Lord Brandon of Oakbrook in Rhesa Shipping Co SA v Edmunds (The Popi M) [1985] 1 WLR 948, where his Lordship said the following (at 955):

My Lords, the late Sir Arthur Conan Doyle in his book The Sign of Four, describes his hero, Mr Sherlock Holmes, as saying to the latter's friend, Dr. Watson: “How often have I said to you that, when you have eliminated the impossible, whatever remains, however improbable, must be the truth?” …

In my view there are three reasons why it is inappropriate to apply the dictum of Mr Sherlock Holmes, to which I have just referred, to the process of fact-finding which a judge of first instance has to perform at the conclusion of a case of the kind here concerned.

The first reason is one which I have already sought to emphasise as being of great importance, namely, that the judge is not bound always to make a finding one way or the other with regard to the facts averred by the parties. He has open to him the third alternative of saying that the party on whom the burden of proof lies in relation to any averment made by him has failed to discharge that burden. No judge likes to decide cases on burden of proof if he can legitimately avoid having to do so. There are cases, however, in which, owing to the unsatisfactory state of the evidence or otherwise, deciding on the burden of proof is the only just course for him to take….

47    This reflects the legal reality of how the onus of proof works and the logical reality that a failure to prove a state of affairs does not necessarily mean the state of affairs does not exist. Hence, an acquittal by a jury of a person accused of a criminal charge does not amount to a declaration by the jury of the accused’s innocence (although this state of affairs is taken at law to be the case by reason of the operation of a different presumption of the criminal law).

48    I come back to this important notion in Section J when I deal with the appropriateness of answering the common questions as initially framed.

B.4    The “Weight of Evidence” and the Scientific Studies

49    With metronomic regularity, Mr McNickle asserted that in determining whether glyphosate and/or GBFs can cause or increase the risk of NHL, “it is appropriate to take a weight of evidence approach” to the following three streams of scientific evidence, examined in detail below, being: (1) epidemiological studies of the effects of glyphosate and/or GBFs on humans; (2) long-term studies on experimental animals relating to whether glyphosate and/or GBFs can cause cancer in animals; and (3) mechanistic evidence, that is, evidence relating to whether there is a biologically plausible mechanism by which glyphosate and/or GBFs can induce cancer in humans.

50    The concept of the “weight of evidence” is a common term in published scientific literature, most often seen in the context of risk assessment. The concept is referred to in various conclave reports: Conclave A Joint Report (JRA) (at [19], [125]); Conclave B Joint Report (JRB) (at [72]–[78]); Conclave G Joint Report (JRG) (at [35], [39], [49]–[50], [102]); and Conclave I Joint Report (JRI) (at [13]). But this does not mean the concept is entirely clear; nor does its mere recitation explain how it is called in aid by Mr McNickle.

51    Dr Douglas Weed, an epidemiologist (who was the former Chief, Office of Preventive Oncology, and the Director of the Cancer Prevention Fellowship Program[me] at the National Cancer Institute (NCI), being an agency of the Executive Government of the United States) has published a systematic review of the scientific literature to characterise the “weight of evidence” concept. As I noted during final oral submissions, it is used apparently in three senses:

(1)    metaphorical, where weight of evidence refers to a collection of studies or to an unspecified methodological approach;

(2)    methodological, where weight of evidence points to established interpretative methodologies or where weight of evidence means that “all” rather than some subset of the evidence is examined, or rarely, where it points to methods using quantitative weights for evidence; and

(3)    theoretical, where weight of evidence serves as a label for a conceptual framework. Dr Weed identifies several problems associated with the frequent lack of definition of the term and a lack of consensus about its meaning: see Weed, D L “Weight of Evidence: A Review of Concept and Methods” (2005) 25(6) Risk Analysis 1545 (at 1545–57).

52    Following clarification made during the course of oral submissions, it became apparent that the parties employed the term in the second, methodological sense in that the principled approach requires attention to be given to “all”, rather than some subset, of the scientific evidence.

53    Further, it became apparent during oral submissions that the assessment is suggested to take place on two different levels.

54    First, as is evident from the various conclave reports within each scientific stream, it is common ground that an expert cannot cherry-pick from the peer-reviewed studies those that support the expert’s opinion, but rather carry out a comprehensive assessment of the weight of evidence of regulatory and peer-reviewed studies within that stream of scientific evidence.

55    Secondly, Mr McNickle submits, and Monsanto expressly accepts, that the Court needs to carry out a weight of evidence analysis to “weigh up the three streams of evidence”.

56    I understand the way the concept is used within each scientific stream and its utility, but to the extent the concept is used in this second way, it needs to be approached with some caution and its discrete analytical utility is elusive. As I will explain, scientific causation and the legal question of general causation are different notions and it is trite that my job as a fact finder is to determine the central issue, being a factual issue, according to law, that is, by having regard to all the evidence admitted (and only the evidence admitted).

57    Subject to one matter to which I will come, evidence was admitted because the parties, through their non-objection to its adduction, accepted it was relevant because, if it was accepted, it could rationally affect (directly or indirectly) the assessment of the probability of the existence of a fact in issue at the initial trial: see ss 55 and 56 EA. As explained above, in assessing whether I have reached the level of reasonable satisfaction on the preponderance of probabilities such as to sustain Mr McNickle’s case on the central issue, statute and authority demand I have regard, and only have regard, to the whole of the evidence admitted. My reasoning, inferential and otherwise, can only be based on the evidence admitted and must take account of the entirety of the evidence in reaching conclusions.

58    There is one qualification to what I have said about admissibility. I referred earlier to the scientific articles and studies placed before the Court. There is a lot of law about how a court is to deal with representations of opinion made in such material. As Heydon J explained in Dasreef Pty Limited v Hawchar [2011] HCA 21; (2011) 243 CLR 588 (at 615 [69], 631–632 [110]), there has long existed a qualification to the proof of assumption rule (under a common law exception to the hearsay rule), that experts may give evidence of hearsay matters which go to demonstrate their expertise – what is said in the writings of others in the relevant area of expertise they have read as a basis for their opinion, or what has been said to them in discussions they have had with colleagues and taken into account.

59    In short, experts are generally entitled to rely upon publications and material produced by others in the area in which they have expertise as a basis for their opinions and may give evidence of fact which is based on them. As is explained in Wigmore on Evidence (Little, Brown and Company, 3rd ed) Vol 2 (at 784–785):

The data of every science are enormous in scope and variety. No one professional man can know from personal observation more than a minute fraction of the data which he must every day treat as working truths. Hence a reliance on the reported data of fellow-scientists, learned by perusing their reports in books and journals. The law must and does accept this kind of knowledge from scientific men. On the one hand, a mere layman, who comes to court and alleges a fact which he has learned only by reading a medical or a mathematical book, cannot be heard. But, on the other hand, to reject a professional physician or mathematician because the fact or some facts to which he testifies are known to him only upon the authority of others would be to ignore the accepted methods of professional work and to insist on finical and impossible standards. Yet it is not easy to express in usable form that element of professional competency which distinguishes the latter case from the former. In general, the considerations which define the latter are (a) a professional experience, giving the witness a knowledge of the trustworthy authorities and the proper source of information, (b) an extent of personal observation on the general subject, enabling him to estimate the general plausibility, or probability of soundness, of the views expressed, and (c) the impossibility of obtaining information on the particular technical detail except through reported data in part or entirely.

60    In Karpik v Carnival plc (The Ruby Princess) (Evidential Ruling) [2022] FCA 1318, Stewart J addressed this topic and, after referring to the above extract, noted (at [5]) that “[i]t can thus be seen that expert reliance on such learned publications is not excluded by the rule against hearsay (s 59) or the opinion rule (s 76)”. His Honour was dealing with a case, like here, where the expert reports were replete with references to scientific articles. Having noted the need for caution in making an order under s 136 EA limiting the use of such materials, such a limitation order was not made generally but was made in relation to an article which was: (1) not referred to in any of the expert reports; (2) only referred to in cross-examination; and (3) “deal[t] with an area of expertise not possessed by any of the many expert witnesses in the case” and therefore “there is no other evidence against which to compare, or weigh, the opinions and conclusions expressed in the article”, leaving the Court “in an invidious position with regard to trying to assess the weight to give to those opinions and conclusions” (at [24]).

61    His Honour explained that “[f]or that reason, taken together with the absence of any opportunity for the respondents the test or challenge those opinions and conclusions, I am satisfied that to not limit the use to which the article can be put under s 136 would be unfairly prejudicial to the respondents” (at [24]).

62    To their great credit, the parties to this case (well advised by experienced solicitors and junior counsel), took a sensible and constructive approach to all issues as to admissibility. One aspect of their agreement related to the reception and use of the scientific articles. Taking a slightly different approach to Ruby Princess, I had proposed to the parties a limitation that would restrict reception of material to those articles and studies in the experts’ relevant area of expertise that were read and were said to form a basis for otherwise admissible opinion evidence or were used in cross-examination, accompanied by a general limitation under s 136 EA that the evidence was to be used for the purposes of providing a basis for, or understanding, the expert opinions adduced by the witnesses called.

63    The parties did not embrace this approach. On 19 October 2023, I was informed by the parties that a “joint position” had been reached to the following effect:

All reports or decisions of regulatory authorities and authoritative scientific bodies, as well as scientific articles and reviews and other learned scientific publications, which have been referred to or relied upon by the expert witnesses who participated in the joint conclave should be admitted into evidence, with the parties being at liberty to make submissions in the course of closing submissions as to the weight that the court should place on any such document admitted into evidence.

64    As I noted when informed of this agreement, to the extent it operates to expand the admissibility and use of these scientific materials beyond the principled application of the Pt 3.3 EA exception to the Pt 3.2 hearsay rule, it was open to the parties to dispense, by consent, with the application of Pts 3.2 and 3.3 in relation to particular evidence (see s 190(1)(c) EA).

65    Accordingly, I made orders in accordance with this agreement and will proceed to deal with this aspect of the evidence in accordance with the express agreement of the parties.

B.5    The Scientific and Legal Method and Causation

66    Before passing from fact finding and causation, it is important to make a further point that was somewhat obscured during submissions.

67    In ACQ Pty Ltd v Cook [2009] HCA 28; (2009) 237 CLR 656 (at 661 [14]), French CJ, Gummow, Heydon, Crennan and Bell JJ observed, albeit in a different context, that causation is one of the most difficult fields of debate in the law and one about which “abstract discussion is seldom valuable”. Perhaps recognising this, abstract discussion about causation was largely absent from the parties’ submissions. But one must not gloss over important differences between the scientific and legal method in relation to causal connexions between facts.

68    These differences have been explored in many articles including, by way of example, Korn, H L “Law, Fact, and Science in the Courts” (1966) 66 Columbia Law Review 1080; Brennan, T A “Causal Chains and Statistical Links: The Role of Scientific Uncertainty in Hazardous-Substance Litigation” (1988) 73 Cornell L. Rev. 469; and Geistfeld, M “Scientific Uncertainty and Causation in Tort Law” (2001) 54 Vanderbilt Law Review 1011. It is beyond the scope of this judgment to canvass such a large topic in anything other than the most superficial way, and this judgment should not be unnecessarily diverted into epistemological or philosophical questions, but the nature of differences must be understood, at least in broad terms, so the scientific evidence is properly understood and its relationship to the legal issue of general causation is not distorted.

69    Justice Jonathan Beach has written extra-curially on this topic in a way that repays close attention. Drawing upon his earlier work (“Scientific Evidence: A Need for Caution in Decision-Making” (2010) 42(1) Australian Journal of Forensic Sciences 49 and “Indeterminacy: The Uncertainty Principle of Negligence” (2005) 13 Torts Law Journal 129), in a recent article “Causation: The Interface Between the Scientific and Legal Methods” (2022) 49(1) University of Western Australia Law Review 113, Justice Beach explained (at 113) how the objectives of legal and scientific inquiry differ as science “searches for increased knowledge, with truth as its ideal; its coverage [being] more comprehensive and not time sensitive. But the law’s “objective is to resolve conflicts; its coverage is limited and time sensitive”. In this way, science describes explains and predicts while the law “is concerned with regulating human affairs in accordance with values and objectives”.

70    It is simplistic to say there is one “scientific method” because it depends upon the nature of the scientific task, but speaking generally, these different objectives mean there are critical contextual differences between science and the law concerning methodologies and in the expression of conclusions of their differently focused enquires. Moreover, when it comes to causation, the role of the law might be said to involve an optimistic search for truth but, as my self-directions make clear, the aim is not to discover objective truth, but rather whether the relevant factual issue (in this case of general factual causation) has been proved to the requisite standard.

71    As Justice Beach, with respect, cogently explains in his recent article (at 117–8), while recognising the differences between the legal method and the scientific method:

… when lawyers analyse scientific evidence, the epistemic values that need to be applied correlate with values underpinning the scientific method. A lens must be used which provides a sophisticated picture of the content and reliability of the scientific evidence. And not to analyse scientific evidence through the appropriate lens can lead to outcomes ‘determined by intuitive perceptions of the weight of authority rather than by reasoning from evidence’.

Of course, the language, premises and analytical styles between the scientific method and the legal method have their differences. But when the latter is required to evaluate the former, significant correspondence of epistemic values arises. Lawyers must ‘retrace and evaluate the technical analysts’ logic, from empirical data to subjective judgment’. And complex issues of science require reasoning among purely technical facts analysed through complex models, statistical inference and mathematical instruments that have no comparator in legal reasoning.

Now there are different modes for judicial analysis of scientific evidence. The judge might hear expert evidence in the usual way or through the mechanism of concurrent evidence sessions, which is my preferred option in patent litigation. The judge might appoint his own expert. The judge might obtain assistance from sitting with an assessor who may assist with any explanation of the scientific evidence. The judge might receive a report on the scientific question from a special referee, although this last option is remote from any direct judicial analysis of the scientific question. But whatever the procedural mode, this does not change the epistemic values and methods needed to assess scientific evidence. Moreover, judges are not obliged to accept the ipse dixit of the expert. They are obliged to test propositions, to assess one expert’s opinion against another, and to synthesise a position taking into account all expert, lay and documentary evidence.

72    This is what I have attempted to do. To rely acontextually upon general statements made by scientists as to causation, or to assume that scientific expressions as to possibility or probability necessarily translate to the legal task of considering probability through the lens of general factual causation at law is simplistic. Put in more simple terms, the factual issue before me is, at bottom, a common law jury question and, as was explained by Dixon CJ, McTiernan, Kitto, Taylor and Windeyer JJ in Ramsay v Watson (1961) 108 CLR 642 (at 645):

That some medical witnesses should go into the witness box and say only that in his opinion something is more probable than not does not conclude the case. A qualified medical practitioner may, as an expert, express his opinion as to the nature and cause, or probable cause, of an ailment. But it is for the jury to weigh and determine the probability. In so doing they may be assisted by the medical evidence. But they are not simply to transfer their task to the witnesses. They must ask themselves ‘are we on the whole of the evidence satisfied on the balance of probabilities of the fact?’

C    FACTUAL BACKGROUND

73    To limit the scope of the evidence, I required the legal representatives for both parties to attempt to agree upon all relevant non-contentious facts prior to preparing evidence that would be the subject of dispute.

74    Although for reasons that are unnecessary to canvass, this process was somewhat delayed, a statement of agreed facts was prepared, and a final version was provided to the Court. That finalised version of the statement of agreed facts document was admitted into evidence (Agreed Facts or AF) with each fact being an “agreed fact” within the meaning of s 191 EA. The factual findings are made upon an admixture of the Agreed Facts, material drawn from contemporaneous documents, as well as inferences drawn from any communications and documents. Most facts agreed are irrelevant to the scope of the initial trial. I set out some below for the purpose of recording matters which are of relevance (at times, marginal at best relevance) to the present trial.

C.1    Monsanto

75    Between around 1974 and 1987, the first respondent, Huntsman Chemical Company Australia Pty Limited (previously called Monsanto Australia Limited, which I will describe as Monsanto Australia (Old)), was a wholly-owned subsidiary of the fourth respondent (Monsanto Company (Old)), and together with associated and related entities conducted an agricultural chemicals business in Australia.

76    In about March 1988, the agricultural chemicals business of Monsanto Australia (Old) was transferred to the second respondent, Monsanto Australia Pty Ltd, (Monsanto Australia (New)) and was thereafter conducted by Monsanto Australia (New) and associated and related entities. Monsanto Australia (Old) ceased to conduct that business.

77    Pursuant to a Share Sale Agreement dated 21 March 1988, all of the assets and liabilities of Monsanto Australia (Old) (which were primarily used prior to 31 March 1988 in the agricultural chemicals business conducted by it) were transferred to an entity which became known as Monsanto Australia Pty Ltd (that is, Monsanto Australia (New)).

78    Until September 2000, Monsanto Australia (New) was a wholly-owned subsidiary of Monsanto Company (Old). From September 2000 until 2018, it was a wholly-owned subsidiary of the third respondent (Monsanto Company (New)). In 2018 Monsanto Australia (New) became a wholly-owned subsidiary of Bayer Aktiengesellschaft (Bayer AG).

79    Monsanto Company (Old) and Monsanto Company (New) are entities incorporated in Delaware in the United States. Until approximately September 2000, Monsanto Company (Old) conducted, among other things, an agricultural chemicals business. Following a corporate restructure, that business including as to Roundup Products, was conducted by Monsanto Company (New) after approximately 1 September 2000.

80    In or around early 2000, Monsanto Company (Old) merged with a publicly owned pharmaceuticals company. Monsanto Company (Old) was the surviving corporation from the merger.

81    Monsanto Company (New) was created on 9 February 2000 as a wholly-owned subsidiary of Monsanto Company (Old). Effective on 1 September 2000, Monsanto Company (New) received the operations, assets and liabilities of the agricultural business previously conducted by Monsanto Company (Old), including in respect of the Roundup Products. Upon completion of the merger Monsanto Company (Old) changed its name to Pharmacia Corporation.

C.2    Glyphosate and Roundup

82    The International Union of Pure and Applied Chemistry name for glyphosate is N- (phosphonomethyl) glycine.

83    Glyphosate was first synthesised by a chemist employed by Monsanto Company (Old) in 1970 and patented by Monsanto Company (Old) in 1974 for use as an herbicide.

84    The manufacture of Roundup Products involves the conversion of intermediate products to glyphosate acid (also known as glyphosate technical) which, in turn, is further converted to glyphosate salts for use in the formulation of Roundup Products.

C.3    Roundup Herbicide and Roundup Biactive in Australia

85    Roundup Herbicide is a herbicide product which has been available for sale in Australia since 1976. In the time periods set out below, the entity identified manufactured Roundup Herbicide and/or Roundup Biactive:

(1)    Roundup Herbicide

(a)    1979 to 1988 – Monsanto Australia (Old);

(b)    1988 to 2000 – Monsanto Australia (New) and Monsanto Company (Old);

(c)    2000 to 2019 – Monsanto Australia (New) and Monsanto Company (New)

(2)    Roundup Biactive

(a)    1996 and 2000 – Monsanto Australia (New) and Monsanto Company (Old);

(b)    2000 to 4 February 2002 – Monsanto Australia (New) and Monsanto Company (New);

(c)    2 February 2002 to 2010 – Monsanto Australia (New);

(d)    2010 to 2019 – Monsanto Australia (New) and Monsanto Company (New).

86    The Roundup Herbicide and Roundup Biactive used in Australia during the Relevant Period was manufactured in accordance with specifications (including as to the formulations) and quality assurances manual set by Monsanto Company (Old) for the period 1976 to 2000, and by Monsanto Company (New) for the period 2000 to 2022.

87    The Roundup Herbicide and Roundup Biactive used in Australia during the Relevant Period was packaged (including labelling) in accordance with specifications and quality assurances manuals set by Monsanto Company (Old) for the period 1976 to 2000, and by the Monsanto Company (New) for the period 2000 to 2020.

88    During the claim period:

(1)    Monsanto Company (Old) for the period 1976 to 2000 in respect of Roundup Herbicide and for the period 1996 to 2000 in respect of Roundup Biactive;

(2)    Monsanto Company (New) for the period 2000 to 2019;

(3)    Monsanto Australia (Old) for the period 1976 to 1988 in respect of Roundup Herbicide; and

(4)    Monsanto Australia (New) for the period 1988 to 2020 in respect of Roundup Herbicide and for the period 1996 to 2020 in respect of Roundup Biactive,

directly or through other parties and agencies promoted and marketed Roundup Herbicide and Roundup Biactive in Australia and engaged in activities related to marketing including conducting promotions, educational publications and programmes, product demonstration programmes and print advertisements and in accordance with product use and safety information provided by Monsanto Company (Old) for the period 1976 to 2000 and by Monsanto Company (New) for the period 2000 to 2020.

89    Information concerning the safe use of Roundup Herbicide and Roundup Biactive was communicated by Monsanto to consumers in Australia during the Relevant Period by two predominant means:

(1)    the product label including the “Directions for Use” booklet, both affixed to the container of the products; and

(2)    the safety data sheets.

D    STRUCTURE OF CONSIDERATION OF THE SCIENTIFIC EVIDENCE

90    As noted above (at [49]), the scientific evidence relevant to determining the central issue was divided into three streams, broadly as follows:

(1)    epidemiological studies of the effects of glyphosate and/or GBFs on humans (Conclaves C, D and E) (epidemiology stream);

(2)    long-term studies on experimental animals relating to whether glyphosate and/or GBFs can cause cancer in animals (Conclaves H, I and J) (animal studies stream); and

(3)    mechanistic evidence, that is evidence relating to whether there is a biologically plausible mechanism by which glyphosate and/or GBFs can induce cancer in humans (Conclaves A, B, G) (mechanistic evidence stream).

91    Consistently with the need to have regard to the whole of the evidence, the parties accepted that it is necessary to make factual findings in relation to the evidence adduced in each conclave. It is then appropriate to consider each conclave in the context of the relevant scientific streams and then have regard to all the evidence from each stream. For reasons that will become clear, I will deal separately with Conclave F, which concerned absorption and exposure and the discrete question of dose.

92    Accordingly, I will address each stream separately (although there is necessarily some overlap). In doing so, I will first explain the nature of the stream of scientific evidence; secondly, make general credit findings with respect to each the relevant experts; thirdly, summarise the evidence given in each conclave relevant to each stream; and fourthly, summarise my conclusions in relation to each stream of evidence building upon and supplementing the findings that have been made during the process of setting out the conclave evidence. I will then draw all the threads together in reaching my conclusions.

E    THE FIRST STREAM: EPIDEMIOLOGY (CONCLAVES C, D AND E)

E.1    Nature of the Evidence

I    Introduction

93    In 1854, Dr John Snow witnessed an outbreak of cholera in London. Determined to discover its source, he plotted the location of cholera deaths on a dot-map. The dots congregated a short distance from Broad Street (now Broadwick Street) in Soho. There, the local council had installed a water pump. Suspecting that the pump was the source of the outbreak, Dr Snow persuaded the council to remove the handle. The outbreak subsided.

94    But that was not the end of the story: had the handle been contaminated? Was it the water itself? Or was the outbreak due to something else? Dr Snow thought it was the water. He continued to collect data on cholera deaths in the city and identified three water companies that supplied water to districts with the highest mortality rates. It was not clear within each district, however, which company supplied water to households where a cholera death occurred. By collecting data from individual households, Dr Snow inferred that two of the water companies were responsible for transmitting cholera poison and, by 1857, legislation had been enacted to mandate the filtering of all water in London (see Harris, S A “Epidemiology: Theory, Study Design and Planning for Education” (2000) 20 Journal of Continuing Education in the Health Profession 133).

95    Albeit in nascent form, Dr Snow was engaged in epidemiology. For his efforts, he enjoys the not inconsiderable distinction of having a pub named after him.

96    Expressed in broad terms, epidemiology is the study of distribution and determinants of disease in humans. As Spigelman CJ explained in Seltsam Pty Ltd v McGuiness [2000] NSWCA 29; (2000) 49 NSWLR 262 (at 271–272 [59]–[62]), epidemiology is based on the assumption that a disease is not distributed among populations at random: subgroups may be identified which are at increased risk of contracting particular diseases. Epidemiological evidence identifies associations between specific forms of exposure and the risk of disease in groups of individuals.

97    Epidemiologists make judgments about whether a statistical association represents a cause-effect relationship. However, those judgments focus on what is called in the epidemiological literature (as well as the law) general causation, that is, whether the factor is capable of causing the disease. Epidemiologists are not concerned with specific causation, that is, whether the factor caused the disease in an individual case.

98    For the purposes of determining whether exposure to a particular substance is the legal cause of a particular disease, epidemiology only provides evidence of possibility, which (for reasons I have explained), when expressed in opinion form and in epidemiological research, is admissible and part of the whole of the evidence to which the court must have regard: Seltsam v McGuiness (at 274–275 [78]–[79] per Spigelman CJ).

99    At bottom, this is a trial directed to the issue of general factual causation. Obviously enough, given its nature, epidemiological evidence is of real importance. Indeed, as it turns out, successfully demonstrating that the epidemiological evidence adduced in this proceeding supports the conclusion that glyphosate and/or GBFs can increase the risk of and/or cause NHL in humans is a hurdle in accepting Mr McNickle’s case.

100    But it remains one hurdle. As senior counsel for Mr McNickle, Mr Clements KC, submitted in his opening (T10.32–37):

Despite there being very broad agreement between the experts retained by each side, as to the appropriateness of the weight of evidence approach, the respondents, in their written opening, have submitted that the epidemiological evidence should be given primacy by the court. We contend that approach should be rejected. We contend that it’s important for the court to closely consider and weigh up the evidence in each of the three separate streams of scientific evidence.

101    Put another way, the importance of the epidemiological stream does not mean that my conclusion as to whether I have reached a level of reasonable satisfaction on the preponderance of probabilities to sustain Mr McNickle’s case on the central issue is the product of giving primacy to one aspect of the evidence and reducing others to some form of secondary role or, more specifically, placing unwavering and disproportionate focus upon the epidemiological evidence in a vacuum. It is not. Indeed, focussing too intently on one part of the evidence at the expense of proper consideration of another part and then assessing the accumulated whole, would not only be a legal but a logical error.

102    Of course, some parts of the evidence may emerge as being more persuasive or important than other aspects, but any conclusion on the central issue must be made with regard to the overall weight of all of the evidence.

II    Methodology

103    It is largely uncontroversial among epidemiologists as to the principles or postulates which should be applied to assess evidence of a statistical correlation or association between the exposure in question and health outcome of interest: see Seltsam v McGuiness (at 285 [139] per Spigelman CJ).

104    The first step in assessing the epidemiological evidence and whether it indicates an etiologic relationship (that is, a causal relationship) between exposure to glyphosate and/or GBFs and development of NHL in humans is to identify the relevant peer-reviewed literature. The second step is to identify strengths and limitations of each publication, which, in turn, allows an identification of publications that are most relevant to the specific topic of concern, and an evaluation of the weight that should be placed on such publications (Conclave C Joint Report (JRC) (at [63])).

105    In evaluating the relevant literature, epidemiologists may attempt to distinguish between causal and non-causal explanations (at JRC [64]–[65]). While there is no agreed set of criteria to assess causation that exists in the field, lists of criteria have become popular. The most well-known of these among epidemiologists is one devised by Sir Austin Bradford Hill, then Professor Emeritus of Medical Statistics at the University of London, in his Presidential address to the Section of Occupational Medicine: “The Environment and Disease: Association or Causation” (1965) 58 Proc R Soc Medicine 295.

106    The following summary of the “Bradford Hill criteria” is extracted from French CJ’s judgment in Amaca v Booth Pty Ltd [2011] HCA 53; (2011) 246 CLR 36 (at 54–55 [44]):

[44]    … The [Bradford Hill] criteria were expressed as the aspects of an association between two variables that should be considered before inferring that the most likely interpretation of the association is causation. In summary, they are:

strength of association – eg, reflected in the ratio of the death rates between groups exposed to a suspected agent and those not so exposed;

consistency in the observed association – eg, has it been repeatedly observed by different persons in different places, circumstances and times;

the specificity of the association – if the association is limited to specific workers and particular sites and types of disease and there is no association between the work and other modes of dying, that is a strong argument in favour of causation;

temporality – the temporal relationship of the variables;

biological gradient – whether the association reveals a biological gradient or dose-response curve;

plausibility – whether the expected causation is biologically plausible – a consideration which depends upon the biological knowledge of the day;

coherence – the cause and effect interpretation of the data should not seriously conflict with the generally known facts of the natural history and biology of the disease;

experiment – whether experimental or semi-experimental evidence supports a causation hypothesis;

analogy – eg, given the effects of thalidomide and rubella it is easier to accept slighter but similar evidence with another drug or another viral disease in pregnancy.

107    These factors are not intended as a necessary condition for a cause-and-effect relationship, nor a checklist or set of hard and fast rules to be applied in evaluating causation among epidemiologists. Rather, the Bradford Hill criteria constitutes a guide or set of commonsense propositions which may be taken into account in determining whether or not the Court should infer, on the balance of probabilities, that a particular exposure caused injury (see Amaca v Booth Pty Ltd (at 54–55 [44]–[45] per French CJ); Seltsam v McGuiness (at 285 [139]–[141] per Spigelman CJ); (at JRC [66]–[71])). As was noted in a discussion of the application of the Bradford Hill criteria in the Restatement Third, Torts: Liability for Physical and Emotional Harm §28, Comment c(3) (at 406–407):

Whether an inference of causation based on an association is appropriate is a matter of informed judgment, not scientific methodology, as is a judgment whether a study that finds no association is exonerative or inconclusive. No algorithm exists for applying the Hill guidelines to determine whether an association truly reflects a causal relationship or is spurious. Because the inferential process involves assessing multiple unranked factors, some of which may be more or less appropriate with regard to a specific causal assessment, judgment is required.

108    Before going further, and in an attempt to make what follows less unreadable, it is worth identifying some key terminology and demystifying some jargon frequently deployed in the epidemiological literature, which is also relevant to the two other streams of evidence.

III    Terminology

Relative risk, odds ratios and confidence

109    As Dr Ian Freckelton AO KC explained in Expert Evidence: Law, Practice, Procedure and Advocacy (Thomson Reuters, 6th ed, 2021) (at 1199), epidemiologists identify the strength of an association between a particular form of exposure and the risk of contracting a particular disease by a measure described as “relative risk” (RR). The RR describes the rate of a disease of interest occurring in persons who have been exposed to a particular environmental factor divided by the corresponding rate in persons without a history of that exposure. For example, a RR of 2.0 indicates a two-fold risk in exposed people compared to non-exposed people.

110    Another term frequently used in the literature is “odds ratio” (OR). The OR represents the odds that an outcome will occur given a particular form of exposure, compared to the odds of the outcome occurring without that exposure. Although the OR can indicate the strength of an association between exposure and outcome, it is not the same as RR. For example, when there is no association between exposure and outcome, the OR and RR are identical (1.0). However, when there is an association between an exposure and an outcome, the OR can exaggerate the estimate of their relationship.

111    ORs can be statistically significant or not. As Morling J explained in Re Australian Federation of Consumer Organisations Inc v Tobacco Institute of Australia Ltd (1991) 27 FCR 149 (at 163–164), statistical significance is an expression of the relative confidence that a particular result (say, an OR of 2.0), is a true effect and not due to chance.

112    Statistical significance is frequently expressed in the scientific literature in terms of a “confidence level” (CL) which, conventionally, is the 95% CL. That is, if a result (or a range of results) is statistically significant at the 95% CL, the probability that result is due to chance is approximately 1 in 20. An example is usefully provided by Morling J (at 163):

[A]n investigation of the association between exposure (to an agent) and outcome (disease), having opted for a confidence level of 95%, may produce a point estimate of 2.5, with upper and lower confidence limits of 3.5 and 1.5. This means that based on the raw data, the best estimate is that those exposed have a 2 times greater chance of experiencing the disease than those not exposed. Further, one can be 95% sure that the true relative risk is somewhere between 1.5 and 3.5.

113    The range between the upper and lower confidence limits is known as the “confidence interval” (CI). If the CI is wide, then the point estimate is less reliable than if it is narrow. Ordinarily, wide CIs indicate a small number of observations.

114    It is also worth noting the relationship between CIs and statistical significance.

115    The relative risk of disease normally associated with pure chance is 1.0: that is, no increased risk, no decreased risk. Using the 95% CL, if the value 1.0 lies between the reported upper and lower CLs, then the probability that the result is due to chance has not been satisfactorily excluded at the 95% CL. Whenever the confidence interval straddles 1.0, then the result is not statistically significant. To borrow Morling J’s example (at 164), assuming a 95% CL, Study A reports a point estimate of 2.5 with a CI of 1.5 to 3.5. Study B reports a point estimate of 2.5 with a CI of 0.8 to 4.0. The results in Study A are statistically significant, whereas the results in Study B are not. Importantly, a CI which includes the value of 1.0 is frequently referred to in the literature as including the “null hypothesis” (that is, no association between the particular form of exposure and risk of contracting the disease).

116    Decisions in the United States have demanded a relative OR of greater than 2.0 to establish causation on the balance of probabilities: see Daubert v Merrell Dow Pharmaceuticals Inc 43 F 3d 1311 (9th Cir, 1995); Re Joint Eastern and Southern District Asbestos Litigation 758 F Supp 199 (1991) (at 202–203). This is not the law in Australia. As Spigelman CJ explained in Seltsam v McGuiness (at 285 [137]), the test of actual persuasion does not require epidemiological studies to reach the level of a relative risk of 2.0, notwithstanding that the closer the ratio approaches 2.0, the “greater the significance that can be attached to the studies”. The “strands in the cable” must be capable of bearing the weight of the ultimate inference: see also Re Australian Federation of Consumer Organisations (at 211 per Morling J).

Types of studies

117    It is also worth briefly explaining some key differences between the types of studies found in the epidemiological literature.

1.    Cohort studies

118    In a cohort study, a group of individuals is identified and then divided into those who have been exposed to the event in question (say, exposure to glyphosate and/or GBFs) and those who have not. The group is then followed for a number of years, and individuals who contract the disease are identified in order to determine if there is any difference in incidence between the groups (at JRC [16]).

119    Associations in cohort studies are typically expressed in terms of RR.

2.    Case-control studies

120    In contrast, case-control studies identify individuals with the disease in question (the cases) from the outset, which are then matched to a group of individuals who do not have the disease (the controls). Data about exposure to a risk factor (or several risk factors) are then collected retrospectively, typically by interview, abstraction from records, or questionnaires.

121    Controls can be individually matched to cases or matched on a group basis in which distributions of age, sex, and other covariates are similar. Matching is conducted to reduce confounding by causes of the disease of interest other than those being studied that are correlated with the exposures under investigation (at JRC [15]).

122    Associations in case-control studies are typically expressed as ORs.

3.    Pooled studies

123    Pooled analysis is a method which provides for the combining of results of multiple studies. It is often used when the results of individual studies are too small or not definitive enough to draw any firm conclusion as to causation. In contrast to meta-analyses, pooled analyses can only be conducted if the included studies used the same study design and statistical models, and if their respective populations were homogeneous. If done properly, a pooled study can provide more statistically precise estimates of risks associated with the exposures of concern and allow for the evaluation of associations among subpopulations or subgroups or subtypes of disease (at JRC [17]).

Other relevant terminology

124    It is useful here to set out some other relevant terminology, some of which will be relevant to the other streams. What follows is largely drawn from definitions provided in Section D of the JRC:

(1)    Environmental exposures – chemical, biological, and physical agents that occur in the air, water, or soil that can be hazardous to health. Chemical agents of particular concern globally are air pollution (for example, particulate matter), water pollution (for example, toxic metals), and food contaminants (for example, E coli bacteria).

(2)    Meta-analysis – a meta-analysis statistically summarises the risk estimates from multiple studies and can provide an overall estimate of risk. When properly conducted, a meta-analysis can combine evidence from individual studies and can provide more stable summary estimates of effect. Typically, systematic reviews of the literature and meta-analysis of the results are conducted to obtain summary risk estimates that can help to inform causal inference.

(3)    Confounding – confounders are variables that are associated with the health outcome of interest and are correlated with exposures being studied as risk factors. Confounders may be demographic variables (for example, age, sex, ethnicity) or other exposures that are risk factors for the health outcomes of interest. For example, findings from studies of a particular pesticide and a particular health outcome may be confounded if the exposures of interest are correlated with other exposures that are etiologically related to the outcome of interest. Results that are confounded can lead to erroneous results and conclusions.

(4)    Exposure assessment – the method of exposure assessment involves estimating study participants’ exposures to environmental or lifestyle factors. The assessment can range from binary precision (“yes” or “no”); to high / medium / low (ordinal); to quantitative levels (continuous) (for example, micrograms of the chemical exposure per milligram of soil).

(5)    Etiologic association – a causal relationship between the exposure and the health outcome being studied.

(6)    P-value – the probability that a finding was due to chance. A p-value of less than 0.05 is often regarded as an indicator of a “real” (non-chance) finding. The underlying assumption is that 5% of findings are likely to be due merely to chance. For example, if p=0.01, then there is no more than a 1% possibility that the finding was merely a chance observation.

(7)    Dose-response trend – the pattern of relative risk estimates in relation to increasing exposure levels. An upward (“positive”) consistent trend is indicative of an etiologic relationship between exposure and disease risk.

(8)    Recall bias – a systematic error that occurs when study participants do not remember previous events or experiences accurately.

(9)    Selection bias – a distortion in a measure of association due to a sample selection that does not accurately reflect the target population.

125    I turn now to the relevant experts in the epidemiology stream.

E.2    The Relevant Experts

I    Overview

126    The epidemiology stream spanned Conclaves C, D and E and involved the following expert witnesses:

(1)    Conclave C – Epidemiology (epidemiology concerning glyphosate)

Professor Checkoway and Associate Professor Harris

(2)    Conclave D – Epidemiology and biostatistics (the correct allocation of person years in the cohort studies which have been conducted into glyphosate)

Professor Gordon and Associate Professor Harris

(3)    Conclave E – Biostatistics (statistical analysis of epidemiology studies)

Professor Gordon and Dr Crump

127    As noted earlier, in addition to the giving of concurrent evidence in open court and the preparation of joint reports in respect of each conclave, several individual reports were prepared by each of the experts and relied upon by the parties for the purposes of the epidemiology stream, as set out in the table below:

128    In the following section, I will introduce each of the experts who participated in the epidemiology stream, before turning to making some general observations as to credit. In doing so, I am conscious that the parties have made submissions as to credit which loom large in respect of some witnesses and not others. I have had regard to all these submissions, but it would add further to an already lengthy and oppressive judgment to set them out in full, except where I consider them to be of especial importance. As will be seen, in this section I make repeated references to various aspects of the evidence and various studies. This evidence and the studies referred to are dealt with comprehensively later in my judgment.

129    With that said, for reasons that will already be obvious, a significant amount of time was expended by both sides in mounting attacks upon witnesses (not limited to the epidemiology stream) whose evidence, it is said, was marred by partisanship. As I explained in Russell v Australian Broadcasting Corporation (No 3) [2023] FCA 1223 (at [438]), many experienced judges have expressed the caution that any criticisms of a witness, which go beyond the legitimate necessities of the occasion, should be avoided. Indeed, unnecessary credit findings should be avoided partly because of a body of research casting doubt on the ability of judges to make accurate credibility findings based on demeanour: see Fox v Percy [2003] HCA 22; (2003) 214 CLR 118 (at 129 [31] per Gleeson CJ, Gummow and Kirby JJ).

130    Despite an adverse credit finding with respect to one witness, issues going to credit are not the centrepiece of this case. As the High Court explained in Fox v Percy (at 129 [31]), the Court’s task is to reason its conclusions, as far as possible, on the basis of contemporary materials, objectively established facts and the apparent logic of events (in this initial trial, the scientific materials). This is not to say that my assessment of the scientific evidence in the three streams is not partly informed by impressions I have gained as to the relevant expert’s overall demeanour and credit. It is. But there is an extent to which such impressions can distract the Court from the principal task of assessing the underlying scientific merit of the evidence adduced in this proceeding.

II    Credit Findings

Professor Checkoway

131    Professor Harvey Checkoway is an applied epidemiologist specialising in occupational and environmental risk factors for multiple different malignant diseases and neurological disorders, especially Parkinson’s disease. Professor Checkoway was engaged by Mr McNickle.

132    Professor Checkoway holds a bachelor’s degree in psychology from Boston University (1971), an MPH in epidemiology from Yale University (1975), and a PhD in epidemiology from the University of North Carolina (1979). Since 2013, he has been a Professor at the University of California, San Diego Departments of Family Medicine and Public Health. Professor Checkoway previously held faculty positions at the University of North Carolina in the Department of Epidemiology and the University of Washington in the Departments of Environmental and Occupational Health Sciences and Epidemiology.

133    Much of the contest concerning Professor Checkoway’s evidence went to the basis for his conclusion in the JRC (at [132]) that there is “generally supportive evidence that environmental GBF exposures are etiologically related to NHL” (emphasis added). It is important to note from the outset that Professor Checkoway, in reaching that conclusion, stated that he relied predominantly on four individual case-control studies, namely Hardell and Eriksson (1999); McDuffie (2001); Eriksson (2008); and Meloni (2021). I will return to these studies below.

134    Mr McNickle submits that Professor Checkoway gave evidence in a careful and considered manner and that the Court should find that he was a reliable witness, for the following reasons.

135    First, Professor Checkoway responded directly to questions posed to him and made appropriate concessions without hesitation. He did not seek to dismiss, nor set aside, epidemiological studies which did not support his conclusion that there is generally supportive evidence that GBF exposures are etiologically related to the risk of humans developing NHL.

136    Secondly, and relatedly, Professor Checkoway gave evidence that it is important for the Court to have regard to all of the relevant studies, and that he did not consider the studies which did not indicate an etiologic relation as inferior to those which did, but that those studies were inconsistent with other studies (namely, the four individual studies) on which he relied and found to be “especially meaningful” (T866.11–13; T866.36–39).

137    Subject to matters to which I will return momentarily, my impression of Professor Checkoway is that he generally did his best to give candid answers when challenged on critical aspects of his evidence. Having said that, it is worth mentioning two examples which were reflective of a tendency on the part of Professor Checkoway, from time to time, to not explain persuasively the reasons why he placed less weight upon studies which did not support his ultimate conclusion that there is an etiologic association between exposure to glyphosate and/or GBFs and NHL.

138    First, and most significantly, Professor Checkoway omitted reference to the Agricultural Health Study (AHS) (to which I will return) in the JRC (at [128]–[132]) when setting out his conclusions as to whether there exists evidence for an etiologic association between glyphosate exposure and development of NHL in humans. Notwithstanding that Professor Checkoway addressed the AHS and its strengths and weaknesses (at JRC [102]–[104]), for reasons that will become all too apparent, its omission when offering his ultimate conclusion as to the central issue is, at best, puzzling. This is highlighted by Professor Checkoway ultimately accepting that the AHS is an “excellent study” and “[i]n terms of study design and methodology, I think [the AHS is] perhaps the best study we have” (T959.14–15; T959.44–45).

139    Secondly, in the JRC (at [95]), Professor Checkoway sets out in Table 1 a summary of findings from publications that address potential etiologic relationships between glyphosate and/or GBFs and NHL. It does not, however, include the case-control study of Cantor (1992) (to which I will return). At some risk of oversimplification, that study suggests that there is no significant risk elevation of NHL for exposure to glyphosate. When asked about why he chose not to include Cantor (1992) in Table 1, Professor Checkoway explained that it was an accidental omission and that it should have been included in the table (T1031.44–1032.3). It is regrettable the omission of the Cantor study in Table 1 was the result of an inadvertent slip given that it is included in Table 2 of the JRC (at [113]), albeit as part of a list of limitations.

140    Thirdly, and relatedly, Professor Checkoway did not resile from his position in the JRC that greater weight ought to be placed upon the individual studies, as opposed to the pooled studies, for the conclusion that there is an etiologic link between glyphosate and/or GBF exposure and development of NHL in humans. Despite this, when asked about the weight that should be attributed to Cantor (1992), compared to De Roos (2003) (a pooled study, of which the Cantor study contributes 26 of the 36 exposed cases and 49 of the 61 controls), Professor Checkoway’s evidence was that greater weight should be placed on De Roos (2003) (T1032.19–36).

141    These examples may not be particularly significant in isolation and do not detract from my overall impression that Professor Checkoway was a generally impressive witness who sometimes made concessions. It is necessary, however, to approach Professor Checkoway’s evidence with increased scrutiny in the light of his tendency to place weight upon studies which support his conclusion and downplay weaknesses which detract from that conclusion.

Associate Professor Harris

142    Associate Professor Shelly Harris (AP Harris) is an epidemiologist and an associate professor in the Division of Epidemiology and the Division of Occupational and Environmental Health in the Dalla Lana School of Public Health, University of Toronto. AP Harris was engaged by Monsanto.

143    AP Harris holds a PhD in epidemiology in the Department of Preventive Medicine and Biostatistics (now Public Health Sciences), Faculty of Medicine, University of Toronto. AP Harris is a scientist in the Occupational Cancer Research Centre and previously was appointed as a scientist in the Division of Prevention and Cancer Control at Cancer Care Ontario.

144    Relevantly for present purposes, AP Harris’ ultimate conclusion (at JRC [233]–[236]) is that based upon her critical evaluation of all the epidemiological studies, despite some case-control studies indicating some statistically significant elevations in ORs for NHL, the evidence does not support a causal connexion between exposures to glyphosate and/or GBFs and increased risk of NHL overall (or its subtypes).

145    Mr McNickle submits that the Court should reject AP Harris’ evidence and conclusion that the scientific literature does not indicate a causal connexion between exposure to glyphosate and/or GBFs and development of NHL because AP Harris was not a wholly credible witness and, at least at times, did not approach her role as an expert witness impartially and strayed into advocacy. In the light of that tendency, Mr McNickle submits that the Court should prefer the evidence of Professor Checkoway and Professor Gordon.

146    Four points are made.

147    First, in the First Harris Report, the JRC, and the Conclave D Joint Report (JRD), it is said that AP Harris expressed opinions and conclusions that are inconsistent with opinions that she had expressed previously in peer-reviewed, published papers and in presentations to the wider scientific community, and that accordingly, the opinions expressed by AP Harris in this proceeding are more favourable to Monsanto than conveyed previously to the scientific community.

148    Secondly, AP Harris sought to defend and endorse the AHS with a degree of impartiality that was inappropriate for an expert witness. Three points are developed in relation to this submission, namely: (1) AP Harris produced five reports (totalling, remarkably, 335 pages) in respect of the NCI data in De Roos (2005) (despite the orders of the Court providing for one report); (2) AP Harris prepared a “sensitivity analysis” (that is, the AP Harris Sensitivity Report) on her own initiative in respect of Professor Gordon’s conclusions in the Third Gordon Report because it remained a live issue in the proceeding; and (3) AP Harris told the Court that the conclusions in the AHS papers “must be relied upon” for the evaluation of the association between glyphosate use and the risk of NHL as their study design offers many advantages and strengths over other studies. By contrast, it is said that AP Harris referred to aspects of Chang and Andreotti (2023), which was prepared by many of the same authors and used the same AHS data but reported an association between glyphosate exposure and oxidative stress, as “misleading”.

149    Thirdly, Mr McNickle submits that AP Harris strayed into ad hominem attacks upon Professor Gordon’s professional expertise and character, including an assertion that Professor Gordon’s opinion was “purposively misleading or demonstrates a lack of knowledge of the fundamentals of analysing data from epidemiological cohort studies” (Supplementary Harris Report (at [19])).

150    Fourthly, it is said that in contrast to Professor Checkoway and Professor Gordon, AP Harris often gave unresponsive answers to questions put to her in cross-examination and that she had to be asked several times to answer the question asked.

151    On balance, I consider these criticisms to be overstated. I do not think that AP Harris was intent on avoiding concessions that she considered adverse to the case of Monsanto, and, to the extent that she strayed into any perceived advocacy, I consider that that was a product of a genuine attempt to get at the truth of the scientific evidence, as opposed to an attempt to bolster the case of Monsanto or her own conclusions in her primary report or joint reports.

152    One example will suffice. When giving oral evidence in Conclave D, AP Harris was asked about the cross-sectional study Blair (2011) which, in summary, assessed the impact of exposure misclassification on estimates of relative risks in the AHS papers, using the range of correlation coefficients observed between measured post-application urinary levels of 2,4-D and chlorpyrifos (two herbicides) and exposure estimates. Blair (2011) did not include data on exposure to glyphosate and, therefore, was of general relevance only to an understanding of the AHS papers. However, when questioned about the correlation coefficient in the study, AP Harris opined that glyphosate may “behave similarly” to 2,4-D or other herbicides (T799.4). The exchange continued (T799.19):

MS SZYDZIK: Just a couple of questions arising out of that. Associate Professor Harris, you just said that you would expect glyphosate might behaviour similarly to 2,4-D, are you just able to explain why you think that’s the case?

ASSOC PROF HARRIS: It’s a possibility. I think it’s because it’s not a highly volatile compound and often they’re spraying and using similar types of application methods. So they might be spraying it off a tractor or something like that. But they’re not going to be putting it in a backpack and spraying it by hand. So there’s some similarities in the way these compounds are applied and also their physical characteristics, as I understand.

153    Having had the benefit of seeing this evidence unfold, I have difficulties in accepting the submission that AP Harris was intent on supporting the case of Monsanto at the expense of her own impartiality. There can be little doubt that if AP Harris was so minded, she would have been reluctant to proffer a view that glyphosate may behave similarly to other herbicides, such as 2,4-D. As noted earlier, this was indicative of what I perceived to be a desire on the part of AP Harris to be faithful to her oath.

154    While I accept far too much material was produced contrary to the directions of the Court, I suspect that had more to do with over-enthusiasm (including of lawyers) rather than any lack of objectivity on the part of AP Harris.

Professor Gordon

155    Professor Ian Gordon is an applied statistician specialising in biostatistics and epidemiology, including methodological issues in randomised controlled trials, cohort studies, case-control studies, meta-analysis, and survival analysis. Professor Gordon was engaged by Mr McNickle.

156    Professor Gordon holds a bachelor’s degree in statistics from the University of Melbourne, a MSc (by thesis) from La Trobe University concerning conditioning in statistical inference, and a PhD from the University of Melbourne on sample size determination for discrete data. He is currently a Professor of Statistics at the University of Melbourne and the director of the Statistical Consulting Centre.

157    It will suffice to note for present purposes that much of the controversy over Professor Gordon’s evidence concerned his scrutiny of the methodology used in the AHS papers which, in summary, were that first, Andreotti (2018) did not obtain exposure data up to the point in time that it obtained data about cancer incidence; and secondly, about the way in which Andreotti (2018) utilised multiple imputations to address missing data. As it turns out, these criticisms were largely unfounded or, if they applied, had a negligible effect on the ultimate conclusions expressed by the authors of the AHS papers. For reference, I deal with this topic in the context of addressing the AHS in Section E.3 below.

158    Mr McNickle submits that notwithstanding this, the Court should find that Professor Gordon was a credible and reliable witness who gave evidence in a careful and considered manner. It is said that his willingness to change and revise his opinions in the light of new information and data indicates that he kept an open mind throughout the trial, which is demonstrative of his impartiality as an expert.

159    It was difficult not to gain the impression that Professor Gordon, at times, appeared inflexibly wedded to his criticisms of the AHS papers. Several examples may be cited, and I do not intend to set them all out here.

160    With that said, I do not accept that this was the product of a deliberate attempt on Professor Gordon’s part to obfuscate the scientific literature to bolster his own conclusions or the case of Mr McNickle. When challenged on his criticisms of the methodologies adopted in the AHS papers, Professor Gordon, albeit belatedly, made appropriate concessions and accepted when the evidence plainly tended against his conclusions (see, for example, at JRD [66], [230]). At one point, during cross-examination, Professor Gordon even went so far as to concede that he could not comment upon the issue of latency between exposure and clinical manifestation of NHL because he lacked expertise in cancer processes (T675.24–31).

161    Overall, my impression is that Professor Gordon approached the task of giving evidence with integrity, and that if his evidence was less than satisfactory in respects, it was not for want of candour.

Dr Crump

162    Dr Kenny Crump is a professional consultant and biostatistician. He was engaged by Monsanto.

163    Dr Crump holds a bachelor’s degree in electrical engineering from Louisiana Tech University, a bachelor’s degree in mathematics from the University of Denver, and a PhD in mathematics with an emphasis in probability theory from Montana State University. In 1980, Dr Crump started a consulting company focussed on providing advice and conducting research on problems related to health risk assessment, and worked for several consulting companies in which he performed quantitative risk assessments, including for exposure to ethylene oxide and inorganic arsenic.

164    Little can be gained by way of an assessment of Dr Crump’s credibility in giving oral evidence because it was obvious during the Conclave E concurrent evidence session that Dr Crump had trouble hearing and comprehending questions put to him in cross-examination. I do not propose to speculate as to why this was the case, but I raised that it may have been partly due to a combination of jetlag and fatigue.

165    The consequence was that senior counsel for Mr McNickle, Ms Szydzik SC, was constrained in her ability to effectively cross-examine Dr Crump on matters squarely falling within his area of expertise and the subject matter of his primary report and published article, Crump (2020). Monsanto, on the other hand, contends that this concern is overstated, and further submitted (in my view, surprisingly) that Dr Crump had little difficulty engaging with what he was asked by Ms Szydzik. I do not accept that submission. The transcript is an incomplete guide to the difficulties as I perceived them unfold while closely observing what was occurring in Court, but Dr Crump’s significant hesitations and evident vexation caused me to ask Dr Crump to leave the witness box and then intervene as follows (T1201.18–23):

HIS HONOUR: … I just raise this issue because I’m just wondering how utile this process is. I’m not convinced that Dr Crump is really understanding [Ms Szydzik’s] questions. Whether he’s tired or jetlagged, or whatever, but he’s obviously having difficulty hearing and I don’t want there to be an unfairness associated with that. Now, I don’t know what the position is. There has been no – I mean, I raise it because I’m concerned about his ability to comprehend what’s being put to him.

166    I did not make this comment without carefully reflecting upon whether it was necessary to intervene. Sometimes, these impressions can only be gauged if one observes what is happening in real time and I thought what was happening could occasion a real unfairness to Dr Crump.

167    I will return to Dr Crump’s evidence below in relation to Conclave E, but I hasten to note that I must bear in mind that there was a limited ability for Mr McNickle to test Dr Crump’s evidence, which must be borne in mind steadily when evaluating his evidence and assessing the ultimate weight the Court ought to place upon it.

168    With these general observations in mind, I come to my findings as to the contested aspects of the epidemiological and biostatistical evidence.

E.3    The Epidemiological and Biostatistical Evidence

169    Mr McNickle contends that the body of epidemiological and biostatistical literature provides generally supportive evidence that exposure to glyphosate and/or GBFs increase an individual’s risk of developing NHL, and therefore that glyphosate and/or GBFs are carcinogenic to humans.

170    As noted earlier (at [133]), Mr McNickle relies upon Professor Checkoway’s conclusion in the JRC (at [132]) that there is “generally supportive evidence that environmental GBF exposures are etiologically related to NHL” (emphasis added). Subject to a matter to which I will return, Professor Checkoway primarily relied upon four individual case-control studies which, in his opinion, indicate an “etiologic contribution” of exposure to GBFs to NHL; namely Hardell and Eriksson (1999); McDuffie (2001); Eriksson (2008); and Meloni (2021) (primary case-control studies). Professor Checkoway also relied upon the pooled studies of Hardell (2002); De Roos (2003); Pahwa (2019); Lee (2004); and De Roos (2022) (pooled case-control studies) in formulating his conclusion in the JRC. It will also be necessary to deal with some other articles referred to in the course of the concurrent evidence session.

171    In dealing with the various studies relied upon by the parties in the epidemiological stream, it is convenient first to set out, at a high level, the key conclusions of each study; secondly, summarise the findings of each of the experts in respect of each study; and thirdly, make findings as to the weight to be attributed or assistance derived from the various studies and expert opinion evidence.

172    In doing so, I will adopt the following structure:

    Section I will detail my findings as to the primary case-control studies;

    Section II will set out my findings in relation to the pooled case-control studies;

    Section III will address some other studies referred to in Conclave C;

    Section IV will set out my findings as to the AHS in the context of Conclave D;

    Section V will detail my findings as to recall and selection bias in Conclave E; and

    Section VI will set out my conclusion as to the epidemiology stream.

I    Primary case-control studies

Hardell and Eriksson (1999)

173    This study was a population-based study in northern and central Sweden encompassing 442 cases and approximately twice as many controls. Exposure data was obtained by questionnaires, which were supplemented by telephone interviews. In total, 404 cases and 741 controls answered the questionnaire.

174    The authors found there was increased risk for NHL for subjects exposed to herbicides, specifically an OR of 2.3 (95% CI: 0.4–13) for ever/never exposure to GBFs (that is, subjects who were ever exposed to GBFs compared to those who were never exposed).

175    Professor Checkoway considered that despite the wide CI, the OR “in terms of its direction” was nonetheless suggestive of an etiologic relation between exposure to GBFs and development of NHL in humans (T917.46). In his analysis, CIs are only a measure of statistical “precision” and are not a “necessary criterion” for assessing causation (at JRC [128]).

176    Further, in cross-examination, Professor Checkoway said that it could be inferred from the fact that the multivariate analysis produced much higher ORs than the univariate analysis that glyphosate exposure was not confounded (that is, a variable whose presence affects the variables being studied so that the results do not reflect the actual relationship) by other pesticides (T914.22–25) (although he acknowledged that it was unclear what variables were controlled for in the multivariate analysis (T915.20)). While Professor Checkoway placed less weight on the multivariate analysis, he nonetheless considered that the data presented was valid and he did not disregard that it showed an elevated OR (T916.3–4; T916.21–22).