1. [2005] CSOH 69.

2. Quite apart from public understanding of the strength of the link between smoking and lung cancer, it is somewhat ironic that a Scottish judge should come to a contrary conclusion at a time when the Scottish Parliament was passing legislation which, to protect public health, would ban smoking in public places.

3. See, eg, Huber, PW Galileo’s Revenge: Junk Science in the Courtroom (New York: Basic Books, 1993).Google Scholar

4. McTear v Imperial Tobacco Limited, above n 1, at 6.184.

5. A Bradford Hill ‘The environment and disease: association or causation?’ (1965) 58 Proc R Soc Med 295.

6. Reay and Hope v BNFL (1995) 5 Med LR 1.

7. The author’s understanding of the continuing debate over rival (frequentist or epistemic) conceptions of probability has been influenced primarily by Ayer’s, AJ Probability and Evidence (Basingstoke: Macmillan, 1972), where, at p 27 Google Scholar, he distinguishes a ‘judgement of credibility’ from the a priori statements of probability (there is a 50% probability that a true coin will land ‘heads’) and from ‘statistical judgements’ such as ‘the probability that a man will die of lung cancer is increased if he is a heavy smoker’.

8. McTear v Imperial Tobacco Limited, above n 1. Section 5 below discusses his decision on ‘individual causation’ within the context of the recent cases which have added to the incoherence in causation in personal injury. It is not suggested that his ruling on ‘general causation’ (or his finding that the defendants were not in breach of duty, ibid, at 7.167–7.181) is part of that incoherence. The quality of the evidence presented in court may indeed have left him with little choice.

9. [2005] UKHL 2, [2005] 2 AC 176. This author argues in ‘Gregg v Scott: loss of chance revisited’ [2006] Law, Probability and Risk doi:10.1093/lpr/mgl002 that Gregg was ultimately concerned with damages for anticipated but, as yet, unrealised harm (viz premature death) and, as such, does little to elucidate the incoherence which this paper seeks to identify. However, Lord Hoffmann and Baroness Hale of Richmond, two of the majority which denied Gregg compensation, argued that any extension of liability for loss of chance should, given the cost implications for insurance companies and the National Health Service, be made by legislation. For another comment, see J Stapleton ‘Loss of chance of a cure from cancer’ (2005) 68 MLR 996.

10. Barker v Corus (UK) plc (formerly Saint Gobain Pipelines plc); Murray v British Shipbuilders (Hydrodynamics) Ltd and Others; Patterson v Smiths Dock Ltd and Others [2006] UKHL 20 (unreported). By abandoning the insistence on joint and several liability for multiple, negligent employers, as applied in Fairchild (below n 24), and by making damages for non-cumulative (or ‘indivisible’) injury proportional to each tortfeasor’s contribution to the aggregate risk (of an actualised mesothelioma), this latest House of Lords’ ruling on personal injury represents a short step towards proportionate liability – a step which is analogous to recent concessions in damages for cumulative (or ‘divisible’) injury; see below nn 64 and 84. But in Barker, the Lords also re-iterated that liability for increased risk of injury should be confined to situations characterised by a ‘single agent of risk’ (see text at n 69 et seq). The implications of this part of the ruling are discussed in s 5 below.

11. [1987] AC 750.

12. J Stapleton ‘The gist of negligence’ [1988] LQR 212 and [1988] LQR 389.

13. [1972] 3 All ER 1008.

14. It is necessary to stress, from the outset, that this paper is not concerned with ‘near misses’, that is, increased risks which do not result in actual harm. Civil liability for ‘pure risk’, where, through good fortune, an injury is avoided, is an irrelevance and only generates further incoherence. It has proved less attractive to judges than to academics; see G Robinson ‘Probabilistic causation and compensation for tortious risk’ (1985) 14 J Leg Stud 779 and JG Fleming ‘Probabilistic causation in tort law’ (1989) 68 La Revue du Barreau Canadien 661.

15. These are Professor Stapleton’s preferred terms; see her article ‘Cause-in-fact and the scope of liability for consequences’ [2003] LQR 388.

16. Bradford Hill, above n 5.

17. R Doll and AB Hill ‘The mortality of doctors in relation to their smoking habits: a preliminary report’ [1954] 1 BMJ 1451.

18. M Susser ‘The logic of Sir Karl Popper and the practice of epidemiology’ (1986) 124 Am J of Epidemiology 711.

19. Bradford Hill, above n 5, at 299. But as an anonymous (epidemiologist) referee points out, it would be hard to accept any study in which the fourth criterion in table 1 (temporality) was not satisfied: a cause must precede its effects.

20. Ibid.

21. [1990] 1 Med LR 117.

22. XYZ and Others v Schering Health Care Ltd [2002] EWHC 1420 (QB), 70 BMLR 88 at [32].

23. Ibid, para [344].

24. Fairchild v Glenhaven Funeral Services Ltd and Others; Fox v Spousal (Midlands) Ltd; Matthews v Associated Portland Cement Manufacturers (1978) Ltd and Another [2002] UKHL 22, [2003] 1 AC 32.

25. In a letter to The Times (22 October 2002), Martyn Day, the claimants’ solicitor criticised the judge’s confidence in one particular industry-funded study ‘published in a second division journal [whose] only scrutiny in the scientific literature eschewed its methodology and conclusions’. In addition, an expert witness called by the claimants wrote a commentary in The Lancet (21 September 2002), which pointed to the statistically significant (p < 0.001) difference between the risk estimates emerging from independently funded studies and those funded by the pharmaceutical industry.

26. Daubert v Merrell Dow Pharmaceuticals, Inc 43 F. 3d 1311 (9th Cir 1995). Congenital deformities were alleged to have been caused by the ‘Bendectin’ drug taken by pregnant women to reduce the effects of morning sickness.

27. See, eg, Edmond, G and Mercer, D Litigation life: law–science knowledge construction in (Bendectin) mass toxic tort litigation’ (2000) 30 Social Studies of Science 265.CrossRefGoogle Scholar

28. S Gold ‘Causation in toxic torts: burdens of proof, standards of persuasion, and statistical evidence’ (1986) 96 Yale LJ 376. Among the many other articles in this area, the author has particularly benefited from M Thompson ‘Causal inference in epidemiology: implications for toxic tort litigation’ (1992) 71 NCL Rev 247.

29. Barnes, DW Too many probabilities: statistical evidence of tort causation’ (2001) 64 Law and Contemporary Problems 191.CrossRefGoogle Scholar

30. A two-sided p value of less than 0.05 is normally equivalent to a 95% confidence interval’s not including the hypothesised null value (viz, in the studies considered in this paper, RR = 1.0) for the population parameter under consideration.

31. J McBaine, in ‘Burden of proof: degrees of belief’ (1944) 32 Cal L Rev 242, stresses the different degrees of belief required for verdicts in civil and criminal trials. However, his claim that ‘the chemist and the physicist in many areas of knowledge can accurately ascertain facts’, at 246, serves as a reminder that his article predates, and would be contested by, the sociology of scientific knowledge.

32. The foregoing paraphrase of Barnes’ description of ‘sample error probability’ needs a little amendment when applied to relative risk and odds ratio, that is, the statistics usually employed in case-control studies (such as those in dispute in both Loveday and Schering). Here, the aim is to identify a number of ‘controls’ whose personal characteristics are, save for the absence of the pathological condition under investigation, as close as possible to that of each ‘case’. Where it is necessary to select from among a number of potential controls, that selection is made randomly. But that does not mean that a case-control study relies on taking random samples from a larger population. Nevertheless, it remains true to say that an odds ratio and its standard error (from which the p value and the confidence interval are derived) are independent and that the latter is inversely proportional to the numbers of cases and controls (see Appendix 1).

33. If we calculate the 95% confidence interval for each sample parameter, then, on average, 95% of such confidence intervals will include the population parameter.

34. See Lewis, D Philosophical Papers vol II (Oxford: Oxford University Press, 1986) p 83 Google Scholar for a discussion of ‘credence about chance’.

35. Merrell-Dow Pharmaceuticals Inc v Havner 953 SW.2d 706 (Tex 1997).

36. Re Agent Orange Liability Litigation 611 F Supp 1223 (1985).

37. Hall v Baxter Healthcare Corp 947 F Supp 1387 (D Or 1996).

38. Manko v United States 636 F Supp 1419 (WD Mo 1986).

39. Marder v GD Searle & Co 630 F Supp 1087 (D Md 1986).

40. Daubert, above n 26.

41. Merrell-Dow, above n 35, at 718.

42. Ibid, at 726.

43. Re Joint Eastern & Southern District Asbestos Litigation 52 F.3d 1124, 1134 (2nd Cir 1995). See also Allen v US 588 F Supp 247 (D Utah 1984); Grassis v Johns-Manville Corp 248 NJ Super 446 (1991).

44. See Freckelton, I Epilogue: dilemmas in proof of causation’ in Freckelton, I and Mendelson, D Causation in Law and Medicine (Dartmouth: Ashgate, 2002) p 450 Google Scholar for an account of Chief Justice Spigelman’s ‘uncomplicated statements of commonsense propositions’ in the interpretation of epidemiological evidence in Seltsam Pty Ltd v McGuiness[2000] NSWCA 29.

45. (1983) 664 P.2d 474.

46. Wright, R Causation, responsibility, risk, probability, naked statistics and proof: pruning the bramble bush by clarifying the concepts’ (1988) 73 Iowa L Rev 1001.Google Scholar

47. Fallibility of the eye-witness of the cab accident is sometimes used as the textbook introduction to Bayes Theorem; see P Hawkins and A Hawkins ‘Lawyer’s probability misconceptions and the implications for legal education’ (1998) 18 LS 316 at 324. For a more advanced discussion, see R Allen and M Redmayne ‘Bayesianism and juridical proof’ (1997) 1 E &P 253 et seq (Special Issue).

48. A mechanical lie-detector’s analysis of a confession would play an analogous role to Professor Wright’s ‘ballistics test’ (which identifies the single ‘guilty’ bullet from among the 100 fired) as the quintessentially ‘particularistic’ evidence on which causation should be based; see above n 46, at 1057.

49. Tribe, LH Trial by mathematics: precision and ritual in the legal process’ (1971) 84 Harvard Law Review 1329 CrossRefGoogle Scholar at 1330 fn 2 quoted by M Parascandola in ‘Chances, individuals and toxic torts’ (1997) 14 J of Applied Philosophy 147 at 153.

50. Response 2 is sometimes referred to as ‘threshold liability’; see Young, R, Faure, M and Fenn, P ‘Causality and causation in tort law’ (2004) 24 Int Rev of Law and Economics 507 CrossRefGoogle Scholar at 517.

51. David Fischer has argued, in ‘Proportional liability: statistical evidence and the probability paradox’ (1993) 46 Vand L Rev 1201 at fn 49, that ‘statistical evidence alone would produce a belief in a proposition only if the probability were extremely high’. Imagine that the blue cab company operates all 20 cabs in the city; the proprietor then gives one cab to his brother-in-law on the condition that he paints it yellow and operates as a separate company. If the particularists maintain their opposition to naked statistics per se, they cannot object to the blue cab company having created, by this ruse, a virtual immunity for itself.

52. See Kaye, D The limits of the preponderance of the evidence standard: justifiably naked statistical evidence and multiple causation 1982 American Bar Foundation Research Journal 487.Google Scholar

53. Ibid, at 500.

54. 26 Cal.3d 588.

55. Hotson v East Berkshire Area Health Authority, above n 11.

56. Stapleton, above n 12.

57. Among the others, the author readily acknowledges his debt to M Stauch ‘Causation, risk and loss of chance in medical negligence’ (1997) 17 OJLS 205 and to Goldberg, R Causation and Risks in the Law of Torts (Oxford: Hart, 1999).Google Scholar

58. Gregg v Scott, above n 9.

59. Newman and Others v Secretary of State for Health (No 1) 54 BMLR 85. Limited space means that we cannot, in this paper, give this case the attention it merits; for a comment, see M Mildred ‘The human growth hormone (Creutzfeldt Jakob Disease) litigation’ [1998] JPIL 251.

60. McGhee v National Coal Board, above n 13, at 1012.

61. Ibid.

62. Ibid, at 1011. Lord Reid, and three of the other four judges, relied upon the ruling in Bonnington Castings Ltd v Wardlaw [1956] 1 AC 613, in which it had agreed that causation was established once a claimant demonstrated that the defendant’s wrongful conduct materially contributed to injury. But in Bonnington the uncertainty related to the amount of the harm (respiratory dysfunction) caused; there was no doubt that some derived from the defendant’s negligence. McGhee can claim to be the more important ‘policy’ decision because it was made in the full awareness that the defendant’s negligence might not, in fact, have been the cause of the harm.

63. Section 323 of the ‘Negligent performance of undertaking to render services Restatement (Second) Torts vol 2 (Philadelphia: American Law Institute, 1965) (cited in Herskovits, above n 45) might be said to offer a similarly unqualified approach to liability for increased risks: ‘One who undertakes, gratuitously or for consideration, to render services to another which he should recognise as necessary for the protection of the other’s person or things, is subject to liability to the other for physical harm resulting from his failure to exercise reasonable care to perform his undertaking, if (a) his failure to exercise such care increases the risk of such harm . . .’. But this wording does not appear in the current draft of the Third Restatement, which sees loss of chance as a specialised issue within medical negligence and one to be considered in future restatements.

64. See above n 10. Although Bonnington, above n 62, strictly applies only to cumulative (non-stochastic) injury, it should be noted that it became the source of a pragmatic tradition (undisturbed by doctrinal issues of probability and statistics) which has, in recent cases, allowed apportionment of damages for non-tortious contribution to occupational harms such as ‘vibration white finger’, in Allen and Others v British Rail Engineering Ltd and Another [2001] EWCA Civ 242, [2001] ICR 942, and asbestosis, in Holtby v Brigham & Cowan (Hull) Ltd [2000] 3 All ER 421. In 2002, the Court of Appeal held, in a case involving psychiatric harm from occupational stress, ‘that it is for the defendant to raise the question of apportionment’; see Hatton v Sutherland [2002] 2 All ER 1 at 20.

65. Hotson v East Berkshire Area Health Authority, above n 11, at 786.

66. Ibid.

67. [1988] 1 AC 1074.

68. McGhee v National Coal Board, above n 13, at 1012. The allegedly offending phrase – ‘unless [the defendant] shows that it had some other cause’ – does not necessarily imply that Lord Wilberforce was proposing that the defendant should bear the burden of proof from the outset of any McGhee-like case. It could be interpreted as simply pointing to the obvious counter which any alert defendant would make, once he senses that the claimant’s account of the cause of her injury appears persuasive to the fact-finder.

69. Wilsher v Essex Area Health Authority [1987] QB 730 at 779.

70. Ibid, at 772.

71. Fairchild v Glenhaven Funeral Services Ltd, above n 24.

72. For example, Summers v Tice (1948) 199 P 2d 1.

73. The ‘single fibre’ theory appeals more to politicians than to medical scientists: ‘We must therefore assume that a single fibre could do real damage which may not even be seen for 20 years or more’: Gummer, J (as Under Secretary of State for Employment on 28 July 1983) ‘Asbestos (Power Stations) Hansard 46 (1983-84) 1411.Google Scholar

74. Fairchild v Glenhaven Funeral Services Ltd, above n 24, para [72]. In Barker, Lord Hoffmann declares his unease in Fairchildto have been ‘wrong’ and he is now, as Lord Rodger of Earlsferry was in Fairchild (ibid, para [170]), more concerned with ‘what counts as being the same agent’: above n 10, para [23].

75. Fairchild v Glenhaven Funeral Services Ltd, above n 24, para [170].

76. There was clearly only one agent of risk in the vCJD case (human growth hormone), above n 59. The House in Barker (above n 10, para [64]) continues to see Hotson as failing the single agent test. When the House originally heard this case (above n 11), it took the view that Hotson’s injury (avascular necrosis of the neck of the femur) resulted from the loss of blood vessels ruptured when he fell from a tree and from those lost later when the bleeding of ruptured blood vessels into the hip joint compressed and blocked some of the vessels that had survived the fall. In other words, there was a pathological condition (contributing to the risk of the injury which ultimately eventuated) which existed prior to the negligent omission; with the passage of time and the absence of potentially remedial treatment following the tort, the original condition generated other contributions to the overall risk. But does the hypothetical possibility of distinguishing between the pre-tort (broken blood vessels) and post-tort (blocked blood vessels) sources of risk justify denying Hotson the benefit enjoyed by McGhee, the logical structure (see text below at n 93) of whose case seems so very similar?

77. Phrases such as ‘in cases like this’ (Lord Reid, above n 13, at 1011), ‘in the circumstances of the present case’ (Lord Salmon at 1017 and 1018) and ‘in this type of case’ (Lord Simon of Glaisdale at 1014) do appear but their contexts do not suggest any express desire to qualify the ‘risk added under conditions of scientific uncertainty’ essence of the ratio.

78. Fitzgerald v Lane [1987] 1 QB 781.

79. Fairchild v Glenhaven Funeral Services Ltd, above n 24, para [170].

80. Fitzgerald v Lane, above n 78, at 800.

81. Page v Smith (unreported) 22 December 1992 (QBD).

82. When this case was heard, for the second time, in the Court of Appeal, Lord Bingham of Cornhill MR criticised Otton J’s reliance on McGhee. He made no reference to a single or similar agent but opined that the case was not ‘concerned with risk at all’: Page v Smith (No 2) [1996] 3 All ER 272 at 275.

83. Re British Coal Respiratory Disease Litigation (unreported) 28 January 1998 (QBD); see also C Miller ‘Coal dust, causation and common sense’ (2000) 63 MLR 763.

84. Ibid, para 119. Turner, J’s ruling is very pragmatic: in citing Thompson v Smiths Shiprepairers (North Shields) [1984] 1 QB 405 as justifying reduced damages to take account of the effect of smoking and non-tortious dust, he is relying on a case concerned with occupational deafness. Since this condition is cumulative in effect (hearing loss is proportional to duration of exposure to noise as well as its intensity) Thompson’s case was another in the line of descent from Bonnington (above nn 62 and 64) for which apportionment of damages is now unproblematic.

85. Nor by this author’s understanding of coal dust as finely divided particles of (chemically inert) silica and tobacco smoke as a suspension of liquid droplets containing a selection of some, reportedly, very toxic chemicals.

86. XYZ and Others v Schering Health Care Ltd, above n 22, para [20].

87. Claimants in the vCJD litigation, had they been given the opportunity, would have been able to pass the single agent test. An attempt to invoke McGhee appears (Mildred, above n 59, at 261) to have been disallowed by the judge considering causation, even though the tortious risk was increased as soon as one (post-breach) injection was given. Note that some claimants were denied because, some 20 years earlier, a meeting, from which a warning letter did not issue, occurred a month or so too late. Arguably, the very raison d’être of McGhee was that such happenstance should play no part in the attribution of liability where a breach of duty has increased the risk of an injury which has occurred.

88. In Hymowitz v Eli Lilly & Co 541 NYS 2d 94 it was stated that diethylstilbestrol (DES, the drug also at issue in Sindell) ‘mimics’ the effect of the natural hormone oestrogen.

89. A similar argument might show that it was not necessary to base background on the risk of VTE posed by second-generation COC takers. If VTE can arise from some naturally occurring malfunction in the endocrine system, then all women carry that risk simply by the fact that they produce female hormones. If the side-effects of third-generation COCs are taken to arise in substantially the same way as the natural malfunctions, then the background risk is the risk of VTE in all women.

90. The incongruity of excluding ‘outcome similarity’ can be illustrated by removing the single agent character from the ‘negligent sportsmen’ paradox. A loaded revolver, a carpenter’s chisel and a knitting needle have very little in common, but each can be used to good (and, let us assume, indistinguishable) effect in puncturing a human artery. If these objects were to be so used in a simultaneous assault by three assassins, then, unless the ‘substantially similar’ rule was deemed to be satisfied, Wilsher rather than Fairchild would be applied in any subsequent civil action and none of the three held liable (because the prob >½ rule would apply and ⅓ < ½).

91. Stapleton, J Lords a’leaping evidentiary gaps’ (2002) 10 Torts Law Journal 276 Google Scholar at 281. The circumstances of Margereson and Hancock v JW Roberts Ltd[1996] Env LR 304 may have been exceptional but these mesothelioma victims recovered for what was unquestionably non-occupational exposure to asbestos.

92. Wright, RW Causation in tort law’ (1985) 73 Cal L Rev 1735 which formalised an approach proposed earlier by Hart, HLA and Honoré, A Causation in the Law (Oxford: Oxford University Press, 2nd edn, 1985).Google Scholar For a recent discussion of the NESS test, see

93. Although when one thinks of the many causes of breathlessness – tortious and non-tortious exposure to occupational dust; allergic reaction to house dust; cigarette smoke (both passive and active exposure); genetic predisposition – the BCRDL case, above n 83, perhaps might appear close to our ‘generic’ case. But table 2 is strictly concerned with stochastic or non-cumulative injury, where increased exposure increases the risk, not the degree, of the injury. If the various causes of breathlessness add to its severity, BCRDL passes into the orbit of Bonnington rather than McGhee.

94. To borrow Professor Stapleton’s phrase, above n 91, at 292.

95. During Mr Barker’s period of self-employment, above n 10.

96. Hubbard, R The aetiology of mesothelioma: are risk factors other than asbestos exposure important?’ (1997) 52 Thorax 496.CrossRefGoogle ScholarPubMed

97. Reading McGhee with the hindsight of Wilsher’s latest re-affirmation (in Barker), it might also be argued that the Lords should not have found in James McGhee’s favour had there been evidence that his dermatitis could have been an allergic reaction to, say, the soap used when he eventually washed off the brick dust.

98. McTear v Imperial Tobacco Limited, above n 1, at 6.171.

99. Ibid, at 6.184 and the word ‘fallacious’ appears in the conclusions (on individual causation) at 9.10.

100. See text at n 66 above.

101. McTear v Imperial Tobacco Limited, above n 1, at 6.180.

102. Sindell v Abbott Labs, above n 54.

103. The tendency of some American courts to treat causation as ‘a given’ whenever ‘a rare tumor is closely associated with a toxic substance’ is criticised by T Brennan in ‘Causal chains and statistical links: the role of scientific uncertainty in hazardous-substance litigation’ (1988) 73 Cornell L Rev 469 at 500. Goldberg, above n 57, at 56–57, fnn 3–4, discusses some of the epidemiological evidence establishing (and questioning) the link between DES and vaginal cancer. He cites one of the early studies (Herbst, Ulfeder and Poskanzer ‘Association of maternal stilbestrol therapy with tumor appearance in young women’ (1971) 284 New England J Med 878 at 878–879) in which 7 out of 8 mothers of patients with vaginal adenocarcinoma had taken DES during early pregnancy.

104. ‘Insofar as [McTear] might in any sense be a special case, it fell within the reasoning of Wilsher’: above n 1, at 6.25.

105. McTear follows an unsuccessful action against Imperial Tobacco Ltd in the English High Court, Hodgson v Imperial Tobacco Ltd (No 3) [1999] CLY 459. These failures in British courts stand in marked contrast to the success of the Attorneys General of 46 (US) states in securing the 1998 Master Settlement Agreement, under which the four largest tobacco companies in the USA will pay, over the next 25 years, in excess of $200 billion to cover the estimated costs of treating smoking-related illnesses. See also M Day ‘Tobacco litigation’ (2006) 1 JPIL 1.

106. One of Professor Wright’s objection to Sindell is that, because there were relatively few DES daughters, ‘there will be only a very rough correlation between the injuries actually caused in the aggregate by a particular firm and those for which it is held liable under the probabilistic market-share approach’ (above n 92, at 1819). This would not apply if our hypothetical scheme were to be adopted in the UK, where smoking is implicated in 46,500 cancers (of all types) per annum, according to official statistics cited in McTear, above n 1, at 2.67.

107. The figure of 10% refers to the proportion of all lung cancers in the population not related to smoking. The proportion of lung cancers suffered by smokers but which are not attributable to their smoking habit will be less than 10%. Any apportionment should be based upon this lower figure. The author is once again grateful to an anonymous (epidemiologist) referee for this point.

108. Barker v Corus, above n 10, para [44].

109. Fairchild v Glenhaven Funeral Services Ltd, above n 24, para [74], where he suggests that Sindell falls outside the remit of McGhee because ‘the existence of additional manufacturers did not materially increase the risk of injury’. But mutatis mutandis the risk borne by a mesothelioma victim is not increased if, rather than working for Y years for one negligent employer, he works Y ÷ E years for E employers, each one of whose control over exposure to asbestos is negligent to the same extent. Sindell was seen as a close analogue of the ‘negligent sportsmen’ paradigm, which implicitly involves a ‘single agent’. So it is hardly surprising that Fairchild, which also had elements of the negligent sportsmen paradox, should similarly invite consideration of its ‘single agent’ characteristics and, in turn, of its affinity with McGhee (as interpreted in Wilsher).

110. See above n 103.

111. Barker v Corus, above n 10, para [13].

112. Armitage, P and Berry, G Statistical Methods in Medical Research (Oxford: Blackwell, 3rd edn, 1994) p 131.Google Scholar