SUPREME COURT OF CANADA

Appeal heard: April 28, 2016 Judgment rendered: November 10, 2016 Docket: 36291

Between:

Albert Benhaim and Michael O’Donovan

Appellants

and

Cathie St-Germain, personally

and in her capacity as tutor to her minor son,

whose name is being kept confidential,

and in her capacity as universal legatee

of the late Marc Émond

Respondent

Coram: McLachlin C.J. and Abella, Karakatsanis, Wagner, Gascon, Côté and Brown JJ.

Reasons for Judgment: (paras. 1 to 87) Dissenting Reasons: (paras. 88 to 135)

Wagner J. (McLachlin C.J. and Karakatsanis and Gascon JJ. concurring) Côté J. (Abella and Brown JJ. concurring)

Benhaim v. St‑Germain, 2016 SCC 48, [2016] 2 S.C.R. 352

Albert Benhaim and

Michael O’Donovan Appellants

v.

Cathie St‑Germain, personally and

in her capacity as tutor to her minor son,

whose name is being kept confidential, and in her capacity

as universal legatee of the late Marc Émond Respondent

Indexed as: Benhaim v. St‑Germain

2016 SCC 48

File No.: 36291.

2016: April 28; 2016: November 10.

Present: McLachlin C.J. and Abella, Karakatsanis, Wagner, Gascon, Côté and Brown JJ.

on appeal from the court of appeal for quebec

Civil liability — Medical malpractice — Negligence — Causation — Evidence — Presumption of fact — Physicians negligently delaying patient’s cancer diagnosis — Patient dying of lung cancer — Physicians’ negligence undermining plaintiff’s ability to prove causation — Trial judge refusing to apply presumption of fact in favour of causation and finding causation not established — Whether trial judge required to draw adverse inference of causation or apply presumption of fact where defendant’s negligence undermines plaintiff’s ability to prove causation and where at least some evidence of causation adduced — Whether Court of Appeal justified in reversing trial judge’s decision on basis of error of law — Whether trial judge committed palpable and overriding error in appreciation of facts — Civil Code of Québec, art. 2849 .

E, a non‑smoker who exercised regularly and took care of himself, died tragically of lung cancer at the age of 47. His partner, in her own name, in her capacity as tutor to her son, and as E’s universal legatee, brought an action against E’s physicians. She alleged that the negligent delay in diagnosing E’s cancer caused his death. The physicians argued that the cancer would likely have taken E’s life even if he had been promptly diagnosed, and therefore, that the delay in diagnosing him was not the cause of his death. At trial, the three expert witnesses formed opinions on the basis of incomplete information and each opinion involved some degree of speculation and estimation as to the staging of E’s lung cancer. The trial judge allowed the action in part on the basis that while E’s physicians were both negligent, their negligence did not cause E’s death. In coming to this conclusion, the trial judge recognized that she could draw an adverse inference of causation against the physicians because their negligence made it impossible to prove causation, but she drew no such inference. Damages were only awarded to E’s partner personally and in her capacity as universal legatee for the anguish caused by the physicians’ negligent handling of E’s treatment. The Quebec Court of Appeal reversed that decision. The majority held that the trial judge erred in law by failing to draw an adverse inference of causation. The concurring judge concluded that the trial judge should have found that causation had been established.

Held (Abella, Côté and Brown JJ. dissenting): The appeal should be allowed.

Per McLachlin C.J. and Karakatsanis, Wagner and Gascon JJ.: A trier of fact is not required to draw an adverse inference of causation or apply a presumption of fact as defined in art. 2849 of the Civil Code of Québec in medical liability cases where the defendant’s negligence undermines the plaintiff’s ability to prove causation and where the plaintiff adduces at least some evidence of causation. This Court’s decisions in Snell v. Farrell , 1990 70 (SCC) , [1990] 2 S.C.R. 311, and St‑Jean v. Mercier , 2002 SCC 15 , [2002] 1 S.C.R. 491, make it clear that in such circumstances, an adverse inference of causation is one that trial judges are permitted to draw. It is not one they are required to draw. In Quebec civil law, the adverse inference discussed in Snell constitutes nothing more than a presumption of fact as defined in art. 2849 , and the principles laid down in Snell for drawing inferences as part of the ordinary fact‑finding process apply to triers of fact making determinations on causation. The principles must not be interpreted in a manner that alters the burden of proof or departs from the criteria for establishing presumptions of fact. In weighing the evidence, trial judges may consider the ability of the parties to produce evidence. Whether the inference or presumption arises on the facts must be assessed according to a legal, not a scientific, standard. Furthermore, because an adverse inference of causation is a component of the fact‑finding process, the decision as to whether the inference is warranted in a particular case falls within the discretion of the trier of fact, to be determined with reference to all of the evidence, and is reviewable on the stringent standard of palpable and overriding error.

In this case, despite using permissive language to describe the adverse inference discussed in Snell , the Court of Appeal failed to give effect to its discretionary nature. Indeed, by reversing the trial judge’s decision on the basis of an error of law the court wrongly treated the inference as compulsory. The trial judge did not commit an error of law in applying the rules of evidence. She applied St‑Jean , pursuant to which presumptions of causation can be applied only when they are serious, precise and concordant. She did not think that these criteria were met in this case, as she chose to believe the physicians’ expert over the plaintiff’s experts. She was not required by law to apply a presumption of fact against the physicians simply because (i) it was impossible to prove causation as a result of the physicians’ fault; and (ii) the plaintiff adduced some affirmative evidence that the physicians’ fault was linked to the loss.

The Court of Appeal also failed to show deference to the trial judge’s weighing of the evidence. The trial judge did not improperly rely on the speculative expert evidence adduced in this case at the expense of statistical evidence. Trial judges are empowered to make legal determinations even where medical experts are not able to express an opinion with certainty. Moreover, while courts may take statistics into account when determining causation, statistical evidence should be approached with caution — it is not determinative. It is also for the trial judge to decide what weight, if any, to give to statistical evidence, and drawing an inference from such evidence is an inherent, and often implicit, part of the fact‑finding process. It must be interpreted in light of the whole of the evidence, and that interpretation is entitled to considerable deference on appeal. In this case, the trial judge did not commit a palpable and overriding error in relying on the opinion of an expert who acknowledged the uncertainty in his opinion. She carefully weighed the evidence as a whole, including the statistical evidence, the evidence specific to E, and the three expert opinions, all of which involved some speculation. She made no palpable and overriding error in finding that the plaintiff had failed to establish causation on a balance of probabilities.

Per Abella, Côté and Brown JJ. (dissenting): There is agreement with the majority that the Court of Appeal erred in characterizing as an error of law the trial judge’s failure to draw an adverse inference or to apply a factual presumption under the Code . However, the Court of Appeal did not misstate the rule described in Snell and St‑Jean . Rather, it correctly summarized Snell and held that judges are permitted to draw an unfavourable inference in some circumstances.

Nevertheless, the Court of Appeal’s mischaracterization of the trial judge’s failure to draw an adverse inference is of no consequence to the outcome of the case, because that failure constituted a palpable and overriding error. The trial judge committed three errors in her understanding of the evidence. First, she misconstrued the physicians’ expert’s testimony. Second, she omitted key objective evidence, namely the fact that E survived more than 31 months even though the life expectancy of patients diagnosed with stage III to stage IV lung cancer is 8 to 12 months. This evidence was supported by uncontested statistical data, and while such data should not be the sole basis for drawing an inference or applying a presumption of fact, it may help to confirm the trial judge’s factual determinations. Third, she erred in the resulting inference‑drawing process itself. Had the trial judge disregarded the highly speculative facts on which the physicians’ expert’s testimony was based, and had she taken into account E’s survival period, she would have drawn an inference of causation. It was a palpable and overriding error not to apply the presumption in art. 2849 of the Code . In light of the trial judge’s errors, it was the role of the Court of Appeal to intervene and reweigh the evidence.

Cases Cited

By Wagner J.

Referred to: Snell v. Farrell , 1990 70 (SCC) , [1990] 2 S.C.R. 311; St‑Jean v. Mercier , 2002 SCC 15 , [2002] 1 S.C.R. 491; Housen v. Nikolaisen , 2002 SCC 33 , [2002] 2 S.C.R. 235; Clements v. Clements , 2012 SCC 32 , [2012] 2 S.C.R. 181; Ediger v. Johnston , 2013 SCC 18 , [2013] 2 S.C.R. 98; South Yukon Forest Corp. v. R. , 2012 FCA 165 , 4 B.L.R. (5th) 31; J.G. v. Nadeau , 2016 QCCA 167 ; McGhee v. National Coal Board , [1973] 1 W.L.R. 1 ; Wilsher v. Essex Area Health Authority , [1988] 2 W.L.R. 557 ; Laferrière v. Lawson , 1991 87 (SCC) , [1991] 1 S.C.R. 541; Sentilles v. Inter‑Caribbean Shipping Corp. , 361 U.S. 107 (1959); Blatch v. Archer (1774), 1 Cowp. 63, 98 E.R. 969 ; Prud’homme v. Prud’homme , 2002 SCC 85 , [2002] 4 S.C.R. 663; Hinse v. Canada (Attorney General) , 2015 SCC 35 , [2015] 2 S.C.R. 621.

By Côté J. (dissenting)

St‑Jean v. Mercier , 2002 SCC 15 , [2002] 1 S.C.R. 491; Housen v. Nikolaisen , 2002 SCC 33 , [2002] 2 S.C.R. 235; H.L. v. Canada (Attorney General) , 2005 SCC 25 , [2005] 1 S.C.R. 401; Snell v. Farrell , 1990 70 (SCC) , [1990] 2 S.C.R. 311; Ediger v. Johnston , 2013 SCC 18 , [2013] 2 S.C.R. 98; Martel v. Hôtel‑Dieu St‑Vallier , 1969 3 (CSC) , [1969] S.C.R. 745; Schreiber Brothers Ltd. v. Currie Products Ltd. , 1980 11 (SCC) , [1980] 2 S.C.R. 78; Laurentide Motels Ltd. v. Beauport (City) , 1989 81 (SCC) , [1989] 1 S.C.R. 705; J.G. v. Nadeau , 2016 QCCA 167 .

Statutes and Regulations Cited

Civil Code of Québec , arts. 2846 to 2849.

Code of Civil Procedure , CQLR, c. C‑25.01.

Authors Cited

Cohen, L. Jonathan. The Probable and the Provable . Oxford: Clarendon Press, 1977.

Jutras, Daniel. “Expertise scientifique et causalité”, dans Congrès annuel du Barreau du Québec (1992) . Montréal: Service de la formation permanente, Barreau du Québec, 1992, 897.

Khoury, Lara. Uncertain Causation in Medical Liability . Cowansville, Qué.: Yvon Blais, 2006.

Larombière, M. L. Théorie et pratique des obligations , nouv. éd., t. 7. Paris: A. Durand et Pedone‑Lauriel, 1885.

Linden, Allen M., and Bruce Feldthusen. Canadian Tort Law , 10th ed. Toronto: LexisNexis, 2015.

Wright, Richard W. “Causation, Responsibility, Risk, Probability, Naked Statistics, and Proof: Pruning the Bramble Bush by Clarifying the Concepts” (1988), 73 Iowa L. Rev. 1001.

APPEAL from a judgment of the Quebec Court of Appeal (Kasirer, Fournier and Bélanger JJ.A.), 2014 QCCA 2207 , 16 C.C.L.T. (4th) 190, [2014] AZ‑51130892, [2014] J.Q. n o 13772 (QL), 2014 CarswellQue 12131 (WL Can.), setting aside in part a decision of Marcotte J., 2011 QCCS 4755 , [2011] AZ‑50785409, [2011] J.Q. n o 12173 (QL), 2011 CarswellQue 10033 (WL Can.). Appeal allowed, Abella, Côté and Brown JJ. dissenting.

David E. Platts and Élisabeth Brousseau , for the appellants.

Gordon Kugler and Stuart Kugler , for the respondent.

The judgment of McLachlin C.J. and Karakatsanis, Wagner and Gascon JJ. was delivered by

Wagner J. —

I. Introduction

[ 1 ] In some professional liability cases, the defendant’s negligence may undermine the plaintiff’s ability to prove causation. The plaintiff may nonetheless lead some affirmative evidence of causation. In these circumstances, is the trier of fact required to draw an adverse inference of causation against the defendant? That is the legal issue at the heart of this appeal.

[ 2 ] Marc Émond died tragically of lung cancer at the age of 47, leaving behind his partner, Cathie St-Germain, and their young son. Ms. St-Germain brought an action for damages against the appellants, two of Mr. Émond’s physicians.

[ 3 ] The trial judge concluded that, while the appellant physicians were negligent, the evidence did not establish on a balance of probabilities that their negligence caused Mr. Émond’s death. The Quebec Court of Appeal reversed that decision. The majority held that the trial judge erred in law by failing to draw an adverse inference of causation. The concurring judge concluded that the trial judge should have found that causation had been established.

[ 4 ] The Civil Code of Québec and the Code of Civil Procedure , CQLR, c. C-25.01, provide a comprehensive set of evidentiary rules. The rules on presumptions set out in arts. 2846 to 2849 of the Civil Code govern the inferences drawn by triers of fact. In my opinion, no rule of law requires the trier of fact to draw an adverse inference of causation where the defendant’s negligence has undermined the plaintiff’s ability to prove causation, even where there is some evidence of causation.

[ 5 ] In this case, the trial judge committed no error of law in her causation analysis, nor did she commit a palpable and overriding error of fact. I would, therefore, allow the appeal.

II. Facts

[ 6 ] In November 2005, Mr. Émond appeared to be in excellent health. He was then a 44-year-old non-smoker who exercised regularly and took care of himself. He also had annual physicals at the Clinique Physimed, where he was under the care of the appellant Dr. Albert Benhaim, a general practitioner.

[ 7 ] Mr. Émond attended his annual physical on November 9, 2005. Dr. Benhaim suggested that Mr. Émond undergo a chest X-ray, even though he did not have symptoms of a chest problem.

[ 8 ] The appellant Dr. Michael O’Donovan, a radiologist, reviewed Mr. Émond’s chest X-ray. According to Dr. O’Donovan’s report, there was “a 1.5 to 2 cm ill-defined opacity” in Mr. Émond’s right lung, “the etiology of which is uncertain”. There were no other associated abnormalities on the X-ray. Dr. O’Donovan suggested that, for purposes of comparison, Dr. Benhaim consult Mr. Émond’s previous chest X-rays if they were available. Otherwise, Dr. O’Donovan suggested a follow-up X-ray and “probably a CT scan”.

[ 9 ] Contrary to Dr. O’Donovan’s suggestion, Dr. Benhaim did not attempt to find Mr. Émond’s previous chest X-rays. Nor did he consult Mr. Émond’s medical file, which contained reports on chest X-rays taken in 1994, 1998 and 1999. However, Dr. Benhaim had a follow-up chest X-ray performed on January 17, 2006.

[ 10 ] Dr. O’Donovan reviewed the follow-up X-ray, and noted no change in the opacity in Mr. Émond’s right lung. He suspected that the opacity was a chronic lesion and suggested that Mr. Émond undergo a third X-ray in four months’ time. That X-ray was never taken.

[ 11 ] Mr. Émond had his next annual physical on December 4, 2006, and an X-ray of his chest was taken. Dr. O’Donovan observed that the lesion in Mr. Émond’s lung had increased in size to approximately 2.5 cm. Dr. O’Donovan suspected that the lesion might be cancerous, and he suggested further tests.

[ 12 ] Mr. Émond underwent a CT scan on December 19, 2006 and a PET scan in January 2007. Those tests confirmed that he had stage IV lung cancer. He was diagnosed in January 2007. The cancer was incurable. At that time, however, Mr. Émond still displayed no symptoms of his disease.

[ 13 ] Beginning in February 2007, Mr. Émond received palliative chemotherapy with the hope of slowing the cancer’s spread. The chemotherapy treatments were suspended between June and November 2007, as Mr. Émond was feeling well. He exercised regularly and biked over 5,000 km that summer. Mr. Émond began to experience symptoms of his disease in the fall of 2007. The chemotherapy treatments resumed in November 2007, but were stopped again in February 2008 because they were unsuccessful.

[ 14 ] Mr. Émond died on June 6, 2008. He was survived by his partner, Ms. St-Germain, and their eight-year-old son. Ms. St-Germain brought the present action against Drs. Benhaim and O’Donovan in her own name, in her capacity as tutor to her son, and as Mr. Émond’s universal legatee.

[ 15 ] At trial, the plaintiff argued that, in November 2005 or at least in January 2006, Drs. Benhaim and O’Donovan should have informed Mr. Émond that the opacity on the X-ray of his right lung could be cancerous, and should have conducted further tests to determine whether that was the case. The plaintiff’s expert witnesses testified that, in November 2005, the cancer was at stage I or IIA. Had Mr. Émond been promptly diagnosed and treated at that time, he would likely have been cured. However, due to the defendants’ errors, Mr. Émond was not diagnosed until January 2007, when the cancer was at stage IV and incurable. Therefore, the plaintiff argued, the negligent delay in diagnosing Mr. Émond’s cancer caused his death.

[ 16 ] The defendants’ expert, Dr. Ferraro, testified that Mr. Émond’s cancer was already at stage III or IV in November 2005, and his chances of survival at that time were low. Therefore, the cancer would likely have taken Mr. Émond’s life even if he had been promptly diagnosed. The delay in diagnosing Mr. Émond was not the cause of his death.

III. Judgments Below

A. Quebec Superior Court (Marcotte J., 2011 QCCS 4755 )

[ 17 ] After thoroughly reviewing the evidence in her well-crafted reasons, the trial judge, Marcotte J. (as she then was), allowed the action in part.

[ 18 ] She found that Drs. Benhaim and O’Donovan were both negligent in failing to more thoroughly investigate the opacity on Mr. Émond’s chest X-rays of November 2005 and January 2006. Because of the negligence, Mr. Émond’s cancer diagnosis was delayed to January 2007. The trial judge’s findings of negligence are not contested in this appeal.

[ 19 ] However, the trial judge concluded that the evidence did not establish on a balance of probabilities that the negligence of Drs. Benhaim and O’Donovan caused Mr. Émond’s death. She was not convinced that, but for the delay in diagnosing Mr. Émond’s cancer, he would likely have survived. She found that, by November 2005 and January 2006, Mr. Émond’s cancer was likely already at stage III, and was likely incurable.

[ 20 ] The trial judge stated that [ translation] “[m]oreover, no presumption of causation can be drawn on the basis that the defendants’ fault made it impossible for the plaintiff to prove causation” (para. 92 ()). Instead, she recognized that the court may draw [ translation] “an adverse inference” against the defendant in such circumstances (para. 100).

[ 21 ] The trial judge rejected the plaintiff’s expert evidence that the cancer was at stage I in November 2005. This evidence was based on three factors: the lesion in Mr. Émond’s lung was less than 3 cm in size, Mr. Émond was asymptomatic, and he was in a good state of health. The trial judge found that these three factors were also present in December 2006 and January 2007, when Mr. Émond was diagnosed with stage IV cancer. They were unreliable indicators of the cancer’s progression.

[ 22 ] The trial judge accepted the opinion of the defendants’ expert, Dr. Ferraro, who testified that the cancer was at least at stage III in November 2005. Dr. Ferraro’s opinion was based on three considerations. First, Mr. Émond’s lung cancer could not have progressed from stage I to stage IV in approximately 12 months (November 2005 to December 2006) because lung cancer evolves slowly. Second, when the November 2005 X-ray was viewed with the benefit of hindsight — that is, with knowledge of the results of the tests performed in December 2006 and January 2007 — it was apparent that there were shadows on the November 2005 X-ray that were consistent with stage III or IV cancer. Third, the majority of patients diagnosed with cancer are at an advanced stage.

[ 23 ] The trial judge considered the evidence that patients with untreated stage III or IV lung cancer typically live for one year or less. One of the plaintiff’s experts testified that if Mr. Émond had stage III or IV cancer in November 2005, he would not have been alive for his next annual physical in December 2006. However, the trial judge emphasized Dr. Ferraro’s evidence that Mr. Émond’s case defied statistics.

[ 24 ] Accordingly, the trial judge concluded that the faults of Drs. Benhaim and O’Donovan did not cause Mr. Émond’s death. However, she awarded $70,000 in damages to Ms. St-Germain personally and in her capacity as universal legatee for the anguish caused by the defendants’ negligent handling of Mr. Émond’s treatment.

B. Quebec Court of Appeal (Kasirer, Fournier and Bélanger JJ.A., 2014 QCCA 2207 , 16 C.C.L.T. (4th) 190)

[ 25 ] The Court of Appeal allowed the appeal and concluded that the faults of Drs. Benhaim and O’Donovan caused Mr. Émond’s death.

(1) Majority Reasons (Kasirer and Bélanger JJ.A.)

[ 26 ] The majority of the Court of Appeal, Kasirer and Bélanger JJ.A., concluded in a well-crafted decision that the trial judge erred in law in her causation analysis. Her error of law was her failure to draw an unfavourable or adverse inference of causation that, according to Kasirer and Bélanger JJ.A., was required by this Court’s decisions in Snell v. Farrell , 1990 70 (SCC) , [1990] 2 S.C.R. 311, and St-Jean v. Mercier , 2002 SCC 15 , [2002] 1 S.C.R. 491.

[ 27 ] In the view of Kasirer and Bélanger JJ.A., an adverse inference of causation arises when two criteria are established. First, the defendant physician’s negligence must have undermined the plaintiff’s ability to prove causation. [1] Second, the plaintiff must adduce at least “some” or “very little affirmative evidence” of causation. These two criteria form the basis for an adverse inference against the defendant that discharges, prima facie , the plaintiff’s burden of proving causation. It is then open to the defendant to rebut the inference of causation by leading evidence to the contrary. The majority stressed that drawing an adverse inference of causation is not the same as shifting the onus of disproving causation to the defendant. The plaintiff retains the “ultimate burden of showing causation on the balance of probabilities” (C.A. reasons, at para. 168).

[ 28 ] In this case, the majority of the Court of Appeal concluded that the criteria for drawing an adverse inference of causation were established.

[ 29 ] First, the defendants’ negligence undermined the plaintiff’s ability to prove causation. The defendants’ negligence consisted of their failure to investigate the opacity on Mr. Émond’s chest X-rays. Had they investigated and conducted the appropriate tests, they would have determined the stage of Mr. Émond’s cancer in November 2005, which would in turn have determined whether his cancer could likely have been cured with prompt treatment. Therefore, because of the defendants’ negligence, “it was impossible for the [plaintiff] to show scientifically, by direct evidence of staging of cancer, that the fault resulted in a delay in the treatment of the disease that ultimately caused Mr. Émond’s death” (C.A. reasons, at para. 142).

[ 30 ] Second, the plaintiff led some affirmative evidence that the cancer was at an early stage in November 2005, and thus could likely have been cured with prompt diagnosis and treatment. This evidence consisted of a statistic that 78 percent of cancers discovered fortuitously — as Mr. Émond’s cancer was — are at stage I. Stage I cancer has a cure rate of 70 percent.

[ 31 ] According to Kasirer and Bélanger JJ.A., these two factors gave rise to an adverse inference against the defendants that their negligence caused Mr. Émond’s death. It was open to the defendants to lead evidence to rebut the adverse inference — evidence establishing that Mr. Émond was among the 22 percent of persons whose fortuitously discovered cancer is at a late, incurable stage — but they failed to do so. In particular, the majority of the Court of Appeal found Dr. Ferraro’s evidence about the slow progression of lung cancer, and his retrospective reading of the November 2005 X-ray, to be speculative and unreliable.

[ 32 ] The majority of the Court of Appeal concluded that the trial judge erred in law by failing to apply an adverse inference of causation. Had she done so, she would have found that causation had been established on a balance of probabilities.

(2) Concurring Reasons (Fournier J.A.)

[ 33 ] Fournier J.A. wrote concurring reasons. He found that the trial judge committed a palpable and overriding error when she concluded that causation was not established. He relied principally on the evidence that untreated stage III or IV cancer is generally fatal within 12 months. Mr. Émond received no treatment between November 2005 and January 2007. Therefore, if he had stage III or IV cancer in November 2005, he would likely have died by January 2007. In fact, Mr. Émond survived to June 2008. As a result, in November 2005, Mr. Émond must have had stage I or II cancer, which would likely have been curable with prompt treatment.

[ 34 ] Fournier J.A. acknowledged Dr. Ferraro’s testimony that, when the November 2005 X-ray was viewed with the benefit of hindsight, there were shadows that were consistent with stage III or IV cancer. However, neither the plaintiff’s experts nor Dr. O’Donovan could see these shadows. In the view of Fournier J.A., the shadows alone were not enough to undermine the plaintiff’s theory of causation, as X-rays are of limited use in diagnosing the staging of cancer.

[ 35 ] Accordingly, Fournier J.A. found that the plaintiff had proved on a balance of probabilities that the defendants’ fault caused Mr. Émond’s death.

IV. Analysis

A. Standard of Review

[ 36 ] The standard of review is correctness for questions of law, and palpable and overriding error for findings of fact and inferences of fact: Housen v. Nikolaisen , 2002 SCC 33 , [2002] 2 S.C.R. 235, at paras. 8 , 10 and 19; St-Jean , at paras. 33-36 . Causation is a question of fact, and so the trial judge’s finding on causation is owed deference on appeal: St-Jean , at paras. 104-5 ; Clements v. Clements , 2012 SCC 32 , [2012] 2 S.C.R. 181, at para. 8 ; Ediger v. Johnston , 2013 SCC 18 , [2013] 2 S.C.R. 98, at para. 29 .

[ 37 ] It may be useful to recall the many reasons why appellate courts defer to trial courts’ findings of fact, which were described at length in Housen , at paras. 15-18 . Deference to factual findings limits the number, length and cost of appeals, which in turn promotes the autonomy and integrity of trial proceedings. Moreover, the law presumes that trial judges and appellate judges are equally capable of justly resolving disputes. Allowing appellate courts free rein to overturn trial courts’ factual findings would duplicate judicial proceedings at great expense, without any concomitant guarantee of more just results. Finally, according deference to a trial judge’s findings of fact reinforces the notion that they are in the best position to make those findings. Trial judges are immersed in the evidence, they hear viva voce testimony, and they are familiar with the case as a whole. Their expertise in weighing large quantities of evidence and making factual findings ought to be respected. These considerations are particularly important in the present case because it involves a large quantity of complex evidence.

[ 38 ] It is equally useful to recall what is meant by “palpable and overriding error”. Stratas J.A. described the deferential standard as follows in South Yukon Forest Corp. v. R. , 2012 FCA 165 , 4 B.L.R. (5th) 31, at para. 46 :

Palpable and overriding error is a highly deferential standard of review . . . . “Palpable” means an error that is obvious. “Overriding” means an error that goes to the very core of the outcome of the case. When arguing palpable and overriding error, it is not enough to pull at leaves and branches and leave the tree standing. The entire tree must fall.

[ 39 ] Or, as Morissette J.A. put it in J.G. v. Nadeau , 2016 QCCA 167 , at para. 77 (), [ translation] “a palpable and overriding error is in the nature not of a needle in a haystack, but of a beam in the eye. And it is impossible to confuse these last two notions.”

[ 40 ] However, the majority of the Court of Appeal did not review the trial judge’s finding on causation for palpable and overriding error. After acknowledging that causation is a question of fact, the majority stated that the trial judge’s failure to apply an adverse inference of causation constituted an error of law. I now turn to that alleged error of law.

B. Did the Trial Judge Err in Law by Failing to Draw an Adverse Inference of Causation?

[ 41 ] The majority of the Court of Appeal found that the trial judge erred in law by failing to draw an adverse inference of causation against the defendants in accordance with this Court’s decisions in Snell and St-Jean . According to the majority, these decisions require a trier of fact to draw an adverse inference of causation against a defendant physician where the defendant’s negligence has undermined the plaintiff’s ability to prove causation and where the plaintiff adduces at least “some” or “very little affirmative evidence” of causation. This adverse inference serves, prima facie , to discharge the plaintiff’s burden of proving causation. The defendant may resist the inference by leading evidence to the contrary.

[ 42 ] This was an error. This Court held in Snell that, in such circumstances, an adverse inference of causation may discharge the plaintiff’s burden of proving causation. Those circumstances do not trigger such an inference. Whether an inference of causation is warranted, and how it is to be weighed against the evidence, are matters for the trier of fact. Here, the trial judge’s reasons show that she did not draw that inference, although she was aware that it was available (para. 100). That conclusion is a question of fact and deserves deference from a court of appeal.

[ 43 ] The majority of the Court of Appeal appropriately used permissive language to explain this Court’s decision in Snell . By reversing the trial judge’s decision on the basis of an error of law, however, the majority treated the inference described in Snell and St-Jean as compulsory once certain criteria were established.

[ 44 ] The majority of the Court of Appeal evaluated the evidence in a manner that illustrates this error. It approached the evidence from the perspective that, since the defendants had negligently created the uncertainty, it was for them to rebut the single piece of statistical evidence relating to the fortuitous discovery of cancer. Since it found that the defendants’ evidence was speculative as a consequence of that very uncertainty, the majority of the Court of Appeal held that the adverse inference had to result in a finding of causation. The majority of the Court of Appeal did not consider the weaknesses in the plaintiff’s expert evidence, as the trial judge had, because it rested the adverse inference on the statistical evidence. With respect for the contrary view, Snell and St-Jean do not support this approach. Rather, the decision to draw an adverse inference must be based on an evaluation of all of the evidence. To do otherwise has the same effect as impermissibly reversing the burden of proof. I return to this issue below.

[ 45 ] As I will now explain, Snell and St-Jean held that the ordinary rules of causation must be applied in medical malpractice cases. As prime examples of how the ordinary rules of causation operate in medical liability cases, these decisions are equally relevant in Quebec.

[ 46 ] I will begin by discussing Snell , in which Sopinka J. examined developments in English tort law that purported to reverse the onus of proving causation in some circumstances. [2] Traditionally, the plaintiff in a common law negligence claim had to prove on a balance of probabilities that, but for the defendant’s negligent conduct, the plaintiff would not have been injured. Sopinka J. stated that his task was “to determine whether a departure from well-established principles is necessary for the resolution of this appeal” (p. 320). Sopinka J. concluded that such a departure was not warranted, provided that the traditional principles are not applied in an overly rigid manner (p. 328).

[ 47 ] First, Sopinka J. held that it is not necessary that the plaintiff adduce expert scientific or medical evidence definitively supporting the plaintiff’s theory of causation, as “[c]ausation need not be determined by scientific precision” (p. 328; see also pp. 330-31). This is because the law requires proof of causation only on a balance of probabilities, whereas scientific or medical experts often require a higher degree of certainty before drawing conclusions on causation (p. 330). Simply put, scientific causation and factual causation for legal purposes are two different things. Factual causation for legal purposes is a matter for the trier of fact, not for the expert witnesses, to decide: Laferrière v. Lawson , 1991 87 (SCC) , [1991] 1 S.C.R. 541, at pp. 607-8; see also Sentilles v. Inter-Caribbean Shipping Corp. , 361 U.S. 107 (1959), at pp. 109-10.

[ 48 ] Second, in medical malpractice cases, the defendant is often in a better position than the plaintiff to determine the cause of the injury (p. 322). Sopinka J. held that, in such circumstances, the trier of fact may take into account the relative ability of each party to present evidence on a fact in issue:

The legal or ultimate burden remains with the plaintiff, but in the absence of evidence to the contrary adduced by the defendant, an inference of causation may be drawn although positive or scientific proof of causation has not been adduced. If some evidence to the contrary is adduced by the defendant, the trial judge is entitled to take account of Lord Mansfield’s famous precept. [p. 330]

This precept, stated by Lord Mansfield in Blatch v. Archer (1774), 1 Cowp. 63, 98 E.R. 969, at p. 970, is that evidence should be “weighed according to the proof which it was in the power of one side to have produced, and in the power of the other to have contradicted” (quoted in Snell , at p. 328).

[ 49 ] An inference of causation is available to trial judges by virtue of the ordinary operation of these principles in the medical malpractice context:

In many malpractice cases, the facts lie particularly within the knowledge of the defendant. In these circumstances, very little affirmative evidence on the part of the plaintiff will justify the drawing of an inference of causation in the absence of evidence to the contrary.

( Snell , at pp. 328-29)

[ 50 ] The majority of the Court of Appeal relied in large part on this passage to justify its conclusion that Snell created a rule of law that requires an adverse inference of causation in certain circumstances. The inference of causation Sopinka J. described in Snell is one that trial judges are permitted to draw even in the absence of positive or scientific proof. It is not one that they are required to draw once certain criteria are established. The decision on whether to draw such an inference is left to the discretion of the trial judge. Despite using permissive language to describe the adverse inference in Snell , the decision of the majority of the Court of Appeal failed to give effect to the permissive, discretionary nature of that inference.

[ 51 ] By overturning the trial judge’s decision on the basis of an error of law, the majority of the Court of Appeal implicitly transformed the permissive inference described by this Court in Snell into one that is compulsory once certain facts are established. The majority’s decision would have the effect of creating a novel legal rule governing presumptions. And yet it is apparent that Sopinka J. was not purporting to create such a rule in Snell . Rather, he was simply describing how the usual fact-finding process works in the medical malpractice context:

It is not strictly accurate to speak of the burden shifting to the defendant when what is meant is that evidence adduced by the plaintiff may result in an inference being drawn adverse to the defendant. Whether an inference is or is not drawn is a matter of weighing evidence. The defendant runs the risk of an adverse inference in the absence of evidence to the contrary. This is sometimes referred to as imposing on the defendant a provisional or tactical burden. . . . In my opinion, this is not a true burden of proof, and use of an additional label to describe what is an ordinary step in the fact-finding process is unwarranted. [Emphasis added; pp. 329-30.]

[ 52 ] The adverse inference of causation described in Snell is permissive precisely because it is a component of the fact-finding process. For the same reason, the question of whether an inference is warranted in a particular case falls within the discretion of the trier of fact, to be determined with reference to all of the evidence. This principle was recently reaffirmed by this Court in Ediger , at para. 36 . It was therefore not open to the majority to substitute its own decision to draw an unfavourable inference. Rather, the majority of the Court of Appeal would have been bound to find a palpable and overriding error in the trial judge’s decision not to draw an adverse inference. I return to this point below.

[ 53 ] Snell itself provides an example of the circumstances in which an inference of causation may be drawn. The plaintiff sued the defendant ophthalmologist after she became blind in her right eye following cataract surgery. Before the surgery, the defendant noticed some bleeding when he injected anaesthetic, yet he negligently continued the surgery. The blindness could have been caused by the bleeding or by natural causes. The expert witnesses could not definitively determine the cause. In concluding that the plaintiff had established that the defendant’s negligence caused her blindness, Sopinka J. stated:

The [defendant] was present during the operation and was in a better position to observe what occurred. Furthermore, he was able to interpret from a medical standpoint what he saw. In addition, by continuing the operation which has been found to constitute negligence, he made it impossible for the [plaintiff] or anyone else to detect the bleeding which is alleged to have caused the injury. In these circumstances, it was open to the trial judge to draw the inference that the injury was caused by the retrobulbar bleeding. There was no evidence to rebut this inference. . . .

. . . it is not essential to have a positive medical opinion to support a finding of causation. Furthermore, it is not speculation but the application of common sense to draw such an inference where, as here, the circumstances, other than a positive medical opinion, permit. [pp. 335-36]

[ 54 ] In sum, the Court held in Snell that “the plaintiff in medical malpractice cases — as in any other case — assumes the burden of proving causation on a balance of the probabilities”: Ediger , at para. 36 . Causation need not be proven with scientific or medical certainty, however. Instead, courts should take a “robust and pragmatic” approach to the facts, and may draw inferences of causation on the basis of “common sense”: Snell , at pp. 330-31; Clements , at paras. 10 and 38 . The trier of fact may draw an inference of causation even without “positive or scientific proof”, if the defendant does not lead sufficient evidence to the contrary. If the defendant does adduce evidence to the contrary, then, in weighing that evidence, the trier of fact may take into account the relative ability of each party to produce evidence: Ediger , at para. 36 .

[ 55 ] I conclude that Snell did not create the rule of law attributed to it by the majority of the Court of Appeal. This reading is consistent with the position taken by authors Allen M. Linden and Bruce Feldthusen in Canadian Tort Law (10th ed. 2015), at p. 129: “There is nothing unusual about [the holding in Snell ]; an inference can be drawn in causation fact-finding just the same as in other fact-finding situations. This is not, therefore, a retreat from ‘but for’ analysis, but an example of it, using ordinary logic and reasoning as in other contexts.”

[ 56 ] In any event, I am of the view that any doubt which might have existed over the appropriate interpretation of Snell in the Quebec civil law context was assuaged by this Court’s decision in St-Jean . As the Court explained at para. 116 of its judgment:

To the extent that such a notion is a separate means of proof with a less stringent standard to satisfy, Snell , supra , and definitely Laferrière , supra , should have put an end to such attempts at circumventing the traditional rules of proof on the balance of probabilities. There may be a misapprehension of what I said in Laferrière , supra , at p. 609: “In some cases, where a fault presents a clear danger and where such a danger materializes, it may be reasonable to presume a causal link, unless there is a demonstration or indication to the contrary” . . . . This is merely a reiteration of the traditional approach on presumptions, and does not create another means of proof in Quebec civil law in the establishment of the causal link. The Court of Appeal correctly interpreted this passage as pertaining to presumptions within the traditional rules of causation. [Emphasis added; emphasis in original deleted.]

[ 57 ] The teachings of St-Jean are clear. Properly applied in civil law, Snell results simply in the application of the rules of evidence set out in the Civil Code and the Code of Civil Procedure , more specifically those relating to presumptions ( Civil Code , arts. 2846 to 2849). This is consistent with the principle that, in Quebec, the Civil Code and the Code of Civil Procedure are comprehensive and contain all rules of evidence necessary to decide a case. As this Court explained in Prud’homme v. Prud’homme , 2002 SCC 85 , [2002] 4 S.C.R. 663, “[i]t is important to recall . . . that the new Code does not simply lay down a body of private law rules, or ‘a law of exception’. As stated in its preliminary provision, it is the jus commune of Quebec” (para. 28 (emphasis in original)).

[ 58 ] While courts must at times interpret or develop the provisions of the Civil Code to adapt its principles to novel situations, it is not necessary to do so in this appeal. The discretion afforded to trial judges in drawing presumptions of fact, exercised in accordance with the principles set out in Snell , provides us with all the tools necessary to resolve this case. When a Quebec court must decide whether to draw a presumption or an inference in a given case, requiring it to stay within the boundaries of arts. 2846 to 2849 of the Civil Code , as those boundaries have been explained in the jurisprudence, ensures the continued stability and predictability of the law.

[ 59 ] In Quebec civil law, the adverse inference discussed in Snell constitutes nothing more than a presumption of fact, as defined in art. 2849 of the Civil Code . Under that provision, “[p]resumptions which are not established by law are left to the discretion of the court which shall take only serious, precise and concordant presumptions into consideration”, and “[t]he refusal to make presumptions [of fact] is as much an evidentiary decision as is any other acceptance or non-acceptance of methods of proof”: St-Jean , at para. 114 . As such, a trial judge’s decision to draw an adverse inference is a factual decision which will depend solely on the proper application of art. 2849 to the circumstances of the case.

[ 60 ] As stated in art. 2849 , only serious , precise and concordant presumptions can be taken into consideration by courts. According to Professor M. L. Larombière, Théorie et pratique des obligations (new ed. 1885), vol. 7, at p. 216, whose explanation of serious, precious and concordant presumptions was recently quoted by this Court in Hinse v. Canada (Attorney General) , 2015 SCC 35 , [2015] 2 S.C.R. 621, at para. 71 :

[ translation ] Presumptions are serious when the connection between the known fact and the unknown fact is such that the existence of one establishes the existence of the other in a clear and obvious manner. . . .

Presumptions are precise when the conclusions that flow from the known fact tend to establish the contested unknown fact in a direct and specific manner. If it were also possible to draw different and even contrary results, to infer the existence of various and contradictory facts, the presumptions would not be precise in nature and would give rise only to doubt and uncertainty.

Finally, they are concordant, whether or not they each spring from a common or different source, when they tend[, as a whole and in how they accord with one another,] to establish the fact to be proven. . . . If, on the contrary, they contradict each other . . . and cancel each other out, they are no longer concordant, and create only doubt in the magistrate’s mind.

[ 61 ] If the criteria giving rise to a presumption are not met, “the plaintiff must actually establish the unknown fact rather than the trier of fact being permitted to draw an inference from the known fact to the unknown fact”: St-Jean , at para. 113 .

[ 62 ] The reasoning in Snell complements the application of art. 2849 , particularly in the medical malpractice context. I have said that the inference described in Snell constitutes a presumption of fact under the Civil Code . The principles laid down in Snell for drawing inferences as part of the ordinary fact-finding process apply to triers of fact making determinations of causation within the framework of art. 2849. These principles must not be interpreted in a manner that alters the burden of proof or departs from the criteria for establishing presumptions of fact ( St-Jean , at para. 116 ), as the majority of the Court of Appeal did in this case.

[ 63 ] However, Snell does provide guidance on how evidence is to be weighed and evaluated in medical malpractice cases in the civil law context. First, when weighing the evidence, trial judges are entitled to consider the ability of the parties to produce evidence ( Snell , at p. 330). Second, the distinction between legal and scientific fact-finding articulated in Snell (at p. 328) and Laferrière (at p. 609) is equally important in the context of presumptions under the Civil Code . Whether a serious, precise and concordant presumption arises on the facts must be assessed according to a legal, and not a scientific, standard. There may be cases where little affirmative evidence could be sufficient to establish causation on a balance of probabilities. Moreover, medical uncertainty will not in itself defeat a presumption of fact. For example, if the trial judge in this case had rejected the testimony of the defendants’ expert, Dr. Ferraro, and accepted that of the plaintiff’s experts, Dr. Langleben and Dr. Agulnik, the principles in Snell would have made a presumption of fact available despite the speculative character of their opinions and the existence of a competing explanation.

[ 64 ] Though the majority of the Court of Appeal was attempting to redress an injustice which it perceived, its decision was inconsistent with these teachings, as it purported to lower the threshold for drawing presumptions of fact. The majority’s view leads to the conclusion that, because of the concern in medical malpractice cases regarding a plaintiff’s inability to produce scientific or direct evidence of causation by reason of a physician’s fault, a trial judge would be required to draw a presumption against a defendant physician even where the presumption could not be said to be serious, precise and concordant. Rather, greatly lowering the threshold, the majority held that a judge is required to draw a presumption where there is “little affirmative evidence” adduced by the plaintiff.

[ 65 ] Additionally, though the holding of the majority of the Court of Appeal seems, at first glance, to be applicable only to a small number of medical malpractice cases, if we were to accept its position, there is no principled reason why its reasoning could not extend to many other professional liability cases where a defendant’s negligence precludes proof of causation with scientific certainty.

[ 66 ] In cases of causal uncertainty, both parties face the difficulty of attempting to establish facts in the absence of complete information. This case raises the issue of how that difficulty ought to be distributed between plaintiffs and defendants in cases involving what Prof. Lara Khoury calls “negligently created causal uncertainty”: Uncertain Causation in Medical Liability (2006), at p. 223 (emphasis deleted). That distribution must balance two considerations: ensuring that defendants are held liable for injuries only where there is a substantial connection between the injuries and their fault, on the one hand, and preventing defendants from benefitting from the uncertainty created by their own negligence, on the other. In Snell , this Court struck a balance by clarifying that an adverse inference may be available in such circumstances, while leaving the decision on whether to draw that inference to the trial judge as part of the fact-finding process, which is governed by ordinary principles of causation.

[ 67 ] The approach taken by the majority of the Court of Appeal disrupts this balance. As Professor Khoury points out:

In cases fraught with scientific uncertainty, rebutting a factual presumption of causation becomes a difficult task for the defendant unless he benefits from some special knowledge unavailable to the plaintiff. When the defendant must cope with the same evidential difficulties as the plaintiff, calling upon the former to demonstrate alternative cause ultimately imposes the consequences of evidential uncertainty on him. The drawing of an automatic presumption from scarce evidence may lead in effect to the same consequences as pure reversal of the burden of proof. [Footnotes omitted; p. 226.]

This is precisely what results from the approach taken by the majority of the Court of Appeal. Despite its insistence that the burden of proof remained with the plaintiff (paras. 145 and 167), its analysis of the evidence indicates otherwise. Once it found that the statistic that 78 percent of fortuitously discovered cancers are at stage I gave rise to an adverse inference, the majority of the Court of Appeal reviewed the remainder of the evidence in terms of whether it was sufficiently concrete to rebut that statistic: “The real question is . . . whether the respondents brought evidence to demonstrate that he was in the 22% of persons whose cancer was discovered fortuitously at a later stage” (para. 191). The majority of the Court of Appeal found that the defendants had not done so because their expert evidence was speculative. However, as the Court of Appeal recognized, any opinion would have been hypothetical in the absence of a biopsy (para. 162). And yet the statistical evidence that gave rise to the adverse inference insulated the testimony of the plaintiff’s experts from that same criticism.

[ 68 ] As Professor Khoury points out, an approach that triggers an adverse inference on the basis of scarce evidence has the same effect as reversing the burden of proof, and should generally be avoided (p. 226). She suggests, however, that cases of negligently created causal uncertainty may be recognized as an exception for which a rebuttable legal inference — in the civil law context, a judicial presumption — is warranted (pp. 227 and 236-37). With respect for the contrary view, I do not agree. Shifting the consequences of causal uncertainty in this manner risks turning defendant professionals into insurers. This is something which the Quebec regime of civil liability does not contemplate. In any event, since it would have the same effect as reversing the burden of proof — as the majority of the Court of Appeal’s approach to the evidence demonstrates — I find that Snell and particularly St-Jean do not support such a development in the law.

[ 69 ] The trial judge did not commit an error of law in applying the rules of evidence. She applied St-Jean , pursuant to which presumptions of causation can be drawn only when they are serious, precise and concordant. She did not think that these criteria were met because she chose to believe Dr. Ferraro, the defendants’ expert, over the plaintiff’s experts. She was not required by law to draw a presumption of fact which was for the defendants to rebut simply because (i) it was impossible to prove causation as a result of the defendants’ fault; and (ii) the plaintiff adduced “some affirmative evidence” that the defendants’ fault was linked to the loss.

[ 70 ] Therefore, in my respectful view, the trial judge committed no error of law in her causation analysis, and the majority of the Court of Appeal should not have intervened on that basis. The trial judge’s finding on causation was thus reviewable solely on the stringent standard of palpable and overriding error.

C. Did the Trial Judge Commit a Palpable and Overriding Error in Her Appreciation of the Facts?

[ 71 ] Fournier J.A. found that, in evaluating the evidence adduced, the trial judge committed a palpable and overriding error, which justified the intervention of the Court of Appeal. The majority of the Court of Appeal was of the view that the trial judge improperly relied on the speculative evidence offered by the defence expert, Dr. Ferraro, at the expense of statistical evidence that, according to the majority, established that Mr. Émond’s cancer was likely at stage I in November 2005 (para. 207). My colleague Justice Côté agrees that the adverse inference described in Snell is permissive, but finds that the trial judge’s failure to draw it in this case amounted to palpable and overriding error (paras. 95 and 105-7). In my respectful opinion, none of these approaches give sufficient deference to the trial judge’s weighing of the evidence.

[ 72 ] The majority of the Court of Appeal characterized the task facing the trial judge in weighing the evidence as follows: “The choice was between the concrete statistical evidence, on the one hand, and the speculative explanation for the losses offered by the defence” (para. 212). With respect, I cannot agree with this framing of either the expert evidence or the statistical evidence. Each of the experts formed an opinion on the basis of incomplete information, and, as the trial judge recognized, each of these opinions necessarily involved some degree of speculation and estimation (paras. 120, 123, 125 and 147). As this Court recognized in Snell , trial judges are empowered to make legal determinations even where medical experts are not able to express an opinion with certainty (p. 330). The trial judge did not commit a palpable and overriding error in relying on the opinion of an expert who acknowledged the inherent uncertainty in his opinion. Indeed, it was her role to make a legal determination of causation in spite of this. To do so, she familiarized herself with the relevant literature. She was aware of the strengths and weaknesses of all of the evidence, including both the expert and the statistical evidence. Upon weighing the whole of the evidence, she was entitled to accept Dr. Ferraro’s conclusion.

[ 73 ] The reliance on statistical evidence in this case warrants further comment. Two statistics have featured prominently in appellate review of the trial judge’s decision: the average survival period of 8 to 12 months for a patient with stage III or IV lung cancer (trial court decision, at para. 143) and the finding that 78 percent of cancers discovered fortuitously are at stage I ( ibid. , at para. 119). Fournier J.A.’s finding of palpable and overriding error with respect to the progression of Mr. Émond’s cancer rested in large part on the question of life expectancy (paras. 92-93). In the view of the majority of the Court of Appeal, the 78 percent statistic provided the “little affirmative evidence” necessary for an adverse inference of causation (para. 143).

[ 74 ] In my view, statistical evidence of this sort should be approached with some caution. Statistical generalizations are not determinative in particular cases. An example from legal theory — L. Jonathan Cohen’s well-known gatecrasher paradox — illustrates the risk of reliance on pure statistical evidence. In a case where it is established that only 499 of 1,000 rodeo spectators paid for admission, and where there is no evidence available of payment or non-payment, it would be unjust to rely on the 50.1 percent probability that a randomly selected attendee is a gatecrasher in order to hold him or her liable for non-payment: The Probable and the Provable (1977), at p. 75. Even if a higher probability is available that is more closely tailored to the generic circumstances of the particular rodeo attendee — for example, by age or gender — there is still a risk of injustice where the person nevertheless falls into the minority for whom the generalization does not hold: Cohen, at p. 78. Such a statistic alone does not establish on a balance of probabilities that any specific attendee is a gatecrasher:

Regardless of what rule governs the required quantum or preponderance of proof, naked statistics, which are merely reports of accidental groupings, do not count at all as proof of what actually happened on a particular occasion. To determine what actually happened — including how it happened and who did it — we must match particularistic evidence from the particular occasion against possibly applicable causal generalizations . . . .

The problem is not, as some have supposed, that it ordinarily is improper to rely solely on naked statistics. Rather, the problem is that naked statistics are not probative at all on the issues of what actually happened, how, and by whom. [Emphasis in original; footnote omitted.]

(R. W. Wright, “Causation, Responsibility, Risk, Probability, Naked Statistics, and Proof: Pruning the Bramble Bush by Clarifying the Concepts” (1988), 73 Iowa L. Rev. 1001, at pp. 1056-57)

[ 75 ] Appellate courts should generally not interfere with a trial judge’s decision not to draw an inference from a general statistic to a particular case. Statistics themselves are silent about whether the particular parties before the court would have conformed to the trend or been an exception from it. Without an evidentiary bridge to the specific circumstances of the plaintiff, statistical evidence is of little assistance. For this reason, such general trends are not determinative in particular cases: Laferrière , at p. 609. What inferences follow from such evidence — whether the generalization that a statistic represents is instantiated in the particular case (Wright, at p. 1057) — is a matter for the trier of fact. This determination must be made with reference to the whole of the evidence.

[ 76 ] Statistics are derived from observations of a large number of cases. The relationships they reveal between conditions and outcomes are a function of repetition on a large scale: D. Jutras, “Expertise scientifique et causalité”, in Congrès annuel du Barreau du Québec (1992) , 897, at pp. 908-9. This is another instance in which we must recognize that scientific and legal fact-finding are distinct processes: Laferrière , at p. 606. Legal fact-finding must be concerned with what actually happened between the parties before the court, and not with what happens in most cases, most of the time. As Gonthier J. held in Laferrière :

As far as possible, the court must consider the question of responsibility with the particular facts of the case in mind, as they relate concretely to the fault, causation and actual damage alleged in the case. While probabilities are unquestionably a part of the assessment of these elements in the finding of responsibility, I am very reluctant to remove the analysis from the concrete to the probabilistic plane. [p. 606]

[ 77 ] Of course, courts may take statistics into account when determining causation: Laferrière , at p. 607. Indeed, causation in this case turned on the stage of Mr. Émond’s cancer in November 2005 precisely because the trial judge relied on statistical cure rates to determine the point at which his chances of survival would have dropped below 50 percent, as she was entitled to do (paras. 113-15 and 154-56). The probative value of statistics will vary according to several factors, such as their methodology, how clearly they reveal a trend, and the resemblance between their underlying conditions and the position of the plaintiff: Jutras, at p. 909.

[ 78 ] Drawing an inference from a general statistic in a particular case is an inherent, and often implicit, part of the fact-finding process. A statistic alone reveals nothing about a particular case. It must be interpreted in light of the whole of the evidence. This interpretation is the role of the trial judge, and it is entitled to considerable deference on appeal. Respectfully, the Court of Appeal in this case failed to show such deference.

[ 79 ] The trial judge in this case acknowledged the statistic that 78 percent of fortuitously discovered cancers are at stage I. She recognized that Mr. Émond’s cancer was fortuitously discovered (para. 119). In considering the evidence as a whole, she declined to infer from those facts that Mr. Émond’s cancer was probably at stage I, as she was entitled to do. By contrast, the majority of the Court of Appeal simply treated the statistic as evidence in itself of a 78 percent probability that Mr. Émond’s cancer was at stage I (paras. 204, 207-8 and 211-12) simply because it was fortuitously discovered. In doing so, the majority usurped the role of the trial judge.

[ 80 ] It was for the trial judge to decide what weight, if any, to give to that statistic in Mr. Émond’s case. She had a reasonable basis for giving it very little. As originally put to Dr. Ferraro, the defence expert, during cross-examination, the statistic related to cancers fortuitously discovered by CT screening, a more precise diagnostic tool. CT screening would be logically expected to reveal a higher proportion of early stage cancers than fortuitous discovery resulting from a mass initially identified on an X-ray, which is why it was the recommended follow-up to Mr. Émond’s November 2005 X-ray in the first place. Moreover, Mr. Émond remained asymptomatic throughout his progression to stage IV cancer: trial court decision, at para. 130. Mr. Émond could easily have had his cancer discovered fortuitously at any stage in its progression. As a result, it was open to the trial judge to find that the general correlation between fortuitous discovery and early stage cancer was a poor basis from which to draw an inference in this case. She committed no palpable and overriding error in disregarding it.

[ 81 ] The same principle of deference applies to the trial judge’s consideration of the average life expectancy of patients diagnosed with stage III or IV lung cancer. According to the plaintiff’s expert, Dr. Langleben, a patient diagnosed with stage III lung cancer generally survives 12 months without treatment, and a patient diagnosed with stage IV lung cancer generally survives 8 months without treatment: trial court decision, at para. 143. In this case, from the time he had stage III cancer, Mr. Émond survived 14 months without treatment (from November 2005 to January 2007), and an additional 17 months with treatment (from February 2007 to June 2008). In Fournier J.A.’s view, these figures ought to have precluded a finding that Mr. Émond’s cancer was already at an advanced stage at the time the defendants negligently failed to diagnose it (paras. 92-93). As with the 78 percent statistic, the experts’ views of the average survival periods for patients with advanced stage lung cancer were available to the trial judge as a basis for inferring that Mr. Émond’s cancer could not have been at an advanced stage in November 2005. But they were not determinative. It was open to her to find that Mr. Émond’s survival for longer than average between November 2005 and June 2008 was a departure from statistics that could be explained by his overall good health (para. 144).

[ 82 ] I will now turn to the facts on which the trial judge did expressly rely. She carefully considered the evidence before her. Faced with competing expert testimony, she chose to believe Dr. Ferraro, the defendants’ expert. Dr. Ferraro’s opinion, according to the trial judge, was based on three elements. One of these elements was that a great majority of patients suffering from lung cancer are diagnosed at an advanced stage: trial court decision, at para. 134. This statistic is subject to the same caution I have explained. Indeed, because it does not refer to patients whose cancer is discovered fortuitously, it is even less applicable in Mr. Émond’s case, as the trial judge recognized (para. 119). In reaching her conclusion on causation, the trial judge did not return to this figure.

[ 83 ] The trial judge relied on two additional elements of Dr. Ferraro’s opinion. She accepted his view that, when the November 2005 chest X-ray was examined in conjunction with the December 2006 CT scan and the January 2007 PET scan, there was a [ translation ] “prominence” of the hilum and the mediastinum. This supported a finding that Mr. Émond’s cancer was at stage III in November 2005, as at stage III the cancer has generally spread to the ganglions of the mediastinum: trial court decision, at para. 108. In accepting this reading, the trial judge noted that Dr. Ferraro, a thoracic surgeon who regularly relied on X-rays and scans to conduct his surgeries, could give a reliable reading of the November 2005 X-ray (para. 150).

[ 84 ] In concluding that Mr. Émond’s cancer had at least reached stage III by the time of the November 2005 X-ray, Dr. Ferraro’s expert report refers only to a hilar prominence. During most of Dr. Ferraro’s testimony, he also did not refer to the presence of a mass in the mediastinum. This may appear troublesome, as, according to the trial judge, the presence of a mass in the hilar region (with no presence in the mediastinum) is consistent only with a finding of stage II cancer, and does not indicate stage III cancer. Dr. Ferraro was aware of this difference. However, in re-examination, he stated that, on the November 2005 X-ray, there was a [ translation] “mass at the level of the hilum and the mediastinum” , which is consistent with a finding of stage III cancer. My colleague Justice Côté concludes that this was a mistake on Dr. Ferraro’s part (para. 122) and that the trial judge’s reliance on it gave rise to a palpable and overriding error. With respect for the contrary view, the trial judge’s reliance on Dr. Ferraro’s re-examination testimony on this point is not the sort of error that meets the high threshold for appellate intervention. Appellate courts must be cognizant of the risk of “tunnel vision” in reviewing medical evidence at trial for palpable and overriding error. In this case, the trial judge would have weighed Dr. Ferraro’s reference to the mediastinum against all of the evidence, including her assessment of the strengths and weaknesses of each expert’s testimony and the reliability of Dr. Ferraro’s retrospective reading of the November 2005 X-ray. Respectfully, the fact that she did not specifically address this inconsistency when she concluded that the cancer had likely spread to the mediastinum by November 2005 does not amount to a palpable and overriding error in her evaluation of the evidence.

[ 85 ] Finally, the trial judge highlighted Dr. Ferraro’s opinion that the cancer could not have progressed from stage I to stage IV between November 2005 and December 2006, as lung cancer generally progresses slowly and can take up to three years to reach stage IV (para. 134). She cited medical literature confirming that lung cancer develops particularly slowly among non-smokers like Mr. Émond. My colleague Justice Côté observes that studies revealing a slower progression among non-smokers do not necessarily indicate that the progression is slow (para. 116). However, the trial judge would not have understood these studies in isolation. She was entitled to find that being a non-smoker could place Mr. Émond at the upper end of the three-year range she identified from Dr. Ferraro’s testimony. She also observed that the opacity in Mr. Émond’s lung did not increase in size between November 2005 and January 2006 and grew only 0.5 to 1 cm by December 2006 (paras. 138-39). These observations are consistent with her conclusion that the slow progression of Mr. Émond’s lung cancer meant that it would already have been beyond a curable stage in November 2005. It was open to the trial judge to exercise her discretion to make this finding on the basis of the evidence before her. To find otherwise requires reweighing the evidence in a manner that, in my respectful opinion, goes beyond the standard of palpable and overriding error.

[ 86 ] It could be said that it would have been open to the trial judge to find in favour of the plaintiff, particularly if individual components of the evidence had been examined in isolation. However, the trial judge carefully weighed the evidence as a whole, including both the statistical evidence and the evidence specific to Mr. Émond. Against that backdrop, she considered and evaluated three expert opinions, all of which necessarily involved some speculation. Her causation analysis was based on all of this evidence. She made no palpable and overriding error in finding that the plaintiff had failed to establish causation on a balance of probabilities, and deference to her conclusion is in order.

V. Disposition

[ 87 ] I would allow the appeal with costs throughout.

The reasons of Abella, Côté and Brown JJ. were delivered by

[ 88 ] Côté J. (dissenting) — In this matter, we must decide whether the Quebec Court of Appeal erred in reversing the trial judge’s finding that the appellants’ faulty conduct had not, on a balance of probabilities, caused the death of the late Marc Émond.

I. Whether the Majority of the Quebec Court of Appeal Applied the Correct Standard of Review

[ 89 ] Kasirer and Bélanger JJ.A., writing for the majority of the Court of Appeal, held that the trial judge’s failure to draw a negative inference against the defendants in this case constituted an error of law and was reviewable as such: 2014 QCCA 2207 , 16 C.C.L.T. (4th) 190.

[ 90 ] It is trite law to say that factual findings made by a trial judge should not be interfered with by an appellate court absent a palpable and overriding error in the trial judge’s understanding of the evidence: St-Jean v. Mercier , 2002 SCC 15 , [2002] 1 S.C.R. 491, at para. 36 ; Housen v. Nikolaisen , 2002 SCC 33 , [2002] 2 S.C.R. 235, at paras. 10-18 .

[ 91 ] This is also true for factual inferences drawn by trial judges: Housen , at paras. 19-25 . Indeed, where there is no palpable and overriding error regarding the facts underlying an inference, it is open to an appellate court to interfere with factual inferences made by a trial judge only where the inference-drawing process itself constitutes a palpable and overriding error: Housen , at paras. 22-23 . It is useful to recall this Court’s discussion of the standard of review applicable to factual findings in H.L. v. Canada (Attorney General) , 2005 SCC 25 , [2005] 1 S.C.R. 401, at paras. 52-55 :

Fact finding in the litigation context involves a series of cerebral operations, some simple, others complex, some sequential, others simultaneous. The entire process is generally reserved in Canada to courts of first instance. In the absence of a clear statutory mandate to the contrary, appellate courts do not “rehear” or “retry” cases. They review for error.

The standard of review for error has been variously described. In recent years, the phrase “palpable and overriding error” resonates throughout the cases. Its application to all findings of fact — findings as to “what happened” — has been universally recognized; its applicability has not been made to depend on whether the trial judge’s disputed determination relates to credibility, to “primary” facts, to “inferred” facts or to global assessments of the evidence.

Nor has the standard been said to vary according to whether we are concerned with what Hohfeld long ago described as “evidential” or “constitutive” facts (see W. N. Hohfeld, Fundamental Legal Conceptions as Applied in Judicial Reasoning and Other Legal Essays (1923), at p. 32). Nor, put differently, has the standard been said to vary according to whether our concern is with direct proof of a fact in issue, or indirect proof of facts from which a fact in issue has been inferred.

“Palpable and overriding error” is at once an elegant and expressive description of the entrenched and generally applicable standard of appellate review of the findings of fact at trial. But it should not be thought to displace alternative formulations of the governing standard. In Housen , for example, the majority (at para. 22) and the minority (at para. 103) agreed that inferences of fact at trial may be set aside on appeal if they are “clearly wrong”. Both expressions encapsulate the same principle: an appellate court will not interfere with the trial judge’s findings of fact unless it can plainly identify the imputed error, and that error is shown to have affected the result. [Emphasis added.]

[ 92 ] Causation, without a doubt, is a question of fact ( St-Jean , at paras. 98 and 103‑5 ):

Causation here is a question of fact. . . .

The confusion on this issue perhaps stems from an inability to distinguish between cause in the pure physical sense and cause as it is cognizable in law. The latter is a question of law only insofar as we are looking at facts through the lens of the law. However, the inconsequentiality of this observation is made obvious by the truth that everything in judicial decision making is looked at through the lens of the law. This does not make everything a question of law. For example, even questions of pure fact like whether a person was present in a certain place, or whether a person committed a certain act are determined according to the probability of that being so (or according to certainty beyond a reasonable doubt in the criminal domain). This use of the legal rule of evidence of proof on a balance of probabilities to ascertain facts does not transform the question of fact into one of law.

In the determination of fault one applies norms of behaviour required by law to a set of facts. This obviously makes the question one of mixed law and fact. In contrast, in the determination of causation one is inquiring into whether something happened between the fault and the damage suffered so as to link the two. That link must be legally significant in an evidentiary sense, but it is rendered no less a question of fact.

The difficulty on this point may also arise from the fact that the analysis of causation sometimes depends on presumptions that can be drawn. However, as indicated by the legislator in the Civil Code of Québec in making “Presumptions” a chapter in the title on proof, and as L’Heureux-Dubé J. discussed in Quebec (Public Curator) , supra , at para. 47, these too are means of proof and so are properly within the realm of fact. [Emphasis deleted.]

[ 93 ] It should also be remembered that refusal to draw an inference or to apply a presumption under the Civil Code of Québec “is as much an evidentiary decision as is any other acceptance or non-acceptance of methods of proof”: St-Jean , at para. 114 .

[ 94 ] As a result, I agree with my colleague, Justice Wagner, that the majority of the Court of Appeal erred in finding that the trial judge’s decision not to draw a negative inference or to apply a factual presumption under the Civil Code was a question of law and therefore reviewable on a correctness standard.

[ 95 ] Nonetheless, to say that the standard of review is a stringent one does not mean that it can never be satisfied. Here, in my opinion, Justice Fournier of the Court of Appeal showed the required level of deference and rightly concluded that the court’s intervention was justified. Moreover, regardless of the fact that the majority of the Court of Appeal erred in characterizing the trial judge’s failure to draw an adverse inference as an error of law, in my opinion, that failure by the trial judge in the present case constitutes a palpable and overriding error. The majority’s mischaracterization of the type of error is therefore of no consequence to the outcome of the case at bar.

[ 96 ] But before I turn to the issue of whether the trial judge committed a palpable and overriding error in her understanding of the evidence or in the inference-drawing process itself, it is important to clarify that, in my view, and contrary to my colleague’s opinion, the majority of the Quebec Court of Appeal did not misstate the rule described by this Court in Snell v. Farrell , 1990 70 (SCC) , [1990] 2 S.C.R. 311, and in St‑Jean . Rather, its reasoning constitutes an example of the inference-drawing process described in those cases.

II. Whether the Quebec Court of Appeal Misstated the Law of Adverse Inference and Presumptions in Causation

[ 97 ] According to my colleague’s opinion, the majority of the Quebec Court of Appeal created a new rule according to which a trier of fact “must” draw an adverse inference of causation against a defendant when the defendant’s negligence has undermined the plaintiff’s ability to prove causation, and when the plaintiff has adduced “some” or “very little affirmative evidence” of causation (paras. 27 and 41). In my view, this is a mischaracterization of what the Court of Appeal really held.

[ 98 ] In fact, relying on Snell and St-Jean , the Court of Appeal held that “[w]here a physician, by reason of his or her fault, undermines the plaintiff’s ability to prove causation, courts have rightly held that the rules of proof that place the full onus on the plaintiff should be relaxed” (para. 165) and that this is not achieved “by shifting the ultimate burden of proving causation from the plaintiff but, instead, by recognizing that an unfavourable inference may be drawn against the defendant physician where some affirmative evidence of causation is adduced by the plaintiff” (para. 166 (emphasis added)). The Court of Appeal used such permissive language throughout its reasons (see paras. 145, 213 and 218):

Applying the ultimate burden of proof — which, of course, always rests with the plaintiffs — we are of the respectful view that the judge should have found the unanswered statistical proof advanced by the appellants allowed causation to be inferred, to the disadvantage of the respondents, given the absence of concrete evidence to the contrary.

This is an instance in which courts should be on guard against eviscerating the advantages of the evidentiary rules on proving causation spoken to in Snell and St-Jean . The danger lies in circumstances where, as in this case, the defence raises various hypothetical explanations, as opposed to concrete factual evidence, as to how the damage came about in order to neutralize the “unfavourable inference” spoken to by the Supreme Court. It is safe to say that the defence will often be able to find some hypothesis to advance in the guise of “evidence to the contrary”, whatever its probative value, and thereby seek to undermine the unfavourable inference that might weigh against it. . . .

In emphasizing the factors that allow the unfavourable inference of causation to be drawn against the respondents in this case, it is instructive to contrast the present circumstances with those prevailing in St-Cyr c. Fisch and Aristorenas v. Comcare Heath Services , respectively Quebec and Ontario cases in which the Snell inference was not applied. [Emphasis added; footnotes omitted.]

[ 99 ] In my view, the Court of Appeal correctly summarized Snell , in which this Court held, at pp. 329-30, that “[i]t is not strictly accurate to speak of the burden shifting to the defendant when what is meant is that evidence adduced by the plaintiff may result in an inference being drawn adverse to the defendant” (emphasis added). This statement of the law is also consistent with the discussion in Ediger v. Johnston , 2013 SCC 18 , [2013] 2 S.C.R. 98, at para. 36 , in which this Court recently confirmed that “ [t]he trier of fact may , upon weighing the evidence, draw an inference against a defendant who does not introduce sufficient evidence contrary to that which supports the plaintiff’s theory of causation” (emphasis added). In other words, the Court of Appeal correctly held that trial judges are permitted to draw an unfavourable inference in some circumstances (“may”), as opposed to being required to do so (“must”). The Court of Appeal did not hold that judges should automatically draw an unfavourable inference whenever the above conditions are met, as my colleague suggests it did.

[ 100 ] In my opinion, the Quebec Court of Appeal did not create a new legal presumption, nor did it reverse the burden of proof. Although the phrasing it used was sometimes questionable, it seems to me that the majority applied the principles regarding presumptions of fact that were put forward by this Court in Snell and recognized to be applicable under Quebec law in St-Jean . All the Court of Appeal did was apply a presumption of fact under art. 2849 of the Civil Code , which provides that such presumptions are “left to the discretion of the court which shall take only serious, precise and concordant presumptions into consideration”.

[ 101 ] In the absence of direct evidence, the Court of Appeal simply grounded its finding of causation in indirect evidence and drew a presumption under art. 2849 of the Civil Code . It is useful to recall Justice Pigeon’s discussion in Martel v. Hôtel-Dieu St-Vallier , 1969 3 (CSC) , [1969] S.C.R. 745, at p. 749:

[ translation ] The Quebec courts, like those of the other provinces and Great Britain, have for quite some time, and for good reason, rejected a certain theory to the effect that direct evidence of fault is necessary in medical malpractice cases. The provisions that allow for proof by presumption of fact (arts. 1238, 1242 C.C.) make no distinction, and there is no reason to introduce one arbitrarily.

The only point to be considered, therefore, is whether the evidence was sufficient to support the conclusion that, in all probability, what happened would not have occurred, in the absence of fault. I use the words “in all probability”, because it is clear that when Taschereau J., in the above quoted passage, said “it is evident”, he was not intending to require a degree of certainty beyond the standard used in civil cases, i.e., reasonable probability. We are not dealing here with the certainty beyond reasonable doubt which is required only in criminal matters. Much less, may we require mathematical certainty, a demonstration excluding all other probabilities. In Montreal Tramways v. Léveillé , this Court accepted a presumption of fact as sufficient evidence of a cause‑and‑effect relationship between a pregnant woman’s fall and her child’s deformity.

[ 102 ] Drawing an unfavourable inference or applying a presumption is a discretionary decision and is but one of many methods of proof available to courts in making factual findings: St-Jean , at para. 114 . Ultimately, whether or not an inference of causation should be drawn is “a matter of weighing evidence” ( Snell , at p. 330). Trial judges may draw adverse inferences of causation against a defendant by applying a presumption under the Civil Code based on their assessment and weighing of the evidence as a whole. In my opinion, that is what the Court of Appeal did in this case.

III. Whether the Trial Judge Committed a Palpable and Overriding Error in Her Understanding of the Evidence

[ 103 ] I also part ways with my colleague on the issue of whether the trial judge’s understanding of the evidence before her involved palpable and overriding errors. In my opinion, the judge committed three such errors: (1) she misapprehended the evidence before her by misconstruing Dr. Ferraro’s testimony; (2) she omitted key objective evidence; and (3) she erred in the resulting inference-drawing process itself.

[ 104 ] A Court of Appeal will be justified in intervening where the trial judge’s factual findings are “based on a failure to consider relevant evidence or on a misapprehension of the evidence”: Schreiber Brothers Ltd. v. Currie Products Ltd. , 1980 11 (SCC) , [1980] 2 S.C.R. 78, at p. 84; see also Laurentide Motels Ltd. v. Beauport (City) , 1989 81 (SCC) , [1989] 1 S.C.R. 705, at p. 794; Housen , at para. 72 .

[ 105 ] First, in the case at bar, the trial judge relied on the defence expert’s testimony: 2011 QCCS 4755 . However, in my opinion, she misapprehended Dr. Ferraro’s evidence in several respects. She misconstrued his expert opinion about the rate of progression of lung cancer and his alleged detection of shadows in the hilar and mediastinal regions on the November 2005 X-ray. This misapprehension of the evidence constitutes a palpable and overriding error.

[ 106 ] Second, in her reasons, the trial judge ignored essential evidence relating to the condition of the deceased in November 2005: the fact that Mr. Émond survived more than 31 months, i.e. 14 months without treatment and an additional 17 months with treatment, even though the life expectancy of patients diagnosed with stage III to stage IV lung cancer is 8 to 12 months. The failure to consider this highly important key evidence, which was in line with undisputed statistical data showing that 78 percent of patients fortuitously diagnosed with lung cancer — which was undeniably the case of Mr. Émond — are at stage I, also constitutes a palpable and overriding error.

[ 107 ] Third, as I said above, I agree with my colleague that no trial judge is automatically compelled to draw an inference as soon as certain criteria are met and that this process, instead, is a matter of weighing evidence. To say this, however, does not amount to immunizing the inference-drawing process from appellate scrutiny; in some circumstances, drawing an inference of causation or the failure to do so will constitute a palpable and overriding error. And, as this Court recognized in H.L. , at para. 4 :

Like other appellate courts across the country, [an appellate court] may substitute its own view of the evidence and draw its own inferences of fact where the trial judge is shown to have committed a palpable and overriding error or made findings of fact that are clearly wrong, unreasonable or unsupported by the evidence . [Emphasis in original.]

The factual evidence before her undoubtedly led to a “serious, precise and concordant presumption” ( art. 2849 of the Civil Code ) that Mr. Émond’s cancer was not at stage III or higher in November 2005. In other words, in the case at bar, the inference-drawing process itself — the trial judge’s failure to draw a negative inference and apply the relevant presumption — also amounts to a palpable and overriding error.

[ 108 ] While the majority and the minority of the Court of Appeal took different routes to conclude that the appeal should be allowed, both sets of reasons acknowledged and relied on these errors, which in my opinion are palpable and overriding. Indeed, they are obvious from a simple reading of the trial judge’s reasons. As Morissette J.A. put it in J.G. v. Nadeau , 2016 QCCA 167 , at para. 77 (), these errors fall within the [ translation ] “beam-in-the-eye” category as opposed to the “needle in a haystack” category.

[ 109 ] For these reasons, I am of the view that the appeal should be dismissed. I will now discuss each of these errors in more depth.

A. The Trial Judge Misapprehended the Expert Evidence

[ 110 ] The task of the trial judge in relation to the causation issue in this case was a limited one. The only thing she had to decide was whether the lung cancer of the deceased was at stage II or lower in November 2005. If it was, the plaintiff would have satisfied her burden of proof, since the deceased would likely have survived had he been diagnosed in a timely manner.

[ 111 ] It is not contested that it was impossible from a medical standpoint to confirm the exact stage of Mr. Émond’s cancer in November 2005 as a result of the defendants’ fault. As a matter of fact, the only way to stage his lung cancer would have been to diagnose him in a timely manner, i.e. in November 2005, which the defendants regrettably failed to do. The evidence the trial judge was left with to make a finding in this regard was: the X-ray taken in November 2005, which admittedly could not be the basis of a diagnosis; the results of the December 2006 CT scan and the January 2007 PET scan, which were indicative of the condition of the deceased in December 2006 only; and the length of time it took for the disease to take Mr. Émond’s life.

[ 112 ] Three experts testified on the basis of this evidence. The trial judge found that causation had not been proven. She preferred Dr. Ferraro’s evidence to that of the plaintiff’s experts, Drs. Agulnik and Langleben, for three reasons:

• Since lung cancer grows slowly, it was “impossible” for Mr. Émond’s cancer to have progressed from stage I to stage IV in 12 months as the plaintiff suggested;

• Dr. Ferraro allegedly identified a shadow in the hilar and mediastinal regions on the November 2005 X-ray, which was consistent with cancer that had metastasized; and

• She considered the statistics showing that the majority of patients with lung cancer are diagnosed at an advanced stage of the disease.

[ 113 ] With regard to the first of these three elements, Dr. Ferraro never actually said that it was “impossible” for Mr. Émond’s cancer to have progressed from stage I to stage IV in 12 months. What he actually said was that such a scenario was, according to him, [ translation ] “possible but . . . unlikely” or “ very very unlikely ”. As noted by the trial judge, Dr. Ferraro also admitted that

[ translation ] there is no medical literature that confirms that it is impossible for cancer to progress from stage I to stage IV in less than 12 months if it is not treated, since it would be unethical not to treat a patient with stage I lung cancer in order to assess the cancer’s rate of progression. [para. 137 ()]

[ 114 ] Furthermore, Dr. Ferraro never said that lung cancer necessarily grows slowly. First, he acknowledged that this was not an absolute rule and that in the specific case of Mr. Émond, he did not know if his lung cancer actually grew slowly because it was diagnosed too late:

[ translation ]

A. As a general rule, lung cancer progresses slowly, as a general rule .

Q. And do you have objective proof that, that it progressed slowly in Mr. Émond’s case, or is that an assumption on your part?

A. Um, the only way to assess how quickly it progresses, but when we talked about the natural history, and unfortunately we don’t have the natural history in the, for the majority of patients, because we haven’t observed the patient for five (5) years without treatment.

So it’s an assumption, because we don’t have a formal staging at the start and a formal staging at the end that, then, allows us to make a comparison. So we — what I’m saying is — it’s an assumption. [Emphasis added; A.R., vol. III, at pp. 139-40.]

[ 115 ] With regard to the rate of progression of the disease, he also said that it often starts with a latent period followed by a period where the cancer becomes more aggressive. In fact, Dr. Ferraro’s own expert report states that lung cancer progresses quickly:

[ translation ] Generally speaking, when lung cancer is discovered in a patient, the long‑term recovery rate is 10‑15%. These poor results are caused by two main factors: (1) the cancer is already an advanced, stage III or IV, cancer at the time of the diagnosis; and (2) the cancer progresses quickly and spreads no matter what treatment is recommended (surgery vs. chemotherapy/radiation therapy vs. combined treatment) . [Emphasis added; A.R., vol. VII, at p. 117.]

[ 116 ] The trial judge therefore misconstrued Dr. Ferraro’s evidence regarding the rate of progression of lung cancer. She tried to supplement her finding that lung cancer progresses slowly, noting that [ translation ] “the medical literature supports the view that the progression of lung cancer is slow, particularly in non‑smoking patients” (para. 140; see also para. 152). However, this literature only supported the fact that lung cancer progresses more slowly in non-smoking patients than it does in smoking patients. It said nothing about the rate of progression of lung cancer per se ; relatively slower is not necessarily “slow” .

[ 117 ] As for the second of these elements, the trial judge’s statement that Dr. Ferraro identified a shadow in the hilar and mediastinal regions on the November 2005 X-ray is simply wrong. The identified shadow referred to by Dr. Ferraro throughout his expert report, evidence-in-chief, and cross-examination was a shadow in the hilar region only.

[ 118 ] For example, and contrary to what my colleague suggests, Dr. Ferraro always referred to [ translation ] “a shadow” as opposed to “shadows” when discussing the 2005 X-ray:

[ translation ]

Q. O.K. In your report, you give another reason: the review of the X‑ray films.

A. Yes. When you look at the films, and obviously I — I looked at them in retrospect, I saw all the films. I had the report in mind, and in my hands, the report from the CT scan, so we knew there was stage 4 cancer. We knew there were biopsies that had proved it was stage 4 cancer, so we had proof that it was stage 4 cancer. And when you look at the X‑ray in retrospect, well, you see that at the level of the pulmonary hilum, there’s a sort of mass. Yesterday we spoke of a prominence. It’s not obvious, I’m not a radiologist, it’s — I think it’s clear, it’s not obvious, but in retrospect, you can see it, the prominence .

From a prospective standpoint, I don’t know what a radiologist would have seen, um, I don’t know what I would have seen, and I’m saying this in all honesty. But in retrospect, knowing what was there one (1) year later, and looking at the series of X‑rays, well, in December two thousand six (2006), it was evident at the level of the hilum, and when you go back and compare, you can see that yes indeed it was there. So in retrospect, since the X‑ray, in retrospect it’s, by the way, it’s very very easy because you — you of course have the luxury of seeing the result twelve (12) months later, and when you compare, you say, oh yeah, it was there. [Emphasis added; A.R., vol. III, at pp. 74-75.]

[ 119 ] This point was reiterated during Dr. Ferraro’s cross-examination:

[ translation ] . . . when you review the X‑rays in retrospect and you see that there was already a nodal mass at the level of the hilum, near the mediastinum, well, when I put it all together, well, my hypothesis, eh! then I remain convinced that Mr. Émond unfortunately had stage 3 cancer when he came in. [Emphasis added; A.R., vol. III, at p. 139.]

[ 120 ] In his expert report, Dr. Ferraro makes no mention of the mediastinal region when discussing the ill-defined opacity on the 2005 X-ray:

[ translation ] In the present case, when the pulmonary lesion was first identified (November 2005), it was most likely already at stage III or IV. There are two main arguments that support this hypothesis: First, a review of the X‑ray films confirms the prominence of the right pulmonary hilum in November 2005 , which would be consistent with lymph node metastases. [Emphasis added; A.R., vol. VII, at p. 117.]

[ 121 ] Only once, and only during his re-examination, did Dr. Ferraro refer to [ translation ] “a mass at the level of the hilum and the mediastinum”.

[ 122 ] My reading of the record leads me to conclude that this last comment, made for the first time during his re-examination, was a mistake that should not be used to replace or supplant the whole of his report and his evidence. Dr. Ferraro was referring to a shadow only in the hilar region all along. Indeed, in light of the uniformly singular reference to the hilar region throughout Dr. Ferraro’s written report, examination-in-chief, and cross-examination, the only plausible reading of the record is that the reference to the mediastinal region was a mistake and that Dr. Ferraro was neither resiling from nor supplementing his consistent evidence of a shadow in the hilar region. The trial judge therefore also misconstrued the expert evidence by finding that Dr. Ferraro had identified a shadow in the hilar and mediastinal regions on the November 2005 X-ray.

[ 123 ] This is a troubling error. Indeed, experts from both sides explained in their respective testimonies that a shadow in the mediastinal region was consistent with cancer that had metastasized (stage III or more) while a shadow in the hilar region alone was consistent with stage I or II lung cancer. The trial judge accepted this evidence as a fact:

[ translation ] All the medical experts who testified agreed on the definitions of the various stages of lung cancer. They described stage I lung cancer as being limited to a lesion on the lung. They recognized that, at stage II, the cancer has spread to the lymph nodes both in and outside the lung, including the hilar region. At stage III, the nodes have reached the level of the mediastinum. At stage IV, the cancer is generalized and has metastasized.

They agreed on the pattern of progression of a cancer, namely that the cancer cells start in the lung and then attack the lymph nodes before progressing to the hilum and then to the mediastinum. [Footnotes omitted; paras. 108-9.]

Accordingly, the trial judge’s reliance on Dr. Ferarro’s error in re-examination that the mediastinal region was also affected, led her to the conclusion that the cancer was at a later stage (Stage III or IV) in November 2005 (paras. 132-34). This finding amounts to a palpable and overriding error because it was not and could not reasonably be supported by the evidence .

[ 124 ] As for the third of the above elements, the fact that the vast majority of patients are diagnosed at an advanced stage of lung cancer does not take into account the context in which patients are diagnosed. The majority of the patients to whom Dr. Ferraro was referring in his testimony were consulting him for diagnosis and were not patients for whom cancer is diagnosed fortuitously at a routine checkup . Mr. Émond fell into the latter category, not the former. No one disputes that factual finding by the trial judge: paras. 59-60. The evidence that the majority of lung cancer patients are diagnosed at an advanced stage is of limited value because it says nothing about people who, like Mr. Émond, are diagnosed fortuitously. Instead, t he statistical data that 78 percent of fortuitously discovered cancers are at stage I is of greater probative value, as it is more specific to Mr. Émond’s circumstances.

[ 125 ] In sum, the fact is that — as pointed out by the majority of the Court of Appeal — the whole of Dr. Ferraro’s testimony was speculative: paras. 201-12. To treat speculations and hypotheticals as evidence, let alone as proven facts, the way the trial judge did, constitutes a further misapprehension of the evidence: Martel , at p. 750.

[ 126 ] In my respectful view, these misapprehensions of the evidence by the trial judge amount to a palpable and overriding error.

B. The Trial Judge Omitted Key Objective Evidence

[ 127 ] Additionally, the trial judge disregarded key objective evidence, namely the fact that the deceased survived more than 31 months — i.e. 14 months without treatment and an additional 17 months with treatment — even though the life expectancy of patients diagnosed with stage III to stage IV lung cancer is 8 to 12 months. This essential evidence was in line with the statistical data showing that 78 percent of patients fortuitously diagnosed with lung cancer — which was the case of Mr. Émond — have a stage I cancer. While statistical data should not be the sole basis for drawing (or not drawing) an inference or applying (or not applying) a presumption of fact, such statistical data may help to confirm the trial judge’s factual determinations.

[ 128 ] Justice Fournier of the Court of Appeal properly dealt with the fact that Mr. Émond survived more than 14 months without treatment and an additional 17 months with treatment after the November 2005 X-ray:

[ translation ] According to the plaintiffs’ experts, it is possible for a person with stage IV cancer to survive 17 months while being treated, as the treatment has the effect of slowing the cancer. As Dr. Langleben clearly explained, if Marc Émond’s cancer was at stage IV in January 2007, the length of time he survived was average: “I’m sorry to say this, he died exactly on schedule”.

Dr. Langleben also expressed the opinion that, if Marc Émond’s cancer had been at an advanced stage in November 2005, he would have died within the year or, at least, he would not have been able to bike thousands of kilometres in the summer of 2007. Dr. Agulnik, an expert in oncology, expressed the same opinion: patients with stage IV cancer who receive no treatment live for less than 12 months.

On the second point, I do not think it can be said that the fact that Marc Émond was in good health accounts for a slow progression of the disease, especially in the period before he began chemotherapy. In the instant case, Dr. Langleben explained, rather, that a stage IV cancer patient who is in excellent health will survive about 8 months. He added that chemotherapy can extend the patient’s life by 6 months or so and that some people live for an additional period of up to 4 months. In Marc Émond’s case, he should have survived 14 to 18 months from the start of stage IV of the disease.

But if we accept the defence’s theory, Marc Émond survived nearly 30 months although he had advanced cancer and received no treatment for the first 13 months. There is nothing in the evidence to suggest that living for such a long time with advanced cancer is likely, especially with such a delay in treatment.

The three physicians agreed that cancer that has progressed to stage III or IV is fatal within twelve months if untreated and that palliative chemotherapy may extend the patient’s life by a few months. I mentioned this above.

The evidence also shows that Marc Émond was not treated between November 2005 and January 2007.

Therefore, in accordance with the unanimous opinion of the physicians and in all probability, Marc Émond would have been dead by the time the biopsies confirming that his cancer was at stage IV were performed.

Marc Émond died 31 months later, on June 6, 2008. It is unlikely that he would have survived for 31 months in the circumstances described in the evidence and the expert assessments. I find on a balance of probabilities that the possibility that Marc Émond had stage III or IV cancer must be ruled out, which shows, on the same balance of probabilities, that he had stage I or II cancer in November 2005. [paras. 73-93]

[ 129 ] I agree with this reasoning.

[ 130 ] As for the supporting statistical data disregarded by the trial judge, again, experts from both sides testified that 78 percent of patients fortuitously diagnosed with lung cancer have a stage I cancer. According to Dr. Langleben, a solitary pulmonary nodule discovered fortuitously will be a stage I cancer 80 percent of the time. Dr. Ferraro’s own evidence was that [ translation ] “as a general rule, if it’s discovered fortuitously, it’s at an earlier stage”. We know from his testimony that “stage I cancers are early cancers”. The trial judge consequently accepted this statistic as a finding of fact:

[ translation ] All the experts agreed that, as a general rule, only 10 to 15% of lung cancer patients manage to survive beyond the pivotal five‑year period, after which they are considered cured.

They stated that this low percentage can be explained by the fact that more than 75% of all cancers are diagnosed at an advanced stage and are incurable by the time they are discovered, although they acknowledged that 78% of cancers that are detected fortuitously are stage I cancers. . . . [paras. 118-19]

[ 131 ] In my opinion, the trial judge’s disregard of the key objective evidence, which was supported by uncontested statistical data, also amounts to a palpable and overriding error.

C. The Trial Judge Should Have Drawn a Negative Inference by Applying a Presumption Under the Civil Code

[ 132 ] In light of these errors, it was the role of the Court of Appeal to intervene and reweigh the evidence. After doing so, the three judges of the Court of Appeal rightly concluded that, had the trial judge disregarded the speculations on which Dr. Ferraro’s testimony was based, and had she taken into account Mr. Émond’s survival period, an indisputable fact supported by the undisputed statistical data indicating that 78 percent of the patients diagnosed fortuitously are at an early stage, she would have drawn an inference of causation.

[ 133 ] The plaintiff did not have to prove that it was impossible for the cancer of the deceased to be at stage III or IV. She had only to prove that Mr. Émond’s lung cancer was likely at stage I or II in November 2005, which she did.

[ 134 ] In my opinion, the fact that Mr. Émond survived 31 months after the November 2005 X-ray — which is entirely inconsistent with the life expectancy of stage III and IV lung cancer patients — supported by the undisputed statistical data indicating that 78 percent of patients fortuitously diagnosed with lung cancer, as he was, are at stage I, should have led, in light of the highly speculative opposing evidence, to a “serious, precise and concordant” presumption that he was not at stage III or higher in November 2005: art. 2849 of the Civil Code . It was a palpable and overriding error not to draw such a presumption.

IV. Conclusion

[ 135 ] For all these reasons, I would dismiss the appeal with costs.

Appeal allowed with costs, Abella , Côté and Brown JJ. dissenting.

Solicitors for the appellants: McCarthy Tétrault, Montréal.

Solicitors for the respondent: Kugler Kandestin, Montréal.

[1] Elsewhere in its reasons, the majority described this criterion as requiring that the defendant’s negligence make it impossible for the plaintiff to prove causation, or impossible for the plaintiff to prove causation scientifically (paras. 142, 144 and 162).

[2] McGhee v. National Coal Board , [1973] 1 W.L.R. 1 (H.L.) ; Wilsher v. Essex Area Health Authority , [1988] 2 W.L.R. 557 (H.L.) .