11:03 PM Kee En Chong (Attorney-General’s Chambers, Singapore)

    Causation in medical negligence: Armstrong, Carol Ann v Quest Laboratories Pte Ltd [2020] 1 SLR 133; [2019] SGCA 75


    The philosopher Immanuel Kant once observed that while human minds cannot help but think in terms of cause-and-effect, the true nature of causation remains shrouded from our understanding. What is true in philosophy appears true in medical negligence, where despite causation’s seeming simplicity, the law and science interweave in tricky and thorny ways.

    In Armstrong, Carol Ann (executrix of the estate of Peter Traynor, deceased and of behalf of the dependents of Peter Traynor, deceased) v Quest Laboratories Pte Ltd and another and other appeals [2020] 1 SLR 133 (“Armstrong”), a five-judge coram of the Court of Appeal took the opportunity to throw some light on the law of causation. In particular, the Court considered aspects of causation in medical negligence, expert testimony, and statistical and probabilistic evidence. This commentary highlights and discusses these principles.


    Brief Facts

    In September 2009, the deceased Mr Peter Traynor (“Mr Traynor”) discovered a bloodstain on his shirt. He underwent a shave biopsy of a mole on his back and the specimen was sent to Quest Laboratories Pte Ltd (“Quest”) and Dr Tan Hong Wui (“Dr Tan”) (collectively “the Respondents”). After examining a glass slide of the specimen (“the original slide”), the Respondents returned a report (“the 2009 pathology report”) concluding that there was no malignancy. Misled by the Respondents, Mr Traynor did not undertake further action even though the mole was really a malignant melanoma.

    Some two years later, Mr Traynor discovered a lump under his armpit. A core biopsy of the lymph nodes under his armpits revealed metastatic melanoma. The original slide from 2009 was recalled in January 2012. Several pathologists reviewed the original slide and they differed from Dr Tan. Instead, they diagnosed the specimen as a “malignant melanoma with ulceration” or an atypical “Spitz tumour” with “high risk features for aggressive behaviour”. Significantly, Dr Tan himself reviewed a deeper portion of the specimen in 2012, this time Dr Tan concluded that it was “suggestive of a melanoma”.

    In April 2012, Mr Traynor’s lymph nodes were surgically removed. Although five of his lymph nodes were infected with melanoma, CT and PET scans did not detect metastasis of the cancer to his distant organs. However, in April and August 2013, signs of metastasis to his distal regions emerged. Despite chemotherapy, Mr Traynor passed away, leaving behind his wife, and their two children.

    Mr Traynor’s wife, Mrs Carol Ann Armstrong (“the Appellant”) sued the Respondents alleging a misdiagnosis (“the Breach Question”), and that the Respondent’s breach had caused Mr Traynor to lose his full life expectancy (“the Causation Question”). At the trial, the High Court Judge (“the Judge”) found the Respondents’ breach “straightforward and obvious”. However, the Judge’s answer to the Causation Question was nuanced and complex. The Judge rejected the Respondents’ defence that Mr Traynor would have died anyway because the cancer had already spread into his distant organs by the time of misdiagnosis in September 2009. However, the Judge also rejected the Appellant’s claim that Mr Traynor would have been cured. Instead, the Judge appears to have relied on the minority Law Lords’ decision in Gregg v Scott [2005] 2 AC 176 on the “loss of a chance” doctrine and held that the Respondents’ breach cost Mr Traynor four years of his life expectancy.


    When does Bolam-Bolitho apply to breach?

    Although the main focus of this commentary is on causation, it is worthwhile to discuss the Court of Appeal’s holdings on breach. The assessment of breach is encapsulated in the tests set out in Bolam v Friern Hospital Management Committee [1957] 1 WLR 582 (“Bolam”) and Bolitho v City and Hackney Health Authority [1998] 1 AC 232 (“Bolitho”).

    The Bolam-Bolitho test is often thought to give significant leeway to medical practitioners. Bolam only holds a doctor to the medical profession’s consensus on the reasonable standard and no higher. Bolitho goes a step further, recognising that there may be a diversity of views as to what is a reasonable standard. This commentary considers that it is not unprincipled for Bolam-Bolitho test to afford doctors latitude. As the Court explained in Hii Chii Kok v Ooi Peng Jin London Lucien and another [2017] 2 SLR 492 (“Hii Chii Kok”) a medical practitioner should not be punished if there is a genuine difference of opinion as to what she ought to have done. Hence, if a doctor adheres to a logical body of opinion, she should not be held liable for negligence. That said, the Court of Appeal’s decision in Armstrong helpfully illustrates that the Bolam-Bolitho test is not applicable in two ways.

    First, findings of fact apart from a diagnosis are not part of the Bolam test. The Respondents had argued that because the shave biopsy was superficial, the cancerous features on the original slide from 2009 could not be visualised. The Court rejected this argument. Determining “what could be observed” was a pure finding of fact, whilst determining the correct diagnosis was a finding about the reasonable interpretation of those facts. Put simply, the former is about what clinical features can be seen, whereas the latter is about what those clinical features mean. The Court’s reasoning here follows a through-line from its decision in Noor Azlin bte Abdul Rahman v Changi General Hospital Pte Ltd and others [2019] 1 SLR 834 (“Noor Azlin”). This is sensible. Determining what can be seen on a slide does not require medical expertise, whereas interpreting and diagnosing those features do.

    Second, the Bolitho addendum is engaged only when there is a genuine difference in opinion. If the result of applying Bolam is that there is only one professional consensus on the correct course of action, then the Bolitho is not engaged. In the present case, the professional consensus was that the specimen was straightforwardly a malignant melanoma, or at least one suggestive of melanoma. Although the Respondents tried to argue that there was a “reasonable variance in interpretation”, this argument did not find favour with the Court, as the variance appears to have been whether the specimen was possibly malignant, or definitively malignant. None of the pathologists and experts (not even Dr Tan himself in 2012) was prepared to diagnose the specimen as unequivocally benign, which was exactly what Dr Tan had done in issuing the 2009 pathology report.


    A standard of care for pathologists

    The Court in Armstrong pointed out that pathologists had to undertake diagnoses with diligence because lives depended upon accurate diagnose. In this regard, the Court outlined a standard of care – if a pathologist is unable to rule out the worst-case scenario, they should state so. Dr Tan not only failed to state that he could not rule out the malignancy, he also delivered a report stating the exact opposite – that there was no malignancy. Moreover, the Respondents could have obtained deeper cross-sections of the specimen for a further check, but unreasonably chose not to do so.

    This commentary takes the view that the Court is not imposing an unreasonably high bar, but a minimum floor. This standard of care is a practical one. As the Court highlighted, a pathologist is not expected to get it right all the time, and there can be cases where a slide truly shows equivocal features for cancer. But if a pathologist harbours uncertainty about whether the worst-case scenario is present, it will cost her almost no trouble to pen an additional line or two in her report stating so. The benefits on the other hand are potentially enormous for the patient concerned. A second opinion might lead to the early detection of cancer, and in the case of Mr Traynor, could have saved his life.

    This standard of care is also a principled one. In Hii Chii Kok, the Court considered that the time for the patient to assume an active role was after a working diagnosis was formed. But for the patient to assume this active role, it is absolutely necessary that the diagnosis be a sound one. After all, an informed choice requires the patient be adequately informed.


    Factual backdrop to the Causation Question

    The Respondents sought to evade liability on the Causation Question by suggesting that the cancerous melanoma had already spread through Mr Traynor’s blood in September 2009 and into his distant organs. Hence, Mr Traynor’s fate was already biologically sealed by the time of their misdiagnosis. Even if the cancer had been detected early, nothing could have been done to save him.

    However, the Respondents’ hypothesis had to account for the fact that cancer in Mr Traynor’s distant organs was not detected by the CT and PET scans even 43 months after the time of misdiagnosis. To explain this, the Respondents suggested that melanoma had already seeded in Mr Traynor’s distant organs by September 2009, but had remained “dormant” throughout, neither growing, nor being capable of elimination by the immune system. Several large scale studies were adduced by the Respondents’ expert to support this theory of “dormancy” and to argue that Mr Traynor suffered a poor prognosis and would have relapsed anyway.

    Conversely, the Appellant argued that Mr Traynor could have been cured as the cancer had not yet spread into his blood. The Appellant’s expert suggested that if the cancer had been properly detected in 2009, Mr Traynor would have surgically removed the cancerous lymph nodes via a procedure known as sentinel lymph node biopsy (“SLNB”). Because all of the cancer was confined to his sentinel lymph nodes, SLNB would have eradicated all traces of melanoma, preventing a relapse.

    In the alternative, the Appellant also relied on large scale studies which showed that 77% of patients at Mr Traynor’s stage of cancer (at the time of misdiagnosis in September 2009), would go on to survive for the next ten years. The Appellant relied on this 77% figure, suggesting this was equivalent to a “cure” entitling her to a full award of damages. In the further alternative, she suggested the 77% figure showed that Mr Traynor had lost a 77% chance of survival, entitling her a proportionate award of damages.


    Should Bolam-Bolitho apply to causation?

    The Court in Armstrong made several observations. The first concerned the Respondents’ argument that the Bolam-Bolitho test should apply in causation. The Court rejected this argument, noting that factual causation is about “what had actually happened” – a finding of fact; whereas Bolam-Bolitho is about “what ought to have been done” – a finding concerning a diversity of professional views. As the Court put it, the former is a descriptive inquiry, whereas the latter is a quasi-normative one.

    Interestingly, the Court went a step further, explaining this on the basis of the policy rationales animating Bolam-Bolitho. Although Bolam-Bolitho has been criticised as being doctor-friendly, one of the concerns justifying its retention was that the practice of “defensive medicine” might hinder the medical profession’s fulfilment of its duties to its patients. However, concerns about “defensive medicine” only apply to breach and are “fundamentally irrelevant” in the context of causation. Hence, the Court held that Bolam-Bolitho ought not apply in the context of factual causation.

    The Court’s observations are eminently logical. While the argument of defensive medicine might make some sense in the context of breach, it is nonsensical in the context of causation. A doctor might practice medicine “defensively” to avoid breaching his duties. If he is afraid the standard of care is unreasonably high, the argument on defensive medicine runs, he might prescribe unnecessary treatments just in case. Hence, when looking at breach, a court might ask whether a defendant’s actions were “reasonable”. But when looking at causation, the court is asking “what really happened” and whether despite having breached his duties, the doctor should escape liability because the harm was ultimately not caused by him, or would otherwise have occurred.

    Put another way, questions of causation ask whether the incorrect diagnosis or treatment, once administered, did or did not kill the patient. This is not an inquiry about a “reasonable decision”, but a matter of fact. It is also not an inquiry that influences a doctor to practice medicine “defensively” because by this time the doctor has already been found to have been in breach. The doctor is not justifying what he ought or ought not have done (breach), he is trying to excuse himself from liability by that claiming despite his breach, some other event was the reason for the patient’s death (causation).


    What is causation?

    The court’s second preliminary observation relates to the philosophical difficulties alluded to at the start of this commentary. In asking what happened to Mr Traynor, the Court was concerned with what was the cause(s) of Mr Traynor’s death and the Respondents’ role in it. As a preliminary inquiry, the Court was concerned with factual causation – the physical connection between the defendant’s action and the plaintiff’s damage.

    In the ordinary case, a simple application of the “but-for” causation test will yield an intuitive answer. The test asks “but for” the defendant’s actions or omission, would the plaintiff have suffered the damage alleged? In other words, the “but for” test narrows down the necessary conditions for the damage. But even after this narrowing exercise, we may still be left none the wiser as there can be many necessary conditions left in the mix. A negligent driver may have sent the plaintiff to the hospital, but the negligent doctor’s prescription of the wrong medication led to the loss of the leg. Who can be said to have “caused” the loss of the plaintiff’s leg? In some cases, the “but for” test might not even tell us whether the defendant’s action was or was not a necessary condition. Several employers might have exposed a plaintiff to asbestos over different periods of time, but on whose work-site did the single asbestos fibre that triggered the mesothelioma originate from?

    It is for this reason that aside from factual causation, the courts develop tests of legal causation, which seeks to attribute one or more necessary conditions as legally responsible for the harm that has befallen the plaintiff. This is a question of blameworthiness. Interestingly, the loss of chance doctrine as outlined in the minority’s decision in Gregg v Scott and in the High Court’s decision in Armstrong is one such theory of legal causation. The theoretical trickiness lies in situations where the odds of recovery or improvements are less than even (eg., where the patient is diagnosed according to large scale studies to have a less than 50% outcome of survival due to a pre-existing condition at the time of misdiagnosis). In causing this loss of a “chance”, could the breach be said to have caused the damage alleged under the “but-for” test of causation?

    The answer depends, in large part, on what we are prepared to consider as “damage”. For Lady Hale and Lord Hoffman in the majority in Gregg v Scott, the damage had to be “defined in terms of the outcome”. In other words, in respect of tangible matters such as avoiding death or saving a limb. But for Lord Nicholls in the minority, in situations of significant medical uncertainty, the damage should include “diminution of prospects”. As this commentary will go on to elaborate, the “damage” here might be more appropriately conceived as the intangible opportunity to try for treatment.

    However, the Court of Appeal in Armstrong was able to straightforwardly find under “but-for” causation that the Respondents’ misdiagnosis was the only necessary condition for Mr Traynor’s death. The Court of Appeal therefore did not need to discuss whether the loss of chance doctrine should be recognised. On the facts, the evidence pointed to the cancer being confined only to Mr Traynor’s lymph nodes at the time of misdiagnosis. If diagnosed correctly, SLNB would have eradicated all of the melanoma, curing him completely.


    The role of probability and statistical evidence in causation

    The Court’s third preliminary observation concerned probabilistic and statistical evidence. The Court drew distinctions between “fact probability” – the statistical evidence presented by the parties – and “belief probability”, which is the overall degree of strength and credibility to be attributed to that piece of statistical evidence. In this regard, the Court was of the view that the large-scale statistical studies cited by the Respondents’ expert were not relevant or appropriate. This was because it was possible to find that in Mr Traynor’s specific situation he would have been cured.

    Although the Court appears to have left open the question of recognising the loss of chance doctrine, it is the Court’s critique of the majority’s decision in Gregg v Scott that is worth highlighting. In Gregg v Scott, a three-two majority of the UK House of Lords dismissed the patient’s (Mr Gregg) claim in medical negligence, ruling that his cancer would have relapsed anyway. The majority’s ruling appears to be predicated on a statistic that 58% of patients in Mr Gregg’s position at the time of the misdiagnosis eventually perished from their cancers. This was taken to mean, on a balance of probabilities, that Mr Gregg would have died anyway even if he had been properly diagnosed.

    The Court in Armstrong pointed out that the majority’s decision in Gregg v Scott was problematic. The large scale studies tracked patients who were not misdiagnosed. Mr Gregg, on the other hand, was misdiagnosed. The course of causation had therefore shifted in Mr Gregg’s case, and it was not necessarily appropriate to rely on what had happened to other patients to determine whether the same would have happened to Mr Gregg.

    The majority had also conflated the statistical figure of 58% with the balance of probabilities test. While 58% of patients had perished, this did not mean it was more likely than not that Mr Gregg would also perish, especially since both cohorts of the 58% who perished and the 42% which did survive presented with exactly the same characteristics. This sort of reasoning conflated the tools of analysis (the statistics) with the object of analysis (the damage). The course of Mr Gregg’s cancer was something particular and unique to him, and an event independent of the statistical outcomes of other patients.

    The Court’s observations uncover a fundamental flaw in Gregg v Scott’s majority’s reasoning that has been overlooked by most commentators on the loss of chance doctrine. While other commentators have pointed out that majority’s reasoning in Gregg v Scott is flawed as a matter of legal principle and practical policy, the Court in Armstrong’s real contribution is that it exposes the logical leap the majority in Gregg v Scott had taken in dismissing Mr Gregg’s appeal.

    Medical negligence cases are really concerned with two inquiries. First, a court is often attempting to figure out the state of the plaintiff’s body at the time of the misdiagnosis (namely, whether he or she was suffering from a pre-existing condition that would have rendered him or her non-susceptible to treatment anyway). Second, assuming the patient was not suffering from a pre-existing condition, the court may also be concerned with what might have happened if treatment had been timely (namely, whether he or she would still have a chance of survival). Lord Nicholls who was in the minority in Gregg v Scott termed this latter inquiry “hypothetical facts” because they concerned alternative future timelines. The former inquiry is usually resolved on the balance of probabilities, whereas the latter inquiry – being a hypothetical – does not require the same “more likely than not” proof, but merely requires the plaintiff to show a real and substantial chance of survival.

    The majority’s error in Gregg v Scott was to collapse parts of the former inquiry in on itself. The inquiry of “what has actually happened” is something to be determined on a more likely than not standard. But the 58% statistic cannot be a shorthand for a finding that it was “more likely than not” that Mr Gregg had a fatal pre-existing condition. While 58% of patients did die from non-Hodgkin’s lymphoma, the evolution of their cancers was probably specific to each patient in that 58% figure. As the Court in Armstrong alluded to, 42% of the patient did not die from that same cancer. This implies that Mr Gregg’s disease did not necessarily have pre-existing features which doomed him. At most, the 58% figure might suggest that it was more likely than not that Mr Gregg had that sort of pre-determined fatal condition, but in and of itself the 58% figure was not dispositive. Of course, whether Mr Gregg truly did or did not suffer from such a pre-existing condition could never be known, precisely because the defendant’s breach had obscured the true course of events.

    This leads to the other error committed by the majority in Gregg v Scott, which was to collapse the latter inquiry of hypothetical timelines into the former inquiry of pre-determined facts. Not only did the 58% figure not show that patients with Mr Gregg’s condition died from some pre-existing condition, the 58% figure was almost certain to have contained possibilities that had nothing to do with any pre-existing condition and might have differed entirely from Mr Gregg’s specific situation. Some of these patients might have caught a flu, which weakened their immune system causing them to die from their lymphoma. Others might have genetic anomalies and would have acquired driver mutations causing their cancer to progress more rapidly. Still others might have developed immunities to chemotherapy that might have been effective for Mr Gregg if his cancer was treated when the tumour was smaller. All of this was unknown, again precisely because the defendant’s breach had prevented any of these alternative future timelines from ever materialising.

    Interestingly, the Court in Armstrong left open the question of whether the minority’s decision in Gregg v Scott on the reduction of prospects (as reasoned by Lord Nicholls and with whom Lord Hope was prepared to agreed) should be recognised. It should be noted that unlike the majority’s reasoning, the minority’s reasoning on the loss of a chance doctrine does not substitute the statistical evidence for the factual inquiry of what had happened to the patient. Instead, the doctrine focuses on the fact that the breach has caused the patient to lose an opportunity to try for treatment (regardless of whether the treatment would ultimately be successful or not), and that opportunity is valuable. The statistical evidence therefore serves, if at all, as a proxy for quantifying the remedy.

    The Court of Appeal’s decision in Armstrong brings much needed light on aspects of causation in the law. However, it has left the doctrine of loss of chance for future determination, and it appears that this area of the law of causation remains a mystery to be solved on another day.

    * This blog entry may be cited as Kee En Chong, “Causation in medical negligence: Armstrong, Carol Ann v Quest Laboratories [2020] 1 SLR 133; [2019] SGCA 75” (

    ** A PDF version of this blog entry may be downloaded here

Comment Section