From the Editor
A century after his death, is Dr. Osler still relevant?
This week, there are three selections. First, we start with a look back with an essay on Dr. William Osler. We then look forward: with papers on AI and ketamine.
In the first selection, Drs. Charles S. Bryan (the University of South Carolina) and Scott H. Podolsky (Harvard University) write in The New England Journal of Medicine about Dr. Osler on the 100th anniversary of his death. Contemplating his life and views, they note that he “gave physicians what certain national historians gave their countries: warm feelings of togetherness, pride, and purpose.”
In the second selection, we look at a CMAJ paper. Considering AI and health care, University of Strasbourg’s Dr. Thierry Pelaccia and his co-authors write about the reasoning of mind and machine. They see a bright future: “AI can assume its place as a routine tool in medical practice.”
Finally, for the third selection, we consider a new paper on ketamine and safety from the Journal of Affective Disorders. Drawing on several studies, NIMH’s Elia E. Acevedo-Diaz and her co-authors conclude: “The results indicate that a single intravenous subanesthetic-dose ketamine infusion was relatively safe for the treatment of [treatment-resistant depression].”
Selection 1: “Sir William Osler (1849-1919) – The Uses of History and the Singular Beneficence of Medicine”
Charles S. Bryan and Scott H. Podolsky
The New England Journal of Medicine, 5 December 2019
Sir William Osler died on December 29, 1919, at his home in Oxford from hemorrhage after surgery for loculated empyema. Two days later, Richard C. Cabot (1868-1939) wrote in the New York Evening Post: ‘I doubt if any single man has ever so deeply influenced any other profession.’ Cabot’s was among the first of more than 400 obituaries and posthumous tributes, including a five-page obituary in the Journal. Few, if any, physicians have been more widely loved by their contemporaries. A century later, Osler is revered for his efforts to place clinical medicine on a rational foundation (in part through the multiple editions of his Principles and Practice of Medicine, first published in 1892), his transformation of graduate medical education (with the 1910 Flexner Report later formalizing the Hopkins model of critical, hospital-based teaching), and his attempt to keep medicine informed by a sense of humanism, even as it became ever more scientific.
Osler was many things to many people, but his ultimate gift to physicians was the sense of belonging to a splendid profession committed to the public interest.
So begins an essay by Drs. Bryan and Podolsky.
Dr. Osler had a storied career in Canada, before leaving for Hopkins and then settling at Oxford. In this essay, they seem to summarize so much of his beliefs and career in just a half sentence: “Osler nurtured the notion of the medical profession as a global force for human betterment.”
Reviewing some of his essays and addresses, they note his vision of education, commenting on how the “higher” education of the physician is best achieved through “the silent influence of character on character and in no way more potently than in the contemplation of the lives of the great and good of the past.” Dr. Osler saw the development of physicians as being part of a significant task for the “great features of the guild” of medicine has “noble ancestry, remarkable solidarity, progressive character, and singular beneficence.”
Of course, the authors also note that the test of time hasn’t been fully kind to Dr. Osler. By the 1970s, they write, “the very notion of the noble physician would be critiqued amid claims of physician venality, iatrogenic illnesses, research malfeasance, and the excesses of the ‘medical-industrial complex.’”
Still they argue for the relevance of Dr. Osler, particularly in his emphasis on the history of medicine. “Properly taught, the history of medicine can foster among students and practitioners a feeling of belonging and solidarity as members of a profession, a sense of civic responsibility, and ongoing self-reflection.”
They go on to argue that Dr. Osler’s views about the importance medicine and humanity are decidedly modern. “If Osler were alive today, he would be thrilled that two physician-led groups have received the Nobel Peace Prize for outspoken opposition to nuclear proliferation. He might start a movement to canonize Carlo Urbani (1956-2003), the Italian physician who identified the severe acute respiratory distress syndrome (SARS) and sacrificed his own life in the effort to contain it.”
The essay is concise and thoughtful, and makes a strong case that many of Dr. Osler’s views are of value.
Of course, we should always hesitate a bit when trying to extrapolate the views of historical figures to today’s issues, which may reflect more our biases than their beliefs.
For those interested in more on Dr. Osler, the Postgraduate Medical Journal (a BMJ journal) has several papers in its new issue.
Selection 2: “Deconstructing the diagnostic reasoning of human versus artificial intelligence”
Thierry Pelaccia, Germain Forestier and Cédric Wemmert
CMAJ, 2 December 2019
Artificial intelligence (AI) is often presented as the future of medical practice. The concept of AI was developed in the 1950s and has been defined as ‘the use of a computer to model intelligent behaviour with minimal human intervention.’ It is an alternative to human intelligence, particularly as a replacement for the diagnostic skill of physicians. For several years, the scientific literature and lay media have commented that nonhuman intelligence could equal or even exceed human intelligence in diagnostic tasks…
We analyze the differences in the ways humans and AI approach diagnostic reasoning to argue that human reasoning will not become obsolete in medical diagnosis.
So begins a paper by Pelaccia et al.
The authors describe the way we as humans learn, think and diagnose, in contrast to machines.
Physicians mainly use a hypothetico-deductive approach to make diagnoses. After generating diagnostic hypotheses early, they spend most of their diagnostic time testing them by collecting more data. This approach is underpinned by cognitive processes that, according to the dual-process theory, can be either intuitive or analytical. Intuition – sometimes referred to as ‘pattern recognition’ – is a process that works automatically and subconsciously. It allows humans to generate diagnostic hypotheses early by taking a few pieces of information, associating them and comparing the result with patterns stored in long-term memory… Intuition allows humans to consider only a few solutions…
In AI, acquired knowledge can be stored in different ways. Deep neural networks are composed of layers of interconnected artificial neurons forming a ‘model.’ The architecture of the network and the weights associated with each connection represent a ‘decision function.’ From an input (e.g., a histopathological image), the neural network provides a prediction as an output (e.g., cancer or not cancer). To learn, the algorithm automatically optimizes its solution by calculating an evaluation metric function, which is basically the difference between the output proposed by the algorithm and the ground truth.
They then turn their attention to diagnosis: “The rate of diagnostic errors in medical practice is estimated at about 5%-15%, depending on the specialty. This translates into more than 12 million misdiagnoses annually in the United States alone. Cognitive biases are considered to be the cause of most diagnostic errors and many biases have been reported in the medical scientific literature.”
They note, for example, several possible errors, including “the tendency to stop considering other hypotheses after reaching a diagnosis…”
How does AI perform today?
- Dermatology. AI matched dermatologists after training on 130 000 images in distinguishing “between 2 common cancers and between a benign and a malignant lesion.”
- Radiology. AI outperformed radiologist after training on 100 000 images in detecting pneumonia using chest radiographs.
- Ophthalmology. AI matched eight ophthalmologists after training on 130 000 images in detecting diabetic retinopathy while “providing more consistent interpretation, high sensitivity and specificity, and an instantaneous result.”
Because human and artificial intelligences are different and complementary, it is unlikely that AI will entirely replace the physician in the resolution of clinical problems. Artificial intelligence will be among the tools available to physicians seeking to make a diagnosis, to help with reasoning, reduce diagnostic uncertainty and augment shared decision-making, which also involves other health professionals and the patient.
This is a strong and timely paper that summarizes well the AI literature. The final conclusion – that AI has a role, not in replacing clinicians, but in augmenting our work – seems fair. The authors don’t consider psychiatry, but the conclusion would be applicable.
A handful of mental health apps employ AI (e.g., conversational agents like Woebot and Wyza), though the clinical utility of such programs is questionable. These are early days, but there would appear to be a role for AI in treatment (i.e., psychological interventions), perhaps in the near future.
For a consideration of conversational agents in mental health, see this Canadian Journal of Psychiatry review:
Selection 3: “Comprehensive assessment of side effects associated with a single dose of ketamine in treatment-resistant depression”
Elia E. Acevedo-Diaz, Grace W. Cavanaugh, Dede Greenstein, Christoph Kraus, Bashkir Kadric, Carlos A. Zarate, Lawrence T. Park
Journal of Affective Disorders, 10 November 2019 Online First
On March 5, 2019, the U.S. Food and Drug Administration (FDA) approved the glutamatergic modulator esketamine for treatment-resistant depression (TRD). Esketamine, the S-enantiomer of ketamine, is delivered intranasally and represents the first rapid-acting antidepressant drug treatment based on this novel mechanism of action. In its New Drug Application (NDA) review, the FDA identified sedation, dissociation, and increased blood pressure, as well as the potential for misuse and abuse, as the primary safety concerns associated with esketamine. In order to mitigate risk, the FDA required that a Risk Evaluation and Mitigation Strategy (REMS) accompany esketamine administration. The REMS consists of prescriber training; esketamine administration only in certain health care settings; monitoring by a healthcare provider for two hours after administration; the certification of dispensing pharmacies, practitioners, or healthcare settings; and the creation of a registry for all esketamine patients to monitor risk.
So begins the Elia E. Acevedo-Diazet al. paper.
Here’s what they did:
- They used data from 188 participants drawing from four placebo-controlled, crossover ketamine trials and one open-label study.
- Side effects were actively monitored in a standardized fashion over the time-course of each study.
- Statistical analyses were done, and included a look at CADSS total scores and side effects.
Here’s what they found:
- “Forty-four of 120 side effects occurred in at least 5% of participants over all trials.”
- Thirty-three of these 44 side effects were significantly associated with active drug administration (versus placebo).
- The most common side effect: “feeling strange/weird/loopy.” See figure below.
- There are no noted adverse events.
This study was an exploratory meta-analysis of secondary measures drawn from five substudies examining the antidepressant effects of ketamine conducted at the NIH over a period of 13 years. The results indicate that a single intravenous subanesthetic-dose ketamine infusion was relatively safe for the treatment of TRD.
This is a good and needed paper. Ketamine is growing in popularity in both research and clinical circles. Studies have covered different populations, including inpatients, and have been published by weighty journals, including The American Journal of Psychiatry.
Effectiveness doesn’t necessarily mean safety. This study thus nicely complements the literature. But while the authors did look at different side effects, it should be noted that ketamine studies in mental health are still relatively few and far between. The authors speak of a “comprehensive assessment” (perhaps referring to the statistical analyses) but the overall n was 188, and only after a single dose of ketamine.
Reading of the Week. Every week I pick articles and papers from the world of Psychiatry.