health warning: this talk was published in 2005 and delivered in 2002 so things may have moved on since then
1: “The more we understand about many diseases, the simpler the models get”
I don’t think I’ve ever come across such an impressive critique of the biopsychosocial model before, and it’s astonishing that this talk from epidemiologist George Davey Smith was delivered way back in 2002.
The talk was part of a conference titled “Biopsychosocial Medicine: An integrated approach to understanding illness”. And Davey Smith thinks this is the wrong approach – at least when it comes to understanding what causes illness. He argues that much of the evidence is based on spurious associations between illness and psychosocial factors. And that psychosocial treatments for non-psychological illnesses have been disappointing, largely because they are based on false premises. That might ring a bell for mecfs watchers.
Davey Smith agrees that psychosocial factors have important causal roles in mental illness. And they also affect mortality and disease outcomes for other illnesses, he says, through their effects on health-related behaviours such as smoking, drinking and adherence to medical care. But he doubts that psychosocial factors play a direct causal role in most illnesses.
Davey Smith quotes Engel’s seminal 1977 Science paper (pdf) that launched the formal biopsychosocial movement:
Davey Smith isn’t impressed, arguing that instead, “the more we understand about many diseases, the simpler the models get”. He cites the way cholera was viewed in the first half of the 19th century, contrasting “the complex, highly theoretical language of those who opposed the view that diseases like cholera were contagious was contrasted with the crude, reductionist tone of the contagionists”.
He prefers Susan Sontag’s view:
Returning to cholera, Davey Smith points out that many psychosocial factors were linked to cholera, but they didn’t cause it. [Miserable people were more at risk, because people who lived in insanitary conditions where cholera was rife were more likely to be miserable.] In a modern-day example, he points out that many psychosocial factors are associated with lung cancer through their association with smoking. But take away cigarettes, and the same psychosocial factors will not result in lung cancer.
[Cholera may be an old example for Davey Smith to choose, but it will have resonated with the audience because John Snow’s work on an outbreak of cholera in London in 1854 helped found the modern science of epidemiology. Not only did Snow’s careful statistical work on cholera cases help map the outbreak to the Broad Street water pump, but he used an experiment too. Snow persuaded the local council to remove the handle of the suspect Broad Street pump, forcing residents to get water from other sources. The outbreak stopped. The importance of experimental evidence to back up theories based simply on association is one of the main themes of Davey Smith’s talk.]
The heart of Davey Smith’s case is highlighting how often in general associations between ‘risk’ factors and disease are spurious, and how the biopsychosocial movement had some major examples of this. He implies they should be much more careful in future about claiming psychosocial causes for poorly-understood illnesses.
Exhibit A is the famous case of peptic ulcers, long argued to be caused by stress until Barry Marshall showed them to be mostly caused by a bacterium, Heliobacter Pylori, work that won him a Nobel Prize. The other principal causes of peptic ulcers are now thought to be smoking and anti-inflammatory pain killers such as aspirin – but not stress.
Davey Smith points out that epidemiological work repeatedly fingered stress, including one seemingly well-designed study that gave an Odds Ratio of 20 for ‘life events’ for men as a ‘risk’ factor for peptic ulcers. He suggests the false results was because those who reported symptoms of peptic ulcers were also tended to over-report problems in other aspects of their lives; such reporting bias is a big problem in epidemiological work.
He highlighted another study that found peptic ulcers were more common among those who came from large families. Now this would be interpreted as evidence for transmission of infection (hygiene can be harder to maintain and disease can spread more easily with lots of young children living together). Instead, because of the prevailing belief in a biopsychosocial-like model, it was taken as evidence of psychosocial factors were at play and taken as “yet more ‘proof’ of the pre-eminent role of stress in peptic ulcer.” Other evidence from Birth cohorts that indicated infection in early life was instead interpreted as more support for the prevailing stress theory.
Some people at the time argued against this mind-set, says Davey Smith, including one Richard Asher: ‘It is now fashionable to put forward mental causes for those illnesses where physical causes have not yet been found, such as peptic ulcer’. Asher gave a colleague an article to read about peptic ulcer and its cause by psychological stress and asked, ‘Do you think that's a fair account of what people think?' The colleague replied, ‘Well, it's a bit stilted language, but yes, it’s a perfectly fair account’. Asher replied. ‘Actually, it was written in 1850 about general paralysis of the insane. And I merely substituted peptic ulcer for general paralysis of the insane’. Which just highlights how long this has been an issue in medicine.
Davey Smith doesn’t hold back in his criticism of the peptic ulcer debacle, pointing out how psychological treatments were claimed to be the answer but well-designed trials showed no benefit, including one that concluded that:
And Davey Smith argues that the focus on psychosocial theories actually held back progress:
[In the case of mecfs, in the UK particularly, the medical establishment’s acceptance of a biopsychosocial model made it harder to get funding for biomedical research, and perhaps more importantly discouraged researchers from pursuing the biomedical challenge.]
Davey Smith gave plenty of examples of the problems with epidemiological evidence for biomedical work too, including the striking case of β-carotene, which is converted in the body to Vitamin A and had been repeatedly linked to heart disease. People with higher levels of β-carotene consistently have lower rates of heart disease. So β-carotene prevents heart disease? It turns out not: studies found that taking β-carotene over up to 13 years led to a small but statistically significant increase in heart disease.
It may well be that people with higher levels of β-carotene have them as a result of a healthy diet, and that the diet, not the β-carotene protects against heart disease. The point is the risk of ‘confounding’ in observational studies, where you are measuring a ‘bystander’ that just happens to be associated with the real cause of an illness (the real cause ‘confounds’ the study results). The answer, says Davey Smith, is to use experimental approaches to back up – or refute - the observational findings. HRT is another case in point: again it was linked with protection against heart disease in observational studies, again randomised controlled trials showed no benefit.
[diagram showing the problem of confounding, in this case how the link between coffee and cancer was actually down to the association of coffee-drinking with smoking, where the smoking caused cancer]
Another telling example is the link between car ownership and heart disease. Originally, car-ownership status was measured as part of the huge ‘Whitehall’ epidemiological study, with researchers suspecting car ownership would be associated with less exercise and so a risk factor for heart disease. By the time the data was collected, researchers (including Davey Smith) realised that car ownership was a marker for social class and this drove health risk (through, for example, behaviours such as promptly seeking treatment).
Car ownership turned out to ‘protect’ against heart disease, but again the problem is confounding, where car ownership is simply a marker for the real factor linked to disease. Davey Smith argues that many psychosocial factors suffer from the same confounding problem as car ownership:
Such confounding explains why health advice has changed so much over the years – on salt and polyunsaturated fats amongst others – as early observational data suggesting a causal role for dietary changes wasn’t backed up by experimental evidence.
On to the final example, one of the Crown Jewels of the biopsychosocial model: the link between stress and heart disease. A series of studies had shown that higher levels of stress are linked with higher rates of heart disease – more specifically of angina (chest pains). However, when researchers instead focused on objective markers of disease (electrocardiogram abnormalities and death from heart attack), the association was no longer statistically significant. It seems stressed people are more likely to report chest pains, but not more likely to have heart disease.
And the evidence of psychosocial interventions reducing heart disease is similarly weak, according to the evidence reviewed by Davey Smith.
As you might imagine at a conference organised for biopsychosocial enthusiasts, there was a lively discussion following Davey Smith’s talk.
The Chair, Professor Sir Simon Wessely, suggested Davey Smith had presented a popular but simplistic view of medical history:
@Woolie has pointed out that list of disease that were seen as organic and now are known to be psychological is still under debate.
Davey Smith’s response was focused on the evidence he’s presented: “My main point was about disease aetiology. As a disease epidemiologist I want to get the right answers about this”.
Professor Sir Michael Marmot made a lot of points which struck me as marginal, but he also said:
Davey Smith and Marmot are two heavyweight epidemiologists, and collaborators on the Whitehall 2 study: I’ve no idea what accounts for the very different views of the literature on this.
The final comment was from Professor Glyn Elwyn, who suggested the case for causal biopsychosocial factors may be weaker:
Overall, I thought Davey Smith's critique was fascinating as well as highly readable (online), and just wish I'd come across it years ago. I've always been frustrated by the lack of robust intellectual scrutiny of the biopsychosocial model in the research literature but that's exactly what Davey Smith has delivered here.
Phoenix Rising discussion thread about Davey Smith's talk. (Thanks to @Dolphin for spotting this talk and starting the thread).
1: “The more we understand about many diseases, the simpler the models get”
I don’t think I’ve ever come across such an impressive critique of the biopsychosocial model before, and it’s astonishing that this talk from epidemiologist George Davey Smith was delivered way back in 2002.
The talk was part of a conference titled “Biopsychosocial Medicine: An integrated approach to understanding illness”. And Davey Smith thinks this is the wrong approach – at least when it comes to understanding what causes illness. He argues that much of the evidence is based on spurious associations between illness and psychosocial factors. And that psychosocial treatments for non-psychological illnesses have been disappointing, largely because they are based on false premises. That might ring a bell for mecfs watchers.
Davey Smith agrees that psychosocial factors have important causal roles in mental illness. And they also affect mortality and disease outcomes for other illnesses, he says, through their effects on health-related behaviours such as smoking, drinking and adherence to medical care. But he doubts that psychosocial factors play a direct causal role in most illnesses.
Davey Smith quotes Engel’s seminal 1977 Science paper (pdf) that launched the formal biopsychosocial movement:
Davey Smith isn’t impressed, arguing that instead, “the more we understand about many diseases, the simpler the models get”. He cites the way cholera was viewed in the first half of the 19th century, contrasting “the complex, highly theoretical language of those who opposed the view that diseases like cholera were contagious was contrasted with the crude, reductionist tone of the contagionists”.
He prefers Susan Sontag’s view:
in "Illness as a Metaphor" Sontag said:
Returning to cholera, Davey Smith points out that many psychosocial factors were linked to cholera, but they didn’t cause it. [Miserable people were more at risk, because people who lived in insanitary conditions where cholera was rife were more likely to be miserable.] In a modern-day example, he points out that many psychosocial factors are associated with lung cancer through their association with smoking. But take away cigarettes, and the same psychosocial factors will not result in lung cancer.
[Cholera may be an old example for Davey Smith to choose, but it will have resonated with the audience because John Snow’s work on an outbreak of cholera in London in 1854 helped found the modern science of epidemiology. Not only did Snow’s careful statistical work on cholera cases help map the outbreak to the Broad Street water pump, but he used an experiment too. Snow persuaded the local council to remove the handle of the suspect Broad Street pump, forcing residents to get water from other sources. The outbreak stopped. The importance of experimental evidence to back up theories based simply on association is one of the main themes of Davey Smith’s talk.]
2: Cautionary Tales – Peptic ulcers
The heart of Davey Smith’s case is highlighting how often in general associations between ‘risk’ factors and disease are spurious, and how the biopsychosocial movement had some major examples of this. He implies they should be much more careful in future about claiming psychosocial causes for poorly-understood illnesses.
Exhibit A is the famous case of peptic ulcers, long argued to be caused by stress until Barry Marshall showed them to be mostly caused by a bacterium, Heliobacter Pylori, work that won him a Nobel Prize. The other principal causes of peptic ulcers are now thought to be smoking and anti-inflammatory pain killers such as aspirin – but not stress.
Davey Smith points out that epidemiological work repeatedly fingered stress, including one seemingly well-designed study that gave an Odds Ratio of 20 for ‘life events’ for men as a ‘risk’ factor for peptic ulcers. He suggests the false results was because those who reported symptoms of peptic ulcers were also tended to over-report problems in other aspects of their lives; such reporting bias is a big problem in epidemiological work.
He highlighted another study that found peptic ulcers were more common among those who came from large families. Now this would be interpreted as evidence for transmission of infection (hygiene can be harder to maintain and disease can spread more easily with lots of young children living together). Instead, because of the prevailing belief in a biopsychosocial-like model, it was taken as evidence of psychosocial factors were at play and taken as “yet more ‘proof’ of the pre-eminent role of stress in peptic ulcer.” Other evidence from Birth cohorts that indicated infection in early life was instead interpreted as more support for the prevailing stress theory.
Some people at the time argued against this mind-set, says Davey Smith, including one Richard Asher: ‘It is now fashionable to put forward mental causes for those illnesses where physical causes have not yet been found, such as peptic ulcer’. Asher gave a colleague an article to read about peptic ulcer and its cause by psychological stress and asked, ‘Do you think that's a fair account of what people think?' The colleague replied, ‘Well, it's a bit stilted language, but yes, it’s a perfectly fair account’. Asher replied. ‘Actually, it was written in 1850 about general paralysis of the insane. And I merely substituted peptic ulcer for general paralysis of the insane’. Which just highlights how long this has been an issue in medicine.
Davey Smith doesn’t hold back in his criticism of the peptic ulcer debacle, pointing out how psychological treatments were claimed to be the answer but well-designed trials showed no benefit, including one that concluded that:
And Davey Smith argues that the focus on psychosocial theories actually held back progress:
[In the case of mecfs, in the UK particularly, the medical establishment’s acceptance of a biopsychosocial model made it harder to get funding for biomedical research, and perhaps more importantly discouraged researchers from pursuing the biomedical challenge.]
3: Cautionary tales - Heart disease and bias (and the importance of objective measures)
Davey Smith gave plenty of examples of the problems with epidemiological evidence for biomedical work too, including the striking case of β-carotene, which is converted in the body to Vitamin A and had been repeatedly linked to heart disease. People with higher levels of β-carotene consistently have lower rates of heart disease. So β-carotene prevents heart disease? It turns out not: studies found that taking β-carotene over up to 13 years led to a small but statistically significant increase in heart disease.
It may well be that people with higher levels of β-carotene have them as a result of a healthy diet, and that the diet, not the β-carotene protects against heart disease. The point is the risk of ‘confounding’ in observational studies, where you are measuring a ‘bystander’ that just happens to be associated with the real cause of an illness (the real cause ‘confounds’ the study results). The answer, says Davey Smith, is to use experimental approaches to back up – or refute - the observational findings. HRT is another case in point: again it was linked with protection against heart disease in observational studies, again randomised controlled trials showed no benefit.
[diagram showing the problem of confounding, in this case how the link between coffee and cancer was actually down to the association of coffee-drinking with smoking, where the smoking caused cancer]
Another telling example is the link between car ownership and heart disease. Originally, car-ownership status was measured as part of the huge ‘Whitehall’ epidemiological study, with researchers suspecting car ownership would be associated with less exercise and so a risk factor for heart disease. By the time the data was collected, researchers (including Davey Smith) realised that car ownership was a marker for social class and this drove health risk (through, for example, behaviours such as promptly seeking treatment).
Car ownership turned out to ‘protect’ against heart disease, but again the problem is confounding, where car ownership is simply a marker for the real factor linked to disease. Davey Smith argues that many psychosocial factors suffer from the same confounding problem as car ownership:
Such confounding explains why health advice has changed so much over the years – on salt and polyunsaturated fats amongst others – as early observational data suggesting a causal role for dietary changes wasn’t backed up by experimental evidence.
On to the final example, one of the Crown Jewels of the biopsychosocial model: the link between stress and heart disease. A series of studies had shown that higher levels of stress are linked with higher rates of heart disease – more specifically of angina (chest pains). However, when researchers instead focused on objective markers of disease (electrocardiogram abnormalities and death from heart attack), the association was no longer statistically significant. It seems stressed people are more likely to report chest pains, but not more likely to have heart disease.
And the evidence of psychosocial interventions reducing heart disease is similarly weak, according to the evidence reviewed by Davey Smith.
4. Discussion: "I want to get the right answers”
As you might imagine at a conference organised for biopsychosocial enthusiasts, there was a lively discussion following Davey Smith’s talk.
The Chair, Professor Sir Simon Wessely, suggested Davey Smith had presented a popular but simplistic view of medical history:
@Woolie has pointed out that list of disease that were seen as organic and now are known to be psychological is still under debate.
Davey Smith’s response was focused on the evidence he’s presented: “My main point was about disease aetiology. As a disease epidemiologist I want to get the right answers about this”.
Professor Sir Michael Marmot made a lot of points which struck me as marginal, but he also said:
Davey Smith and Marmot are two heavyweight epidemiologists, and collaborators on the Whitehall 2 study: I’ve no idea what accounts for the very different views of the literature on this.
The final comment was from Professor Glyn Elwyn, who suggested the case for causal biopsychosocial factors may be weaker:
Overall, I thought Davey Smith's critique was fascinating as well as highly readable (online), and just wish I'd come across it years ago. I've always been frustrated by the lack of robust intellectual scrutiny of the biopsychosocial model in the research literature but that's exactly what Davey Smith has delivered here.
Phoenix Rising discussion thread about Davey Smith's talk. (Thanks to @Dolphin for spotting this talk and starting the thread).
