Hans Eysenck’s contribution to cognitive behavioral therapy for physical health problems: fraudulent data

  • The centenary of the birth of Hans Eysenck is being marked by honoring his role in bringing clinical psychology to the UK and pioneering cognitive behavior therapy (CBT).
  • There is largely silence about his publishing fraudulent data, editorial misconduct, and substantial undeclared conflicts of interest.
  • The articles in which Eysenck used fraudulent data are no longer cited much, but the influence of his claims which depended on these data remains profound.
  • Eysenck used fraudulent data to argue that CBT could prevent cancer and cardiovascular disease and extend the lives of persons with advanced cancer.
  • He similarly used fraudulent data to advance the claim that psychoanalysis is, unlike smoking, carcinogenic and has other adverse effects on health.
  • Ironically, Eysenck incorporated into his explanations for how CBT works elements of the psychoanalytic thinking that he seemingly detested.

If there is sufficient interest, a follow-up blog post will discuss:

  • Because of Eysenck’s influence, CBT in the UK exaggerates the role of early childhood adversity and much less to functional behavioral analysis than the American behavior therapy and cognitive behavior therapy.
  • Both CBT in the UK and some quack therapy approaches make assumptions about mechanism tied to Eysenck’s use of fraudulent data.
  • Consistent with Eysenck’s influence, CBT for physical problems in the UK largely focuses on self-report questionnaire assessments of mechanism of change and of outcome, rather than functional behavioral and objective physical health outcome variables.

8th-chocolate-happy-birthday-cake-for-HansHappy Birthday, Hans Eysenck

March 12, 2016 was the centenary of the birth of psychologist Hans Eysenck. The British Psychological Society’s  The Psychologist marked the occasion with release of a free app by which BPS members can access a collection of articles about Hans Eysenck from the archives.  Nonmembers can access the articles here.

The introduction to the collection, Philip Corr’s The centenary of a maverick states

Eysenck’s contributions were many, varied and significant, including: the professional development of clinical psychology; the slaying of the psychoanalytical dragon; pioneering behaviour therapy and, thus, helping to usher in the era of cognitive behavioural therapy…

Corr also wrote in the March 30 2016 Times Higher Education:

in defence corr

hans ensenck portraitThe articles collected in The Psychologist were written over many years. Together they present an unflattering picture of a controversial man who was shunned by his colleagues, blocked from getting awards, and who would humiliate those with whom he disagreed rather than acknowledge any contradictory evidence. Particularly revealing are Roderick Buchanan’s   Looking back: The controversial Hans Eysenck and a review of Buchanan’s book by Eysenck’s son Michael, Playing with fire: The controversial career of Hans J. Eysenck.

However, the collection stops short of acknowledging what was revealed in the early 90s in The BMJ: Eysenck knowingly published fraudulent data to back outrageous claims that CBT prevented cancer and extended the lives of patients with terminal cancer, whereas psychoanalysis was carcinogenic. He published his claims in journals he had founded, liberally self-plagiarizing and duplicate publishing with undeclared conflicts of interest. Eysenck received salary supplements and cash awards from German tobacco companies and from lawyers for the American tobacco companies for these activities.

slide 2 r smith should editors slide1 R Smith EysenckThe BMJ gave psychiatrists Anthony Pelosi and Louis Appleby a forum in the early nineties for criticizing Eysenck, even though the articles they attacked had been published elsewhere. The BMJ Editor Richard Smith followed up,  citing Eysenck as an example in raising the question whether editors should publish research articles in their own journal. Pelosi filed formal charges against Eysenck with the British Psychological Society. But, according to Buchanan’s book:

The BPS investigatory committee deemed it “inappropriate” to set up an investigatory panel to look into the material Pelosi had sent them, and henceforth considered the matter closed. Pelosi disagreed, of course, but was left with little recourse.

In an editorial in The Times Simon Wessely acknowledged Pelosi and Appleby’s criticism of Eysenck, but said “It would take more than a couple of psychiatrists to ruffle Eysenck.”

Simon on EysenckWessely suggested that the matter be dropped: the controversy was distracting everyone from the real progress being made in psychological approaches to cancer, like showing a fighting spirit extends the lives of cancer patients.  There was apparently no further mention in the UK press. Read more here.

Eysenck’s articles involving fraudulent data are seldom cited in the contemporary literature, but the claims the data were used to back remain quite influential. For instance, Eysenck claimed psychological factors presented more risk for cancer than many well-established biological factors. Including Eysenck’s data probably allowed one of the most cited meta-analyses of psychological factors in cancer to pass the threshold of hazard ratios strong enough for publication in the prestigious journal, Nature Clinical Practice: Oncology. Without the inclusion of Eysenck’s data, hazard ratios from methodologically weak studies cluster slightly higher than 1.0, suggesting little association that cannot be explained by confounds. A later blog post will document the broader influence of the Eysenck fraud on psychoneuroimmunology.

Eysenck’s claims concerning effects of CBT on physical health conditions now similarly go uncited.  However, the idiosyncratic definition he gave to CBT and his claims about the presumed mechanism by which it improved physical health pervade both CBT as defined in the UK and a number of quack treatments in the UK and elsewhere.

It is important to establish the connection between fraudulent data, distinctive features of CBT in the UK, and presumed mechanisms of action in order to open for re-examination the forms that CBT for physical health problems take in the UK and the way in which claims of efficacy are evaluated.

Fraudulent Data

Eysenck repeated tables and text in a number of places, but I will mainly draw on data as he presented them in the journal he founded, Behaviour Research and Therapy [1,   2], which correspond with what he presents elsewhere.

Eysenck’s Croatian collaborator Grossarth-Maticek conducted the therapy and collected the predictor and outcome data. A personality inventory  was used to classify participants receiving therapy into four types , a cancer-prone type (Type 1), a coronary heart disease (CHD)-prone type (Type 2), and 2 healthy types (Type 3 and Type 4). The typology was derived from quadrants in a 2×2 dichotomization of high versus low and rationality versus anti-emotionality, quite different from the dimensions and item content of the Eysenck Personality Questionnaire. Indeed, Roderick Buchanan noted in his biography that “Eysenck had struggled to banish typological concepts in favour of continuous dimensions for most of his career.” Grossarth-Maticekis questionnaire and typology has been sharply criticized later by Eysenck son Michael, among many others.

Eysenck and Grossarth-Maticek reported results of individually delivered “creative novation behaviour therapy”:

… Effects of prophylactic behaviour therapy on the cancer-prone and the CHD-prone probands respectively after 13 yr. It will be clear that treatment by means of creative novation behaviour therapy has had a highly significant prophylactic effect, preventing deaths from cancer in probands of Type 1, and death from coronary heart disease in probands of Type 2.

table 3 prophylactic effectsFor creative novation behaviour therapy delivered in a group format:

It will be seen that both cancer and CHD mortality are very significantly higher in the control group, as is death from other causes. Incidence rates are also very significantly higher in the control group for cancer, but with a difference below our selected P = 0.01 level of significance for CHD. Most telling is the difference regarding those ‘still living’-79.9% in the therapy group, 23.9% in the control group. The results of the group therapy study support those of the individual therapy group in demonstrating the value of behaviour therapy in preventing death from cancer and CHD, and in lowering the incidence from cancer and possibly from CHD.

table 4 group therapyStrong effects were reported even when the treatment was delivered as a discussion of a brief pamphlet. The companion paper  described this bibliotherapy and provided the pamphlet as an appendix,  which is reproduced here.

This statement is given to the proband, who also receives an introductory 1-hr treatment in which the meaning of the statement is explained, application considered, and likely advantages discussed. After the patient has been given time to consider the statement, and apply it to his/her own problems, the therapist spends a further 3-5 hr with the patient, suggesting specific applications of the principles in the statement to the needs of the patient, and his/her particular circumstances.

Six hundred probands received the bibliotherapy and a control group of 500 matched for personality type, smoking, age and sex received no treatment. Another 100 matched patients received a placebo condition in which they met with interviewers to discuss a pamphlet with “psychoanalytic explanation and suggestions.”

I encourage readers to take a look at the pamphlet, which is less than a page long. It ends with:

The most important aims of autonomous self-activation: your aim should always be to produce conditions would make it possible for you to lead a happy and contented life.

The results were:

There are no statistically significant differences between the control group and the placebo group, which may therefore be combined and considered a single control group. Compared with this control group, the treatment group fared significantly better. In the control group, 128 died of cancer, 176 of CHD; in the treatment group only 27 died of cancer, and 47 of CHD. For ‘death from other causes’, the figures are 192 and 115. Clearly the bibliographic method had a very strong prophylactic effect.

table 5 group and biblioEysenck and Grossarth-Maticek reported numerous other studies, including one in which 24 matched pairs of patients with inoperable cancer were assigned to either creative novation behaviour therapy or a control group. The patients receiving the behaviour therapy lived five years versus the three years of those in the control group, a difference which was highly significant.

Keep in mind that in these studies that all of the creative novation behaviour therapy sessions were solely provided by Grossarth-Maticek.

But let’s jump to a final in a series of tables constructed to make the argument that psychoanalysis was harmful to physical health.

We are here dealing with three groups. Group I is constituted of patients who terminated their  psychoanalytical treatment after 2 yr or less, and were then treated with behaviour therapy.

Group 2 is a control group matched with the members of group I on age, sex, smoking and personality type. Group 3 is a control group which discontinued psychoanalysis, like Group I, but did not receive behaviour therapy. Members of Group I and 2 do not differ significantly in mortality, but Group 3 has significantly greater mortality than either. Looking again at the percentage of patients still living, we find for Group 1 92, 95 and 95%, for Group 2 96, 89 and 95%, for Group 3 the figures are: 72, 63 and 61%. Clearly behaviour therapy can reverse the negative impact psychoanalysis has on survival.

table 15 psychoanalysisIn a number of places, this is explained in identical words:

Theoretically, this conclusion is not unreasonable. We have shown that stress is a powerful factor in causing cancer and CRD, and it is widely agreed, even among psychoanalysts, that their treatment imposes a considerable strain on patients. The hope is often expressed that finally the treatment will resolve these strains, but there is no evidence to suggest that this is true (Rachman & Wilson, 1980; Eysenk & Martin, 1987). Indeed, there is good evidence that even in cases of mental disorder psychoanalysis often does considerable harm (Mays & Franks, 1985). A theoretical model to account for these negative outcomes of psychoanalysis and psychotherapy generally has been presented elsewhere (Eysenck, 1985); it would apply equally well in the psychosomatic as in the purely psychiatric field.

dog breakfastCBT for physical health problems: a dog’s breakfast approach

Grossarth-Maticek had already formulated his approach and delivered all psychotherapy before Eysenck began co-authored papers and promoting him. In a 1982 article without Eysenck as an author, Grossarth-Maticek is quite explicit about the psychoanalytic theory behind his approach:

A central proposition of our research program is that cancer patients are either preoccupied with traumatic events of early childhood or with excessive expectations of the parents during their whole life. They are characterized by intensive internal inhibitions toward expressing feelings and desires. Therefore, we speak of a chronic blockade of expression of feelings and desires. We assume that parents of cancer patients did not respond adequately to the child’s cries for help and these children were obliged very early to do non-conforming daily task. Cancer patients have never learned to express persistent cries for help…

The specific family dynamics in the special educational pattern which block hysterical reactions determine the behavior, which in turn is characterized by excessive persistence of performance of the daily task, disregard of symptoms and lack of aggressiveness in behavior. Through the currents of negative life events (i.e., death of closely connected persons) expressions of loneliness and reactive depression can appear intensively and chronically.

If this is not clear enough:

In our approach we try not to deny the psycho analytic propositions but to integrate the psychoanalytic research program with social psychological and sociological factors, hereby assuming that they have interactive effects on carcinogenesis.

Strangely, Grossarth-Maticek suggests in this article, that the psychoanalytic factors interact with “organic risk factors such as cigarette smoking in the case of lung cancer.” Grossarth-Maticek and Eysenck would soon be receiving tens of thousands of dollars in support from the German tobacco companies and lawyers from the American tobacco companies to promote the idea that personality caused smoking and lung cancer, but any connection between smoking and lung cancer was spurious. Product liability suits against tobacco companies should therefore be dismissed.

In the articles co-authored by Grossarth-Maticek and Eysenck, these roots of what Eysenck repackaged as creative novation behaviour therapy are only hinted at, but are noticeable to the observant reader in references to the role of dependency and autonomy. Fraudulent data are mustered to show the powerful positive effects of this behaviour therapy versus the toxicity of psychoanalysis.

On page 8 of this article, ten  explicitly labeled behavioural techniques are identified as occurring across individual, group, and bibliotherapy:

  • Training for reduction of the planned behaviors initiation of autonomous behavior.
  • Training for cognitive alteration under conditions of relaxation
  • Training for alternative reactions.
  • Training for the integration of cognition, emotionality and intuition.
  • Training to achieve stable expression of feelings.
  • Training for potentiating social behavioral control
  • Training to suppress stress-creating ideas
  • Training to achieve a behavior-directing hierarchic value structure
  • Training in the suppression of stress-creating thought.
  • Abolition of dependence reactions.

This approach has only superficial resemblance to American behavioral therapy and CBT. The emphasis on expression of emotional feelings and abolition of dependent reactions is incomprehensible when it is detached from its psychoanalytic roots. The paper refers to behavioral analysis, but interviews about the past, including childhood experiences are emphasized, rather than applied behavioral analysis. The hierarchies of behavior do not correspond to operant approaches, but to a value structure of autonomy versus dependence.

There is also considerable reference to the use of hypnosis to achieve these goals.

In short, neither the goals nor the methods have much relationship to learning theory at the time that Eysenck was writing nor to contemporary developments in operant conditioning. His approach is a tortured extension of classical conditioning. Outside of the fraudulent data that Grossarth-Maticek developed and that he published with Eysenck, there is little basis for assuming that psychological factors were related to physical health in the way the treatment approach postulated.

It should be kept in mind that Eysenck was not a psychotherapist. He actually detested psychotherapy and generated considerable controversy earlier by arguing that any apparent effects of psychotherapy were due to natural remission. It should also be noted that Eysenck was claiming creation novation behaviour therapy modified personality traits, even when delivered in a brief pamphlet, in ways that could not be anticipated by his other writings about personality. Finally, the particular personality characteristics that Eysenck was talking about modifying were very different than what he assessed with the Eysenck Personality Inventory.

Only “controversial” and “too good to be true” or fraud?

 Before Eysenck began collaborating with Grossarth-Maticek, there was widespread doubts about the validity of Grossarth-Maticek’s work.  In 1973, Grossarth-Maticek’s work had been submitted to the University of Heidelberg as a Habilitation, a second doctoral degree required for a full professorship. It was rejected. One member of the committee, Manfred Amelung, declared the results “too good to be true.” He retained a copy and would later put his knowledge of its details into a devastating critique. According to Buchanan’s biography, Eysenck demanded of Grossarth-Maticek “you must let me check your data, for if you deceive me I will never forgive you.”

Eysenck gained access to the data set, sometimes directing reanalyses by Grossarth-Maticek and his statistician. Other analyses were done by Eysenck’s statisticians in London. Eysenck’s biographer Buchanan noted “there were ample opportunities to select, tease out, or redirect attention – given a data set that was apparently sprawling chaotic but rich and ambitious….From the mid-1980s, Eysenck did virtually all of the writing for publication in English and presumably exerted a strong editorial control.” Buchanan also notes that tobacco companies became skeptical of the strength of findings that were reported, but also their inconsistency. They refused to continue to support Eysenck unless an independent team was set up to check analyses and the conclusions that Eysenck was drawing from them.

Eysenck single-authored a target article for Psychological Inquiry that reproduced many of the tables that we have been discussing. More than a dozen commentators included the members of the independent team, but also others who did not have access to the data, but who examined the tables with forensic attention. The commentary started off with Manfred Manfred Amelung who made use of what he had learned from Grossarth-Maticek’s doctoral work.

Many of the commentators suggested that the intervention studies presented conclusions that were “too good to be true,” not only in terms of the efficacy claim for the intervention, but for the negative outcomes claimed for the control group. But other commentators pointed to gross inconsistencies across different reports in terms of methods and results, clear evidence of manipulation of data, including some patients being counted a number of times, other patients dying twice, Eysenck and Grossarth-Maticek’s improbable ability to obtain matching of intervention patients and controls, and too perfect predictions. In the end, even Grossarth-Maticek’s Heidelberg statistician expressed concerns that there had been tampering with the data.

Both Grossarth-Maticek and Eysenck got opportunities to respond and were defensive and dismissive of the overwhelming evidence of exaggeration of the results and even fraud.

The exchanges in Psychological Inquiry occurred over two issues. Taken together, the critical commentaries are devastating, but the criticisms became diffuse because commentators focused on different problems. It took a more succinct, pithy critique by Anthony Pelosi and Louis Appleby in The BMJ to bring the crisis of credibility to a head.

Anthony Pelosi and Louis Appleby in The BMJ

 In the first round of their two-part attack, Pelosi and Appleby centered on Eysenck and Grossarth-Maticek’s  two articles in Behaviour Research and Therapy, but referenced the critiques in Psychological Inquiry. The incredible effectiveness of these two psychiatrists depended largely on their pointing  out what was hiding in plain sight in the two Behaviour Research and Therapy articles. For instance:

After 13 years, 16 of 50 untreated type 1 subjects had died of a carcinoma. Not one of the 50 cancer prone subjects receiving the psychotherapy died of cancer. The therapy was a genuine panacea, giving equivalent results for type 2 subjects and heart disease. The all cause mortality was over 60% in untreated and 15% in treated subjects. The death rate in the untreated subjects was truly alarming as they began the trial healthy and most were between 40 and 60 years of age.

I encourage readers to compare the Pelosi and Appleby paper to the tables I presented here and see what they missed.

Pelosi and Appleby calculated the effort required by Grossarth-Maticek if he had – as Eysenck insisted- single-handedly carried out all of the treatment.

It is striking that all the individual and group therapy was given by Professor Grossarth-Maticek. The trials were undertaken between 1972 and 1974 and involved 96 subjects (or perhaps 192 subjects, see below) in at least 20 hours of individual work, and at least 10 groups (245 subjects with 20-25 in each) for six to 15 sessions each. Add to this Grossarth-Maticek’s explanatory introduction to bibliotherapy for 600 people, and it can be seen that the amount of time spent by this single senior academic on his experimental psychotherapies is huge and certainly unprecedented.

They summarized inconsistencies and contradictions reported in the Psychological Inquiry, but then added their own observation that a matching of 192 pairs of intervention and control patients had only produced a sample of 192! They suggested that in the two Behaviour Research and Therapy articles there were at least  “10 elaborate misprints or misstatements in the description of the methods” that the editor or reviewers should have caught.

At no point, does the word “fraud” or “fraudulent” appear in Pelosi and Appleby’s first article. Rather, they suggest that  “Eysenck and Grossarth-Maticek… are:

making claims which, if correct, would make creative novation therapy a vital part of public health policy throughout the world.”

They conclude with

For these reasons there should be a total reexamination and proper analysis of the original data from this research in an attempt to answer the questions listed above. The authors give their address as the Institute of Psychiatry in London, which must be concerned about protecting its reputation. Therefore the institute should, in our view, assist in this clarification of the meaning of the various studies. There should also be some stern questions asked of the editors of the various journals involved, especially those concerned among the editorial staff of Behaviour Research and Therapy who, in our opinion, have done a disservice to their scientific disciplines, and indeed to Professors Eysenck and Grossarth-Maticek, in allowing this ill considered presentation of research on such a serious topic.

Eysenck’s reply and Pelosi and Appleby’s response

 Readers can consult Eysenck’s reply  for themselves, but it strikes me as evasive and dismissive. Specific criticisms are not directly answered, but Eysenck points to consistency between his results and those of David Spiegel, who had claimed to get even stronger effects in his small study of supportive expressive therapy for women with metastatic breast cancer. Rather than demolishing the credibility of his work with Grossarth-Maticek, Eysenck argues that Pelosi and Appleby only point to the need for funding of a replication. Eysenck closes with:

Their critical review, however incorrect, full of errors and misunderstandings, and lacking in objectivity, may have been useful in drawing attention to a large body of work, of both scientific and social relevance, that has been overlooked for too long.

Pelosi and Appleby took Eysenck’s reply as an opportunity to get even more specific in the criticisms:

We are accused of being vague in mentioning many errors, inappropriate analyses, and missing details in the publications on this research programme. We value this opportunity to be more specific, to clarify just a few of the questions raised by ourselves and others, which Eysenck has failed to answer, and to outline additional findings from these authors’ investigations.

After a detailed reply, they wrap up with references to the criticisms that Eysenck received in Psychological Inquiry, in an ironic note, turning Eysenck’s attacks on proponents of the link between smoking and lung cancer on to Eysenck himself:

Our concern has been to clarify the methods and analyses of a body of research which, if accurate, would profoundly influence public health policies on cancer and heart disease. Other critics have been more challenging in what they have alleged, and in our opinion the controversy which now surrounds one of academic psychology’s most influential figures constitutes a crisis for the subject itself. The seriousness of the detailed allegations by van der Ploeg, although refuted by Eysenck and Grossarth-Maticek, should in themselves prompt these authors to reexamine their own findings after appropriate further training in the methodology of medical research. Perhaps the most skilfully worded criticism on this subject was made not about Eysenck but by him in a debate on the relation between smoking and cancer. In disputing the findings of Doll and Hill’s epidemiological studies on this association he comments: “What we have found are serious methodological weaknesses in the design of the studies quoted in favour of these theories, statistical errors, and unsubstantiated extrapolations from dubious data to unconfirmed conclusions.” Eysenck owes it to himself and to his discipline to reconsider critically his own work on this subject.

In the over 20 years since this exchange, Pelosi and Appleby and their ally editor Richard Smith of The BMJ failed to get an appropriate response from the British Psychological Society, King’s College London or the Institute of Psychiatry, the journal Behaviour Research and Therapy, or the Committee on Publication Ethics (COPE). This situation demonstrates the inability of British academia to correct bad and even fraudulent science. It stands as a cautionary note to those of us now attempting to correct what we perceive as bad science. Efforts are likely to be futile. On the other hand, the editorship of Behaviour Research and Therapy has passed to an American, Michelle Craske, a professor at UCLA. Perhaps she can be persuaded to make a long overdue correction to the scientific record and remove a serious blemish on the credibility of that Journal.

If there is sufficient interest, I will survey the profound influence of the fraudulent work of Eysenck and Grossarth-Maticek in a future blog post.

  • Because of their influence, CBT in the UK gives an exaggerated emphasis to early childhood adversity and much less to functional behavioural analysis than the American behavior therapy and CBT.
  • Consistent with Eysenck’s influence, CBT for physical problems in the UK largely focuses on self-report questionnaire assessments of mechanism of change and of outcome, rather than functional behavioral and objective physical health outcome variables.

Influences can also be seen in:

Contemporary CBT for physical conditions as practiced in UK, including CBT for irritable bowel syndrome (IBS), fibromyalgia, and other “all in the head” conditions that are deemed Medically Unexplained Symptoms (MUS) in the UK, as in PRINCE trial of Trudie Chalder and Simon Wessely.

The “psychosomatic” approach as seen in neurologist Suzanne O’Sullivan’s  recent editorial in The Lancet and her “It’s All in Your Head”, which won the 2016 Wellcome Book Award and her.

Quack treatments, such as Phil Parker’s Lightning Process, which the UK’s Advertising Standards Authority (ASA) ruled against advertising its effectiveness in treatment of chronic fatigue syndrome/ myalgic Encephalopathy,  multiple sclerosis, and irritable bowel syndrome/digestive issues. The Lightning Process is nonetheless implemented in the UK NHS under the direction of University of Bristol Professor Esther Crawley 

Quack cancer treatments such as Simonton visualization method.

More mainstream, but unproven psychological treatments for cancer including David Spiegel’s supportive expressive therapy. Neither Spiegel –nor anyone else– has ever been able to replicate the finding praised by Eysenck, but repeats his claims in a recent non-peer reviewed article in the UK-based Psycho-Oncology and with a closely related article in BPS’ British Journal of Health Psychology.

More mainstream, but unproven psychological approaches to cancer that claim to improve immune functioning by reducing stress.

Some Scottish readers will understand this message concerning Eysenck’s fraud: The ice cream man cometh.

My usual disclaimer: All views that I express are my own and do not necessarily reflect those of PLOS or other institutional affiliations.



Leave a comment

Please note, comments must be approved before they are published