Critical Psychiatry Textbook, Chapter 11: Dementia


Editor’s Note: Over the next several months, Mad in America is publishing a serialized version of Peter Gøtzsche’s book, Critical Psychiatry Textbook. In this blog, he critiques the use of drugs to treat dementia. Each Monday, a new section of the book is published, and all chapters are archived here.

As noted earlier, an editor of one of the textbooks,18 Poul Videbech, wrote in 2014 that depression doubles the risk of dementia,408 but the meta-analysis he cited did not mention with one word which treatments the patients had received.407 Other studies suggest that it is depression pills and other psychiatric drugs that make people demented.557,558

Flat vector illustration of person's head turning into dotsThe information about treatment of dementia was highly misleading. In one book, the chapter about Alzheimer’s disease was written by two psychologists20:341 who went into detail about the drugs, even though this is none of their business, as psychologists are not allowed to prescribe drugs in Denmark. They claimed that acetylcholinesterase inhibitors can dampen the development of symptoms20:351 but none of their 19 references were to research documenting this.

Two other psychologists wrote that the drugs have a better effect on Lewy body dementia and dementia in Parkinson’s disease than in Alzheimer’s disease on cognitive functions, apathy, visual hallucinations, delusions, and other neuropsychiatric symptoms.20:375 These finger-tip sensations are far-fetched for drugs that don’t work (see below), and none of their 38 references were about drug effects whereas several were about psychotherapy and other therapies, e.g. a meta-analysis of the effect of dancing in patients with Parkinson’s disease. Very strange, indeed.

Another book claimed that drugs can inhibit the progression of Alzheimer ‘s disease for months to a few years, and that donepezil, galantamine, and rivastigmine have equal effect.18:48

A third book mentioned that acetylcholinesterase inhibitors may delay the decline in functional level and behaviour.17:243 This became more concrete 424 pages later: Drugs, primarily acetyl-cholinesterase inhibitors, may to some extent re-establish lost cognitive skills as well as postpone further deterioration. The progression in Alzheimer’s can be delayed for 6-12 months.17:667

A fourth book did not pull any punches either.16:127 It claimed that, in a minority, a clear improvement of cognitive functions is experienced, with resumption of earlier activities and possible disappearance of hallucinations or other neuropsychiatric symptoms. The authors also claimed that acetylcholinesterase inhibitors may have a beneficial effect on behavioural and psychological symptoms of dementia and may delay their onset.

Dementia was of course not an issue in the textbook about child and adolescent psychiatry.19

All these statements are totally wrong. There wasn’t a single reference to placebo-controlled trials or meta-analyses, which would have told a story of drugs that don’t work and are harmful.7:197

The small subjective effects registered in drug trials are likely spurious, as they can easily have been caused by unblinding bias because of the drugs’ conspicuous adverse effects.

A 2006 Cochrane review of donepezil, galantamine, and rivastigmine didn’t pay attention to this problem and concluded that, “The three cholinesterase inhibitors are efficacious for mild to moderate Alzheimer’s disease.”559 Even without considering the unblinding problem, this conclusion was unwarranted. The improvement in cognitive function was 2.7 points, in the midrange of a 70-point scale. This is less than the 4 points the FDA considers the minimally relevant clinical change.560 We may also compare with the smallest effect that can be perceived on the Hamilton scale for depression, which is 5-6, although the maximum is only 52.267

The author of the Cochrane review wrote that “donepezil appears to have no serious or common side effects.” This is so egregiously false that I don’t think Pfizer would have dared claim this in one of their advertisements for Aricept (donepezil).

The harms are both common and serious, which the author of the Cochrane review actually demonstrated herself, as 29% of the patients dropped out of the drug groups, as compared to only 18% in the placebo groups, partly because of more adverse events.559 The most common harms of donepezil are nausea, diarrhoea, insomnia, vomiting, muscle cramps, fatigue, and anorexia.561 This is not what we would want for an old person who might already have problems with bad sleep, feeling tired, and eating too little.

The list of frequent adverse effects in Pfizer’s product information for Aricept is very long.561 The drug causes syncope in 1% of the patients and when old people fall, there is a considerable risk that they break their hip and die. A large Canadian cohort study showed that if people with dementia took dementia drugs, they had almost a doubled risk of hospitalisation for syncope, and they had more pacemakers inserted and more hip fractures.562 More than half of the patients who were admitted to hospital for bradycardia were retreated with the same type of drug after discharge.562 This is yet another proof that doctors cannot handle psychotropic drugs safely.

A 2014 study, of 5,406 nursing home residents in the United States with advanced dementia, found that one-third received cholinesterase inhibitors and one-fourth memantine, another dementia drug.563 The title of the paper was appropriate: “Use of Medications of Questionable Benefit in Advanced Dementia.”

It is interesting that no benefits for society have been found,564 as we so often hear about the economic burden of dementia and how important it is to intervene with drugs.

The political sales pitches—which tend to coincide with general elections—are vacuous. A long-term trial of 565 patients with mild to moderate Alzheimer’s disease that compared donepezil with placebo found no meaningful effects, and the authors concluded that donepezil isn’t cost-effective, with benefits below minimally relevant thresholds.565

In contrast to other trials, this trial was publicly funded. It was excluded from the Cochrane review,559 and the author used 511 words on explaining why. The main reasons appeared in a table: “Results for the 5 and 10 mg/day groups were not reported separately. Complex design and high numbers of dropouts made analysis and interpretation difficult.”

It is not acceptable to exclude a study because it combines two dose groups in the results. And that the design was complex is not a valid reason either for its exclusion. Furthermore, as it was a long-term trial, where more people drop out than in short-term trials, the high drop-out rate was also an invalid reason for exclusion.

The outcome after three years was similar on drug and placebo for institutionalisation, progression of disability, and behavioural and psychological symptoms.565

Extremely few trials in psychiatry run for three years but such trials are exactly those we need instead of the thousands of short-term trials we have, which are useless for an assessment of drug effects, as very few patients are treated for only a few weeks.

Six years after the trial was published, TV commercials for Aricept implied that the patients’ cognitive and daily functioning, including attention, focus, orientation, communication, social interaction and engagement, will be restored to normal; “Don’t wait. Talk to your doctor about Aricept.”566 The FDA told the company that—with these huge lies—it had broken the law.

You should not talk to your doctor about dementia drugs because, as the textbooks so clearly showed, your doctor is highly likely to mislead and harm you. These drugs should not be used by anyone to prevent or treat dementia.

Three critical comments have been published on the 2006 Cochrane review, including mine.559 Unfortunately, contrary to good scientific practice, they are undated. The author apologised for an error, which she said would be corrected in the next version, and she replied to me that another error had “also now been corrected.” It has not been corrected. In 2015, I was told that “An update of the review … is in preparation.”559 The review has not been updated. It stands as a gravestone over a once magnificent organisation, which is currently facing big financial trouble because it has not lived up to the expectations of its major funder, the UK National Health Service.146

There are other Cochrane reviews of these drugs, e.g. one in vascular dementia, which is not encouraging either.567 The authors concluded that donepezil and galantamine have a small effect on cognition but that it is unlikely to be clinically important.

One of the textbooks noted that psychosis pills cause considerable harms, e.g. an increased risk of thrombosis in the heart and brain and an increased risk of death.16:127 It claimed that risperidone and olanzapine have a documented minor effect in dementia.16:127 This was in a chapter about dementia written by two doctors who work with these patients. In another chapter, about psychopharmacology, the author contradicted this, as she noted that psychosis pills should be avoided in elderly people with dementia and behavioural disorders due to the lack of evidence for an effect, increased sensitivity to harms, and an increased risk of stroke.16:561 She did not mention the most important reason to avoid these drugs: To avoid killing patients in large numbers (see Chapter 7).

This demonstrated a general issue. People who treat patients become carried away by their “clinical experience” and other biases and are much too positive towards the effects of psychiatric drugs. They are therefore not the most trustworthy textbook authors. They have many vested interests, too, very often financial ones related to the drug industry.

Even people who should know better can be disappointing. A clinical pharmacologist acknowledged at a public meeting that the drugs don’t work but he recommended that they should be tried, as they work better in some people than in others. I asked him if he had never heard about statistical variation. With his argument, we could use whatever we pleased that doesn’t work.

The perspective is chilling. Doctors are like children. They cannot keep their fingers away from dangerous toys, which is why we should take all the ineffective and dangerous psychiatric drugs off the market. I suggested this in a newspaper article in 2014.189

As I doubted it could be true that risperidone and olanzapine work for dementia,16:127 which no drugs do, I browsed the Internet and found a trial of olanzapine.568 I had been duped again. It was not about having an effect on dementia but about calming down disturbing Alzheimer patients with a major tranquilliser, and the patients became somnolent and developed gait disturbances. I also found a Cochrane review, but this was also not about treating dementia but about treating aggression and psychosis in people with dementia. Everything I found was about this.

This book noted that the effect of depression pills is very limited and added that a minority without depression develop depression after discontinuation.16:131 This is interesting because it is an iatrogenic harm, an abstinence depression (see Chapter 8, Part Twelve).

Yet again, this was not about treating dementia, it was about treating depression in people with dementia. I found a Cochrane review, which was also discouraging.569 It noted that the data were of variable quality and unsupportive: “On the only measure of efficacy for which we had high‐quality evidence (depression rating scale scores), antidepressants showed little or no effect.”

As noted earlier, it is likely that all psychotropic drugs can cause chronic brain damage,5,135 which may be permanent. A hallmark of this is impaired cognitive function. Chronic brain damage is related to the length of drug exposure and often worsens when the dose is increased, whereas it will usually improve considerably when drugs are tapered off. If it had been the disease that caused the problems, the patients should have become worse when the drugging was reduced.135 A 17-year follow up of the Framingham Heart Study found that use of depression pills increased the risk of developing dementia by about 50%,570 and benzodiazepines seem to double the risk of dementia.571

We should avoid drugging demented people. We should care for them. A systematic review of 33 trials of agitated demented people showed pretty large effects of care, e.g. communication skills training, activities, music, touch, massage and talking to people.572


To see the list of all references cited, click here.


Mad in America hosts blogs by a diverse group of writers. These posts are designed to serve as a public forum for a discussion—broadly speaking—of psychiatry and its treatments. The opinions expressed are the writers’ own.


Mad in America has made some changes to the commenting process. You no longer need to login or create an account on our site to comment. The only information needed is your name, email and comment text. Comments made with an account prior to this change will remain visible on the site.


  1. In my days of specializing in clinical and research neuropsychology I remember that AZ patients’ motor skills were quite well preserved, at least in the mild to moderate stages, while memory and higher level skills progressively worsened. I suspected that supposed drugs to slow down the AZ dementing process were, like psychiatric drugs in general, also cognitively impairing, which would increase the AZ patient’s dementia. However at the time this was only a conjecture without any hard proof. However, I learned from Dr. Gotzsche’s current chapter that these AZ
    “slowing” drugs are associated with increased falls in AZ patients which to me suggests that these drugs provide increased brain impairment to these AZ patients!

    Report comment

    • Interesting to me that it has always been somehwat of a conjecture that medications for AZ help at all, given that there is no solid physiopatologic basis for their effectivenes (some form “of just works”) and their propangadized effectivenes comes from more or less conjectural empiricism.

      And as it seems as maybe you point out, without careful enough or proper balance between harms/risks and benefits. Adding the costs in using and of supervision of using said medications.

      And what is the real effect between the miriad molecules like neurotransmitters, histopathology (tangles, plaques, brain atrophy, etc) and the signs/symptoms is not only conjectural to me, but frankly chaotic, and at least part of it, relevant or not to treatment, fraudulent as has been shown in the past 3 months in major worldwide newspapers.

      And even if all the parts of the disease process were to fit in a grand scheme of non-conjectural physiopathology, in the art and science of therapeutics, given the large number of moving parts with feedback, I doubt will provide enough guidance to even call it conjectural treatment.

      But that is my conjecture, probably not a fact.

      Report comment

  2. That is indeed a chilling picture. In the future, we will all become old and find ourselves facing challenges where our rights could be at risk, potentially entrusted to medical professionals. In such cases, a hierarchical unidirectional power structure can be akin to “007 a license to kill” movie.

    I’ve encountered relatives with memory issues, and I have had difficulty to accept if their diagnosis as dementia or early dementia were accurate. One person constantly recounted childhood stories, and his dear children had distanced themselves. Another attempted to form connections by offering money to anyone, but others simply found her annoying due to her never stopping talking. Someone else had numerous near-death experiences and was aware of his impending death. I witnessed one elderly individual silently weeping at a bustling food table while her children and grandchildren reveled in noisy merriment, when what she likely needed was tranquility and genuine personal attention.

    Interestingly, the elderly relatives I knew who engaged daily in meaningful conversations and were not in imminent danger did not experience significant memory problems.

    As for myself, I’ve noticed that when a particular thought occupies my mind persistently, it becomes challenging to recall other information. This effect happens also when I’m stressed. If I were elderly and faced constant loneliness or health issues, it’s conceivable that someone might label it as dementia, with memory tests potentially confirming this. I can easily see that worsening as time goes on.

    Report comment

    • And as for “our” future, with not enough people to chip in to our future heatlh insurance or retirement health benefits. The proportion of the working aged in relationship to the retirees in some places is approaching 2 to 1, when I think in some places it requires at least 4 to 1, even 8 to 1. Just for funding, human caring, ex robots, might probably require, depending on the complexity 1 to 2 per retirees, or at the best 1 to 4 for retirees.

      Meaning between 20 and 30%, probably ex minors and students, will have to work caring for retirees and that is in direct competition with caring for children for people who work caring for retirees!. Which to me is a big problem now in the US, not stated as I did, but I think the math when done carefully might make my case.

      Report comment

    • As for the change in performance with age, the old rule is that to be signifcant the apparent decline in performance must be around, or at least 2 standard deviations from the mean.

      That is, folks reasonably suspected to have a disease have to have a performance that is at or below the 5% of the rest of the population. Considered somehow “normal” in the statistical sense. Aware that if AZ affects 10% of the population then a better “cuttof” would be 10% not 5%.

      And when it comes to aging, human development at the rigth long tail of the distribution, at around 65yrs, if I remember correctly, is 2 standard deviations below “peak performance”, around 30-40yrs of age. Cuestionable in recent discourse, but useful guidance I think.

      Second consideration that might translate to something like a person suspected to have dementia has to perform 4, four, standard deviations below his or her peak performance, as in individualized maner, to be statistitically significant in those over 65yrs. Again, taking into account the prevalence in over 65s of dementia, not only AZ, which might be more than 5-10-15%.

      Unorthodox, but I think my approach and possibly misunderstanding can give something to compare to, to remove some subjectivity, aware it has some caveats that I ignore.

      Report comment

    • And sorry for extending my comments. But I think it can clarify further.

      As a corollary in diseases/conditions/spectrums that affect less than 5% of the population like the so called bipolar disorder or schizophrenia, performance in any metric, in principle, like in these two cases has to be at or below 1% of the rest of the population, not 5%. And it depends on at least age and gender. Not equal at 20, 30, 50 or 60yrs of age. Frequency of the “disease” and performance are age related.

      To try to match the metric with the “diseased” population. That is excluding people without the “disease” and including people with the “disease”. Ignoring for simplicity that a metric might reflect another condition, a differential diagnosis, that might explain the result of the metric.

      Like medication before or during diagnosis that can increase or decrease a given performance, also not used, not understood or dismissed by psych professionals. And that probably has no reliable research to incorporate in the diagnositc process.

      Concept embodied in the receiver operating characteristic curve or ROC curve. To my mind poorly understood and severy lacking in the practice of psychaitry where subjective scoring, even if 80% agreed among practitioners is used, instead of solid, validated, compared to comparable populations.

      As the case of dementia, and AZ exemplifies with at least regarding age of the suspect.

      To exemplify: manic behaviour has to score at or below 1% of the population, same for hallucinations, poor judgement, inability to do stuff, isolation, etc. Not just subjective creepy crappy as the Bender-Gestalt test, developed by a psychiatrist infamous for giving LSD to minors to treat them for psychosis in childhood. Among others.

      So compared to what I described as approach to statistically and clinicaly sound diagnosis psychiatric practice looks to me like eeny meeny miny moe.

      And even worse, knowing how many, what percentage of a population at a given age has a disease/condition/spectrum requires a GOLD STANDARD, to establish truly, not inventing as in fiction, the real percentage of the population with said “disease”.

      Which psychiatry and clinical psychology never, ever have had. And if they did it would probably make the disease like AZ and many dementias a neurological disease, not a psychiatric one…

      Thanks for the patience.

      Report comment