In an editorial titled "Should journals stop publishing research funded by the drug industry? " in the British Medical Journal the authors make the following claim:

Another dire example is antipsychotics. Many recent drug industry crimes are related to off-label promotion of antipsychotics, and in the United States they were the most sold drugs in 2009. However, they are so dangerous that just one of them, olanzapine (Zyprexa), has probably caused 200 000 deaths.

The citation for this claim is the book "Deadly medicines and organised crime: how big pharma has corrupted health care." by Peter C. Gøtzsche, one of the authors of the editorial.

The number of deaths mentioned in this claim seems extremely large to me, and I wonder how the authors arrived at this number and how credible it is.

  • 1
    Damn, you asked it before I did. Good question an important topic. – matt_black Jan 17 '14 at 15:53
  • 1
    ncbi.nlm.nih.gov/pubmed/22249081 People with schizophrenia have a 10-25 year reduced life expectancy, partly due to medication. So it is plausible that widely used anti psychotics have killed hundreds of thousands. – D J Sims Mar 1 '16 at 20:34

The evidence does not suppor the claim conclusively.

There is a 1% increased risk of death per year due to atypical anti psychotic drugs such as olanzapine:

The risk differences for death in patients treated with [...] olanzapine vs placebo, 0.01 (95% CI, −0.00 to 0.03; P = .07)

olanzapine effect

--Risk of Death With Atypical Antipsychotic Drug Treatment for Dementia

Olanzapine sells about 1,000,000 units per month in the US (Ref.).

enter image description here

This, in a year, will statistically lead to around 10,000 more deaths than placebo per year, assuming that the dementia patient data can be extended to the whole set of olanzapine users, that a unit corresponds to a month's worth of medicine and that the figure is representative of a typical month.

Olanzapine was released in 1996. This means 17 years of use, or 170,000 possible statistical deaths. This is probably an upper limit as it would be surprising to see a figure of 1 million units per month in the first year of sale, for example.

That said, the effect measured is barely statistically significant, so we can't confidently conclude that there were any extra deaths.

In any case the 200,000 figure doesn't seem to be justified by the numbers, as the best upper estimate I can come up with is 170,000, even though, if the effect was confirmed, it would mean that a substantial number of people died because of this molecule.

  • I think you have done a good job of summarising the evidence, but it might be worth adding error bars or confidence intervals to the estimated number. I haven't crunched the numbers but that might look like estimated deaths per year could be between 0 and 15k given the confidence intervals. – matt_black Jan 19 '14 at 14:00
  • @matt_black the most important "error bar" is the one in measuring if the molecule kills people. If we consider the error bar, the lower bound is exactly zero effect whereas the higher bound is about 3 times the effect (this is already in my answer). So that alone would bring our death count to "from zero to half a million people", but this would be misleading, because either the medicine has an deadly effect or it doesn't. Since the second case is not ruled out, "taking an average" between the two values is not warranted. – Sklivvz Jan 19 '14 at 14:12
  • I appreciate the response but I question whether natural deaths should be included in such total. Zyprexa can cause diabetes and health heart problems according to the listed side effects. People with serious mental illnesses are definitely dying younger, so the question becomes where is it coming from. If someone dies from diabetes incredibly young they don't list Zyprexa if that is what cause it. You technique seems like the best approach considering the data. – William Apr 2 '18 at 20:43
  • @William the fact is that this is tested against placebo, which should take care of the number of deaths that would have occurred within the group without the medicine. – Sklivvz Apr 2 '18 at 22:26
  • 1
    @William randomized blind studies do not determine cause and effect of singular cases. They study the statistical effects over groups. If the two groups are statistically similar at the beginning, composition matters much less. – Sklivvz Apr 2 '18 at 23:09

The number of deaths mentioned in this claim seems extremely large to me, and I wonder how the authors arrived at this number and how credible it is.

The authors of the article use the following book as a reference for that number:

Deadly Medicines and Organised Crime: How Big Pharma Has Corrupted Healthcare

Here is the relevant passage from a sample chapter of the book:

Let’s estimate how many people Lilly has killed with Zyprexa. In 2007, it was reported that more than 20 million people had taken Zyprexa.78 A meta- analysis of the randomised trials of olanzapine and similar drugs given to patients with Alzheimer’s disease or dementia showed that 3.5% died on drug and 2.3% on placebo (P = 0.02).79 Thus, for every 100 patients treated, there was one additional death on the drug. Elderly patients are often treated with several drugs and are more vulnerable to their harms, which means that the death rate is likely higher than in younger patients. However, the reviewed trials generally ran for only 10–12 weeks, and most patients in real life are treated for years. Further, drugs like Zyprexa are most used in the elderly, and as deaths are often underreported in trials, the true death rate is likely higher than shown in the meta- analysis. One death in a hundred therefore seems a reasonable estimate to use. I therefore estimate that 200 000 of the 20 million patients treated with Zyprexa have been killed because of the drug’s harms. What is particularly saddening is that many of these patients shouldn’t have been treated with Zyprexa.

The author of the book cites the following two sources when deriving that answer:

The author of the book is Peter C. Gøtzsche:

Peter C. Gøtzsche is a Danish medical researcher, and leader of the Nordic Cochrane Center at Rigshospitalet in Copenhagen, Denmark. He has written numerous reviews within the Cochrane collaboration.

  • Well done for finding the sources. You can't simply multiply 20 million patients and 1% chance. 1% is for patient-year and not simply for patient (it's in your 3rd link). Thus there needs to be another source to support that the average length of usage is exactly 1 year to support that number. As it stands, I see no reason to believe that (even the trials are of 10-12 weeks, so it's possible that this figure is a typical usage length) – Sklivvz Dec 16 '14 at 8:14
  • @Sklivvz a) I didn't "simply multiply": the author did (I simply quote the author). b) I've added a quote of the relevant extract from the book. c) The author IMO hand-waves whatever difference there might be between "patients", "patient-years", and patient-10-or-12-weeks", and between vulnerability of older versus younger patients, but in defence of that he claims that "most patients in real life are treated for years": which I find at least superficially plausible based on my (inexpert) understanding of psychosis and dementia etc. – ChrisW Dec 16 '14 at 12:36
  • My "you" was meant to be a generic you. – Sklivvz Dec 16 '14 at 12:38
  • 1
    In summary, IMO a 1% death rate is disastrous. It might barely be justifiable if the only alternative is untreated psychosis, but IMO difficult to justify off-label use and commercial promotion. – ChrisW Dec 16 '14 at 12:38
  • It's not considered disastrous in the the literature review you provided: "Atypical antipsychotic drugs may be associated with a small increased risk for death compared with placebo". OTOH I totally agree with you, to me it seems high but I admit ignorance on the matter. – Sklivvz Dec 16 '14 at 12:41

The makers of Zyprexa were sued, in a class action suit, by people claiming it gave them diabetes. They paid to settle the suit. 18,000 people. Make of that what you will.

Eli Lilly Faces More Drug Injury Lawsuits Over Zyprexa

Back in 2007, Eli Lilly, the maker of Zyprexa, agreed to pay out $500 million to settle 18,000 lawsuits.

Zyprexa linked to Diabetes, Hyperglycemia

According to The New York Times (January 5, 2007), in 2007, Eli Lilly agreed to pay up to $500 million to settle 18,000 lawsuits alleging patients developed diabetes or other serious side effects after taking Zyprexa. By early 2007, according to the newspaper, Lilly had agreed to pay at least $1.2 billion to approximately 28,500 people who alleged the use of Zyprexa caused them harm.

  • 1
    This doesn't answer the question, which is asking whether it "probably caused 200 000 deaths". – ChrisW Dec 15 '14 at 23:20

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .