Memorandum by Professor David Healy (PI
2. In response to the select committee's
call for comments, I wish to offer an analysis that may help explain
how companies can engineer a clinical consensus that will favour
their product—even in the absence of a scientific basis
for claims for superiority for the new and usually much more expensive
product—and how this process can feed through and shape
even disinterested assessments of the evidence undertaken by NICE.
These processes appear to many to underpin a relatively recent
"capture" of regulatory and other professional domains.
3. I will also comment on how the safety
of pharmaceutical products might be better ensured under current
arrangements and on how current arrangements may be changed to
produce a safer framework for healthcare.
4. I bring to this commentary a background
in consultancy work with pharmaceutical companies, access to company
archives through medico-legal work, (as outlined in the accompanying
conflict of interest statement—appendix 1), and 10 years
of work on mapping the history of the development of psychotropic
drugs and pharmaceutical companies in their present form as published
by Harvard and New York University Presses.
6. Clinical Consensus: The main point
of note is that the scientific literature is being managed to
an ever increasing extent, the most visible sign of which is that
an increasing proportion of the scientific literature on aspects
of therapeutics with pharmacological agents (over 50%) is now
ghost-written. This is associated with a demonstrable failure
to report important safety data on drugs, or with the reporting
of such data in terms that mislead. The risks from ghost-writing
stem from the fact that the data that ghostwriters purport to
represent remain inaccessible to outside scrutiny.
7. As a consequence of the ghost-writing
of articles, efforts to establish clinical consensus in the form
of guidelines and algorithms, which depend on a review of the
entire literature on a drug are compromised and can be shown to
produce outcomes indistinguishable from the outcomes that would
be produced by having a group of employees of the pharmaceutical
company write the guidelines. Even though there may in fact be
no evidence of superior efficacy or safety for newer agents, they
will end being written up as superior to older agents and to be
used in preference to older agents.
8. The effects of ghost-writing and other
literature management strategies feed through into a process of
pharmaco-economic modelling undertaken for new drugs, which can
demonstrate that not only is a drug that is not proven to be in
any way superior its predecessors better than its predecessors
but is better to the extent that it should be in replace older
agents, even though the costs of the new treatment may be 50 times
greater than the older agent.
9. The details of how these processes work
are laid out in appendix 2. Of note here is that I have been a
participant in these processes and party to the generation of
views favouring newer over older agents, unaware that pharmaceutical
companies were keeping key safety data hidden from the scientific
community and that they would refuse to produce such data on request.
10. Safety Data: Current procedures
to manage the entry of drugs onto the market favour the detection
of drug effects and set a higher threshold for safety effects.
This shows up in two ways.
11. First, in order for a drug to be licensed
it has to show superiority to placebo in two controlled trials.
Companies however can run ten or more trials in carefully selected
samples using instruments carefully designed to pick up any effect
in order to demonstrate this, and even if the results show the
drug failing to beat placebo in the clear majority of trials,
this is not held against them. These other trials are commonly
termed failed trials rather than drug failures.
12. In contrast, the demonstration of a
safety issue is not handled in this way. In this case regulators
will only act if the overwhelming preponderance of the data show
13. This issue has at its heart unresolved
philosophical issues about the nature of statistics. Safety data
is typically presented in terms of Confidence Intervals, so that
for instance in recent antidepressant studies the rate of suicides
on drugs compared to placebo is typically of the order of two
times greater but what is termed the confidence interval surrounding
this figure of 2.0 might be for example 0.9 to 4.4.
14. There is a deep philosophical divide
between two ways to interpret such a finding. First according
to a school of thought stemming from R A Fisher is the view that
nothing has in fact been shown unless the confidence interval
does not include 1.0, thus for instance a confidence interval
of 1.1 to 4.4.
15. Second, the Neymann-Pearson school of
thought argues that the best estimate of the effect is 2.0, in
this case. This is the figure regarding which we can be most confident.
16. In practice regulators, adopt the Fisher
approach. This cannot be viewed as a rigorous approach to safety.
Epidemiologists or a drug regulatory process on the other hand,
concerned with safety, would argue that in this case the figure
of 4.4 is potentially consistent with the data—that is the
hazard in question may well happen up to 4.4 times more often
on the drug than on placebo or non-treatment, and that therefore
if the hazard is serious patients and doctors should be warned
about this possibility.
17. Safety Procedures and Informed Consent:
At present when patients enter clinical trials they are asked
to sign informed consent forms, which contain ever longer lists
of potential problems a treatment may cause.
18. These consent forms though never inform
patients or others that neither patient, nor physician nor any
other third party will ever have access to the raw data from this
trial, particularly data on safety issues. They are not told that
pharmaceutical companies may chose only to "market"
the bits of the data that suit them.
19. Changing informed consent forms to make
this explicit might have salutary effects. Alternatively transforming
informed consent forms into contracts between patients and companies
that provided rights of access to experts in the event of safety
concerns might make a big difference.
20. Requiring companies to permit access
to the raw data, given for free by patients as part of a "gift"
arrangement with the rest of the community, would simply require
companies to conform to the norms of science under whose banner
they claim they sell their products.
21. Until some arrangement is put in place
to ensure access, then every patient who enters a clinical trial
in the United Kingdom (or anywhere else) is putting every Member
of Parliament and all the constituents those Members represent
in a state of legal jeopardy. This follows as the absence of publicly
available data on hazards will be used by pharmaceutical companies
to argue that the hazard does not exist, even when the datasets
available to industry establish clearly that the hazard does exist.
22. Regulators, who typically read study
summaries prepared for them by pharmaceutical company employees,
rather than scrutinise and analyse the raw data themselves, can
miss these issues.
23. This evidence is being submitted on
an individual rather than a corporate basis. I would be willing
to give oral evidence on these issues, and can bring some benefit
to any hearings by virtue of being an active participant in the
processes described above and below.
24. I am a Reader in Psychological Medicine
in the University of Wales College of Medicine (as of August Cardiff
20 July 2004