Population Health Research Institute

McMaster University/Hamilton Health Sciences

David Braley Cardiac, Vascular and Stroke Research Institute
237 Barton Street East
Hamilton, Ontario
Canada L8L 2X2
www.phri.ca

June 30, 2010 Salim Yusuf, MBBS, DPhil,

FRCP(UK), FRCPC, FACC, FRSC
Dr. Mary Parks Telephone: (905) 527-7327
FDA Fax: (905) 297-3781
mary.parks@fda.hhs.gov

Re: The urgent need to use reliable methods to assess the true effects of glucose lowering drugs and other
therapies on major cardiovascular outcomes.

Dear Dr. Parks,

Please excuse me for writing to you and to your committee spontaneously. First, permit me to introduce
myself. Second, allow me to indicate my interest in the issue and third, I would like to make a few relevant
points that may be valuable to your committee.

I am a Professor of Medicine, Executive Director of the Population Health Research Institute at Hamilton
Health Sciences and McMaster University in Canada. I have been involved in developing the methodology
of both meta-analysis, as well as large, simple, randomized trials since 1976 while I was at Oxford
University in England, with Prof Sir Richard Peto and Prof Peter Sleight. Subsequently, from 1984 to 1992,
I worked at the National, Heart, Lung and Blood Institute in the clinical trials branch in conjunction with
Dr. Curt Furberg and Dr. Larry Friedman, and eventually became the acting Branch Chief of the Clinical
Trials Division. I have been the Principal Investigator or chairman of over 60 major trials of several
therapies in cardiovascular disease and in diabetes, some of which are now commonly used in clinical
practice. I have published extensively on meta-analyses, the misuse and the dangers of using database
analyses for the evaluation of therapies, and I have been instrumental in developing a significant part of the
methodology that is now widely used for the conduct of large trials.

Along with Dr. Hertzel Gerstein, I am the joint Principal Investigator of the TIDE study, which is
evaluating two thiazolidinediones (roziglitazone and pioglitazone), as well as vitamin D in high risk
diabetes patients. Please note that this is one of several trials in diabetes that we are involved in, and one of
over 50 trials that the Population Health Research Institute has been involved with, or is currently
conducting. I do not write to you and your committee to express my views on the TIDE study. Instead, I
write to you because of my growing concern that the debate in the media, as well as in the literature, is
based on poor science. It is essential that any recommendation that your committee and FDA come to is
based on ensuring that we continue to support the best scientific methods. Poor methods will lead to
unreliable conclusions and we will be left forever with considerable uncertainty.

Below, I will list my concerns with the current debate:

Director: Salim Yusuf D.Phil,
Associate Directors: Koon Teo MB, PhD, Sonia Anand MD, PhD,
Chief Operating Officer: Deidre L. Henne CA,
Khursh Ahmed MSc, PhD, Heather M. Arthur RN, PhD, Jackie Bosch MSc, BSc.OT, Susan Chrolavicius RN, BA, Stuart Connolly MD, Beena Cracknell,
Catherine Demers MD, MSc, P.J. Devereaux MD, John Eikelboom MBBS, Ernest Fallen MD, Hertzel Gerstein MD, MSc, Jeff Healey MD, MSc, Andre Lamy MD,MHSc, Eva Lonn MD, MSc,
Patrick Magloire MD, Kamil Malikov MD, MSc, MBA, Robert Mckelvie MD, PhD, Matthew McQueen MB, ChB, PhD, Shamir Mehta MD, Carlos A. Morillo MD, Katherine Morrison MD, Madhu
Natarajan MD, Martin O’Donnell MD, Janice Pogue MSc, Parminder Raina PhD, Arya M. Sharma MD, Tej Sheth MD, Susan Stansbury,
Ellison Themeles MSc
 Population Health Research Institute
McMaster University/Hamilton Health Sciences
David Braley Cardiac, Vascular and Stroke Research Institute
237 Barton Street East
Hamilton, Ontario
Canada L8L 2X2
www.phri.ca
1. Small meta-analyses are not reliable: The meta-analysis by Nissen et al published in the New
England Journal of Medicine (2007; 356(24): 2457-71) has created considerable interest in the issue
of whether rosiglitazone has adverse effects on cardiovascular outcomes. However, this meta-
analysis is based on very few events, other meta-analyses have not confirmed the conclusions, and
most notably there is no justification as to why this meta-analysis focused solely on rosiglitazone,
but was not based on all thiazolidinediones. In particular, most meta-analyses that have been
conducted in the literature (such as beta-blockers in myocardial infarction or heart failure, ACE-
inhibitors in the same conditions, thrombolysis, cholesterol lowering agents, etc) have always been
based on examining a class of drugs as a whole, unless there was specific reasons to believe that
specific drugs in a class markedly differed from other drugs. In the case of the glitazones, there is
no compelling evidence that one glitazone differs materially from the other on major clinical
outcomes. Indeed, an AHRQ commissioned systematic TZD drug class review concluded that data
were “not sufficient to determine the comparative effectiveness of pioglitazone and rosiglitazone on
microvascular or macrovascular complications of diabetes as there are no head- to-head data and
indirect data are sparse” (http://www.ncbi.nlm.nih.gov/bookshelf/br.fcgi?book=tzd). Thus, the most
relevant analysis would be a meta-analysis involving ALL the trials of glitazones versus placebo.
Such information is available to the FDA, and I would urge you to use such an approach based on
the TOTALITY of all data, in preference to a meta-analysis done on a single agent within a class.
Such an approach would avoid the potential data-dependent emphasis on one or another agent that
leads to “selection” biases.

As explained in an article that Janice Pogue and I wrote some 12 years back on overcoming the
limitations of meta-analysis (Lancet 1998; 351(9095):47-52), meta-analyses based on small
numbers of events can be substantially misleading. Practical examples of these come from our
own experience where we have been misled by meta-analyses of trials based on scant data. This
includes a meta-analysis of magnesium in acute MI, where we observed a 50% reduction in
mortality (p<0.001) (Teo et al BMJ 1991; 303(6816): 1499-503). Instead of relying on this as
conclusive evidence, despite the extreme statistical significance, we conducted a large prospective
study (the ISIS-4 study on about 60,000 people, Lancet 1995; 345(8951): 669-85), which
demonstrated absolutely no benefit from magnesium, but instead, a trend towards harm. Therefore,
even directionally reliable information cannot be obtained from meta-analysis based on scant
data. This instance is not isolated. Similar misleading information has come from a meta-analysis
of beta-blockers of small trials in avoiding peri-operative cardiovascular events. The apparent
benefit claimed by this meta-analysis was contradicted by the large POISE study that we conducted
(Lancet 2008; 371(9627): 1839-47). Conversely, suggestions of harm with calcium channel
blockers that were based on small trials were shown to be wrong by large prospective trials that
demonstrated benefit. Therefore, small meta-analysis, at best, can provide a hypothesis that needs to
be tested in prospective well-designed large trials.

Director: Salim Yusuf D.Phil,

Associate Directors: Koon Teo MB, PhD, Sonia Anand MD, PhD,
Chief Operating Officer: Deidre L. Henne CA,
Khursh Ahmed MSc, PhD, Heather M. Arthur RN, PhD, Jackie Bosch MSc, BSc.OT, Susan Chrolavicius RN, BA, Stuart Connolly MD, Beena Cracknell,
Catherine Demers MD, MSc, P.J. Devereaux MD, John Eikelboom MBBS, Ernest Fallen MD, Hertzel Gerstein MD, MSc, Jeff Healey MD, MSc, Andre Lamy MD,MHSc, Eva Lonn MD, MSc,
Patrick Magloire MD, Kamil Malikov MD, MSc, MBA, Robert Mckelvie MD, PhD, Matthew McQueen MB, ChB, PhD, Shamir Mehta MD, Carlos A. Morillo MD, Katherine Morrison MD, Madhu
Natarajan MD, Martin O’Donnell MD, Janice Pogue MSc, Parminder Raina PhD, Arya M. Sharma MD, Tej Sheth MD, Susan Stansbury,
Ellison Themeles MSc
 Population Health Research Institute
McMaster University/Hamilton Health Sciences
David Braley Cardiac, Vascular and Stroke Research Institute
237 Barton Street East
Hamilton, Ontario
Canada L8L 2X2
www.phri.ca
2. Observational analysis of databases can be extremely misleading in evaluating therapies: Some 20
years ago, I along with Drs. Curt Furberg, Kent Bailey and Janet Wittes, published an article in
Circulation titled “Digitalis: a new controversy about an old drug. The pitfalls of inappropriate
methods” (Circulation 1986; 73(1): 14-8). At that time, five database analyses claimed an increased
mortality with digitalis and yet when a large prospective trial was conducted it showed a neutral
effect on mortality, but an important reduction in hospitalizations for heart failure, which was
clinically relevant. We know of many other instances where database analyses are misleading. This
includes the recent story of hormone replacement therapy, where several analyses of observational
studies suggested substantial reduction in cardiovascular events. However, the large Women’s
Health Initiative trial on hormone replacement therapy indicated substantial harm. Similar
misleading results have been obtained with other therapies as well. Again, these two examples
(digitalis and HRT) indicate that database analyses can lead to even directionally incorrect answers.

Therefore, irrespective of their results, observational analyses cannot generally be used for anything
other than the generation of hypotheses which in turn need to be tested in large clinical trials. An
exception to this would be the occurrence of rare outcomes such as thrombocytopenia or aplastic
anemia, or when large relative risks (in excess of 4 or 5) are observed.

3. Reliable results can only be obtained by large long-term trials that accrue over 1000 and ideally
several thousand events: The need for large trials to evaluate therapies reliably is well known (Stat
Med 1984; 3(4): 409-22). It is unfortunate that until recently very few large (e.g. over 10,000
people) and long term trials (e.g. over 5 yrs of exposure) in the field of diabetes have been
conducted. Unless trials are both large and prolonged, at best only uncommon early side effects
will be detected, as any intervention that is expected to reduce major vascular events through an
impact on atherosclerosis needs prolonged treatment of several thousands of individuals for a few
years for the real effects of the treatment to emerge. Thus any conclusion based on relatively few
patients followed for less than a year or two could well be misleading. A year or two ago, the FDA
took the very progressive step of demanding large studies of all new antidiabetic agents. However,
this leaves us with the dilemma of not knowing what the true long term effects of most currently
used glucose lowering agents on major, vascular and other events. This includes metformin, the
sulfonylureas, and insulin. Fortunately, at least for insulin and many of the new agents, large trials
are underway. Until such large and long term trials have been completed, it would be impossible for
us to obtain reliable information about any of the agents. The same holds true for the
thiazolidinediones.

Thus, given that the existing data are scant and unreliable, there is every reason to conduct well
designed trials, which are both large and prolonged that could generate reliable data (e.g. TIDE). I
would therefore hope that your committee will be supportive of rigorous methodology, the conduct
of such trials and the conclusions that arise from it.
Director: Salim Yusuf D.Phil,
Associate Directors: Koon Teo MB, PhD, Sonia Anand MD, PhD,
Chief Operating Officer: Deidre L. Henne CA,
Khursh Ahmed MSc, PhD, Heather M. Arthur RN, PhD, Jackie Bosch MSc, BSc.OT, Susan Chrolavicius RN, BA, Stuart Connolly MD, Beena Cracknell,
Catherine Demers MD, MSc, P.J. Devereaux MD, John Eikelboom MBBS, Ernest Fallen MD, Hertzel Gerstein MD, MSc, Jeff Healey MD, MSc, Andre Lamy MD,MHSc, Eva Lonn MD, MSc,
Patrick Magloire MD, Kamil Malikov MD, MSc, MBA, Robert Mckelvie MD, PhD, Matthew McQueen MB, ChB, PhD, Shamir Mehta MD, Carlos A. Morillo MD, Katherine Morrison MD, Madhu
Natarajan MD, Martin O’Donnell MD, Janice Pogue MSc, Parminder Raina PhD, Arya M. Sharma MD, Tej Sheth MD, Susan Stansbury,
Ellison Themeles MSc
 Population Health Research Institute
McMaster University/Hamilton Health Sciences
David Braley Cardiac, Vascular and Stroke Research Institute
237 Barton Street East
Hamilton, Ontario
Canada L8L 2X2
www.phri.ca
As you may well recognize, my views are not isolated, and a large part of the scientific community shares
this view. My hope is that good sense and good science will prevail and ultimately allow the conduct of
well designed studies so that reliable answers will emerge. Without this, science and our patients will be
“losers”, and we will never know whether the treatments that patients are receiving or denied are of any
particular value.

I trust these thoughts are of value to you and your committee. If you have any questions, I would be happy
to discuss them with you by telephone. At present I do not intend to participate in your two day hearings in
Washington, DC on July 13 and 14th.

Best wishes,
Yours sincerely,

Salim Yusuf, DPhil, FRCPC, FRSC

Professor of Medicine, McMaster University
Executive Director, Population Health Research Institute
McMaster University, Hamilton Health Sciences
Vice President Research, Chief Scientific Officer, Hamilton Health Sciences
Heart & Stroke Foundation of Ontario/Marion W. Burke Chair in Cardiovascular Diseases
SY:jl

Director: Salim Yusuf D.Phil,

Associate Directors: Koon Teo MB, PhD, Sonia Anand MD, PhD,
Chief Operating Officer: Deidre L. Henne CA,
Khursh Ahmed MSc, PhD, Heather M. Arthur RN, PhD, Jackie Bosch MSc, BSc.OT, Susan Chrolavicius RN, BA, Stuart Connolly MD, Beena Cracknell,
Catherine Demers MD, MSc, P.J. Devereaux MD, John Eikelboom MBBS, Ernest Fallen MD, Hertzel Gerstein MD, MSc, Jeff Healey MD, MSc, Andre Lamy MD,MHSc, Eva Lonn MD, MSc,
Patrick Magloire MD, Kamil Malikov MD, MSc, MBA, Robert Mckelvie MD, PhD, Matthew McQueen MB, ChB, PhD, Shamir Mehta MD, Carlos A. Morillo MD, Katherine Morrison MD, Madhu
Natarajan MD, Martin O’Donnell MD, Janice Pogue MSc, Parminder Raina PhD, Arya M. Sharma MD, Tej Sheth MD, Susan Stansbury,
Ellison Themeles MSc