1. Comparative effectiveness research:
understanding and designing the
geometry of the research agenda
John P.A. Ioannidis, MD, DSc
C.F. Rehnborg Chair in Disease Prevention
Professor of Medicine and Professor of Health Research and Policy
Director, Stanford Prevention Research Center
Stanford University School of Medicine
Professor of Statistics (by courtesy)
Stanford University School of Humanities and Sciences

2. I want to make big money
• My company, MMM (Make More Money, Inc.),
has successfully developed a new drug that is
probably a big loser
• At best, it may be modestly effective for one or
two diseases/indications for one among many
outcomes
• If I test it in RCTs, even for this one or two
indications, it may seem not to be worth it
• But still I want to make big money
• Please tell me: What should I do?

3. The answer
• Run many trials (this is the gold standard of research) with many
outcomes on each of many different indications
• Ideally against placebo (this is the gold standard for regulatory
agencies) or straw man comparators
• Test 10 indications and 10 outcomes for each, just by chance you
get 5 indications with statistically significant beneficial results
• A bit of selective outcome and analysis will help present
“positive” results for 7-8, maybe even for all 10 indications
• There are meta-analysts out there who work for free and who will
perform a meta-analysis based on the published data
SEPARATELY for each indication proving the drug works for
all 10 indications
• We can depend also on electronic databases to give us more
evidence on additional collateral indications. People working on
them also do it for free, this research is tremendously
underfunded.
• We love that all these people work for us and don’t even know it!
• With $ 1 billion market share per approved indication, we can
make $ 10 billion a year out of an (almost) totally useless drug

4. We probably all agree
• It is stupid to depend on the evidence of a
single study
• when there are many studies and a meta-
analysis thereof on the same treatment
comparison and same indication

5. Similarly
• It is stupid to depend on a single meta-analysis
• when there are many outcomes
• when there are many indications the same
treatment comparison has been applied to
• when there are many other treatments and
comparisons that have been considered for each of
these indications in randomized and non-
randomized evidence

6. Networks and their geometry
• Networks can be defined as diverse pieces of data
that pertain to research questions that belong to a
wider agenda
• Information on one research question may
indirectly affect also evidence on and inferences
from other research questions
• In the typical application, data come from trials on
different comparisons of different interventions,
where many interventions are available to
compare

7. Network geometry offers the big picture: e.g.
Size of each node proportional to the
making sense of 700 trials of advanced breast
Figure 2a amount of information (sample size)
cancer treatment
AT SD
Tc
AN SD T+tzmb
A+tzmb SD Ts+lpnb
Os
A c SD Ts
A s SD
NT
ANT SD
Ns
A s LD Oc
N+lpnb
M c SD
Nc N+bmab
M c LD
A c LD
M s SD
Mauri et al, JNCI 2008

8. Main types of network geometry
Polygons
Stars
Lines
Complex figures
Salanti, Higgins, Ades, Ioannidis, Stat Methods Med Res 2008

9. Diversity and co-occurrence
• Diversity = how many treatments are
available and have they been equally well
studied
• Co-occurrence = is there preference or
avoidance for comparisons between specific
treatments
Salanti, Kavvoura, Ioannidis, Ann Intern Med 2008

10. Diversity and co-occurrence can be easily measured and
statistically tested

13. Homophily
• OΜOΦΙΛΙΑ = Greek for “love of the same” =
birds of a feather flock together
• Testing for homophily examines whether
agents in the same class are disproportionately
more likely to be compared against each other
than with agents of other classes.

14. For example: Antifungal agents
agenda
• Old classes: polyenes, old azoles
• New classes: echinocandins, newer azoles

15. Rizos et al, 2010

17. • Among polyene and azole groups, agents were
compared within the same class more often than
they did across classes (homophily test p<0.001
for all trials).
• Lipid forms of amphotericin B were compared
almost entirely against conventional amphotericin
formulations (n=18 trials), with only 4
comparisons against azoles.

18. Figure 2
posaconazole
1
3
lipid amphotericin B
fluconazole 1 1
11 1
18
3
17
1 4 amphotericin B
2
itraconazole 2
2
voriconazole ketoconazole

19. • There was strong evidence of avoidance of
head-to-head comparisons for newer agents.
Only one among 14 trials for echinocandins
has compared head-to-head two different
echinocandins (p<0.001 for co-occurrence).
Of 11 trials on newer azoles, only one
compared a newer azole with an
echinocandin (p<0.001 for co-occurrence).

20. Figure 3
anidulafungin
2
other caspofungin
8
3 1
micafungin

21. e4
12
other echinocandins
10 1
voriconazole or posaconazole

22. Auto-looping
Design of clinical research: an open world or isolated city-states (company-states)?
Lathyris et al., Eur J Clin Invest, 2010

23. Reversing the paradigm
Design networks prospectively
– Data are incorporated prospectively
– Geometry of the research agenda is pre-
designed
– Next study is designed based on enhancing,
improving geometry of the network, and
maximizing the informativity given the network

24. This may be happening
already?
Agenda-wide meta-analyses
BMJ 2010

25. Anti-TNF agents: $ 10 billion and 43 meta-analyses,
all showing significant efficacy for single indications
5 FDA-approved anti-TNF agents
Infliximab
Etanercept
Adalimumab
1998 Golimumab
Certolizumab pegol
Indications Psoriatic
2003 Psoriasis
arthritis
1998 Juvenile
RA Ankylosing
idiopathic
spondylitis
arthritis
Crohn’s Ulcerative
disease colitis

26. 1200 (and counting) clinical trials of
bevacizumab

27. Fifty years of research with 2,000 trials:
9 of the 14 largest RCTs on systemic steroids
claim statistically significant mortality benefits
Contopoulos-Ioannidis and Ioannidis EJCI 2011

28. How about non-randomized
evidence?
• Epidemiology
• Cohort studies
• Electronic record databases
• Registries
• Propensity adjusted effects
• Biobanks
• Patient-centered outcomes research
• You name it

29. Comparisons between randomized and
non-randomized evidence
Ioannidis J. et al. JAMA 2001;286:821-830.
7 pairs discrepancies beyond chance

30. Inflated effects for cardiovascular
biomarkers in observational datasets
Tzoulaki, Siontis, Ioannidis, BMJ 2011

32. Inflation in statistically significant treatment
effects of meta-analyses of randomized trials
Ioannidis, Epidemiology 2008

33. Some systemic changes
• Public upfront availability of protocols
• Public eventual availability of raw data
• Public upfront availability of research
agendas
• Reproducible research movement

35. Science 2011

36. So, what the next study should do?
• Maximize diversity
• Address comparisons that have not been addressed
• Minimize co-occurrence
• Break (unwarranted) homophily
• Be powered to find an effect or narrow the credible or
predictive interval for a specific comparison of interest
• Maximize informativity across the network (entropy
concept)
• Some/all of the above

37. Bottomline
We need to think about how to design
prospectively large agendas of clinical
studies and their respective networks
This requires a paradigm shift about the
nature and practice of comparative
effectiveness research, compared with
current standards