1. Department of OUTCOMES RESEARCH

2. Lost in Translation www.or.org Daniel I. Sessler, M.D.
Michael Cudahy Professor and Chair
Department of OUTCOMES RESEARCH
Anesthesiology Institute
Cleveland Clinic
No personal financial interests
related to this presentation

3. Time held me green and dying, But I sang in my chains like the sea.
Dylan Thomas, 1954
Time held me green and dying, But I sang in my chains like the sea.

4. Dylan Thomas, shortly before dying
Do not go gentle into that good night, Old age should burn and rave at close of day; Rage, rage against the dying of the light.

5. Perioperative Mortality
Intraoperative mortality rare
10-fold reduction in 3 decades
Thirty-day postoperative mortality
1% nationwide in United States
2% for inpatients ≥45 years old
Half during initial hospitalization
1,000-fold more than intraop mortality
“Myocardial Injury after Non-cardiac Surgery” = MINS

6. Causes of Death
Bartels, Anesthesiology 2013

7. Common, Silent, and Deadly
MI incidence 8% among inpatients >45 years
≈10 million postoperative infarctions per year
Most MIs only detected by troponin
Only 15% report chest pain
65% entirely asymptomatic
Mortality identical after apparent & silent MIs
It’s not just “troponitis”
Mortality is 10% at 30 days
Twice as high as non-operative infarctions
VISION Investigators, JAMA 2012 and Anesthesiology 2014

8. Troponin T Predicts Mortality
“Prognosis define diagnosis”
Even slight troponin elevations predict death
Population attributable risk = 34%
Peak Troponin (ng/ml)
30-day Mortality (%)
Time to death (days)
≤0.01 1 —
0.02 4 13 9
≥0.3 17 6
Population attributable risks: Age – 40%, MINS – 34%, and sepsis – 30%.

9. Universal Definition of MI
“Most patients who have a perioperative MI will not experience ischemic symptoms. Nevertheless, asymptomatic perioperative MI is as strongly associated with 30-day mortality as symptomatic MI. Routine monitoring of cardiac biomarkers in high-risk patients … after major surgery is therefore recommended.”
Thygeson, Circulation 2012

10. Causes of Long-term Mortality
≈10% one-year mortality in patients ≥65 yrs
Cause of Death %
Cancer
52%
Cardiovascular
17%
Renal/Liver Failure 5%
Respiratory Failure 4%
Pulmonary Emboli 2%
Sepsis/Infection
Monk, A&A 2005

11. Postoperative Mortality
Mortality is the big problem facing anesthesia
Short-term: myocardial injury
Long-term: cancer
Most undoubtedly due to underlying disease
Modifiable by anesthetic management?
Better hope so!
Everyone needs to help
Basic scientists
Translational investigators
Clinicians

12. Basic Science Hubble image
All chemists start by trying to blow themselves up —
Hubble image

13. Andrew M. Sessler

14. Humans are a Poor Model for Rats
Nutrition
Many interventions supported by animal studies
Virtually all nutritional studies failed in humans
Vitamins E and C, olive oil, margarine, red wine, etc.
Genetics
58% of 2016 NIH budget on genetics & stem cells
29,000 papers on genetics & stem cells in 2014
Not even a single targeted therapy
Not even for sickle cell, 60 yr after identification of gene

15. Therapeutic Hypothermia
Every model in every species shows benefit
2-3°C provides profound benefit
Except in humans!
Failed in humans for:
Brain trauma
Aneurism surgery
Myocardial infarction
Out-of-hospital cardiac arrest
In-hospital cardiac arrest
Routine cardiac surgery
Status epilepticus

16. Nitrous Oxide Interferes with vitamin B12 Prevents folate metabolism
Increases plasma homocysteine
Impairs endothelial function
Prevents protein synthesis
Why were 9,000 patients needed to show that biochemical effects were not clinically important?
Myles, Lancet 2014

17. Why the Disconnect? Basic science is interesting and important
Provides little guidance to clinical investigators
Which findings are likely to translate?
Distinguish “significant” from “important”
There will never by many clinical trials
Help investigators determine what to study

18. Clinical Research Literature clogged with irreproducible results
Small studies about 90% positive
Large trials about two-thirds negative
Provides little guidance to clinicians
Statistics
Estimate random error
Compensate poorly for systematic error

19. Statistical Association
These errors seem obvious and easy to avoid. They are not!

20. Placebo Effect
Kaptchuk, PLoS ONE, 2010

21. Example of Measurement Bias
Reported parental history
Arthritis (%)
No arthritis (%)
Neither parent 27 50
One parent 58 42
Both parents 15 8
P = 0.003
Schull & Cobb, J Chronic Dis, 1969

22. Treatment infarctions
Fragile Results
Consider two identical trials of treatment for infarction
N=200 versus n=8,000
Which result do you believe? Which is biologically plausible?
What happens if you add two events to each Rx group?
Study A p=0.13
Study B p=0.02
Trial N
Treatment infarctions
Placebo infarctions RR P A
200 1 9
0.11
0.02 B
4,000
250
0.80

23. Sample Size and 95% Confidence Intervals
Intervention reduces risk from 10% to 5%

24. Replication of Studies
H0: m1-m2=0
No difference in means
P=0.05 X
Assume the true effect size is the estimate from the first study.
Ha: m1-m2 = D observed
Power=0.5

25. Replication of Studies
H0: m1-m2=0
No difference in means
P=0.0003 X
About 3.5 times as many patients needed move P from 0.05 to
Ha: m1-m2 = D observed
Power=0.95

26. Clinicians — Not Off the Hook!
Evidence-based practice
Little evidence for most of what we do
Clinical judgement is poor substitute for evidence
Well-established evidence poorly implemented
PONV prophylaxis
Practice changes made without evidence
Eliminating nitrous oxide in entire hospitals

27. Troponin screening Inpatients >45 years
Number needed to test <15
10% 30-day mortality
Positive values trigger intervention
Cardiology consult
Aspirin ± statins or ACEIs
Heart rate and hypertension control
Lifestyle
Smoking cessation
Healthful diet
Exercise

28. Looking Beyond the OR
Sun, Anesth Analg, 2015

29. Postoperative Blood Pressure
Heart attacks occur postoperatively
87% within two days
Ward hypotension is common and prolonged
Continuous ward monitoring will be routine
We should push for it
Develop protocols for interpretation and response
Fellowships in perioperative care

30. We All Need to Contribute
Basic Scientists
Identify results suitable for clinical trials
Consensus panels?
Translational investigators
Do well-powered robust trials
Credit investigators for participation
Clinicians
Extend care beyond the operating room
Become perioperative physicians
Anesthesia is at a crossroads…

31. Rabbi Hillel
If I am not for myself, who will be for me? But if I am only for myself, who am I? And if not now, when?

32. Dory Previn
i can't go on... i really can't go on; i swear i can't go on; so i guess i'll get up and go on.

33. Department of OUTCOMES RESEARCH

34. Study Conduct N=7 mice “because we’ve always used that many”
Undefined primary and secondary outcomes
Post hoc decisions about comparisons to feature
Multiple comparisons without compensation
Data selection
“Get another mouse”
Multiple “looks” as data accrue
Hoefauf 2016 ASA abstract: only 16% of animal studies in ten major anesthesia journals mentioned a sample size estimate. Only 50% blinded; 50% randomized.

35. Multiple “Looks” Informal evaluations count Number of “looks”
Alpha error 1
0.05 2
0.10 3
0.14 4
0.19 5
0.23 10
0.40
Can’t get to ten “looks” unless you study at least ten mice. But by the time you get to ten “looks,” you might as well be flipping coins.
Informal evaluations count