Make EBM Individualized and Smarter - vghtpefdc.weebly.com · 5 Outcomes : myocardial infarction, stroke or systemic embolism, intracranial hemorrhage, gastrointestinal bleeding,

Make EBM Individualized and Smarter

Hao-min Cheng, M.D., PhD. Department of Medical Education, Taipei Veterans General Hospital

Division of Cardiology, Taipei Veterans General HospitalFaculty of Medicine, National Yang-Ming University, Taiwan

1

Presentation Outline

Core epistemological principle of EBM

The principal of EBM application

Shared Decision Making

Make EBM Individualized

Why and how?

Maker EBM smarter

Learning on demand

Medical calculation in EMR

A better EBM inquiry system

Take Home Messages

3






Why and how?

Maker EBM smarter

Learning on demand



Take Home Messages

4

Benchmark of Clinical Research

5

Benchmark of Clinical Research

6

3 Epistemological Principles of EBM

Djulbegovic B, Guyatt GH. Progress in evidence-based medicine: a quarter century on. The Lancet; 390(10092):415-423.

Not all evidence is created equal

The pursuit of truth is best accomplished by evaluating the totality of the evidence

Clinical decision making requires consideration of patients’ values and preferences

Practice of medicine should be

based on the best available

evidence.

Health claims should be based on

systematic reviews that summarise

the best available evidence

Effective decision making has to

address the consequences of

importance to the decision maker






Why and how?

Maker EBM smarter

Learning on demand



Take Home Messages

8

實證醫學應用模式

9

Progress in evidence-based medicine: a quarter century on.The Lancet.390(10092):415-423.

10

Values &

Context

Scientific

Evidence

*HCP’s

Expertise

Patient-centered

*HCP: Health Care Providers

The SHARE ApproachEssential Steps of Shared Decision Making

Step 1: Seek your patient’s participation

Step 2: Help your patient explore and compare treatment options You need the help of SR for decision aids

Step 3: Assess your patient’s values and preferences

Step 4: Reach a decision with your patient

Step 5: Evaluate your patient’s decision

The SHARE Approach

12

Discuss the benefits and risks of each treatment option.

Know the benefits and risks of each option

Understand how they relate to your patient’s situation and condition.

Use evidence-based decision-making resources to compare the treatment options.

Step 2: Help your patient explore and compare treatment options

13

Patient-centered outcomes research can help.

Effective Health Care Program activities

Evidence synthesis

Development of systematic reviews of existing evidence to compare treatment/intervention effectiveness and identifying relevant knowledge gaps

• Compare the benefits and harms of treatment options

• Explain what is known and what is not known

14

Health Decisions

Good decisions Informed

Supported by best evidence

Compatible with patients values

Considers patients preferences

Weigh pros and cons

Practical

Poor decisions Objective data inadequate

Too few options considered

Alternatives unclear

Values and preferences unexplored

Roles unclear

Communication is poor

Cornelia Rulandhttp://www.dbmi.columbia.edu/homepages/cmr7001/sdm/html/decision_support.htm

Decision Aids

Designed to Provide information on

options Help people participate

in decision making Help clarify and

communicate personal values

NOT designed to Advise people to choose

one option over another Not meant to replace

physician consultation

Cornelia Rulandhttp://www.dbmi.columbia.edu/homepages/cmr7001/sdm/html/decision_support.htm

Prepare Patients to Make Informed, Value-based Decisions with Their Physicians






Why and how?

Maker EBM smarter

Learning on demand



Take Home Messages

17

What do you mean by patient value?

The patients’ preference can be related to Intervention options: it is subjective but important!

Multiple outcomes resulting from intervention: it can be objective

18

Let us do a opinion poll!

Which of the following outcomes you rate the most critical to you:

A. Myocardial infarction 心肌梗塞

B. Stroke or systemic embolism 缺血性中風

C. Intracranial hemorrhage 顱內出血

D. Gastrointestinal bleeding 胃出血

E. All-cause death 死亡

19

Patients and clinical trialists did not weigh individual components of a composite end point equally.

Whereas trialists are most concerned about avoiding death, patients place equal or greater importance on reducing myocardial infarction or stroke

20Circulation. 2014;130:1254-1261

Adjusted end point weights according to patient age, race, and annual income

21

Circulation. 2014;130:1254-1261

End point weights of clinical trialist

22Circulation. 2014;130:1254-1261

Stroke Prevention for Atrial Fibrillation

23

Effective aNticoaGulation with factor xAnext GEneration in Atrial Fibrillation

Edoxaban30 mg QD regimen

Warfarin (INR 2.0–3.0)

Edoxaban60 mg QD regimen

PATIENTSAF on electrical recording within last 12 months

Intended oral anticoagulant

CHADS2 ≥2

N=21,105

Median duration of follow up 2.8 years

Study design: ENGAGE AF-TIMI 48

*Dose reduced by 50% if CrCl 30–50 mL/min, body weight ≤60 kg

or patient receiving verapamil, quinidine or dronedarone

AF=atrial fibrillation; CrCl=creatinine clearance

INR=International Normalized Ratio; QD=once daily

RANDOMIZATION1:1:1 randomization is stratified by CHADS2 score 2–3 versus 4–6

and need for edoxaban dose reduction*

Randomized,double-blind,

double-dummy,event-driven study

Giugliano et al. N Engl J Med 2013; DOI: 10.1056/NEJMoa1310907

Primary Endpoint: Stroke / SEE(2.8 years median f/u)

Noninferiority Analysis (mITT, On Treatment)

0.79

0.50 1.00 2.0

Edoxaban 60* mg QD

vs warfarin

Edoxaban 30* mg QD

vs warfarin

P Values

Non-inferiority Superiority

P<0.0001

P=0.005

Hazard ratio (97.5% CI)

1.07

1.38

P=0.017

P=0.44

edoxaban noninferior

0.87P=0.08

P=0.10

Hazard ratio (97.5% CI)

1.13

0.50 1.00 2.0

P Value for Superiority

Edoxaban 60* mg QD

vs warfarin

Edoxaban 30* mg QD

vs warfarin

edoxaban better warfarin better

Superiority Analysis (ITT, Overall)

Warfarin TTR 68.4%

*Dose reduced by

50% in selected ptsGiugliano RP. NEJM 2013; 369:2093-2104

Key Secondary Outcomes

edoxaban better warfarin better

Warfarin TTR 68.4%

*Dose reduced by 50%

in selected pts

2° EP: Stroke, SEE, CV death

Death or ICH

All-cause mortality

CV death

Myocardial infarction

HR (95% CI)

Hem. Stroke

Ischemic Stroke

0.25 1.00 2.00.5

Edoxaban 60* mg QD

vs warfarin

0.94

1.411.00

0.540.33

0.87

0.87

0.92

0.86

Edoxaban 30* mg QD

vs warfarin

0.95

0.82

0.87

0.85

1.19

P vs

warfarin

E-60 E-30

<0.001

<0.001

0.97

<0.001

0.005

0.32

0.004

<0.001

0.08

0.006

0.013

0.008

0.60

0.13

Giugliano RP. NEJM 2013; 369:2093-2104

Key Safety Results- Safety Cohort on Treatment -

P Value

vs warfarin

Safety cohort=all patients who received at least 1 dose study drug

*Dose reduced by

50% in selected pts

Warfarin TTR 68.4%HR (95% CI)

Edoxaban 60* mg QD

vs warfarin

Edoxaban 30* mg QD

vs warfarin

edoxaban better warfarin better0.25 1.00.5

P<0.001

P<0.001

ISTH Major Bleeding0.80

0.47

2.0

Fatal Bleeding 0.55

0.35P=0.006

P<0.001

Intracranial Hemorrhage

0.47

0.30P<0.001

P<0.001

Gastrointestinal Bleeding1.23

0.67P=0.03P<0.001


You have so many choices and so many outcomes!!!

Net Clinical Outcomes

31

Warfarin TTR 68.4% Hazard ratio

(95% CI)

Edoxaban 60* mg QD

vs warfarin

edoxaban better Warfarin better

0.5 1.00.71*Dose reduced by 50% in selected pts

SEE=systemic embolic event

P Value

vs warfarin

Stroke, SEE, death, major bleeding0.89

P=0.003

P<0.001

Disabling stroke, life-threatening

bleeding, death

0.88P=0.008

P<0.001

Stroke, SEE, life-threatening

bleeding, death

0.88P=0.003

P=0.007

Edoxaban 30* mg QD

vs warfarin

0.83

0.83

0.89


Gap In the Shared Decision Process

33

BMJ 2016;353:i2452.

Problems faced Scenarios

ComorbiditiesRCT aimed for single condition

Real patients have multiple conditions

Power struggleWhat mean to you doesn’t necessary

mean to the patients

Viewpoints

Risks, benefits, and downsides of

management options may be viewed

differently at the level of the population

than from the perspective of an individual

SDMSDM is not clearly enabled in

contemporary practice

When applying population based evidence to individuals

Decision aids should be published in tandem with guidelines

Guideline itself should be created and evaluated in real-world

Short pressurised consultations may not be the best place for making choices: Coaching system

35

Users’ Guides to the Medical LiteratureXX. Integrating Research Evidence With the Care of the Individual Patient

36JAMA. 2000 Jun 7;283(21):2829-36.

Threshold NNTIntegrate the evidence on benefit or harm with patient values to reach treatment recommendations

Challenges:• Value data not available

• Substantial variation in values between individuals

• Decision analyses that rely on group averages for values may not always

be applicable to a particular patient: close examination of the utility sensitivity analyses of a decision analysis may provide some

guidance

Aim

Desirability of differential weighting of clinical trial end points, but a widely acceptable weighting method has not been advanced

We propose a novel method to generate a value-weighted composite end point

Data Extraction 9 RCTs that compared NOACs with VKAs in patients with nonvalvular AF 5 Outcomes : myocardial infarction, stroke or systemic embolism, intracranial hemorrhage,

gastrointestinal bleeding, and all-cause death

Preparation of End Point Weights

The disability-adjusted life-years (DALYs) The lifetime cost of medical care The clinical end points preferences survey

Meta-analysisMantel-Haenszel fixed-effects model DerSimonian and Laird random-effects model

Summarization

Incorporation of Preference Weights into Individual End Points

Standardized preference-weight for individual end points Age-standardized DALYs per patient

Lifetime medical spending per patient

Importance of individual end points for each subject

Relative weight

Constructing a preference-weight integrated composite end point

1000 resampling iterations

Mean predicted cost per case in subacute phase

Mean predicted cost per case in chronic phase+ +

Mean predicted cost per case in acute phase

Data Extraction

9 RCTs that compared NOACs with VKAs in patients with nonvalvular AF

5 Outcomes

myocardial infarction

stroke or systemic embolism

intracranial hemorrhage

gastrointestinal bleeding

all-cause death

Main characteristics of included studies

40

Study, year Trial design Sample size, n(Int/Comp)

Intervention Comparator Duration of treatment

Duration of follow-up

TTR, % Mean age, year

Male, %

Mean CHADS2

score

ARISTOTLE, 2011

Multicenter & multinational (1034 centers in 39 countries), double-blind, phase III RCT

18,201(9120 / 9081)

Apixaban 5 mg BID (or 2.5 mg BID if ≥2 criteria: age ≥80 years, body weight ≤60 kg, serum creatinine ≥1.5 mg/dl)

Warfarin (dose adjusted, target INR 2-3)

1.8 years (median)

1.8 years (median)

62.2 (mean) 70 (median)

64.7 2.1

ARISTOTLE-J, 2011

Multicenter (23 centers in Japan), double-blind, open-label, phase II RCT

222(74 for 2.5 mg, 74 for 5 mg / 74)

Apixaban 2.5 mg BID, 5 mg BID

Warfarin (dose adjusted, target INR 2-3 if age ≤70 years, or 2-2.6 if age >70 years)

12 weeks (median)

NA Overall ≥60% of patients had INR within the 2-3 range for 60% of the treatment period

70.3 82.9 1.9

CHUNG, 2011

Multinational (4 Asian countries), double-blind, open-label, phase II RCT

235(79 for 30 mg, 80 for 60 mg / 76)

Edoxaban 30 mg OD, 60 mg OD


3 months 3 months 45.1 65.1 65.4 1.9

ENGAGE AF-TIMI 48, 2013

Multicenter & multinational (1393 centers in 46 countries), double-blind, double-dummy, phase III RCT

21,105(7034 for 30 mg, 7035 for 60 mg / 7036)



907 days (medium)

1022 days (medium)

64.9 (mean) 72 (median)

61.9 2.8

RE-LY, 2009

Multicenter & multinational (951 centers in 44 countries), open-label, phase III RCT

18,113(6015 for 110 mg, 6076 for 150 mg / 6022)

Dabigatran 110 mg BID, 150 mg BID


2 years (median)

2 years (median)

64 (mean) 71.5 63.6 2.1

41





Male, %

Mean CHADS2

score

ROCKET AF, 2011

Multicenter & multinational (1178 clinical sites in 45 countries), double-blind, double-dummy, phase III RCT

14,264(7131 / 7133)

Rivaroxaban 20 mg OD (or 15 mg OD if CrCl 30-49 ml/min)


590 days (median)

707 days (median)

55 (mean) 73 (median)

60.3 3.5

J-ROCKET AF, 2012

Multicenter (167 sites in Japan), double-blind, double-dummy, phase III RCT

1,280(640 / 640)

Rivaroxaban 15 mg OD (or 10 mg OD if CrCl 30-49 ml/min)

Warfarin (dose adjusted, target INR 2-3 if age <70 years, or 1.6-2.6 if age≥70 years)

499 days for Rivaroxaban, 481 days for Warfarin (median)

1.3 years (median)

65 71.1 80.6 3.25

WEITZ, 2010 Multicenter & multinational (91 centers in 12 countries), double-blind, open-label, phase II RCT

1,146 (235 for 30 mg OD, 245 for 30 BID, 235 for 60 mg OD, 180 for 60 BID / 251)

Edoxaban 30 mg OD, 30 mg BID, 60 mg OD, 60 mg BID


12 weeks 12 weeks 49.7 65.1 62.1 % CHADS2

score of 2 = 63.3%

YAMASHITA, 2012


536(135 for 30 mg, 135 for 45 mg, 132 for 60 mg / 134)

Edoxaban 30 mg OD, 45 mg OD, 60 mg OD

Warfarin (dose adjusted, target INR 2-3 if age <70 years, or 1.6-2.6 if age ≥70 years)

12 weeks 12 weeks 73% for age <70 years, 83% for age ≥70 years

69.0 82.5 2.1

Main characteristics of included studies (Cont'd)

BID, twice daily; OD, once daily; INR, international normalized ratio; TTR, time in therapeutic range; CrCl, creatinine clearance; Int, intervention; Comp, comparator


Three metrics

The disability-adjusted life-years (DALYs) developed by WHO Global Burden of Disease Project (GBDP)

The lifetime cost of medical care reimbursed by Taiwan's National Health Insurance

The clinical endpoints preferences surveyed by questionnaire in the field

Disability-adjusted life-years (DALYs)

A summary measure of population health widely used to quantify burden of disease

1 DALY = 1 lost year of "healthy" life in given population, due to combined effects of disability and premature death

Formula

(Stroke. 2011;42:1722-1729)

New versionhttp://ghdx.healthdata.org/gbd-results-tool

DALYs

Open database

Reference

Clinical endpoints preferences survey

Demographic characteristics

Disease history

Importance of each end point to subjects themselves

Incorporation of Preference Weights into Individual End Points

STEP 1: Standardized preference-weight for individualend points

Age-standardized DALYs per patient

Lifetime medical spending per patient

Clinical endpoints preferences for each subject

Mean predicted cost per case in

acute phase

Mean predicted cost per case in subacute phase

Mean predicted cost per case in chronic phase

+ +

Data Extraction 9 RCTs that compared NOACs with VKAs in patients with nonvalvular AF 5 Outcomes : myocardial infarction, stroke or systemic embolism, intracranial hemorrhage,

gastrointestinal bleeding, and all-cause death


The disability-adjusted life-years (DALYs) The clinical end points preferences survey

Meta-analysisMantel-Haenszel fixed-effects model DerSimonian and Laird random-effects model

Summarization

Incorporation of Value Weights into Individual End Points

Standardized value-weight for individual end points


Importance of individual end points for each subject

Relative weight

Constructing a value-weight integrated composite end point

1000 resampling iterations

Age−standardized DALY rate (per 100,000)

Age−standardized prevalence (per 100,000)

Weight in each outcome

Mean of weights of five outcomes

Relative weight𝑖× Number of events𝑖

Data Extraction

9 RCTs that compared NOACs with VKAs in patients with nonvalvular AF

5 Outcomes

myocardial infarction

stroke or systemic embolism

intracranial hemorrhage

gastrointestinal bleeding

all-cause death

Main characteristics of included studies





Male, %

Mean CHADS2

score

ARISTOTLE, 2011

Multicenter & multinational (1034 centers in 39 countries), double-blind, phase III RCT

18,201(9120 / 9081)

Apixaban 5 mg BID (or 2.5 mg BID if ≥2 criteria: age ≥80 years, body weight ≤60 kg, serum creatinine ≥1.5 mg/dl)


1.8 years (median)

1.8 years (median)

62.2 (mean) 70 (median) 64.7 2.1

ARISTOTLE-J, 2011


222(74 for 2.5 mg, 74 for 5 mg / 74)

Apixaban 2.5 mg BID, 5 mg BID

Warfarin (dose adjusted, target INR 2-3 if age ≤70 years, or 2-2.6 if age >70 years)

12 weeks (median)

NA Overall ≥60% of patients had INR within the 2-3 range for 60% of the treatment period

70.3 82.9 1.9

CHUNG, 2011 Multinational (4 Asian countries), double-blind, open-label, phase II RCT

235(79 for 30 mg, 80 for 60 mg / 76)



3 months 3 months 45.1 65.1 65.4 1.9

ENGAGE AF-TIMI 48, 2013

Multicenter & multinational (1393 centers in 46 countries), double-blind, double-dummy, phase III RCT

21,105(7034 for 30 mg, 7035 for 60 mg / 7036)



907 days (medium)

1022 days (medium)

64.9 (mean) 72 (median) 61.9 2.8

RE-LY, 2009 Multicenter & multinational (951 centers in 44 countries), open-label, phase III RCT

18,113(6015 for 110 mg, 6076 for 150 mg / 6022)

Dabigatran 110 mg BID, 150 mg BID


2 years (median)

2 years (median)

64 (mean) 71.5 63.6 2.1

Main characteristics of included studies (Cont'd)





Male, %

Mean CHADS2

score

ROCKET AF, 2011

Multicenter & multinational (1178 clinical sites in 45 countries), double-blind, double-dummy, phase III RCT

14,264(7131 / 7133)

Rivaroxaban 20 mg OD (or 15 mg OD if CrCl30-49 ml/min)


590 days (median)

707 days (median)

55 (mean) 73 (median) 60.3 3.5

J-ROCKET AF, 2012

Multicenter (167 sites in Japan), double-blind, double-dummy, phase III RCT

1,280(640 / 640)

Rivaroxaban 15 mg OD (or 10 mg OD if CrCl30-49 ml/min)


499 days for Rivaroxaban, 481 days for Warfarin (median)

1.3 years (median)

65 71.1 80.6 3.25

WEITZ, 2010 Multicenter & multinational (91 centers in 12 countries), double-blind, open-label, phase II RCT

1,146 (235 for 30 mg OD, 245 for 30 BID, 235 for 60 mg OD, 180 for 60 BID / 251)

Edoxaban 30 mg OD, 30 mg BID, 60 mg OD, 60 mg BID


12 weeks 12 weeks 49.7 65.1 62.1 % CHADS2 score of 2 = 63.3%

YAMASHITA, 2012


536(135 for 30 mg, 135 for 45 mg, 132 for 60 mg / 134)

Edoxaban 30 mg OD, 45 mg OD, 60 mg OD


12 weeks 12 weeks 73% for age <70 years, 83% for age ≥70 years

69.0 82.5 2.1

BID, twice daily; OD, once daily; INR, international normalized ratio; TTR, time in therapeutic range; CrCl, creatinine clearance; Int, intervention; Comp, comparator


Two metrics

The disability-adjusted life-years (DALYs) developed by WHO Global Burden of Disease Project (GBDP)

The clinical endpoints preferences surveyed by questionnaire in the field

Disability-adjusted life-years (DALYs)

A summary measure of population health widely used to quantify burden of disease

1 DALY = 1 lost year of "healthy" life in given population, due to combined effects of disability and premature death

Formula

New versionhttp://ghdx.healthdata.org/gbd-results-tool

DALYs

Open database

Reference

Global age-standardized rates (per 100,000) of DALYs, YLDs,

YLLs, prevalence, and mortality for five causes in 2013

DALYs

(95% UI)

YLDs

(95% UI)

YLLs

(95% UI)

Prevalence

(95% UI)

Mortality

(95% UI)

Ischemic heart

disease

2371.7

(2138.6-2560.6)

95.1

(66.4-129.4)

2276.6

(2051.9-2454.9)

1518.7

(1472.5-1566.5)

137.8

(123.9-148.2)

Ischemic

stroke

789.9

(676.7-867.2)

43.0

(30.5-56.6)

746.8

(633.5-825.9)

299.1

(290.2-309.2)

57.3

(49.3-62.9)

Hemorrhagic

stroke

1013.9

(921.7-1161.2)

17.2

(12.0-22.6)

996.7

(905.7-1141.3)

116.6

(113.1-120.5)

52.8

(48.0-62.3)

Peptic ulcer

disease

128.3

(106.2-148.3)

20.4

(14.1-28.5)

107.9

(88.6-126.3)

810.7

(786.1-838.9)

4.9

(4.1-5.7)

Cardiovascular

diseases- -

4863.4

(4591.1-5120.8)-

293.2

(280.4-306.1)

UI, uncertainty interval; DALYs, disability-adjusted life years; YLDs, years lived with disability; YLLs, years of life lost

Clinical endpoints preferences survey

• Demographic characteristics• Disease history• Importance of each endpoint to subjects themselves

63

Distribution of endpoints preferences (N=172)

Myocardial

infarction

Ischemic

stroke

Hemorrhagic

stroke

Gastrointestinal

bleeding

All-cause

death

Importance Score N (%) N (%) N (%) N (%) N (%)

Minimal 1 2 (1.16) 2 (1.16) 2 (1.16) 5 (2.91) 15 (8.72)

2 4 (2.33) 2 (1.16) 1 (0.58) 3 (1.74) 10 (5.81)

3 2 (1.16) 2 (1.16) 0 (0.00) 13 (7.56) 9 (5.23)

4 0 (0.00) 1 (0.58) 2 (1.16) 12 (6.98) 8 (4.65)

5 10 (5.81) 8 (4.65) 6 (3.49) 34 (19.77) 12 (6.98)

6 6 (3.49) 4 (2.33) 4 (2.33) 28 (16.28) 6 (3.49)

7 4 (2.33) 11 (6.40) 9 (5.23) 21 (12.21) 0 (0.00)

8 25 (14.53) 32 (18.60) 28 (16.28) 24 (13.95) 7 (4.07)

9 22 (12.79) 35 (20.35) 31 (18.02) 4 (2.33) 4 (2.33)

Maximal 10 97 (56.40) 75 (43.60) 89 (51.74) 28 (16.28) 101 (58.72)

E(score) 8.7093 8.5930 8.8488 6.3430 7.5116

Incorporation of Value Weights into Individual End Points

STEP 1: Standardized value-weight for individual end points


Clinical endpoints preferences for each subject

Age−standardized DALY rate (per 100,000)

Age−standardized prevalence (per 100,000)

STEP 2: Relative weight for individual end points

STEP 3: Constructing a value-weight integrated composite end point



Relative weight𝑖 × Number of events𝑖

Comparison of end point weights from two metrics, DALYs and

clinical endpoints preferences

Weights Relative Weights

MI SSE ICH GIBAll-cause

deathMI SSE ICH GIB

All-cause

death

DALYs,

year (per

event)

1.56 (1.42-1.70)

2.64 (2.31-3.00)

8.66 (7.53-9.69)

0.16 (0.13-0.19)

16.59 (15.47-17.80)

0.264 (0.236-0.293)

0.446 (0.387-0.501)

1.462 (1.314-1.603)

0.027 (0.022-0.032)

2.802(2.667-2.940)

Endpoints

preference

s, points

(per

subject)

8.70 (8.38-8.99)

8.59 (8.29-8.84)

8.84(8.55-9.08)

6.35(5.99-6.71)

7.50(6.99-8.00)

1.088 (1.051-1.123)

1.074 (1.040-1.107)

1.106 (1.071-1.137)

0.794 (0.753-0.834)

0.938 (0.882-0.993)

Values are expressed as mean (95% uncertainty interval).DALYs, disability-adjusted life years; MI, Myocardial infarction; SSE, Stroke or systemic embolism; ICH, Intracranial hemorrhage; GIB, Gastrointestinal bleeding

1000 resampling iterations based on the appropriate distribution of weighting parameter

STEP 2: Relative weight for individual end points

STEP 3: Constructing a value-weight integrated composite end point



Relative weight𝑖 × Number of events𝑖

Meta-analysis

Two types of models were used to statistically combine results

Mantel-Haenszel fixed-effects model

DerSimonian and Laird random-effects model

Summarization

1000 pooled estimates calculated by meta-analysis

DALYs

NOACs VKAs DALYs-weighted (1000 bootstraps)Events Total Events Total RR (95% CI) RR, range 95% LL, range 95% UL, range

Apixaban

ARISTOTLE 1062 9120 1277 9081 0.83 (0.77, 0.89)

ARISTOTLE-J 1 148 4 75 0.13 (0.01, 1.11)

Subtotal 1063 9268 1281 9156

Fixed-effects 0.83 (0.76, 0.89) (0.844, 0.860) (0.797, 0.816) (0.892, 0.906)

Random-effects 0.45 (0.08, 2.52) (0.845, 0.861) (0.799, 0.817) (0.894, 0.907)

Dabigatran

RE-LY 1782 12091 979 6022 0.91 (0.84, 0.97)

Subtotal 1782 12091 979 6022

Fixed-effects 0.91 (0.84, 0.97) (0.848, 0.867) (0.803, 0.824) (0.896, 0.913)

Edoxaban

CHUNG 1 159 2 75 0.24 (0.02, 2.56)

ENGAGE AF 2954 14069 1639 7036 0.90 (0.85, 0.95)

WEITZ 23 469 11 250 1.11 (0.55, 2.25)

YAMASHITA 7 260 4 125 0.84 (0.25, 2.82)

Subtotal 2985 14957 1656 7486

Fixed-effects 0.90 (0.86, 0.95) (0.866, 0.881) (0.834, 0.850) (0.900, 0.913)

Random-effects 0.90 (0.86, 0.95) (0.866, 0.880) (0.833, 0.849) (0.900, 0.913)

Rivaroxaban

J-ROCKET AF 44 640 57 640 0.77 (0.53, 1.13)

ROCKET AF 1260 7111 1318 7125 0.96 (0.89, 1.03)

Subtotal 1304 7751 1375 7765

Fixed-effects 0.95 (0.89, 1.02) (0.901, 0.909) (0.853, 0.864) (0.952, 0.957)

Random-effects 0.93 (0.82, 1.07) (0.901, 0.909) (0.853, 0.864) (0.952, 0.956)

Total 7134 44067 5291 30429

Fixed-effects 0.90 (0.87, 0.93) (0.865, 0.879) (0.844, 0.859) (0.887, 0.899)

Random-effects 0.90 (0.85, 0.94) (0.866, 0.879) (0.844, 0.860) (0.887, 0.900)

72

Cumulative ranking curve for DALYs-weighted results

0

50

100

Best 2nd 3rd Worst0

50

100

Best 2nd 3rd Worst

0

50

100

Best 2nd 3rd Worst0

50

100

Best 2nd 3rd Worst

Apixaban Dabigatran

Edoxaban Rivaroxaban

Cum

ula

tive p

robabili

ty (

%)

SUCRA = 1.0 SUCRA = 0.7


SUCRA, surface under the cumulative ranking

NOACs VKAs Preferences-weighted (1000 bootstraps)

Events Total Events Total RR (95% CI) RR, range 95% LL, range 95% UL, range

Apixaban

ARISTOTLE 1062 9120 1277 9081 0.83 (0.77, 0.89)

ARISTOTLE-J 1 148 4 75 0.13 (0.01, 1.11)

Subtotal 1063 9268 1281 9156

Fixed-effects 0.83 (0.76, 0.89) (0.811, 0.822) (0.750, 0.762) (0.878, 0.886)

Random-effects 0.45 (0.08, 2.52) (0.370, 0.407) (0.044, 0.060) (2.768, 3.093)

Dabigatran

RE-LY 1782 12091 979 6022 0.91 (0.84, 0.97)

Subtotal 1782 12091 979 6022

Fixed-effects 0.91 (0.84, 0.97) (0.885, 0.897) (0.823, 0.834) (0.952, 0.965)

Edoxaban

CHUNG 1 159 2 75 0.24 (0.02, 2.56)

ENGAGE AF 2954 14069 1639 7036 0.90 (0.85, 0.95)

WEITZ 23 469 11 250 1.11 (0.55, 2.25)

YAMASHITA 7 260 4 125 0.84 (0.25, 2.82)

Subtotal 2985 14957 1656 7486

Fixed-effects 0.90 (0.86, 0.95) (0.897, 0.902) (0.849, 0.855) (0.946, 0.952)

Random-effects 0.90 (0.86, 0.95) (0.897, 0.902) (0.849, 0.855) (0.946, 0.952)

Rivaroxaban

J-ROCKET AF 44 640 57 640 0.77 (0.53, 1.13)

ROCKET AF 1260 7111 1318 7125 0.96 (0.89, 1.03)

Subtotal 1304 7751 1375 7765

Fixed-effects 0.95 (0.89, 1.02) (0.932, 0.941) (0.869, 0.878) (1.000, 1.010)

Random-effects 0.93 (0.82, 1.07) (0.933, 0.941) (0.869, 0.877) (1.001, 1.017)

Total 7134 44067 5291 30429

Fixed-effects 0.90 (0.87, 0.93) (0.885, 0.892) (0.856, 0.863) (0.915, 0.922)

Random-effects 0.90 (0.85, 0.94) (0.883, 0.890) (0.837, 0.846) (0.930, 0.938)

74

Cumulative ranking curve for preferences-weighted results

0

50

100

Best 2nd 3rd Worst0

50

100

Best 2nd 3rd Worst

0

50

100

Best 2nd 3rd Worst0

50

100

Best 2nd 3rd Worst

Apixaban Dabigatran

Edoxaban Rivaroxaban

Cum

ula

tive p

robabili

ty (

%)



SUCRA, surface under the cumulative ranking

In Clinical Practice

S c enar io

MISS

E

IC

H

GI

B

All-

caus

e

death

Relative Weights Apixaban Dabigatran Edoxaban Rivaroxaban

X Y Z P Q X Y Z P Q RR (95% CI) RR (95% CI) RR (95% CI)* RR (95% CI)

1 10 1 1 1 1 3.57 0.36 0.36 0.36 0.36 0.847 (0.768, 0.936)1.065 (0.971, 1.168)1.193 (0.634, 2.242)0.941 (0.864, 1.025)

2 1 10 1 1 1 0.36 3.57 0.36 0.36 0.36 0.799 (0.741, 0.861)0.825 (0.768, 0.887)0.967 (0.919, 1.019)0.901 (0.844, 0.961)

3 1 1 10 1 1 0.36 0.36 3.57 0.36 0.36 0.637 (0.575, 0.706)0.659 (0.597, 0.727)0.786 (0.515, 1.198)0.832 (0.754, 0.918)

4 1 1 1 10 1 0.36 0.36 0.36 3.57 0.36 0.852 (0.775, 0.937)1.106 (1.020, 1.198)0.919 (0.777, 1.085)1.170 (1.084, 1.264)

5 1 1 1 1 10 0.36 0.36 0.36 0.36 3.57 0.885 (0.843, 0.928)0.905 (0.863, 0.948)0.900 (0.872, 0.929)0.931 (0.889, 0.975)

63 3 3 3 3

1 1 1 1 1 0.825 (0.765, 0.890)0.907 (0.844, 0.974)0.902 (0.855, 0.951)0.950 (0.887, 1.018)7 7 7 7 7

*Estimates were calculated with a random-effects model because more than three studies were included.

Future Plans

Establish interactive decision support system incorporating our proposed preference-weighted SR method

76

Summary

Diseases and their treatment can influence many organs in diverse ways

Composite end points (Net clinical outcomes) are increasingly used in RCTs

To capture the number of patients who have 1 or more of several events of interest

Can index the overall impact of therapeutic interventions and reduce sample size requirements

Have well-recognized limitations

Common practice of weighting all end point components equally, irrespective of their relative impact on the life of the patient






Why and how?

Maker EBM smarter

Learning on demand



Take Home Messages

78

院內建置UpToDate實證連結

79

有效節省臨床工作時間！

Learning on Demand 功能擴充

80






Why and how?

Maker EBM smarter

Learning on demand



Take Home Messages

81

82

醫學計算機展示-首頁

83

選擇計算機

84

自動抓取病歷資料1

85


86


87

參數有誤提醒

88

計算結果與儲存

89

儲存紀錄瀏覽1

90

儲存紀錄瀏覽2






Why and how?

Maker EBM smarter

Learning on demand



Take Home Messages

91

實證醫學與教學研究發展的嶄新思維

臺北榮總實證醫學中心的核心理念與前瞻遠見，不只是展現出實證醫學5A的流程需

求 , 更洞見了實證醫學在臨床和教研上的相互關聯，這將開創出國內首創的實證醫

學與互動交流的創新改革。

臨床實務

實證醫學中心

同儕互動

前瞻願景

規劃理念

實證醫學模組

Future Leaders in EBM

知識互動模組

Clinicians & Healthcare Professionals

Resources, Tools, Case Studies…

Discussion, Interaction,Analysis….

02

使用者(會員)

實證醫學模組。• 5A流程系統的具體實踐。

Ask: 臨床情境、PICO查詢。

Acquire: 查詢策略、結果去重、書目收藏、 SFX全文鏈結。

Appraise:評讀及發表工具下載、結果PDF匯入、評讀表單。

Apply: 決策輔助工具下載、PDAs表單。

Audit: 評估表單填寫及儲存。

Case Studies: 成果紀錄單預覽、下載、E-mail及公開。






Why and how?

Maker EBM smarter

Learning on demand



Take Home Messages

102

Take Home Message

EBM is a useful and powerful clinical tool for clinicians

Shared decision making is the core of patient-centered care

In the era of AI, technology and environment embedded with augmented intelligence is important for a better clinical care

Facing the many clinical outcomes, including benefit and risk endpoints, it is hard to incorporate patients’ value in the decision making process.

We propose a novel method to generate a preference-weighted composite endpoint which may better the informed decision process

103

Q & A

謝謝聆聽敬請指教

Documents

Make EBM Individualized and Smarter - vghtpefdc.weebly.com · 5 Outcomes : myocardial infarction, stroke or systemic embolism, intracranial hemorrhage, gastrointestinal bleeding,