Quality Challenges and Pitfalls in the Evaluation of Patients with Suspected Disease

Joseph A. Ladapo, MD, PhD Assistant Professor of Medicine Department of Population Health NYU School of Medicine

Disclosures: K23 HL116787 Award from National Heart, Lung, and Institute; Consultant for CardioDx, Inc

Overview

• Health and economic burden of (CAD)

• Pitfalls in care of patients suspected of having CAD

• Approaches to improve quality of care

• Introduction to Corus CAD, a gene expression test for CAD

2

1 Coronary Artery Disease (CAD) is a Major Public Health Challenge

Epidemiology

• 15.4 million adults in the US live with CAD • 6.4% of adults overall; 7.9% of men and 5.1% of women

Mortality

• Accounts for over 386,000 deaths • Nearly one million new or recurrent annually (1 in 6 of all deaths) heart attacks each year

• More than half of sudden cardiac • Early diagnosis is important deaths occur in people with no prior because treatment & preventive history of heart disease practices significantly reduce morbidity and mortality

National Health Interview Survey, 2010; Go, Circulation 20133

Healthcare Utilization and Costs Attributable to CAD are Substantial

Healthcare utilization and costs

• CAD costs $195.2 billion in direct and indirect costs annually (2009) • In 2006, Medicare spent $11.7 billion on inpatient care for CAD • By 2030, medical costs for CAD (real 2010$) projected to increase ≈100%

Ambulatory care Hospital care

• In 2009, there were over 14 million • 1.3 million hospital admissions with ambulatory care visits with CAD as the CAD listed as the first diagnosis first-listed diagnosis • 954,000 percutaneous coronary interventions (PCIs), 397,000 cardiac bypass surgeries, 1.03 million diagnostic cardiac catheterizations

National Health Interview Survey, 2010; Go, Circulation 20134

2 Cardiac Testing

Source of picture: American College of , CardioSmart website

Primary Care Physicians Routinely Manage Patients With Suspected or Diagnosed CAD

~4M Stable, Symptomatic Patients Suggestive of Coronary Disease Annually Diagnostic Tools

Clinical Factors Primary Care EKG Treadmill

Stress Echo Often Repeat Cardiology Nuclear Imaging Testing CT Angiography

Invasive Obstructive CAD is not found in Angiography ~60% patients1

~$4.5B annual expenditures2

1 Patel et al, N Engl J Med 2010;362:886-95; COMPASS study. 2 IMV Market Reports 6

3 Ambulatory Care: Cardiac Stress Tests In Patients Without CAD, From 1993 To 2010

5,000,000

4,000,000

3,000,000

2,000,000 Any cardiac stress test

No. of Tests Stress imaging 1,000,000 Avg. Avg. annual annual cost in cost in - 1993-1995 2008-2010 All cardiac $820M $2.0B stress tests

Stress $550M $1.9B imaging

7 Source: Ladapo et al, Annals of Internal Medicine 2014

Opportunities For Quality Improvement

Presentation Title Goes Here 8

4 Despite Wide Array of Available Tests for Working Up CAD, Patients Routinely Encounter…

1. Diagnostic uncertainty

2. Unnecessary testing and procedures

3. Unnecessary radiation exposure

9

Challenges to Clinical Care: Diagnostic Uncertainty

Presentation Title Goes Here 10

5 Uncertainty About How to Interpret Diagnostic Test Results is Common

• Bayes’ theorem defines how pretest disease risk and diagnostic test performance can be used to guide interpretation of test results

Missed or Correctly misdiagnosed identified

Reverend Thomas Bayes

11

Stress Test Effectiveness in Diagnosing CAD: Results From Meta-analyses

Stress MPI Stress ECHO

Sens Spec # studies Sens Spec # studies

Garber et 88% 77% 10 76% 88% 6 al

Fleischma 87% 64% 27 85% 77% 24 nn et al

Mowatt et 76% 65% 10 --- al*

Mowatt et 92% 74% 4 --- al**

*includes or **excludes patients with prior myocardial infarction

Garber, Annals 1999; Fleischmann, JAMA 1998; Mowatt, HTA 2004 12

6 Physicians Interpret Test Results in a Bayesian Manner

Stress MPI example

• Most frequently performed imaging stress test • Sensitivity = 87%, Specificity = 64%

Normal test result Abnormal test result

• With a pretest probability of 20%, • With a pretest probability of 20%, post-test probability of CAD is 5% post-test probability of CAD is after a normal test result 38% after an abnormal test result

• With a pretest probability of 60%, • With a pretest probability of 60%, post-test probability of CAD is 23% post-test probability of CAD is after a normal test result 78% after an abnormal test result

13

Referral Bias also Influences Diagnostic Accuracy and Test Interpretation

Referral bias Biases clinical • Sometimes called “verification bias” or “workup bias” decision-making • Occurs because higher-risk patients are preferentially referred to • Bayesian methods needed to adjust • Because it biases test diagnostic test performance for referral performance, it may also bias clinical decision-making

• Studies of stress test performance do not adjust Most studies do for this phenomenon so not account for estimates of sensitivity and referral specificity biased

Begg, Biometrics 1983 14

7 Presentation Title Goes HereLadapo et al, JAHA 2013 15

Cardiac Catheterization Referral Rates After Normal Exercise ECHO or MPI Results

Referral rates are low after a normal study • Range of ~1% to 5% generally

Homogenous • Geographic location and patient characteristics vary but little variation in referral rates

Ladapo et al, JAHA 2013 16

8 Cardiac Catheterization Referral Rates After Abnormal Exercise ECHO or MPI Results

Referral rates higher after an abnormal test • Range of ~20% to 50% generally

Heterogeneity • Significant variation in referral rates

Ladapo et al, JAHA 2013 17

Observed vs. True Diagnostic Performance of Exercise ECHO

99% 100% 91% 81% 80%

60% 40% 40%

20%

0% Sensitivity Specificity Observed performance True performance

True performance = observed performance after adjustment for referral18

9 Observed vs. True Diagnostic Performance of Exercise MPI

97% 100% 91%

80% 67% 60% 44% 40%

20%

0% Sensitivity Specificity Observed performance True performance

True performance = observed performance after adjustment for referral19

Challenges to Clinical Care: Unnecessary Testing and Procedures

Presentation Title Goes Here 20

10 Unnecessary Testing is Common

Our Threshold for Testing Patients is Falling

• 39,515 patients undergoing stress MPI between 1991- 2009 at Cedars Sinai

• Incidence of abnormal scans fell from 41%→9%

• Incidence of ischemic scans fell from 30%→5%

• Only 3% of patients who did not have typical and could exercise had an abnormal scan

Rozanski, JACC 2013 21

Unnecessary Procedures are Common

Cardiac Catheterization Frequently Performed Needlessly

• 398,978 patients in American College of Cardiology (ACC) National Cardiovascular Data Registry (NCDR)

• Noninvasive cardiac testing performed in 84% of patients prior to catheterization

• Only 1 in 3 patients were found to have obstructive coronary disease

• No coronary artery disease was reported in 39% of patients

Patel, NEJM 2010 22

11 Inappropriate Imaging is Common

What is appropriate imaging?

• Imaging that provides (1) accurate information, (2) influences behavior, and (3) yields benefits that outweigh risks

• In general, imaging in intermediate to high-risk patients is appropriate (diabetics, AAA, PVD, angina/ischemic equiv., ECG uninterpretable)

• Number of inappropriate stress tests in patients evaluated for CAD has risen

• Primary care doctors fare worse than cardiologists

Presented at American College of Cardiology 63rd Annual Scientific Session, March 23 2014 in Washington, DC

Challenges to Clinical Care: Unnecessary Radiation

Presentation Title Goes Here 24

12 25

Unnecessary Radiation Exposure is a Growing Problem

Potential harms related to radiation exposure are poorly understood

• Stress MPI accounts for 22% of cumulative effective radiation from medical sources

• One MPI ≈ 1,000 chest x-rays ≈ 10-15 mSv

• Persons at risk for repeated radiation exposure, such as healthcare workers and the nuclear industry, typically restricted to max 100 mSv every 5 years

Columbia University Medical Center Medicare population

• 1,097 consecutive patients, 8-10 years of • Between 1993-2001, 34% of enrollees follow-up underwent repeat testing • Multiple MPIs performed in 424 patients (39%) • Median cumulative effective dose from MPI US nonelderly population was 29 mSv • Median cumulative effective dose from • 952,420 adults in 5 US markets (2005-2007) medical sources was 64 mSv • Among patients undergoing cardiac imaging, mean cumulative dose 16.4 mSv (1.5-190 mSv) • MPI accounted for 74% of cumulative dose

Fazel, NEJM 2009; Einstein, JAMA 2010; Gerber, Circulation 2009; Lucas, Circulation 2006; Chen, JACC 201026

13 Improving Quality Of Care

Presentation Title Goes Here 27

Improving Quality of Care: Reducing Unnecessary Testing

Appropriate use criteria

• Growing physician awareness of appropriate use criteria for diagnostic testing in patients suspected of having CAD

• ACC and United Healthcare registry reported that 34% of stress MPIs were inappropriate or of uncertain appropriateness

Insurer policies and regulation Patient/Professional education • Prior authorization by radiology benefits managers • Informed decision-making • Reductions in reimbursement

• Medicare released national coverage • Shared decision-making decision requesting more evidence for coronary CT angiography (CTCA) • Professional society scrutiny

…Impact on health is unknown and needs to be studied

Shaw, JACC: CV Imaging 2010; Gibbons, JACC 2008 28

14 Improving Quality of Care: Comparative Effectiveness Research

Comparative effectiveness research (CER)

• Because tests vary in diagnostic accuracy, initial test choices may influence outcomes, independent of regional/physician care patterns and patients’ risk factors

• Large data set analyses and clinical trials

Stress MPI vs. ECHO (unpublished data) Stress MPI vs. ECHO vs. CTCA

• 11,794 patients underwent stress testing and • 282,830 Medicare patients undergoing were followed for 12 months stress testing and followed for 6 months

• Mean age 53 yrs, 62% male • Mean age 74 yrs, 46% men

• Risk of major adverse cardiovascular events • Risk of acute MI lower after stress ECHO lower after stress ECHO (aOR 0.6, p<0.001) (aOR 0.8, p<0.03) and CTCA (aOR 0.6, compared to MPI p<0.04) compared to MPI

Ladapo et al, in development; Shreibati, JAMA 2011 29

Improving Quality of Care: Clinical Decision Analysis

Clinical decision analysis Patient preferences

• Integrates mathematical modeling, • Impact of false positive tests uncertainty, and patient preferences to identify optimal care strategies • Preferences over medical management vs. invasive management of CAD • Informs health policy and clinical practice • Radiation concerns

No models of patient care have integrated (1) patients’ preferences about false positives and (2) accurate measures of diagnostic accuracy (referral-adjusted)

Ladapo et al, JACC 2009 30

15 Improving Quality of Care: Optimizing Diagnostic Accuracy

Research on diagnostic accuracy of stress testing that accounts for referral patterns to cardiac catheterization is needed

• Wider dissemination of accurate information about diagnostic performance

Coronary CT angiography Blood-based gene-expression test for diagnosing • High-resolution visualization of obstructive CAD (Corus CAD) coronary anatomy

• Limitation: Ischemic heart disease occurs in absence of coronary stenosis

31

Gene expression score (GES) Measures Expression of 23 Genes From Peripheral Blood Cells

32

16 Blood-Based Genomic Test: A New Diagnostic Test to “Rule-out” CAD

Gene expression score (GES)

• First clinically validated gene expression test for assessing obstructive CAD (>50% stenosis in a major coronary artery) • Algorithm based on the peripheral blood gene expression of 23 genes, gender and age • Molecular basis includes genes and pathways with known involvement in and its progression • GES quantifies risk of CAD

• Targeted application as a first line, rule-out test, prior to additional stress testing or cardiac catheterization • Stress testing often perceived as gatekeeper to cardiac catheterization • GES may be gatekeeper to stress testing • Nondiabetic patients only

33

COMPASS Trial Overview

Purpose • Independent validation of Corus CAD in stable symptomatic patients referred for MPI (intended use population) • Also evaluated MPI performance in the real-world clinical setting

Study Design Patient Inclusion/Exclusion • Prospective, multi-center, blinded study • Symptomatic (chest pain or anginal • 19 U.S. sites, 431 patients equivalent) who have been referred for MPI • QCA Core Lab: Cardiovascular Research for the workup of suspected obstructive Foundation, NY CAD • CTA Core Lab Readers: Szilard Voros, MD, James • Non-diabetic Adams, MD • No known obstructive CAD, prior • MPI Core Lab Reader: Timothy M Bateman, MD myocardial infarction, or prior revascularization procedure • Steering committee: Greg Thomas, MD, MPH, John McPherson, MD, Alexandra Lansky, MD, Szilard Voros, MD

Thomas, Circ Genetics 2013 34

17 COMPASS Trial Design

COMPASS (Coronary Obstruction Detection by Molecular Personalized Gene Expression)

• Primary Endpoint: GES performance by ROC analysis • Steering Committee: Greg Thomas, MD, MPH, John McPherson, MD, Alexandra Lansky, MD, Szilard Voros, MD • 19 U.S. sites, 431 patients

35

1 - Specificity 36

18 GES Performance in Independent Validation Studies

GES COMPASS PREDICT Performance* (N=431)** (N=526)

Sensitivity 89% 85%

NPV 96% 83%

Specificity 52% 43%

Prevalence 15% 37%

*Performance associated with a score threshold of 15 **ROC curve area (AUC) for GES was 0.79 compared to site/core-lab AUCs for MPI were 0.59 and 0.63

37

IMPACT-Cardiology & -Primary Care Providers Trial Designs

IMPACT-CARD Patients Referred from Primary Care Cardiology N=83 GES-driven Decision: • Medical Management GES • Non-Invasive Test IMPACT-PCP • Cardiac Cath Patients Presenting with Angina Primary Care Related Symptoms N=251 Initial Decision: • Medical Management • Non-Invasive Test • Cardiac Cath

Primary Analysis: Change in diagnostic test utilization with vs without GES Secondary Analysis: Change in testing vs. historical controls

Herman L et al. JABFM 2014;27(2):258-267. McPherson JA et al. Crit Pathw Cardiol 2013;12(2):37-42. 38

19 Clinical Utility of GES in Evaluation of Patients with Suspected CAD

IMPACT-Cardiology (n=83): 63% of patients had GES ≤ 15 (52/83) and 37% of patients had scores > 15 (31/83) 58% of patients saw a change in diagnostic testing* following GES (p<0.001) IMPACT-PCP (n=251): 51% of patients had GES ≤ 15 (127/251) and 49% of patients had scores > 15 (124/251) 58% of patients saw a change in diagnostic testing* following GES (p<0.001)

Percentage of Percentage of Patients Percentage of Patients Patients with with No Change with Increased Decreased Testing in Testing Testing

GES ≤15 56% 44% 0% 23/52 Cardiology 29/52 0/52 GES >15 10% 39% 52% 3/31 12/31 16/31 GES ≤15 60% 38% 2% PCP 76/127 48/127 3/127 GES >15 14% 47% 39% 17/124 58/124 49/124

* E.g., myocardial imaging, CT Angiography (CTA) and/or invasive angiography

Herman L et al. JABFM 2014;27(2):258-267. McPherson JA et al. Crit Pathw Cardiol 2013;12(2):37-42. 39

Conclusions

• Despite the high prevalence and substantial economic burden of CAD, significant challenges to effectively diagnosing and treating patients exist

• Major quality issues include diagnostic inaccuracy, unnecessary testing and procedures, and radiation exposure

• Opportunities exist to improve care through multiple health policy, research, and patient-oriented dimensions

• A gene expression test for CAD, has promising diagnostic characteristics and may play a role in improving care and reducing unnecessary testing

40

20 Thank you! [email protected]

646-501-2561Presentation Title Goes Here 41

21