Utilizing Biostatistics in Diagnosis, Screening, and Prevention Bindu Swaroop, MD Hospitalist Program Department of Medicine University of California, Irvine Learning Objectives • Review important statistical concepts • Understand how to apply these concepts to support high value care decisions in both the inpatient and outpatient setting • Develop an individualized approach to screening, diagnosis and treatment recommendations for your patients Essential Biostatistical Concepts • Sensitivity and Specificity Disease Present • Pretest and Posttest probability Disease Absent • Positive/Negative Predictive Values Test A Positive B • Likelihood Ratios Test C Negative D • Number Needed to Treat/Harm Role of Diagnostic Tests • To reduce uncertainty regarding a specific patient’s diagnosis • Generally most appropriate for patients you feel have an intermediate pre-test probability of a disease • Test characteristics (i.e. likelihood ratios) should be considered before ordering a test to help determine whether a given test would significantly alter your pre-test probability (and thus affect management) Ex. if you estimate a patient’s pretest probability to be 10% but the LR of a diagnostic test is only 2, then a positive test result will still only yield around a 25% posttest probability Likelihood Ratios Using likelihood ratios: 1. Use the estimated pretest probability of disease as an anchor on the left side of the graph 2. Draw a straight line through the known likelihood ratio, either (+) or (-) 3. Where this line intersects the graph on the right represents the posttest probability of disease Likelihood Ratios Using likelihood ratios: 1. Use the estimated pretest probability of disease as an anchor on the left side of the graph 2. Draw a straight line through the known likelihood ratio, either (+) or (-) 3. Where this line intersects the graph on the right represents the posttest probability of disease General Rules for Interpreting Likelihood Ratios A LR greater than 1 increases the probability that the target disease is present, and a LR less than 1 decreases the probability that the target disease is present. LR Interpretation 10 Increases the pretest probability of disease by ~ 45% 5 Increases the pretest probability of disease by ~ 30% 2 Increases the pretest probability of disease by ~ 15% 1 No change in the likelihood of disease 0.5 Decreases the pretest probability of disease by ~ 15% 0.2 Decreases the pretest probability of disease by ~ 30% 0.1 Decreases the pretest probability of disease by ~ 45% Common Diseases, Tests and Likelihood Ratios Disease Test/Result Acute cholecystitis Abdominal ultrasound Acute pulmonary embolism Pulmonary CT angiography Breast cancer Mammogram SLE Antinuclear antibody Likelihood Ratio LR(+)= 23.8 LR(−)= 0.05 LR(+) = 29.1 LR(−) = 0.05 LR(+)= 8.7 LR(−)= 1.4 LR(+)= 4.5 LR(−)= 0.125 Scenario 1: Approach To Chest Pain • A 32 y/o woman presents with a 2 day history of dull, aching, left-sided chest pain. She is physically active and exercises regularly. She first noted her symptoms after exercise. Her pain has been fairly constant since onset, but does seem to worsen with certain positional changes and improves with acetaminophen. There is no exertional component. She is worried because her father had a heart attack at age 55. • Past medical history is unremarkable with no active medical issues. She is a non-smoker and a lipid profile 2 years ago showed: total cholesterol of 185, LDL of 122, and HDL of 42. Physical examination shows a BP of 122/68 mm Hg and a heart rate of 68/min. Her lungs are clear and the heart examination is completely normal. Her chest pain is reproducible to palpation across the left costosternal margin. • A resting ECG is completely normal. Scenario 1: Focus on the diagnostic process: ▫ Estimating the pretest probability of disease in your patient What is the pre test probability this patient has ischemic heart disease? ▫ Evaluate how cardiac stress testing in your patient would influence your pretest probability of disease Will you perform additional testing to evaluate for ischemic heart disease in this patient? ▫ Assess whether you believe that cardiac stress testing would be helpful in your patient Use likelihood ratios to assess how testing might influence your pre-test probability of the disease Clinical Assessment of Pretest Probability of Ischemic Coronary Artery Disease Pretest Probability Nonanginal Chest Paina Atypical Chest Painb Typical Chest Painc Age (y) Men Women Men Women Men Women 30-39 4 2 34 12 76 26 40-49 13 3 51 22 87 55 50-59 20 7 65 31 93 73 60-69 27 14 72 51 94 86 History and Physical Examination Features and Ischemic Coronary Artery Disease History LR Physical Exam LR Pain in chest or left arm LR(+) = 2.7 Third heart sound on auscultation LR(+) = 3.2 Radiation to left arm LR(+) = 2.3 Hypotension LR(+) = 3.1 Radiation to right shoulder LR(+) = 2.9 Crackles on lung examination LR(+) = 2.1 Radiation to both arms LR(+) = 7.0 Chest pain reproducible to palpation Chest pain is primary symptom LR(+) = 2.0 History of myocardial infarction LR(+) = 1.5-3.0 Nausea/vomiting LR(+) = 1.9 Diaphoresis LR(+) = 2.0 Nature of chest pain: Pleuritic LR(-) = 0.2 Sharp or stabbing LR(-) = 0.3 Positional LR(-) = 0.3 LR(-) = 0.2-0.4 Clinical Assessment of Pretest Probability of Ischemic Coronary Artery Disease Pretest Probability Nonanginal Chest Paina Atypical Chest Painb Typical Chest Painc Age (y) Men Women Men Women Men Women 30-39 30-39 4 2 34 12 76 26 40-49 13 3 51 22 87 55 50-59 20 7 65 31 93 73 60-69 27 14 72 51 94 86 History and Physical Examination Features and Ischemic Coronary Artery Disease History LR Physical Exam LR Pain in chest or left arm LR(+) = 2.7 Third heart sound on auscultation LR(+) = 3.2 Radiation to left arm LR(+) = 2.3 Hypotension LR(+) = 3.1 Radiation to right shoulder LR(+) = 2.9 Crackles on lung examination LR(+) = 2.1 Radiation to both arms LR(+) = 7.0 Chest pain reproducible to palpation Chest pain is primary symptom LR(+) = 2.0 History of myocardial infarction LR(+) = 1.5-3.0 Nausea/vomiting LR(+) = 1.9 Diaphoresis LR(+) = 2.0 Nature of chest pain: Pleuritic LR(-) = 0.2 Sharp or stabbing LR(-) = 0.3 Positional LR(-) = 0.3 LR(-) = 0.2-0.4 Clinical Assessment of Pretest Probability of Ischemic Coronary Artery Disease Pretest Probability Nonanginal Chest Paina Atypical Chest Painb Typical Chest Painc Age (y) Men Women Men Women Men Women 30-39 30-39 4 2 34 12 76 26 40-49 13 3 51 22 87 55 50-59 20 7 65 31 93 73 60-69 27 14 72 51 94 86 History and Physical Examination Features and Ischemic Coronary Artery Disease History LR Physical Exam LR Pain in chest or left arm LR(+) = 2.7 Third heart sound on auscultation LR(+) = 3.2 Radiation to left arm LR(+) = 2.3 Hypotension LR(+) = 3.1 Radiation to right shoulder LR(+) = 2.9 Crackles on lung examination LR(+) = 2.1 Radiation to both arms LR(+) = 7.0 Chest pain reproducible to palpation Chest pain is primary symptom LR(+) = 2.0 History of myocardial infarction LR(+) = 1.5-3.0 Nausea/vomiting LR(+) = 1.9 Diaphoresis LR(+) = 2.0 Nature of chest pain: Pleuritic LR(-) = 0.2 Sharp or stabbing LR(-) = 0.3 Positional LR(-) = 0.3 LR(-) = 0.2-0.4 Clinical Assessment of Pretest Probability of Ischemic Coronary Artery Disease Pretest Probability Nonanginal Chest Paina Atypical Chest Painb Typical Chest Painc Age (y) Men Women Men Women Men Women 30-39 30-39 4 2 34 12 76 26 40-49 13 3 51 22 87 55 50-59 20 7 65 31 93 73 60-69 27 14 72 51 94 86 History and Physical Examination Features and Ischemic Coronary Artery Disease History LR Physical Exam LR Pain in chest or left arm LR(+) = 2.7 Third heart sound on auscultation LR(+) = 3.2 Radiation to left arm LR(+) = 2.3 Hypotension LR(+) = 3.1 Radiation to right shoulder LR(+) = 2.9 Crackles on lung examination LR(+) = 2.1 Radiation to both arms LR(+) = 7.0 Chest pain reproducible to palpation Chest pain is primary symptom LR(+) = 2.0 History of myocardial infarction LR(+) = 1.5-3.0 Nausea/vomiting LR(+) = 1.9 Diaphoresis LR(+) = 2.0 Nature of chest pain: Pleuritic LR(-) = 0.2 Sharp or stabbing LR(-) = 0.3 Positional LR(-) = 0.3 LR(-) = 0.2-0.4 Cardiac Stress Test Characteristics Test Sensitivity Specificity ECG Stress Test 68% 77% Echocardiographic Stress Test 85% 77% Nuclear Medicine Stress Test 88% 74% Likelihood Ratios LR(+) = 2.9 LR(-) = 0.42 LR(+) = 3.7 LR(-) = 0.19 LR(+) = 3.4 LR(-) = 0.16 Scenario 2: Approach to Chest Pain • A 47 y/o man presents with a 10 day history of left-sided chest pain. He describes his symptoms as a mild ache that is ‘deep’ and does not change with position. It comes and goes, and does not appear to be related to exertion. There is no associated nausea or diaphoresis. He cannot relate the onset of his pain to any acute event or activity. He has not had to limit his work as a construction worker because of his symptoms, and says that at times he forgets about his symptoms when he is busy at work. • Past medical history is significant for mild overweight (BMI 28) and hypertension that is well controlled with a daily dose of lisinopril. He has a 15 pack-year smoking history but has not smoked since age 25. A lipid profile 2 years ago showed: total cholesterol of 225, LDL of 142, and HDL of 38. Physical examination shows a blood pressure of 128/80 mm Hg and a heart rate of 78/min. His lungs are clear and the heart examination is normal. The remainder of his examination is unremarkable. • A resting ECG shows mild nonspecific ST-T wave changes but is otherwise normal. Clinical Assessment of Pretest Probability of Ischemic Coronary Artery Disease Pretest Probability Nonanginal Chest Paina Atypical Chest Painb Typical Chest Painc Age (y) Men Women Men Women Men Women 30-39 4 2 34 12 76 26 40-49 13 3 51 22 87 55 50-59 20 7 65 31 93 73 60-69 27 14 72 51 94 86 History and Physical Examination Features and Ischemic Coronary Artery Disease History LR Physical Exam LR Pain in chest or left arm LR(+) = 2.7 Third heart sound on auscultation LR(+) = 3.2 Radiation to left arm LR(+) = 2.3 Hypotension LR(+) = 3.1 Radiation to right shoulder LR(+) = 2.9 Crackles on lung examination LR(+) = 2.1 Radiation to both arms LR(+) = 7.0 Chest pain reproducible to palpation Chest pain is primary symptom LR(+) = 2.0 History of myocardial infarction LR(+) = 1.5-3.0 Nausea/vomiting LR(+) = 1.9 Diaphoresis LR(+) = 2.0 Nature of chest pain: Pleuritic LR(-) = 0.2 Sharp or stabbing LR(-) = 0.3 Positional LR(-) = 0.3 LR(-) = 0.2-0.4 Clinical Assessment of Pretest Probability of Ischemic Coronary Artery Disease Pretest Probability Nonanginal Chest Paina Atypical Chest Painb Typical Chest Painc Age (y) Men Women Men Women Men Women 30-39 4 2 34 12 76 26 40-49 40-49 13 3 51 22 87 55 50-59 20 7 65 31 93 73 60-69 27 14 72 51 94 86 History and Physical Examination Features and Ischemic Coronary Artery Disease History LR Physical Exam LR Pain in chest or left arm LR(+) = 2.7 Third heart sound on auscultation LR(+) = 3.2 Radiation to left arm LR(+) = 2.3 Hypotension LR(+) = 3.1 Radiation to right shoulder LR(+) = 2.9 Crackles on lung examination LR(+) = 2.1 Radiation to both arms LR(+) = 7.0 Chest pain reproducible to palpation Chest pain is primary symptom LR(+) = 2.0 History of myocardial infarction LR(+) = 1.5-3.0 Nausea/vomiting LR(+) = 1.9 Diaphoresis LR(+) = 2.0 Nature of chest pain: Pleuritic LR(-) = 0.2 Sharp or stabbing LR(-) = 0.3 Positional LR(-) = 0.3 LR(-) = 0.2-0.4 Clinical Assessment of Pretest Probability of Ischemic Coronary Artery Disease Pretest Probability Nonanginal Chest Paina Atypical Chest Painb Typical Chest Painc Age (y) Men Women Men Women Men Women 30-39 4 2 34 12 76 26 40-49 40-49 13 3 51 22 87 55 50-59 20 7 65 31 93 73 60-69 27 14 72 51 94 86 History and Physical Examination Features and Ischemic Coronary Artery Disease History LR Physical Exam LR Pain in chest or left arm LR(+) = 2.7 Third heart sound on auscultation LR(+) = 3.2 Radiation to left arm LR(+) = 2.3 Hypotension LR(+) = 3.1 Radiation to right shoulder LR(+) = 2.9 Crackles on lung examination LR(+) = 2.1 Radiation to both arms LR(+) = 7.0 Chest pain reproducible to palpation Chest pain is primary symptom LR(+) = 2.0 History of myocardial infarction LR(+) = 1.5-3.0 Nausea/vomiting LR(+) = 1.9 Diaphoresis LR(+) = 2.0 Nature of chest pain: Pleuritic LR(-) = 0.2 Sharp or stabbing LR(-) = 0.3 Positional LR(-) = 0.3 LR(-) = 0.2-0.4 Focus on the diagnostic process: ▫ Estimating the pretest probability of disease in your patient -moderate (51%) ▫ Evaluate how cardiac stress testing in your patient would influence your pretest probability of disease -LR for treadmill stress test = 2.9 ▫ Assess whether you believe that cardiac stress testing would be helpful in your patient -increase post test probability by ? Scenario 3: Approach to Chest Pain • A 68 y/o man presents with a 5 day history of left-sided chest pressure. He has a history of coronary artery disease with stent placement in the RCA and an LAD diagonal artery 3 years ago. He has had no cardiac symptoms since that time. His current symptoms are similar to his prior chest pain symptoms. His chest pain is exertional and responds to sublingual nitroglycerin. • Past medical history is significant for hypertension, diabetes mellitus type 2, hyperlipidemia, and sleep apnea, which is treated with nocturnal CPAP. Medications include metformin, lisinopril, atorvastatin, isosorbide mononitrate, aspirin, clopidogrel, and as-needed sublingual nitroglycerin. He is a non-smoker and a lipid profile 6 months ago (on treatment): total cholesterol of 120, LDL of 65, and HDL of 32. • Physical examination shows a BP of 122/82 mm Hg and heart rate of 72/min. BMI is 26. The lungs are clear and his heart examination shows a regular rate with normal heart sounds and a 2/6 systolic murmur at the right upper sterna border. There is no lower extremity edema, and the remainder of his examination is unremarkable. • A resting ECG show normal sinus rhythm, with small Q waves in leads II, III and aVF. There are no acute or ischemic changes. Focus on the diagnostic process: ▫ Estimate the pretest probability of disease in your patient Pretest Probability Nonanginal Chest Paina Men Women Age (y) 30-39 40-49 50-59 60-69 4 13 20 27 Atypical Chest Painb Men Women 2 3 7 14 34 51 65 72 12 22 31 51 Typical Chest Painc Men Women 76 87 93 94 26 55 73 86 ▫ Evaluate how cardiac stress testing in your patient would influence your pretest probability of disease Test ECG Stress Test Sensitivity 68% Specificity 77% Echocardiographic Stress Test 85% 77% Nuclear Medicine Stress Test 88% 74% Likelihood Ratios LR(+) = 2.9 LR(-) = 0.42 LR(+) = 3.7 LR(-) = 0.19 LR(+) = 3.4 LR(-) = 0.16 LR Interpretation 10 Increases the pretest probability of disease by ~ 45% 5 Increases the pretest probability of disease by ~ 30% 2 Increases the pretest probability of disease by ~ 15% 1 No change in the likelihood of disease 0.5 Decreases the pretest probability of disease by ~ 15% 0.2 Decreases the pretest probability of disease by ~ 30% 0.1 Decreases the pretest probability of disease by ~ 45% ▫ Assess whether you believe that cardiac stress testing would be helpful in your patient Preventive Practice: The Periodic Health Examination • A 57-year-old woman presents for her periodic health examination. She has not been seen by a doctor for ten years • She has no past medical history except for a hysterectomy for fibroids 10 years ago. She takes no medications • She is a retired teacher, lives alone, smokes 1 pack per day (38 pack years), does not use alcohol or illicit drugs, and has not been sexually active since her husband died 8 yrs ago • She has no family history of cancer, vascular disease, or osteoporosis • Her BP 135/83 and her BMI is 24, the remainder of her exam is unremarkable What Should be Included in her Preventive Care? • Rank the top five preventive services you might offer this patient. Use the categories below to help you. ▫ Immunizations ▫ Screening ▫ Behavioral counseling to motivate lifestyle changes ▫ Chemoprevention Preventive Care Menu • • • • • • • • • • • Influenza vaccine Pneumococcal vaccine Td/TDAP Zoster vaccine Smoking cessation Weight loss Exercise Alcohol misuse Aspirin Statin therapy Hepatitis C screening • • • • • • • • • • • Calcium/Vitamin D Tamoxifen Mammogram Colonoscopy Pap smear HIV screening Lung cancer screening (LDCT) DEXA scan HbA1C Fasting lipid panel Domestic violence screening Role of Screening Tests • To detect asymptomatic and early stage disease • Should be highly sensitive and highly specific to pick up most cases of true disease and avoid false positives • Targeted toward populations with a higher disease prevalence (high positive predictive value) • Should be relatively safe and cost-effective • Should screen for diseases in which early identification and treatment have been demonstrated to improve clinical outcomes Cost-effectiveness “Cost-saving” Reduces cost, Improves Health 1 Costs money, Improves Health Costs money, Worsens Health • Measures that cost money but improve health can be further categorized by their cost, often measured in dollars per QALY (quality-adjusted lifeyear) • QALYs incorporate an estimate of the quantity of life gained by the intervention, coupled with a more subjective assessment of the quality of that life affected by the intervention • Historically, payers have considered any intervention that has a costeffectiveness ratio of <$100K per QALY as acceptable Cohen J, Neumann P, Weinstein M NEJM 2008:358;72 Examples of relative cost-effectiveness of different preventive measures. • Colonoscopy for average risk patients costs $1000-3000 but provides a 20-30% absolute mortality reduction (USPSTF grade A) • Mammography for average risk women ages 40-49 prevents only one death for every 1000 women screened annually for 10 yrs (USPSTF grade B) Common Harms Associated with Screening False positive results • Can lead to incorrect labeling, inconvenience, expense, and physical harm in follow-up tests “Overdiagnosis3” and “Pseudodisease” (Length-time bias) USPSTF Recommendation Grades, Electronic Preventive Services Selector (ePSS), and Vaccine Scheduler • USPSTF provides a tool for clinicians (ePSS) to search for graded recommendations. epss.ahrq.gov/PDA/index.jsp • Vaccine recommendations are available in the second tool. www.cdc.gov/vaccines/schedules/Schedulers/adultscheduler.html • The ACP also offers a downloadable Immunization Advisor. immunization.acponline.org/app/ USPSTF Recommendation Grades Back to the Case: • A 57-year-old woman presents for her periodic health examination. She has not been seen by a doctor for ten years • She has no past medical history except for a hysterectomy for fibroids 10 years ago. She takes no medications • She is a retired teacher, lives alone, smokes 1 pack per day (38 pack years), does not use alcohol or illicit drugs, and has not been sexually active since her husband died 8 yrs ago • She has no family history of cancer, vascular disease, or osteoporosis • Her BP 135/83 and her BMI is 24, the remainder of her exam is unremarkable Back to the Case: What 5 preventive services should have been offered? 1. Aspirin for chemoprevention of CVA – grade A 2. Colorectal cancer screening – grade A 3. Lipid screening- grade A 4. Tobacco use counseling and intervention – grade A 5. Influenza/pneumococcal vaccine Balancing Benefits and Harms 4 • Daily ASA use results in a 17% absolute risk reduction in ischemic CVA for women and 32% absolute risk reduction for MI in men • Major risk of daily ASA use is serious upper GI bleed • Both the risk of major GI bleed (net harm) and the risk of ischemic CVA (net benefit) increase with age • Using data from large randomized controlled trials, USPSTF has provided tables that estimate the absolute # of strokes and MI prevented per 1000 women and men, respectively, stratified by baseline stroke risk and age • Baseline 10-yr CVA risk can be calculated using: www.westernstroke.org/PersonalStrokeRisk1.xls = 3.5% for our patient Balancing benefits and harms 4 Balancing benefits and harms 4 ASA for the Prevention of Cardiovascular Disease: U.S. Preventive Services Task Force Statement What about for this patient? • Mr. T is a 50 y/o male who comes to your office for routine followup • He has heard that taking an aspirin may decrease his risk of a heart attack and is wondering whether he should start taking daily aspirin • Past medical history ▫ ▫ ▫ ▫ ▫ No family history of CAD No personal history of CAD or gastrointestinal issues His only medical history of HTN, well controlled with medication His recent lipid profile shows total cholesterol of 160, HDL 50 PE in office with normal cardiac and general exam, BP 110/74 • How would you counsel him? What is the data? • Use Framingham risk assessment calculator (10 year MI risk) at hp2010.nhlbihin.net/atpiii/calculator.asp = 2% Balancing benefits and harms 4 Summary • Diagnostic tests should only be used if the result is likely to significantly affect your certainty of a disease (posttest probability) • Preventive healthcare must be individualized with the help of expert recommendations (USPSTF) to offer patients interventions that are most likely to positively impact their long-term healthcare goals • Recommendations are not prescriptive, but rather the beginning of an open dialogue with patients to create a prioritized plan of preventive health maintenance References 1. Owens D, et al. High-value, cost-conscious health care: concepts for clinicians to evaluate the benefits harms and costs of medical interventions. Ann Intern Med. 2011; 154: 174-80 2. Cohen JT, et al. Does preventive care save money? Health economics and the presidential candidates. N Engl J Med. 2008; 358:661-3. 3. Moynihan R, et al. Preventing overdiagnosis: how to stop harming the healthy. BMJ. 2012; 344: e3502. 4. U.S. Preventive Services Task Force; Aspirin for the Prevention of Cardiovascular Disease: U.S. Preventive Services Task Force Recommendation Statement. Annals of Internal Medicine. 2009 Mar;150(6):396-404.