principles_of_screening.pdf

Principles & Biases in Screening & Preventive Services Doug Einstadter, MD, MPH Center for Health Care Research & Policy CWRU at MetroHealth Medical Center Block One - 2024  Overview – Screening – Criteria and Principles – Potential Biases in Screening Lead Time...

Principles & Biases in Screening & Preventive Services Doug Einstadter, MD, MPH Center for Health Care Research & Policy CWRU at MetroHealth Medical Center Block One - 2024  Overview – Screening – Criteria and Principles – Potential Biases in Screening Lead Time Bias Length Bias Overdiagnosis Bias  Why Preventive Services? A Public Health Perspective Because preventive services are applicable to everyone, the population- based improvement in outcomes associated with their use may be larger than even more effective therapies targeted to much smaller groups (e.g., patients with specific diseases).  Therapeutic Interventions vs. Preventive Services Deaths/Yr from Risk Reduction* Total Deaths Intervention Target Condition from Intervention Prevented Drug 10 50% 5 Therapy Prev. Service 100,000 1% 1000 *Risk Reduction = proportionate reduction in death and disability associated with intervention Classification of Preventive Services Type of Intervention: – Immunizations, Counseling, Screening Types of Preventive Services: – Primary: Asymptomatic persons who are “Healthy” – Secondary: Persons with known risk factors – Tertiary: Preventing complications in persons with known illness Classification of Preventive Services Types of Preventive Services: Who are the targets? What are the goals? Disease Illness Goals present present Primary No No Stop disease before it begins Detect disease before Secondary Yes No signs/symptoms Managing disease to prevent Tertiary Yes Yes progression SCREENING  Screening Procedures performed to detect asymptomatic disease. The same procedures performed among patients with symptoms or signs are not considered screening; e.g., colonoscopy for bleeding (= “case finding” or “diagnostic”) For most screened conditions, the Pre-test probability (i.e., prevalence) is usually on the order of 0.1 - 1%.  Types of Diseases Appropriate for Screening  Critical Point in Disease for Screening Early Usual Biologic Diagnosis Clinical Onset Possible Diagnosis Outcome # Dx +/- * 1 2 3 Too early; Critical point: No advantage False Screening over waiting reassurance potentially for signs valuable & symptoms  Screening at the Critical Point Say a disease is detectable with tests but not by usual clinical examination Why might we Not Recommend Screening for the Disease?  Why Not Screen? Causes of False Positives with Screening False Positives from screening tests are a common problem because of: Imperfect specificity of most tests results in many false positives. And Low prevalence (pre-test probability) of diseases for which screening typically is undertaken e.g., 0.1 – 1 %,  False Positives with Screening How many false positive test results are likely to occur among 100,000 people tested for a disease with a prevalence of 1% and using a test with a sensitivity & specificity of 90%?  Size of Population : 100,000 Sensitivity of Test : 90% Specificity of Test: 90% Disease Test Result Present Absent + 900 9900 TP FP - 100 89,100 FN TN 1000 99,000 100,000 Prevalence = 1% 900 of 1000 Cases detected, but 9900 are mislabeled PPV = (900  10,800) x 100 = 8.3%  False Positives with Screening: Possible Consequences Follow-up tests with cost/pain/risk: – Biopsy for suspected cancer – Prophylactic mastectomy, prostate biopsy Initiation of Rx with resultant cost/pain/risk: – Chemotherapy for “cancer;” Antibiotic for “strep” Unnecessary anxiety or mental anguish Medico-legal consequences of complications  Can you Stomach it? 11 cousins have stomachs removed to avoid cancer risk. LOS ANGELES, California (AP) June 18, 2006 -- Mike Slabaugh doesn't have a stomach. Neither do his 10 cousins. The CDH1 gene mutation was first discovered in 1998 in a large New Zealand family with a history of stomach cancer. About 22,000 Americans will be diagnosed with stomach cancer this year and half will die. But the form that runs in the Bradfield family, called hereditary diffuse gastric cancer, is extremely rare with about 100 families diagnosed worldwide. How predictive does a genetic test need to be before we recommend prophylactic gastrectomy?  True Positives with Screening: Possible Adverse Consequences True Positive results create problems if labeling, or treatment in the pre-symptomatic phase causes more harm than benefit. Examples: Multiple sclerosis, + Gene test establishing risk later in life; ? Prostate cancer. Our diagnostic toolkit has outstripped our therapeutic toolkit: – when should we test for these conditions? What Kind of Evidence is Required to Recommend Screening? Expected Benefits > Expected Harms  Criteria to Justify Screening for a Disease Frame PS, Carlson SJ: A critical review of periodic health screening using specific screening criteria. Part 2: Selected endocrine, metabolic and gastrointestinal diseases. J Fam Pract 2:123 -129, 1975  Criteria to Justify Screening for a Disease 1. The disease must have a significant effect on quantity or quality of life. 2. The frequency of the disease must be sufficient to justify the cost and risk of screening. 3. Tests must be available at reasonable cost and risk to detect the disease in the asymptomatic phase.  Criteria to Justify Screening for a Disease 4. Acceptable methods of treatment must be available. 5. The disease must have an asymptomatic phase during which treatment yields a result superior to that obtained by delaying treatment until symptoms appear. Diseases for Which Screening is Not Recommended: Why Not? Ovarian Cancer Pancreatic Cancer Multiple Sclerosis Alzheimer’s Disease Parkinson’s Disease  Independent source for Screening Evidence: U.S. Preventive Services Task Force (USPSTF) http://www.uspreventiveservicestaskforce.org/  Overarching Questions for USPSTF Recommendations 1. What is the net benefit? Net Benefit = (Benefits – Harms) 2. How certain is the evidence? Certainty of Evidence Magnitude of Net Benefit Substantial Moderate Small Zero/Negative High A B C D Moderate B B C D Low Insufficient (I statement)  USPSTF Grades Grade Suggestions for Practice A Offer or provide this service. B Offer or provide this service. Offer or provide this service for selected C patients depending on individual circumstances. D Discourage the use of this service. Read the clinical considerations section of USPSTF Recommendation Statement. If the I service is offered, patients should understand the uncertainty about the balance of benefits and harms. Should we Screen for Ovarian Cancer? What about Lung Cancer? Before public policy recommendations are crafted, the cost- effectiveness of low-dose CT screening must be rigorously analyzed. The reduction in lung-cancer mortality must be weighed against the harms from positive screening results and over diagnosis, as well as the costs. Deaths from lung CA per 100k person-years: CT: 247 X-ray: 309 RR = 0.8 (95% CI, 0.73 – 0.93)  20% reduction in CT group Absolute Risk Difference: 309 – 247 = 62 / 100,000  The Number Needed to Screen to prevent one lung cancer death: 100,000 / 62 = 1613 per 5 years  Dec. 2013; updated March 21, 2021 https://www.uspreventiveservicestaskforce.org/uspstf/recommendation/lung-cancer-screening What about Colon Cancer? USPSTF: May 18, 2021 Clinical Guidelines: 1 August 2023 Biases in Evaluating Screening Programs Biases in Evaluating New Screening Programs and New Tests Lead time bias Over-diagnosis Bias Length bias  Lead Time Bias Lead Time Bias: The erroneous inference of reduced mortality resulting merely from earlier detection of disease, and not from greater effectiveness of early treatment  Lead Time Bias Usual Biologic Clinical Onset Diagnosis Outcome Dx +/- * Survival Time Usual Biologic Early Clinical Onset Diagnosis Diagnosis Outcome Dx +/- * Lead Time Survival Time  Survival vs. Death In studies of screening, survival can be a misleading metric. We often interpret “better survival” as equivalent to “extended life” or “delayed death,” i.e., a lower death rate. The measurement of survival is affected by the mechanism of diagnosis; measurement of the death rate is not. If the mechanism of diagnosis is changing, survival can increase dramatically, even if nobody had their death delayed.  Lead Time Bias Without Screening 10-year survival = 0% Dies at age 70 y Diagnosed at age 67 y With Screening 10-year survival = 100% Dies at age 70 y Diagnosed at age 59 y  Lead Time Bias: Examples & How to Avoid It Examples of Lead Time: – PSA/Prostate Ca, Mammography/Breast Ca, MRI/MS, Gene Tests for untreatable disease “X” How to Avoid: – RCT of Screening Tests (e.g., mammography) – Examine age/sex-specific disease mortality rate If age-specific mortality has not improved, screening has only incurred cost, side effects, complications, and anxiety.  Length Bias Length Bias* The erroneous inference of reduced mortality associated with screening caused by the disproportionate detection of indolent (less aggressive) disease among screened patients. *Also called length-time bias  Screen * # Dx +/- * # Dx +/- * # Dx +/- * # Dx +/- * # Dx +/- * # Dx +/- Time  Overdiagnosis Bias Overestimation of survival duration among screen-detected cases caused by inclusion of pseudo-disease i.e., subclinical disease that would not become overt before the patient dies of other causes. Can be considered as an extreme form of length bias.  Overdiagnosis Bias Without Screening 900 1000 10 years later patients patients are dead with lung cancer 100 patients are alive 100 10 year Survival = = 10% 1000 With Screening 10 years later 4000 patients with 4000 patients “pseudo-disease” are Alive 900 1000 patients patients are dead with lung cancer 100 patients are alive 4100 10 year Survival = = 82% 5000 A 10–percentage point increase in screening is associated with a 16% mean increase in breast cancer incidence (RR, 1.16; 95% CI, 1.13-1.19, JAMAorIntern 35-49 Med. cases per 100 000 as an absolute difference [AD]). However, there is no doi:10.1001/jamainternmed.2015.3043 commensurate change in 10-year breast cancer mortality (RR, 1.01; 95% CI, 0.96-1.06, or –2 to +3 deaths Published perJuly online 100 000 as an AD). 6, 2015.  Survival vs. Death Lead-time, length, and over- diagnosis bias combine to inflate the survival rate, even if the mortality rate is unchanged. Thus, the Survival Rate is an unreliable measure to evaluate progress against cancer over time.  Length Bias: Examples & How to Avoid It Examples of Length / Overdiagnosis Bias: – Prostate Ca, Breast Ca, Lung Ca How to Avoid It: – RCT of screening process – Use Age- & Sex-specific mortality rates – Stratify Outcomes of screened and unscreened subjects by pathologic markers of aggressiveness.  What We’ve Covered Screening Criteria and Principles Biases – how to detect and avoid: – Lead Time Bias – Length Bias – Over-diagnosis Bias  Summary Screening is an important part of medicine. Not all diseases are good candidates for screening.  Summary Even the best tests may perform poorly when used to screen for uncommon conditions. Evaluation of the effectiveness of screening must consider lead-time and length bias as well as possible over- diagnosis bias.