Over-diagnosis in pathology testing and the impact on patients Sally Lord MBBS, MS(Epi) School of Medicine, The University of Notre Dame NHMRC Clinical Trials Centre, The University of Sydney
Landmark paper
Questions Ever increasing number of laboratory tests.need rigorous examination of the clinical value in diagnosis of the tests. 3
Outline Epidemiologist perspective Problem: over-diagnosis, over-treatment Patient impact Principles Progress
21 st century More tests More testing More costly benefit:harm? MBS services per 100,000 population, Australia, 2000-2012 5
Problem: diminishing returns? Over-diagnosis Diagnosis of disease in asymptomatic patients who would not otherwise experience any associated clinical events (& more sensitive tests for differential dx ) Prognosis (health outcomes) question: Are these extra cases clinically significant?
Problem: potential for harm Over-treatment Using existing treatment strategies for milder forms of disease with lower risk of disease events despite no or uncertain treatment benefit Treatment question: Do these cases benefit from treatment?
Key principles 1. Natural history of the condition should be adequately understood i) Pathology vs. physiology ii) Prognosis (health outcomes) 2. Effective treatment same for any new test!
Example: 25 (OH)D testing 4.5 million tests, 2006-2010, 2000 vs 2010 Tests x94 (x2.5 for FBC) MBS cost $1.0 M - $96M Impact on health outcomes? Small increase in bone tests Good practice?? K Bilinski et al BMJ Open 2013 Over-diagnosis? Vitamin D tests, Australia 2006-2010 MBS data 9
Evidence clinical cut-off? Prognosis study Bone fracture risk (trial) Osteoid (autopsy) PTH (lab survey) Optimal cut-point? Seasonal variation Age and sex diff [inter-lab variability] fracture risk N= 42 279, 12 trials, 25(OH)D, non-vertebral # Bischoff-Ferrari et al 2009 Arch Intern Med
Evidence treatment effective? Treatment benefit - strategy? Mortality, fracture Treatment harm?? * cancer, other+ Demands definitive trial B: H Past failures: B-carotene, Vit E Inform clinical cut-off D-Health Trial, QIMR Population: 60-79yr Vitamin D Outcomes Placebo Outcomes N= 25,000, 5-year Rx, 10-year f/up CIA R Neale, NHMRC Project Grant R
Risk of disease events More sensitive tests *same concern Expand disease definition: mild disease, low risk New test positive Existing test positive Improved outcomes if treatment benefit>harm mild Disease spectrum severe 12
More sensitive thyroid imaging Cancer incidence Classic example of over-dx nodules 1cm Mortality stable 0.5/100,000 Evidence requirements 1. Prognosis? small nodules Autopsy: subclinical ~ 30% 2. Treatment? monitor vs treat Incidence Improved Ultrasound Mortality stable Davies & Welch JAMA 2006 Missed opportunity refine disease classification & Mx
More sensitive CT imaging for detecting pulmonary embolism Incidence Mortality sl Wiener, R. S. et al. Arch Intern Med 2011 14
Risk of anticoagulation complications Any anticoagulation complication - Gastro h age - 2 thrombocytopenia - Intracranial h age Wiener, R. S. et al. Arch Intern Med 2011 OverDx, OverRx = Unfavourable Benefit:Harm
Test benefit Improve diagnosis or risk classification Alter management (further tests, Rx) to improve outcomes +/- Improve process of care eg. time to diagnosis, pt adherence, clinician workload, costs without compromising management Diagnosis Screen Prognosis Monitor Prediction Population Classification Management Outcomes Existing tests New test/cut-off reclassifies Standard treatment Alter management for this subgroup 16
Test harms All relative to existing test and treatment pathway Misclassification Less sensitive ( False Neg) - Delayed diagnosis - Delayed treatment Less specific ( False Pos) - psychological: anxiety - physical: iatrogenic (tests, Rx) - health system: time & cost Mislabelling =Unfavourable benefit:harm Over-diagnosis Over-treatment OR recognise opportunity to refine/stratify disease classification and management personalised medicine
Risk of disease events Rx benefit? Good practice ***Proactive: seek clinical data to refine disease classification, management 1. Prognosis 2. Treatment effect H > B? New cases Benefit > Harm Existing cases Rx harm Disease spectrum 18
Progress.. 19
Risk of disease events Example: Troponin testing Chest Pain: suspected Acute Coronary Syndrome (ACS) hs Troponin positive standard Troponin positive Disease spectrum 20
Collaboration IFCC Task Force on clinical applications of cardiac biomarkers = clinical chemists, clinicians Progress demands efficient clinical study design to inform clinical practice & guidelines 21
1. Cross-sectional study Retrospective cohort study Both tests: ctnt hstnt 81 neg neg 161 patients extra hstnt 44 Diagnosis at discharge 36 pos pos Extra Tests? Extra Dx? Change in Rx? Process benefits? Nil unsuspected Type I AMI 2 cases unstable angina reclassified as NSTEMI 8 cases Type 2 NSTEMI Prognosis: 90 Day f/up Jairam et al 2011 Emerg Med Aust
2. Prognosis study Test population Existing test (T1) New test (T2) N=1124 ctnt n=198 T1-, T2- T1+, T2- T1-, T2+ T1+, T2+ follow-up hstntt n=44 Outcomes: event rate, time-to-event Reichlin et al 2012 Am. J Med 23
3. Clinical trial Optimal management for hs Tn-defined myocardial injury? - Consequences for patients with non-cardiac cause (harm) - New patient group, new strategy eg. exercise trial! - Improved outcomes provide the measure of test clinical value Existing test New test myocardial injury R New strategy Old strategy Outcomes
Clinical consequences Sum-up over-dx Prognosis studies No action Lord et al BMJ 2011 over-rx Monitor Basic Rx Treatment trials Treat Disease spectrum If new test more sensitive, look beyond test accuracy Seek data to define risk Drive toward optimal riskstratified Rx strategies to avoid over-dx, over-rx Evidence critical if high stakes: harms, costs Better tests, Better Rx
Acknowledgements Collaborators test evaluation methods & clinical trials Better Tests Better Outcomes Better Treatment John Eisman University of Notre Dame & Garvan Institute Rita Horvath Uni NSW, & Chair, EFLM Test Evaluation Methods Working Group Chee Lee NHMRC Clinical Trials Centre, Sydney University John Simes NHMRC Clinical Trials Centre, Sydney University Lukas Staub MEM Research Centre, University of Bern, Switzerland Martin Than Christchurch Hospital, University of Otago, New Zealand 26