HOW TO BIAS A KAPLAN-MEIER CURVE

Slides:



Advertisements
Similar presentations
Impact of cytokeratin-positive cells in bone marrow on survival in patients with non-metastatic prostate cancer treated by radiotherapy PURPOSE To evaluate.
Advertisements

Agency for Healthcare Research and Quality (AHRQ)
Surviving Survival Analysis
Lecture 3 Survival analysis. Problem Do patients survive longer after treatment A than after treatment B? Possible solutions: –ANOVA on mean survival.
Measures of disease frequency (I). MEASURES OF DISEASE FREQUENCY Absolute measures of disease frequency: –Incidence –Prevalence –Odds Measures of association:
EVIDENCE BASED MEDICINE
1Statistics 741, Chappell - Spring 2012 Rick Chappell, Ph.D. Professor, Department of Biostatistics and Medical Informatics Department of Statistics University.
Controversies in the management of PSA-only recurrent disease Stephen J. Freedland, MD Associate Professor of Urology and Pathology Durham VA Medical Center.
Survival analysis Brian Healy, PhD. Previous classes Regression Regression –Linear regression –Multiple regression –Logistic regression.
Marshall University School of Medicine Department of Biochemistry and Microbiology BMS 617 Lecture 10: Survival Curves Marshall University Genomics Core.
IN THE NAME OF GOD By: Dr malek. References Am J Clin Pathol. 2008;130(5): © 2008 American Society for Clinical Pathology Bryant J, Picot J,
Eleni Galani Medical Oncologist
Lecture 3 Survival analysis.
A/Prof Brian Cox Cancer Epidemiologist Dunedin. Research Associate Professor Brian Cox Hugh Adam Cancer Epidemiology Unit Department of Preventive and.
Estimation of overweight-attributable deaths Katherine M. Flegal, Ph.D. Centers for Disease Control and Prevention National Center for Health Statistics.
Design of Clinical Trials for Select Patients With a Rising PSA following Primary Therapy Anthony V. D’Amico, MD, PhD Professor of Radiation Oncology Harvard.
D:/rg/folien/ms/ms-USA ppt F 1 Assessment of prediction error of risk prediction models Thomas Gerds and Martin Schumacher Institute of Medical.
Prospective Phase I/II Trial of Carbon Ion Radiotherapy for Locally Advanced Non-small-cell Lung Cancer (NSCLC) Abstract title: CIRT for Locally Advanced.
Prostate Cancer Screening in African American Men Mark H. Kawachi, MD FACS Director, Prostate Cancer Center City of Hope, National Medical Ctr.
Updated 5-year Biochemical Relapse-Free Survival after Prostate Brachytherapy Jenny P. Nobes St. Luke’s Cancer Centre, The Royal Surrey County Hospital,
Ten Year Outcomes In Men Under 60 Treated With Iodine-125 Permanent Brachytherapy As Monotherapy GU - Prostate Cancer: Novel Imaging (MRI,PET) & Brachytherapy.
1 Lecture 6: Descriptive follow-up studies Natural history of disease and prognosis Survival analysis: Kaplan-Meier survival curves Cox proportional hazards.
1 Study Design Issues and Considerations in HUS Trials Yan Wang, Ph.D. Statistical Reviewer Division of Biometrics IV OB/OTS/CDER/FDA April 12, 2007.
High Dose Rate Brachytherapy Boost for Prostate Cancer: Comparison of Two Different Fractionation Schemes Tania Kaprealian 1, Vivian Weinberg 3, Joycelyn.
Hazlina Hamdan 31 March Modelling survival prediction in medical data By Hazlina Hamdan Dr. Jon Garibaldi.
Date of download: 5/31/2016 Copyright © 2016 American Medical Association. All rights reserved. From: Outcome of Salvage Total Laryngectomy Following Organ.
Debra Freeman, MD – Naples Christopher King, MD, PhD - Stanford.
ABSTRACT Purpose This retrospective review was conducted to determine if delay in the start of radiotherapy after conservative breast surgery had any detrimental.
Risk Stratification in Stage II Colon Cancer Patients Ramzi Amri, MD, PhD; Liliana G Bordeianou, MD, MPH; and David L Berger, MD Massachusetts General.
Methods and Statistical analysis. A brief presentation. Markos Kashiouris, M.D.
Date of download: 7/9/2016 Copyright © 2016 American Medical Association. All rights reserved. From: Development of Antidrug Antibodies Against Adalimumab.
Date of download: 9/20/2016 Copyright © 2016 American Medical Association. All rights reserved. From: Effects of a Palliative Care Intervention on Clinical.
Inference for a Single Population Proportion (p)
From: Anesthetic Technique for Radical Prostatectomy Surgery Affects Cancer Recurrence:A Retrospective Analysis Anesthes. 2008;109(2): doi: /ALN.0b013e31817f5b73.
Neoadjuvant twice daily chemoradiotherapy for esophageal cancer: Treatment-related mortality and long-term outcomes  Stuart E. Samuels, MD, PhD, Matthew.
Volume 62, Issue 3, Pages (September 2012)
Survival Analysis Rick Chappell, Ph.D. Professor,
The impact of ipsilateral breast tumor recurrence on cause specific survival in DCIS patients treated with breast conserving therapy or mastectomy  C.
Quality of Life Assessment
Copyright © 2012 American Medical Association. All rights reserved.
Compassionate People World Class Care
Treatment With Continuous, Hyperfractionated, Accelerated Radiotherapy (CHART) For Non-Small Cell Lung Cancer (NSCLC): The Weston Park Hospital Experience.
New perioperative risk factors for biochemical recurrence after robotic assisted radical prostatectomy: A single surgeon experience in high volume Canadian.
Claude C. Schulman, Jacques Irani, Juan Morote, Jack A
Kaplan-Meier survival curves comparing survival between both time periods according to management strategy. Survival in patients with infective endocarditis.
External Beam Radiotherapy as Curative Treatment of Prostate Cancer
Outcome and predictors of treatment failure in early post-surgical prosthetic joint infections due to Staphylococcus aureus treated with debridement 
Kaplan-Meier curves comparing: (A) overall survival for patients treated on trial compared to those outside of a trial; (B) progression-free survival for.
LRC-CPPT and MRFIT Content Points:
• Kaplan-Meier analyses of (A) overall survival (OS) of the whole cohort (n=173), (B) OS from time of diagnosis of high-risk melanoma among those who remained.
Kaplan-Meier curves of OS by dose of nivolumab for (A) all patients (n=47), (B) PD-L1-positive patients (n=13) and (C) PD-L1-negative or unknown patients.
Where are we?.
Kaplan-Meier-estimated PFS and OS are presented, with PFS in c-Met high and low patients shown in (A), OS in c-Met high and low patients in (B), PFS in.
JAMA Internal Medicine 2018;178(11):
Identifying and validating surrogate endpoints for overall survival (OS) in metastatic castration-resistant prostate cancer (CRPC) Xiaowei Guan, De Phung,
Kaplan-Meier curves for overall survival in patients with adenocarcinoma and time since first-line therapy of
Kaplan-Meier survival curve for all catheterisations.
Kaplan-Meier plot of overall survival from the time of first dose of intrathecal interleukin-2 (IT IL-2) for all patients (A, n=43) and based on the extent.
Kaplan-Meier plot presenting PFS for patients with BRAFV600-mutated ctDNA at first visit (
Efficacy of nivolumab in Japanese patients with advanced non-squamous non-small cell lung cancer (A) Kaplan-Meier curve for PFS, (B) Kaplan-Meier curve.
Kaplan-Meier curves for PFS (panel A) and OS (panel B) of patients with mTCC receiving an anti-EGFR based therapy. mTCC, metastatic transverse colon cancer;
• Kaplan-Meier analyses of (A) time-to-recurrence (TTR) for the subgroup of patients with high-risk melanoma who had relapsed (n=82), (B) recurrence-free.
Progression-free (a) and overall (b) survival by age subgroup, Kaplan-Meier plots. Progression-free (a) and overall (b) survival by age subgroup, Kaplan-Meier.
The ROC curves analyzing the sensitivity and specificity of rCBVmax values in astrocytomas for 1-year survival (A) and recurrence (B) show the optimal.
A and B, The ROC curves analyzing the sensitivity and specificity of the WHO grading system (I-IV) for the 1-year survival (A) and recurrence (B) show.
Scatter plot of serum OPN levels and its correlation with patient survival. Scatter plot of serum OPN levels and its correlation with patient survival.
Kaplan-Meier estimates for survival in metastatic disease for the whole patient cohort (A) and in patients with or without history of adjuvant trastuzumab.
Kaplan-Meier event-free survival from ipsilateral stroke for patients with near occlusion, pooled data from the NASCET and ECST. The 3-year intention-to-treat.
Kaplan-Meier (K-M) curves of progression-free survival (PFS) of the entire cohort of patients with metastatic gastric cancer treated with RAD001. Kaplan-Meier.
Kaplan-Meier (K-M) curves of progression-free survival (PFS) in 54 patients with metastatic gastric cancer treated with RAD001. Kaplan-Meier (K-M) curves.
Presentation transcript:

HOW TO BIAS A KAPLAN-MEIER CURVE University of Wisconsin – Madison Department of Biostatistics and Medical Informatics HOW TO BIAS A KAPLAN-MEIER CURVE Rick Chappell, Ph.D. BMI 542 Week6 Spring 2017

Outline I. Obviously Dependent Censoring “Retiring to Arizona” or “Going back to the farm”  Dependent Administrative Censoring Induced by time trends The ASTRO Definition of Biochemical Recurrence in Prostate Cancer  

I. Obviously Dependent Censoring Consider a simple situation without censoring: x x x x Time

The K-M curve is the empirical CDF: 1 x x x x Time

If the healthiest in terms of remaining life are selectively censored (negatively dependent censoring) then the K-M curve is biased downward: 1 x x O x O x Time

If the sickest are selectively censored (positively dependent censoring) then the K-M curve is biased upward: 1 o x x x o x Time

II. Dependent Administrative Censoring Even when the sole source of censoring is administrative (event hasn’t yet occurred at the time of analysis), it can be dependent with failure time. Pointed out by Kaplan & Meier (1958), credited to Sartwell and Merrell (1952), Am. J. Pub. Health 42, “Influence of the dynamic character of chronic disease on the interpretation of morbidity rates”.

“For example, in a study of survival after an operation, a change in surgical technique five years before the data are analyzed will affect the survival times only of those with observation limit less than five years [p. 470].” Consider an extreme example: 1990 200 accrued 50% failure in 1991 1993 2000 accrued 50% failure in 1996 The rest are cured. An analysis is performed in 1995.

+ = K-M estimate for 1990 cohort of 200 analyzed in 1995 .5 K-M estimate for 1990 cohort of 200 analyzed in 1995 + 0 years 5 1 .5 K-M estimate for 1993 cohort of 2000 analyzed in 1995 0 years 5 = 1 .5 K-M estimate for combined sample of 2200 analyzed in 1995 0 years 5

Thus, even though the long-term failure rate in both cohorts is 50%, the K-M curve remains near 100%. This is not a sample-size issue: the confidence intervals for the previous example are narrow (and can be made arbitrarily narrower by choosing higher sample sizes). Note that censoring is solely administrative.

III. The ASTRO Definition of Biochemical Failure (BF) in Prostate Cancer The American Society for Therapeutic Radiology and Oncology consensus statement on guidelines for PSA following radiation therapy (1997): “Three consecutive rises in prostate-specific antigen (PSA) after reaching the PSA nadir constitute BF. The date of failure is the midpoint between the nadir and the first of the three consecutive rises in PSA.”

A hypothetical PSA curve after radiation treatment PSA assay times PSA level treatment observed nadir backdated BF “at call” BF Time

Why Backdating is a Problem Problems with definition quickly noticed by Vicini et al., attributed to inadequate followup. They examined a series of prostate cancer patients treated with radiation and followed for up to 12 years. They artificially censored patients at a range of followups, recalculated backdated BF times, and plotted K-M curves.

Vicini, F. A. , Kestin, L. L. , and Martinez, A. A Vicini, F.A., Kestin, L.L., and Martinez, A.A. The importance of adequate follow-up in defining treatment success after external beam irradiation for prostate cancer. IJROBP 1999; 45:553-561.

Their conclusion: need more followup – at least 5, preferably 10 years. Vicini and others recommended that most or all patients be followed “at least beyond the time point at which actuarial results are examined”. This is problematic considering the lengthy progress of the disease and the frail patient population.

In what other medical specialties could backdating outcomes be problematic?