Individual Prediction

Transcription

1 Individual Prediction Michael W. Kattan, Ph.D. Professor of Medicine, Epidemiology and Biostatistics, Cleveland Clinic Lerner College of Medicine of Case Western Reserve University Chairman, Department of Quantitative Health Sciences, Cleveland Clinic

2 Disclosures Consultant for GSK and Merck

3 Hodgkin s Disease Prognosis 1 I II III Survival IV 0 Time

4 Many clinical counseling tools are not designed to predict accurately Problems with my prediction: Didn t feel very tailored! Not adjusted for age, comorbidities Categories (e.g., extent of disease) were very broad Was this staging system really optimized for prediction?

5 When The Patient Wants A Prediction, What Options Does The Clinician Have? Predict based on knowledge and experience Quote an overall average to all patients Assign the patient to a risk group, i.e. high, intermediate, or low Apply a model

6 Preoperative Nomogram for Prostate Cancer Recurrence Points PSA T2a T2c T3a Clinical Stage T1c T1ab T2b ? Biopsy Gleason Grade Total Points Month Rec. Free Prob Instructions for Physician: Locate the patient s PSA on the PSA axis. Draw a line straight upwards to the Points axis to determine how many points towards recurrence the patient receives for his PSA. Repeat this process for the Clinical Stage and Biopsy Gleason Sum axes, each time drawing straight upward to the Points axis. Sum the points achieved for each predictor and locate this sum on the Total Points axis. Draw a line straight down to find the patient s probability of remaining recurrence free for 60 months assuming he does not die of another cause first. Instruction to Patient: Mr. X, if we had 100 men just like you, we would expect <predicted percentage from nomogram> to remain free of their disease at 5 years following radical prostatectomy, and recurrence after 5 years is very rare. Kattan MW et al: JNCI 1998; 90: Michael W. Kattan and Peter T. Scardino

7 CaPSURE Heterogeneity within Risk Groups Preoperative Nomogram Predicted Probability Nomogram Values by Prostate Cancer Risk Group Low Intermediate High Risk Group J Urol Apr;173(4):

8 The consequence of risk stratification relative to a statistical model Mr. X, from the Cleveland Clinic: PSA=6, clinical stage = T2c, biopsy Gleason sum=9, planned dose of 66.6 Gy without neoadjuvant hormones Radiation risk stratification: 5 yr. Surgery nomogram: 5yr. Radiation therapy nomogram: 5yr. Kattan MW, et al., J Clin. Oncol., 2000.

9 Heterogeneity within stages for gastric cancer AJCC Percent of Patients within AJCC Stage IV (32) IIIB (24) IIIA (69) II (117) IB (115) IA (102) Nomogram Predicted Probability of 5-Year Disease-Specific Survival

10 Continuous Models vs. Staging/Grouping Systems Model Comparator CI (M vs C) Preop L/I/H Risk Groups 0.67 vs Preop + IL6/TGFβ1 L/H Risk Groups 0.84 vs Pre XRT L/I/H Risk Groups 0.76 vs Melanoma SLN+ AJCC Stage 0.69 vs Pancreatic Ca AJCC Stage 0.64 vs Gastric Ca AJCC Stage 0.77 vs Breast Ca NPI Groups 0.69 vs Sarcoma CART Groups 0.77 vs. 0.74

11 When The Patient Wants A Prediction, What Options Does The Clinician Have? Predict based on knowledge and experience Quote an overall average to all patients Assign the patient to a risk group, i.e. high, intermediate, or low Apply a model

12 Urologists vs. Preoperative Nomogram 10 case descriptions from 1994 MSKCC patients presented to 17 urologists In addition to PSA, biopsy Gleason grades, and clinical stage, urologists were provided with patient age, systematic biopsy details, previous biopsy results, and PSA history. Preoperative nomogram was provided. Urologists were asked to make their own predictions of 5 year progression-free probabilities with or without use of the preoperative nomogram. Concordance indices: Nomogram = 0.67 Urologists = 0.55, p<0.05 Ross P et al., Semin Urol Oncol, 2002.

13 Nomogram for predicting the likelihood of additional nodal metastases in breast cancer patients with a positive sentinel node biopsy Vanzee K, et al., Ann Surg Oncol., 2003.

14 Breast Cancer Prediction: 17 Clinicians vs. Nomogram on 33 Patients Clinician AUC 0.54 Nomogram AUC 0.72 Areas Nomogram Specht et al., Ann Surg Oncol., 2005.

15 Clinical Gestalt is Highly Variable Dendreon: Proprietary and Confidential

16 Survey Results Concordance index Dendreon: Proprietary and Confidential nomo doc2 doc8 doc11 doc18 doc14 doc24 doc13 doc22 doc5 doc1 doc7 doc9 doc17 doc12 doc15 doc16 doc4 doc21 doc20 doc23 doc10 doc3 doc19 doc6

17 Comparative Effectiveness Treatment options Benefits T1 T2 B1 B2 Harms H1 H2 Must tailor the probabilities to the individual patient. Kattan MW, Med Decis Making, 2009.

18 Risk Calculator - using EHR data from EPIC

19 Risk Calculator

20

21

22

23 EPIC Integration

24 Conclusions Data from the EHR, while imperfect, yield reasonably accurate statistical prediction models. The many modeling options should be compared with respect to predictive accuracy. Patients should be provided with individual predictions for choices with no easy answers (i.e., involving trade offs). The most accurate predictions presently available should be used, and these are likely from regression models. Clinician judgment, risk groups, risk factor counting, overall treatment effects from RCTs, are all less helpful. Free web software should help with dissemination. Followup data collection is the biggest challenge.