Doctor, How Am I Doing? Conditional Survival Analyses

Transcription

1 Doctor, How Am I Doing? Conditional Survival Analyses

2 Background Survival rates usually reported from time of diagnosis only Doesn't reflect changing hazard rates over time: early: higher hazard rate late: lower hazard rate

3 What is Conditional Survival? The probability of survival given that the patient has already survived a certain period of time since diagnosis.

4 Scenarios A patient asks, "Doctor, now that I have survived 2 years, what are the chances I will survive another 5 years? As a physician, what is an appropriate schedule for follow-up surveillance, based on changing risk? As a researcher, when designing a clinical trial, what is a sufficient followup time for reporting end points?

5 Conditional Survival Let S(t) be Kaplan-Meier survival at time t. Conditional Survival, CS(y x), is probability of surviving y yrs, given already survived x yrs. S( x y) CS( y x) S( x) e.g., CS(5 2) = S(5+2) / S(2)

6 Overall Survival Example Ten-Year Overall Survival 100% 90% 80% 70% S(5) CS(5 2) 60% 50% 40% Stage I Stage II Stage III Stage IV 30% 20% 10% 0% Years after Diagnosis

7 As time since diagnosis passes, changes in conditional survival reflect changes in the instantaneous hazard rate ("slope" of K-M curve).

8 Overall Survival Example 2 Ten-Year Overall Survival 100% 90% 80% 70% 60% 50% 40% Stage I Stage II Stage III Stage IV 30% 20% 10% 0% Years after Diagnosis

9 Conditional Survival for Patients with Colon Cancer: An Analysis of NSABP C-03 to C-06

10

11 Specific Aim To examine Conditional Survival of patients enrolled in the NSABP C-03 to C-06 colon cancer trials

12 Methods NSABP C-03 to C-06 Included all Fluorouracil-equivalent arms Retrospectively analyzed 10 years of disease-free (DFS) & overall survival (OS) data Subset by stage, # positive nodes, location, age, race, sex, performance status

13 Included Study Arms Trial Arm Randomized Ineligible Consent Withdraw No Follow-up Included C-03 LV + 5FU C-04 LV + 5FU C-04 LV+5FU+LEV C-05 LV + 5FU C-05 LV+5FU+IFN C-06 LV + 5FU C-06 LV + UFT TOTAL

14 Kaplan-Meier Overall Survival

15 5-Year Conditional Overall Survival 5-Yr Conditional OS 100% 80% 60% 40% 20% 0% OS 76% 75% 78% 81% 83% 85% After Recurrence 76% 7% 11% 10% 19% 11% DFS 76% 79% 85% 89% 90% 90% Years after Diagnosis

16 OS Conditional on DFS Pts with recurrence are known to have poorer prognosis After a pt has survived without disease for x years, what is their 5-yr OS? Variation of "pure" Conditional Survival Better estimate of prognosis for pts who are disease-free after x years

17 5-year overall Conditional Survival for all pas as a function of (1) the total number of yrs survived since diagnosis ( Alive ) and (2) the # yrs survived without recurrence or second primary cancer. Zamboni B A et al. JCO 2010;28:

18 5-Year Conditional Overall Survival By Stage 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Dukes B 87% 87% 90% 92% 91% 92% Dukes C 68% 72% 81% 87% 89% 88% Disease-Free Years after Diagnosis

19 5-yr overall Conditional Survival stratified by stage, as a function of the number of disease-free years survived.

20 5-Year Conditional Overall Survival Number of Positive Nodes 100% 80% 60% 40% 20% 0% Nodes 0 87% 87% 90% 92% 91% 92% Nodes % 77% 84% 88% 90% 89% Nodes % 63% 74% 84% 87% 88% Nodes % 47% 67% 76% 78% 81% Disease-Free Years after Diagnosis

21 5-yr overall Conditional Survival stratified by # positive nodes, as a function of the number of disease-free yrs survived.

22 5-Year Conditional Overall Survival By Age 100% 80% 60% 40% 20% 0% Age <50 78% 82% 90% 95% 95% 95% Age % 79% 86% 91% 92% 92% Age % 78% 83% 86% 87% 88% Age >70 71% 75% 80% 82% 76% 73% Disease-Free Years after Diagnosis

23 5-yr overall Conditional Survival stratified by age, as a function of the number of disease-free years survived. Zamboni B A et al. JCO 2010;28:

24 5-Year Conditional Overall Survival By Sex 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Male 76% 77% 83% 87% 88% 88% Female 77% 81% 88% 92% 92% 93% Disease-Free Years after Diagnosis

25 5-Year Conditional Overall Survival # of Resected Nodes 100% 80% 60% 40% 20% 0% Resected % 78% 84% 88% 89% 89% Resected % 77% 83% 88% 88% 88% Resected % 81% 86% 90% 90% 90% Resected >18 79% 81% 87% 91% 92% 93% Disease-Free Years after Diagnosis

26 5-Year Conditional Overall Survival Tumor Location 100% 80% 60% 40% 20% 0% Right Colon 75% 79% 86% 91% 91% 90% Left Colon 77% 79% 85% 89% 90% 90% Recto-Sigmoid 79% 80% 84% 88% 89% 90% Multiple 68% 71% 80% 83% 91% 90% Disease-Free Years after Diagnosis

27 5-Year Conditional Overall Survival Performance Status 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% PS 0 77% 80% 86% 90% 90% 91% PS 1 72% 74% 83% 87% 88% 86% PS 2 62% 65% 47% 52% 63% 56% Disease-Free Years after Diagnosis

28 5-yr overall Conditional Survival stratified by ECOG PS, as a function of the number of disease-free yrs survived.

29

30 Limitations of CS Calculation As more time elapses from Dx, # pts in each group decreases as pts recur, develop 2 nd cancer, or expire. Hence some of the trends may, in part, be due to smaller # of pts in the modal over time. Wider C.I. may be noted over time Exploratory analyses over eras may not reflect current diagnostic & Rx approaches

31 Summary Conditional Survival for colon cancer patients changes over time Largest CS increases seen for advanced stage, more positive nodes, younger age, females CS decreases for older age, poor performance status Number of positive nodes is prognostic, but number resected is not

32 Summary Known prognostic factors that are important at baseline (Stage, # positive nodes) appear to lose significance as disease-free period increases (CS curves converge)

33 Gastric Cancer Conditional Survival

34 Overall Survival (Relative) Csurvival, Gastric Cancer, SEER analysis, Wang et al: Gastric Cancer 10:153-8, 2007 Overall Survival 100% 90% 80% 70% 60% 50% 40% 30% 20% Localized Regional- Direct Ext Regional- Nodes Regional- Both Distant Mets 10% 0% Years after Diagnosis

35 5-Year Relative Conditional Survival Figure 2 All Patients By SEER Summary Stage 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Localized 61% 71% 78% 82% 85% 85% Regional-Direct Extension 26% 45% 57% 72% 77% 79% Regional-Lymph Nodes 34% 41% 53% 63% 72% 79% Regional-Both 15% 22% 37% 51% 63% 67% Distant Mets 2% 12% 29% 46% 58% 64% Years after Diagnosis

36 5-Year Relative Conditional Survival Figure 3 All Patients By Race 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% White 20% 40% 56% 68% 75% 78% Black 22% 45% 60% 69% 75% 79% Am Indian/Alaskan 16% 32% 41% 53% 61% 77% Asian/Pacific Islander 31% 51% 68% 78% 83% 86% Years after Diagnosis

37 Ovarian Cancer Conditional Survival

38 Ovarian Cancer Conditional Survival

39 Five-Year Conditional Survival By Histology 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% Epithelial - Clear Cell Epithelial - Endometrioid Epithelial - Mucinous Epithelial - Serous Epithelial - Undifferentiated Germ Cell Sex Cord / Stromal 0% Years after Diagnosis

40 HNSCC Conditional Survival

41 Rectal Cancer Conditional Survival

42 Rectal Cancer CS using SEER 17 database submitted Nov 2005, SEER Stat software

43 Rectal Cancer CS using SEER 17 database submitted Nov 2005, SEER Stat software

44 Wang et al, Gastroint Cancer Res 1:84-89, 2007

45 Wang et al, Gastroint Cancer Res 1:84-89, 2007

46 Wang et al, Gastroint Cancer Res 1:84-89, 2007

47 Nomogram for Conditional Survival

48 Nomogram for Conditional Survival

49 Additional Demographic & Tumor-related Variables for Conditional Survival Nomogram Calculations

50 Limitations of CS Calculation As more time elapses from Dx, # pts in each group decreases as pts recur, develop 2 nd cancer, or expire. Hence some of the trends may, in part, be due to smaller # of pts in the modal over time. Wider C.I. may be noted over time Exploratory analyses over eras may not reflect current diagnostic & Rx approaches

51 Recent Conditional Survival Analyses Gastric Dikken, van de Velde et al, 2011 GI Cancers Symposium HNSCC Van der Schroeff et al, Erasmus Med Ctr, Rotterdam, Head & Neck 32:1613, 2010 NSCLC Stage I s/p Lobectomy Groth et al, UMinn, Ann Thorac Surg 90:375, 2010 Liver mets Melanoma Nathan et al, JACS 210:755, 2010 Rueth et al, Uminn, Ann Surg Oncol 17:1668, 2010 European database analysis Janssen-Heijnen et al, JCO 28:2520, 2010 NHL Diffuse-large cell Moller et al, Odense Univ, Dennmark Review Cancer 106:2165, 2006 Merrill et al, The Oncologist 15:873, 2010

52 Conditional Survival Applications At annual follow-ups, patients can be given more accurate survival estimates Clinicians can use to plan more appropriate follow-up surveillance Researchers can use to determine follow-up time in clinical trial design

53 Conclusions Conditional Survival quantifies changing risk profile over time CS is a useful adjunct to traditional survival statistics CS has practical relevance to patients, clinicians, & researchers

54 Acknowledgements OHSU Samuel J. Wang, MD, PhD Mehee Choi, MD UTHSCSA C. Dave Fuller, MD, PhD-C NSABP Beth A. Zamboni, MS H. Samuel Wieand, PhD Greg Yothers, PhD James J. Dignam, PhD Peter C. Raich, MD Michael J. O'Connell, MD Norman Wolmark, MD Portland State University Jong-Sung Kim, PhD Northwest Kaiser Permanente Applied Res. Dean F. Settig, PhD MDACC Randal Weber & David Rosenthal