The Length of Time a Cancer Patient Survival After Diagnosis Discrete or Continuous Data
Ann Surg Oncol. Author manuscript; available in PMC 2011 Aug 16.
Published in final edited form as:
PMCID: PMC3156394
NIHMSID: NIHMS312341
An Interactive Tool for Individualized Estimation of Conditional Survival in Rectal Cancer
Samuel J Wang
1Department of Radiation Medicine, Oregon Health & Science University, Portland, OR
4Department of Medical Informatics and Clinical Epidemiology, Oregon Health & Science University, Portland, OR
Amanda R. Wissel
1Department of Radiation Medicine, Oregon Health & Science University, Portland, OR
Join Y Luh
2Department of Radiation Oncology, St. Joseph Hospital, Eureka, CA
3Department of Radiation Oncology, University of Texas Health Science Center, San Antonio, TX
C David Fuller
3Department of Radiation Oncology, University of Texas Health Science Center, San Antonio, TX
Jayashree Kalpathy-Cramer
4Department of Medical Informatics and Clinical Epidemiology, Oregon Health & Science University, Portland, OR
Charles R Thomas, Jr
1Department of Radiation Medicine, Oregon Health & Science University, Portland, OR
Abstract
Background
For rectal cancer patients who have already survived a period of time after diagnosis, survival probability changes and is more accurately depicted by conditional survival. The specific aim of this study was to develop an interactive tool for individualized estimation of changing prognosis for rectal cancer patients.
Methods
A multivariate Cox proportional hazards (CPH) survival model was constructed using data from rectal cancer patients diagnosed from 1994 to 2003 from the Surveillance, Epidemiology, and End Results (SEER) database. Age, race, sex, and stage were used as covariates in the survival prediction model. The primary outcome variable was overall survival conditional on having survived up to 5 years from diagnosis.
Results
Data from 42,830 rectal cancer patients met the inclusion criteria. The multivariate CPH model showed age, race, sex, and stage as significant independent predictors of survival. The survival prediction model demonstrated good calibration and discrimination, with a bootstrap-corrected concordance index of 0.75. A web-based prediction tool was built from this regression model that can compute individualized estimates of changing prognosis over time.
Conclusions
An interactive prediction modeling tool can estimate prognosis for rectal cancer patients who have already survived a period of time after diagnosis and treatment. Having more accurate prognostic information can empower both patients and clinicians to be able to make more appropriate decisions regarding follow-up, surveillance testing, and future treatment.
Background
A cancer patient's prognosis is often based on aggregate data from large heterogeneous groups of patients in the published oncology literature, and these estimates are usually only made at the time of diagnosis. Survival probability however, changes over time for cancer patients, and estimates of prognosis made at the time of diagnosis are no longer applicable after a patient has already survived for a period after diagnosis and treatment [1]. Conditional survival (CS) is a more accurate estimate of prognosis for these cancer survivors because it accounts for the continuously changing hazard rates over time [2, 3]. CS estimates can be helpful for patients and providers who seek more accurate prognostic estimates to help guide health-related decisions. This information enables cancer survivors to more accurately assess whether their prognosis is improving or declining over time. It can also be used by providers to determine more appropriate therapy and follow-up surveillance testing.
Several authors have published studies on CS for breast, colon, central nervous system (CNS), lung, head and neck, and other advanced carcinomas [2–19]. We have also previously published our work on CS for various disease sites, including breast, head and neck, gynecologic, gastrointestinal and other malignancies[20–31].
There are a number of important cancer risk prediction models being used today for various cancer sites, including prostate, breast, colon, gallbladder, and other disease sites [6, 26, 32–39]. While rectal cancer has some similar characteristics to colon cancer, there are also important distinctions in how rectal cancer is treated because of its retroperitoneal location in the low pelvis. To date, there have been no prediction models available to specifically depict predict how prognosis changes for rectal cancer patients after they have survived a time interval after diagnosis and treatment. Such individualized prognostic information is extremely valuable to both patients and providers.
In our previous work, we demonstrated how conditional survival changes over time for rectal cancer patients using basic Kaplan-Meier survival analyses [25]. In this follow-up study, we constructed a Cox proportional hazards survival regression model from these data to enable estimation for individual patients. Using this regression model, we built an interactive web-based prediction tool can provide customized conditional survival predictions for rectal cancer patients who have already survived a period of time following their diagnosis and treatment.
Methods
CS is derived from the concept of conditional probability in biostatistics. CS can be calculated from traditional Kaplan-Meier or actuarial life table survival data. The mathematical definition of CS [1] can be expressed as follows: CS(y|x), is the probability of surviving an additional y years, given that the person has already survived x years. Let S(t) be the traditional actuarial life table survival at time t. CS can be expressed as:
For example, to compute the 5-year CS for a patient that has already survived 2 years, the survival at 5+2 years, S(7), is divided by the survival at 2 years, S(2). When a survival curve has a changing hazard rate over time, this will be reflected as a change in CS as more time elapses from the time of diagnosis.
The Surveillance, Epidemiology, and End Results (SEER 17) database from the National Cancer Institute is a population-based cancer registry covering approximately 26% of the US population across several disparate geographic regions and is the largest publicly available domestic cancer dataset[40]. SEER program registries collect data on patient demographics, cancer type and stage, first course of treatment, and follow up vital status. We analyzed 42,830 patients diagnosed with rectal cancer between 1994 and 2003, with follow-up through 2006, using the SEER 17 database (April 2009 release) using SEER*Stat version 6.6.2 software[40,41]. Patients who had a "site recode" data field of "Rectum" were selected. The Case Selection options in SEER*Stat were used to restrict cases to "Actively Followed", "Malignant Behavior," "Known Age," and "Cases in Limited Use Database". Excluded were "All Death Certificate Only and Autopsy Only," "Second and Later Primaries" and "Alive with No Survival Time."
A Cox proportional hazards (CPH) multivariate regression model was constructed using these data. The primary outcome variable was overall survival, conditional upon 0 to 5 years already survived. Covariates included in the model were age, race (White, Black, Asian/Pacific Islander, Alaskan/American Indian), sex, and stage (AJCC TNM 3rd edition). All covariates were treated as discrete and converted to binary variables, except for age, which was modeled as a continuous variable and fitted to a smoothed restricted cubic spline function as per Harrell [42]. The model was internally validated for both discrimination and calibration using bootstrap correction with 100 resamples.
The CPH regression model was incorporated into a web-browser based software application. Utilizing this tool, users can enter specific tumor and patint characteristics, and the system will calculate a customized conditional survival curve specific to that individual patient. This software tool was built using JavaScript.
This research study was determined to be exempt from requiring Institutional Review Board approval by the authors' home institution.
Results
A total of 42,380 rectal cancer patients were included in the analysis. Numbers of patients in each stage according to sex, age, and race are shown in Table 1. Ten year actuarial survival data were used to calculate 5-year observed CS in categories of stage, age, gender and race. Although higher-stage patients had lower CS overall at diagnosis compared to lower-stage patients, higher-stage patients had greater increases in CS as more time elapsed from diagnosis. As the time survived since diagnosis increased from 0 to 5 years, the 5-year observed conditional survival for Stage I patients increased minimally from 71% to 74%, while CS for Stage II increased from 55% to 67%. Stage III increased from 50% to 66%, and Stage IV increased from 7% to 53%. Patients aged 65 years and over at diagnosis had lower CS than those under 65 years, both at diagnosis (44.3% vs. 65.7%) and at five years from diagnosis (60.3% vs. 82.1%). For all stages of disease, men had slightly lower 5-year survival than women, both at diagnosis (52.0% vs. 55.1%) and after 5 years (69.9% vs. 72.1%). The largest differences by sex were in stage IV patients. Among race, the largest disparities were in stage IV disease, with African American patients having lower CS compared to White and "other" patients, particularly with longer time intervals from diagnosis.
Table 1
Patient & Tumor Characteristics (N=42,830)
| Median Age (range) | 67 (14–107) | Stage | |
| Female Sex (%) | 18,278 (43%) | I | 17,922 (42%) |
| Race | II | 9,309 (22%) | |
| White | 35,332 (82%) | III | 9,073 (21%) |
| African-American | 3,350 (8%) | IV | 6,526 (15%) |
| Native American/Alaskan | 281 (<1%) | ||
| Asian/Pacific Islander | 3,673 (9%) | ||
| Other/Unknown | 194 (<1%) |
A CPH model was built from these data using age, race, sex, and stage as covariates. The beta coefficients for the CPH model are shown in Table 2. All covariates were found to be statistically significant prognostic factors in this multivariate analysis. The C-index for discrimination for this CPH model was 0.75. The calibration curve showed good agreement between predicted and observed outcomes.
Table 2
Cox proportional hazards multivariate regression analysis. These beta coefficients were used to construct the nomogram.
| Covariate | Beta Coefficient | Hazard Ratio | Lower 95% CI | Upper 95% CI |
|---|---|---|---|---|
| Age | 0.00348 | - | - | - |
| Age′ | 0.04854 | - | - | - |
| Age″ | −0.08818 | - | - | - |
| Sex | ||||
| Female | 0 | |||
| Male | 0.1519 | 1.16 | 1.13 | 1.20 |
| Race | ||||
| White | 0 | |||
| African American | 0.28916 | 1.34 | 1.27 | 1.40 |
| Asian or 1 Pacific Islander | −0.09861 | 0.91 | 0.86 | 0.95 |
| American Indian/Alaska Native | 0.21855 | 1.24 | 1.06 | 1.46 |
| Other | −0.76471 | 0.47 | 0.15 | 1.48 |
| Uknown | −0.99218 | 0.37 | 0.25 | 0.56 |
| Stage | ||||
| Stage I | 0 | |||
| Stage II | 0.49242 | 1.64 | 1.57 | 1.70 |
| Stage III | 0.73063 | 2.08 | 2.00 | 2.16 |
| Stage IV | 2.24072 | 9.40 | 9.05 | 9.76 |
Figure 3 shows a screenshot of the interactive web-based prediction tool that was built from the CPH survival model. The user enters specific information about a rectal cancer patient, including time already survived since diagnosis. The tool will then estimate the percent likelihood the patient will survive up to an additional 5 years from that point in time. The interactive nature of the tool allows the user to see how a patient's prognosis often improves with additional time survived since diagnosis. This web-based prediction tool is available for public use and can be found at http://skynet.ohsu.edu/nomograms.
Web-Based Conditional Survival Prediction Tool
To use this tool, the user can enter specific patient information including time since diagnosis, and the application will calculate an individualized estimate of prognosis starting from that point in time.
Discussion
Conditional survival quantifies changes in prognosis as time progresses from diagnosis. With this interactive web-based tool, patients and providers will have the ability to more accurately quantify changes in prognosis over time. This information is potentially of great interest to patients, clinicians and researchers. Patients should be given an accurate risk assessment and prognosis that accounts for time already survived since diagnosis. Clinicians could also use this tool to implement a more evidence-driven approach to planning post-therapy surveillance, based on a patient's changing risk. Physicians often arbitrarily taper follow-up visit frequency after 2 to 3 years, but it may be more appropriate to schedule follow-up testing frequency and duration based on an individual patient's actual disease risk, rather than based on generalized guidelines.
Conditional Survival may also assist in detecting differences in subgroup survival patterns [8, 14]. In other studies of CS, it has been found that patients with poorer initial prognoses exhibit the greatest increases in CS as they survive for longer periods of time from diagnosis [4, 8, 11, 12]. Our analysis indicates that this is also true for rectal cancer, with CS appearing to increase most for patients over 65 years of age and those with advanced-stage disease.
Although CS generally increased over time for all patient groups examined in this study, CS increases are not necessarily observed for all cancer types. In analyses of disease at other sites, a CS decline has been observed for certain population subsets, such as the elderly, patients with unfavorable tumor histology, and those with poor performance status [11, 12, 31]. Consequently, CS trends over time cannot be predicted from point estimates made at the time of diagnosis. A full Kaplan-Meier survival plot must be examined in order to determine the trend of CS over time; in essence, the shape of the Kaplan-Meier plot influences whether CS increases or decreases over time. CS quantifies prognostic changes over time in a manner readily comprehended by both providers and patients.
There are several limitations to this study. Treatment data in SEER are limited (ie, no chemotherapy regimen or radiotherapy dose is recorded), so we did not attempt a sub-analysis by treatment modality. In addition, some subgroups had small sample size, which precludes our ability to make more definite generalizations regarding the observed differences between some subgroups. Since SEER does not include information on disease recurrence, and since cause of death information is not always reliable, we are unable to analyze other important end points, such as time to recurrence and disease-free survival. Finally, historic survival data collected over an extended period of time may not reflect current practices in oncology. Despite these limitations, however, the SEER dataset is the largest population-based cancer registry in the U.S., and it allows us to make reasonable estimates of CS that are generally applicable for most patients with rectal cancer.
Conditional survival data and this prediction calculator should be considered within the individual context of patient care. Although CS provides a more accurate quantification of prognosis for long-term survivors, it is important that this tool be used in combination with other factors that may influence mortality risk that may not be accounted for in this model.
Conclusions
Conditional survival is a method for quantifying a changing risk profile in terms that are meaningful and easily understood by patients and providers. We have constructed a prediction model that estimates conditional survival for rectal cancer patients. This web-based interactive prediction tool can be used by patients and providers to make individualized estimates of updated prognosis for patients who have already survived from 1 to 5 years after diagnosis. We believe that more accurate estimation of CS will permit improved risk assessment and follow up care for rectal cancer survivors.
Ten-Year Kaplan-Meier Overall Survival By Stage
These data were used to calculate the 5-year observed conditional survival probabilities.
Five-year Observed Conditional Survival By Stage
Each bar represents the 5-year conditional survival for a specific stage after the patient has already survived the indicated number of years since diagnosis.
Acknowledgments
Supported in part by the N.L. Tartar Research Fund.
Footnotes
Disclosures: None
Competing interests
The authors disclosed no potential conflicts of interest.
Authors' contributions
SJW, JYL, CDF, JKC, and CRT carried out the conception and design of this study. SJW and AWS coordinated the collection and assembly of data. SJW, AWS, CDF and JKC participated in data analysis and interpretation. SJW, AWS, JYL and CDF each helped draft the manuscript. SJW and CRT provided final approval of the manuscript.
References
1. Henson DE, Ries LA. On the estimation of survival. Semin Surg Oncol. 1994;10(1):2–6. [PubMed] [Google Scholar]
2. Gloeckler Ries LA, Reichman ME, Lewis DR, Hankey BF, Edwards BK. Cancer survival and incidence from the Surveillance, Epidemiology, and End Results (SEER) program. Oncologist. 2003;8(6):541–552. [PubMed] [Google Scholar]
3. Merrill RM, Hunter BD. Conditional survival among cancer patients in the United States. Oncologist. 2010;15(8):873–882. [PMC free article] [PubMed] [Google Scholar]
4. Henson DE, Ries LA, Carriaga MT. Conditional survival of 56,268 patients with breast cancer. Cancer. 1995;76(2):237–242. [PubMed] [Google Scholar]
5. Merrill RM, Henson DE, Ries LA. Conditional survival estimates in 34,963 patients with invasive carcinoma of the colon. Dis Colon Rectum. 1998;41(9):1097–1106. [PubMed] [Google Scholar]
6. Chang GJ, Hu CY, Eng C, Skibber JM, Rodriguez-Bigas MA. Practical application of a calculator for conditional survival in colon cancer. J Clin Oncol. 2009;27(35):5938–43. [PMC free article] [PubMed] [Google Scholar]
7. Nathan H, de Jong MC, Pulitano C, et al. Conditional survival after surgical resection of colorectal liver metastasis: an international multi-institutional analysis of 949 patients. J Am Coll Surg. 2010;210(5):755–64. 764–6. [PubMed] [Google Scholar]
8. Davis FG, McCarthy BJ, Freels S, Kupelian V, Bondy ML. The conditional probability of survival of patients with primary malignant brain tumors: surveillance, epidemiology, and end results (SEER) data. Cancer. 1999;85(2):485–491. [PubMed] [Google Scholar]
9. Hwang SL, Yang YH, Lieu AS, et al. The conditional survival statistics for survivors with primary supratentorial astrocytic tumors. J Neurooncol. 2000;50(3):257–264. [PubMed] [Google Scholar]
10. Lin CL, Lieu AS, Lee KS, et al. The conditional probabilities of survival in patients with anaplastic astrocytoma or glioblastoma multiforme. Surg Neurol. 2003;60(5):402–6. discussion 406. [PubMed] [Google Scholar]
11. Merrill RM, Henson DE, Barnes M. Conditional survival among patients with carcinoma of the lung. Chest. 1999;116(3):697–703. [PubMed] [Google Scholar]
12. Skuladottir H, Olsen JH. Conditional survival of patients with the four major histologic subgroups of lung cancer in Denmark. J Clin Oncol. 2003;21(16):3035–3040. [PubMed] [Google Scholar]
13. Yang YH, Liu SH, Ho PS, et al. Conditional survival rates of buccal and tongue cancer patients: how far does the benefit go? Oral Oncol. 2009;45(2):177–83. [PubMed] [Google Scholar]
14. Kato I, Severson RK, Schwartz AG. Conditional median survival of patients with advanced carcinoma: surveillance, epidemiology, and end results data. Cancer. 2001;92(8):2211–2219. [PubMed] [Google Scholar]
15. Janssen-Heijnen ML, Houterman S, Lemmens VE, et al. Prognosis for long-term survivors of cancer. Ann Oncol. 2007;18(8):1408–13. [PubMed] [Google Scholar]
16. Janssen-Heijnen ML, Gondos A, Bray F, et al. Clinical relevance of conditional survival of cancer patients in europe: age-specific analyses of 13 cancers. J Clin Oncol. 2010;28(15):2520–8. [PubMed] [Google Scholar]
17. Bowles TL, Xing Y, Hu CY, et al. Conditional survival estimates improve over 5 years for melanoma survivors with node-positive disease. Ann Surg Oncol. 2010;17(8):2015–23. [PubMed] [Google Scholar]
18. Xing Y, Chang GJ, Hu CY, et al. Conditional survival estimates improve over time for patients with advanced melanoma: results from a population-based analysis. Cancer. 2010;116(9):2234–41. [PMC free article] [PubMed] [Google Scholar]
19. Rueth NM, Groth SS, Tuttle TM, et al. Conditional survival after surgical treatment of melanoma: an analysis of the Surveillance, Epidemiology, and End Results database. Ann Surg Oncol. 2010;17(6):1662–8. [PubMed] [Google Scholar]
20. Wang SJ, Luh JY, Thomas CR. Conditional Survival of Breast Cancer Patients. American Society of Breast Diseases, 29th Annual Symposium; Las Vegas, NV. [Google Scholar]
21. Wang SJ, Luh JY, Fuller CD, Thomas CR. Impact of ethnicity on conditional survival of breast cancer patients: analysis from the SEER database. 28th Annual San Antonio Breast Cancer Symposium; San Antonio, TX. [Google Scholar]
22. Luh JY, Wang SJ, Fuller CD, Thomas CR. A SEER database analysis of conditional survival for prostate cancer patients. American Society of Clinical Oncology, 42nd Annual Meeting; Atlanta, GA. [Google Scholar]
23. Wang SJ, Fuller CD, Luh JY, Thomas CR, Bleyer WA. Older Adolescents and Young Adults with Cancer: Conditional Survival Deficit. American Society for Therapeutic Radiology and Oncology, 48th Annual Meeting; Philadelphia, PA. [Google Scholar]
24. Fuller CD, Wang SJ, Thomas CR, et al. Conditional survival in head and neck squamous cell carcinoma: Results from the SEER Dataset 1973–1998. Cancer. 2007;109(7):1331–1343. [PubMed] [Google Scholar]
25. Wang SJ, Fuller CD, Emery R, Thomas CR. Conditional Survival in Rectal Cancer: A SEER Database Analysis. Gastrointest Cancer Res. 2007;1(3):84–89. [PMC free article] [PubMed] [Google Scholar]
26. Wang SJ, Fuller CD, Sittig DF, Holland JM, Thomas CR. A regression model for predicting conditional survival for head & neck cancer patients: a SEER analysis. ASCO Annual Meeting; Chicago, IL. [Google Scholar]
27. Wang SJ, Emery R, Fuller CD, et al. Conditional survival in gastric cancer: a SEER database analysis. Gastric Cancer. 2007;10(3):153–158. [PubMed] [Google Scholar]
28. Wang SJ, Fuller CD, Thomas CR., Jr Ethnic disparities in conditional survival of patients with non-small cell lung cancer. J Thorac Oncol. 2007;2(3):180–90. [PubMed] [Google Scholar]
29. Choi M, Fuller CD, Thomas CR, Jr, Wang SJ. Conditional survival in ovarian cancer: results from the SEER dataset 1988–2001. Gynecol Oncol. 2008;109(2):203–9. [PubMed] [Google Scholar]
30. Bleyer A, Choi M, Fuller CD, Thomas CR, Jr, Wang SJ. Relative lack of conditional survival improvement in young adults with cancer. Semin Oncol. 2009;36(5):460–7. [PubMed] [Google Scholar]
31. Zamboni BA, Yothers G, Choi M, et al. Conditional survival and the choice of conditioning set for patients with colon cancer: an analysis of NSABP trials C-03 through C-07. J Clin Oncol. 2010;28(15):2544–8. [PMC free article] [PubMed] [Google Scholar]
32. Kattan MW, Zelefsky MJ, Kupelian PA, et al. Pretreatment nomogram for predicting the outcome of three-dimensional conformal radiotherapy in prostate cancer. J Clin Oncol. 2000;18(19):3352–3359. [PubMed] [Google Scholar]
33. Kattan MW. When and how to use informatics tools in caring for urologic patients. Nat Clin Pract Urol. 2005;2(4):183–190. [PubMed] [Google Scholar]
34. Ravdin PM. A computer based program to assist in adjuvant therapy decisions for individual breast cancer patients. Bull Cancer. 1995;82 (Suppl 5):561s–564s. [PubMed] [Google Scholar]
35. Rouzier R, Pusztai L, Delaloge S, et al. Nomograms to predict pathologic complete response and metastasis-free survival after preoperative chemotherapy for breast cancer. J Clin Oncol. 2005;23(33):8331–8339. [PubMed] [Google Scholar]
36. Wang SJ, Fuller CD, Kim JS, et al. Prediction model for estimating the survival benefit of adjuvant radiotherapy for gallbladder cancer. J Clin Oncol. 2008;26(13):2112–7. [PubMed] [Google Scholar]
37. Kattan MW, Karpeh MS, Mazumdar M, Brennan MF. Postoperative nomogram for disease-specific survival after an R0 resection for gastric carcinoma. J Clin Oncol. 2003;21(19):3647–3650. [PubMed] [Google Scholar]
38. Gross ND, Patel SG, Carvalho AL, et al. Nomogram for deciding adjuvant treatment after surgery for oral cavity squamous cell carcinoma. Head Neck. 2008;30(10):1352–60. [PubMed] [Google Scholar]
39. Abu-Rustum NR, Zhou Q, Gomez JD, et al. A nomogram for predicting overall survival of women with endometrial cancer following primary therapy: toward improving individualized cancer care. Gynecol Oncol. 2010;116(3):399–403. [PMC free article] [PubMed] [Google Scholar]
40. Surveillance, Epidemiology, and End Results (SEER) Program. National Cancer Institute, DCCPS, Surveillance Research Program, Cancer Statistics Branch; SEER*Stat Database: Incidence - SEER 17 Regs Limited-Use + Hurricane Katrina Impacted Louisiana Cases, Nov 2008 Sub (1973–2006 varying) -Linked to County Attributes - Total U.S., 1969–2006 Counties. ( www.seer.cancer.gov) released April 2009, based on the November 2008 submission. [Google Scholar]
41. Surveillance Research Program, National Cancer Institute SEER*Stat software. ( www.seer.cancer.gov/seerstat) version 6.6.2.
42. Harrell FE. Regression Modeling Strategies. New York: Springer-Verlag; 2001. [Google Scholar]
fullwoodwentented.blogspot.com
Source: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3156394/
0 Response to "The Length of Time a Cancer Patient Survival After Diagnosis Discrete or Continuous Data"
Post a Comment