1 Impact of Age, Tumor Risk, and Comorbidity on Competing Risks for Survival in a U.S.

2 Population-Based Cohort of Men with Prostate Cancer

3 Running Title: Age, Comorbidity, and Tumor Risk in CaP
4
5 Timothy J. Daskivich, MD1
6 Kang-Hsien Fan, MS2
7 Tatsuki Koyama, PhD 2
8 Peter C. Albertsen, MD3
9 Michael Goodman, MD, MPH4
10 Ann S. Hamilton, PhD5
11 Richard M. Hoffman, MD, MPH6
12 Janet L. Stanford, PhD, MPH7
13 Antoinette M. Stroup, PhD8
14 Mark S. Litwin, MD, MPH1
15 David F. Penson, MD, MPH9,10
16
1
17 Department of Urology, David Geffen School of Medicine, University of California, Los Angeles,
18 Los Angeles, CA
2
19 Department of Biostatistics, Vanderbilt University, Nashville, TN
3
20 Department of Urology, University of Connecticut, Farmington, CT
4
21 Department of Epidemiology, Emory University, Atlanta, GA
5
22 Department of Preventative Medicine, Keck School of Medicine, University of Southern
23 California, Los Angeles, CA
6
24 Department of Internal Medicine, University of New Mexico, Albuquerque, NM
7
25 Department of Urology, Fred Hutchinson Cancer Center, Seattle, WA
8
26 Department of Internal Medicine, University of Utah, Salt Lake City, UT
9
27 Department of Urology, Vanderbilt University, Nashville, TN
10
28 VA Tennessee Valley Geriatric Research, Education, and Clinical Center (GRECC), Nashville,
29 TN
30
31
32 Corresponding author:
33 David F. Penson, MD, MPH
34 Professor of Urologic Surgery
35 Director, Center for Surgical Quality and Outcomes Research
36 Vanderbilt University Medical Center
37 2525 West End Avenue, Suite 600
38 Nashville, TN 37203-1738
39 (615) 343-1529

1
1 fax (615) 936-8291
2
3 david.penson @vanderbilt.edu
4
5 Word Count: 3604 Abstract: 299 Number of Tables: 2 Number of Figures: 3
6

2
ABSTRACT

Background: Accurate estimation of life expectancy is essential to offering appropriate

care to men with early-stage prostate cancer, but mortality risks associated with

comorbidity remain poorly defined.

Objective: To determine the impact of age, comorbidity, and tumor risk on other-cause

and prostate cancer-specific mortality in men with early-stage disease

Design: Prospective cohort study

Patients: 3,183 men with non-metastatic prostate cancer at diagnosis

Setting: A nationally representative, population-based cohort

Measurements and analyses: Baseline self-reported comorbidity scored as a count of

twelve major comorbid conditions, tumor characteristics, initial treatment, and overall

and disease-specific mortality through 14-year follow up. Survival analyses accounting

for competing risks.

Results: Fourteen-year cumulative other-cause mortality was 24%, 33%, 46%, and 57%,

for men with comorbidity counts of 0, 1, 2, and 3+, respectively. For men diagnosed at

age 65, subhazard ratios (95%CI) for other-cause mortality among men with 1, 2, or 3+

comorbidities (vs. none) were 1.2 (1.0–1.5), 2.0 (1.6–2.4), and 2.6 (2.1–3.2), respectively.

With each decade increase in age, the subhazard of other-cause mortality associated

with each comorbidity count doubled; in men with 3+ comorbidities, subhazard ratios

(95%CI) for ages 65 and 75 (vs. 55) were 2.0 (1.6–2.5) and 4.0 (2.7–6.1). In men with

3+ comorbidities, ten-year other-cause mortality rates were 26%, 40%, and 71% for

those aged ≤60, 61–74, and ≥75 at diagnosis, respectively. Prostate cancer mortality

was minimal in low-(3%) and intermediate-risk(7%) disease but appreciable in high-risk

3
disease(18%) and did not vary by co-morbidity count will all groups having 10-11%

prostate cancer specific mortality.

Limitations: Self-report of comorbid conditions

Conclusions: Older men with multiple major comorbidities are at high risk of other-cause

mortality within ten years of diagnosis. Such men should consider this information when

deciding between conservative management and aggressive treatment for low- or

intermediate-risk prostate cancer.

4
INTRODUCTION

Men with a new diagnosis of clinically localized prostate cancer are faced with numerous

treatment options, ranging from no initial therapy (watchful waiting or active surveillance)

to aggressive therapy with surgery and radiation. The first question a newly diagnosed

man should consider is whether immediate aggressive treatment is necessary. Because

the survival benefits of aggressive treatment for low- and intermediate-risk disease are

delayed for eight to ten years (1), clinical guidelines recommend that men with a life

expectancy of shorter than ten years be spared the morbidity and expense of aggressive

treatment (2,3).

Despite the recognized importance of estimating life expectancy in medical decision

making for men with prostate cancer, physicians remain poor judges of prognosis (4);

this often leads to inappropriate treatment decisions. Recent retrospective data have

shown that men with Charlson scores of 3 or greater are treated aggressively with

surgery or radiation more often than not, despite a 70% other-cause mortality rate at

eight years after diagnosis (5). This practice may be due to several reasons. First, the

long-term risk of other-cause mortality associated with different ages and comorbidity

states remains unclear. The current data on risk of other-cause mortality associated with

comorbidity are either from institutional case series (6-9) or from populations treated

with only one treatment modality(10); to our knowledge, to date there has been no true

population-based assessment of risk of other-cause mortality associated with

comorbidity in U.S. prostate cancer patients. Second, the current instruments used to

determine risk of other-cause mortality are prohibitively cumbersome to use in the

clinical setting. Lastly, little data simultaneously integrate risk assessment for both

5
cancer-specific and other-cause mortality and include relevant variables to predict both,

including age, comorbidity, and tumor risk.

In this study, we sought to characterize the association of comorbidity, age, and tumor

features with long-term other-cause and prostate cancer-specific mortality in a large,

population-based sample of men with clinically localized prostate cancer. We simplified

our comorbidity assessment to a count of twelve major conditions to allow for easy

translation to the clinical setting. We hoped to identify groups of men that have a high

risk of other-cause mortality and low risk of prostate cancer-specific mortality, so that

these men might better understand the risks and benefits of various therapeutic

strategies and make a truly informed decision regarding treatment.

6
METHODS

Study Participants

The Prostate Cancer Outcomes Study (PCOS) is based on a population-based cohort of

men diagnosed with prostate cancer ascertained from the National Cancer Institute

Surveillance, Epidemiology, and End Results (SEER) program. Details of the PCOS

have been published previously (11). In brief, the study includes men diagnosed with

prostate cancer between October 1, 1994 and October 31, 1995 who resided in an area

covered by one of six SEER tumor registries: the states of Connecticut; Utah; and New

Mexico; and the metropolitan areas of Atlanta, Georgia; Los Angeles County, California;

and King County, Washington (Seattle). All men aged 39 to 89 years were eligible

except in King County, where inclusion was limited to men aged 60 to 89 years. Subjects

were identified within 6 months of diagnosis using a Rapid Case Ascertainment System.

Population attrition is reported in Appendix Figure 1. A total of 3,533 (62%) of the eligible

men completed the 6 or 12 month survey. The institutional review board of each

participating institution approved the study.

For the current analysis, we included all men in PCOS with non-metastatic prostate

cancer at diagnosis. We excluded men with nodal or distant metastases, those without

information on comorbidities at diagnosis, and those diagnosed incidentally at the time of

cystoprostatectomy. Our final analytic sample comprised 3,183 men.

Data Collection

7
All patients included in this analysis completed a baseline survey within 6 months of

diagnosis that included sociodemographic and clinical information (including presence or

absence of comorbid conditions) as well as self-reported function and quality of life.

Medical record data. All participants underwent a review of their medical records at one

and five years after diagnosis. Abstractors obtained demographic information, clinical

symptoms, diagnostic examinations, tumor-related information (diagnostic Prostate

Specific Antigen (PSA), Gleason score, and clinical stage), and primary treatment type.

Treatment types were defined as aggressive therapy (surgery, external beam radiation

therapy, or brachytherapy) or non-aggressive therapy (androgen deprivation therapy

(ADT) or watchful waiting). In addition, information on tumor characteristics, primary

treatment, and vital status were collected from the SEER registries.

Comorbidity. We assessed comorbidity using a modified version of the Charlson

Comorbidity Index (12). In the current analysis, comorbidity was expressed as a count of

the following twelve major conditions at the time of diagnosis: Diabetes, bleeding

gastrointestinal ulcer, chronic lung disease, congestive heart failure (CHF), stroke,

myocardial infarction, angina/chest pain, cirrhosis/liver disease, arthritis, inflammatory

bowel disease, hypertension, and depression. Subjects answered yes/no for each of

these comorbidities on the baseline survey. Responses were not validated by chart

review.

Tumor characteristics. Tumors were stratified by clinical and pathological features using

the widely accepted D’Amico criteria, which utilize diagnostic PSA level, Gleason score,

and clinical stage at diagnosis to predict risk of progression, overall mortality, and
8
cancer-specific mortality. Tumors are classified as low- (PSA < 10, clinical stage ≤ T2a,

and Gleason score ≤ 6), intermediate- (PSA 10-20, clinical stage T2b, or Gleason score

7), or high-risk (PSA > 20, clinical stage ≥ T2c, or Gleason score ≥ 8) (13, 14). Tumors

are categorized into higher risk strata by having at least one of the characteristics of that

stratum.

Vital status. Vital status and underlying cause of death were determined through 14

years following diagnosis using SEER data.

Statistical Analysis

We first grouped patients by comorbidity count (0, 1, 2, and 3+) and compared baseline

characteristics using the analysis of variance for continuous variables and chi-square

test for categorical variables. Cumulative incidence rates were computed for overall,

prostate cancer specific, and other-cause mortality.

Other-cause mortality was modeled with the proportional subdistribution hazards

regression as described by Fine and Gray treating prostate cancer death as a

competing risk (15). A competing event eliminates a possibility of the primary event of

interest, and treating it as a mere censored observation would violate the non-

informative censoring assumption of the Cox proportional-hazard model. When modeling

competing risks, the conclusion may be based on cause-specific hazard or sub-

distributional hazard. As summarized by Dignam et. al, both approaches are valid and

informative, and the choice often depends on questions of interest (16). We choose to

use Fine and Gray model, which applies a regression modeling directly on a cumulative

9
incidence function (CIF) and allows for estimation of the effect of covariates on the

CIF.We included comorbidity count, age, race, SEER site, D’Amico tumor risk, and

treatment type as covariates. An interaction term between age and comorbidity count

was included to allow age-specific effects of comorbidity counts on survival. Prostate

Cancer-specific mortality was modelled using a similar approach. We checked the

proportional hazard assumption using scaled Schoenfeld residuals (17) as suggested by

Fine and Gray (15). A weak evidence of violation of the assumption on the categorized

comorbidity counts was found for other-cause mortality (p=0.04); however, closer

graphical examination of the Schoenfeld residuals revealed that the proportionality

seems to hold except for the very end of the study period (13+ years) where the data are

scarce. Similar trend was found for D’Amico tumor risk (p=0.02); only 94 patients were

categorized into the unknown group that causes assumption violation. For prostate

cancer mortality, proportional hazard assumption was likely to be violated for comorbidity

counts (p < 0.01), as well as age, SEER site, tumor risk and treatment type; however, for

this secondary endpoint we still present average subhazard ratios (18) to denote its

relationship with comorbidity count.

To determine if aggressive treatment was associated with better prostate cancer-specific

survival among men with higher comorbidity counts, we further analyzed the potential

interaction between treatment (aggressive/non-aggressive) and comorbidity. An alpha

level of 0.05 was used to denote statistical significance, and all tests were two-sided.

Statistical analyses were performed using R 2.14 (19), with cmprsk package (20) for

Fine and Gray modeling.

Funding Source

10
This study was funded by grant R01CA114524 from the National Cancer Institute of the

National Institutes of Health. Dr. Daskivich is supported by grants from the Robert Wood

Johnson Clinical Scholars Foundation and American Urological Association. The various

funding sources had no role in the design, conduct, analysis, or decision to publish the

study.

11
RESULTS

Baseline characteristics of the sample after stratification by comorbidity count are shown

in Table 1. Older men, and African American race/ethnicity, and those with higher tumor

risks tended to have higher comorbidity counts. Men with higher comorbidity counts

tended to be treated aggressively less often than those with lower comorbidity counts.

However, 256 of 419 men (61%) with comorbidity counts of 3 or greater were treated

aggressively for clinically localized disease. Counts of specific co-morbid conditions

across comorbidity groups are reported in Table 1 of the appendix.

Cumulative other-cause mortality rates by comorbidity count are shown in Appendix

Table 2. At fourteen years of follow-up, other-cause mortality estimates were 24%, 33%,

46%, and 57% for men with comorbidity counts of 0, 1, 2, and 3+, respectively.

Cumulative incidence curves depicting cumulative other-cause and prostate cancer-

specific mortality by comorbidity count are shown in Figure 1.

Proportional subdistributional hazards models investigating the association between

other-cause mortality and comorbidity count also showed an increasing subhazard of

other-cause mortality with higher comorbidity count. Accounting for age, race, SEER site,

tumor risk stratum, and type of treatment, the subhazards of other-cause mortality for

men with 1, 2, or 3+ comorbidities were 1.2 (95% CI 1.0–1.4), 1.7 (95% CI 1.4–2.0), and

2.4 (95% CI 2.0–2.8) respectively, compared with men with no comorbidities.

12
In order to illustrate the impact of increasing age on risk of other-cause mortality in men

with the most severe comorbidity, we calculated ten-year cumulative other-cause

mortality rates for men with three or more comorbidities across different age strata.

Among those with three or more comorbidities, men aged <60 years, 61-74 years, and

>75 years at diagnosis had ten-year other-cause mortality rates of 26% (17 of 65 men),

40% (108 of 272 men), and 71% (58 of 82 men), respectively.

Cumulative incidence of prostate cancer and other-cause mortality by D’Amico tumor

risk categories and comorbidity count—after stratification into younger (age <60),

intermediate (age 60–70), and older (age >70) groups—are illustrated in Figures 2 and 3.

In all age groups, the risk of other-cause mortality increased with greater comorbidity

count, and the risk of prostate cancer mortality increased with higher tumor risk-stratum.

In concordance with our multivariable models, the absolute impact of comorbidity on risk

of other-cause mortality was markedly less pronounced in men younger than 60,

compared with older men. Across the entire cohort, prostate cancer mortality was low in

men with low- (3%) and intermediate-risk (7%) disease but appreciable in men with high-

risk disease (18%). Proportions of men treated aggressively vs. non-aggressively within

each age/tumor risk subgroup are reported in Appendix Table 3. The risk of prostate

cancer mortality associated with tumor risk was similar across age groups.

A competing risks model analyzing the subhazard of prostate cancer mortality

associated with non-aggressive treatment among comorbidity groups suggested that

there may be a benefit to aggressive therapy in men with little or no co-morbid

disease. These treatments may, however, not be of as much value in men with more

comorbid conditions. Men with zero or one comorbidity who were managed

13
conservatively had 2.4 (95% CI 1.6–3.5) and 2.2-fold (95% CI 1.5–3.3) increase

respectively in the subhazard of prostate cancer mortality compared with men treated

aggressively. Men with two or three or more comorbidities who were managed

conservatively did not have a statistically significant increase in prostate cancer mortality

(HR:1.6, 95%CI=1.0-, 2.7 and 1.5, 95%CI=0.9-2.5 respectively).

In an effort to further inform clinical decision-making in non-metastatic prostate cancer,

we recapitulated our analysis including men who elected ADT as primary therapy in the

aggressive management group, effectively exploring the decision of any therapy vs none

whatsoever. The results were virtually the same, with the exception the competing risks

model analyzing the subhazard of prostate cancer mortality associated with no treatment

whatsoever. While men with no comorbid conditions still had a increased risk (hazard

ratio: 2.0, 95% CI 1.3-3.0), men with 1, 2 or 3+ co-morbid conditions were not at

increased risk of prostate cancer mortality (HR:1.1, 95%CI=0.7-1.9; 1.2, 95%CI=0.7-2.2

and; 1.1, 95%CI=0.6-2.0 respectively)

14
DISCUSSION

These results illustrate the contemporary long-term risks of other-cause and prostate

cancer mortality associated with age at diagnosis, comorbidity, and tumor risk in a

population-based cohort of men diagnosed with clinically localized prostate cancer.

These population-based data provide more accurate estimates of long-term prognosis

for men with newly diagnosed prostate cancer, since they are not limited by the selection

bias incurred by using single institutional (6-9) or modality-specific data (10). In fact,

several studies have shown that the risk of other-cause mortality associated with a given

comorbidity score increases in a stepwise manner when comparing men treated with

radical prostatectomy, radiation therapy, and watchful waiting (6, 21). Since men who

undergo surgery are generally younger, have better functional status, and have less

severe manifestations of disease than men who choose radiation therapy (and even

more so compared with those who choose watchful waiting), other-cause mortality rates

among men receiving surgery may be lower for a given comorbidity score than those

who choose radiation or watchful waiting. Using a population-based sample of men who

are balanced in age, tumor characteristics, and primary treatment will provide a more

accurate estimate of mortality risk associated with comorbidity. We feel that a

population-based cohort best mimics the relevant clinical scenario: counseling a man

with newly diagnosed with clinically localized prostate cancer who has not yet made a

treatment decision.

Our data also verify the strong prognostic utility of comorbidity in predicting risk of other-

cause mortality and show that its prognostic strength persists even when comorbidity

assessment is reduced to a simple count of twelve major comorbid conditions. It is

unknown whether more complicated, weighted systems of comorbidity assessment,

15
such as the Charlson comorbidity index (12), provide better risk stratification than a mere

count of comorbidities, but our data suggest that a count may be sufficient for assessing

prognosis in men at the highest risk. Such a simplification of comorbidity assessment

may be essential for widespread application of these findings to the clinical setting.

We also found that the subhazard of other-cause mortality associated with a given

comorbidity count doubles with each decade of age at diagnosis above 55 years.

Therefore, a comorbidity count of 3+ for a man aged 65 years at diagnosis is associated

with double the risk compared with a man having an equal number of comorbidities but

aged 55 years at diagnosis. A man aged 75 years at diagnosis with 3+ comorbidities has

quadruple the risk of a similar 55-year-old. A simple frequency analysis mirrors this

relationship, showing 10-year other-cause mortality rates of 26%, 40%, and 71% for men

with 3 or greater comorbidities aged <60, 61 to 74, and >75 years at diagnosis,

respectively.

These data provide benchmarks for evaluating risk of prostate and other-cause mortality

among men with different baseline comorbidity, age, and tumor features. The decision of

whether to pursue aggressive treatment is preference sensitive, affected by an

individual's knowledge of and personal attitudes about potential benefits and harms;

therefore, the optimal risk ratio of prostate cancer vs. other-cause mortality may differ

from individual to individual. Regardless, our data provide a framework upon which all

patients can apply their own attitudes and assess their likelihood of treatment-related

benefit based on their age, comorbidity, and disease characteristics. However, from a

public health perspective, the precedent of the widely-recognized 2008 US Preventative

Services Task Force recommendations on cessation of screening and treatment for men

greater than 75 years (22) has established that men with less than a 10-year life
16
expectancy (i.e. populations with greater than 50% other-cause mortality at ten years)

garner little or no benefit from aggressive treatment. Applying a similar cutoff for

comorbidity and age, our data show that men with 3+ comorbidities aged 60 and older

approach 50% mortality at 10 years after diagnosis. Therefore, given the low likelihood

of short-term prostate cancer mortality and high likelihood of other-cause mortality, older

men with more than 3 major comorbidities should strongly weigh the risk of dying from

other causes prior to realizing any potential survival benefit from aggressive therapy

when choosing primary treatment.

A corollary analysis investigating whether non-aggressive treatment is associated with

an increased risk of prostate cancer mortality among comorbidity groups suggested that

men with two or more comorbidities may derive little survival benefit from aggressive

treatment. Men with two or more comorbidities did not have a statistically significant

increase in risk of prostate cancer mortality with non-aggressive treatment compared

with men treated aggressively, although we had limited power to discern differences.

Men with fewer than two comorbidities, however, did have a significant increase in risk.

This stands in contrast to a previous claims-based study comparing overall survival in

men older than 65 who were treated aggressively vs. conservatively, which suggested a

survival benefit with aggressive treatment that persisted after correction for comorbidity

(23). This study also showed a benefit in prostate cancer-specific survival with

aggressive treatment after correction for propensity scoring alone, but propensity scores

were not balanced by number of comorbidities. Although thought-provoking, our results

should be treated as exploratory given that the analysis has limited power, that there is

likely a selection bias between groups treated aggressively and non-aggressively and

that there does appear to be some benefit in other studies.

17
Our study is limited by several methodological issues. First, the comorbidity assessment

relied on self-report by participants and was not confirmed by medical records or

physicians. However, numerous studies have shown that patient report of co-morbid

conditions, particularly common ones like those assessed here, is fairly reliable (24).

Second, because self-reporting of conditions does not allow for a detailed assessment of

comorbidity severity, a broad spectrum of disease is subsumed under each of the twelve

major comorbidity categories. Men with mild (or severe) disease manifestations may

have lower (or higher) rates of other-cause mortality than those estimated by our raw

comorbidity counts. Furthermore, information regarding the presence or absence of

other cancers at the time of diagnosis was not collected at the time of enrollment in

PCOS, so It is not included in our comorbidity count. Third, since the intake

questionnaire was completed within six months of diagnosis, it is possible that some

comorbid conditions may not have been present before treatment. However, the majority

of our twelve designated conditions are chronic in nature and should have been

apparent at the time of diagnosis. Fourth, a small proportion of men (32 men) were too

sick or lacked the capacity to fill out the intake questionnaire and were excluded from

PCOS; this may result in underestimation of other-cause mortality in the sickest men.

Fifth, since the majority of the cohort was treated aggressively for prostate cancer, rates

of prostate cancer mortality may be lower than if these men had been managed

conservatively, especially at more distant time points and for men with high-risk tumors.

However, for men with low- and intermediate-risk tumors, our estimates will

underestimate prostate cancer-specific mortality by less than 10% at ten years, based

on survival differences observed between aggressively and non-aggressively treated

men in a randomized controlled trial (1). Sixth, we did not collect information on length of

time from diagnosis to baseline survey and could not, therefore, conduct sensitivity

analyses on the assumption that no temporal bias related to comorbidity reporting was
18
introduced into the study. Lastly, although current randomized controlled trial data

suggests that there is no significant survival benefit with aggressive treatment of early-

stage prostate cancer until 8 to 10 years after local therapy (1), further data on efficacy

of aggressive treatment—especially for men with intermediate-risk disease—will help to

define appropriate cut points for triage of care.

When considering these data, the reader may be inclined to inappropriately apply the

results of these findings to the decision as to whether or not to screen for prostate

cancer, particularly given the recent USPSTF assignment of a grade D to prostate

cancer screening (25). Unfortunately, these data do not inform this decision and should

not be used to decide whether or not to screen. While the risk of overtreatment (which is

driven primarily by the morbidity associated with aggressive therapy) is cited as one of

the risks of screening that led the panel to their conclusion, they do not state in their

recommendation that men who have been diagnosed with prostate cancer should or

should not undergo treatment. We strongly believe that the decisions to screen or to

treat prostate cancer are two separate entities and should always be considered

individually.

In summary, because the potential for morbidity is high, men with a new diagnosis of

prostate cancer should first understand the likely benefit vs. potential harm associated

with aggressive treatment. These data provide a basis upon which to counsel men

regarding their risk of prostate cancer-specific and other-cause mortality. Our data are

based on simple variables that are commonly available to the clinician at the time of

treatment decision: age, number of major comorbidities at diagnosis, and tumor risk.

Older men with multiple major comorbid conditions should be informed of their higher

probability of death from other causes before deriving a survival benefit from surgery or
19
radiation therapy for low- and intermediate-risk disease. The information provided herein

aims to make the competing risks of mortality clearer to both patients and their

physicians.

20
ACKNOWLEDGEMENTS

We wish to thank the men who participated in PCOS who, by their participation, have

contributed to a better understanding of the effects of prostate cancer on men’s lives.

We also thank the physicians in the six Surveillance, Epidemiology and End Results

areas who assisted us in the collection of data from their patients and from medical

records. We thank all the study managers and chart abstractors for their outstanding

efforts in data collection. Finally, we thank all the staff in the six cancer registries for their

help with the study.

21
Address for Reprint Requests:

David F. Penson, MD, MPH

Professor of Urologic Surgery
Director, Center for Surgical Quality and Outcomes Research
Vanderbilt University Medical Center
2525 West End Avenue, Suite 600
Nashville, TN 37203-1738

Current Mailing Address for All Authors:

Timothy J. Daskivich, MD
Robert Wood Johnson Clinical Scholars
University of California, Los Angeles
7th Floor, Suite 710
10940 Wilshire Blvd
Los Angeles, CA 90024

Kang-Hsien Fan, MS
Vanderbilt University
571 Preston Research Building
Nashville, Tennessee 37232-6848

Tatsuki Koyama, PhD

Assistant Professor of Biostatistics
571 Preston Research Building
Nashville, TN 37232-6848

Peter C. Albertsen, MD
Chief and Program Director, Division of Urology
University of Connecticut Health Center
263 Farmington Avenue
Farmington, CT 06030– 3955

Michael Goodman, MD, MPH

Associate Professor, Department of Epidemiology
Emory University Rollins School of Public Health
1518 Clifton Rd. NE
Atlanta, GA 30322

Ann S. Hamilton, PhD

22
Associate Professor of Research, University of Southern California
Health Services Campus
SSB 318E
M/C 9239
Los Angeles, CA 90089-9239

Richard M. Hoffman, MD, MPH

Professor, Department of Internal Medicine
Department of Veterans Affairs Medical Center
General Internal Medicine 111GIM
1501 San Pedro Dr., S.E.
Albuquerque, NM 87108

Janet L. Stanford, PhD, MPH

Research Professor, Department of Epidemiology
Fred Hutchinson Cancer Research Center
Box 358080 M4-B874
1100 Fairview Ave N., Building M
P.O. Box 19024
Mailstop: M4-B874
Seattle, WA 98195-1024

Antoinette M. Stroup, PhD

Associate Professor of Research, Department of Clinical Epidemiology
University of Utah
Utah Cancer Registry
650 Komas Drive, Suite 106B
Salt Lake City, UT 84108

Mark S. Litwin, MD, MPH

Professor and Chair, Department of Urology
University of California, Los Angeles
BOX 957383
924 Westwood Blvd, Ste 1000
Los Angeles, CA 90095-7383

David F. Penson, MD, MPH

Professor of Urologic Surgery
Vanderbilt University Medical Center
2525 West End Avenue, Suite 600
23
Nashville, TN 37203-1738

Reproducible Research Statement: The authors of this study are commited to the
concept of reproducibility of medical research. Our methods section details our research
protcol and our statistical code was provided to the statisticians at Annals in order to
validate our results. Our detailed research protocol and statistical code may be made
available at the discretion of the corresponding author. Use of our dataset is restricted to
members of the PCOS study team and their designates.

24
TABLE AND FIGURE LEGENDS

Table 1. Sample characteristics by comorbidity count, % (No of patients)

Footnote: Percentages are reported across rows and may not sum to 100% due to
rounding.
PSA is summarized with median and quartiles.
* Chi-squared tests except ANOVA for age and PSA.
** Aggressive = surgery or radiation therapy
*** ADT = Androgen deprivation therapy

Figure 1. Cumulative Incidence Curves for Other-cause Specific Mortality and

Prostate Cancer Specific Mortality by Comorbidity Count
Figure 2. Competing Risks Model Depicting Cumulative Incidence of Other-Cause
and Prostate Cancer-Specific Mortality by Comorbidity Count for Men Ages (a) <
60 Years, (b) 60-70 Years, and (c) > 70 Years
Figure 3. Competing Risks Model Depicting Cumulative Incidence of Other-Cause
and Prostate Cancer-Specific Mortality by D’Amico risk criteria for Men Ages (a) <
60 Years, (b) 60-70 Years, and (c) > 70 Years

APPENDIX TABLES AND FIGURE LEGENDS

Appendix Table 1: Frequency of Comorbid Conditions Across Comorbidity Count

Groups

Appendix Table 2. Fourteen-year cumulative other-cause mortality by comorbidity

count
Footnote: OCM = Other-cause Mortality.

Appendix Table 3: Type of Treatment by Age/Tumor Risk Subgroups

Appendix Figure 1: CONSORT Diagram for PCOS

25
REFERENCES

1. Bill-Axelson A, Holmberg L, Ruutu M, Garmo H, Stark JR, Busch C, Nordling S,

Häggman M, Andersson SO, Bratell S, Spångberg A, Palmgren J, Steineck G, Adami
HO, Johansson JE; SPCG-4 Investigators. Radical prostatectomy versus watchful
waiting in early prostate cancer. N Engl J Med. 2011 May 5;364(18):1708-17.
2. Mohler J, Bahnson RR, Boston B, Busby JE, D'Amico A, Eastham JA, Enke CA,
George D, Horwitz EM, Huben RP, Kantoff P, Kawachi M, Kuettel M, Lange PH,
Macvicar G, Plimack ER, Pow-Sang JM, Roach M 3rd, Rohren E, Roth BJ, Shrieve DC,
Smith MR, Srinivas S, Twardowski P, Walsh PC. NCCN clinical practice guidelines in
oncology: prostate cancer. J Natl Compr Canc Network 2010 Feb;8(2):162-200.
3. Thompson I, Thrasher JB, Aus G, Burnett AL, Canby-Hagino ED, Cookson MS,
D'Amico AV, Dmochowski RR, Eton DT, Forman JD, Goldenberg SL, Hernandez J,
Higano CS, Kraus SR, Moul JW, Tangen CM; AUA Prostate Cancer Clinical Guideline
Update Panel. Guideline for the management of clinically localized prostate cancer:
2007 update. J Urol. 2007 Jun;177(6):2106-31.
4. Walz J, Suardi N, Shariat SF, et al. Accuracy of life tables in predicting overall survival
in patients after radical prostatectomy. BJU Int. 2008;102:33-8.
5. Daskivich TJ, Chamie K, Kwan L, Labo J, Palvolgyi R, Dash A, Greenfield S, Litwin
MS. Overtreatment of men with low-risk prostate cancer and significant comorbidity.
Cancer. 2011 May 15;117(10):2058-66.
6. Tewari A, Johnson CC, Divine G, et al. Long-term survival probability in men with
clinically localized prostate cancer: a case-control, propensity modeling study stratified
by race, age, treatment and comorbidities. J Urol. 2004;171:1513-9
7. Walz J, Gallina A, Saad F, et al. A nomogram predicting 10-year life expectancy in
candidates for radical prostatectomy or radiotherapy for prostate cancer. J Clin Oncol.
2007;25:3576-81.
8. Cowen ME, Halasyamani LK, Kattan MW. Predicting life expectancy in men with
clinically localized prostate cancer. J Urol. 2006;175:99-103.
9. Daskivich TJ, Chamie K, Kwan L, Labo J, Dash A, Greenfield S, Litwin MS.
Comorbidity and competing risks for mortality in men with prostate cancer. Cancer. 2011
Apr 8.
10. Albertsen PC, Moore DF, Shih W, Lin Y, Li H, Lu-Yao GL. Impact of comorbidity on
survival among men with localized prostate cancer. J Clin Oncol. 2011 Apr
1;29(10):1335-41.
11. Potosky AL, Harlan LC, Stanford JL, Gilliland FD, Hamilton AS, Albertsen PC, et al.
Prostate cancer practice patterns and quality of life: the Prostate Cancer Outcomes
Study. J Natl Cancer Inst 1999;91:1719–24.
12. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying
prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis.
1987;40:373-83.
13. D'Amico AV, Whittington R, Malkowicz SB, et al. Biochemical outcome after radical
prostatectomy, external beam radiation therapy, or interstitial radiation therapy for
clinically localized prostate cancer. JAMA. 1998;280:969-74.

26
14. Boorjian SA, Karnes RJ, Rangel LJ, Bergstralh EJ, Blute ML. Mayo Clinic validation
of the D'amico risk group classification for predicting survival following radical
prostatectomy. J Urol. 2008;179:1354-60.
15. Fine, J. P., and R. J. Gray. A proportional hazards model for the subdistribution of a
competing risk. Journal of the American Statistical Association 1999;94:496-509.
16. Dignam JJ, Zhang Q, and Kocherginsky MN. The use and interpretation of
competing risks regression models. Clinical Cancer Research 2012; 18: 2301-2308.
17. Grambsh P, Therneau T. Proportional hazards tests and diagnostics based on
weighted residuals. Biometrika 1994; 81: 515-526.
18. Grambauer N, Schumacher M, Bayersmann J. Proportional subdistribution hazards
modeling offers a summary analysis, even if misspecified. Statistics in Medicine 2010;
29: 875-84.
19. R Development Core Team (2011). R: A language and environment for statistical
computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-
0, URL http://www.R-project.org/.
20. Bob Gray. cmprsk: Subdistribution Analysis of Competing Risks. 2011; R package
version 2.2-2.
21. Fowler JE Jr, Terrell FL, Renfroe DL. Co-morbidities and survival of men with
localized prostate cancer treated with surgery or radiation therapy. J Urol.
1996;156:1714-8.
22. U.S. Preventive Services Task Force. Screening for prostate cancer: U.S. Preventive
Services Task Force recommendation statement. Ann Intern Med. 2008;149(3):185-91.
23. Wong YN, Mitra N, Hudes G, Localio R, Schwartz JS, Wan F, Montagnet C,
Armstrong K. Survival associated with treatment vs. observation of localized prostate
cancer in elderly men. JAMA. 2006; 296(22): 2683-93.
24. Klabunde CN, Reeve BB, Harlan LC, Davis WW, Potosky AL. Do patients
consistently report comorbid conditions over time?: results from the prostate cancer
outcomes study. Med Care 2005; 43(4): 391-400.
25. Moyer, VA and the U.S. Preventive Services Task Force: Screening for Prostate
Cancer: US Preventive Services Task Force recommendation statement. Ann Intern
Med 2012; 157(2): 120-34

27