change in health-related quality of life of newly diagnosed cancer patients, cancer survivors, and...

Change in Health-related Quality ofLife of Newly Diagnosed CancerPatients, Cancer Survivors,and ControlsFrank Baker, PhD1; Maxine Denniston, MSPH2; Samuel C. Haffer, PhD3; and Penny Liberatos, PhD1

BACKGROUND: Data from the 1998 Health Outcomes Survey (HOS) of patients who were enrolled in Medi-

care managed care and follow-up data from the 2000 HOS resurvey were analyzed to examine changes in

health-related quality of life (HRQOL) of newly diagnosed cancer patients, cancer survivors, and patients

without cancer. METHODS: In 1998, the HOS was mailed to a random sample of 279,135 beneficiaries, and

167,096 respondents (60%) returned completed surveys. Those who were diagnosed with cancer (22,747)

were frequency age-matched to an equal number of patients with no cancer. In 2000, the HOS was mailed

to the same cohort of beneficiaries. Complete data at both baseline and follow-up were available on 16,850

individuals for inclusion in the current study. RESULTS: After 2 years, respondents who had been diagnosed

with cancer at baseline continued to have lower scores on all but 3 scales of the 36-item short-form

HRQOL measure. However, there was no evidence that they were declining any faster than or catching up

with noncancer patients. Those who had been newly diagnosed with cancer since the baseline survey had

lower mean scale scores than the no-cancer group on all scales and lower mean scores than the cancer

survivors on all subscales except Bodily Pain, Vitality, and Mental Health. CONCLUSIONS: This study dem-

onstrated that, after 2 years, cancer survivors continued to have poorer HRQOL than the no-cancer group.

Newly diagnosed cancer patients had poorer quality of life than both the longer term cancer survivors and

the no-cancer group. Cancer 2009;115:3024–33. Published 2009 by the American Cancer Society*.

KEY WORDS: health-relatedqualityof life, oncology, cancer patients, canceroutcomes,Medicaremanagedcare.

The term ‘‘cancer survivor’’ has been defined by the American Cancer Society (ACS) and the NationalCancer Institute as including individuals from the time of their diagnosis with cancer and for the balanceof their lives. By using this definition, it has been estimated that 11.1 million cancer survivors currently arealive in the United States.1 During recent decades, as the number of patients surviving cancer has increased,more attention has been given to health-related quality of life (HRQOL) as an important patient-reported mea-sure of cancer care outcomes. Although it is recognized that assessment of quality of life has the potential for

Received: September 24, 2008; Revised: November 18, 2008; Accepted: December 19, 2008

Published online: April 28, 2009 VC 2009 American Cancer Society

DOI: 10.1002/cncr.24330, www.interscience.wiley.com

Corresponding author: Frank Baker, PhD, Department of Behavioral Sciences and Community Health, School of Public Health, New York Medical

College, Valhalla, NY 10595; Fax: (914) 594-3481; [email protected]

1School of Public Health, New York Medical College, Valhalla, New York; 2Behavioral Research Center, American Cancer Society, Atlanta, Georgia;3Centers for Medicare and Medicaid Services, Baltimore, Maryland

Ms. Denniston is no longer employed by the American Cancer Society.

*This article is a US government work and, as such, is in the public domain in the United States of America.

3024 Cancer July 1, 2009

Original Article

identifying the changing sequelae of cancer and its treat-ments, possibly providing important data for treatmentdecisions and improvement of health outcomes, meth-odological problems have made the interpretation of suchdata difficult. Usually, it is not possible to compareHRQOL before and after diagnosis, because the measure-ment of HRQOL typically comes only after the diagnosisof cancer already has occurred. Furthermore, such studiesusually do not include a no-cancer control group that isassessed with the same measures and at the same time asthe group of cancer patients being studied.

Studies in the literature comparing the quality of life

of cancer survivors with individuals who have never had

cancer have produced mixed results. For example, a study

conducted in Sweden compared the HRQOL of cancer

survivors 3 years after diagnosis with the scores of a nor-

mative population and observed no overall differences

except for the Role-Physical subscale of the Medical Out-

comes Study (MOS) 36-item short form (SF-36). How-

ever, those authors also observed differences between the

sexes: Women in that study had better HRQOL than con-

trols, but it was the opposite for men.2 Other studies that

have focused on breast and colon cancers,3 testicular can-

cer,4 Hodgkin disease,5 and breast cancer6 have not

reported any significant differences in overall HRQOL

between cancer survivors and comparison groups. Other

studies have reported some differences in HRQOL for

survivors of breast cancer,7,8 colorectal cancer,9 and other

cancers.10-12 Many of those studies did not include US

samples and were limited in sample size. Moreover, in

many studies, adults of all ages are included.2,7,9 More

recently, 2 studies13,14 focused on older adults in large

national samples and observed that cancer survivors had

poorer HRQOL than individuals without cancer.

A previous article by the current authors also

reported significant differences in HRQOL scores

between large groups of US cancer survivors and non-

cancer patients. Baker et al15 compared a sample of

22,747 cancer survivors with an equal number of fre-

quency age-matched noncancer patients from the Health

Care Financing Administration’s 1998 Medicare Health

Outcomes Survey (HOS) of patients who were receiving

managed care services through Medicare. Cancer survi-

vors had statistically significantly lower scores than the

noncancer patients on all 8 of the subscales and on the 2

summary measures of the MOS SF-36. Among the

patients who were in treatment at the time of that study

for 4 cancer types, the poorest quality of life was observed

for those with lung cancer, followed by those with more

than 1 of the 4 cancer types examined. The data suggested

that cancer has negative effects on HRQOL that were not

explained by simple age effects, because the groups were

frequency age-matched.

The current report builds on those initial data and

takes advantage of the finding that a follow-up HOS sur-

vey was sent in 2000 to the respondents from the 1998

survey. During that 2-year interval, some of the respond-

ents who had reported no cancer diagnosis in 1998 said in

2000 that they had developed cancer. Thus, it is possible

to examine changes in HRQOL scores from 1998 to 2000

in individuals who were diagnosed with cancer during the

interim, longer term cancer survivors, and individuals

who have never had cancer.

MATERIALS AND METHODS

Measures

The HOS includes a 95-item core instrument and a 5-

item variable module.16 The core instrument includes a

measure of HRQOL and demographic questions. The

HRQOL measure that we used was the MOS SF-36, a

36-item, well tested, valid, and reliable self-report tool

that has been used in thousands of studies worldwide to

measure HRQOL.17 This measure includes 8 scales:

Physical Functioning, Role-Physical, Bodily Pain, Gen-

eral Health, Vitality, Social Functioning, Role-Emo-

tional, and Mental Health. In addition, 2 summary

measures—the Physical Component Summary (PCS)

and the Mental Component Summary (MCS)—are cal-

culated based on the 8 scales. The SF-36 is scored so that

higher scores represent better functioning, and scores are

standardized using normative values for the general US

population, with a score of 50 representing the national

average and 10 points above or below the mean represent-

ing a difference of 1 standard deviation from the national

average.

Procedures

The 1998 baseline survey was sent to participants who

were in Medicare managed care plans that had a Medicare

contract as of January 1, 1997 and were enrolled

HRQOL in Elderly With or Without Cancer/Baker et al

Cancer July 1, 2009 3025

continuously for at least 6 months. In each of the larger

plans, a random sample of 1000 beneficiaries was selected.

In plans with �1000 enrollees, all eligible members were

surveyed. The survey was administered in the spring of

1998 and again to the same cohort in the spring of 2000.

The baseline survey included 279,135 beneficiaries

who were enrolled in 268 managed care plans that covered

287 market areas. Completed surveys were received from

167,096 respondents for a raw response rate of 60%. Two

separate data files were created for this study from the

cleaned data. The first consisted of all respondents who indi-

cated that a physician ever had told them that they had any

cancer (except skin cancer). This yielded 22,747 respond-

ents who we defined as cancer survivors using the ACS defi-

nition. The second file consisted of respondents who

indicated that no physician ever had told them that they had

cancer (except skin cancer). A frequency age-matched sam-

ple of noncancer patients was selected randomly for inclu-

sion in the analytic dataset using 13 age categories (ages

<30 years, 30-39 years, 40-49 years, 50-59 years, 60-64

years, 65-69 years, 70-74 years, 75-79 years, 80-84 years,

85-89 years, 90-94 years, 95-99 years, and>99 years), thus

allowing for better assessment of differences between cancer

survivors and those without a cancer diagnosis.

Of the 43,757 respondents to the 1998 survey who

were included in our initial dataset, 10,639 were ineligible

for follow-up in that they were disenrolled involuntarily

because of changes in their original plan; 2906 respond-

ents had died, 8163 voluntarily had disenrolled from

Medicare managed care, and 897 were no longer eligible

for the program. Most of these groups split about equally

between those who had cancer at baseline and those who

did not, except for those who died during the 2-year inter-

val. A little more than two-thirds (64.5%) of those who

had died (2906) had indicated a cancer diagnosis at the

1998 survey.

In addition, data on cancer diagnoses were missing

for 2966 individuals at follow-up, and 1336 individuals

provided inconsistent responses by reporting a cancer di-

agnosis at baseline but no cancer diagnosis ever at follow-

up. This reduced the final number of respondents to

16,850 for analysis in the current study. All of these indi-

viduals were aged �65 years, because those who were

younger had disabilities and were excluded both from the

current analysis and from earlier analysis of the baseline

1998 data.15

Data Analysis

A single SAS analytic data file was created that included

baseline data and follow-up data for the 16,850 respond-

ents for this study. One-way and 2-way analyses of var-

iance performed in PROC GLM were used to compare

mean SF-36 standardized scores; Tukey (honestly signifi-

cant differences) tests were used in post hoc testing. Chi-

square tests in PROC FREQ were used for demographic

comparisons. All tests were 2-sided.

RESULTS

Sample Characteristics

Table 1 presents demographic characteristics of the

respondents who were cancer survivors in 1998 (previous

cancer), those who did not have cancer in 1998 or in 2000

(no cancer), and those who were newly diagnosed with

cancer in 2000 (new cancer). Chi-square tests across all 3

groups were significant for all characteristics that were

examined (all P¼ .0001).

Pairwise chi-square tests indicated the following:

Although the no-cancer group and the previous-cancer

group did not differ significantly with regard to age (ages

65-74 years and �75 years), they differed significantly on

the other characteristics. The no-cancer group included

proportionally more women, racial minorities, Hispanics,

and individuals who were unmarried, had less than a high

school education, or had lower household incomes. The

new-cancer group was older and included fewer women

compared with the previous-cancer group. Comparing

the no-cancer group with the new-cancer group, the latter

group was older and included fewer women.

Although more statistically significant differences

were observed in pairwise testing between the previous-

cancer and no-cancer groups than between the new-cancer

group and the other 2 groups, the new-cancer group

included only 469 patients, whereas the other 2 groups

were much larger (n ¼ 7369 and n ¼ 9012, respectively).

It is noteworthy that nearly two-thirds of respondents

(64.6%) in the new-cancer group were aged �75 years

compared with the previous-cancer group and the no-can-

cer group, in which approximately 55% of each group was

included in the older age group. The differences in sex

across the groups also were sizeable (women comprised

48.4% of the new-cancer group vs 55.5% of the previous-

Original Article


cancer group and 59.8% of the no-cancer group). Differ-

ences in other characteristics were much smaller and,

although statistically significant, may not be of great prac-

tical importance.

Similarity of Follow-up

to Baseline Respondents

To assess whether the individuals who responded to the

follow-up survey (n ¼ 21,152) were similar on HRQOL

to those in the entire initial cohort (n¼ 43,757), the base-

line data for mean SF-36 scores for both groups were cal-

culated, as shown in Table 2. The respondents in 2000

had higher baseline scores on the 2 summary scales and on

all 8 subscales of the SF-36 than the total 1998 cohort.

This was true both for individuals who were diagnosed

with cancer and for individuals who were not diagnosed

with cancer in 1998. In addition, the difference in mean

scores between cancer and noncancer patients was smaller

for the 2000 respondents than for the entire 1998 cohort

Table 1. Sample Characteristics by Cancer Status at Follow-up: Health Outcomes Study 1998 Baseline and 2000

Follow-up Data (n ¼ 16,850)

New Cancer,n 5 469

Previous Cancer,n 5 7369

No Cancer,n 5 9012

Characteristic No. % No. % No. %

Age at follow-up, y65-74 166 35.4*,y,z 3379 45.8*,y 4083 45.3*,z‡75 303 64.6 3990 54.2 4929 54.7

SexMen 238 51.6*,y,z 3235 44.5*,y,§ 3575 40.2*,z,§Women 223 48.4 4037 55.5 5324 59.8

Race/EthnicityAmerican Indian 3 0.7* 28 0.4*,§ 50 0.6*,§

Asian/Pacific Islander 9 2 95 1.3 209 2.4

Black/African American 26 5.7 286 4 505 5.7

White 412 90 6737 93 7902 89.3

Other/multiracial 8 1.8 95 1.3 179 2

Hispanic

Yes 17 3.7* 235 3.3*,§ 406 4.7*,§

No 439 96.3 6911 96.7 8208 95.3

Current marital statusMarried 262 56.7* 4250 58.4*,§ 4966 55.9*,§

Divorced/separated 30 6.5 630 8.7 671 7.5

Widowed 156 33.8 2228 30.6 3002 33.8

Single, never married 14 3 167 2.3 248 2.8

Highest education<HS graduate 146 31.6* 1825 25.1*,§ 2722 30.8*,§

HS graduate/GED 158 34.2 2576 35.5 3075 34.8

Some college/2-y degree 102 22.1 1664 22.9 1792 20.2

4-y degree 24 5.2 536 7.4 572 6.5

>4-y degree 32 6.9 657 9.1 678 7.7

Household income, $US<$10,000 57 15.5* 747 12.6*,§ 1113 15.7*,§

$10,000-$19,999 126 34.3 1790 30.2 2213 31.3

$20,000-$49,999 148 40.3 2655 44.8 2970 42

$50,000-$79,999 23 6.3 486 8.2 526 7.4

‡$80,000 13 3.6 248 4.2 251 3.6

HS indicates high school; GED, general education degree.

*P ¼ .0001 (chi-square test across all 3 cancer status groups).

yP ¼ .0001 for age and P ¼ .0028 for sex (chi-square tests comparing the new cancer group with the previous cancer group).

zP ¼ .0001 (chi-square test comparing the new cancer group with the no cancer group).

§P ¼ .0001 (chi-square test comparing the previous cancer group with the no cancer group).



on all scales; however, these differences were statistically

significant within each survey. Differences in mean

scores between the entire 1998 cohort and the 2000

respondents could not be tested, because the latter are a

subgroup of the former. The higher scores among the

respondents in 2000 compared with the entire 1998

cohort might reflect the deaths of individuals by 2000

who had been in poor health when they initially were sur-

veyed in 1998.

Change in SF-36 Scores From 1998 to

2000 by Cancer Status

Table 3 presents the standardized scores at baseline and at

follow-up and the change scores for the 2 summary meas-

ures and the 8 subscales of the SF-36 for the 3 groups that

we assessed. At baseline, the new-cancer group and the

no-cancer group differed only on the Physical Function-

ing and Role-Emotional subscales, and the no-cancer

group had higher mean scores on both. The previous-can-

cer group had a significantly lower mean score on 5 of the

8 subscales and on PCS compared with the other 2 groups

at baseline. The no-cancer group had significantly higher

mean scores on the Physical Functioning subscale than

the other 2 groups, whereas the new-cancer group had a

significantly lower mean score on the Role-Emotional

subscale compared with the no-cancer group. The 3

groups did not differ on the MCS scale or on the Mental

Health subscale.

Comparing the previous-cancer group with the no-

cancer group at follow-up indicated that the patients with

cancer still had significantly lower HRQOL scores after

2 years on all scales except the MCS, Role-Emotional, and

Mental Health scales. However, there was no evidence

that the patients with cancer declined any faster or that

they were catching up with regard to HRQOL scores with

patients in the no-cancer group. For example, the average

decline across the 8 subscales for the previous-cancer and

no-cancer groups was almost identical (�1.63 vs �1.65,

respectively). It should be noted that all mean scores for

the SF-36 declined for all 3 groups. However, the greatest

decline was for the new-cancer group, in which there was

an average change across the 8 scales of�3.87.

The new-cancer group had significantly lower mean

scores at follow-up than the no-cancer group on all scales

and also had lower mean scores than the previous-cancer

Table 2. Mean Baseline SF-36 Standardized Scores by

Baseline Cancer Status: Health Outcomes Study 1998Baseline and 2000 Follow-up Respondents

Mean (SD) for SF-36Score or Difference*

Baseline (1998)Cancer Status:SF-36 Subscale

Total BaselineSample,n 5 43,757

2000RespondentsOnly, n 5 21,152

PCSyCancer 38.5 (12) 39.6 (11.6)

No cancer 41.2 (11.8) 41.6 (11.6)

Difference 2.7 2

MCSyCancer 51.4 (10.7) 52.2 (10.2)

No cancer 52.6 (10) 53 (9.7)

Difference 1.2 0.8

Physical FunctionyCancer 38.3 (13.1) 39.8 (12.4)

No cancer 40.5 (12.9) 41.2 (12.5)

Difference 2.2 1.4

Role-PhysicalyCancer 40.6 (12.9) 41.6 (12.8)

No cancer 43 (12.8) 43.5 (12.7)

Difference 2.4 1.9

Bodily PainyCancer 43.2 (11.4) 44 (11)

No cancer 45 (11.2) 45.3 (11)

Difference 1.8 1.3

General HealthyCancer 43 (11.4) 44.2 (10.8)

No cancer 46 (10.8) 46.6 (10.5)

Difference 3 2.2

VitalityyCancer 45.5 (11.2) 46.5 (10.9)

No cancer 47.9 (10.8) 48.4 (10.6)

Difference 2.4 1.9

Social FunctionyCancer 45.7 (12.8) 47.3 (11.8)

No cancer 48 (11.7) 48.8 (11)

Difference 2.3 1.5

Role-EmotionalyCancer 47.2 (12.1) 48.1 (11.6)

No cancer 48.3 (11.5) 48.7 (11.2)

Difference 1.1 0.6

Mental HealthyCancer 50.2 (10.6) 51 (10.1)

No cancer 51.3 (10.2) 51.7 (9.9)

Difference 1.1 0.7

SD indicates standard deviation; SF-36, the 36-item Medical Outcomes

Study short form; PCS, Physical Component Summary; MCS, Mental Com-

ponent Summary.

* Difference indicates the difference between the no cancer group and the

cancer group.

yP < .001 for the difference between the cancer group and the no cancer

group on mean baseline scores within both the total baseline sample (n ¼43,757) and the 2000 follow-up respondents only (n ¼ 21,152).

Original Article


group on all scales except the Bodily Pain, Vitality, and

Mental Health subscales. The mean change scores for

those who were newly diagnosed with cancer were signifi-

cantly greater than those for the other 2 groups except for

the Mental Health scale, on which the 3 groups did not

differ significantly.

The overall pattern of change for the 3 groups from

baseline to follow-up for the PCS and MCS is illustrated

in Figure 1. It is apparent that the trajectories of the no-

cancer and previous-cancer groups are the same. That is,

they both declined at the same rate over the 2-year period,

and the previous-cancer group began at a lower level of

HRQOL. The new-cancer group was better off than the

previous-cancer group at baseline but then declined pre-

cipitously by the follow-up point. This pattern was the

same for the PCS and MCS scales but was more marked

for the physical domain.

Given the large sample sizes used in the current

study, some of the comparisons presented above were stat-

istically significant even when the magnitude of the differ-

ences was somewhat small. Using an estimate of effect size

of 2.2 points to indicate an approximation for clinical sig-

nificance,15 fewer of these significant differences poten-

tially may represent clinically important differences.

Table 3. Mean SF-36 Standardized Scores and Change

Scores by Cancer Status: Health Outcomes Study 1998Baseline and 2000 Follow-up Data (n ¼ 16,850)

Mean Standardized Scoreor Change Score

2000 Cancer Status:SF-36 Subscale

Baseline Follow-up Change

PCSNew cancer 41.3* 36.4* �4.86*

Previous cancer 39.8y 37.7y �2.03yNo cancer 42.1* 40.1z �2.02y

MCSNew cancer 52.8* 50.2* �2.54*

Previous cancer 52.7* 51.7y �0.95yNo cancer 53.3* 52.3y �1.04y

Physical FunctionNew cancer 40.2* 36.4* �3.86*

Previous cancer 40.1* 37.7y �2.44yNo cancer 41.7y 39.5z �2.18y

Role-PhysicalNew cancer 43.3* 37.9* �5.44*


Bodily PainNew cancer 45.1* 41.8* �3.25*

Previous cancer 44.3y 42.8* �1.26yNo cancer 45.4* 44.8y �1.36y

General HealthNew cancer 46.3* 41.6* �4.62*

Previous cancer 44.4y 42.9y �1.60yNo cancer 47* 45.3z �1.70y

VitalityNew cancer 48.4* 44.5* �3.90*

Previous cancer 46.5y 45.1* �1.42yNo cancer 48.7* 47.2y �1.45y

Social FunctionNew cancer 48.7* 43.4* �5.31*


Role-EmotionalNew cancer 47.7* 44.7* �3.03*

Previous cancer 48.6*,y 46.9y �1.65yNo cancer 49y 47.5y �1.45y

Mental HealthNew cancer 51.7* 50.1* �1.52*

Previous cancer 51.5* 50.5*,y �0.93*

No cancer 52.1* 51.1y �0.95*

SF-36 indicates the Medical Outcomes Study 36-item short form; PCS,

Physical Component Summary; MCS, Mental Component Summary.

* Tukey grouping within each SF-36 subscale for each score group: Mean

values of the baseline, follow-up, and change scores on each subscale for

each group with the same symbol did not differ significantly from others in

the same group for that subscale.

yTukey grouping.

zTukey grouping.

FIGURE 1. Physical Component Summary (PCS) and Mental

Component Summary (MCS) scores at baseline and at fol-

low-up are illustrated by cancer status. Solid line indicates

new cancer; dashed line with squares, previous cancer;

dashed line with triangles, no cancer.



Change in SF-36 Scores by Age and Cancer

Status at Baseline

Table 4 presents change scores between baseline and fol-

low-up by age group and baseline cancer status for 2000

respondents only. Regardless of cancer status, change

scores illustrated a greater decline for older respondents

than for their younger counterparts. These changes are

statistically significant for 7 of the 8 subscales (except

Bodily Pain) and for the MCS scale but not the PCS scale.

Furthermore, the average decline for the 8 subscales of the

SF-36 for the total sample and for the 2 age groups (ages

65-74 years and �75 years) are very similar both for the

previous-cancer group and for the no-cancer group (total,

�1.63 vs �1.65, respectively; ages 65-74 years, �1.27 vs

�1.29, respectively; aged �75 years, �1.94 vs �1.95,

respectively) (see Fig. 2). Differences on mean scale scores

by age were presented in our previous report15; thus, in

the current article, only change scores are included.

DISCUSSION

This follow-up of a large sample of cancer patients with a

frequency age-matched no-cancer group demonstrates

that cancer survivors continue to have poorer HRQOL

than patients without cancer. In the current study, there is

no evidence that cancer survivors return to the same level

of quality of life as those who have not had cancer, at least

within the 2-year time frame that was examined. This is

consistent with some recent studies13,14 but is in contrast

to some earlier studies in which it was suggested that can-

cer survivors return to a ‘‘normal’’ level of quality of life

over time or that there is no difference in their quality of

life compared with individuals who do not have cancer.2-6

This difference in findings may be attributed to several

differences in study design.

Table 4. Mean SF-36 Change Scores by Age and Cancer

Status at Baseline: Health Outcomes Study 1998 Baselineand 2000 Follow-up Data (n ¼ 16,850)

Mean Change Score

Baseline CancerStatus: SF-36Subscale

TotalSample,N 5 16,850

Ages65-74 Years,n 5 7628

Aged‡75 Years,n 5 9222

PCSPrevious cancer �2.03 �1.85 �2.18

No cancer �2.02 �1.97 �2.05

MCS*Previous cancer �0.95 �0.50 �1.34

No cancer �1.04 �0.44 �1.57

Physical Function*Previous cancer �2.44 �2.03 �2.79

No cancer �2.18 �1.98 �2.34

Role-PhysicalyPrevious cancer �2.04 �1.71 �2.33

No cancer �2.10 �1.88 �2.28

Bodily PainPrevious cancer �1.26 �1.14 �1.36

No cancer �1.36 �1.29 �1.42

General Health*Previous cancer �1.60 �1.39 �1.78

No cancer �1.70 �1.31 �2.02

Vitality*Previous cancer �1.42 �1.13 �1.67

No cancer �1.45 �1.07 �1.76

Social Function*Previous cancer �1.70 �1.09 �2.21

No cancer �2.02 �1.27 �2.65

Role-Emotional*Previous cancer �1.65 �1.03 �2.19

No cancer �1.45 �0.94 �1.88

Mental Health*Previous cancer �0.93 �0.63 �1.19

No cancer �0.95 �0.60 �1.23

SF-36 indicates the Medical Outcomes Study 36-item short form; PCS,

Physical Component Summary; MCS, Mental Component Summary.

*P < .001 for the difference in mean change score between age groups

within cancer status.

yP ¼ .01 for the difference in mean change score between age groups

within cancer status. There were no statistically significant differences by

cancer status in mean change scores.

FIGURE 2. Mean change scores (from baseline to follow-up)

are illustrated for 8 Medical Outcomes Study 36-item short

form subscales by age and cancer status at baseline. Gray

bars indicate previous cancer; black bars, no cancer.

Original Article


First, the current study included a much larger sam-

ple than most previous studies, especially cancer survivors

(almost 8000 compared with the usual number of<200).

Second, this study was restricted to individuals aged �65

years, whereas many other studies have included younger

cancer survivors, who may have a different quality-of-life

trajectory over time.2,7,9 In fact, 1 study reported differen-

ces in HRQOL between cancer survivors and controls

among an older population (aged�65 years) but not nec-

essarily among a younger population.2 Third, follow-up

after cancer diagnosis in several previous studies has

occurred over a considerable time (13 or more years). It is

reasonable to assume that the longer an individual has

been cancer-free, especially with cancers that occur at a

younger age, the greater the likelihood that quality of life

will improve and will be more similar to that of individu-

als who have never had cancer. The current study included

individuals with various (and unknown) lengths of survi-

vorship and may have included individuals with recent

diagnoses, which would increase the likelihood of finding

poorer quality of life. Fourth, some studies have assessed

the quality of life of cancer survivors with no cancer recur-

rence6 and observed that it was similar to the quality of

life of controls, and other studies that compared controls

with survivors who had recurrent cancer have reported

much poorer quality of life within the latter group.6,11

Because cancer recurrence was not available in our dataset,

those with cancer recurrence may have been included

within the ‘‘previous-cancer’’ group, thereby lowering the

reported quality of life of this group.

It is noteworthy that quality of life declined for all

individuals who were included in the current study regard-

less of their cancer status. Among individuals aged �65

years, a decline in HRQOL was observed over a 2-year pe-

riod and may have been related to aging. For example,

according to Deimling and colleagues (among others),

there is an increased likelihood of chronic conditions,

functional limitations, and self-reported fair/poor

health,18-20 and these may be reflected in declining

HRQOL scores over time. However, analyses of cancer

status that included major comorbidities were performed

on our baseline data, and, regardless of comorbidities,

cancer survivors still had poorer HRQOL. This decline in

HRQOL, as measured by the SF-36 with increasing age,

has been documented in other studies20-23 and is sup-

ported further by our current findings that the largest

declines in HRQOL were among the older respondents

(aged�75 years) regardless of their cancer status. Further-

more, for each of the 3 groups that we assessed in this

study, respondents reported larger declines in PCS scores

compared with MCS scores over the 2-year period that

was examined. This is consistent with other studies in

which patients were more likely to report significant

declines in physical functioning but not necessarily in

mental functioning.7,11,20 One study in which signifi-

cantly lower scores on the PCS and MCS were reported

among cancer survivors compared with controls also indi-

cated that, for the MCS, this was not true after 2 years

past the cancer diagnosis.7 This suggests that, after a cer-

tain amount of time has passed after diagnosis, cancer sur-

vivors may adjust to their diagnosis mentally; however,

physical functioning decrements still may persist.

One of the distinctive aspects of the current study

was the availability of HRQOL data on a group of indi-

viduals before they had cancer. Assessment of quality of

life generally is available only for cancer patients after they

have been diagnosed or, at the earliest, when it is sus-

pected that they have cancer.24 Our study indicated that

there is a sharp decline in both Physical and Mental

Health HRQOL scores for individuals recently diagnosed

(within a 2-year period) with cancer relative to their

reported HRQOL before diagnosis and relative to those

without cancer and longer term survivors.

It is noteworthy that, at baseline, before patients in

the new-cancer group were diagnosed, their reported

HRQOL was lower numerically on all subscales and was

significantly lower on the Physical Functioning and Role-

Emotional subscales of the SF-36 compared with the

reported HRQOL for the no-cancer group. Because many

cancers tend to develop slowly over time, these individuals

may have been asymptomatic and may have had subtle

negative effects on their HRQOL even before diagnosis.

This finding warrants further study.

There are several limitations to this study. First, this

sample of individuals enrolled in Medicare managed care

is not necessarily representative of the Medicare popula-

tion in the United States as a whole, because those who

participated in the managed care program constituted

approximately 17% of the total in 2000.25 Furthermore,

African Americans are somewhat under represented

(4.8% vs 8.9%), whereas whites (89.3% vs 84.5%) and

Latinos (3.9% vs 2.7%) are somewhat over represented in



the study sample compared with all beneficiaries who

were enrolled in Medicare managed care plans in 1998.26

Second, the current study is limited to older individuals

and does not include younger individuals with cancer,

whose HRQOL trajectory may differ from that of the

group we studied. Third, because information regarding

the time since cancer diagnosis is not available, those with

previous cancer potentially may include a very wide range

of individuals in terms of time of diagnosis. Clearly a pre-

ponderance of recent survivors could have a negative

influence on HRQOL scores. Similarly, during the 2-year

interval that we studied, we did not identify the point at

which the new-cancer group was diagnosed. Patients who

were diagnosed closer to the time of follow-up probably

reported much poorer HRQOL. Fourth, the types of can-

cer diagnoses that respondents received were available

only for those who still were in treatment when the HOS

was completed. Because the health consequences of differ-

ent types of cancers differ, the corresponding HRQOL

also may differ. Finally, the number of patients with new

cancer (469) was much smaller than the number with pre-

vious cancer (7369) and no cancer (9012); however, this

number is much larger than the number of cancer patients

included in most other studies.

Despite these limitations, the current study has sev-

eral important strengths. The size of the sample was

uniquely large. The sample was drawn from a geographi-

cally diverse national population, including urban, subur-

ban, and rural residents. The no-cancer group that we

used for comparison was drawn from the same population

as the previous-cancer and new-cancer groups. Unlike

many studies, the current investigation was not focused

on a single type of cancer but included a wide range of

cancers. Most previous studies have tended to focus on

cross-sectional comparisons as opposed to changes over

time in HRQOL. Also, unlike most other studies, data on

HRQOL were available for this study before some individ-

uals had been diagnosed with cancer. This allowed us to

take a unique look at HRQOL before the diagnosis of a

major illness.

On the basis of the current findings, we recom-

mend that future work should consider the potential dif-

ferential effects of age when examining the outcomes of

cancer survivors, because it appears that older individuals

may have different trajectories in HRQOL than their

younger counterparts. To the extent possible, the time

since cancer diagnosis and the documentation of any

cancer recurrence should be considered when assessing

the HRQOL of cancer survivors. The evidence does not

appear to support the notion that cancer survivors’

HRQOL in terms of physical health will return to a

‘‘normal’’ (precancer) level, at least among older adults.

Whether this occurs in terms of mental health function-

ing still requires further study.

Conflict of Interest Disclosures

Analyses were supported by intramural funding from the Ameri-can Cancer Society.

References

1. Reis LAG, Melbert D, Krapcho M, et al, eds. SEER Can-cer Statistics Review, 1975-2005. Baltimore, Md: NationalCancer Institute; 2008. Available at: http://seer.cancer.gov/csr/1975_2005/. Accessed May 31, 2008.

2. Hammerlid E, Taft C. Health-related quality of life inlong-term head and neck cancer survivors: a comparisonwith general population norms. Br J Cancer. 2001;84:149-156.

3. Mosconi P, Apolone G, Barni S, Secondino S, Sbanotto A,Filiberti A. Quality of life in breast and colon cancer long-term survivors: an assessment with the EORTC QLQ-C30and SF-36 questionnaires. Tumori. 2002;88:110-116.

4. Joly F, Heron JF, Kalusinski L, et al. Quality of life inlong-term survivors of testicular cancer: a population-basedcase-control study. J Clin Oncol. 2002;20:73-80.

5. Gil-Fernandez J, Ramos C, Tamayo T, et al. Quality of lifeand psychological well-being in Spanish long-term survivorsof Hodgkin’s disease: results of a controlled pilot study.Ann Hematol. 2003;82:14-18.

6. Helgeson VS, Tomich PL. Surviving cancer: a comparisonof 5-year disease-free breast cancer survivors with healthywomen. Psychooncology. 2005;14:307-317.

7. Trentham-Dietz A, Sprague BL, Klein R, et al. Health-related quality of life before and after a breast cancer diag-nosis. Breast Cancer Res Treat. 2008;109:379-387.

8. Robb C, Haley WE, Balducci L, et al. Impact of breastcancer survivorship of quality of life in older women. CritRev Oncol Hematol. 2007;62:84-91.

9. Arndt V, Merx H, Stegmaier C, Ziegler H, Brenner H.Quality of life in patients with colorectal cancer 1 year afterdiagnosis compared with the general population: a popula-tion-based study. J Clin Oncol. 2004;22:4829-4836.

10. Thome B, Dykes AK, Hallberg IR. Quality of life in oldpeople with and without cancer. Qual Life Res. 2004;13:1067-1080.

Original Article


11. Boini S, Briancon S, Guillemin F, Galan P, Hercberg S.Impact of cancer occurrence on health-related quality oflife: a longitudinal prepost assessment [serial online]. HealthQual Life Outcomes. 2004;2:4.

12. Kopp M, Holzner B, Meraner V, et al. Quality of life inadult hematopoietic cell transplant patients at least 5 yearsafter treatment: a comparison with healthy controls. Eur JHaematol. 2005;74:304-308.

13. Clauser SB, Arora NK, Bellizzi KM, Haffer SC, Topor M,Hays RD. Disparities in HRQOL of cancer survivors andnon-cancer managed care enrollees. Health Care Financ Rev.2008;29:23-40.

14. Smith AW, Reeve BB, Bellizzi KM, et al. Cancer,comorbidities, and health-related quality of life of older adults[serial online].Health Care Financ Rev. 2008;29:41-56.

15. Baker F, Haffer SC, Denniston M. Health-related qualityof life of cancer and noncancer patients in Medicare man-aged care. Cancer. 2003;97:674-681.

16. National Committee for Quality Assurance, ed. Health ofSeniors Survey Manual. HEDIS 3.0, No. 6. Washington,DC: National Committee for Quality Assurance; 1998.

17. Turner-Bowker DM, Bartley PJ, Ware JE Jr. SF-36 HealthSurvey and ‘‘SF’’ Bibliography, 3rd ed (1988-2000). Lin-coln, RI: QualityMetric Inc.; 2002.

18. Deimling GT, Sterns S, Bowman KF, Kahana B. Thehealth of older-adult, long-term cancer survivors. CancerNurs. 2005;28:415-424.

19. Pleis JR, Lethbridge-Cejku M, eds. Summary health statis-tics for U.S. adults: National Health Interview Survey,2006. Vital Health Stat 10. 2007;235:1-153.

20. Trentham-Dietz A, Remington PL, Moinpour CM, Hamp-ton JM, Sapp AL, Newcomb PA. Health-related quality oflife in female long-term colorectal cancer survivors. Oncolo-gist. 2003;8:342-349.

21. Ganz PA, Guadagnoli E, Landrum MB, Lash TL, Rakow-ski W, Silliman RA. Breast cancer in older women: qualityof life and psychosocial adjustment in the 15 months afterdiagnosis. J Clin Oncol. 2003;21:4027-4033.

22. Ware JE, Gandek B, Sinclair SJ, Kosinski M. Measuringand Improving Health Outcomes: An SF-36 Primer for theMedicare Health Outcomes Survey. Waltham, Mass:Health Assessment Lab and QualityMetric Inc.; 2004.

23. Walters SJ, Munro JF, Brazier JE. Using the SF-36 witholder adults: a cross-sectional community-based survey. AgeAgeing. 2001;30:337-343.

24. Montazeri A, Hole DJ, Milroy R, McEwen J, Gillis CR.Does knowledge of cancer diagnosis affect quality of life? Amethodological challenge. BMC Cancer. 2004;4:21-26.

25. Kaiser Family Foundation. Medicare Advantage Plan Pene-tration: 2000. Available at: http://www.kff.org/medicare/healthplantracker/topicresults.jsp? Accessed July 11, 2008.

26. Thorpe KE, Atherly A. MedicareþChoice: current role andnear-term prospects [serial online]. Health Affairs. 2002;17:W242-W252.



change in health-related quality of life of newly diagnosed cancer patients, cancer survivors, and...

Documents