Association Between Time Since Cancer Diagnosis and Health-Related Quality of Life: A Population-Level Analysis

Abstract

Objectives

To examine the association between time since cancer diagnosis and health-related quality of life (HRQOL) among cancer survivors in remission.

Methods

Analyzing data from 3,610 cancer survivors and 59,539 individuals without cancer in the Medical Expenditure Panel Survey, we examined the relationship between time since cancer diagnosis and HRQOL, taking remission status into account and controlling for patient demographics and comorbidities. HRQOL measurements included the six-dimensional health state short form (SF-6D) utility scores, the Physical Component Summary (PCS), and the Mental Component Summary (MCS).

Results

The relation between the time since cancer diagnosis and HRQOL varied substantially across cancer types. Compared with individuals without cancer, survivors of breast, prostate, or poor-prognosis cancer had statistically lower SF-6D scores within 2 years of diagnosis (-0.044, −0.062, and −0.088, respectively). Breast cancer survivors had SF-6D scores similar to non-cancer individuals after 2 years, as did patients with poor-prognosis cancer after 5 years. However, even after a period of 10 years, survivors of prostate or cervical cancer had a lower level of SF-6D scores (−0.027 and −0.042, respectively). The comparisons of physical health between cancer survivors and individuals without cancer were similar to those of SF-6D. In contrast, most cancer survivors did not experience poorer mental health; however, survivors of prostate or cervical cancer had lower MCS after 10 years of diagnosis.

Conclusions

The level of HRQOL among cancer survivors depends on time since cancer diagnosis and cancer type. Some cancer survivors have lower HRQOL after a decade of diagnosis, even in remission.

Introduction

In 2014, approximately 14.5 million individuals in the US were alive with a history of cancer.¹ The 2005 Institute of Medicine report, From Cancer Patient to Cancer Survivor: Lost in Transition, emphasized the necessity of additional research on cancer survivorship.² One of the critical issues in survivorship care is to maintain and improve survivors’ health-related quality of life (HRQOL).

Many population-based studies have examined the effect of cancer and its treatment on HRQOL.^3–15 Extant research, however, has not reached consensus on whether the HRQOL of cancer survivors in remission after a long period of time returns to the same level of that of those who have not had cancer.^{3,9–11,16,17} Few studies have simultaneously accounted for cancer type and time since diagnosis; thus, the interplay between these two factors and HRQOL may be masked. Furthermore, prior research did not control for remission status. A recurrence event can conceivably lead to a decrease in HRQOL,¹⁶ which could influence the relationship between time since cancer diagnosis and HRQOL. Finally, most research measured scores for individual health domains, such as pain, and physical or mental health. Few studies examined an overall, preference-based HRQOL score among cancer survivors.^4,6,11 Such a single health index concerning multiple domains could be incorporated in cost-effectiveness analysis (CEA) for economic evaluation of cancer care.¹⁸ As more people are surviving cancer, their preference-based HRQOL is urgently needed to help providers and policy makers appropriately allocate resources.

To address these knowledge gaps, we analyzed data from a large-scale national survey, the Medical Expenditure Panel Survey (MEPS), to examine the relation between time since cancer diagnosis and HRQOL among cancer survivors taking remission status into account. Using a validated multi-attribute health-state classification system, we aimed to develop a nationally representative “off-the-shelf” catalog of preference-based HRQOL scores. We also reported the HRQOL scores of physical and mental health.

Methods

Design and data

We conducted a cross-sectional study using the MEPS data. The MEPS, a nationally representative survey of the US civilian non-institutionalized population, is considered a comprehensive and reliable data source to evaluate national estimates of health expenditures and health status. It provides information for a 2-year reference period with an overlapping panel survey, in which a new cohort (“panel”) is initiated each year and consists of individuals who are interviewed in-person 5 times (“round”) over two and half consecutive years. To avoid duplicate observations from the same panel and maximize our sample size, we used MEPS data from years 2008, 2010, and 2012, selecting observations from six survey panels entering between 2007 and 2012.

Sample

We identified cancer survivors based on the response to the survey. MEPS includes a series of questions about a cancer diagnosis for adults aged 18 and older. Adult cancer survivors were identified based on their response to a question about whether a doctor or other health professional had ever told them that they had cancer or a malignancy of any kind. Anyone who answered “yes” is further asked about “what kind of cancer,” and “age of diagnosis.” In the data years we selected, MEPS included variables that indicated whether each reported cancer was in remission. Specifically, the MEPS asked individuals whether the condition of cancer is “in remission, that is, the condition (cancer) is under control,” allowing us to identifying cancer survivors who were or were not in remission. Individuals diagnosed solely with non-melanoma skin cancer were not classified as cancer survivors. To limit the effect of multiple cancers on the estimated relationship between time since cancer diagnosis and HRQOL, we selected adult cancer survivors who had only one cancer diagnosis. We included respondents with no cancer history as our comparison group, and excluded respondents who were younger than 18 or who did not have complete data. Acknowledging that cancer survivors might interpret “in remission” incorrectly, we also excluded cancer survivors in remission who had cancer for more than 2 years but received chemotherapy or radiotherapy in the survey year. We included cancer survivors in remission who had cancer for less than 2 years but received chemotherapy or radiotherapy in the survey year because cancer survivors may receive these treatments as an adjuvant therapy to prevent recurrence events. Including the respondents whom we thought not to be in remission, approximately 2.5% of our sample of cancer survivors, reached similar results. Institutional review board review was not required because data are publicly available.

HRQOL Measures

HRQOL was assessed by the Medical Outcomes Study Short Form 12-Item Health Status Survey version 2 (SF-12v2) instrument, with data collected via the self-administered questionnaire to adults aged 18 years old or older participating in the MEPS. Our primary HRQOL measure was the Short Form-6D (SF-6D) score, a preference-based single index. The SF-6D score is generated by converting the elements of the SF-12v2 with a validated utility-based algorithm.¹⁹ MEPS also imputes HRQOL scores in physical and mental health domains, Physical Component Summary (PCS) and Mental Component Summary (MCS). The PCS and MCS scores have been rescaled with averages of 50 and standard deviations of 10 with respect to a proprietary US national data set.²⁰

Covariates

To examine the association between HRQOL and time since cancer diagnosis among cancer survivors in remission, we classified time since cancer diagnosis into four groups: <2 years, 2–4 years, 5–9 years, and ≥10 years. Survivors who were not in remission were placed in one group. We did not classify this group based on time since cancer diagnosis for two reasons: first, the sample size was small, and second, survivors who were not in remission may have an advanced disease or a recurrence event (either of which could have substantial impact on HRQOL). Such an impact on HRQOL among not-in-remission survivors may mask the association between time since cancer diagnosis and HRQOL. Patient cancer type was categorized as breast, prostate, colorectal, melanoma, cervical, hematologic, poor-prognosis, or non-specified cancer. Consistent with prior literature,¹¹ poor-prognosis cancer is defined as cancer of the liver, lung, pancreas, esophagus, or stomach. Survivors diagnosed with cancer at sites other than those classified were combined into a single category of non-specified cancer because of a small sample size for each cancer site. Covariates included individual’s demographic characteristics, such as age, sex, race/ethnicity (white, black, Asian, Hispanic, and others), education, marital status, income level, insurance coverage, metropolitan residence, and geographic region (Northwest, Midwest, South, and West).^4,21 We also controlled for each respondent’s comorbidity based on the response to a series of questions in MEPS, including hypertension, stroke, emphysema, asthma, diabetes, arthritis, vision problems, and hearing problems. Individuals with a history of coronary artery disease, angina, heart attack, or “other heart disease,” were classified as having a history of heart disease.

Analyses

We described cancer survivors by cancer type and time since cancer diagnosis. We calculated unadjusted HRQOLs for cancer survivors as well as individuals without cancer. Descriptive statistics were stratified by history of cancer and were compared using chi-square statistics. All estimates were weighted to account for the MEPS complex survey design and survey nonresponse.

For each HRQOL measure, we conducted a multivariate linear regression with the key independent variables of both time since cancer diagnosis and cancer type, and the covariates. In short, we created 32 dummy variables for cancer survivors in remission indicating eight cancer types and four categories of time since cancer diagnosis. For cancer survivors who were not in remission, we also created 8 dummy variables by cancer type. The estimated coefficient on the dummy variable represents the difference in HRQOL compared to individuals without cancer. For each cancer type, we used an adjusted Wald test regarding the joint significance; that is, we examined the coefficient across four categories of time since diagnosis was jointly equal to zero. We also examined the relation of time since cancer diagnosis and cancer types to HRQOL separately. Because SF-6D scores in MEPS exhibited a ceiling effect and 5.8% of eligible respondents rated themselves in full health, ordinary least squares may result in biased and inconsistent estimates.²² We also applied a Tobit model with upper censoring at 1.0 for SF-6D as a sensitivity analysis.

As a sensitivity analysis, we applied a matching approach to select a group of respondents without cancer. Based on age, sex, and race, non-cancer respondents were randomly selected to match cancer survivors of each cancer type at a 4:1 ratio. For each HRQOL measure and each cancer type, a linear regression was used to estimate the association with time since diagnosis, adjusting for all covariates except age, sex, and race. The model was clustered by matched sets to allow for the correlation between matched pairs of individuals with and without cancer. All statistical analyses were completed using SAS (version 9.3, SAS Institute, Inc., Cary, NC) or Stata, version 12 (StataCorp, College Station, TX). Two-tailed tests with a 0.05 level of significance were used to determine statistical significance.

Results

The sample included 3,610 cancer survivors and 59,539 non-cancer respondents. Among survivors, 20.7% had breast cancer, 15.0% had prostate cancer, 10.2% had cervical cancer, and 3.8% had poor-prognosis cancer (Table 1). Approximately 39.1% of cancer survivors were in remission with a cancer diagnosed greater than 10 years prior, 10.2% were in remission with cancer diagnosed within 2 years prior to taking the survey, and 8.4% were not in remission. Compared with those without cancer, cancer survivors had lower levels of HRQOL; the average scores of SF-6D, PCS, and MCS were 0.737 (standard error, SE = 0.002), 43.0 (SE = 0.23), and 50.1 (SE = 0.21), respectively. For those without cancer, the average scores of SF-6D, PCS, and MCS were 0.806 (SE = 0.001), 50.1 (SE = 0.08), and 51.3 (SE = 0.06), respectively (all P values <.001; Table 2). Respondents with a history of cancer were more likely to be older, female, white, and insured, and to have greater educational attainment. They were also more likely to have comorbidities.

Table 1.

Cancer type and time since cancer diagnosis among cancer survivors

	Breast Cancer	Prostate Cancer	Colorectal Cancer	Poor-prognosis Cancer*	Hematological Malignancy	Melanoma	Cervical Cancer	Non-specified* *	Total (Weighted %)
In remission, <2 years	66	50	28	20	12	42	30	121	369 (10.2)
In remission, 2–4 years	129	132	44	32	34	56	51	195	673 (18.6)
In remission, 5–9 years	190	161	64	34	45	60	64	232	850 (23.5)
In remission, ≥10 years	315	158	83	24	60	129	197	447	1,413 (39.1)
Not in remission	48	41	22	27	16	17	25	109	305 (8.4)
Total (Weighted %)	748 (20.7)	542 (15.0)	241 (6.7)	137 (3.8)	167 (4.6)	304 (8.4)	367 (10.2)	1,104 (30.6)	3,610 (100)

^*Poor-prognosis cancers included cancer of lung, liver, pancreas, esophagus, and stomach.

^**Non-specified cancers included all other cancers except those we specified above.

Table 2.

Health-Related Quality of Life (HRQOL) Outcomes and Demographic Characteristics of Individuals With and Without a History of Cancer

	Individuals With History of Cancer (N=3,610)		Individuals Without History of Cancer (N=59,539)		P Value
HRQOL, Mean (S3)
SF-6D	0.737 (0.002)		0.806 (0.001)		<.001
Physical component summary	43.0 (0.23)		50.1 (0.08)		<.001
Mental component summary	50.1 (0.21)		51.3 (0.06)		<.001
Characteristics	n	Weighted %	n	Weighted %
Age					<.001
18–29	134	3.7	14,005	22.8
30–34	104	2.9	5,879	9.1
35–39	133	3.7	5,858	9.2
40–44	180	5.0	5,631	9.1
45–49	245	6.8	5,819	9.4
50–54	305	8.4	5,624	9.8
55–59	367	10.2	4,854	8.7
60–64	452	12.5	3,856	7.1
65–69	463	12.8	2,837	5.1
70–74	382	10.6	1,931	3.5
75–79	335	9.3	1,387	2.6
≥80	510	14.1	1,858	3.6
Sex					<.001
Male	1,240	37.7	27,604	46.4
Female	2,065	62.3	31,935	53.6
Race/ethnicity					<.001
White	2,474	68.5	26,644	44.8
Hispanic	392	10.9	15,730	26.4
Black	563	15.6	11,472	19.3
Others	181	5.0	5,693	9.6
Year of education					.008
<12	668	18.5	12,823	21.5
12	1,165	32.3	18,771	31.5
>12	1,777	49.2	27,945	46.9
Marriage status					<.001
Married	1,941	53.8	30,414	51.1
Widowed	570	15.8	3,099	5.2
Divorced/separated	720	20.0	8,224	13.8
Never married	379	10.5	17,802	29.9
Income level					<.001
Poor	578	16.0	10,603	17.8
Near poor	250	6.9	3,611	6.1
Low income	540	15.0	9,695	16.3
Middle income	1,034	28.6	18,065	30.3
High income	1,208	33.5	17,565	29.5
Insurance coverage					<.001
Any private	2,131	59.0	34,886	58.6
Public only	1,217	33.7	11,876	19.9
Uninsured	262	7.3	12,777	21.5
MSA					.008
Non-MSA	638	17.7	7933	13.3
MSA	2,962	82.3	51,606	86.7
Region					.512
Northwest	552	15.3	9,343	15.7
Midwest	834	23.1	11,568	19.4
South	1,388	38.4	22,740	38.2
West	836	23.2	15,888	26.7
Comorbidity
Hypertension	2,036	56.4	18,326	30.8	<.001
Heart disease	1,069	29.6	6,890	11.6	<.001
Stroke	325	9.0	1,833	3.1	<.001
Lung disease	243	6.7	999	1.8	<.001
Diabetes	694	19.2	5,630	9.5	<.001
Arthritis	1,802	49.9	12,968	21.8	<.001
Asthma	474	13.1	5,217	8.8	<.001
Impaired vision or blind	106	2.9	782	1.3	<.001
Impaired hearing or deaf	71	2.0	332	0.6	<.001

MSA: Metropolitan statistical area.

The association between time since cancer diagnosis and preference-based quality of life measures varied substantially across cancer types (Figure 1). For survivors in remission who had a cancer diagnosis within the past 2 years, survivors of breast (−0.044; 95% confidence interval, CI −0.007 to −0.080), prostate (−0.062; 95% CI −0.029 to −0.096), or poor-prognosis cancer (−0.088; 95% CI −0.039 to −0.138) had significantly lower SF-6D scores compared with non-cancer respondents. Breast cancer survivors did not have significantly lower SF-6D scores after 2 years. Poor-prognosis cancer survivors had lower SF-6D scores in 2–4 years (−0.046; 95% CI −0.0002 to −0.093), but did not have significantly lower SF-6D scores after 5 years. While survivors who had prostate cancer diagnosed 2–9 years before had a “normal” level of SF-6D, those who were diagnosed more than 10 years prior had lower SF-6D scores (−0.024, 95% CI −0.002 to −0.046). Cervical cancer survivors did not have lower SF-6D scores when cancer was diagnosed less than 10 years, but had significantly lower SF-6D scores with more than 10 years since diagnosis (−0.042, 95% CI −0.022 to −0.061). Survivors in the non-specified cancer group had a lower level of SF-6D scores and did not show a “normal” level even after 10 years since diagnosis. Compared with non-cancer respondents, most cancer survivors who were not in remission had a significantly lower level of preference-based HRQOL, except for survivors of prostate cancer and melanoma who had similar levels of preference-based HRQOL (Appendix Table A1). Results of sensitivity analyses using a Tobit model were qualitatively similar.

Figure 1.

Adjusted Difference of Preference-Based Health-Related Quality of Life Between Cancer Survivors in Remission and Individuals Without Cancer, According to Cancer Type and Time Since Diagnosis

*: P value of joint significance test <.05; SF-6D: Short Form-6D

The relation of time since diagnosis to physical health also varied considerably across cancer types (Figure 2); the comparison of physical health between cancer survivors and individuals without cancer produced results similar to those of preference-based HRQOL. We found that, during the first two years of cancer diagnosis, survivors of breast (−3.6; 95% CI −0.7 to −6.6), prostate (−4.3; 95% CI −0.7 to −7.2), or poor-prognosis cancer (−8.5; 95% CI −4.2 to −12.7) had a significantly lower level of PCS. Survivors of hematologic malignancies had a significantly lower PCS between 2 and 9 years. In contrast, cervical cancer survivors did not have a lower PCS until 10 years after cancer diagnosis; their PCS was lower by −2.2 (95% CI −0.7 to −3.6). Similar to the results of preference-based HRQOL, cancer survivors who were not in remission had a significantly lower level of PCS, except those with prostate cancer or melanoma (Appendix Table A2).

Figure 2.

Adjusted Difference of Physical Health Between Cancer Survivors in Remission and Individuals Without Cancer, According to Cancer Type and Time Since Diagnosis

*:P value of joint significance test <.05; PCS: Physical component summary

Unlike the results for physical health, most cancer survivors did not experience poorer mental health within the first 9 years (Figure 3). However, after 10 years of cancer diagnosis, cervical cancer survivors had a significantly low level of MCS (−2.9; 95% CI −1.1 to −4.8), and prostate cancer survivors had a marginally low level of MCS (−1.7; 95% CI −3.5 to 0.005). Detailed regression results were reported in the Appendix Tables A1–3. Results of sensitivity analyses using a matching approach were qualitatively similar to the findings we reported above (Appendix Table A4).

Figure 3.

Adjusted Difference of Mental Health Between Cancer Survivors in Remission and Individuals Without Cancer, According to Cancer Type and Time Since Diagnosis

*:P value of joint significance test <.05; MCS: Mental component summary

Discussion

Using a cross-sectional research design, this population-based study examined the association between time since cancer diagnosis and HRQOL across various types of cancer survivors. A lower level of HRQOL among cancer survivors in remission in the United States indicated long-term impacts attributable to cancer and associated cancer treatments on HRQOL. Compared with individuals without cancer, cancer survivors in remission of some types of cancer, such as breast cancer, colorectal cancer and melanoma, may have a similar level of HRQOL after 10 years, while survivors of prostate or cervical cancer still had lower levels of HRQOL. Our findings highlight the heterogeneity of cancer survivorship and identify high-risk populations in extended survivorship that may benefit from additional resources and interventions.

Building upon prior work, our findings contribute to literature in important ways. First, in analyses of national data across different cancer types, we demonstrated not only the time-sensitive nature of HRQOL differences but also the variation of these time-sensitive changes across cancer types. While prior literature did identify time-sensitive HRQOL changes in cancer survivorship,^9–11 these studies either grouped different types of cancer together^9,11 or examined only a single cancer type.¹⁰ For example, using data from the Surveillance, Epidemiology, and End Results-Medicare Health Outcomes Survey, researchers demonstrated the time-sensitive nature of decline in HRQOL of prostate cancer patients.¹⁰ We reported HRQOL differences by examining time since cancer diagnosis and cancer type simultaneously, shedding light on the intricate HRQOL changes among cancer survivors.

Second, the finding of a lower HRQOL level in survivors with cervical or prostate cancer after 10 years is particularly intriguing, suggesting that even if cancer is in remission, survivors still suffered from the long-term side effects. Research has shown that cervical or prostate cancer survivors may have long-term bowel, urinary, and sexual dysfunction as well as psychosocial consequences.^23–26 Among literature concerning cervical cancer, a French population-based study reported that 15-year but not 5- or 10-year cervical cancer survivors had impaired psychoemotional HRQOL.²⁷ Lymphedema, the most disabling sequelae of treatment, can worsen over time and was speculated as the main cause.^27–29 Our results corroborated with their findings, indicating that cervical cancer survivors may have long-term mental health issues and additional check-ups on these individuals may be necessary. Regarding prostate cancer, researchers have observed that bowel, urinary, or sexual dysfunction varied, depending on the duration after treatment.³⁰ Prior literature reported that the disease-specific quality of life measures among prostate cancer survivors decreased abruptly after treatment, rebounded (but did not necessarily return to pre-treatment levels) within 2 years, and then decreased after 15 years since diagnosis.³⁰ The variations in these functional outcomes may explain the patterns of HRQOL observed in our study. Additionally, our results that prostate cancer survivors do not have low HRQOL between 2–10 years are consistent with the findings of an early study reporting no difference in general HRQOL between non-cancer controls and prostate cancer survivors at median follow-up of approximately 5.5 years.³¹ Nevertheless, our findings suggest that novel treatments for prostate cancer (such as stereotactic body radiotherapy) may have potential for long-term side effects; thus HRQOL of these patients requires long-term monitoring.

Third, the long-term impact of cancer on mental health differed from that on physical health. Most cancer survivors usually had a lower level of physical HRQOL during the early period but the negative impact of cancer on physical health diminished over time. In contrast, mental health changes did not follow this pattern. Therefore, skillful and nuanced survivorship care is needed. Furthermore, our findings underscore the utility of preference-based HRQOL, which aims to capture both physical and mental health. Also, the results do not seem to be driven by age. Our results showed that as people aged, physical health declined but mental health conversely improved (see Appendix Tables 2–3), which is consistent with the analyses of populations from both the United States and the United Kingdom.³² We observed a non-monotonic relationship between age and SF-6D (see Appendix Table 1), potentially due to the opposite effect of age on physical and mental health. Future studies validating our observations are warranted.

Fourth, our results regarding the relation of time since cancer diagnosis to preference-based HRQOL scores could be used in cancer-related CEA. Future CEAs may need to account for time-sensitive preference-based HRQOL changes in their models. Quality-adjusted life years, incorporating with a preference-based HRQOL score, have been recommended as the most appropriate means of evaluating the effectiveness of health interventions.³³ Based on our results, researchers can assign time-dependent HRQOL weights for a health state of cancer survivors in remission. For example, researchers may assume that disutility of prostate cancer is −0.062 in the first 2 years, and this decrease would change to −0.015 between 2 and 5 years, −0.012 between 5 and 10 years, and −0.024 after 10 years.

Cancer survivors who were not in remission may have an advanced disease or experience a recurrent event. Therefore, it is not surprising that they had a lower level of preference-based HRQOL, with the exception of survivors with prostate cancer or melanoma. Since prostate cancer patients may receive active surveillance without treatments, these not-in-remission survivors might not have a lower HRQOL. Future research is necessary to examine HRQOL among melanoma survivors. Additionally, the results of long-term HRQOL differences among poor-prognosis cancer survivors should be interpreted in light of the study design. Because our sample of poor-prognosis cancer survivors was limited to those in remission, these survivors should have their cancer early diagnosed and appropriately treated so that they can have life expectancy more than 10 years. Our findings indicate that poor-prognosis cancer survivors who are in remission may have their HRQOL similar to the level of individuals without cancer. Future research is needed.

We acknowledge the limitations of our study. First, this is a cross-sectional study that does not allow for longitudinal inferences. The MEPS data is based on self-report; hence errors in reporting on cancer diagnoses, timing, and/or type of cancer may exist. MEPS does not collect information about cancer stage and treatment modalities that preceded MEPS participation. Thus, HRQOL of cancer survivors in the same cancer category may be heterogeneous. We also lacked information on the severity of each comorbidity. Thus, our analyses did not control for these factors, which may influence HRQOL. Second, the MEPS does not include cancer survivors living in institutions such as nursing homes or those too ill to participate who may have had worse HRQOL; thus, cancer survivors’ HRQOL may be overestimated. Third, there may be a cohort effect because patients with longer duration since cancer diagnosis may have received more invasive or toxic treatment modalities compared to those in current use, and these may have more side effects and thus decrease patient HRQOL. If this were the case, a lower level of HRQOL for survivors with longer duration since diagnosis may be overestimated when applying our results to patients who received treatment more recently. Additionally, the progress in screening for some of these cancers, such as prostate or breast cancer, has changed over time and may confound our time-based findings. Finally, we may not have sufficient power to detect significant differences for certain types of cancer. For example, prior literature suggests a minimal important difference (MID) of 0.03 for SF-6D and 2 for PCS.^34,35 Our study showed that the effect sizes of the decreases in SF-6D and PCS among survivors of colorectal cancer or hematologic malignancies were larger than the MIDs but remained insignificant. Similarly, we found non-specified cancer survivors, such as those with brain or bladder cancer, had a significantly lower level of HRQOL after cancer diagnosis. However, we were unable to analyze non-specified cancers separately due of the small sample size of each specific cancer type. Future research using a cohort design with long-term follow-up to examine the HRQOL issues in cancer survivorship is warranted.

In conclusion, we evaluated the relation of time since cancer diagnosis to HRQOL among cancer survivors in remission. Compared with individuals without cancer, cancer survivors, even in remission, may still have a lower level of HRQOL. Therefore, cancer-related CEAs may need to account for time-sensitive HRQOL changes in their models.